arno/hashcat
1
0
Fork 0
mirror of https://github.com/hashcat/hashcat.git synced 2024-11-24 08:58:13 +00:00
Code Issues Releases Wiki Activity
e69c385b4d
hashcat/OpenCL/rp.c

4356 lines
161 KiB
C
Raw Normal View History

Initial commit
2015-12-04 14:47:52 +00:00
/**
Allow and support vector-width 16, which is current maximum for OpenCL. Closes #226.
2016-02-16 15:42:08 +00:00
* Authors.....: Jens Steube <jens.steube@gmail.com>
* magnum <john.magnum@hushmail.com>
*
Initial commit
2015-12-04 14:47:52 +00:00
* License.....: MIT
*/
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 apply_rule (const u32 name, const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len);
inline u32 apply_rules (const __global u32 *cmds, u32 buf0[4], u32 buf1[4], const u32 len);
inline u32x apply_rules_vect (const u32 pw_buf0[4], const u32 pw_buf1[4], const u32 pw_len, const __global kernel_rule_t *rules_buf, const u32 il_pos, u32x w0[4], u32x w1[4]);
Fixed compiler warnings (no previous prototype for function)
2016-01-25 12:35:58 +00:00
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 generate_cmask (u32 buf)
Initial commit
2015-12-04 14:47:52 +00:00
{
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
const u32 rmask = ((buf & 0x40404040) >> 1)
& ~((buf & 0x80808080) >> 2);
Initial commit
2015-12-04 14:47:52 +00:00
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
const u32 hmask = (buf & 0x1f1f1f1f) + 0x05050505;
const u32 lmask = (buf & 0x1f1f1f1f) + 0x1f1f1f1f;
Initial commit
2015-12-04 14:47:52 +00:00
return rmask & ~hmask & lmask;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline void truncate_right (u32 w0[4], u32 w1[4], const u32 len)
Initial commit
2015-12-04 14:47:52 +00:00
{
const u32 tmp = (1 << ((len % 4) * 8)) - 1;
switch (len / 4)
{
case 0: w0[0] &= tmp;
w0[1] = 0;
w0[2] = 0;
w0[3] = 0;
w1[0] = 0;
w1[1] = 0;
w1[2] = 0;
w1[3] = 0;
break;
case 1: w0[1] &= tmp;
w0[2] = 0;
w0[3] = 0;
w1[0] = 0;
w1[1] = 0;
w1[2] = 0;
w1[3] = 0;
break;
case 2: w0[2] &= tmp;
w0[3] = 0;
w1[0] = 0;
w1[1] = 0;
w1[2] = 0;
w1[3] = 0;
break;
case 3: w0[3] &= tmp;
w1[0] = 0;
w1[1] = 0;
w1[2] = 0;
w1[3] = 0;
break;
case 4: w1[0] &= tmp;
w1[1] = 0;
w1[2] = 0;
w1[3] = 0;
break;
case 5: w1[1] &= tmp;
w1[2] = 0;
w1[3] = 0;
break;
case 6: w1[2] &= tmp;
w1[3] = 0;
break;
case 7: w1[3] &= tmp;
break;
}
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline void truncate_left (u32 w0[4], u32 w1[4], const u32 len)
Initial commit
2015-12-04 14:47:52 +00:00
{
const u32 tmp = ~((1 << ((len % 4) * 8)) - 1);
switch (len / 4)
{
case 0: w0[0] &= tmp;
break;
case 1: w0[0] = 0;
w0[1] &= tmp;
break;
case 2: w0[0] = 0;
w0[1] = 0;
w0[2] &= tmp;
break;
case 3: w0[0] = 0;
w0[1] = 0;
w0[2] = 0;
w0[3] &= tmp;
break;
case 4: w0[0] = 0;
w0[1] = 0;
w0[2] = 0;
w0[3] = 0;
w1[0] &= tmp;
break;
case 5: w0[0] = 0;
w0[1] = 0;
w0[2] = 0;
w0[3] = 0;
w1[0] = 0;
w1[1] &= tmp;
break;
case 6: w0[0] = 0;
w0[1] = 0;
w0[2] = 0;
w0[3] = 0;
w1[0] = 0;
w1[1] = 0;
w1[2] &= tmp;
break;
case 7: w0[0] = 0;
w0[1] = 0;
w0[2] = 0;
w0[3] = 0;
w1[0] = 0;
w1[1] = 0;
w1[2] = 0;
w1[3] &= tmp;
break;
}
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline void lshift_block (const u32 in0[4], const u32 in1[4], u32 out0[4], u32 out1[4])
Initial commit
2015-12-04 14:47:52 +00:00
{
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
out0[0] = __byte_perm_S (in0[0], in0[1], 0x4321);
out0[1] = __byte_perm_S (in0[1], in0[2], 0x4321);
out0[2] = __byte_perm_S (in0[2], in0[3], 0x4321);
out0[3] = __byte_perm_S (in0[3], in1[0], 0x4321);
out1[0] = __byte_perm_S (in1[0], in1[1], 0x4321);
out1[1] = __byte_perm_S (in1[1], in1[2], 0x4321);
out1[2] = __byte_perm_S (in1[2], in1[3], 0x4321);
out1[3] = __byte_perm_S (in1[3], 0, 0x4321);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
out0[0] = amd_bytealign_S (in0[1], in0[0], 1);
out0[1] = amd_bytealign_S (in0[2], in0[1], 1);
out0[2] = amd_bytealign_S (in0[3], in0[2], 1);
out0[3] = amd_bytealign_S (in1[0], in0[3], 1);
out1[0] = amd_bytealign_S (in1[1], in1[0], 1);
out1[1] = amd_bytealign_S (in1[2], in1[1], 1);
out1[2] = amd_bytealign_S (in1[3], in1[2], 1);
out1[3] = amd_bytealign_S ( 0, in1[3], 1);
Initial commit
2015-12-04 14:47:52 +00:00
#endif
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline void rshift_block (const u32 in0[4], const u32 in1[4], u32 out0[4], u32 out1[4])
Initial commit
2015-12-04 14:47:52 +00:00
{
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
out1[3] = __byte_perm_S (in1[2], in1[3], 0x6543);
out1[2] = __byte_perm_S (in1[1], in1[2], 0x6543);
out1[1] = __byte_perm_S (in1[0], in1[1], 0x6543);
out1[0] = __byte_perm_S (in0[3], in1[0], 0x6543);
out0[3] = __byte_perm_S (in0[2], in0[3], 0x6543);
out0[2] = __byte_perm_S (in0[1], in0[2], 0x6543);
out0[1] = __byte_perm_S (in0[0], in0[1], 0x6543);
out0[0] = __byte_perm_S ( 0, in0[0], 0x6543);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
out1[3] = amd_bytealign_S (in1[3], in1[2], 3);
out1[2] = amd_bytealign_S (in1[2], in1[1], 3);
out1[1] = amd_bytealign_S (in1[1], in1[0], 3);
out1[0] = amd_bytealign_S (in1[0], in0[3], 3);
out0[3] = amd_bytealign_S (in0[3], in0[2], 3);
out0[2] = amd_bytealign_S (in0[2], in0[1], 3);
out0[1] = amd_bytealign_S (in0[1], in0[0], 3);
out0[0] = amd_bytealign_S (in0[0], 0, 3);
Initial commit
2015-12-04 14:47:52 +00:00
#endif
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline void lshift_block_N (const u32 in0[4], const u32 in1[4], u32 out0[4], u32 out1[4], const u32 num)
Initial commit
2015-12-04 14:47:52 +00:00
{
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Initial commit
2015-12-04 14:47:52 +00:00
switch (num)
{
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
case 0: out0[0] = in0[0];
Initial commit
2015-12-04 14:47:52 +00:00
out0[1] = in0[1];
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[2] = in0[2];
out0[3] = in0[3];
out1[0] = in1[0];
out1[1] = in1[1];
out1[2] = in1[2];
out1[3] = in1[3];
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 1: out0[0] = __byte_perm_S (in0[0], in0[1], 0x4321);
out0[1] = __byte_perm_S (in0[1], in0[2], 0x4321);
out0[2] = __byte_perm_S (in0[2], in0[3], 0x4321);
out0[3] = __byte_perm_S (in0[3], in1[0], 0x4321);
out1[0] = __byte_perm_S (in1[0], in1[1], 0x4321);
out1[1] = __byte_perm_S (in1[1], in1[2], 0x4321);
out1[2] = __byte_perm_S (in1[2], in1[3], 0x4321);
out1[3] = __byte_perm_S (in1[3], 0, 0x4321);
break;
case 2: out0[0] = __byte_perm_S (in0[0], in0[1], 0x5432);
out0[1] = __byte_perm_S (in0[1], in0[2], 0x5432);
out0[2] = __byte_perm_S (in0[2], in0[3], 0x5432);
out0[3] = __byte_perm_S (in0[3], in1[0], 0x5432);
out1[0] = __byte_perm_S (in1[0], in1[1], 0x5432);
out1[1] = __byte_perm_S (in1[1], in1[2], 0x5432);
out1[2] = __byte_perm_S (in1[2], in1[3], 0x5432);
out1[3] = __byte_perm_S (in1[3], 0, 0x5432);
break;
case 3: out0[0] = __byte_perm_S (in0[0], in0[1], 0x6543);
out0[1] = __byte_perm_S (in0[1], in0[2], 0x6543);
out0[2] = __byte_perm_S (in0[2], in0[3], 0x6543);
out0[3] = __byte_perm_S (in0[3], in1[0], 0x6543);
out1[0] = __byte_perm_S (in1[0], in1[1], 0x6543);
out1[1] = __byte_perm_S (in1[1], in1[2], 0x6543);
out1[2] = __byte_perm_S (in1[2], in1[3], 0x6543);
out1[3] = __byte_perm_S (in1[3], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
case 4: out0[0] = in0[1];
out0[1] = in0[2];
out0[2] = in0[3];
out0[3] = in1[0];
out1[0] = in1[1];
out1[1] = in1[2];
out1[2] = in1[3];
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 5: out0[0] = __byte_perm_S (in0[1], in0[2], 0x4321);
out0[1] = __byte_perm_S (in0[2], in0[3], 0x4321);
out0[2] = __byte_perm_S (in0[3], in1[0], 0x4321);
out0[3] = __byte_perm_S (in1[0], in1[1], 0x4321);
out1[0] = __byte_perm_S (in1[1], in1[2], 0x4321);
out1[1] = __byte_perm_S (in1[2], in1[3], 0x4321);
out1[2] = __byte_perm_S (in1[3], 0, 0x4321);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 6: out0[0] = __byte_perm_S (in0[1], in0[2], 0x5432);
out0[1] = __byte_perm_S (in0[2], in0[3], 0x5432);
out0[2] = __byte_perm_S (in0[3], in1[0], 0x5432);
out0[3] = __byte_perm_S (in1[0], in1[1], 0x5432);
out1[0] = __byte_perm_S (in1[1], in1[2], 0x5432);
out1[1] = __byte_perm_S (in1[2], in1[3], 0x5432);
out1[2] = __byte_perm_S (in1[3], 0, 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 7: out0[0] = __byte_perm_S (in0[1], in0[2], 0x6543);
out0[1] = __byte_perm_S (in0[2], in0[3], 0x6543);
out0[2] = __byte_perm_S (in0[3], in1[0], 0x6543);
out0[3] = __byte_perm_S (in1[0], in1[1], 0x6543);
out1[0] = __byte_perm_S (in1[1], in1[2], 0x6543);
out1[1] = __byte_perm_S (in1[2], in1[3], 0x6543);
out1[2] = __byte_perm_S (in1[3], 0, 0x6543);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
case 8: out0[0] = in0[2];
out0[1] = in0[3];
out0[2] = in1[0];
out0[3] = in1[1];
out1[0] = in1[2];
out1[1] = in1[3];
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 9: out0[0] = __byte_perm_S (in0[2], in0[3], 0x4321);
out0[1] = __byte_perm_S (in0[3], in1[0], 0x4321);
out0[2] = __byte_perm_S (in1[0], in1[1], 0x4321);
out0[3] = __byte_perm_S (in1[1], in1[2], 0x4321);
out1[0] = __byte_perm_S (in1[2], in1[3], 0x4321);
out1[1] = __byte_perm_S (in1[3], 0, 0x4321);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 10: out0[0] = __byte_perm_S (in0[2], in0[3], 0x5432);
out0[1] = __byte_perm_S (in0[3], in1[0], 0x5432);
out0[2] = __byte_perm_S (in1[0], in1[1], 0x5432);
out0[3] = __byte_perm_S (in1[1], in1[2], 0x5432);
out1[0] = __byte_perm_S (in1[2], in1[3], 0x5432);
out1[1] = __byte_perm_S (in1[3], 0, 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 11: out0[0] = __byte_perm_S (in0[2], in0[3], 0x6543);
out0[1] = __byte_perm_S (in0[3], in1[0], 0x6543);
out0[2] = __byte_perm_S (in1[0], in1[1], 0x6543);
out0[3] = __byte_perm_S (in1[1], in1[2], 0x6543);
out1[0] = __byte_perm_S (in1[2], in1[3], 0x6543);
out1[1] = __byte_perm_S (in1[3], 0, 0x6543);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
case 12: out0[0] = in0[3];
out0[1] = in1[0];
out0[2] = in1[1];
out0[3] = in1[2];
out1[0] = in1[3];
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
case 13:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
out0[0] = __byte_perm_S (in0[3], in1[0], 0x4321);
out0[1] = __byte_perm_S (in1[0], in1[1], 0x4321);
out0[2] = __byte_perm_S (in1[1], in1[2], 0x4321);
out0[3] = __byte_perm_S (in1[2], in1[3], 0x4321);
out1[0] = __byte_perm_S (in1[3], 0, 0x4321);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 14: out0[0] = __byte_perm_S (in0[3], in1[0], 0x5432);
out0[1] = __byte_perm_S (in1[0], in1[1], 0x5432);
out0[2] = __byte_perm_S (in1[1], in1[2], 0x5432);
out0[3] = __byte_perm_S (in1[2], in1[3], 0x5432);
out1[0] = __byte_perm_S (in1[3], 0, 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 15: out0[0] = __byte_perm_S (in0[3], in1[0], 0x6543);
out0[1] = __byte_perm_S (in1[0], in1[1], 0x6543);
out0[2] = __byte_perm_S (in1[1], in1[2], 0x6543);
out0[3] = __byte_perm_S (in1[2], in1[3], 0x6543);
out1[0] = __byte_perm_S (in1[3], 0, 0x6543);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
case 16: out0[0] = in1[0];
out0[1] = in1[1];
out0[2] = in1[2];
out0[3] = in1[3];
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 17: out0[0] = __byte_perm_S (in1[0], in1[1], 0x4321);
out0[1] = __byte_perm_S (in1[1], in1[2], 0x4321);
out0[2] = __byte_perm_S (in1[2], in1[3], 0x4321);
out0[3] = __byte_perm_S (in1[3], 0, 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 18: out0[0] = __byte_perm_S (in1[0], in1[1], 0x5432);
out0[1] = __byte_perm_S (in1[1], in1[2], 0x5432);
out0[2] = __byte_perm_S (in1[2], in1[3], 0x5432);
out0[3] = __byte_perm_S (in1[3], 0, 0x5432);
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 19: out0[0] = __byte_perm_S (in1[0], in1[1], 0x6543);
out0[1] = __byte_perm_S (in1[1], in1[2], 0x6543);
out0[2] = __byte_perm_S (in1[2], in1[3], 0x6543);
out0[3] = __byte_perm_S (in1[3], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
out1[1] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
case 20: out0[0] = in1[1];
out0[1] = in1[2];
out0[2] = in1[3];
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 21: out0[0] = __byte_perm_S (in1[1], in1[2], 0x4321);
out0[1] = __byte_perm_S (in1[2], in1[3], 0x4321);
out0[2] = __byte_perm_S (in1[3], 0, 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[0] = 0;
out1[1] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 22: out0[0] = __byte_perm_S (in1[1], in1[2], 0x5432);
out0[1] = __byte_perm_S (in1[2], in1[3], 0x5432);
out0[2] = __byte_perm_S (in1[3], 0, 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[3] = 0;
out1[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[1] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[2] = 0;
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 23: out0[0] = __byte_perm_S (in1[1], in1[2], 0x6543);
out0[1] = __byte_perm_S (in1[2], in1[3], 0x6543);
out0[2] = __byte_perm_S (in1[3], 0, 0x6543);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
case 24: out0[0] = in1[2];
out0[1] = in1[3];
out0[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 25: out0[0] = __byte_perm_S (in1[2], in1[3], 0x4321);
out0[1] = __byte_perm_S (in1[3], 0, 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 26: out0[0] = __byte_perm_S (in1[2], in1[3], 0x5432);
out0[1] = __byte_perm_S (in1[3], 0, 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 27: out0[0] = __byte_perm_S (in1[2], in1[3], 0x6543);
out0[1] = __byte_perm_S (in1[3], 0, 0x6543);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[1] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[2] = 0;
out1[3] = 0;
break;
case 28: out0[0] = in1[3];
out0[1] = 0;
out0[2] = 0;
out0[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 29: out0[0] = __byte_perm_S (in1[3], 0, 0x4321);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[1] = 0;
out0[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 30: out0[0] = __byte_perm_S (in1[3], 0, 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[1] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 31: out0[0] = __byte_perm_S (in1[3], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
out0[1] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
Initial commit
2015-12-04 14:47:52 +00:00
switch (num)
{
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
case 0: out0[0] = in0[0];
Initial commit
2015-12-04 14:47:52 +00:00
out0[1] = in0[1];
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[2] = in0[2];
out0[3] = in0[3];
out1[0] = in1[0];
out1[1] = in1[1];
out1[2] = in1[2];
out1[3] = in1[3];
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 1: out0[0] = amd_bytealign_S (in0[1], in0[0], 1);
out0[1] = amd_bytealign_S (in0[2], in0[1], 1);
out0[2] = amd_bytealign_S (in0[3], in0[2], 1);
out0[3] = amd_bytealign_S (in1[0], in0[3], 1);
out1[0] = amd_bytealign_S (in1[1], in1[0], 1);
out1[1] = amd_bytealign_S (in1[2], in1[1], 1);
out1[2] = amd_bytealign_S (in1[3], in1[2], 1);
out1[3] = amd_bytealign_S ( 0, in1[3], 1);
break;
case 2: out0[0] = amd_bytealign_S (in0[1], in0[0], 2);
out0[1] = amd_bytealign_S (in0[2], in0[1], 2);
out0[2] = amd_bytealign_S (in0[3], in0[2], 2);
out0[3] = amd_bytealign_S (in1[0], in0[3], 2);
out1[0] = amd_bytealign_S (in1[1], in1[0], 2);
out1[1] = amd_bytealign_S (in1[2], in1[1], 2);
out1[2] = amd_bytealign_S (in1[3], in1[2], 2);
out1[3] = amd_bytealign_S ( 0, in1[3], 2);
break;
case 3: out0[0] = amd_bytealign_S (in0[1], in0[0], 3);
out0[1] = amd_bytealign_S (in0[2], in0[1], 3);
out0[2] = amd_bytealign_S (in0[3], in0[2], 3);
out0[3] = amd_bytealign_S (in1[0], in0[3], 3);
out1[0] = amd_bytealign_S (in1[1], in1[0], 3);
out1[1] = amd_bytealign_S (in1[2], in1[1], 3);
out1[2] = amd_bytealign_S (in1[3], in1[2], 3);
out1[3] = amd_bytealign_S ( 0, in1[3], 3);
Initial commit
2015-12-04 14:47:52 +00:00
break;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
case 4: out0[0] = in0[1];
out0[1] = in0[2];
out0[2] = in0[3];
out0[3] = in1[0];
out1[0] = in1[1];
out1[1] = in1[2];
out1[2] = in1[3];
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 5: out0[0] = amd_bytealign_S (in0[2], in0[1], 1);
out0[1] = amd_bytealign_S (in0[3], in0[2], 1);
out0[2] = amd_bytealign_S (in1[0], in0[3], 1);
out0[3] = amd_bytealign_S (in1[1], in1[0], 1);
out1[0] = amd_bytealign_S (in1[2], in1[1], 1);
out1[1] = amd_bytealign_S (in1[3], in1[2], 1);
out1[2] = amd_bytealign_S ( 0, in1[3], 1);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 6: out0[0] = amd_bytealign_S (in0[2], in0[1], 2);
out0[1] = amd_bytealign_S (in0[3], in0[2], 2);
out0[2] = amd_bytealign_S (in1[0], in0[3], 2);
out0[3] = amd_bytealign_S (in1[1], in1[0], 2);
out1[0] = amd_bytealign_S (in1[2], in1[1], 2);
out1[1] = amd_bytealign_S (in1[3], in1[2], 2);
out1[2] = amd_bytealign_S ( 0, in1[3], 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 7: out0[0] = amd_bytealign_S (in0[2], in0[1], 3);
out0[1] = amd_bytealign_S (in0[3], in0[2], 3);
out0[2] = amd_bytealign_S (in1[0], in0[3], 3);
out0[3] = amd_bytealign_S (in1[1], in1[0], 3);
out1[0] = amd_bytealign_S (in1[2], in1[1], 3);
out1[1] = amd_bytealign_S (in1[3], in1[2], 3);
out1[2] = amd_bytealign_S ( 0, in1[3], 3);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
case 8: out0[0] = in0[2];
out0[1] = in0[3];
out0[2] = in1[0];
out0[3] = in1[1];
out1[0] = in1[2];
out1[1] = in1[3];
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 9: out0[0] = amd_bytealign_S (in0[3], in0[2], 1);
out0[1] = amd_bytealign_S (in1[0], in0[3], 1);
out0[2] = amd_bytealign_S (in1[1], in1[0], 1);
out0[3] = amd_bytealign_S (in1[2], in1[1], 1);
out1[0] = amd_bytealign_S (in1[3], in1[2], 1);
out1[1] = amd_bytealign_S ( 0, in1[3], 1);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 10: out0[0] = amd_bytealign_S (in0[3], in0[2], 2);
out0[1] = amd_bytealign_S (in1[0], in0[3], 2);
out0[2] = amd_bytealign_S (in1[1], in1[0], 2);
out0[3] = amd_bytealign_S (in1[2], in1[1], 2);
out1[0] = amd_bytealign_S (in1[3], in1[2], 2);
out1[1] = amd_bytealign_S ( 0, in1[3], 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 11: out0[0] = amd_bytealign_S (in0[3], in0[2], 3);
