arno/hashcat
1
0
Fork 0
mirror of https://github.com/hashcat/hashcat.git synced 2024-11-13 19:28:56 +00:00
Code Issues Releases Wiki Activity
1d3795a3ab
hashcat/OpenCL/m01420_a3.cl

795 lines
35 KiB
Common Lisp
Raw Normal View History

Initial commit
2015-12-04 14:47:52 +00:00
/**
* Author......: Jens Steube <jens.steube@gmail.com>
* License.....: MIT
*/
#define _SHA256_
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
#define NEW_SIMD_CODE
Initial commit
2015-12-04 14:47:52 +00:00
#include "include/constants.h"
#include "include/kernel_vendor.h"
#define DGST_R0 3
#define DGST_R1 7
#define DGST_R2 2
#define DGST_R3 6
#include "include/kernel_functions.c"
Fix path to includes
2016-01-03 00:48:05 +00:00
#include "OpenCL/types_ocl.c"
#include "OpenCL/common.c"
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
#include "OpenCL/simd.c"
Initial commit
2015-12-04 14:47:52 +00:00
Replace the substring GPU to a more appropriate "device" or "kernel" substring depending on the context
2016-01-05 07:26:44 +00:00
static void m01420m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 bfs_cnt, const u32 digests_cnt, const u32 digests_offset)
Initial commit
2015-12-04 14:47:52 +00:00
{
/**
* modifier
*/
const u32 gid = get_global_id (0);
const u32 lid = get_local_id (0);
/**
* salt
*/
u32 salt_buf0[4];
salt_buf0[0] = salt_bufs[salt_pos].salt_buf[ 0];
salt_buf0[1] = salt_bufs[salt_pos].salt_buf[ 1];
salt_buf0[2] = salt_bufs[salt_pos].salt_buf[ 2];
salt_buf0[3] = salt_bufs[salt_pos].salt_buf[ 3];
u32 salt_buf1[4];
salt_buf1[0] = salt_bufs[salt_pos].salt_buf[ 4];
salt_buf1[1] = salt_bufs[salt_pos].salt_buf[ 5];
salt_buf1[2] = salt_bufs[salt_pos].salt_buf[ 6];
salt_buf1[3] = salt_bufs[salt_pos].salt_buf[ 7];
u32 salt_buf2[4];
salt_buf2[0] = 0;
salt_buf2[1] = 0;
salt_buf2[2] = 0;
salt_buf2[3] = 0;
u32 salt_buf3[4];
salt_buf3[0] = 0;
salt_buf3[1] = 0;
salt_buf3[2] = 0;
salt_buf3[3] = 0;
const u32 salt_len = salt_bufs[salt_pos].salt_len;
const u32 pw_salt_len = pw_len + salt_len;
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
/**
* prepend salt
*/
u32 w0_t[4];
u32 w1_t[4];
u32 w2_t[4];
u32 w3_t[4];
w0_t[0] = swap32_S (w0[0]);
w0_t[1] = swap32_S (w0[1]);
w0_t[2] = swap32_S (w0[2]);
w0_t[3] = swap32_S (w0[3]);
w1_t[0] = swap32_S (w1[0]);
w1_t[1] = swap32_S (w1[1]);
w1_t[2] = swap32_S (w1[2]);
w1_t[3] = swap32_S (w1[3]);
w2_t[0] = swap32_S (w2[0]);
w2_t[1] = swap32_S (w2[1]);
w2_t[2] = swap32_S (w2[2]);
w2_t[3] = swap32_S (w2[3]);
w3_t[0] = swap32_S (w3[0]);
w3_t[1] = swap32_S (w3[1]);
w3_t[2] = swap32_S (w3[2]);
w3_t[3] = swap32_S (w3[3]);
switch_buffer_by_offset_le_S (w0_t, w1_t, w2_t, w3_t, salt_len);
w0_t[0] |= salt_buf0[0];
w0_t[1] |= salt_buf0[1];
w0_t[2] |= salt_buf0[2];
w0_t[3] |= salt_buf0[3];
w1_t[0] |= salt_buf1[0];
w1_t[1] |= salt_buf1[1];
w1_t[2] |= salt_buf1[2];
w1_t[3] |= salt_buf1[3];
w2_t[0] |= salt_buf2[0];
w2_t[1] |= salt_buf2[1];
w2_t[2] |= salt_buf2[2];
w2_t[3] |= salt_buf2[3];
w3_t[0] |= salt_buf3[0];
w3_t[1] |= salt_buf3[1];
w3_t[2] |= salt_buf3[2];
w3_t[3] |= salt_buf3[3];
w0_t[0] = swap32_S (w0_t[0]);
w0_t[1] = swap32_S (w0_t[1]);
w0_t[2] = swap32_S (w0_t[2]);
w0_t[3] = swap32_S (w0_t[3]);
w1_t[0] = swap32_S (w1_t[0]);
w1_t[1] = swap32_S (w1_t[1]);
w1_t[2] = swap32_S (w1_t[2]);
w1_t[3] = swap32_S (w1_t[3]);
w2_t[0] = swap32_S (w2_t[0]);
w2_t[1] = swap32_S (w2_t[1]);
w2_t[2] = swap32_S (w2_t[2]);
w2_t[3] = swap32_S (w2_t[3]);
w3_t[0] = swap32_S (w3_t[0]);
w3_t[1] = swap32_S (w3_t[1]);
w3_t[2] = swap32_S (w3_t[2]);
w3_t[3] = swap32_S (w3_t[3]);
Initial commit
2015-12-04 14:47:52 +00:00
/**
* loop
*/
- 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 w0l = w0[0];
Initial commit
2015-12-04 14:47:52 +00:00
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
for (u32 il_pos = 0; il_pos < bfs_cnt; il_pos += VECT_SIZE)
