arno/hashcat
1
0
Fork 0
mirror of https://github.com/hashcat/hashcat.git synced 2024-12-27 17:08:12 +00:00
Code Issues Releases Wiki Activity
68bff94980
hashcat/OpenCL/inc_vendor.cl

247 lines
4.2 KiB
Common Lisp
Raw Normal View History

Initial commit
2015-12-04 14:47:52 +00:00
/**
Added docs/credits.txt Added docs/team.txt
2016-09-11 20:20:15 +00:00
* Author......: See docs/credits.txt
Initial commit
2015-12-04 14:47:52 +00:00
* License.....: MIT
*/
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
/**
* device type
*/
#define DEVICE_TYPE_CPU 2
#define DEVICE_TYPE_GPU 4
#define DEVICE_TYPE_ACCEL 8
#if DEVICE_TYPE == DEVICE_TYPE_CPU
#define IS_CPU
#elif DEVICE_TYPE == DEVICE_TYPE_GPU
#define IS_GPU
#elif DEVICE_TYPE == DEVICE_TYPE_ACCEL
#define IS_ACCEL
Test fix for broken AMD support after https://github.com/hashcat/oclHashcat/pull/303
2016-04-20 19:19:15 +00:00
#endif
Added hash-mode 14600 = LUKS
2017-01-21 14:37:44 +00:00
#if DEVICE_TYPE == DEVICE_TYPE_CPU
#elif DEVICE_TYPE == DEVICE_TYPE_GPU
#define REAL_SHM
#elif DEVICE_TYPE == DEVICE_TYPE_ACCEL
#endif
Do not use __local memory for whirlpool if running on a device without physical shared memory
2017-09-05 14:45:20 +00:00
#ifdef REAL_SHM
#define SHM_TYPE __local
Do not use __local memory for -m 7500 if running on a device without physical shared memory
2017-09-05 15:37:20 +00:00
#define SCR_TYPE __local
Do not use __local memory for whirlpool if running on a device without physical shared memory
2017-09-05 14:45:20 +00:00
#else
#define SHM_TYPE __constant
Do not use __local memory for -m 7500 if running on a device without physical shared memory
2017-09-05 15:37:20 +00:00
#define SCR_TYPE
Do not use __local memory for whirlpool if running on a device without physical shared memory
2017-09-05 14:45:20 +00:00
#endif
Initial commit
2015-12-04 14:47:52 +00:00
/**
* vendor specific
*/
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
#if VENDOR_ID == (1 << 0)
Initial commit
2015-12-04 14:47:52 +00:00
#define IS_AMD
Add fine-tuned AMD GCN control macros
2017-09-07 18:33:43 +00:00
#define AMD_GCN 0
Fixed detection of AMD_GCN version in case the rocm driver is used
2018-07-21 09:52:54 +00:00
#if AMD_ROCM == 1
#if defined __gfx600__ || defined __gfx601__
#undef AMD_GCN
Add fine-tuned AMD GCN control macros
2017-09-07 18:33:43 +00:00
#define AMD_GCN 1
Fixed detection of AMD_GCN version in case the rocm driver is used
2018-07-21 09:52:54 +00:00
#endif
#if defined __gfx700__ || defined __gfx701__ || defined __gfx702__ || defined __gfx703__
#undef AMD_GCN
Add fine-tuned AMD GCN control macros
2017-09-07 18:33:43 +00:00
#define AMD_GCN 2
Fixed detection of AMD_GCN version in case the rocm driver is used
2018-07-21 09:52:54 +00:00
#endif
#if defined __gfx800__ || defined __gfx801__ || defined __gfx802__ || defined __gfx803__ || defined __gfx804__ || defined __gfx810__
#undef AMD_GCN
Add fine-tuned AMD GCN control macros
2017-09-07 18:33:43 +00:00
#define AMD_GCN 3
According to AMD docs, GCN 3 and 4 are the same
2017-09-08 10:46:00 +00:00
// According to AMD docs, GCN 3 and 4 are the same
Fixed detection of AMD_GCN version in case the rocm driver is used
2018-07-21 09:52:54 +00:00
#endif
#if defined __gfx900__ || defined __gfx901__ || defined __gfx902__ || defined __gfx903__
#undef AMD_GCN
Add fine-tuned AMD GCN control macros
2017-09-07 18:33:43 +00:00
#define AMD_GCN 5
Add VEGA specific inline assembly to improve all MD4, MD5, SHA1 and SHA256 based kernels
2017-08-27 17:36:07 +00:00
#endif
Allow OpenCL kernel inline assembly if ROCm drivers was detected
2017-08-22 16:47:53 +00:00
#endif
Added SIMD code for WPA/WPA2
2016-05-14 17:45:51 +00:00
#elif VENDOR_ID == (1 << 1)
#define IS_APPLE
#define IS_GENERIC