out0[1] = amd_bytealign_S (in1[0], in0[3], 3);
out0[2] = amd_bytealign_S (in1[1], in1[0], 3);
out0[3] = amd_bytealign_S (in1[2], in1[1], 3);
out1[0] = amd_bytealign_S (in1[3], in1[2], 3);
out1[1] = amd_bytealign_S ( 0, in1[3], 3);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
case 12: out0[0] = in0[3];
out0[1] = in1[0];
out0[2] = in1[1];
out0[3] = in1[2];
out1[0] = in1[3];
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 13: out0[0] = amd_bytealign_S (in1[0], in0[3], 1);
out0[1] = amd_bytealign_S (in1[1], in1[0], 1);
out0[2] = amd_bytealign_S (in1[2], in1[1], 1);
out0[3] = amd_bytealign_S (in1[3], in1[2], 1);
out1[0] = amd_bytealign_S ( 0, in1[3], 1);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 14: out0[0] = amd_bytealign_S (in1[0], in0[3], 2);
out0[1] = amd_bytealign_S (in1[1], in1[0], 2);
out0[2] = amd_bytealign_S (in1[2], in1[1], 2);
out0[3] = amd_bytealign_S (in1[3], in1[2], 2);
out1[0] = amd_bytealign_S ( 0, in1[3], 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 15: out0[0] = amd_bytealign_S (in1[0], in0[3], 3);
out0[1] = amd_bytealign_S (in1[1], in1[0], 3);
out0[2] = amd_bytealign_S (in1[2], in1[1], 3);
out0[3] = amd_bytealign_S (in1[3], in1[2], 3);
out1[0] = amd_bytealign_S ( 0, in1[3], 3);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
case 16: out0[0] = in1[0];
out0[1] = in1[1];
out0[2] = in1[2];
out0[3] = in1[3];
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 17: out0[0] = amd_bytealign_S (in1[1], in1[0], 1);
out0[1] = amd_bytealign_S (in1[2], in1[1], 1);
out0[2] = amd_bytealign_S (in1[3], in1[2], 1);
out0[3] = amd_bytealign_S ( 0, in1[3], 1);
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 18: out0[0] = amd_bytealign_S (in1[1], in1[0], 2);
out0[1] = amd_bytealign_S (in1[2], in1[1], 2);
out0[2] = amd_bytealign_S (in1[3], in1[2], 2);
out0[3] = amd_bytealign_S ( 0, in1[3], 2);
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 19: out0[0] = amd_bytealign_S (in1[1], in1[0], 3);
out0[1] = amd_bytealign_S (in1[2], in1[1], 3);
out0[2] = amd_bytealign_S (in1[3], in1[2], 3);
out0[3] = amd_bytealign_S ( 0, in1[3], 3);
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
out1[1] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
case 20: out0[0] = in1[1];
out0[1] = in1[2];
out0[2] = in1[3];
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 21: out0[0] = amd_bytealign_S (in1[2], in1[1], 1);
out0[1] = amd_bytealign_S (in1[3], in1[2], 1);
out0[2] = amd_bytealign_S ( 0, in1[3], 1);
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[0] = 0;
out1[1] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 22: out0[0] = amd_bytealign_S (in1[2], in1[1], 2);
out0[1] = amd_bytealign_S (in1[3], in1[2], 2);
out0[2] = amd_bytealign_S ( 0, in1[3], 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[3] = 0;
out1[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[1] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[2] = 0;
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 23: out0[0] = amd_bytealign_S (in1[2], in1[1], 3);
out0[1] = amd_bytealign_S (in1[3], in1[2], 3);
out0[2] = amd_bytealign_S ( 0, in1[3], 3);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
case 24: out0[0] = in1[2];
out0[1] = in1[3];
out0[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 25: out0[0] = amd_bytealign_S (in1[3], in1[2], 1);
out0[1] = amd_bytealign_S ( 0, in1[3], 1);
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 26: out0[0] = amd_bytealign_S (in1[3], in1[2], 2);
out0[1] = amd_bytealign_S ( 0, in1[3], 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 27: out0[0] = amd_bytealign_S (in1[3], in1[2], 3);
out0[1] = amd_bytealign_S ( 0, in1[3], 3);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[1] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[2] = 0;
out1[3] = 0;
break;
case 28: out0[0] = in1[3];
out0[1] = 0;
out0[2] = 0;
out0[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 29: out0[0] = amd_bytealign_S ( 0, in1[3], 1);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[1] = 0;
out0[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 30: out0[0] = amd_bytealign_S ( 0, in1[3], 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[1] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 31: out0[0] = amd_bytealign_S ( 0, in1[3], 3);
Initial commit
2015-12-04 14:47:52 +00:00
out0[1] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[2] = 0;
out0[3] = 0;
out1[0] = 0;
out1[1] = 0;
out1[2] = 0;
out1[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
#endif
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline void rshift_block_N (const u32 in0[4], const u32 in1[4], u32 out0[4], u32 out1[4], const u32 num)
Initial commit
2015-12-04 14:47:52 +00:00
{
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Initial commit
2015-12-04 14:47:52 +00:00
switch (num)
{
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
case 0: out1[3] = in1[3];
Initial commit
2015-12-04 14:47:52 +00:00
out1[2] = in1[2];
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = in1[1];
out1[0] = in1[0];
out0[3] = in0[3];
out0[2] = in0[2];
out0[1] = in0[1];
out0[0] = in0[0];
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 1: out1[3] = __byte_perm_S (in1[2], in1[3], 0x6543);
out1[2] = __byte_perm_S (in1[1], in1[2], 0x6543);
out1[1] = __byte_perm_S (in1[0], in1[1], 0x6543);
out1[0] = __byte_perm_S (in0[3], in1[0], 0x6543);
out0[3] = __byte_perm_S (in0[2], in0[3], 0x6543);
out0[2] = __byte_perm_S (in0[1], in0[2], 0x6543);
out0[1] = __byte_perm_S (in0[0], in0[1], 0x6543);
out0[0] = __byte_perm_S ( 0, in0[0], 0x6543);
break;
case 2: out1[3] = __byte_perm_S (in1[2], in1[3], 0x5432);
out1[2] = __byte_perm_S (in1[1], in1[2], 0x5432);
out1[1] = __byte_perm_S (in1[0], in1[1], 0x5432);
out1[0] = __byte_perm_S (in0[3], in1[0], 0x5432);
out0[3] = __byte_perm_S (in0[2], in0[3], 0x5432);
out0[2] = __byte_perm_S (in0[1], in0[2], 0x5432);
out0[1] = __byte_perm_S (in0[0], in0[1], 0x5432);
out0[0] = __byte_perm_S ( 0, in0[0], 0x5432);
break;
case 3: out1[3] = __byte_perm_S (in1[2], in1[3], 0x4321);
out1[2] = __byte_perm_S (in1[1], in1[2], 0x4321);
out1[1] = __byte_perm_S (in1[0], in1[1], 0x4321);
out1[0] = __byte_perm_S (in0[3], in1[0], 0x4321);
out0[3] = __byte_perm_S (in0[2], in0[3], 0x4321);
out0[2] = __byte_perm_S (in0[1], in0[2], 0x4321);
out0[1] = __byte_perm_S (in0[0], in0[1], 0x4321);
out0[0] = __byte_perm_S ( 0, in0[0], 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
case 4: out1[3] = in1[2];
out1[2] = in1[1];
out1[1] = in1[0];
out1[0] = in0[3];
out0[3] = in0[2];
out0[2] = in0[1];
out0[1] = in0[0];
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 5: out1[3] = __byte_perm_S (in1[1], in1[2], 0x6543);
out1[2] = __byte_perm_S (in1[0], in1[1], 0x6543);
out1[1] = __byte_perm_S (in0[3], in1[0], 0x6543);
out1[0] = __byte_perm_S (in0[2], in0[3], 0x6543);
out0[3] = __byte_perm_S (in0[1], in0[2], 0x6543);
out0[2] = __byte_perm_S (in0[0], in0[1], 0x6543);
out0[1] = __byte_perm_S ( 0, in0[0], 0x6543);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 6: out1[3] = __byte_perm_S (in1[1], in1[2], 0x5432);
out1[2] = __byte_perm_S (in1[0], in1[1], 0x5432);
out1[1] = __byte_perm_S (in0[3], in1[0], 0x5432);
out1[0] = __byte_perm_S (in0[2], in0[3], 0x5432);
out0[3] = __byte_perm_S (in0[1], in0[2], 0x5432);
out0[2] = __byte_perm_S (in0[0], in0[1], 0x5432);
out0[1] = __byte_perm_S ( 0, in0[0], 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 7: out1[3] = __byte_perm_S (in1[1], in1[2], 0x4321);
out1[2] = __byte_perm_S (in1[0], in1[1], 0x4321);
out1[1] = __byte_perm_S (in0[3], in1[0], 0x4321);
out1[0] = __byte_perm_S (in0[2], in0[3], 0x4321);
out0[3] = __byte_perm_S (in0[1], in0[2], 0x4321);
out0[2] = __byte_perm_S (in0[0], in0[1], 0x4321);
out0[1] = __byte_perm_S ( 0, in0[0], 0x4321);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
case 8: out1[3] = in1[1];
out1[2] = in1[0];
out1[1] = in0[3];
out1[0] = in0[2];
out0[3] = in0[1];
out0[2] = in0[0];
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 9: out1[3] = __byte_perm_S (in1[0], in1[1], 0x6543);
out1[2] = __byte_perm_S (in0[3], in1[0], 0x6543);
out1[1] = __byte_perm_S (in0[2], in0[3], 0x6543);
out1[0] = __byte_perm_S (in0[1], in0[2], 0x6543);
out0[3] = __byte_perm_S (in0[0], in0[1], 0x6543);
out0[2] = __byte_perm_S ( 0, in0[0], 0x6543);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 10: out1[3] = __byte_perm_S (in1[0], in1[1], 0x5432);
out1[2] = __byte_perm_S (in0[3], in1[0], 0x5432);
out1[1] = __byte_perm_S (in0[2], in0[3], 0x5432);
out1[0] = __byte_perm_S (in0[1], in0[2], 0x5432);
out0[3] = __byte_perm_S (in0[0], in0[1], 0x5432);
out0[2] = __byte_perm_S ( 0, in0[0], 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 11: out1[3] = __byte_perm_S (in1[0], in1[1], 0x4321);
out1[2] = __byte_perm_S (in0[3], in1[0], 0x4321);
out1[1] = __byte_perm_S (in0[2], in0[3], 0x4321);
out1[0] = __byte_perm_S (in0[1], in0[2], 0x4321);
out0[3] = __byte_perm_S (in0[0], in0[1], 0x4321);
out0[2] = __byte_perm_S ( 0, in0[0], 0x4321);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
case 12: out1[3] = in1[0];
out1[2] = in0[3];
out1[1] = in0[2];
out1[0] = in0[1];
out0[3] = in0[0];
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 13: out1[3] = __byte_perm_S (in0[3], in1[0], 0x6543);
out1[2] = __byte_perm_S (in0[2], in0[3], 0x6543);
out1[1] = __byte_perm_S (in0[1], in0[2], 0x6543);
out1[0] = __byte_perm_S (in0[0], in0[1], 0x6543);
out0[3] = __byte_perm_S ( 0, in0[0], 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 14: out1[3] = __byte_perm_S (in0[3], in1[0], 0x5432);
out1[2] = __byte_perm_S (in0[2], in0[3], 0x5432);
out1[1] = __byte_perm_S (in0[1], in0[2], 0x5432);
out1[0] = __byte_perm_S (in0[0], in0[1], 0x5432);
out0[3] = __byte_perm_S ( 0, in0[0], 0x5432);
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
out0[1] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 15: out1[3] = __byte_perm_S (in0[3], in1[0], 0x4321);
out1[2] = __byte_perm_S (in0[2], in0[3], 0x4321);
out1[1] = __byte_perm_S (in0[1], in0[2], 0x4321);
out1[0] = __byte_perm_S (in0[0], in0[1], 0x4321);
out0[3] = __byte_perm_S ( 0, in0[0], 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[1] = 0;
out0[0] = 0;
break;
case 16: out1[3] = in0[3];
out1[2] = in0[2];
out1[1] = in0[1];
out1[0] = in0[0];
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 17: out1[3] = __byte_perm_S (in0[2], in0[3], 0x6543);
out1[2] = __byte_perm_S (in0[1], in0[2], 0x6543);
out1[1] = __byte_perm_S (in0[0], in0[1], 0x6543);
out1[0] = __byte_perm_S ( 0, in0[0], 0x6543);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[3] = 0;
out0[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[1] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 18: out1[3] = __byte_perm_S (in0[2], in0[3], 0x5432);
out1[2] = __byte_perm_S (in0[1], in0[2], 0x5432);
out1[1] = __byte_perm_S (in0[0], in0[1], 0x5432);
out1[0] = __byte_perm_S ( 0, in0[0], 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 19: out1[3] = __byte_perm_S (in0[2], in0[3], 0x4321);
out1[2] = __byte_perm_S (in0[1], in0[2], 0x4321);
out1[1] = __byte_perm_S (in0[0], in0[1], 0x4321);
out1[0] = __byte_perm_S ( 0, in0[0], 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
case 20: out1[3] = in0[2];
out1[2] = in0[1];
out1[1] = in0[0];
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 21: out1[3] = __byte_perm_S (in0[1], in0[2], 0x6543);
out1[2] = __byte_perm_S (in0[0], in0[1], 0x6543);
out1[1] = __byte_perm_S ( 0, in0[0], 0x6543);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[1] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 22: out1[3] = __byte_perm_S (in0[1], in0[2], 0x5432);
out1[2] = __byte_perm_S (in0[0], in0[1], 0x5432);
out1[1] = __byte_perm_S ( 0, in0[0], 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[0] = 0;
out0[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 23: out1[3] = __byte_perm_S (in0[1], in0[2], 0x4321);
out1[2] = __byte_perm_S (in0[0], in0[1], 0x4321);
out1[1] = __byte_perm_S ( 0, in0[0], 0x4321);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
case 24: out1[3] = in0[1];
out1[2] = in0[0];
out1[1] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 25: out1[3] = __byte_perm_S (in0[0], in0[1], 0x6543);
out1[2] = __byte_perm_S ( 0, in0[0], 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
out1[1] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 26: out1[3] = __byte_perm_S (in0[0], in0[1], 0x5432);
out1[2] = __byte_perm_S ( 0, in0[0], 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[1] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 27: out1[3] = __byte_perm_S (in0[0], in0[1], 0x4321);
out1[2] = __byte_perm_S ( 0, in0[0], 0x4321);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[1] = 0;
out0[0] = 0;
break;
case 28: out1[3] = in0[0];
out1[2] = 0;
out1[1] = 0;
out1[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 29: out1[3] = __byte_perm_S ( 0, in0[0], 0x6543);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[2] = 0;
out1[1] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 30: out1[3] = __byte_perm_S ( 0, in0[0], 0x5432);
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[1] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 31: out1[3] = __byte_perm_S ( 0, in0[0], 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
out1[2] = 0;
Fix bug in rule-engine for NV, left shifts and right shifts were switched
2016-01-16 12:34:54 +00:00
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
Initial commit
2015-12-04 14:47:52 +00:00
switch (num)
{
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
case 0: out1[3] = in1[3];
Initial commit
2015-12-04 14:47:52 +00:00
out1[2] = in1[2];
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = in1[1];
out1[0] = in1[0];
out0[3] = in0[3];
out0[2] = in0[2];
out0[1] = in0[1];
out0[0] = in0[0];
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 1: out1[3] = amd_bytealign_S (in1[3], in1[2], 3);
out1[2] = amd_bytealign_S (in1[2], in1[1], 3);
out1[1] = amd_bytealign_S (in1[1], in1[0], 3);
out1[0] = amd_bytealign_S (in1[0], in0[3], 3);
out0[3] = amd_bytealign_S (in0[3], in0[2], 3);
out0[2] = amd_bytealign_S (in0[2], in0[1], 3);
out0[1] = amd_bytealign_S (in0[1], in0[0], 3);
out0[0] = amd_bytealign_S (in0[0], 0, 3);
break;
case 2: out1[3] = amd_bytealign_S (in1[3], in1[2], 2);
out1[2] = amd_bytealign_S (in1[2], in1[1], 2);
out1[1] = amd_bytealign_S (in1[1], in1[0], 2);
out1[0] = amd_bytealign_S (in1[0], in0[3], 2);
out0[3] = amd_bytealign_S (in0[3], in0[2], 2);
out0[2] = amd_bytealign_S (in0[2], in0[1], 2);
out0[1] = amd_bytealign_S (in0[1], in0[0], 2);
out0[0] = amd_bytealign_S (in0[0], 0, 2);
break;
case 3: out1[3] = amd_bytealign_S (in1[3], in1[2], 1);
out1[2] = amd_bytealign_S (in1[2], in1[1], 1);
out1[1] = amd_bytealign_S (in1[1], in1[0], 1);
out1[0] = amd_bytealign_S (in1[0], in0[3], 1);
out0[3] = amd_bytealign_S (in0[3], in0[2], 1);
out0[2] = amd_bytealign_S (in0[2], in0[1], 1);
out0[1] = amd_bytealign_S (in0[1], in0[0], 1);
out0[0] = amd_bytealign_S (in0[0], 0, 1);
Initial commit
2015-12-04 14:47:52 +00:00
break;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
case 4: out1[3] = in1[2];
out1[2] = in1[1];
out1[1] = in1[0];
out1[0] = in0[3];
out0[3] = in0[2];
out0[2] = in0[1];
out0[1] = in0[0];
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 5: out1[3] = amd_bytealign_S (in1[2], in1[1], 3);
out1[2] = amd_bytealign_S (in1[1], in1[0], 3);
out1[1] = amd_bytealign_S (in1[0], in0[3], 3);
out1[0] = amd_bytealign_S (in0[3], in0[2], 3);
out0[3] = amd_bytealign_S (in0[2], in0[1], 3);
out0[2] = amd_bytealign_S (in0[1], in0[0], 3);
out0[1] = amd_bytealign_S (in0[0], 0, 3);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 6: out1[3] = amd_bytealign_S (in1[2], in1[1], 2);
out1[2] = amd_bytealign_S (in1[1], in1[0], 2);
out1[1] = amd_bytealign_S (in1[0], in0[3], 2);
out1[0] = amd_bytealign_S (in0[3], in0[2], 2);
out0[3] = amd_bytealign_S (in0[2], in0[1], 2);
out0[2] = amd_bytealign_S (in0[1], in0[0], 2);
out0[1] = amd_bytealign_S (in0[0], 0, 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 7: out1[3] = amd_bytealign_S (in1[2], in1[1], 1);
out1[2] = amd_bytealign_S (in1[1], in1[0], 1);
out1[1] = amd_bytealign_S (in1[0], in0[3], 1);
out1[0] = amd_bytealign_S (in0[3], in0[2], 1);
out0[3] = amd_bytealign_S (in0[2], in0[1], 1);
out0[2] = amd_bytealign_S (in0[1], in0[0], 1);
out0[1] = amd_bytealign_S (in0[0], 0, 1);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
case 8: out1[3] = in1[1];
out1[2] = in1[0];
out1[1] = in0[3];
out1[0] = in0[2];
out0[3] = in0[1];
out0[2] = in0[0];
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 9: out1[3] = amd_bytealign_S (in1[1], in1[0], 3);
out1[2] = amd_bytealign_S (in1[0], in0[3], 3);
out1[1] = amd_bytealign_S (in0[3], in0[2], 3);
out1[0] = amd_bytealign_S (in0[2], in0[1], 3);
out0[3] = amd_bytealign_S (in0[1], in0[0], 3);
out0[2] = amd_bytealign_S (in0[0], 0, 3);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 10: out1[3] = amd_bytealign_S (in1[1], in1[0], 2);
out1[2] = amd_bytealign_S (in1[0], in0[3], 2);
out1[1] = amd_bytealign_S (in0[3], in0[2], 2);
out1[0] = amd_bytealign_S (in0[2], in0[1], 2);
out0[3] = amd_bytealign_S (in0[1], in0[0], 2);
out0[2] = amd_bytealign_S (in0[0], 0, 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 11: out1[3] = amd_bytealign_S (in1[1], in1[0], 1);
out1[2] = amd_bytealign_S (in1[0], in0[3], 1);
out1[1] = amd_bytealign_S (in0[3], in0[2], 1);
out1[0] = amd_bytealign_S (in0[2], in0[1], 1);
out0[3] = amd_bytealign_S (in0[1], in0[0], 1);
out0[2] = amd_bytealign_S (in0[0], 0, 1);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
case 12: out1[3] = in1[0];
out1[2] = in0[3];
out1[1] = in0[2];
out1[0] = in0[1];
out0[3] = in0[0];
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 13: out1[3] = amd_bytealign_S (in1[0], in0[3], 3);
out1[2] = amd_bytealign_S (in0[3], in0[2], 3);
out1[1] = amd_bytealign_S (in0[2], in0[1], 3);