Initial commit
2015-12-04 14:47:52 +00:00
{
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
const u32x w0r = w0r_create_bft (bfs_buf, il_pos);
const u32x w0lr = w0l | w0r;
u32x wx[16];
wx[ 0] = w0_t[0];
wx[ 1] = w0_t[1];
wx[ 2] = w0_t[2];
wx[ 3] = w0_t[3];
wx[ 4] = w1_t[0];
wx[ 5] = w1_t[1];
wx[ 6] = w1_t[2];
wx[ 7] = w1_t[3];
wx[ 8] = w2_t[0];
wx[ 9] = w2_t[1];
wx[10] = w2_t[2];
wx[11] = w2_t[3];
wx[12] = w3_t[0];
wx[13] = w3_t[1];
wx[14] = w3_t[2];
wx[15] = w3_t[3];
overwrite_at_be (wx, w0lr, salt_len);
u32x w0_t = wx[ 0];
u32x w1_t = wx[ 1];
u32x w2_t = wx[ 2];
u32x w3_t = wx[ 3];
u32x w4_t = wx[ 4];
u32x w5_t = wx[ 5];
u32x w6_t = wx[ 6];
u32x w7_t = wx[ 7];
u32x w8_t = wx[ 8];
u32x w9_t = wx[ 9];
u32x wa_t = wx[10];
u32x wb_t = wx[11];
u32x wc_t = wx[12];
u32x wd_t = wx[13];
u32x we_t = 0;
u32x wf_t = pw_salt_len * 8;
u32x a = SHA256M_A;
u32x b = SHA256M_B;
u32x c = SHA256M_C;
u32x d = SHA256M_D;
u32x e = SHA256M_E;
u32x f = SHA256M_F;
u32x g = SHA256M_G;
u32x h = SHA256M_H;
Initial commit
2015-12-04 14:47:52 +00:00
SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w0_t, SHA256C00);
SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w1_t, SHA256C01);
SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, w2_t, SHA256C02);
SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, w3_t, SHA256C03);
SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, w4_t, SHA256C04);
SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, w5_t, SHA256C05);
SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, w6_t, SHA256C06);
SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, w7_t, SHA256C07);
SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w8_t, SHA256C08);
SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w9_t, SHA256C09);
SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, wa_t, SHA256C0a);
SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, wb_t, SHA256C0b);
SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, wc_t, SHA256C0c);
SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, wd_t, SHA256C0d);
SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, we_t, SHA256C0e);
SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, wf_t, SHA256C0f);
w0_t = SHA256_EXPAND (we_t, w9_t, w1_t, w0_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w0_t, SHA256C10);
w1_t = SHA256_EXPAND (wf_t, wa_t, w2_t, w1_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w1_t, SHA256C11);
w2_t = SHA256_EXPAND (w0_t, wb_t, w3_t, w2_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, w2_t, SHA256C12);
w3_t = SHA256_EXPAND (w1_t, wc_t, w4_t, w3_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, w3_t, SHA256C13);
w4_t = SHA256_EXPAND (w2_t, wd_t, w5_t, w4_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, w4_t, SHA256C14);
w5_t = SHA256_EXPAND (w3_t, we_t, w6_t, w5_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, w5_t, SHA256C15);
w6_t = SHA256_EXPAND (w4_t, wf_t, w7_t, w6_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, w6_t, SHA256C16);
w7_t = SHA256_EXPAND (w5_t, w0_t, w8_t, w7_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, w7_t, SHA256C17);
w8_t = SHA256_EXPAND (w6_t, w1_t, w9_t, w8_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w8_t, SHA256C18);
w9_t = SHA256_EXPAND (w7_t, w2_t, wa_t, w9_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w9_t, SHA256C19);
wa_t = SHA256_EXPAND (w8_t, w3_t, wb_t, wa_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, wa_t, SHA256C1a);
wb_t = SHA256_EXPAND (w9_t, w4_t, wc_t, wb_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, wb_t, SHA256C1b);
wc_t = SHA256_EXPAND (wa_t, w5_t, wd_t, wc_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, wc_t, SHA256C1c);