#elif VENDOR_ID == (1 << 2)
#define IS_INTEL_BEIGNET
#define IS_GENERIC
#elif VENDOR_ID == (1 << 3)
#define IS_INTEL_SDK
#define IS_GENERIC
#elif VENDOR_ID == (1 << 4)
#define IS_MESA
#define IS_GENERIC
#elif VENDOR_ID == (1 << 5)
Initial commit
2015-12-04 14:47:52 +00:00
#define IS_NV
Added SIMD code for WPA/WPA2
2016-05-14 17:45:51 +00:00
#elif VENDOR_ID == (1 << 6)
#define IS_POCL
#define IS_GENERIC
Added support to utilize multiple different OpenCL platforms in parallel, ex: AMD + NV
2016-01-13 16:10:40 +00:00
#else
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
#define IS_GENERIC
Prepare for allow other OpenCL platforms thans AMD and NVidia
2016-01-04 12:17:20 +00:00
#endif
OpenCL Kernels: Add general function declaration keyword (static inline)
2018-02-06 18:12:24 +00:00
/**
* function declarations can have a large influence depending on the opencl runtime
*/
Some syntax error
2018-02-08 08:49:59 +00:00
#if defined IS_CPU
Re-enable CPU optimizations and some CPU case in thread management
2018-02-16 17:56:21 +00:00
#define DECLSPEC inline
Some syntax error
2018-02-08 08:49:59 +00:00
#elif defined IS_GPU
#if defined IS_AMD
DECLSPEC for CPU
2018-02-08 08:42:59 +00:00
#define DECLSPEC inline
#else
NV JiT doesn't like static inline keywords
2018-02-06 21:05:15 +00:00
#define DECLSPEC
DECLSPEC for CPU
2018-02-08 08:42:59 +00:00
#endif
NV JiT doesn't like static inline keywords
2018-02-06 21:05:15 +00:00
#else
DECLSPEC for CPU
2018-02-08 08:42:59 +00:00
#define DECLSPEC
NV JiT doesn't like static inline keywords
2018-02-06 21:05:15 +00:00
#endif
OpenCL Kernels: Add general function declaration keyword (static inline)
2018-02-06 18:12:24 +00:00
Testing: Workaround some AMD OpenCL runtime segmentation faults
2018-08-09 11:03:22 +00:00
#if (defined IS_AMD && AMD_GCN < 3)
#define MAYBE_VOLATILE volatile
#else
#define MAYBE_VOLATILE
#endif
Initial commit
2015-12-04 14:47:52 +00:00
/**
* AMD specific
*/
#ifdef IS_AMD
Rename some hashcat specific OpenCL functions to avoid conflicts with existing OpenCL functions from OpenCL runtime
2018-07-22 10:20:20 +00:00
#if defined(cl_amd_media_ops)
Prepare for allow other OpenCL platforms thans AMD and NVidia
2016-01-04 12:17:20 +00:00
#pragma OPENCL EXTENSION cl_amd_media_ops : enable
Rename some hashcat specific OpenCL functions to avoid conflicts with existing OpenCL functions from OpenCL runtime
2018-07-22 10:20:20 +00:00
#endif
#if defined(cl_amd_media_ops2)
Initial commit
2015-12-04 14:47:52 +00:00
#pragma OPENCL EXTENSION cl_amd_media_ops2 : enable
#endif
Rename some hashcat specific OpenCL functions to avoid conflicts with existing OpenCL functions from OpenCL runtime
2018-07-22 10:20:20 +00:00
#endif
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
/**
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
* Unrolling is generally enabled, for all device types and hash modes
* There's a few exception when it's better not to unroll
Add a shared section in inc_vendor.cl
2016-11-13 19:58:28 +00:00
* Some algorithms run into too much register pressure due to loop unrolling
- 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
*/
Initial commit
2015-12-04 14:47:52 +00:00
Add a shared section in inc_vendor.cl
2016-11-13 19:58:28 +00:00
// generic vendors: those algos have shown that they produce better results on both amd and nv when not unrolled
// so we can assume they will produce better results on other vendors as well
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
Add a shared section in inc_vendor.cl
2016-11-13 19:58:28 +00:00
#if KERN_TYPE == 1420
#undef _unroll
#endif
#if KERN_TYPE == 1450
#undef _unroll
#endif
#if KERN_TYPE == 1460
#undef _unroll
#endif
#if KERN_TYPE == 1720
#undef _unroll
#endif