out1[0] = amd_bytealign_S (in0[1], in0[0], 3);
out0[3] = amd_bytealign_S (in0[0], 0, 3);
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 14: out1[3] = amd_bytealign_S (in1[0], in0[3], 2);
out1[2] = amd_bytealign_S (in0[3], in0[2], 2);
out1[1] = amd_bytealign_S (in0[2], in0[1], 2);
out1[0] = amd_bytealign_S (in0[1], in0[0], 2);
out0[3] = amd_bytealign_S (in0[0], 0, 2);
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 15: out1[3] = amd_bytealign_S (in1[0], in0[3], 1);
out1[2] = amd_bytealign_S (in0[3], in0[2], 1);
out1[1] = amd_bytealign_S (in0[2], in0[1], 1);
out1[0] = amd_bytealign_S (in0[1], in0[0], 1);
out0[3] = amd_bytealign_S (in0[0], 0, 1);
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
out0[1] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[0] = 0;
break;
case 16: out1[3] = in0[3];
out1[2] = in0[2];
out1[1] = in0[1];
out1[0] = in0[0];
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 17: out1[3] = amd_bytealign_S (in0[3], in0[2], 3);
out1[2] = amd_bytealign_S (in0[2], in0[1], 3);
out1[1] = amd_bytealign_S (in0[1], in0[0], 3);
out1[0] = amd_bytealign_S (in0[0], 0, 3);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[3] = 0;
out0[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[1] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 18: out1[3] = amd_bytealign_S (in0[3], in0[2], 2);
out1[2] = amd_bytealign_S (in0[2], in0[1], 2);
out1[1] = amd_bytealign_S (in0[1], in0[0], 2);
out1[0] = amd_bytealign_S (in0[0], 0, 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 19: out1[3] = amd_bytealign_S (in0[3], in0[2], 1);
out1[2] = amd_bytealign_S (in0[2], in0[1], 1);
out1[1] = amd_bytealign_S (in0[1], in0[0], 1);
out1[0] = amd_bytealign_S (in0[0], 0, 1);
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
case 20: out1[3] = in0[2];
out1[2] = in0[1];
out1[1] = in0[0];
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 21: out1[3] = amd_bytealign_S (in0[2], in0[1], 3);
out1[2] = amd_bytealign_S (in0[1], in0[0], 3);
out1[1] = amd_bytealign_S (in0[0], 0, 3);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[1] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 22: out1[3] = amd_bytealign_S (in0[2], in0[1], 2);
out1[2] = amd_bytealign_S (in0[1], in0[0], 2);
out1[1] = amd_bytealign_S (in0[0], 0, 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[0] = 0;
out0[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 23: out1[3] = amd_bytealign_S (in0[2], in0[1], 1);
out1[2] = amd_bytealign_S (in0[1], in0[0], 1);
out1[1] = amd_bytealign_S (in0[0], 0, 1);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
case 24: out1[3] = in0[1];
out1[2] = in0[0];
out1[1] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 25: out1[3] = amd_bytealign_S (in0[1], in0[0], 3);
out1[2] = amd_bytealign_S (in0[0], 0, 3);
Initial commit
2015-12-04 14:47:52 +00:00
out1[1] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 26: out1[3] = amd_bytealign_S (in0[1], in0[0], 2);
out1[2] = amd_bytealign_S (in0[0], 0, 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[1] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 27: out1[3] = amd_bytealign_S (in0[1], in0[0], 1);
out1[2] = amd_bytealign_S (in0[0], 0, 1);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[1] = 0;
out0[0] = 0;
break;
case 28: out1[3] = in0[0];
out1[2] = 0;
out1[1] = 0;
out1[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out0[3] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 29: out1[3] = amd_bytealign_S (in0[0], 0, 3);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[2] = 0;
out1[1] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[0] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 30: out1[3] = amd_bytealign_S (in0[0], 0, 2);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[2] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
out1[1] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 31: out1[3] = amd_bytealign_S (in0[0], 0, 1);
Initial commit
2015-12-04 14:47:52 +00:00
out1[2] = 0;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
out1[1] = 0;
out1[0] = 0;
out0[3] = 0;
out0[2] = 0;
out0[1] = 0;
out0[0] = 0;
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
#endif
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline void append_block1 (const u32 offset, u32 dst0[4], u32 dst1[4], const u32 src_r0)
Initial commit
2015-12-04 14:47:52 +00:00
{
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tmp[2];
Initial commit
2015-12-04 14:47:52 +00:00
switch (offset & 3)
{
case 0: tmp[0] = src_r0;
tmp[1] = 0;
break;
case 1: tmp[0] = src_r0 << 8;
tmp[1] = src_r0 >> 24;
break;
case 2: tmp[0] = src_r0 << 16;
tmp[1] = src_r0 >> 16;
break;
case 3: tmp[0] = src_r0 << 24;
tmp[1] = src_r0 >> 8;
break;
}
switch (offset / 4)
{
case 0: dst0[0] |= tmp[0];
dst0[1] = tmp[1];
break;
case 1: dst0[1] |= tmp[0];
dst0[2] = tmp[1];
break;
case 2: dst0[2] |= tmp[0];
dst0[3] = tmp[1];
break;
case 3: dst0[3] |= tmp[0];
dst1[0] = tmp[1];
break;
case 4: dst1[0] |= tmp[0];
dst1[1] = tmp[1];
break;
case 5: dst1[1] |= tmp[0];
dst1[2] = tmp[1];
break;
case 6: dst1[2] |= tmp[0];
dst1[3] = tmp[1];
break;
case 7: dst1[3] |= tmp[0];
break;
}
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline void append_block8 (const u32 offset, u32 dst0[4], u32 dst1[4], const u32 src_l0[4], const u32 src_l1[4], const u32 src_r0[4], const u32 src_r1[4])
Initial commit
2015-12-04 14:47:52 +00:00
{
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Initial commit
2015-12-04 14:47:52 +00:00
switch (offset)
{
case 0:
dst0[0] = src_r0[0];
dst0[1] = src_r0[1];
dst0[2] = src_r0[2];
dst0[3] = src_r0[3];
dst1[0] = src_r1[0];
dst1[1] = src_r1[1];
dst1[2] = src_r1[2];
dst1[3] = src_r1[3];
break;
case 1:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[0] = __byte_perm_S (src_l0[0], src_r0[0], 0x6540);
dst0[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
dst0[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
dst0[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x6543);
dst1[0] = __byte_perm_S (src_r0[3], src_r1[0], 0x6543);
dst1[1] = __byte_perm_S (src_r1[0], src_r1[1], 0x6543);
dst1[2] = __byte_perm_S (src_r1[1], src_r1[2], 0x6543);
dst1[3] = __byte_perm_S (src_r1[2], src_r1[3], 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 2:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[0] = __byte_perm_S (src_l0[0], src_r0[0], 0x5410);
dst0[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
dst0[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
dst0[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x5432);
dst1[0] = __byte_perm_S (src_r0[3], src_r1[0], 0x5432);
dst1[1] = __byte_perm_S (src_r1[0], src_r1[1], 0x5432);
dst1[2] = __byte_perm_S (src_r1[1], src_r1[2], 0x5432);
dst1[3] = __byte_perm_S (src_r1[2], src_r1[3], 0x5432);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 3:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[0] = __byte_perm_S (src_l0[0], src_r0[0], 0x4210);
dst0[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
dst0[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
dst0[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x4321);
dst1[0] = __byte_perm_S (src_r0[3], src_r1[0], 0x4321);
dst1[1] = __byte_perm_S (src_r1[0], src_r1[1], 0x4321);
dst1[2] = __byte_perm_S (src_r1[1], src_r1[2], 0x4321);
dst1[3] = __byte_perm_S (src_r1[2], src_r1[3], 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 4:
dst0[1] = src_r0[0];
dst0[2] = src_r0[1];
dst0[3] = src_r0[2];
dst1[0] = src_r0[3];
dst1[1] = src_r1[0];
dst1[2] = src_r1[1];
dst1[3] = src_r1[2];
break;
case 5:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[1] = __byte_perm_S (src_l0[1], src_r0[0], 0x6540);
dst0[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
dst0[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
dst1[0] = __byte_perm_S (src_r0[2], src_r0[3], 0x6543);
dst1[1] = __byte_perm_S (src_r0[3], src_r1[0], 0x6543);
dst1[2] = __byte_perm_S (src_r1[0], src_r1[1], 0x6543);
dst1[3] = __byte_perm_S (src_r1[1], src_r1[2], 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 6:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[1] = __byte_perm_S (src_l0[1], src_r0[0], 0x5410);
dst0[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
dst0[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
dst1[0] = __byte_perm_S (src_r0[2], src_r0[3], 0x5432);
dst1[1] = __byte_perm_S (src_r0[3], src_r1[0], 0x5432);
dst1[2] = __byte_perm_S (src_r1[0], src_r1[1], 0x5432);
dst1[3] = __byte_perm_S (src_r1[1], src_r1[2], 0x5432);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 7:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[1] = __byte_perm_S (src_l0[1], src_r0[0], 0x4210);
dst0[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
dst0[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
dst1[0] = __byte_perm_S (src_r0[2], src_r0[3], 0x4321);
dst1[1] = __byte_perm_S (src_r0[3], src_r1[0], 0x4321);
dst1[2] = __byte_perm_S (src_r1[0], src_r1[1], 0x4321);
dst1[3] = __byte_perm_S (src_r1[1], src_r1[2], 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 8:
dst0[2] = src_r0[0];
dst0[3] = src_r0[1];
dst1[0] = src_r0[2];
dst1[1] = src_r0[3];
dst1[2] = src_r1[0];
dst1[3] = src_r1[1];
break;
case 9:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[2] = __byte_perm_S (src_l0[2], src_r0[0], 0x6540);
dst0[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
dst1[0] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
dst1[1] = __byte_perm_S (src_r0[2], src_r0[3], 0x6543);
dst1[2] = __byte_perm_S (src_r0[3], src_r1[0], 0x6543);
dst1[3] = __byte_perm_S (src_r1[0], src_r1[1], 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 10:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[2] = __byte_perm_S (src_l0[2], src_r0[0], 0x5410);
dst0[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
dst1[0] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
dst1[1] = __byte_perm_S (src_r0[2], src_r0[3], 0x5432);
dst1[2] = __byte_perm_S (src_r0[3], src_r1[0], 0x5432);
dst1[3] = __byte_perm_S (src_r1[0], src_r1[1], 0x5432);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 11:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[2] = __byte_perm_S (src_l0[2], src_r0[0], 0x4210);
dst0[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
dst1[0] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
dst1[1] = __byte_perm_S (src_r0[2], src_r0[3], 0x4321);
dst1[2] = __byte_perm_S (src_r0[3], src_r1[0], 0x4321);
dst1[3] = __byte_perm_S (src_r1[0], src_r1[1], 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 12:
dst0[3] = src_r0[0];
dst1[0] = src_r0[1];
dst1[1] = src_r0[2];
dst1[2] = src_r0[3];
dst1[3] = src_r1[0];
break;
case 13:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[3] = __byte_perm_S (src_l0[3], src_r0[0], 0x6540);
dst1[0] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
dst1[1] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
dst1[2] = __byte_perm_S (src_r0[2], src_r0[3], 0x6543);
dst1[3] = __byte_perm_S (src_r0[3], src_r1[0], 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 14:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[3] = __byte_perm_S (src_l0[3], src_r0[0], 0x5410);
dst1[0] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
dst1[1] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
dst1[2] = __byte_perm_S (src_r0[2], src_r0[3], 0x5432);
dst1[3] = __byte_perm_S (src_r0[3], src_r1[0], 0x5432);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 15:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst0[3] = __byte_perm_S (src_l0[3], src_r0[0], 0x4210);
dst1[0] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
dst1[1] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
dst1[2] = __byte_perm_S (src_r0[2], src_r0[3], 0x4321);
dst1[3] = __byte_perm_S (src_r0[3], src_r1[0], 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 16:
dst1[0] = src_r0[0];
dst1[1] = src_r0[1];
dst1[2] = src_r0[2];
dst1[3] = src_r0[3];
break;
case 17:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[0] = __byte_perm_S (src_l1[0], src_r0[0], 0x6540);
dst1[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
dst1[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
dst1[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 18:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[0] = __byte_perm_S (src_l1[0], src_r0[0], 0x5410);
dst1[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
dst1[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
dst1[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x5432);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 19:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[0] = __byte_perm_S (src_l1[0], src_r0[0], 0x4210);
dst1[1] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
dst1[2] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
dst1[3] = __byte_perm_S (src_r0[2], src_r0[3], 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 20:
dst1[1] = src_r0[0];
dst1[2] = src_r0[1];
dst1[3] = src_r0[2];
break;
case 21:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[1] = __byte_perm_S (src_l1[1], src_r0[0], 0x6540);
dst1[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
dst1[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 22:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[1] = __byte_perm_S (src_l1[1], src_r0[0], 0x5410);
dst1[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
dst1[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x5432);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 23:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[1] = __byte_perm_S (src_l1[1], src_r0[0], 0x4210);
dst1[2] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
dst1[3] = __byte_perm_S (src_r0[1], src_r0[2], 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 24:
dst1[2] = src_r0[0];
dst1[3] = src_r0[1];
break;
case 25:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[2] = __byte_perm_S (src_l1[2], src_r0[0], 0x6540);
dst1[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 26:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[2] = __byte_perm_S (src_l1[2], src_r0[0], 0x5410);
dst1[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x5432);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 27:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[2] = __byte_perm_S (src_l1[2], src_r0[0], 0x4210);
dst1[3] = __byte_perm_S (src_r0[0], src_r0[1], 0x4321);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 28:
dst1[3] = src_r0[0];
break;
case 29:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = __byte_perm_S (src_l1[3], src_r0[0], 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 30:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = __byte_perm_S (src_l1[3], src_r0[0], 0x5410);
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 31:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = __byte_perm_S (src_l1[3], src_r0[0], 0x4210);
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
Initial commit
2015-12-04 14:47:52 +00:00
switch (offset)
{
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 31:
dst1[3] = src_l1[3] | src_r0[0] << 24;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 30:
dst1[3] = src_l1[3] | src_r0[0] << 16;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 29:
dst1[3] = src_l1[3] | src_r0[0] << 8;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 28:
dst1[3] = src_r0[0];
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 27:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst1[2] = src_l1[2] | src_r0[0] << 24;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 26:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst1[2] = src_l1[2] | src_r0[0] << 16;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 25:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst1[2] = src_l1[2] | src_r0[0] << 8;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 24:
dst1[3] = src_r0[1];
dst1[2] = src_r0[0];
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 23:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
dst1[2] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst1[1] = src_l1[1] | src_r0[0] << 24;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 22:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
dst1[2] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst1[1] = src_l1[1] | src_r0[0] << 16;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 21:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
dst1[2] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst1[1] = src_l1[1] | src_r0[0] << 8;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 20:
dst1[3] = src_r0[2];
dst1[2] = src_r0[1];
dst1[1] = src_r0[0];
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 19:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r0[3], src_r0[2], 1);
dst1[2] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
dst1[1] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst1[0] = src_l1[0] | src_r0[0] << 24;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 18:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r0[3], src_r0[2], 2);
dst1[2] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
dst1[1] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst1[0] = src_l1[0] | src_r0[0] << 16;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 17:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r0[3], src_r0[2], 3);
dst1[2] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
dst1[1] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst1[0] = src_l1[0] | src_r0[0] << 8;
Initial commit
2015-12-04 14:47:52 +00:00
break;
case 16:
dst1[3] = src_r0[3];
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst1[2] = src_r0[2];
dst1[1] = src_r0[1];
dst1[0] = src_r0[0];
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 15:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[0], src_r0[3], 1);
dst1[2] = amd_bytealign_S (src_r0[3], src_r0[2], 1);
dst1[1] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