wd_t = SHA256_EXPAND (wb_t, w6_t, we_t, wd_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, wd_t, SHA256C1d);
we_t = SHA256_EXPAND (wc_t, w7_t, wf_t, we_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, we_t, SHA256C1e);
wf_t = SHA256_EXPAND (wd_t, w8_t, w0_t, wf_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, wf_t, SHA256C1f);
w0_t = SHA256_EXPAND (we_t, w9_t, w1_t, w0_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w0_t, SHA256C20);
w1_t = SHA256_EXPAND (wf_t, wa_t, w2_t, w1_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w1_t, SHA256C21);
w2_t = SHA256_EXPAND (w0_t, wb_t, w3_t, w2_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, w2_t, SHA256C22);
w3_t = SHA256_EXPAND (w1_t, wc_t, w4_t, w3_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, w3_t, SHA256C23);
w4_t = SHA256_EXPAND (w2_t, wd_t, w5_t, w4_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, w4_t, SHA256C24);
w5_t = SHA256_EXPAND (w3_t, we_t, w6_t, w5_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, w5_t, SHA256C25);
w6_t = SHA256_EXPAND (w4_t, wf_t, w7_t, w6_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, w6_t, SHA256C26);
w7_t = SHA256_EXPAND (w5_t, w0_t, w8_t, w7_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, w7_t, SHA256C27);
w8_t = SHA256_EXPAND (w6_t, w1_t, w9_t, w8_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w8_t, SHA256C28);
w9_t = SHA256_EXPAND (w7_t, w2_t, wa_t, w9_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w9_t, SHA256C29);
wa_t = SHA256_EXPAND (w8_t, w3_t, wb_t, wa_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, wa_t, SHA256C2a);
wb_t = SHA256_EXPAND (w9_t, w4_t, wc_t, wb_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, wb_t, SHA256C2b);
wc_t = SHA256_EXPAND (wa_t, w5_t, wd_t, wc_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, wc_t, SHA256C2c);
wd_t = SHA256_EXPAND (wb_t, w6_t, we_t, wd_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, wd_t, SHA256C2d);
we_t = SHA256_EXPAND (wc_t, w7_t, wf_t, we_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, we_t, SHA256C2e);
wf_t = SHA256_EXPAND (wd_t, w8_t, w0_t, wf_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, wf_t, SHA256C2f);
w0_t = SHA256_EXPAND (we_t, w9_t, w1_t, w0_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w0_t, SHA256C30);
w1_t = SHA256_EXPAND (wf_t, wa_t, w2_t, w1_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w1_t, SHA256C31);
w2_t = SHA256_EXPAND (w0_t, wb_t, w3_t, w2_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, w2_t, SHA256C32);
w3_t = SHA256_EXPAND (w1_t, wc_t, w4_t, w3_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, w3_t, SHA256C33);
w4_t = SHA256_EXPAND (w2_t, wd_t, w5_t, w4_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, w4_t, SHA256C34);
w5_t = SHA256_EXPAND (w3_t, we_t, w6_t, w5_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, w5_t, SHA256C35);
w6_t = SHA256_EXPAND (w4_t, wf_t, w7_t, w6_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, w6_t, SHA256C36);
w7_t = SHA256_EXPAND (w5_t, w0_t, w8_t, w7_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, w7_t, SHA256C37);
w8_t = SHA256_EXPAND (w6_t, w1_t, w9_t, w8_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w8_t, SHA256C38);
w9_t = SHA256_EXPAND (w7_t, w2_t, wa_t, w9_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w9_t, SHA256C39);
wa_t = SHA256_EXPAND (w8_t, w3_t, wb_t, wa_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, wa_t, SHA256C3a);
wb_t = SHA256_EXPAND (w9_t, w4_t, wc_t, wb_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, wb_t, SHA256C3b);
wc_t = SHA256_EXPAND (wa_t, w5_t, wd_t, wc_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, wc_t, SHA256C3c);
wd_t = SHA256_EXPAND (wb_t, w6_t, we_t, wd_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, wd_t, SHA256C3d);