#if KERN_TYPE == 1750
#undef _unroll
#endif
#if KERN_TYPE == 1760
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
#undef _unroll
#endif
#if KERN_TYPE == 1800
#undef _unroll
#endif
#if KERN_TYPE == 6221
#undef _unroll
#endif
#if KERN_TYPE == 6222
#undef _unroll
#endif
#if KERN_TYPE == 6223
#undef _unroll
#endif
Do not unroll -m 6400 based algorithm on NV, latest windows driver can't compile them otherwise
2016-10-21 07:51:41 +00:00
#if KERN_TYPE == 6400
#undef _unroll
#endif
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
#if KERN_TYPE == 6500
#undef _unroll
#endif
Increase AIX {ssha512} cracking performance Remove unrolling for 7100, 7200, 12100 and 8200: High-End GPUs didn't like it
2016-05-28 10:30:39 +00:00
#if KERN_TYPE == 7100
#undef _unroll
#endif
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
#if KERN_TYPE == 7400
#undef _unroll
#endif
Tune AMD unroll settings for AMD-GPU-PRO 17.40
2017-06-29 13:50:46 +00:00
#if KERN_TYPE == 7900
#undef _unroll
#endif
Do not unroll scrypt based algorithm on NV, latest windows driver can't compile them otherwise
2016-10-20 20:53:47 +00:00
#if KERN_TYPE == 8900
#undef _unroll
#endif
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
#if KERN_TYPE == 10700
#undef _unroll
#endif
Some final fixes for d_return_buf refactorization; Initial kernels vor veracrypts SHA256 KDF (not working)
2016-05-20 22:39:22 +00:00
#if KERN_TYPE == 13721
#undef _unroll
#endif
#if KERN_TYPE == 13722
#undef _unroll
#endif
#if KERN_TYPE == 13723
#undef _unroll
#endif
Do not unroll -m 6800 based algorithm on NV, latest windows driver can't compile them otherwise
2016-10-21 08:14:15 +00:00
#if KERN_TYPE == 13751
#undef _unroll
#endif
#if KERN_TYPE == 13752
#undef _unroll
#endif
#if KERN_TYPE == 13753
#undef _unroll
#endif
Add a shared section in inc_vendor.cl
2016-11-13 19:58:28 +00:00
#if KERN_TYPE == 13800
#undef _unroll
#endif
Update inc_vector.cl for ideal performance with reference GTX1080 for NV and Vega64 for AMD
2017-10-26 11:41:47 +00:00
#if KERN_TYPE == 15700
#undef _unroll
#endif
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
Add a shared section in inc_vendor.cl
2016-11-13 19:58:28 +00:00
// nvidia specific
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
Add a shared section in inc_vendor.cl
2016-11-13 19:58:28 +00:00
#ifdef IS_NV
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
#ifdef IS_GPU
Optimize unrolling for AMD based on AMDGPU-Pro 16.40
2016-11-12 20:20:03 +00:00
#endif
#endif
Add a shared section in inc_vendor.cl
2016-11-13 19:58:28 +00:00
// amd specific
#ifdef IS_AMD
#ifdef IS_GPU
Optimized -m 8000 for ROCm
2017-07-24 11:13:35 +00:00
#if KERN_TYPE == 8000
#undef _unroll
#endif
Update inc_vector.cl for ideal performance with reference GTX1080 for NV and Vega64 for AMD
2017-10-26 11:41:47 +00:00
#if KERN_TYPE == 8200
#undef _unroll
#endif
#if KERN_TYPE == 12300
#undef _unroll
#endif
Optimized 3DES for rocm
2017-08-17 12:03:55 +00:00
#if KERN_TYPE == 14100
#undef _unroll
#endif
Workaround rocm compiler error for -m 15300
2017-08-17 09:25:34 +00:00
#if KERN_TYPE == 15300
#undef _unroll
#endif
Splitted DPAPI kernel in 2 to increase performances
2017-09-20 17:19:23 +00:00
#if KERN_TYPE == 15900
#undef _unroll
#endif
Prepare for a more dynamic #pragma unroll use
2016-05-09 19:32:12 +00:00
#endif
Fix LLVM compilation for OpenCL kernels Fix LLVM compilation for OpenCL kernels Whoops
2016-04-19 21:09:27 +00:00
#endif
Experiment to fix https://github.com/hashcat/hashcat/issues/412
2016-11-14 09:50:47 +00:00
// apple specific
#ifdef IS_APPLE
#if KERN_TYPE == 5000
#undef _unroll
#endif
Fix CL_BUILD_PROGRAM_FAILURE on OSX (#412)
2016-11-14 18:28:07 +00:00
#endif
Reference in New Issue Copy Permalink