dst1[0] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[3] = src_l0[3] | src_r0[0] << 24;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 14:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[0], src_r0[3], 2);
dst1[2] = amd_bytealign_S (src_r0[3], src_r0[2], 2);
dst1[1] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
dst1[0] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[3] = src_l0[3] | src_r0[0] << 16;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 13:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[0], src_r0[3], 3);
dst1[2] = amd_bytealign_S (src_r0[3], src_r0[2], 3);
dst1[1] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
dst1[0] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[3] = src_l0[3] | src_r0[0] << 8;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 12:
dst1[3] = src_r1[0];
dst1[2] = src_r0[3];
dst1[1] = src_r0[2];
dst1[0] = src_r0[1];
dst0[3] = src_r0[0];
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 11:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[1], src_r1[0], 1);
dst1[2] = amd_bytealign_S (src_r1[0], src_r0[3], 1);
dst1[1] = amd_bytealign_S (src_r0[3], src_r0[2], 1);
dst1[0] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
dst0[3] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[2] = src_l0[2] | src_r0[0] << 24;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 10:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[1], src_r1[0], 2);
dst1[2] = amd_bytealign_S (src_r1[0], src_r0[3], 2);
dst1[1] = amd_bytealign_S (src_r0[3], src_r0[2], 2);
dst1[0] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
dst0[3] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[2] = src_l0[2] | src_r0[0] << 16;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 9:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[1], src_r1[0], 3);
dst1[2] = amd_bytealign_S (src_r1[0], src_r0[3], 3);
dst1[1] = amd_bytealign_S (src_r0[3], src_r0[2], 3);
dst1[0] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
dst0[3] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[2] = src_l0[2] | src_r0[0] << 8;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 8:
dst1[3] = src_r1[1];
dst1[2] = src_r1[0];
dst1[1] = src_r0[3];
dst1[0] = src_r0[2];
dst0[3] = src_r0[1];
dst0[2] = src_r0[0];
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 7:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[2], src_r1[1], 1);
dst1[2] = amd_bytealign_S (src_r1[1], src_r1[0], 1);
dst1[1] = amd_bytealign_S (src_r1[0], src_r0[3], 1);
dst1[0] = amd_bytealign_S (src_r0[3], src_r0[2], 1);
dst0[3] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
dst0[2] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[1] = src_l0[1] | src_r0[0] << 24;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 6:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[2], src_r1[1], 2);
dst1[2] = amd_bytealign_S (src_r1[1], src_r1[0], 2);
dst1[1] = amd_bytealign_S (src_r1[0], src_r0[3], 2);
dst1[0] = amd_bytealign_S (src_r0[3], src_r0[2], 2);
dst0[3] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
dst0[2] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[1] = src_l0[1] | src_r0[0] << 16;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 5:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[2], src_r1[1], 3);
dst1[2] = amd_bytealign_S (src_r1[1], src_r1[0], 3);
dst1[1] = amd_bytealign_S (src_r1[0], src_r0[3], 3);
dst1[0] = amd_bytealign_S (src_r0[3], src_r0[2], 3);
dst0[3] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
dst0[2] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[1] = src_l0[1] | src_r0[0] << 8;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 4:
dst1[3] = src_r1[2];
dst1[2] = src_r1[1];
dst1[1] = src_r1[0];
dst1[0] = src_r0[3];
dst0[3] = src_r0[2];
dst0[2] = src_r0[1];
dst0[1] = src_r0[0];
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 3:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[3], src_r1[2], 1);
dst1[2] = amd_bytealign_S (src_r1[2], src_r1[1], 1);
dst1[1] = amd_bytealign_S (src_r1[1], src_r1[0], 1);
dst1[0] = amd_bytealign_S (src_r1[0], src_r0[3], 1);
dst0[3] = amd_bytealign_S (src_r0[3], src_r0[2], 1);
dst0[2] = amd_bytealign_S (src_r0[2], src_r0[1], 1);
dst0[1] = amd_bytealign_S (src_r0[1], src_r0[0], 1);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[0] = src_l0[0] | src_r0[0] << 24;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 2:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[3], src_r1[2], 2);
dst1[2] = amd_bytealign_S (src_r1[2], src_r1[1], 2);
dst1[1] = amd_bytealign_S (src_r1[1], src_r1[0], 2);
dst1[0] = amd_bytealign_S (src_r1[0], src_r0[3], 2);
dst0[3] = amd_bytealign_S (src_r0[3], src_r0[2], 2);
dst0[2] = amd_bytealign_S (src_r0[2], src_r0[1], 2);
dst0[1] = amd_bytealign_S (src_r0[1], src_r0[0], 2);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[0] = src_l0[0] | src_r0[0] << 16;
Initial commit
2015-12-04 14:47:52 +00:00
break;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
case 1:
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
dst1[3] = amd_bytealign_S (src_r1[3], src_r1[2], 3);
dst1[2] = amd_bytealign_S (src_r1[2], src_r1[1], 3);
dst1[1] = amd_bytealign_S (src_r1[1], src_r1[0], 3);
dst1[0] = amd_bytealign_S (src_r1[0], src_r0[3], 3);
dst0[3] = amd_bytealign_S (src_r0[3], src_r0[2], 3);
dst0[2] = amd_bytealign_S (src_r0[2], src_r0[1], 3);
dst0[1] = amd_bytealign_S (src_r0[1], src_r0[0], 3);
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
dst0[0] = src_l0[0] | src_r0[0] << 8;
break;
case 0:
dst1[3] = src_r1[3];
dst1[2] = src_r1[2];
dst1[1] = src_r1[1];
dst1[0] = src_r1[0];
dst0[3] = src_r0[3];
dst0[2] = src_r0[2];
dst0[1] = src_r0[1];
dst0[0] = src_r0[0];
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
#endif
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline void reverse_block (u32 in0[4], u32 in1[4], u32 out0[4], u32 out1[4], const u32 len)
Initial commit
2015-12-04 14:47:52 +00:00
{
rshift_block_N (in0, in1, out0, out1, 32 - len);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
Initial commit
2015-12-04 14:47:52 +00:00
tib40[0] = out1[3];
tib40[1] = out1[2];
tib40[2] = out1[1];
tib40[3] = out1[0];
tib41[0] = out0[3];
tib41[1] = out0[2];
tib41[2] = out0[1];
tib41[3] = out0[0];
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
out0[0] = swap32_S (tib40[0]);
out0[1] = swap32_S (tib40[1]);
out0[2] = swap32_S (tib40[2]);
out0[3] = swap32_S (tib40[3]);
out1[0] = swap32_S (tib41[0]);
out1[1] = swap32_S (tib41[1]);
out1[2] = swap32_S (tib41[2]);
out1[3] = swap32_S (tib41[3]);
Initial commit
2015-12-04 14:47:52 +00:00
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_lrest (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
buf0[0] |= (generate_cmask (buf0[0]));
buf0[1] |= (generate_cmask (buf0[1]));
buf0[2] |= (generate_cmask (buf0[2]));
buf0[3] |= (generate_cmask (buf0[3]));
buf1[0] |= (generate_cmask (buf1[0]));
buf1[1] |= (generate_cmask (buf1[1]));
buf1[2] |= (generate_cmask (buf1[2]));
buf1[3] |= (generate_cmask (buf1[3]));
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_urest (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
buf0[0] &= ~(generate_cmask (buf0[0]));
buf0[1] &= ~(generate_cmask (buf0[1]));
buf0[2] &= ~(generate_cmask (buf0[2]));
buf0[3] &= ~(generate_cmask (buf0[3]));
buf1[0] &= ~(generate_cmask (buf1[0]));
buf1[1] &= ~(generate_cmask (buf1[1]));
buf1[2] &= ~(generate_cmask (buf1[2]));
buf1[3] &= ~(generate_cmask (buf1[3]));
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_lrest_ufirst (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
rule_op_mangle_lrest (p0, p1, buf0, buf1, in_len);
buf0[0] &= ~(0x00000020 & generate_cmask (buf0[0]));
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_urest_lfirst (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
rule_op_mangle_urest (p0, p1, buf0, buf1, in_len);
buf0[0] |= (0x00000020 & generate_cmask (buf0[0]));
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_trest (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
buf0[0] ^= (generate_cmask (buf0[0]));
buf0[1] ^= (generate_cmask (buf0[1]));
buf0[2] ^= (generate_cmask (buf0[2]));
buf0[3] ^= (generate_cmask (buf0[3]));
buf1[0] ^= (generate_cmask (buf1[0]));
buf1[1] ^= (generate_cmask (buf1[1]));
buf1[2] ^= (generate_cmask (buf1[2]));
buf1[3] ^= (generate_cmask (buf1[3]));
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_toggle_at (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 >= in_len) return (in_len);
Fix more missing casts to uint in rule-engine
2016-01-14 19:54:41 +00:00
const u32 tmp = 0x20u << ((p0 & 3) * 8);
Initial commit
2015-12-04 14:47:52 +00:00
switch (p0 / 4)
{
case 0: buf0[0] ^= (tmp & generate_cmask (buf0[0])); break;
case 1: buf0[1] ^= (tmp & generate_cmask (buf0[1])); break;
case 2: buf0[2] ^= (tmp & generate_cmask (buf0[2])); break;
case 3: buf0[3] ^= (tmp & generate_cmask (buf0[3])); break;
case 4: buf1[0] ^= (tmp & generate_cmask (buf1[0])); break;
case 5: buf1[1] ^= (tmp & generate_cmask (buf1[1])); break;
case 6: buf1[2] ^= (tmp & generate_cmask (buf1[2])); break;
case 7: buf1[3] ^= (tmp & generate_cmask (buf1[3])); break;
}
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_reverse (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
reverse_block (buf0, buf1, buf0, buf1, in_len);
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_dupeword (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if ((in_len + in_len) >= 32) return (in_len);
u32 out_len = in_len;
Workaround AMD catalyst optimizer bug in rule-engine
2016-01-19 19:54:25 +00:00
append_block8 (out_len, buf0, buf1, buf0, buf1, buf0, buf1);
Initial commit
2015-12-04 14:47:52 +00:00
out_len += in_len;
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_dupeword_times (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (((in_len * p0) + in_len) >= 32) return (in_len);
u32 out_len = in_len;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
Initial commit
2015-12-04 14:47:52 +00:00
tib40[0] = buf0[0];
tib40[1] = buf0[1];
tib40[2] = buf0[2];
tib40[3] = buf0[3];
tib41[0] = buf1[0];
tib41[1] = buf1[1];
tib41[2] = buf1[2];
tib41[3] = buf1[3];
for (u32 i = 0; i < p0; i++)
{
append_block8 (out_len, buf0, buf1, buf0, buf1, tib40, tib41);
out_len += in_len;
}
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_reflect (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if ((in_len + in_len) >= 32) return (in_len);
u32 out_len = in_len;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
Initial commit
2015-12-04 14:47:52 +00:00
reverse_block (buf0, buf1, tib40, tib41, out_len);
append_block8 (out_len, buf0, buf1, buf0, buf1, tib40, tib41);
out_len += in_len;
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_append (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if ((in_len + 1) >= 32) return (in_len);
u32 out_len = in_len;
append_block1 (out_len, buf0, buf1, p0);
out_len++;
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_prepend (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if ((in_len + 1) >= 32) return (in_len);
u32 out_len = in_len;
rshift_block (buf0, buf1, buf0, buf1);
buf0[0] = buf0[0] | p0;
out_len++;
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_rotate_left (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (in_len == 0) return (in_len);
const u32 in_len1 = in_len - 1;
const u32 sh = (in_len1 & 3) * 8;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
const u32 tmp = (buf0[0] & 0xff) << sh;
Initial commit
2015-12-04 14:47:52 +00:00
lshift_block (buf0, buf1, buf0, buf1);
switch (in_len1 / 4)
{
case 0: buf0[0] |= tmp; break;
case 1: buf0[1] |= tmp; break;
case 2: buf0[2] |= tmp; break;
case 3: buf0[3] |= tmp; break;
case 4: buf1[0] |= tmp; break;
case 5: buf1[1] |= tmp; break;
case 6: buf1[2] |= tmp; break;
case 7: buf1[3] |= tmp; break;
}
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_rotate_right (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (in_len == 0) return (in_len);
const u32 in_len1 = in_len - 1;
const u32 sh = (in_len1 & 3) * 8;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tmp = 0;
Initial commit
2015-12-04 14:47:52 +00:00
switch (in_len1 / 4)
{
case 0: tmp = (buf0[0] >> sh) & 0xff; break;
case 1: tmp = (buf0[1] >> sh) & 0xff; break;
case 2: tmp = (buf0[2] >> sh) & 0xff; break;
case 3: tmp = (buf0[3] >> sh) & 0xff; break;
case 4: tmp = (buf1[0] >> sh) & 0xff; break;
case 5: tmp = (buf1[1] >> sh) & 0xff; break;
case 6: tmp = (buf1[2] >> sh) & 0xff; break;
case 7: tmp = (buf1[3] >> sh) & 0xff; break;
}
rshift_block (buf0, buf1, buf0, buf1);
buf0[0] |= tmp;
truncate_right (buf0, buf1, in_len);
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_delete_first (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (in_len == 0) return (in_len);
const u32 in_len1 = in_len - 1;
lshift_block (buf0, buf1, buf0, buf1);
return in_len1;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_delete_last (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (in_len == 0) return (in_len);
const u32 in_len1 = in_len - 1;
const u32 tmp = (1 << ((in_len1 & 3) * 8)) - 1;
switch (in_len1 / 4)
{
case 0: buf0[0] &= tmp; break;
case 1: buf0[1] &= tmp; break;
case 2: buf0[2] &= tmp; break;
case 3: buf0[3] &= tmp; break;
case 4: buf1[0] &= tmp; break;
case 5: buf1[1] &= tmp; break;
case 6: buf1[2] &= tmp; break;
case 7: buf1[3] &= tmp; break;
}
return in_len1;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_delete_at (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 >= in_len) return (in_len);
u32 out_len = in_len;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
Initial commit
2015-12-04 14:47:52 +00:00
lshift_block (buf0, buf1, tib40, tib41);
const u32 ml = (1 << ((p0 & 3) * 8)) - 1;
const u32 mr = ~ml;
switch (p0 / 4)
{
case 0: buf0[0] = (buf0[0] & ml)
| (tib40[0] & mr);
buf0[1] = tib40[1];
buf0[2] = tib40[2];
buf0[3] = tib40[3];
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 1: buf0[1] = (buf0[1] & ml)
| (tib40[1] & mr);
buf0[2] = tib40[2];
buf0[3] = tib40[3];
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 2: buf0[2] = (buf0[2] & ml)
| (tib40[2] & mr);
buf0[3] = tib40[3];
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 3: buf0[3] = (buf0[3] & ml)
| (tib40[3] & mr);
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 4: buf1[0] = (buf1[0] & ml)
| (tib41[0] & mr);
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 5: buf1[1] = (buf1[1] & ml)
| (tib41[1] & mr);
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 6: buf1[2] = (buf1[2] & ml)
| (tib41[2] & mr);
buf1[3] = tib41[3];
break;
case 7: buf1[3] = (buf1[3] & ml)
| (tib41[3] & mr);
break;
}
out_len--;
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_extract (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 >= in_len) return (in_len);
if ((p0 + p1) > in_len) return (in_len);
u32 out_len = p1;
lshift_block_N (buf0, buf1, buf0, buf1, p0);
truncate_right (buf0, buf1, out_len);
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_omit (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 >= in_len) return (in_len);
if ((p0 + p1) > in_len) return (in_len);
u32 out_len = in_len;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
Initial commit
2015-12-04 14:47:52 +00:00
tib40[0] = 0;
tib40[1] = 0;
tib40[2] = 0;
tib40[3] = 0;
tib41[0] = 0;
tib41[1] = 0;
tib41[2] = 0;
tib41[3] = 0;
lshift_block_N (buf0, buf1, tib40, tib41, p1);
const u32 ml = (1 << ((p0 & 3) * 8)) - 1;
const u32 mr = ~ml;
switch (p0 / 4)
{
case 0: buf0[0] = (buf0[0] & ml)
| (tib40[0] & mr);
buf0[1] = tib40[1];
buf0[2] = tib40[2];
buf0[3] = tib40[3];
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 1: buf0[1] = (buf0[1] & ml)
| (tib40[1] & mr);
buf0[2] = tib40[2];
buf0[3] = tib40[3];
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 2: buf0[2] = (buf0[2] & ml)
| (tib40[2] & mr);
buf0[3] = tib40[3];
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 3: buf0[3] = (buf0[3] & ml)
| (tib40[3] & mr);
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 4: buf1[0] = (buf1[0] & ml)
| (tib41[0] & mr);
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 5: buf1[1] = (buf1[1] & ml)
| (tib41[1] & mr);
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 6: buf1[2] = (buf1[2] & ml)
| (tib41[2] & mr);
buf1[3] = tib41[3];
break;
case 7: buf1[3] = (buf1[3] & ml)
| (tib41[3] & mr);
break;
}
out_len -= p1;
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_insert (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 > in_len) return (in_len);
if ((in_len + 1) >= 32) return (in_len);
u32 out_len = in_len;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
Initial commit
2015-12-04 14:47:52 +00:00
rshift_block (buf0, buf1, tib40, tib41);
const u32 p1n = p1 << ((p0 & 3) * 8);
const u32 ml = (1 << ((p0 & 3) * 8)) - 1;
const u32 mr = 0xffffff00 << ((p0 & 3) * 8);
switch (p0 / 4)
{
case 0: buf0[0] = (buf0[0] & ml) | p1n | (tib40[0] & mr);
buf0[1] = tib40[1];
buf0[2] = tib40[2];
buf0[3] = tib40[3];
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 1: buf0[1] = (buf0[1] & ml) | p1n | (tib40[1] & mr);
buf0[2] = tib40[2];
buf0[3] = tib40[3];
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 2: buf0[2] = (buf0[2] & ml) | p1n | (tib40[2] & mr);
buf0[3] = tib40[3];
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 3: buf0[3] = (buf0[3] & ml) | p1n | (tib40[3] & mr);
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 4: buf1[0] = (buf1[0] & ml) | p1n | (tib41[0] & mr);
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 5: buf1[1] = (buf1[1] & ml) | p1n | (tib41[1] & mr);
buf1[2] = tib41[2];
buf1[3] = tib41[3];
break;
case 6: buf1[2] = (buf1[2] & ml) | p1n | (tib41[2] & mr);
buf1[3] = tib41[3];
break;
case 7: buf1[3] = (buf1[3] & ml) | p1n | (tib41[3] & mr);
break;
}
out_len++;
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_overstrike (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 >= in_len) return (in_len);
const u32 p1n = p1 << ((p0 & 3) * 8);