we_t = SHA256_EXPAND (wc_t, w7_t, wf_t, we_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, we_t, SHA256C3e);
wf_t = SHA256_EXPAND (wd_t, w8_t, w0_t, wf_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, wf_t, SHA256C3f);
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
COMPARE_M_SIMD (d, h, c, g);
Initial commit
2015-12-04 14:47:52 +00:00
}
}
Replace the substring GPU to a more appropriate "device" or "kernel" substring depending on the context
2016-01-05 07:26:44 +00:00
static void m01420s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 bfs_cnt, const u32 digests_cnt, const u32 digests_offset)
Initial commit
2015-12-04 14:47:52 +00:00
{
/**
* modifier
*/
const u32 gid = get_global_id (0);
const u32 lid = get_local_id (0);
/**
* digest
*/
const u32 search[4] =
{
digests_buf[digests_offset].digest_buf[DGST_R0],
digests_buf[digests_offset].digest_buf[DGST_R1],
digests_buf[digests_offset].digest_buf[DGST_R2],
digests_buf[digests_offset].digest_buf[DGST_R3]
};
/**
* salt
*/
u32 salt_buf0[4];
salt_buf0[0] = salt_bufs[salt_pos].salt_buf[ 0];
salt_buf0[1] = salt_bufs[salt_pos].salt_buf[ 1];
salt_buf0[2] = salt_bufs[salt_pos].salt_buf[ 2];
salt_buf0[3] = salt_bufs[salt_pos].salt_buf[ 3];
u32 salt_buf1[4];
salt_buf1[0] = salt_bufs[salt_pos].salt_buf[ 4];
salt_buf1[1] = salt_bufs[salt_pos].salt_buf[ 5];
salt_buf1[2] = salt_bufs[salt_pos].salt_buf[ 6];
salt_buf1[3] = salt_bufs[salt_pos].salt_buf[ 7];
u32 salt_buf2[4];
salt_buf2[0] = 0;
salt_buf2[1] = 0;
salt_buf2[2] = 0;
salt_buf2[3] = 0;
u32 salt_buf3[4];
salt_buf3[0] = 0;
salt_buf3[1] = 0;
salt_buf3[2] = 0;
salt_buf3[3] = 0;
const u32 salt_len = salt_bufs[salt_pos].salt_len;
const u32 pw_salt_len = pw_len + salt_len;
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
/**
* prepend salt
*/
u32 w0_t[4];
u32 w1_t[4];
u32 w2_t[4];
u32 w3_t[4];
w0_t[0] = swap32_S (w0[0]);
w0_t[1] = swap32_S (w0[1]);
w0_t[2] = swap32_S (w0[2]);
w0_t[3] = swap32_S (w0[3]);
w1_t[0] = swap32_S (w1[0]);
w1_t[1] = swap32_S (w1[1]);
w1_t[2] = swap32_S (w1[2]);
w1_t[3] = swap32_S (w1[3]);
w2_t[0] = swap32_S (w2[0]);
w2_t[1] = swap32_S (w2[1]);
w2_t[2] = swap32_S (w2[2]);
w2_t[3] = swap32_S (w2[3]);
w3_t[0] = swap32_S (w3[0]);
w3_t[1] = swap32_S (w3[1]);
w3_t[2] = swap32_S (w3[2]);
w3_t[3] = swap32_S (w3[3]);
switch_buffer_by_offset_le_S (w0_t, w1_t, w2_t, w3_t, salt_len);
w0_t[0] |= salt_buf0[0];
w0_t[1] |= salt_buf0[1];
w0_t[2] |= salt_buf0[2];
w0_t[3] |= salt_buf0[3];
w1_t[0] |= salt_buf1[0];
w1_t[1] |= salt_buf1[1];
w1_t[2] |= salt_buf1[2];
w1_t[3] |= salt_buf1[3];
w2_t[0] |= salt_buf2[0];
w2_t[1] |= salt_buf2[1];
w2_t[2] |= salt_buf2[2];
w2_t[3] |= salt_buf2[3];
w3_t[0] |= salt_buf3[0];
w3_t[1] |= salt_buf3[1];
w3_t[2] |= salt_buf3[2];
w3_t[3] |= salt_buf3[3];
w0_t[0] = swap32_S (w0_t[0]);
w0_t[1] = swap32_S (w0_t[1]);
w0_t[2] = swap32_S (w0_t[2]);
w0_t[3] = swap32_S (w0_t[3]);
w1_t[0] = swap32_S (w1_t[0]);
w1_t[1] = swap32_S (w1_t[1]);
w1_t[2] = swap32_S (w1_t[2]);
w1_t[3] = swap32_S (w1_t[3]);
w2_t[0] = swap32_S (w2_t[0]);
w2_t[1] = swap32_S (w2_t[1]);
w2_t[2] = swap32_S (w2_t[2]);
w2_t[3] = swap32_S (w2_t[3]);
w3_t[0] = swap32_S (w3_t[0]);
w3_t[1] = swap32_S (w3_t[1]);
w3_t[2] = swap32_S (w3_t[2]);
w3_t[3] = swap32_S (w3_t[3]);
Initial commit
2015-12-04 14:47:52 +00:00
/**
* loop
*/
- 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 w0l = w0[0];
Initial commit
2015-12-04 14:47:52 +00:00
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
for (u32 il_pos = 0; il_pos < bfs_cnt; il_pos += VECT_SIZE)
Initial commit
2015-12-04 14:47:52 +00:00
{
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
const u32x w0r = w0r_create_bft (bfs_buf, il_pos);
const u32x w0lr = w0l | w0r;
u32x wx[16];
wx[ 0] = w0_t[0];
wx[ 1] = w0_t[1];
wx[ 2] = w0_t[2];
wx[ 3] = w0_t[3];
wx[ 4] = w1_t[0];
