From ee0b1d2c21e130e3b1ec8cbfbe198603b1e1c8d3 Mon Sep 17 00:00:00 2001 From: Jens Steube Date: Mon, 18 Apr 2016 14:39:52 +0200 Subject: [PATCH] Cleanup -m 112xx kernels to latest standard --- OpenCL/m11200_a0.cl | 54 +++--- OpenCL/m11200_a1.cl | 406 +++++++++++++++++++------------------------- OpenCL/m11200_a3.cl | 46 +++-- 3 files changed, 220 insertions(+), 286 deletions(-) diff --git a/OpenCL/m11200_a0.cl b/OpenCL/m11200_a0.cl index 0be73dd76..1e39af03f 100644 --- a/OpenCL/m11200_a0.cl +++ b/OpenCL/m11200_a0.cl @@ -39,14 +39,12 @@ __kernel void m11200_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, if (gid >= gid_max) return; u32 pw_buf0[4]; + u32 pw_buf1[4]; pw_buf0[0] = pws[gid].i[ 0]; pw_buf0[1] = pws[gid].i[ 1]; pw_buf0[2] = pws[gid].i[ 2]; pw_buf0[3] = pws[gid].i[ 3]; - - u32 pw_buf1[4]; - pw_buf1[0] = pws[gid].i[ 4]; pw_buf1[1] = pws[gid].i[ 5]; pw_buf1[2] = pws[gid].i[ 6]; @@ -204,11 +202,11 @@ __kernel void m11200_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, we_t = rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, we_t); wf_t = rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, wf_t); - u32x plain_sha1_a = a + SHA1M_A; - u32x plain_sha1_b = b + SHA1M_B; - u32x plain_sha1_c = c + SHA1M_C; - u32x plain_sha1_d = d + SHA1M_D; - u32x plain_sha1_e = e + SHA1M_E; + const u32x plain_sha1_a = a + SHA1M_A; + const u32x plain_sha1_b = b + SHA1M_B; + const u32x plain_sha1_c = c + SHA1M_C; + const u32x plain_sha1_d = d + SHA1M_D; + const u32x plain_sha1_e = e + SHA1M_E; /** * sha1 (sha1 ($pass)) @@ -219,7 +217,6 @@ __kernel void m11200_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, w2_t = plain_sha1_c; w3_t = plain_sha1_d; w4_t = plain_sha1_e; - w5_t = 0x80000000; w6_t = 0; w7_t = 0; @@ -504,14 +501,12 @@ __kernel void m11200_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, if (gid >= gid_max) return; u32 pw_buf0[4]; + u32 pw_buf1[4]; pw_buf0[0] = pws[gid].i[ 0]; pw_buf0[1] = pws[gid].i[ 1]; pw_buf0[2] = pws[gid].i[ 2]; pw_buf0[3] = pws[gid].i[ 3]; - - u32 pw_buf1[4]; - pw_buf1[0] = pws[gid].i[ 4]; pw_buf1[1] = pws[gid].i[ 5]; pw_buf1[2] = pws[gid].i[ 6]; @@ -519,18 +514,6 @@ __kernel void m11200_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, const u32 pw_len = pws[gid].pw_len; - /** - * 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 */ @@ -543,6 +526,18 @@ __kernel void m11200_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, salt_buf[3] = swap32_S (salt_bufs[salt_pos].salt_buf[3]); salt_buf[4] = swap32_S (salt_bufs[salt_pos].salt_buf[4]); + /** + * 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] + }; + /** * loop */ @@ -681,11 +676,11 @@ __kernel void m11200_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, we_t = rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, we_t); wf_t = rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, wf_t); - u32x plain_sha1_a = a + SHA1M_A; - u32x