Fix more missing casts to uint in rule-engine
2016-01-14 19:54:41 +00:00
const u32 m = ~(0xffu << ((p0 & 3) * 8));
Initial commit
2015-12-04 14:47:52 +00:00
switch (p0 / 4)
{
case 0: buf0[0] = (buf0[0] & m) | p1n; break;
case 1: buf0[1] = (buf0[1] & m) | p1n; break;
case 2: buf0[2] = (buf0[2] & m) | p1n; break;
case 3: buf0[3] = (buf0[3] & m) | p1n; break;
case 4: buf1[0] = (buf1[0] & m) | p1n; break;
case 5: buf1[1] = (buf1[1] & m) | p1n; break;
case 6: buf1[2] = (buf1[2] & m) | p1n; break;
case 7: buf1[3] = (buf1[3] & m) | p1n; break;
}
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_truncate_at (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 >= in_len) return (in_len);
truncate_right (buf0, buf1, p0);
return p0;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_replace (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Initial commit
2015-12-04 14:47:52 +00:00
for (u32 i = 0; i < in_len; i++)
{
switch (i)
{
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 0: if ((__byte_perm_S (buf0[0], 0, 0x6540)) == p0) buf0[0] = __byte_perm_S (p1, buf0[0], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 1: if ((__byte_perm_S (buf0[0], 0, 0x6541)) == p0) buf0[0] = __byte_perm_S (p1, buf0[0], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 2: if ((__byte_perm_S (buf0[0], 0, 0x6542)) == p0) buf0[0] = __byte_perm_S (p1, buf0[0], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 3: if ((__byte_perm_S (buf0[0], 0, 0x6543)) == p0) buf0[0] = __byte_perm_S (p1, buf0[0], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 4: if ((__byte_perm_S (buf0[1], 0, 0x6540)) == p0) buf0[1] = __byte_perm_S (p1, buf0[1], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 5: if ((__byte_perm_S (buf0[1], 0, 0x6541)) == p0) buf0[1] = __byte_perm_S (p1, buf0[1], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 6: if ((__byte_perm_S (buf0[1], 0, 0x6542)) == p0) buf0[1] = __byte_perm_S (p1, buf0[1], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 7: if ((__byte_perm_S (buf0[1], 0, 0x6543)) == p0) buf0[1] = __byte_perm_S (p1, buf0[1], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 8: if ((__byte_perm_S (buf0[2], 0, 0x6540)) == p0) buf0[2] = __byte_perm_S (p1, buf0[2], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 9: if ((__byte_perm_S (buf0[2], 0, 0x6541)) == p0) buf0[2] = __byte_perm_S (p1, buf0[2], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 10: if ((__byte_perm_S (buf0[2], 0, 0x6542)) == p0) buf0[2] = __byte_perm_S (p1, buf0[2], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 11: if ((__byte_perm_S (buf0[2], 0, 0x6543)) == p0) buf0[2] = __byte_perm_S (p1, buf0[2], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 12: if ((__byte_perm_S (buf0[3], 0, 0x6540)) == p0) buf0[3] = __byte_perm_S (p1, buf0[3], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 13: if ((__byte_perm_S (buf0[3], 0, 0x6541)) == p0) buf0[3] = __byte_perm_S (p1, buf0[3], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 14: if ((__byte_perm_S (buf0[3], 0, 0x6542)) == p0) buf0[3] = __byte_perm_S (p1, buf0[3], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 15: if ((__byte_perm_S (buf0[3], 0, 0x6543)) == p0) buf0[3] = __byte_perm_S (p1, buf0[3], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 16: if ((__byte_perm_S (buf1[0], 0, 0x6540)) == p0) buf1[0] = __byte_perm_S (p1, buf1[0], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 17: if ((__byte_perm_S (buf1[0], 0, 0x6541)) == p0) buf1[0] = __byte_perm_S (p1, buf1[0], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 18: if ((__byte_perm_S (buf1[0], 0, 0x6542)) == p0) buf1[0] = __byte_perm_S (p1, buf1[0], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 19: if ((__byte_perm_S (buf1[0], 0, 0x6543)) == p0) buf1[0] = __byte_perm_S (p1, buf1[0], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 20: if ((__byte_perm_S (buf1[1], 0, 0x6540)) == p0) buf1[1] = __byte_perm_S (p1, buf1[1], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 21: if ((__byte_perm_S (buf1[1], 0, 0x6541)) == p0) buf1[1] = __byte_perm_S (p1, buf1[1], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 22: if ((__byte_perm_S (buf1[1], 0, 0x6542)) == p0) buf1[1] = __byte_perm_S (p1, buf1[1], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 23: if ((__byte_perm_S (buf1[1], 0, 0x6543)) == p0) buf1[1] = __byte_perm_S (p1, buf1[1], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 24: if ((__byte_perm_S (buf1[2], 0, 0x6540)) == p0) buf1[2] = __byte_perm_S (p1, buf1[2], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 25: if ((__byte_perm_S (buf1[2], 0, 0x6541)) == p0) buf1[2] = __byte_perm_S (p1, buf1[2], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 26: if ((__byte_perm_S (buf1[2], 0, 0x6542)) == p0) buf1[2] = __byte_perm_S (p1, buf1[2], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 27: if ((__byte_perm_S (buf1[2], 0, 0x6543)) == p0) buf1[2] = __byte_perm_S (p1, buf1[2], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 28: if ((__byte_perm_S (buf1[3], 0, 0x6540)) == p0) buf1[3] = __byte_perm_S (p1, buf1[3], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 29: if ((__byte_perm_S (buf1[3], 0, 0x6541)) == p0) buf1[3] = __byte_perm_S (p1, buf1[3], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 30: if ((__byte_perm_S (buf1[3], 0, 0x6542)) == p0) buf1[3] = __byte_perm_S (p1, buf1[3], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 31: if ((__byte_perm_S (buf1[3], 0, 0x6543)) == p0) buf1[3] = __byte_perm_S (p1, buf1[3], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
}
#endif
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
const uchar4 tmp0 = (uchar4) (p0);
const uchar4 tmp1 = (uchar4) (p1);
Initial commit
2015-12-04 14:47:52 +00:00
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
uchar4 tmp;
Initial commit
2015-12-04 14:47:52 +00:00
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
tmp = as_uchar4 (buf0[0]); tmp = select (tmp, tmp1, tmp == tmp0); buf0[0] = as_uint (tmp);
tmp = as_uchar4 (buf0[1]); tmp = select (tmp, tmp1, tmp == tmp0); buf0[1] = as_uint (tmp);
tmp = as_uchar4 (buf0[2]); tmp = select (tmp, tmp1, tmp == tmp0); buf0[2] = as_uint (tmp);
tmp = as_uchar4 (buf0[3]); tmp = select (tmp, tmp1, tmp == tmp0); buf0[3] = as_uint (tmp);
tmp = as_uchar4 (buf1[0]); tmp = select (tmp, tmp1, tmp == tmp0); buf1[0] = as_uint (tmp);
tmp = as_uchar4 (buf1[1]); tmp = select (tmp, tmp1, tmp == tmp0); buf1[1] = as_uint (tmp);
tmp = as_uchar4 (buf1[2]); tmp = select (tmp, tmp1, tmp == tmp0); buf1[2] = as_uint (tmp);
tmp = as_uchar4 (buf1[3]); tmp = select (tmp, tmp1, tmp == tmp0); buf1[3] = as_uint (tmp);
Initial commit
2015-12-04 14:47:52 +00:00
#endif
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_purgechar (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
// TODO
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_togglecase_rec (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
// TODO
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_dupechar_first (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if ( in_len == 0) return (in_len);
if ((in_len + p0) >= 32) return (in_len);
u32 out_len = in_len;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
const u32 tmp = buf0[0] & 0xFF;
Initial commit
2015-12-04 14:47:52 +00:00
rshift_block_N (buf0, buf1, buf0, buf1, p0);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Initial commit
2015-12-04 14:47:52 +00:00
switch (p0)
{
case 1: buf0[0] |= tmp;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 2: buf0[0] |= __byte_perm_S (tmp, 0, 0x5400);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 3: buf0[0] |= __byte_perm_S (tmp, 0, 0x4000);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 4: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 5: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
Initial commit
2015-12-04 14:47:52 +00:00
buf0[1] |= tmp;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 6: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x5400);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 7: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x4000);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 8: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 9: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
Initial commit
2015-12-04 14:47:52 +00:00
buf0[2] |= tmp;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 10: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x5400);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 11: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x4000);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 12: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 13: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
Initial commit
2015-12-04 14:47:52 +00:00
buf0[3] |= tmp;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 14: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x5400);
break;
case 15: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x4000);
break;
case 16: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
break;
case 17: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
Initial commit
2015-12-04 14:47:52 +00:00
buf1[0] |= tmp;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 18: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x5400);
break;
case 19: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x4000);
break;
case 20: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
break;
case 21: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
Initial commit
2015-12-04 14:47:52 +00:00
buf1[1] |= tmp;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 22: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[1] |= __byte_perm_S (tmp, 0, 0x5400);
break;
case 23: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[1] |= __byte_perm_S (tmp, 0, 0x4000);
break;
case 24: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
break;
case 25: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
Initial commit
2015-12-04 14:47:52 +00:00
buf1[2] |= tmp;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 26: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[2] |= __byte_perm_S (tmp, 0, 0x5400);
break;
case 27: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[2] |= __byte_perm_S (tmp, 0, 0x4000);
break;
case 28: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[2] |= __byte_perm_S (tmp, 0, 0x0000);
break;
case 29: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[2] |= __byte_perm_S (tmp, 0, 0x0000);
Initial commit
2015-12-04 14:47:52 +00:00
buf1[3] |= tmp;
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 30: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[3] |= __byte_perm_S (tmp, 0, 0x5400);
break;
case 31: buf0[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf0[3] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[0] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[1] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[2] |= __byte_perm_S (tmp, 0, 0x0000);
buf1[3] |= __byte_perm_S (tmp, 0, 0x4000);
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
Initial commit
2015-12-04 14:47:52 +00:00
switch (p0)
{
case 1: buf0[0] |= tmp << 0;
break;
case 2: buf0[0] |= tmp << 0 | tmp << 8;
break;
case 3: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16;
break;
case 4: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
break;
case 5: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0;
break;
case 6: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8;
break;
case 7: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16;
break;
case 8: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
break;
case 9: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0;
break;
case 10: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8;
break;
case 11: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16;
break;
case 12: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
break;
case 13: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0;
break;
case 14: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8;
break;
case 15: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16;
break;
case 16: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
break;
case 17: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0;
break;
case 18: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8;
break;
case 19: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16;
break;
case 20: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
break;
case 21: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[1] |= tmp << 0;
break;
case 22: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[1] |= tmp << 0 | tmp << 8;
break;
case 23: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[1] |= tmp << 0 | tmp << 8 | tmp << 16;
break;
case 24: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
break;
case 25: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[2] |= tmp << 0;
break;
case 26: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[2] |= tmp << 0 | tmp << 8;
break;
case 27: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[2] |= tmp << 0 | tmp << 8 | tmp << 16;
break;
case 28: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
break;
case 29: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[3] |= tmp << 0;
break;
case 30: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[3] |= tmp << 0 | tmp << 8;
break;
case 31: buf0[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf0[3] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[0] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[1] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[2] |= tmp << 0 | tmp << 8 | tmp << 16 | tmp << 24;
buf1[3] |= tmp << 0 | tmp << 8 | tmp << 16;
break;
}
#endif
out_len += p0;
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_dupechar_last (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if ( in_len == 0) return (in_len);
if ((in_len + p0) >= 32) return (in_len);
const u32 in_len1 = in_len - 1;
const u32 sh = (in_len1 & 3) * 8;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tmp = 0;
Initial commit
2015-12-04 14:47:52 +00:00
switch (in_len1 / 4)
{
case 0: tmp = (buf0[0] >> sh) & 0xff; break;
case 1: tmp = (buf0[1] >> sh) & 0xff; break;
case 2: tmp = (buf0[2] >> sh) & 0xff; break;
case 3: tmp = (buf0[3] >> sh) & 0xff; break;
case 4: tmp = (buf1[0] >> sh) & 0xff; break;
case 5: tmp = (buf1[1] >> sh) & 0xff; break;
case 6: tmp = (buf1[2] >> sh) & 0xff; break;
case 7: tmp = (buf1[3] >> sh) & 0xff; break;
}
u32 out_len = in_len;
for (u32 i = 0; i < p0; i++)
{
append_block1 (out_len, buf0, buf1, tmp);
out_len++;
}
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_dupechar_all (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if ( in_len == 0) return (in_len);
if ((in_len + in_len) >= 32) return (in_len);
u32 out_len = in_len;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
Initial commit
2015-12-04 14:47:52 +00:00
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
tib40[0] = __byte_perm_S (buf0[0], 0, 0x1100);
tib40[1] = __byte_perm_S (buf0[0], 0, 0x3322);
tib40[2] = __byte_perm_S (buf0[1], 0, 0x1100);
tib40[3] = __byte_perm_S (buf0[1], 0, 0x3322);
tib41[0] = __byte_perm_S (buf0[2], 0, 0x1100);
tib41[1] = __byte_perm_S (buf0[2], 0, 0x3322);
tib41[2] = __byte_perm_S (buf0[3], 0, 0x1100);
tib41[3] = __byte_perm_S (buf0[3], 0, 0x3322);
Initial commit
2015-12-04 14:47:52 +00:00
buf0[0] = tib40[0];
buf0[1] = tib40[1];
buf0[2] = tib40[2];
buf0[3] = tib40[3];
buf1[0] = tib41[0];
buf1[1] = tib41[1];
buf1[2] = tib41[2];
buf1[3] = tib41[3];
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
Initial commit
2015-12-04 14:47:52 +00:00
tib40[0] = ((buf0[0] & 0x000000FF) << 0) | ((buf0[0] & 0x0000FF00) << 8);
tib40[1] = ((buf0[0] & 0x00FF0000) >> 16) | ((buf0[0] & 0xFF000000) >> 8);
tib40[2] = ((buf0[1] & 0x000000FF) << 0) | ((buf0[1] & 0x0000FF00) << 8);
tib40[3] = ((buf0[1] & 0x00FF0000) >> 16) | ((buf0[1] & 0xFF000000) >> 8);
tib41[0] = ((buf0[2] & 0x000000FF) << 0) | ((buf0[2] & 0x0000FF00) << 8);
tib41[1] = ((buf0[2] & 0x00FF0000) >> 16) | ((buf0[2] & 0xFF000000) >> 8);
tib41[2] = ((buf0[3] & 0x000000FF) << 0) | ((buf0[3] & 0x0000FF00) << 8);
tib41[3] = ((buf0[3] & 0x00FF0000) >> 16) | ((buf0[3] & 0xFF000000) >> 8);
buf0[0] = tib40[0] | (tib40[0] << 8);
buf0[1] = tib40[1] | (tib40[1] << 8);
buf0[2] = tib40[2] | (tib40[2] << 8);
buf0[3] = tib40[3] | (tib40[3] << 8);
buf1[0] = tib41[0] | (tib41[0] << 8);
buf1[1] = tib41[1] | (tib41[1] << 8);
buf1[2] = tib41[2] | (tib41[2] << 8);
buf1[3] = tib41[3] | (tib41[3] << 8);
#endif
out_len = out_len + out_len;
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_switch_first (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (in_len < 2) return (in_len);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
buf0[0] = __byte_perm_S (buf0[0], 0, 0x3201);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
Initial commit
2015-12-04 14:47:52 +00:00
buf0[0] = (buf0[0] & 0xFFFF0000) | ((buf0[0] << 8) & 0x0000FF00) | ((buf0[0] >> 8) & 0x000000FF);
#endif
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_switch_last (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (in_len < 2) return (in_len);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Initial commit
2015-12-04 14:47:52 +00:00
switch (in_len)
{
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 2: buf0[0] = __byte_perm_S (buf0[0], 0, 0x5401);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 3: buf0[0] = __byte_perm_S (buf0[0], 0, 0x4120);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 4: buf0[0] = __byte_perm_S (buf0[0], 0, 0x2310);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 5: buf0[1] = __byte_perm_S (buf0[1], buf0[0], 0x7210);
buf0[0] = __byte_perm_S (buf0[0], buf0[1], 0x4210);