wx[ 5] = w1_t[1];
wx[ 6] = w1_t[2];
wx[ 7] = w1_t[3];
wx[ 8] = w2_t[0];
wx[ 9] = w2_t[1];
wx[10] = w2_t[2];
wx[11] = w2_t[3];
wx[12] = w3_t[0];
wx[13] = w3_t[1];
wx[14] = w3_t[2];
wx[15] = w3_t[3];
overwrite_at_be (wx, w0lr, salt_len);
u32x w0_t = wx[ 0];
u32x w1_t = wx[ 1];
u32x w2_t = wx[ 2];
u32x w3_t = wx[ 3];
u32x w4_t = wx[ 4];
u32x w5_t = wx[ 5];
u32x w6_t = wx[ 6];
u32x w7_t = wx[ 7];
u32x w8_t = wx[ 8];
u32x w9_t = wx[ 9];
u32x wa_t = wx[10];
u32x wb_t = wx[11];
u32x wc_t = wx[12];
u32x wd_t = wx[13];
u32x we_t = 0;
u32x wf_t = pw_salt_len * 8;
u32x a = SHA256M_A;
u32x b = SHA256M_B;
u32x c = SHA256M_C;
u32x d = SHA256M_D;
u32x e = SHA256M_E;
u32x f = SHA256M_F;
u32x g = SHA256M_G;
u32x h = SHA256M_H;
Initial commit
2015-12-04 14:47:52 +00:00
SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w0_t, SHA256C00);
SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w1_t, SHA256C01);
SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, w2_t, SHA256C02);
SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, w3_t, SHA256C03);
SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, w4_t, SHA256C04);
SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, w5_t, SHA256C05);
SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, w6_t, SHA256C06);
SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, w7_t, SHA256C07);
SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w8_t, SHA256C08);
SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w9_t, SHA256C09);
SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, wa_t, SHA256C0a);
SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, wb_t, SHA256C0b);
SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, wc_t, SHA256C0c);
SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, wd_t, SHA256C0d);
SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, we_t, SHA256C0e);
SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, wf_t, SHA256C0f);
w0_t = SHA256_EXPAND (we_t, w9_t, w1_t, w0_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w0_t, SHA256C10);
w1_t = SHA256_EXPAND (wf_t, wa_t, w2_t, w1_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w1_t, SHA256C11);
w2_t = SHA256_EXPAND (w0_t, wb_t, w3_t, w2_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, w2_t, SHA256C12);
w3_t = SHA256_EXPAND (w1_t, wc_t, w4_t, w3_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, w3_t, SHA256C13);
w4_t = SHA256_EXPAND (w2_t, wd_t, w5_t, w4_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, w4_t, SHA256C14);
w5_t = SHA256_EXPAND (w3_t, we_t, w6_t, w5_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, w5_t, SHA256C15);
w6_t = SHA256_EXPAND (w4_t, wf_t, w7_t, w6_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, w6_t, SHA256C16);
w7_t = SHA256_EXPAND (w5_t, w0_t, w8_t, w7_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, w7_t, SHA256C17);
w8_t = SHA256_EXPAND (w6_t, w1_t, w9_t, w8_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w8_t, SHA256C18);
w9_t = SHA256_EXPAND (w7_t, w2_t, wa_t, w9_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w9_t, SHA256C19);
wa_t = SHA256_EXPAND (w8_t, w3_t, wb_t, wa_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, wa_t, SHA256C1a);
wb_t = SHA256_EXPAND (w9_t, w4_t, wc_t, wb_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, wb_t, SHA256C1b);
wc_t = SHA256_EXPAND (wa_t, w5_t, wd_t, wc_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, wc_t, SHA256C1c);
wd_t = SHA256_EXPAND (wb_t, w6_t, we_t, wd_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, wd_t, SHA256C1d);
we_t = SHA256_EXPAND (wc_t, w7_t, wf_t, we_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, we_t, SHA256C1e);
wf_t = SHA256_EXPAND (wd_t, w8_t, w0_t, wf_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, wf_t, SHA256C1f);
