From b477e5c4815b59c7c2549199e1b457b7503e2e81 Mon Sep 17 00:00:00 2001 From: Jens Steube Date: Sat, 6 Feb 2016 10:11:07 +0100 Subject: [PATCH] Converted to new SIMD: -m 4800 -a 0 --- OpenCL/m04800_a0.cl | 126 ++++++++++++-------------------------------- 1 file changed, 34 insertions(+), 92 deletions(-) diff --git a/OpenCL/m04800_a0.cl b/OpenCL/m04800_a0.cl index 145d1ad88..2c0adab03 100644 --- a/OpenCL/m04800_a0.cl +++ b/OpenCL/m04800_a0.cl @@ -5,6 +5,8 @@ #define _MD5_ +#define NEW_SIMD_CODE + #include "include/constants.h" #include "include/kernel_vendor.h" @@ -18,9 +20,7 @@ #include "OpenCL/common.c" #include "include/rp_kernel.h" #include "OpenCL/rp.c" - -#define COMPARE_S "OpenCL/check_single_comp4.c" -#define COMPARE_M "OpenCL/check_multi_comp4.c" +#include "OpenCL/simd.c" __kernel void m04800_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 rules_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max) { @@ -72,37 +72,14 @@ __kernel void m04800_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, * loop */ - for (u32 il_pos = 0; il_pos < rules_cnt; il_pos++) + for (u32 il_pos = 0; il_pos < rules_cnt; il_pos += VECT_SIZE) { - u32 w0[4]; + u32x w0[4] = { 0 }; + u32x w1[4] = { 0 }; + u32x w2[4] = { 0 }; + u32x w3[4] = { 0 }; - w0[0] = pw_buf0[0]; - w0[1] = pw_buf0[1]; - w0[2] = pw_buf0[2]; - w0[3] = pw_buf0[3]; - - u32 w1[4]; - - w1[0] = pw_buf1[0]; - w1[1] = pw_buf1[1]; - w1[2] = pw_buf1[2]; - w1[3] = pw_buf1[3]; - - u32 w2[4]; - - w2[0] = 0; - w2[1] = 0; - w2[2] = 0; - w2[3] = 0; - - u32 w3[4]; - - w3[0] = 0; - w3[1] = 0; - w3[2] = 0; - w3[3] = 0; - - const u32 out_len = apply_rules (rules_buf[il_pos].cmds, w0, w1, pw_len); + const u32 out_len = apply_rules_vect (pw_buf0, pw_buf1, pw_len, rules_buf, il_pos, w0, w1); u32 pw_salt_len = out_len + salt_len; @@ -110,28 +87,28 @@ __kernel void m04800_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, * append salt */ - u32 s0[4]; + u32x s0[4]; s0[0] = salt_buf[0]; s0[1] = salt_buf[1]; s0[2] = salt_buf[2]; s0[3] = salt_buf[3]; - u32 s1[4]; + u32x s1[4]; s1[0] = 0x80; s1[1] = 0; s1[2] = 0; s1[3] = 0; - u32 s2[4]; + u32x s2[4]; s2[0] = 0; s2[1] = 0; s2[2] = 0; s2[3] = 0; - u32 s3[4]; + u32x s3[4]; s3[0] = 0; s3[1] = 0; @@ -174,10 +151,10 @@ __kernel void m04800_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, * md5 */ - u32 a = MD5M_A; - u32 b = MD5M_B; - u32 c = MD5M_C; - u32 d = MD5M_D; + u32x a = MD5M_A; + u32x b = MD5M_B; + u32x c = MD5M_C; + u32x d = MD5M_D; MD5_STEP (MD5_Fo, a, b, c, d, w0[0], MD5C00, MD5S00); MD5_STEP (MD5_Fo, d, a, b, c, w0[1], MD5C01, MD5S01); @@ -247,12 +224,7 @@ __kernel void m04800_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, MD5_STEP (MD5_I , c, d, a, b, w0[2], MD5C3e, MD5S32); MD5_STEP (MD5_I , b, c, d, a, w2[1], MD5C3f, MD5S33); - const u32 r0 = a; - const u32 r1 = d; - const u32 r2 = c; - const u32 r3 = b; - - #include COMPARE_M + COMPARE_M_SIMD (a, d, c, b); } } @@ -326,37 +298,14 @@ __kernel void m04800_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, * loop */ - for (u32 il_pos = 0; il_pos < rules_cnt; il_pos++) + for (u32 il_pos = 0; il_pos < rules_cnt; il_pos += VECT_SIZE) { - u32 w0[4]; + u32x w0[4] = { 0 }; + u32x w1[4] = { 0 }; + u32x w2[4] = { 0 }; + u32x w3[4] = { 0 }; - w0[0] = pw_buf0[0]; - w0[1] = pw_buf0[1]; - w0[2] = pw_buf0[2]; - w0[3] = pw_buf0[3]; - - u32 w1[4]; - - w1[0] = pw_buf1[0]; - w1[1] = pw_buf1[1]; - w1[2] = pw_buf1[2]; - w1[3] = pw_buf1[3]; - - u32 w2[4]; - - w2[0] = 0; - w2[1] = 0; - w2[2] = 0; - w2[3] = 0; - - u32 w3[4]; - - w3[0] = 0; - w3[1] = 0; - w3[2] = 0; - w3[3] = 0; - - const u32 out_len = apply_rules (rules_buf[il_pos].cmds, w0, w1, pw_len); + const u32 out_len = apply_rules_vect (pw_buf0, pw_buf1, pw_len, rules_buf, il_pos, w0, w1); u32 pw_salt_len = out_len + salt_len; @@ -364,28 +313,28 @@ __kernel void m04800_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, * append salt */ - u32 s0[4]; + u32x s0[4]; s0[0] = salt_buf[0]; s0[1] = salt_buf[1]; s0[2] = salt_buf[2]; s0[3] = salt_buf[3]; - u32 s1[4]; + u32x s1[4]; s1[0] = 0x80; s1[1] = 0; s1[2] = 0; s1[3] = 0; - u32 s2[4]; + u32x s2[4]; s2[0] = 0; s2[1] = 0; s2[2] = 0; s2[3] = 0; - u32 s3[4]; + u32x s3[4]; s3[0] = 0; s3[1] = 0; @@ -428,10 +377,10 @@ __kernel void m04800_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, * md5 */ - u32 a = MD5M_A; - u32 b = MD5M_B; - u32 c = MD5M_C; - u32 d = MD5M_D; + u32x a = MD5M_A; + u32x b = MD5M_B; + u32x c = MD5M_C; + u32x d = MD5M_D; MD5_STEP (MD5_Fo, a, b, c, d, w0[0], MD5C00, MD5S00); MD5_STEP (MD5_Fo, d, a, b, c, w0[1], MD5C01, MD5S01); @@ -498,20 +447,13 @@ __kernel void m04800_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, MD5_STEP (MD5_I , b, c, d, a, w3[1], MD5C3b, MD5S33); MD5_STEP (MD5_I , a, b, c, d, w1[0], MD5C3c, MD5S30); - bool q_cond = allx (search[0] != a); - - if (q_cond) continue; + if (MATCHES_NONE_VS (a, search[0])) continue; MD5_STEP (MD5_I , d, a, b, c, w2[3], MD5C3d, MD5S31); MD5_STEP (MD5_I , c, d, a, b, w0[2], MD5C3e, MD5S32); MD5_STEP (MD5_I , b, c, d, a, w2[1], MD5C3f, MD5S33); - const u32 r0 = a; - const u32 r1 = d; - const u32 r2 = c; - const u32 r3 = b; - - #include COMPARE_S + COMPARE_S_SIMD (a, d, c, b); } }