buf0[1] = __byte_perm_S (buf0[1], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 6: buf0[1] = __byte_perm_S (buf0[1], 0, 0x5401);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 7: buf0[1] = __byte_perm_S (buf0[1], 0, 0x4120);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 8: buf0[1] = __byte_perm_S (buf0[1], 0, 0x2310);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 9: buf0[2] = __byte_perm_S (buf0[2], buf0[1], 0x7210);
buf0[1] = __byte_perm_S (buf0[1], buf0[2], 0x4210);
buf0[2] = __byte_perm_S (buf0[2], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 10: buf0[2] = __byte_perm_S (buf0[2], 0, 0x5401);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 11: buf0[2] = __byte_perm_S (buf0[2], 0, 0x4120);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 12: buf0[2] = __byte_perm_S (buf0[2], 0, 0x2310);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 13: buf0[3] = __byte_perm_S (buf0[3], buf0[2], 0x7210);
buf0[2] = __byte_perm_S (buf0[2], buf0[3], 0x4210);
buf0[3] = __byte_perm_S (buf0[3], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 14: buf0[3] = __byte_perm_S (buf0[3], 0, 0x5401);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 15: buf0[3] = __byte_perm_S (buf0[3], 0, 0x4120);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 16: buf0[3] = __byte_perm_S (buf0[3], 0, 0x2310);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 17: buf1[0] = __byte_perm_S (buf1[0], buf0[3], 0x7210);
buf0[3] = __byte_perm_S (buf0[3], buf1[0], 0x4210);
buf1[0] = __byte_perm_S (buf1[0], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 18: buf1[0] = __byte_perm_S (buf1[0], 0, 0x5401);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 19: buf1[0] = __byte_perm_S (buf1[0], 0, 0x4120);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 20: buf1[0] = __byte_perm_S (buf1[0], 0, 0x2310);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 21: buf1[1] = __byte_perm_S (buf1[1], buf1[0], 0x7210);
buf1[0] = __byte_perm_S (buf1[0], buf1[1], 0x4210);
buf1[1] = __byte_perm_S (buf1[1], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 22: buf1[1] = __byte_perm_S (buf1[1], 0, 0x5401);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 23: buf1[1] = __byte_perm_S (buf1[1], 0, 0x4120);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 24: buf1[1] = __byte_perm_S (buf1[1], 0, 0x2310);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 25: buf1[2] = __byte_perm_S (buf1[2], buf1[1], 0x7210);
buf1[1] = __byte_perm_S (buf1[1], buf1[2], 0x4210);
buf1[2] = __byte_perm_S (buf1[2], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 26: buf1[2] = __byte_perm_S (buf1[2], 0, 0x5401);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 27: buf1[2] = __byte_perm_S (buf1[2], 0, 0x4120);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 28: buf1[2] = __byte_perm_S (buf1[2], 0, 0x2310);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 29: buf1[3] = __byte_perm_S (buf1[3], buf1[2], 0x7210);
buf1[2] = __byte_perm_S (buf1[2], buf1[3], 0x4210);
buf1[3] = __byte_perm_S (buf1[3], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 30: buf1[3] = __byte_perm_S (buf1[3], 0, 0x5401);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 31: buf1[3] = __byte_perm_S (buf1[3], 0, 0x4120);
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
Initial commit
2015-12-04 14:47:52 +00:00
switch (in_len)
{
case 2: buf0[0] = ((buf0[0] << 8) & 0x0000FF00) | ((buf0[0] >> 8) & 0x000000FF);
break;
case 3: buf0[0] = (buf0[0] & 0x000000FF) | ((buf0[0] << 8) & 0x00FF0000) | ((buf0[0] >> 8) & 0x0000FF00);
break;
case 4: buf0[0] = (buf0[0] & 0x0000FFFF) | ((buf0[0] << 8) & 0xFF000000) | ((buf0[0] >> 8) & 0x00FF0000);
break;
case 5: buf0[1] = (buf0[0] & 0xFF000000) | buf0[1];
buf0[0] = (buf0[0] & 0x00FFFFFF) | (buf0[1] << 24);
buf0[1] = (buf0[1] >> 24);
break;
case 6: buf0[1] = ((buf0[1] << 8) & 0x0000FF00) | ((buf0[1] >> 8) & 0x000000FF);
break;
case 7: buf0[1] = (buf0[1] & 0x000000FF) | ((buf0[1] << 8) & 0x00FF0000) | ((buf0[1] >> 8) & 0x0000FF00);
break;
case 8: buf0[1] = (buf0[1] & 0x0000FFFF) | ((buf0[1] << 8) & 0xFF000000) | ((buf0[1] >> 8) & 0x00FF0000);
break;
case 9: buf0[2] = (buf0[1] & 0xFF000000) | buf0[2];
buf0[1] = (buf0[1] & 0x00FFFFFF) | (buf0[2] << 24);
buf0[2] = (buf0[2] >> 24);
break;
case 10: buf0[2] = ((buf0[2] << 8) & 0x0000FF00) | ((buf0[2] >> 8) & 0x000000FF);
break;
case 11: buf0[2] = (buf0[2] & 0x000000FF) | ((buf0[2] << 8) & 0x00FF0000) | ((buf0[2] >> 8) & 0x0000FF00);
break;
case 12: buf0[2] = (buf0[2] & 0x0000FFFF) | ((buf0[2] << 8) & 0xFF000000) | ((buf0[2] >> 8) & 0x00FF0000);
break;
case 13: buf0[3] = (buf0[2] & 0xFF000000) | buf0[3];
buf0[2] = (buf0[2] & 0x00FFFFFF) | (buf0[3] << 24);
buf0[3] = (buf0[3] >> 24);
break;
case 14: buf0[3] = ((buf0[3] << 8) & 0x0000FF00) | ((buf0[3] >> 8) & 0x000000FF);
break;
case 15: buf0[3] = (buf0[3] & 0x000000FF) | ((buf0[3] << 8) & 0x00FF0000) | ((buf0[3] >> 8) & 0x0000FF00);
break;
case 16: buf0[3] = (buf0[3] & 0x0000FFFF) | ((buf0[3] << 8) & 0xFF000000) | ((buf0[3] >> 8) & 0x00FF0000);
break;
case 17: buf1[0] = (buf0[3] & 0xFF000000) | buf1[0];
buf0[3] = (buf0[3] & 0x00FFFFFF) | (buf1[0] << 24);
buf1[0] = (buf1[0] >> 24);
break;
case 18: buf1[0] = ((buf1[0] << 8) & 0x0000FF00) | ((buf1[0] >> 8) & 0x000000FF);
break;
case 19: buf1[0] = (buf1[0] & 0x000000FF) | ((buf1[0] << 8) & 0x00FF0000) | ((buf1[0] >> 8) & 0x0000FF00);
break;
case 20: buf1[0] = (buf1[0] & 0x0000FFFF) | ((buf1[0] << 8) & 0xFF000000) | ((buf1[0] >> 8) & 0x00FF0000);
break;
case 21: buf1[1] = (buf1[0] & 0xFF000000) | buf1[1];
buf1[0] = (buf1[0] & 0x00FFFFFF) | (buf1[1] << 24);
buf1[1] = (buf1[1] >> 24);
break;
case 22: buf1[1] = ((buf1[1] << 8) & 0x0000FF00) | ((buf1[1] >> 8) & 0x000000FF);
break;
case 23: buf1[1] = (buf1[1] & 0x000000FF) | ((buf1[1] << 8) & 0x00FF0000) | ((buf1[1] >> 8) & 0x0000FF00);
break;
case 24: buf1[1] = (buf1[1] & 0x0000FFFF) | ((buf1[1] << 8) & 0xFF000000) | ((buf1[1] >> 8) & 0x00FF0000);
break;
case 25: buf1[2] = (buf1[1] & 0xFF000000) | buf1[2];
buf1[1] = (buf1[1] & 0x00FFFFFF) | (buf1[2] << 24);
buf1[2] = (buf1[2] >> 24);
break;
case 26: buf1[2] = ((buf1[2] << 8) & 0x0000FF00) | ((buf1[2] >> 8) & 0x000000FF);
break;
case 27: buf1[2] = (buf1[2] & 0x000000FF) | ((buf1[2] << 8) & 0x00FF0000) | ((buf1[2] >> 8) & 0x0000FF00);
break;
case 28: buf1[2] = (buf1[2] & 0x0000FFFF) | ((buf1[2] << 8) & 0xFF000000) | ((buf1[2] >> 8) & 0x00FF0000);
break;
case 29: buf1[3] = (buf1[2] & 0xFF000000) | buf1[3];
buf1[2] = (buf1[2] & 0x00FFFFFF) | (buf1[3] << 24);
buf1[3] = (buf1[3] >> 24);
break;
case 30: buf1[3] = ((buf1[3] << 8) & 0x0000FF00) | ((buf1[3] >> 8) & 0x000000FF);
break;
case 31: buf1[3] = (buf1[3] & 0x000000FF) | ((buf1[3] << 8) & 0x00FF0000) | ((buf1[3] >> 8) & 0x0000FF00);
break;
}
#endif
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_switch_at (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 >= in_len) return (in_len);
if (p1 >= in_len) return (in_len);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tmp0 = 0;
u32 tmp1 = 0;
Initial commit
2015-12-04 14:47:52 +00:00
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Initial commit
2015-12-04 14:47:52 +00:00
switch (p0)
{
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 0: tmp0 = __byte_perm_S (buf0[0], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 1: tmp0 = __byte_perm_S (buf0[0], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 2: tmp0 = __byte_perm_S (buf0[0], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 3: tmp0 = __byte_perm_S (buf0[0], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 4: tmp0 = __byte_perm_S (buf0[1], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 5: tmp0 = __byte_perm_S (buf0[1], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 6: tmp0 = __byte_perm_S (buf0[1], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 7: tmp0 = __byte_perm_S (buf0[1], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 8: tmp0 = __byte_perm_S (buf0[2], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 9: tmp0 = __byte_perm_S (buf0[2], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 10: tmp0 = __byte_perm_S (buf0[2], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 11: tmp0 = __byte_perm_S (buf0[2], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 12: tmp0 = __byte_perm_S (buf0[3], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 13: tmp0 = __byte_perm_S (buf0[3], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 14: tmp0 = __byte_perm_S (buf0[3], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 15: tmp0 = __byte_perm_S (buf0[3], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 16: tmp0 = __byte_perm_S (buf1[0], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 17: tmp0 = __byte_perm_S (buf1[0], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 18: tmp0 = __byte_perm_S (buf1[0], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 19: tmp0 = __byte_perm_S (buf1[0], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 20: tmp0 = __byte_perm_S (buf1[1], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 21: tmp0 = __byte_perm_S (buf1[1], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 22: tmp0 = __byte_perm_S (buf1[1], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 23: tmp0 = __byte_perm_S (buf1[1], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 24: tmp0 = __byte_perm_S (buf1[2], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 25: tmp0 = __byte_perm_S (buf1[2], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 26: tmp0 = __byte_perm_S (buf1[2], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 27: tmp0 = __byte_perm_S (buf1[2], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 28: tmp0 = __byte_perm_S (buf1[3], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 29: tmp0 = __byte_perm_S (buf1[3], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 30: tmp0 = __byte_perm_S (buf1[3], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 31: tmp0 = __byte_perm_S (buf1[3], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
switch (p1)
{
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 0: tmp1 = __byte_perm_S (buf0[0], 0, 0x6540);
buf0[0] = __byte_perm_S (tmp0, buf0[0], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 1: tmp1 = __byte_perm_S (buf0[0], 0, 0x6541);
buf0[0] = __byte_perm_S (tmp0, buf0[0], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 2: tmp1 = __byte_perm_S (buf0[0], 0, 0x6542);
buf0[0] = __byte_perm_S (tmp0, buf0[0], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 3: tmp1 = __byte_perm_S (buf0[0], 0, 0x6543);
buf0[0] = __byte_perm_S (tmp0, buf0[0], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 4: tmp1 = __byte_perm_S (buf0[1], 0, 0x6540);
buf0[1] = __byte_perm_S (tmp0, buf0[1], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 5: tmp1 = __byte_perm_S (buf0[1], 0, 0x6541);
buf0[1] = __byte_perm_S (tmp0, buf0[1], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 6: tmp1 = __byte_perm_S (buf0[1], 0, 0x6542);
buf0[1] = __byte_perm_S (tmp0, buf0[1], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 7: tmp1 = __byte_perm_S (buf0[1], 0, 0x6543);
buf0[1] = __byte_perm_S (tmp0, buf0[1], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 8: tmp1 = __byte_perm_S (buf0[2], 0, 0x6540);
buf0[2] = __byte_perm_S (tmp0, buf0[2], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 9: tmp1 = __byte_perm_S (buf0[2], 0, 0x6541);
buf0[2] = __byte_perm_S (tmp0, buf0[2], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 10: tmp1 = __byte_perm_S (buf0[2], 0, 0x6542);
buf0[2] = __byte_perm_S (tmp0, buf0[2], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 11: tmp1 = __byte_perm_S (buf0[2], 0, 0x6543);
buf0[2] = __byte_perm_S (tmp0, buf0[2], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 12: tmp1 = __byte_perm_S (buf0[3], 0, 0x6540);
buf0[3] = __byte_perm_S (tmp0, buf0[3], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 13: tmp1 = __byte_perm_S (buf0[3], 0, 0x6541);
buf0[3] = __byte_perm_S (tmp0, buf0[3], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 14: tmp1 = __byte_perm_S (buf0[3], 0, 0x6542);
buf0[3] = __byte_perm_S (tmp0, buf0[3], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 15: tmp1 = __byte_perm_S (buf0[3], 0, 0x6543);
buf0[3] = __byte_perm_S (tmp0, buf0[3], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 16: tmp1 = __byte_perm_S (buf1[0], 0, 0x6540);
buf1[0] = __byte_perm_S (tmp0, buf1[0], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 17: tmp1 = __byte_perm_S (buf1[0], 0, 0x6541);
buf1[0] = __byte_perm_S (tmp0, buf1[0], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 18: tmp1 = __byte_perm_S (buf1[0], 0, 0x6542);
buf1[0] = __byte_perm_S (tmp0, buf1[0], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 19: tmp1 = __byte_perm_S (buf1[0], 0, 0x6543);
buf1[0] = __byte_perm_S (tmp0, buf1[0], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 20: tmp1 = __byte_perm_S (buf1[1], 0, 0x6540);
buf1[1] = __byte_perm_S (tmp0, buf1[1], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 21: tmp1 = __byte_perm_S (buf1[1], 0, 0x6541);
buf1[1] = __byte_perm_S (tmp0, buf1[1], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 22: tmp1 = __byte_perm_S (buf1[1], 0, 0x6542);
buf1[1] = __byte_perm_S (tmp0, buf1[1], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 23: tmp1 = __byte_perm_S (buf1[1], 0, 0x6543);
buf1[1] = __byte_perm_S (tmp0, buf1[1], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 24: tmp1 = __byte_perm_S (buf1[2], 0, 0x6540);
buf1[2] = __byte_perm_S (tmp0, buf1[2], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 25: tmp1 = __byte_perm_S (buf1[2], 0, 0x6541);
buf1[2] = __byte_perm_S (tmp0, buf1[2], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 26: tmp1 = __byte_perm_S (buf1[2], 0, 0x6542);
buf1[2] = __byte_perm_S (tmp0, buf1[2], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 27: tmp1 = __byte_perm_S (buf1[2], 0, 0x6543);
buf1[2] = __byte_perm_S (tmp0, buf1[2], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 28: tmp1 = __byte_perm_S (buf1[3], 0, 0x6540);
buf1[3] = __byte_perm_S (tmp0, buf1[3], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 29: tmp1 = __byte_perm_S (buf1[3], 0, 0x6541);
buf1[3] = __byte_perm_S (tmp0, buf1[3], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 30: tmp1 = __byte_perm_S (buf1[3], 0, 0x6542);
buf1[3] = __byte_perm_S (tmp0, buf1[3], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 31: tmp1 = __byte_perm_S (buf1[3], 0, 0x6543);
buf1[3] = __byte_perm_S (tmp0, buf1[3], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
switch (p0)
{
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 0: buf0[0] = __byte_perm_S (tmp1, buf0[0], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 1: buf0[0] = __byte_perm_S (tmp1, buf0[0], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 2: buf0[0] = __byte_perm_S (tmp1, buf0[0], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 3: buf0[0] = __byte_perm_S (tmp1, buf0[0], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 4: buf0[1] = __byte_perm_S (tmp1, buf0[1], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 5: buf0[1] = __byte_perm_S (tmp1, buf0[1], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 6: buf0[1] = __byte_perm_S (tmp1, buf0[1], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 7: buf0[1] = __byte_perm_S (tmp1, buf0[1], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 8: buf0[2] = __byte_perm_S (tmp1, buf0[2], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 9: buf0[2] = __byte_perm_S (tmp1, buf0[2], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 10: buf0[2] = __byte_perm_S (tmp1, buf0[2], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 11: buf0[2] = __byte_perm_S (tmp1, buf0[2], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 12: buf0[3] = __byte_perm_S (tmp1, buf0[3], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 13: buf0[3] = __byte_perm_S (tmp1, buf0[3], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 14: buf0[3] = __byte_perm_S (tmp1, buf0[3], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 15: buf0[3] = __byte_perm_S (tmp1, buf0[3], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 16: buf1[0] = __byte_perm_S (tmp1, buf1[0], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 17: buf1[0] = __byte_perm_S (tmp1, buf1[0], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 18: buf1[0] = __byte_perm_S (tmp1, buf1[0], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 19: buf1[0] = __byte_perm_S (tmp1, buf1[0], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 20: buf1[1] = __byte_perm_S (tmp1, buf1[1], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 21: buf1[1] = __byte_perm_S (tmp1, buf1[1], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 22: buf1[1] = __byte_perm_S (tmp1, buf1[1], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 23: buf1[1] = __byte_perm_S (tmp1, buf1[1], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 24: buf1[2] = __byte_perm_S (tmp1, buf1[2], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 25: buf1[2] = __byte_perm_S (tmp1, buf1[2], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 26: buf1[2] = __byte_perm_S (tmp1, buf1[2], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 27: buf1[2] = __byte_perm_S (tmp1, buf1[2], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 28: buf1[3] = __byte_perm_S (tmp1, buf1[3], 0x7650);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 29: buf1[3] = __byte_perm_S (tmp1, buf1[3], 0x7604);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 30: buf1[3] = __byte_perm_S (tmp1, buf1[3], 0x7054);
Initial commit
2015-12-04 14:47:52 +00:00
break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 31: buf1[3] = __byte_perm_S (tmp1, buf1[3], 0x0654);
Initial commit
2015-12-04 14:47:52 +00:00
break;
}
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
Initial commit
2015-12-04 14:47:52 +00:00
switch (p0)
{
case 0: tmp0 = (buf0[0] >> 0) & 0xFF;
break;