w0_t = SHA256_EXPAND (we_t, w9_t, w1_t, w0_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w0_t, SHA256C20);
w1_t = SHA256_EXPAND (wf_t, wa_t, w2_t, w1_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w1_t, SHA256C21);
w2_t = SHA256_EXPAND (w0_t, wb_t, w3_t, w2_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, w2_t, SHA256C22);
w3_t = SHA256_EXPAND (w1_t, wc_t, w4_t, w3_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, w3_t, SHA256C23);
w4_t = SHA256_EXPAND (w2_t, wd_t, w5_t, w4_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, w4_t, SHA256C24);
w5_t = SHA256_EXPAND (w3_t, we_t, w6_t, w5_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, w5_t, SHA256C25);
w6_t = SHA256_EXPAND (w4_t, wf_t, w7_t, w6_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, w6_t, SHA256C26);
w7_t = SHA256_EXPAND (w5_t, w0_t, w8_t, w7_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, w7_t, SHA256C27);
w8_t = SHA256_EXPAND (w6_t, w1_t, w9_t, w8_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w8_t, SHA256C28);
w9_t = SHA256_EXPAND (w7_t, w2_t, wa_t, w9_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w9_t, SHA256C29);
wa_t = SHA256_EXPAND (w8_t, w3_t, wb_t, wa_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, wa_t, SHA256C2a);
wb_t = SHA256_EXPAND (w9_t, w4_t, wc_t, wb_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, wb_t, SHA256C2b);
wc_t = SHA256_EXPAND (wa_t, w5_t, wd_t, wc_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, wc_t, SHA256C2c);
wd_t = SHA256_EXPAND (wb_t, w6_t, we_t, wd_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, wd_t, SHA256C2d);
we_t = SHA256_EXPAND (wc_t, w7_t, wf_t, we_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, we_t, SHA256C2e);
wf_t = SHA256_EXPAND (wd_t, w8_t, w0_t, wf_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, wf_t, SHA256C2f);
w0_t = SHA256_EXPAND (we_t, w9_t, w1_t, w0_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w0_t, SHA256C30);
w1_t = SHA256_EXPAND (wf_t, wa_t, w2_t, w1_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w1_t, SHA256C31);
w2_t = SHA256_EXPAND (w0_t, wb_t, w3_t, w2_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, w2_t, SHA256C32);
w3_t = SHA256_EXPAND (w1_t, wc_t, w4_t, w3_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, w3_t, SHA256C33);
w4_t = SHA256_EXPAND (w2_t, wd_t, w5_t, w4_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, w4_t, SHA256C34);
w5_t = SHA256_EXPAND (w3_t, we_t, w6_t, w5_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, w5_t, SHA256C35);
w6_t = SHA256_EXPAND (w4_t, wf_t, w7_t, w6_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, w6_t, SHA256C36);
w7_t = SHA256_EXPAND (w5_t, w0_t, w8_t, w7_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, w7_t, SHA256C37);
w8_t = SHA256_EXPAND (w6_t, w1_t, w9_t, w8_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, a, b, c, d, e, f, g, h, w8_t, SHA256C38);
w9_t = SHA256_EXPAND (w7_t, w2_t, wa_t, w9_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, h, a, b, c, d, e, f, g, w9_t, SHA256C39);
wa_t = SHA256_EXPAND (w8_t, w3_t, wb_t, wa_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, g, h, a, b, c, d, e, f, wa_t, SHA256C3a);
wb_t = SHA256_EXPAND (w9_t, w4_t, wc_t, wb_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, f, g, h, a, b, c, d, e, wb_t, SHA256C3b);
wc_t = SHA256_EXPAND (wa_t, w5_t, wd_t, wc_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, e, f, g, h, a, b, c, d, wc_t, SHA256C3c);
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
if (MATCHES_NONE_VS (d, search[0])) continue;
Initial commit
2015-12-04 14:47:52 +00:00
wd_t = SHA256_EXPAND (wb_t, w6_t, we_t, wd_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, d, e, f, g, h, a, b, c, wd_t, SHA256C3d);