case 1: tmp0 = (buf0[0] >> 8) & 0xFF;
break;
case 2: tmp0 = (buf0[0] >> 16) & 0xFF;
break;
case 3: tmp0 = (buf0[0] >> 24) & 0xFF;
break;
case 4: tmp0 = (buf0[1] >> 0) & 0xFF;
break;
case 5: tmp0 = (buf0[1] >> 8) & 0xFF;
break;
case 6: tmp0 = (buf0[1] >> 16) & 0xFF;
break;
case 7: tmp0 = (buf0[1] >> 24) & 0xFF;
break;
case 8: tmp0 = (buf0[2] >> 0) & 0xFF;
break;
case 9: tmp0 = (buf0[2] >> 8) & 0xFF;
break;
case 10: tmp0 = (buf0[2] >> 16) & 0xFF;
break;
case 11: tmp0 = (buf0[2] >> 24) & 0xFF;
break;
case 12: tmp0 = (buf0[3] >> 0) & 0xFF;
break;
case 13: tmp0 = (buf0[3] >> 8) & 0xFF;
break;
case 14: tmp0 = (buf0[3] >> 16) & 0xFF;
break;
case 15: tmp0 = (buf0[3] >> 24) & 0xFF;
break;
case 16: tmp0 = (buf1[0] >> 0) & 0xFF;
break;
case 17: tmp0 = (buf1[0] >> 8) & 0xFF;
break;
case 18: tmp0 = (buf1[0] >> 16) & 0xFF;
break;
case 19: tmp0 = (buf1[0] >> 24) & 0xFF;
break;
case 20: tmp0 = (buf1[1] >> 0) & 0xFF;
break;
case 21: tmp0 = (buf1[1] >> 8) & 0xFF;
break;
case 22: tmp0 = (buf1[1] >> 16) & 0xFF;
break;
case 23: tmp0 = (buf1[1] >> 24) & 0xFF;
break;
case 24: tmp0 = (buf1[2] >> 0) & 0xFF;
break;
case 25: tmp0 = (buf1[2] >> 8) & 0xFF;
break;
case 26: tmp0 = (buf1[2] >> 16) & 0xFF;
break;
case 27: tmp0 = (buf1[2] >> 24) & 0xFF;
break;
case 28: tmp0 = (buf1[3] >> 0) & 0xFF;
break;
case 29: tmp0 = (buf1[3] >> 8) & 0xFF;
break;
case 30: tmp0 = (buf1[3] >> 16) & 0xFF;
break;
case 31: tmp0 = (buf1[3] >> 24) & 0xFF;
break;
}
switch (p1)
{
case 0: tmp1 = (buf0[0] >> 0) & 0xff;
buf0[0] = (buf0[0] & 0xffffff00) | tmp0 << 0;
break;
case 1: tmp1 = (buf0[0] >> 8) & 0xff;
buf0[0] = (buf0[0] & 0xffff00ff) | tmp0 << 8;
break;
case 2: tmp1 = (buf0[0] >> 16) & 0xff;
buf0[0] = (buf0[0] & 0xff00ffff) | tmp0 << 16;
break;
case 3: tmp1 = (buf0[0] >> 24) & 0xff;
buf0[0] = (buf0[0] & 0x00ffffff) | tmp0 << 24;
break;
case 4: tmp1 = (buf0[1] >> 0) & 0xff;
buf0[1] = (buf0[1] & 0xffffff00) | tmp0 << 0;
break;
case 5: tmp1 = (buf0[1] >> 8) & 0xff;
buf0[1] = (buf0[1] & 0xffff00ff) | tmp0 << 8;
break;
case 6: tmp1 = (buf0[1] >> 16) & 0xff;
buf0[1] = (buf0[1] & 0xff00ffff) | tmp0 << 16;
break;
case 7: tmp1 = (buf0[1] >> 24) & 0xff;
buf0[1] = (buf0[1] & 0x00ffffff) | tmp0 << 24;
break;
case 8: tmp1 = (buf0[2] >> 0) & 0xff;
buf0[2] = (buf0[2] & 0xffffff00) | tmp0 << 0;
break;
case 9: tmp1 = (buf0[2] >> 8) & 0xff;
buf0[2] = (buf0[2] & 0xffff00ff) | tmp0 << 8;
break;
case 10: tmp1 = (buf0[2] >> 16) & 0xff;
buf0[2] = (buf0[2] & 0xff00ffff) | tmp0 << 16;
break;
case 11: tmp1 = (buf0[2] >> 24) & 0xff;
buf0[2] = (buf0[2] & 0x00ffffff) | tmp0 << 24;
break;
case 12: tmp1 = (buf0[3] >> 0) & 0xff;
buf0[3] = (buf0[3] & 0xffffff00) | tmp0 << 0;
break;
case 13: tmp1 = (buf0[3] >> 8) & 0xff;
buf0[3] = (buf0[3] & 0xffff00ff) | tmp0 << 8;
break;
case 14: tmp1 = (buf0[3] >> 16) & 0xff;
buf0[3] = (buf0[3] & 0xff00ffff) | tmp0 << 16;
break;
case 15: tmp1 = (buf0[3] >> 24) & 0xff;
buf0[3] = (buf0[3] & 0x00ffffff) | tmp0 << 24;
break;
case 16: tmp1 = (buf1[0] >> 0) & 0xff;
buf1[0] = (buf1[0] & 0xffffff00) | tmp0 << 0;
break;
case 17: tmp1 = (buf1[0] >> 8) & 0xff;
buf1[0] = (buf1[0] & 0xffff00ff) | tmp0 << 8;
break;
case 18: tmp1 = (buf1[0] >> 16) & 0xff;
buf1[0] = (buf1[0] & 0xff00ffff) | tmp0 << 16;
break;
case 19: tmp1 = (buf1[0] >> 24) & 0xff;
buf1[0] = (buf1[0] & 0x00ffffff) | tmp0 << 24;
break;
case 20: tmp1 = (buf1[1] >> 0) & 0xff;
buf1[1] = (buf1[1] & 0xffffff00) | tmp0 << 0;
break;
case 21: tmp1 = (buf1[1] >> 8) & 0xff;
buf1[1] = (buf1[1] & 0xffff00ff) | tmp0 << 8;
break;
case 22: tmp1 = (buf1[1] >> 16) & 0xff;
buf1[1] = (buf1[1] & 0xff00ffff) | tmp0 << 16;
break;
case 23: tmp1 = (buf1[1] >> 24) & 0xff;
buf1[1] = (buf1[1] & 0x00ffffff) | tmp0 << 24;
break;
case 24: tmp1 = (buf1[2] >> 0) & 0xff;
buf1[2] = (buf1[2] & 0xffffff00) | tmp0 << 0;
break;
case 25: tmp1 = (buf1[2] >> 8) & 0xff;
buf1[2] = (buf1[2] & 0xffff00ff) | tmp0 << 8;
break;
case 26: tmp1 = (buf1[2] >> 16) & 0xff;
buf1[2] = (buf1[2] & 0xff00ffff) | tmp0 << 16;
break;
case 27: tmp1 = (buf1[2] >> 24) & 0xff;
buf1[2] = (buf1[2] & 0x00ffffff) | tmp0 << 24;
break;
case 28: tmp1 = (buf1[3] >> 0) & 0xff;
buf1[3] = (buf1[3] & 0xffffff00) | tmp0 << 0;
break;
case 29: tmp1 = (buf1[3] >> 8) & 0xff;
buf1[3] = (buf1[3] & 0xffff00ff) | tmp0 << 8;
break;
case 30: tmp1 = (buf1[3] >> 16) & 0xff;
buf1[3] = (buf1[3] & 0xff00ffff) | tmp0 << 16;
break;
case 31: tmp1 = (buf1[3] >> 24) & 0xff;
buf1[3] = (buf1[3] & 0x00ffffff) | tmp0 << 24;
break;
}
switch (p0)
{
case 0: buf0[0] = (buf0[0] & 0xffffff00) | tmp1 << 0;
break;
case 1: buf0[0] = (buf0[0] & 0xffff00ff) | tmp1 << 8;
break;
case 2: buf0[0] = (buf0[0] & 0xff00ffff) | tmp1 << 16;
break;
case 3: buf0[0] = (buf0[0] & 0x00ffffff) | tmp1 << 24;
break;
case 4: buf0[1] = (buf0[1] & 0xffffff00) | tmp1 << 0;
break;
case 5: buf0[1] = (buf0[1] & 0xffff00ff) | tmp1 << 8;
break;
case 6: buf0[1] = (buf0[1] & 0xff00ffff) | tmp1 << 16;
break;
case 7: buf0[1] = (buf0[1] & 0x00ffffff) | tmp1 << 24;
break;
case 8: buf0[2] = (buf0[2] & 0xffffff00) | tmp1 << 0;
break;
case 9: buf0[2] = (buf0[2] & 0xffff00ff) | tmp1 << 8;
break;
case 10: buf0[2] = (buf0[2] & 0xff00ffff) | tmp1 << 16;
break;
case 11: buf0[2] = (buf0[2] & 0x00ffffff) | tmp1 << 24;
break;
case 12: buf0[3] = (buf0[3] & 0xffffff00) | tmp1 << 0;
break;
case 13: buf0[3] = (buf0[3] & 0xffff00ff) | tmp1 << 8;
break;
case 14: buf0[3] = (buf0[3] & 0xff00ffff) | tmp1 << 16;
break;
case 15: buf0[3] = (buf0[3] & 0x00ffffff) | tmp1 << 24;
break;
case 16: buf1[0] = (buf1[0] & 0xffffff00) | tmp1 << 0;
break;
case 17: buf1[0] = (buf1[0] & 0xffff00ff) | tmp1 << 8;
break;
case 18: buf1[0] = (buf1[0] & 0xff00ffff) | tmp1 << 16;
break;
case 19: buf1[0] = (buf1[0] & 0x00ffffff) | tmp1 << 24;
break;
case 20: buf1[1] = (buf1[1] & 0xffffff00) | tmp1 << 0;
break;
case 21: buf1[1] = (buf1[1] & 0xffff00ff) | tmp1 << 8;
break;
case 22: buf1[1] = (buf1[1] & 0xff00ffff) | tmp1 << 16;
break;
case 23: buf1[1] = (buf1[1] & 0x00ffffff) | tmp1 << 24;
break;
case 24: buf1[2] = (buf1[2] & 0xffffff00) | tmp1 << 0;
break;
case 25: buf1[2] = (buf1[2] & 0xffff00ff) | tmp1 << 8;
break;
case 26: buf1[2] = (buf1[2] & 0xff00ffff) | tmp1 << 16;
break;
case 27: buf1[2] = (buf1[2] & 0x00ffffff) | tmp1 << 24;
break;
case 28: buf1[3] = (buf1[3] & 0xffffff00) | tmp1 << 0;
break;
case 29: buf1[3] = (buf1[3] & 0xffff00ff) | tmp1 << 8;
break;
case 30: buf1[3] = (buf1[3] & 0xff00ffff) | tmp1 << 16;
break;
case 31: buf1[3] = (buf1[3] & 0x00ffffff) | tmp1 << 24;
break;
}
#endif
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_chr_shiftl (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 >= in_len) return (in_len);
Fix more missing casts to uint in rule-engine
2016-01-14 19:54:41 +00:00
const u32 mr = 0xffu << ((p0 & 3) * 8);
Initial commit
2015-12-04 14:47:52 +00:00
const u32 ml = ~mr;
switch (p0 / 4)
{
case 0: buf0[0] = (buf0[0] & ml) | (((buf0[0] & mr) << 1) & mr); break;
case 1: buf0[1] = (buf0[1] & ml) | (((buf0[1] & mr) << 1) & mr); break;
case 2: buf0[2] = (buf0[2] & ml) | (((buf0[2] & mr) << 1) & mr); break;
case 3: buf0[3] = (buf0[3] & ml) | (((buf0[3] & mr) << 1) & mr); break;
case 4: buf1[0] = (buf1[0] & ml) | (((buf1[0] & mr) << 1) & mr); break;
case 5: buf1[1] = (buf1[1] & ml) | (((buf1[1] & mr) << 1) & mr); break;
case 6: buf1[2] = (buf1[2] & ml) | (((buf1[2] & mr) << 1) & mr); break;
case 7: buf1[3] = (buf1[3] & ml) | (((buf1[3] & mr) << 1) & mr); break;
}
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_chr_shiftr (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 >= in_len) return (in_len);
Fix more missing casts to uint in rule-engine
2016-01-14 19:54:41 +00:00
const u32 mr = 0xffu << ((p0 & 3) * 8);
Initial commit
2015-12-04 14:47:52 +00:00
const u32 ml = ~mr;
switch (p0 / 4)
{
case 0: buf0[0] = (buf0[0] & ml) | (((buf0[0] & mr) >> 1) & mr); break;
case 1: buf0[1] = (buf0[1] & ml) | (((buf0[1] & mr) >> 1) & mr); break;
case 2: buf0[2] = (buf0[2] & ml) | (((buf0[2] & mr) >> 1) & mr); break;
case 3: buf0[3] = (buf0[3] & ml) | (((buf0[3] & mr) >> 1) & mr); break;
case 4: buf1[0] = (buf1[0] & ml) | (((buf1[0] & mr) >> 1) & mr); break;
case 5: buf1[1] = (buf1[1] & ml) | (((buf1[1] & mr) >> 1) & mr); break;
case 6: buf1[2] = (buf1[2] & ml) | (((buf1[2] & mr) >> 1) & mr); break;
case 7: buf1[3] = (buf1[3] & ml) | (((buf1[3] & mr) >> 1) & mr); break;
}
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_chr_incr (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 >= in_len) return (in_len);
Fix more missing casts to uint in rule-engine
2016-01-14 19:54:41 +00:00
const u32 mr = 0xffu << ((p0 & 3) * 8);
Initial commit
2015-12-04 14:47:52 +00:00
const u32 ml = ~mr;
const u32 n = 0x01010101 & mr;
switch (p0 / 4)
{
case 0: buf0[0] = (buf0[0] & ml) | (((buf0[0] & mr) + n) & mr); break;
case 1: buf0[1] = (buf0[1] & ml) | (((buf0[1] & mr) + n) & mr); break;
case 2: buf0[2] = (buf0[2] & ml) | (((buf0[2] & mr) + n) & mr); break;
case 3: buf0[3] = (buf0[3] & ml) | (((buf0[3] & mr) + n) & mr); break;
case 4: buf1[0] = (buf1[0] & ml) | (((buf1[0] & mr) + n) & mr); break;
case 5: buf1[1] = (buf1[1] & ml) | (((buf1[1] & mr) + n) & mr); break;
case 6: buf1[2] = (buf1[2] & ml) | (((buf1[2] & mr) + n) & mr); break;
case 7: buf1[3] = (buf1[3] & ml) | (((buf1[3] & mr) + n) & mr); break;
}
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_chr_decr (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 >= in_len) return (in_len);
Fix more missing casts to uint in rule-engine
2016-01-14 19:54:41 +00:00
const u32 mr = 0xffu << ((p0 & 3) * 8);
Initial commit
2015-12-04 14:47:52 +00:00
const u32 ml = ~mr;
const u32 n = 0x01010101 & mr;
switch (p0 / 4)
{
case 0: buf0[0] = (buf0[0] & ml) | (((buf0[0] & mr) - n) & mr); break;
case 1: buf0[1] = (buf0[1] & ml) | (((buf0[1] & mr) - n) & mr); break;
case 2: buf0[2] = (buf0[2] & ml) | (((buf0[2] & mr) - n) & mr); break;
case 3: buf0[3] = (buf0[3] & ml) | (((buf0[3] & mr) - n) & mr); break;
case 4: buf1[0] = (buf1[0] & ml) | (((buf1[0] & mr) - n) & mr); break;
case 5: buf1[1] = (buf1[1] & ml) | (((buf1[1] & mr) - n) & mr); break;
case 6: buf1[2] = (buf1[2] & ml) | (((buf1[2] & mr) - n) & mr); break;
case 7: buf1[3] = (buf1[3] & ml) | (((buf1[3] & mr) - n) & mr); break;
}
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_replace_np1 (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if ((p0 + 1) >= in_len) return (in_len);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
Initial commit
2015-12-04 14:47:52 +00:00
lshift_block (buf0, buf1, tib40, tib41);
Fix more missing casts to uint in rule-engine
2016-01-14 19:54:41 +00:00
const u32 mr = 0xffu << ((p0 & 3) * 8);
Initial commit
2015-12-04 14:47:52 +00:00
const u32 ml = ~mr;
switch (p0 / 4)
{
case 0: buf0[0] = (buf0[0] & ml) | (tib40[0] & mr); break;
case 1: buf0[1] = (buf0[1] & ml) | (tib40[1] & mr); break;
case 2: buf0[2] = (buf0[2] & ml) | (tib40[2] & mr); break;
case 3: buf0[3] = (buf0[3] & ml) | (tib40[3] & mr); break;
case 4: buf1[0] = (buf1[0] & ml) | (tib41[0] & mr); break;
case 5: buf1[1] = (buf1[1] & ml) | (tib41[1] & mr); break;
case 6: buf1[2] = (buf1[2] & ml) | (tib41[2] & mr); break;
case 7: buf1[3] = (buf1[3] & ml) | (tib41[3] & mr); break;
}
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_replace_nm1 (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 == 0) return (in_len);
if (p0 >= in_len) return (in_len);
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
Initial commit
2015-12-04 14:47:52 +00:00
rshift_block (buf0, buf1, tib40, tib41);
Fix more missing casts to uint in rule-engine
2016-01-14 19:54:41 +00:00
const u32 mr = 0xffu << ((p0 & 3) * 8);
Initial commit
2015-12-04 14:47:52 +00:00
const u32 ml = ~mr;
switch (p0 / 4)
{
case 0: buf0[0] = (buf0[0] & ml) | (tib40[0] & mr); break;
case 1: buf0[1] = (buf0[1] & ml) | (tib40[1] & mr); break;
case 2: buf0[2] = (buf0[2] & ml) | (tib40[2] & mr); break;
case 3: buf0[3] = (buf0[3] & ml) | (tib40[3] & mr); break;
case 4: buf1[0] = (buf1[0] & ml) | (tib41[0] & mr); break;
case 5: buf1[1] = (buf1[1] & ml) | (tib41[1] & mr); break;
case 6: buf1[2] = (buf1[2] & ml) | (tib41[2] & mr); break;
case 7: buf1[3] = (buf1[3] & ml) | (tib41[3] & mr); break;
}
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_dupeblock_first (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 > in_len) return (in_len);
if ((in_len + p0) >= 32) return (in_len);
u32 out_len = in_len;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
Initial commit
2015-12-04 14:47:52 +00:00
tib40[0] = buf0[0];
tib40[1] = buf0[1];
tib40[2] = buf0[2];
tib40[3] = buf0[3];
tib41[0] = buf1[0];
tib41[1] = buf1[1];
tib41[2] = buf1[2];
tib41[3] = buf1[3];
truncate_right (tib40, tib41, p0);
rshift_block_N (buf0, buf1, buf0, buf1, p0);
buf0[0] |= tib40[0];
buf0[1] |= tib40[1];
buf0[2] |= tib40[2];
buf0[3] |= tib40[3];
buf1[0] |= tib41[0];
buf1[1] |= tib41[1];
buf1[2] |= tib41[2];
buf1[3] |= tib41[3];
out_len += p0;
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_dupeblock_last (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
if (p0 > in_len) return (in_len);
if ((in_len + p0) >= 32) return (in_len);
u32 out_len = in_len;
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
Initial commit
2015-12-04 14:47:52 +00:00
rshift_block_N (buf0, buf1, tib40, tib41, p0);
truncate_left (tib40, tib41, out_len);
buf0[0] |= tib40[0];
buf0[1] |= tib40[1];
buf0[2] |= tib40[2];
buf0[3] |= tib40[3];
buf1[0] |= tib41[0];
buf1[1] |= tib41[1];
buf1[2] |= tib41[2];
buf1[3] |= tib41[3];
out_len += p0;
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 rule_op_mangle_title (const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
buf0[0] |= (generate_cmask (buf0[0]));
buf0[1] |= (generate_cmask (buf0[1]));
buf0[2] |= (generate_cmask (buf0[2]));
buf0[3] |= (generate_cmask (buf0[3]));
buf1[0] |= (generate_cmask (buf1[0]));
buf1[1] |= (generate_cmask (buf1[1]));
buf1[2] |= (generate_cmask (buf1[2]));
buf1[3] |= (generate_cmask (buf1[3]));
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#ifdef IS_NV
Initial commit
2015-12-04 14:47:52 +00:00
buf0[0] &= ~(0x00000020 & generate_cmask (buf0[0]));
for (u32 i = 0; i < in_len; i++)
{
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tmp0;
u32 tmp1;
Initial commit
2015-12-04 14:47:52 +00:00
switch (i)
{
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 0: tmp0 = __byte_perm_S (buf0[0], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00002000 & generate_cmask (buf0[0])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 1: tmp0 = __byte_perm_S (buf0[0], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00200000 & generate_cmask (buf0[0])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 2: tmp0 = __byte_perm_S (buf0[0], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x20000000 & generate_cmask (buf0[0])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 3: tmp0 = __byte_perm_S (buf0[0], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00000020 & generate_cmask (buf0[1])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 4: tmp0 = __byte_perm_S (buf0[1], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00002000 & generate_cmask (buf0[1])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 5: tmp0 = __byte_perm_S (buf0[1], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00200000 & generate_cmask (buf0[1])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 6: tmp0 = __byte_perm_S (buf0[1], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x20000000 & generate_cmask (buf0[1])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 7: tmp0 = __byte_perm_S (buf0[1], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00000020 & generate_cmask (buf0[2])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 8: tmp0 = __byte_perm_S (buf0[2], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00002000 & generate_cmask (buf0[2])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 9: tmp0 = __byte_perm_S (buf0[2], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00200000 & generate_cmask (buf0[2])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 10: tmp0 = __byte_perm_S (buf0[2], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x20000000 & generate_cmask (buf0[2])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 11: tmp0 = __byte_perm_S (buf0[2], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00000020 & generate_cmask (buf0[3])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 12: tmp0 = __byte_perm_S (buf0[3], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00002000 & generate_cmask (buf0[3])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 13: tmp0 = __byte_perm_S (buf0[3], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00200000 & generate_cmask (buf0[3])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 14: tmp0 = __byte_perm_S (buf0[3], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x20000000 & generate_cmask (buf0[3])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 15: tmp0 = __byte_perm_S (buf0[3], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00000020 & generate_cmask (buf1[0])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 16: tmp0 = __byte_perm_S (buf1[0], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00002000 & generate_cmask (buf1[0])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 17: tmp0 = __byte_perm_S (buf1[0], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00200000 & generate_cmask (buf1[0])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 18: tmp0 = __byte_perm_S (buf1[0], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x20000000 & generate_cmask (buf1[0])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 19: tmp0 = __byte_perm_S (buf1[0], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00000020 & generate_cmask (buf1[1])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 20: tmp0 = __byte_perm_S (buf1[1], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00002000 & generate_cmask (buf1[1])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 21: tmp0 = __byte_perm_S (buf1[1], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00200000 & generate_cmask (buf1[1])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 22: tmp0 = __byte_perm_S (buf1[1], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x20000000 & generate_cmask (buf1[1])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 23: tmp0 = __byte_perm_S (buf1[1], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00000020 & generate_cmask (buf1[2])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 24: tmp0 = __byte_perm_S (buf1[2], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00002000 & generate_cmask (buf1[2])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 25: tmp0 = __byte_perm_S (buf1[2], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00200000 & generate_cmask (buf1[2])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 26: tmp0 = __byte_perm_S (buf1[2], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x20000000 & generate_cmask (buf1[2])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 27: tmp0 = __byte_perm_S (buf1[2], 0, 0x6543);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00000020 & generate_cmask (buf1[3])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 28: tmp0 = __byte_perm_S (buf1[3], 0, 0x6540);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00002000 & generate_cmask (buf1[3])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 29: tmp0 = __byte_perm_S (buf1[3], 0, 0x6541);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x00200000 & generate_cmask (buf1[3])); break;