we_t = SHA256_EXPAND (wc_t, w7_t, wf_t, we_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, c, d, e, f, g, h, a, b, we_t, SHA256C3e);
wf_t = SHA256_EXPAND (wd_t, w8_t, w0_t, wf_t); SHA256_STEP (SHA256_F0o, SHA256_F1o, b, c, d, e, f, g, h, a, wf_t, SHA256C3f);
Converted _a3 kernels, use SIMD for CPU and GPU
2016-01-23 14:32:31 +00:00
COMPARE_S_SIMD (d, h, c, g);
Initial commit
2015-12-04 14:47:52 +00:00
}
}
Upgrade kernel to support dynamic local work sizes
2016-01-19 15:06:03 +00:00
__kernel void m01420_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 bfs_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
Initial commit
2015-12-04 14:47:52 +00:00
{
/**
* base
*/
const u32 gid = get_global_id (0);
if (gid >= gid_max) return;
- 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 w0[4];
Initial commit
2015-12-04 14:47:52 +00:00
w0[0] = pws[gid].i[ 0];
w0[1] = pws[gid].i[ 1];
w0[2] = pws[gid].i[ 2];
w0[3] = pws[gid].i[ 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
u32 w1[4];
Initial commit
2015-12-04 14:47:52 +00:00
w1[0] = 0;
w1[1] = 0;
w1[2] = 0;
w1[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
u32 w2[4];
Initial commit
2015-12-04 14:47:52 +00:00
w2[0] = 0;
w2[1] = 0;
w2[2] = 0;
w2[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
u32 w3[4];
Initial commit
2015-12-04 14:47:52 +00:00
w3[0] = 0;
w3[1] = 0;
w3[2] = 0;
w3[3] = 0;
const u32 pw_len = pws[gid].pw_len;
/**
* main
*/
m01420m (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, bfs_cnt, digests_cnt, digests_offset);
}
Upgrade kernel to support dynamic local work sizes
2016-01-19 15:06:03 +00:00
__kernel void m01420_m08 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 bfs_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
Initial commit
2015-12-04 14:47:52 +00:00
{
/**
* base
*/
const u32 gid = get_global_id (0);
if (gid >= gid_max) return;
- 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 w0[4];
Initial commit
2015-12-04 14:47:52 +00:00
w0[0] = pws[gid].i[ 0];
w0[1] = pws[gid].i[ 1];
w0[2] = pws[gid].i[ 2];
w0[3] = pws[gid].i[ 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
u32 w1[4];
Initial commit
2015-12-04 14:47:52 +00:00
w1[0] = pws[gid].i[ 4];
w1[1] = pws[gid].i[ 5];
w1[2] = pws[gid].i[ 6];
w1[3] = pws[gid].i[ 7];
- 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 w2[4];
Initial commit
2015-12-04 14:47:52 +00:00
w2[0] = 0;
w2[1] = 0;
w2[2] = 0;
w2[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
u32 w3[4];
Initial commit
2015-12-04 14:47:52 +00:00
w3[0] = 0;
w3[1] = 0;
w3[2] = 0;
w3[3] = 0;
const u32 pw_len = pws[gid].pw_len;
/**
* main
*/
m01420m (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, bfs_cnt, digests_cnt, digests_offset);
}
Upgrade kernel to support dynamic local work sizes
2016-01-19 15:06:03 +00:00
__kernel void m01420_m16 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 bfs_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
Initial commit
2015-12-04 14:47:52 +00:00
{
/**
* base
*/
const u32 gid = get_global_id (0);
if (gid >= gid_max) return;
- 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 w0[4];
Initial commit
2015-12-04 14:47:52 +00:00
w0[0] = pws[gid].i[ 0];
w0[1] = pws[gid].i[ 1];
w0[2] = pws[gid].i[ 2];
w0[3] = pws[gid].i[ 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
u32 w1[4];
Initial commit
2015-12-04 14:47:52 +00:00
w1[0] = pws[gid].i[ 4];
w1[1] = pws[gid].i[ 5];
w1[2] = pws[gid].i[ 6];
w1[3] = pws[gid].i[ 7];
- 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 w2[4];
Initial commit
2015-12-04 14:47:52 +00:00
w2[0] = pws[gid].i[ 8];
w2[1] = pws[gid].i[ 9];
w2[2] = pws[gid].i[10];
w2[3] = pws[gid].i[11];
- 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 w3[4];
Initial commit
2015-12-04 14:47:52 +00:00
w3[0] = pws[gid].i[12];
w3[1] = pws[gid].i[13];