Fix issue in https://github.com/hashcat/oclHashcat/issues/170
2016-01-27 19:01:08 +00:00
case 30: tmp0 = __byte_perm_S (buf1[3], 0, 0x6542);
Initial commit
2015-12-04 14:47:52 +00:00
tmp1 = ~(0x20000000 & generate_cmask (buf1[3])); break;
}
if (i < 3)
{
if (tmp0 == ' ') buf0[0] &= tmp1 ;
}
else if (i < 7)
{
if (tmp0 == ' ') buf0[1] &= tmp1 ;
}
else if (i < 11)
{
if (tmp0 == ' ') buf0[2] &= tmp1 ;
}
else if (i < 15)
{
if (tmp0 == ' ') buf0[3] &= tmp1 ;
}
else if (i < 19)
{
if (tmp0 == ' ') buf1[0] &= tmp1 ;
}
else if (i < 23)
{
if (tmp0 == ' ') buf1[1] &= tmp1 ;
}
else if (i < 27)
{
if (tmp0 == ' ') buf1[2] &= tmp1 ;
}
else if (i < 31)
{
if (tmp0 == ' ') buf1[3] &= tmp1 ;
}
}
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
Fix for -m 1500, -m 3000: Missing "defined" keyword Fix for -m 1000: MD4_H1 copy/paste error Fix for -m 8900, -m 9300: Invalid value for device_processor_cores for CPU devices Fix for -m 9100: Variable salt2 initializer needed some clauses for clearness Temporary limit gpu_accel for CPU devices to 1 for development phase Mark pocl as too bleeding edge for production use, recommend native drivers Remove workarounds for pocl Rename VENDOR_ID_UNKNOWN to VENDOR_ID_GENERIC in host code Rename IS_UNKNOWN to IS_GENERIC in kernel code
2016-01-07 19:14:34 +00:00
#if defined IS_AMD || defined IS_GENERIC
- Dropped all vector code since new GPU's are all scalar, makes the code much easier - Some performance on low-end GPU may drop because of that, but only for a few hash-modes - Dropped scalar code (aka warp) since we do not have any vector datatypes anymore - Renamed C++ overloading functions memcat32_9 -> memcat_c32_w4x4_a3x4 - Still need to fix kernels to new function names, needs to be done manually - Temperature Management needs to be rewritten partially because of conflicting datatypes names - Added code to create different codepaths for NV on AMD in runtime in host (see data.vendor_id) - Added code to create different codepaths for NV on AMD in runtime in kernels (see IS_NV and IS_AMD) - First tests working for -m 0, for example - Great performance increases in general for NV so far - Tested amp_* and markov_* kernel - Migrated special NV optimizations for rule processor
2015-12-15 11:04:22 +00:00
u32 tib40[4];
u32 tib41[4];
const uchar4 tmp0 = (uchar4) (' ');
const uchar4 tmp1 = (uchar4) (0x00);
const uchar4 tmp2 = (uchar4) (0xff);
uchar4 tmp;
tmp = as_uchar4 (buf0[0]); tmp = select (tmp1, tmp2, tmp == tmp0); tib40[0] = as_uint (tmp);
tmp = as_uchar4 (buf0[1]); tmp = select (tmp1, tmp2, tmp == tmp0); tib40[1] = as_uint (tmp);
tmp = as_uchar4 (buf0[2]); tmp = select (tmp1, tmp2, tmp == tmp0); tib40[2] = as_uint (tmp);
tmp = as_uchar4 (buf0[3]); tmp = select (tmp1, tmp2, tmp == tmp0); tib40[3] = as_uint (tmp);
tmp = as_uchar4 (buf1[0]); tmp = select (tmp1, tmp2, tmp == tmp0); tib41[0] = as_uint (tmp);
tmp = as_uchar4 (buf1[1]); tmp = select (tmp1, tmp2, tmp == tmp0); tib41[1] = as_uint (tmp);
tmp = as_uchar4 (buf1[2]); tmp = select (tmp1, tmp2, tmp == tmp0); tib41[2] = as_uint (tmp);
tmp = as_uchar4 (buf1[3]); tmp = select (tmp1, tmp2, tmp == tmp0); tib41[3] = as_uint (tmp);
rshift_block (tib40, tib41, tib40, tib41); tib40[0] |= 0xff;
buf0[0] &= ~(generate_cmask (buf0[0]) & tib40[0]);
buf0[1] &= ~(generate_cmask (buf0[1]) & tib40[1]);
buf0[2] &= ~(generate_cmask (buf0[2]) & tib40[2]);
buf0[3] &= ~(generate_cmask (buf0[3]) & tib40[3]);
buf1[0] &= ~(generate_cmask (buf1[0]) & tib41[0]);
buf1[1] &= ~(generate_cmask (buf1[1]) & tib41[1]);
buf1[2] &= ~(generate_cmask (buf1[2]) & tib41[2]);
buf1[3] &= ~(generate_cmask (buf1[3]) & tib41[3]);
#endif
Initial commit
2015-12-04 14:47:52 +00:00
return in_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 apply_rule (const u32 name, const u32 p0, const u32 p1, u32 buf0[4], u32 buf1[4], const u32 in_len)
Initial commit
2015-12-04 14:47:52 +00:00
{
u32 out_len = in_len;
switch (name)
{
case RULE_OP_MANGLE_LREST: out_len = rule_op_mangle_lrest (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_UREST: out_len = rule_op_mangle_urest (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_LREST_UFIRST: out_len = rule_op_mangle_lrest_ufirst (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_UREST_LFIRST: out_len = rule_op_mangle_urest_lfirst (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_TREST: out_len = rule_op_mangle_trest (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_TOGGLE_AT: out_len = rule_op_mangle_toggle_at (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_REVERSE: out_len = rule_op_mangle_reverse (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_DUPEWORD: out_len = rule_op_mangle_dupeword (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_DUPEWORD_TIMES: out_len = rule_op_mangle_dupeword_times (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_REFLECT: out_len = rule_op_mangle_reflect (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_APPEND: out_len = rule_op_mangle_append (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_PREPEND: out_len = rule_op_mangle_prepend (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_ROTATE_LEFT: out_len = rule_op_mangle_rotate_left (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_ROTATE_RIGHT: out_len = rule_op_mangle_rotate_right (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_DELETE_FIRST: out_len = rule_op_mangle_delete_first (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_DELETE_LAST: out_len = rule_op_mangle_delete_last (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_DELETE_AT: out_len = rule_op_mangle_delete_at (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_EXTRACT: out_len = rule_op_mangle_extract (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_OMIT: out_len = rule_op_mangle_omit (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_INSERT: out_len = rule_op_mangle_insert (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_OVERSTRIKE: out_len = rule_op_mangle_overstrike (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_TRUNCATE_AT: out_len = rule_op_mangle_truncate_at (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_REPLACE: out_len = rule_op_mangle_replace (p0, p1, buf0, buf1, out_len); break;
//case RULE_OP_MANGLE_PURGECHAR: out_len = rule_op_mangle_purgechar (p0, p1, buf0, buf1, out_len); break;
//case RULE_OP_MANGLE_TOGGLECASE_REC: out_len = rule_op_mangle_togglecase_rec (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_DUPECHAR_FIRST: out_len = rule_op_mangle_dupechar_first (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_DUPECHAR_LAST: out_len = rule_op_mangle_dupechar_last (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_DUPECHAR_ALL: out_len = rule_op_mangle_dupechar_all (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_SWITCH_FIRST: out_len = rule_op_mangle_switch_first (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_SWITCH_LAST: out_len = rule_op_mangle_switch_last (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_SWITCH_AT: out_len = rule_op_mangle_switch_at (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_CHR_SHIFTL: out_len = rule_op_mangle_chr_shiftl (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_CHR_SHIFTR: out_len = rule_op_mangle_chr_shiftr (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_CHR_INCR: out_len = rule_op_mangle_chr_incr (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_CHR_DECR: out_len = rule_op_mangle_chr_decr (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_REPLACE_NP1: out_len = rule_op_mangle_replace_np1 (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_REPLACE_NM1: out_len = rule_op_mangle_replace_nm1 (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_DUPEBLOCK_FIRST: out_len = rule_op_mangle_dupeblock_first (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_DUPEBLOCK_LAST: out_len = rule_op_mangle_dupeblock_last (p0, p1, buf0, buf1, out_len); break;
case RULE_OP_MANGLE_TITLE: out_len = rule_op_mangle_title (p0, p1, buf0, buf1, out_len); break;
}
return out_len;
}
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32 apply_rules (const __global u32 *cmds, u32 buf0[4], u32 buf1[4], const u32 len)
Initial commit
2015-12-04 14:47:52 +00:00
{
u32 out_len = len;
for (u32 i = 0; cmds[i] != 0; i++)
{
const u32 cmd = cmds[i];
const u32 name = (cmd >> 0) & 0xff;
const u32 p0 = (cmd >> 8) & 0xff;
const u32 p1 = (cmd >> 16) & 0xff;
out_len = apply_rule (name, p0, p1, buf0, buf1, out_len);
}
return out_len;
}
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
- Added inline declaration to functions from simd.c, common.c, rp.c and types_ocl.c to increase performance - Dropped static declaration from functions in all kernel to achieve OpenCL 1.1 compatibility - Added -cl-std=CL1.1 to all kernel build options - Created environment variable to inform NVidia OpenCL runtime to not create its own kernel cache - Created environment variable to inform pocl OpenCL runtime to not create its own kernel cache
2016-05-01 21:15:26 +00:00
inline u32x apply_rules_vect (const u32 pw_buf0[4], const u32 pw_buf1[4], const u32 pw_len, const __global kernel_rule_t *rules_buf, const u32 il_pos, u32x w0[4], u32x w1[4])
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
{
#if VECT_SIZE == 1
w0[0] = pw_buf0[0];
w0[1] = pw_buf0[1];
w0[2] = pw_buf0[2];
w0[3] = pw_buf0[3];
w1[0] = pw_buf1[0];
w1[1] = pw_buf1[1];
w1[2] = pw_buf1[2];
w1[3] = pw_buf1[3];
return apply_rules (rules_buf[il_pos].cmds, w0, w1, pw_len);
#else
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
u32x out_len = 0;
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
#ifdef _unroll
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
#pragma unroll
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
#endif
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
for (int i = 0; i < VECT_SIZE; i++)
{
u32 tmp0[4];
u32 tmp1[4];
tmp0[0] = pw_buf0[0];
tmp0[1] = pw_buf0[1];
tmp0[2] = pw_buf0[2];
tmp0[3] = pw_buf0[3];
tmp1[0] = pw_buf1[0];
tmp1[1] = pw_buf1[1];
tmp1[2] = pw_buf1[2];
tmp1[3] = pw_buf1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
const u32 tmp_len = apply_rules (rules_buf[il_pos + i].cmds, tmp0, tmp1, pw_len);
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
switch (i)
{
#if VECT_SIZE >= 2
case 0:
w0[0].s0 = tmp0[0];
w0[1].s0 = tmp0[1];
w0[2].s0 = tmp0[2];
w0[3].s0 = tmp0[3];
w1[0].s0 = tmp1[0];
w1[1].s0 = tmp1[1];
w1[2].s0 = tmp1[2];
w1[3].s0 = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.s0 = tmp_len;
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
break;
case 1:
w0[0].s1 = tmp0[0];
w0[1].s1 = tmp0[1];
w0[2].s1 = tmp0[2];
w0[3].s1 = tmp0[3];
w1[0].s1 = tmp1[0];
w1[1].s1 = tmp1[1];
w1[2].s1 = tmp1[2];
w1[3].s1 = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.s1 = tmp_len;
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
break;
#endif
#if VECT_SIZE >= 4
case 2:
w0[0].s2 = tmp0[0];
w0[1].s2 = tmp0[1];
w0[2].s2 = tmp0[2];
w0[3].s2 = tmp0[3];
w1[0].s2 = tmp1[0];
w1[1].s2 = tmp1[1];
w1[2].s2 = tmp1[2];
w1[3].s2 = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.s2 = tmp_len;
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
break;
case 3:
w0[0].s3 = tmp0[0];
w0[1].s3 = tmp0[1];
w0[2].s3 = tmp0[2];
w0[3].s3 = tmp0[3];
w1[0].s3 = tmp1[0];
w1[1].s3 = tmp1[1];
w1[2].s3 = tmp1[2];
w1[3].s3 = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.s3 = tmp_len;
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
break;
#endif
#if VECT_SIZE >= 8
case 4:
w0[0].s4 = tmp0[0];
w0[1].s4 = tmp0[1];
w0[2].s4 = tmp0[2];
w0[3].s4 = tmp0[3];
w1[0].s4 = tmp1[0];
w1[1].s4 = tmp1[1];
w1[2].s4 = tmp1[2];
w1[3].s4 = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.s4 = tmp_len;
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
break;
case 5:
w0[0].s5 = tmp0[0];
w0[1].s5 = tmp0[1];
w0[2].s5 = tmp0[2];
w0[3].s5 = tmp0[3];
w1[0].s5 = tmp1[0];
w1[1].s5 = tmp1[1];
w1[2].s5 = tmp1[2];
w1[3].s5 = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.s5 = tmp_len;
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
break;
case 6:
w0[0].s6 = tmp0[0];
w0[1].s6 = tmp0[1];
w0[2].s6 = tmp0[2];
w0[3].s6 = tmp0[3];
w1[0].s6 = tmp1[0];
w1[1].s6 = tmp1[1];
w1[2].s6 = tmp1[2];
w1[3].s6 = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.s6 = tmp_len;
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
break;
case 7:
w0[0].s7 = tmp0[0];
w0[1].s7 = tmp0[1];
w0[2].s7 = tmp0[2];
w0[3].s7 = tmp0[3];
w1[0].s7 = tmp1[0];
w1[1].s7 = tmp1[1];
w1[2].s7 = tmp1[2];
w1[3].s7 = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.s7 = tmp_len;
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
break;
#endif
Allow and support vector-width 16, which is current maximum for OpenCL. Closes #226.
2016-02-16 15:42:08 +00:00
#if VECT_SIZE >= 16
case 8:
w0[0].s8 = tmp0[0];
w0[1].s8 = tmp0[1];
w0[2].s8 = tmp0[2];
w0[3].s8 = tmp0[3];
w1[0].s8 = tmp1[0];
w1[1].s8 = tmp1[1];
w1[2].s8 = tmp1[2];
w1[3].s8 = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.s8 = tmp_len;
Allow and support vector-width 16, which is current maximum for OpenCL. Closes #226.
2016-02-16 15:42:08 +00:00
break;
case 9:
w0[0].s9 = tmp0[0];
w0[1].s9 = tmp0[1];
w0[2].s9 = tmp0[2];
w0[3].s9 = tmp0[3];
w1[0].s9 = tmp1[0];
w1[1].s9 = tmp1[1];
w1[2].s9 = tmp1[2];
w1[3].s9 = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.s9 = tmp_len;
Allow and support vector-width 16, which is current maximum for OpenCL. Closes #226.
2016-02-16 15:42:08 +00:00
break;
case 10:
w0[0].sa = tmp0[0];
w0[1].sa = tmp0[1];
w0[2].sa = tmp0[2];
w0[3].sa = tmp0[3];
w1[0].sa = tmp1[0];
w1[1].sa = tmp1[1];
w1[2].sa = tmp1[2];
w1[3].sa = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.sa = tmp_len;
Allow and support vector-width 16, which is current maximum for OpenCL. Closes #226.
2016-02-16 15:42:08 +00:00
break;
case 11:
w0[0].sb = tmp0[0];
w0[1].sb = tmp0[1];
w0[2].sb = tmp0[2];
w0[3].sb = tmp0[3];
w1[0].sb = tmp1[0];
w1[1].sb = tmp1[1];
w1[2].sb = tmp1[2];
w1[3].sb = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.sb = tmp_len;
Allow and support vector-width 16, which is current maximum for OpenCL. Closes #226.
2016-02-16 15:42:08 +00:00
break;
case 12:
w0[0].sc = tmp0[0];
w0[1].sc = tmp0[1];
w0[2].sc = tmp0[2];
w0[3].sc = tmp0[3];
w1[0].sc = tmp1[0];
w1[1].sc = tmp1[1];
w1[2].sc = tmp1[2];
w1[3].sc = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.sc = tmp_len;
Allow and support vector-width 16, which is current maximum for OpenCL. Closes #226.
2016-02-16 15:42:08 +00:00
break;
case 13:
w0[0].sd = tmp0[0];
w0[1].sd = tmp0[1];
w0[2].sd = tmp0[2];
w0[3].sd = tmp0[3];
w1[0].sd = tmp1[0];
w1[1].sd = tmp1[1];
w1[2].sd = tmp1[2];
w1[3].sd = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.sd = tmp_len;
Allow and support vector-width 16, which is current maximum for OpenCL. Closes #226.
2016-02-16 15:42:08 +00:00
break;
case 14:
w0[0].se = tmp0[0];
w0[1].se = tmp0[1];
w0[2].se = tmp0[2];
w0[3].se = tmp0[3];
w1[0].se = tmp1[0];
w1[1].se = tmp1[1];
w1[2].se = tmp1[2];
w1[3].se = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.se = tmp_len;
Allow and support vector-width 16, which is current maximum for OpenCL. Closes #226.
2016-02-16 15:42:08 +00:00
break;
case 15:
w0[0].sf = tmp0[0];
w0[1].sf = tmp0[1];
w0[2].sf = tmp0[2];
w0[3].sf = tmp0[3];
w1[0].sf = tmp1[0];
w1[1].sf = tmp1[1];
w1[2].sf = tmp1[2];
w1[3].sf = tmp1[3];
Fixed two major problems 1) SIMD code for all attack-mode Macro vector_accessible() was not refactored and missing completely. Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way. 2) SIMD code for attack-mode 0 With new SIMD code, apply_rules_vect() has to return u32 not u32x. This has massive impact on all *_a0 kernels. I've rewritten most of them. Deep testing using test.sh is still required. Some kernel need more fixes: - Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c - Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this. This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 16:18:54 +00:00
out_len.sf = tmp_len;
Allow and support vector-width 16, which is current maximum for OpenCL. Closes #226.
2016-02-16 15:42:08 +00:00
break;
#endif
Test convert for -m 1000 with -a 0 for SIMD, speed is now on par or faster than hashcat
2016-01-21 15:47:38 +00:00
}
}
return out_len;
#endif
}
Reference in New Issue Copy Permalink