w3[2] = 0;
w3[3] = 0;
const u32 pw_len = pws[gid].pw_len;
/**
* main
*/
m01420m (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, bfs_cnt, digests_cnt, digests_offset);
}
Upgrade kernel to support dynamic local work sizes
2016-01-19 15:06:03 +00:00
__kernel void m01420_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 bfs_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
Initial commit
2015-12-04 14:47:52 +00:00
{
/**
* base
*/
const u32 gid = get_global_id (0);
if (gid >= gid_max) return;
- 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 w0[4];
Initial commit
2015-12-04 14:47:52 +00:00
w0[0] = pws[gid].i[ 0];
w0[1] = pws[gid].i[ 1];
w0[2] = pws[gid].i[ 2];
w0[3] = pws[gid].i[ 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
u32 w1[4];
Initial commit
2015-12-04 14:47:52 +00:00
w1[0] = 0;
w1[1] = 0;
w1[2] = 0;
w1[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
u32 w2[4];
Initial commit
2015-12-04 14:47:52 +00:00
w2[0] = 0;
w2[1] = 0;
w2[2] = 0;
w2[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
u32 w3[4];
Initial commit
2015-12-04 14:47:52 +00:00
w3[0] = 0;
w3[1] = 0;
w3[2] = 0;
w3[3] = 0;
const u32 pw_len = pws[gid].pw_len;
/**
* main
*/
m01420s (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, bfs_cnt, digests_cnt, digests_offset);
}
Upgrade kernel to support dynamic local work sizes
2016-01-19 15:06:03 +00:00
__kernel void m01420_s08 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 bfs_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
Initial commit
2015-12-04 14:47:52 +00:00
{
/**
* base
*/
const u32 gid = get_global_id (0);
if (gid >= gid_max) return;
- 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 w0[4];
Initial commit
2015-12-04 14:47:52 +00:00
w0[0] = pws[gid].i[ 0];
w0[1] = pws[gid].i[ 1];
w0[2] = pws[gid].i[ 2];
w0[3] = pws[gid].i[ 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
u32 w1[4];
Initial commit
2015-12-04 14:47:52 +00:00
w1[0] = pws[gid].i[ 4];
w1[1] = pws[gid].i[ 5];
w1[2] = pws[gid].i[ 6];
w1[3] = pws[gid].i[ 7];
- 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 w2[4];
Initial commit
2015-12-04 14:47:52 +00:00
w2[0] = 0;
w2[1] = 0;
w2[2] = 0;
w2[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
u32 w3[4];
Initial commit
2015-12-04 14:47:52 +00:00
w3[0] = 0;
w3[1] = 0;
w3[2] = 0;
w3[3] = 0;
const u32 pw_len = pws[gid].pw_len;
/**
* main
*/
m01420s (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, bfs_cnt, digests_cnt, digests_offset);
}
Upgrade kernel to support dynamic local work sizes
2016-01-19 15:06:03 +00:00
__kernel void m01420_s16 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 bfs_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
Initial commit
2015-12-04 14:47:52 +00:00
{
/**
* base
*/
const u32 gid = get_global_id (0);
if (gid >= gid_max) return;
- 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 w0[4];
Initial commit
2015-12-04 14:47:52 +00:00
w0[0] = pws[gid].i[ 0];
w0[1] = pws[gid].i[ 1];
w0[2] = pws[gid].i[ 2];
w0[3] = pws[gid].i[ 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
u32 w1[4];
Initial commit
2015-12-04 14:47:52 +00:00
w1[0] = pws[gid].i[ 4];
w1[1] = pws[gid].i[ 5];
w1[2] = pws[gid].i[ 6];
w1[3] = pws[gid].i[ 7];
- 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 w2[4];
Initial commit
2015-12-04 14:47:52 +00:00
w2[0] = pws[gid].i[ 8];
w2[1] = pws[gid].i[ 9];
w2[2] = pws[gid].i[10];
w2[3] = pws[gid].i[11];
- 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 w3[4];
Initial commit
2015-12-04 14:47:52 +00:00
w3[0] = pws[gid].i[12];
w3[1] = pws[gid].i[13];
w3[2] = 0;
w3[3] = 0;
const u32 pw_len = pws[gid].pw_len;
/**
* main
*/
m01420s (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, bfs_cnt, digests_cnt, digests_offset);
}
Reference in New Issue Copy Permalink