plain_sha1_b = b + SHA1M_B; - u32x plain_sha1_c = c + SHA1M_C; - u32x plain_sha1_d = d + SHA1M_D; - u32x plain_sha1_e = e + SHA1M_E; + const u32x plain_sha1_a = a + SHA1M_A; + const u32x plain_sha1_b = b + SHA1M_B; + const u32x plain_sha1_c = c + SHA1M_C; + const u32x plain_sha1_d = d + SHA1M_D; + const u32x plain_sha1_e = e + SHA1M_E; /** * sha1 (sha1 ($pass)) @@ -696,7 +691,6 @@ __kernel void m11200_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, w2_t = plain_sha1_c; w3_t = plain_sha1_d; w4_t = plain_sha1_e; - w5_t = 0x80000000; w6_t = 0; w7_t = 0; diff --git a/OpenCL/m11200_a1.cl b/OpenCL/m11200_a1.cl index 2d6a36445..062ec3b1d 100644 --- a/OpenCL/m11200_a1.cl +++ b/OpenCL/m11200_a1.cl @@ -5,6 +5,8 @@ #define _SHA1_ +#define NEW_SIMD_CODE + #include "include/constants.h" #include "include/kernel_vendor.h" @@ -16,9 +18,7 @@ #include "include/kernel_functions.c" #include "OpenCL/types_ocl.c" #include "OpenCL/common.c" - -#define COMPARE_S "OpenCL/check_single_comp4.c" -#define COMPARE_M "OpenCL/check_multi_comp4.c" +#include "OpenCL/simd.c" __kernel void m11200_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 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max) { @@ -36,154 +36,131 @@ __kernel void m11200_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, if (gid >= gid_max) return; - u32 wordl0[4]; + u32 pw_buf0[4]; + u32 pw_buf1[4]; - wordl0[0] = pws[gid].i[ 0]; - wordl0[1] = pws[gid].i[ 1]; - wordl0[2] = pws[gid].i[ 2]; - wordl0[3] = pws[gid].i[ 3]; - - u32 wordl1[4]; - - wordl1[0] = pws[gid].i[ 4]; - wordl1[1] = pws[gid].i[ 5]; - wordl1[2] = pws[gid].i[ 6]; - wordl1[3] = pws[gid].i[ 7]; - - u32 wordl2[4]; - - wordl2[0] = 0; - wordl2[1] = 0; - wordl2[2] = 0; - wordl2[3] = 0; - - u32 wordl3[4]; - - wordl3[0] = 0; - wordl3[1] = 0; - wordl3[2] = 0; - wordl3[3] = 0; + pw_buf0[0] = pws[gid].i[0]; + pw_buf0[1] = pws[gid].i[1]; + pw_buf0[2] = pws[gid].i[2]; + pw_buf0[3] = pws[gid].i[3]; + pw_buf1[0] = pws[gid].i[4]; + pw_buf1[1] = pws[gid].i[5]; + pw_buf1[2] = pws[gid].i[6]; + pw_buf1[3] = pws[gid].i[7]; const u32 pw_l_len = pws[gid].pw_len; - if (combs_mode == COMBINATOR_MODE_BASE_RIGHT) - { - append_0x80_2x4 (wordl0, wordl1, pw_l_len); - - switch_buffer_by_offset_le (wordl0, wordl1, wordl2, wordl3, combs_buf[0].pw_len); - } - /** * salt */ u32 salt_buf[5]; - salt_buf[0] = swap32 (salt_bufs[salt_pos].salt_buf[0]); - salt_buf[1] = swap32 (salt_bufs[salt_pos].salt_buf[1]); - salt_buf[2] = swap32 (salt_bufs[salt_pos].salt_buf[2]); - salt_buf[3] = swap32 (salt_bufs[salt_pos].salt_buf[3]); - salt_buf[4] = swap32 (salt_bufs[salt_pos].salt_buf[4]); + salt_buf[0] = swap32_S (salt_bufs[salt_pos].salt_buf[0]); + salt_buf[1] = swap32_S (salt_bufs[salt_pos].salt_buf[1]); + salt_buf[2] = swap32_S (salt_bufs[salt_pos].salt_buf[2]); + salt_buf[3] = swap32_S (salt_bufs[salt_pos].salt_buf[3]); + salt_buf[4] = swap32_S (salt_bufs[salt_pos].salt_buf[4]); /** * loop */ - for (u32 il_pos = 0; il_pos < il_cnt; il_pos++) + for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE) { - const u32 pw_r_len = combs_buf[il_pos].pw_len; + const u32x pw_r_len = pwlenx_create_combt (combs_buf, il_pos); - const u32 pw_len = pw_l_len + pw_r_len; + const u32x pw_len = pw_l_len + pw_r_len; - u32 wordr0[4]; + /** + * concat password candidate + */ - wordr0[0] = combs_buf[il_pos].i[0]; - wordr0[1] = combs_buf[il_pos].i[1]; - wordr0[2] = combs_buf[il_pos].i[2]; - wordr0[3] = combs_buf[il_pos].i[3]; + u32x wordl0[4] = { 0 }; + u32x wordl1[4] = { 0 }; + u32x wordl2[4] = { 0 }; + u32x wordl3[4] = { 0 }; - u32 wordr1[4]; + wordl0[0] = pw_buf0[0]; + wordl0[1] = pw_buf0[1]; + wordl0[2] = pw_buf0[2]; + wordl0[3] = pw_buf0[3]; + wordl1[0] = pw_buf1[0]; + wordl1[1] = pw_buf1[1]; + wordl1[2] = pw_buf1[2]; + wordl1[3] = pw_buf1[3]; - wordr1[0] = combs_buf[il_pos].i[4]; - wordr1[1] = combs_buf[il_pos].i[5]; - wordr1[2] = combs_buf[il_pos].i[6]; - wordr1[3] = combs_buf[il_pos].i[7]; + u32x wordr0[4] = { 0 }; + u32x wordr1[4] = { 0 }; + u32x wordr2[4] = { 0 }; + u32x wordr3[4] = { 0 }; - u32 wordr2[4]; - - wordr2[0] = 0; - wordr2[1] = 0; - wordr2[2] = 0; - wordr2[3] = 0; - - u32 wordr3[4]; - - wordr3[0] = 0; - wordr3[1] = 0; - wordr3[2] = 0; - wordr3[3] = 0; + wordr0[0] = ix_create_combt (combs_buf, il_pos, 0); + wordr0[1] = ix_create_combt (combs_buf, il_pos, 1); + wordr0[2] = ix_create_combt (combs_buf, il_pos, 2); + wordr0[3] = ix_create_combt (combs_buf, il_pos, 3); + wordr1[0] = ix_create_combt (combs_buf, il_pos, 4); + wordr1[1] = ix_create_combt (combs_buf, il_pos, 5); + wordr1[2] = ix_create_combt (combs_buf, il_pos, 6); + wordr1[3] = ix_create_combt (combs_buf, il_pos, 7); if (combs_mode == COMBINATOR_MODE_BASE_LEFT) { - append_0x80_2x4 (wordr0, wordr1, pw_r_len); - - switch_buffer_by_offset_le (wordr0, wordr1, wordr2, wordr3, pw_l_len); + switch_buffer_by_offset_le_VV (wordr0, wordr1, wordr2, wordr3, pw_l_len); + } + else + { + switch_buffer_by_offset_le_VV (wordl0, wordl1, wordl2, wordl3, pw_r_len); } - u32 w0[4]; + u32x w0[4]; + u32x w1[4]; + u32x w2[4]; + u32x w3[4]; w0[0] = wordl0[0] | wordr0[0]; w0[1] = wordl0[1] | wordr0[1]; w0[2] = wordl0[2] | wordr0[2]; w0[3] = wordl0[3] | wordr0[3]; - - u32 w1[4]; - w1[0] = wordl1[0] | wordr1[0]; w1[1] = wordl1[1] | wordr1[1]; w1[2] = wordl1[2] | wordr1[2]; w1[3] = wordl1[3] | wordr1[3]; - - u32 w2[4]; - w2[0] = wordl2[0] | wordr2[0]; w2[1] = wordl2[1] | wordr2[1]; w2[2] = wordl2[2] | wordr2[2]; w2[3] = wordl2[3] | wordr2[3]; - - u32 w3[4]; - w3[0] = wordl3[0] | wordr3[0]; w3[1] = wordl3[1] | wordr3[1]; - w3[2] = 0; - w3[3] = 0; + w3[2] = wordl3[2] | wordr3[2]; + w3[3] = wordl3[3] | wordr3[3]; /** * sha1 ($pass) */ - u32 w0_t = swap32 (w0[0]); - u32 w1_t = swap32 (w0[1]); - u32 w2_t = swap32 (w0[2]); - u32 w3_t = swap32 (w0[3]); - u32 w4_t = swap32 (w1[0]); - u32 w5_t = swap32 (w1[1]); - u32 w6_t = swap32 (w1[2]); - u32 w7_t = swap32 (w1[3]); - u32 w8_t = swap32 (w2[0]); - u32 w9_t = swap32 (w2[1]); - u32 wa_t = swap32 (w2[2]); - u32 wb_t = swap32 (w2[3]); - u32 wc_t = swap32 (w3[0]); - u32 wd_t = swap32 (w3[1]); - u32 we_t = 0; - u32 wf_t = pw_len * 8; + u32x w0_t = swap32 (w0[0]); + u32x w1_t = swap32 (w0[1]); + u32x w2_t = swap32 (w0[2]); + u32x w3_t = swap32 (w0[3]); + u32x w4_t = swap32 (w1[0]); + u32x w5_t = swap32 (w1[1]); + u32x w6_t = swap32 (w1[2]); + u32x w7_t = swap32 (w1[3]); + u32x w8_t = swap32 (w2[0]); + u32x w9_t = swap32 (w2[1]); + u32x wa_t = swap32 (w2[2]); + u32x wb_t = swap32 (w2[3]); + u32x wc_t = swap32 (w3[0]); + u32x wd_t = swap32 (w3[1]); + u32x we_t = 0; + u32x wf_t = pw_len * 8; - u32 a = SHA1M_A; - u32 b = SHA1M_B; - u32 c = SHA1M_C; - u32 d = SHA1M_D; - u32 e = SHA1M_E; + u32x a = SHA1M_A; + u32x b = SHA1M_B; + u32x c = SHA1M_C; + u32x d = SHA1M_D; + u32x e = SHA1M_E; #undef K #define K SHA1C00 @@ -281,11 +258,11 @@ __kernel void m11200_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, we_t = rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, we_t); wf_t = rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, wf_t); - u32 plain_sha1_a = a + SHA1M_A; - u32 plain_sha1_b = b + SHA1M_B; - u32 plain_sha1_c = c + SHA1M_C; - u32 plain_sha1_d = d + SHA1M_D; - u32 plain_sha1_e = e + SHA1M_E; + const u32x plain_sha1_a = a + SHA1M_A; + const u32x plain_sha1_b = b + SHA1M_B; + const u32x plain_sha1_c = c + SHA1M_C; + const u32x plain_sha1_d = d + SHA1M_D; + const u32x plain_sha1_e = e + SHA1M_E; /** * sha1 (sha1 ($pass)) @@ -296,7 +273,6 @@ __kernel void m11200_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, w2_t = plain_sha1_c; w3_t = plain_sha1_d; w4_t = plain_sha1_e; - w5_t = 0x80000000; w6_t = 0; w7_t = 0; @@ -552,12 +528,7 @@ __kernel void m11200_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, d ^= plain_sha1_d; e ^= plain_sha1_e; - const u32 r0 = d; - const u32 r1 = e; - const u32 r2 = c; - const u32 r3 = b; - - #include COMPARE_M + COMPARE_M_SIMD (d, e, c, b); } } @@ -585,42 +556,31 @@ __kernel void m11200_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, if (gid >= gid_max) return; - u32 wordl0[4]; + u32 pw_buf0[4]; + u32 pw_buf1[4]; - wordl0[0] = pws[gid].i[ 0]; - wordl0[1] = pws[gid].i[ 1]; - wordl0[2] = pws[gid].i[ 2]; - wordl0[3] = pws[gid].i[ 3]; - - u32 wordl1[4]; - - wordl1[0] = pws[gid].i[ 4]; - wordl1[1] = pws[gid].i[ 5]; - wordl1[2] = pws[gid].i[ 6]; - wordl1[3] = pws[gid].i[ 7]; - - u32 wordl2[4]; - - wordl2[0] = 0; - wordl2[1] = 0; - wordl2[2] = 0; - wordl2[3] = 0; - - u32 wordl3[4]; - - wordl3[0] = 0; - wordl3[1] = 0; - wordl3[2] = 0; - wordl3[3] = 0; + pw_buf0[0] = pws[gid].i[0]; + pw_buf0[1] = pws[gid].i[1]; + pw_buf0[2] = pws[gid].i[2]; + pw_buf0[3] = pws[gid].i[3]; + pw_buf1[0] = pws[gid].i[4]; + pw_buf1[1] = pws[gid].i[5]; + pw_buf1[2] = pws[gid].i[6]; + pw_buf1[3] = pws[gid].i[7]; const u32 pw_l_len = pws[gid].pw_len; - if (combs_mode == COMBINATOR_MODE_BASE_RIGHT) - { - append_0x80_2x4 (wordl0, wordl1, pw_l_len); + /** + * salt + */ - switch_buffer_by_offset_le (wordl0, wordl1, wordl2, wordl3, combs_buf[0].pw_len); - } + u32 salt_buf[5]; + + salt_buf[0] = swap32_S (salt_bufs[salt_pos].salt_buf[0]); + salt_buf[1] = swap32_S (salt_bufs[salt_pos].salt_buf[1]); + salt_buf[2] = swap32_S (salt_bufs[salt_pos].salt_buf[2]); + salt_buf[3] = swap32_S (salt_bufs[salt_pos].salt_buf[3]); + salt_buf[4] = swap32_S (salt_bufs[salt_pos].salt_buf[4]); /** * digest @@ -634,117 +594,105 @@ __kernel void m11200_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, digests_buf[digests_offset].digest_buf[DGST_R3] }; - /** - * salt - */ - - u32 salt_buf[5]; - - salt_buf[0] = swap32 (salt_bufs[salt_pos].salt_buf[0]); - salt_buf[1] = swap32 (salt_bufs[salt_pos].salt_buf[1]); - salt_buf[2] = swap32 (salt_bufs[salt_pos].salt_buf[2]); - salt_buf[3] = swap32 (salt_bufs[salt_pos].salt_buf[3]); - salt_buf[4] = swap32 (salt_bufs[salt_pos].salt_buf[4]); - /** * loop */ - for (u32 il_pos = 0; il_pos < il_cnt; il_pos++) + for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE) { - const u32 pw_r_len = combs_buf[il_pos].pw_len; + const u32x pw_r_len = pwlenx_create_combt (combs_buf, il_pos); - const u32 pw_len = pw_l_len + pw_r_len; + const u32x pw_len = pw_l_len + pw_r_len; - u32 wordr0[4]; + /** + * concat password candidate + */ - wordr0[0] = combs_buf[il_pos].i[0]; - wordr0[1] = combs_buf[il_pos].i[1]; - wordr0[2] = combs_buf[il_pos].i[2]; - wordr0[3] = combs_buf[il_pos].i[3]; + u32x wordl0[4] = { 0 }; + u32x wordl1[4] = { 0 }; + u32x wordl2[4] = { 0 }; + u32x wordl3[4] = { 0 }; - u32 wordr1[4]; + wordl0[0] = pw_buf0[0]; + wordl0[1] = pw_buf0[1]; + wordl0[2] = pw_buf0[2]; + wordl0[3] = pw_buf0[3]; + wordl1[0] = pw_buf1[0]; + wordl1[1] = pw_buf1[1]; + wordl1[2] = pw_buf1[2]; + wordl1[3] = pw_buf1[3]; - wordr1[0] = combs_buf[il_pos].i[4]; - wordr1[1] = combs_buf[il_pos].i[5]; - wordr1[2] = combs_buf[il_pos].i[6]; - wordr1[3] = combs_buf[il_pos].i[7]; + u32x wordr0[4] = { 0 }; + u32x wordr1[4] = { 0 }; + u32x wordr2[4] = { 0 }; + u32x wordr3[4] = { 0 }; - u32 wordr2[4]; - - wordr2[0] = 0; - wordr2[1] = 0; - wordr2[2] = 0; - wordr2[3] = 0; - - u32 wordr3[4]; - - wordr3[0] = 0; - wordr3[1] = 0; - wordr3[2] = 0; - wordr3[3] = 0; + wordr0[0] = ix_create_combt (combs_buf, il_pos, 0); + wordr0[1] = ix_create_combt (combs_buf, il_pos, 1); + wordr0[2] = ix_create_combt (combs_buf, il_pos, 2); + wordr0[3] = ix_create_combt (combs_buf, il_pos, 3); + wordr1[0] = ix_create_combt (combs_buf, il_pos, 4); + wordr1[1] = ix_create_combt (combs_buf, il_pos, 5); + wordr1[2] = ix_create_combt (combs_buf, il_pos, 6); + wordr1[3] = ix_create_combt (combs_buf, il_pos, 7); if (combs_mode == COMBINATOR_MODE_BASE_LEFT) { - append_0x80_2x4 (wordr0, wordr1, pw_r_len); - - switch_buffer_by_offset_le (wordr0, wordr1, wordr2, wordr3, pw_l_len); + switch_buffer_by_offset_le_VV (wordr0, wordr1, wordr2, wordr3, pw_l_len); + } + else + { + switch_buffer_by_offset_le_VV (wordl0, wordl1, wordl2, wordl3, pw_r_len); } - u32 w0[4]; + u32x w0[4]; + u32x w1[4]; + u32x w2[4]; + u32x w3[4]; w0[0] = wordl0[0] | wordr0[0]; w0[1] = wordl0[1] | wordr0[1]; w0[2] = wordl0[2] | wordr0[2]; w0[3] = wordl0[3] | wordr0[3]; - - u32 w1[4]; - w1[0] = wordl1[0] | wordr1[0]; w1[1] = wordl1[1] | wordr1[1]; w1[2] = wordl1[2] | wordr1[2]; w1[3] = wordl1[3] | wordr1[3]; - - u32 w2[4]; - w2[0] = wordl2[0] | wordr2[0]; w2[1] = wordl2[1] | wordr2[1]; w2[2] = wordl2[2] | wordr2[2]; w2[3] = wordl2[3] | wordr2[3]; - - u32 w3[4]; - w3[0] = wordl3[0] | wordr3[0]; w3[1] = wordl3[1] | wordr3[1]; - w3[2] = 0; - w3[3] = 0; + w3[2] = wordl3[2] | wordr3[2]; + w3[3] = wordl3[3] | wordr3[3]; /** * sha1 ($pass) */ - u32 w0_t = swap32 (w0[0]); - u32 w1_t = swap32 (w0[1]); - u32 w2_t = swap32 (w0[2]); - u32 w3_t = swap32 (w0[3]); - u32 w4_t = swap32 (w1[0]); - u32 w5_t = swap32 (w1[1]); - u32 w6_t = swap32 (w1[2]); - u32 w7_t = swap32 (w1[3]); - u32 w8_t = swap32 (w2[0]); - u32 w9_t = swap32 (w2[1]); - u32 wa_t = swap32 (w2[2]); - u32 wb_t = swap32 (w2[3]); - u32 wc_t = swap32 (w3[0]); - u32 wd_t = swap32 (w3[1]); - u32 we_t = 0; - u32 wf_t = pw_len * 8; + u32x w0_t = swap32 (w0[0]); + u32x w1_t = swap32 (w0[1]); + u32x w2_t = swap32 (w0[2]); + u32x w3_t = swap32 (w0[3]); + u32x w4_t = swap32 (w1[0]); + u32x w5_t = swap32 (w1[1]); + u32x w6_t = swap32 (w1[2]); + u32x w7_t = swap32 (w1[3]); + u32x w8_t = swap32 (w2[0]); + u32x w9_t = swap32 (w2[1]); + u32x wa_t = swap32 (w2[2]); + u32x wb_t = swap32 (w2[3]); + u32x wc_t = swap32 (w3[0]); + u32x wd_t = swap32 (w3[1]); + u32x we_t = 0; + u32x wf_t = pw_len * 8; - u32 a = SHA1M_A; - u32 b = SHA1M_B; - u32 c = SHA1M_C; - u32 d = SHA1M_D; - u32 e = SHA1M_E; + u32x a = SHA1M_A; + u32x b = SHA1M_B; + u32x c = SHA1M_C; + u32x d = SHA1M_D; + u32x e = SHA1M_E; #undef K #define K SHA1C00 @@ -842,11 +790,11 @@ __kernel void m11200_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, we_t = rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, we_t); wf_t = rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, wf_t); - u32 plain_sha1_a = a + SHA1M_A; - u32 plain_sha1_b = b + SHA1M_B; - u32 plain_sha1_c = c + SHA1M_C; - u32 plain_sha1_d = d + SHA1M_D; - u32 plain_sha1_e = e + SHA1M_E; + const u32x plain_sha1_a = a + SHA1M_A; + const u32x plain_sha1_b = b + SHA1M_B; + const u32x plain_sha1_c = c + SHA1M_C; + const u32x plain_sha1_d = d + SHA1M_D; + const u32x plain_sha1_e = e + SHA1M_E; /** * sha1 (sha1 ($pass)) @@ -857,7 +805,6 @@ __kernel void m11200_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, w2_t = plain_sha1_c; w3_t = plain_sha1_d; w4_t = plain_sha1_e; - w5_t = 0x80000000; w6_t = 0; w7_t = 0; @@ -1113,12 +1060,7 @@ __kernel void m11200_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, d ^= plain_sha1_d; e ^= plain_sha1_e; - const u32 r0 = d; - const u32 r1 = e; - const u32 r2 = c; - const u32 r3 = b; - - #include COMPARE_S + COMPARE_S_SIMD (d, e, c, b); } } diff --git a/OpenCL/m11200_a3.cl b/OpenCL/m11200_a3.cl index 4fde8171b..52e18e32b 100644 --- a/OpenCL/m11200_a3.cl +++ b/OpenCL/m11200_a3.cl @@ -176,11 +176,11 @@ static void m11200m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_le we_t = rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, we_t); wf_t = rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, wf_t); - u32x plain_sha1_a = a + SHA1M_A; - u32x plain_sha1_b = b + SHA1M_B; - u32x plain_sha1_c = c + SHA1M_C; - u32x plain_sha1_d = d + SHA1M_D; - u32x plain_sha1_e = e + SHA1M_E; + const u32x plain_sha1_a = a + SHA1M_A; + const u32x plain_sha1_b = b + SHA1M_B; + const u32x plain_sha1_c = c + SHA1M_C; + const u32x plain_sha1_d = d + SHA1M_D; + const u32x plain_sha1_e = e + SHA1M_E; /** * sha1 (sha1 ($pass)) @@ -191,7 +191,6 @@ static void m11200m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_le w2_t = plain_sha1_c; w3_t = plain_sha1_d; w4_t = plain_sha1_e; - w5_t = 0x80000000; w6_t = 0; w7_t = 0; @@ -460,18 +459,6 @@ static void m11200s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_le const u32 gid = get_global_id (0); const u32 lid = get_local_id (0); - /** - * digest - */ - - const u32 search[4] = - { - digests_buf[digests_offset].digest_buf[DGST_R0], - digests_buf[digests_offset].digest_buf[DGST_R1], - digests_buf[digests_offset].digest_buf[DGST_R2], - digests_buf[digests_offset].digest_buf[DGST_R3] - }; - /** * salt */ @@ -484,6 +471,18 @@ static void m11200s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_le salt_buf[3] = swap32_S (salt_bufs[salt_pos].salt_buf[3]); salt_buf[4] = swap32_S (salt_bufs[salt_pos].salt_buf[4]); + /** + * digest + */ + + const u32 search[4] = + { + digests_buf[digests_offset].digest_buf[DGST_R0], + digests_buf[digests_offset].digest_buf[DGST_R1], + digests_buf[digests_offset].digest_buf[DGST_R2], + digests_buf[digests_offset].digest_buf[DGST_R3] + }; + /** * loop */ @@ -619,11 +618,11 @@ static void m11200s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_le we_t = rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, we_t); wf_t = rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, wf_t); - u32x plain_sha1_a = a + SHA1M_A; - u32x plain_sha1_b = b + SHA1M_B; - u32x plain_sha1_c = c + SHA1M_C; - u32x plain_sha1_d = d + SHA1M_D; - u32x plain_sha1_e = e + SHA1M_E; + const u32x plain_sha1_a = a + SHA1M_A; + const u32x plain_sha1_b = b + SHA1M_B; + const u32x plain_sha1_c = c + SHA1M_C; + const u32x plain_sha1_d = d + SHA1M_D; + const u32x plain_sha1_e = e + SHA1M_E; /** * sha1 (sha1 ($pass)) @@ -634,7 +633,6 @@ static void m11200s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_le w2_t = plain_sha1_c; w3_t = plain_sha1_d; w4_t = plain_sha1_e; - w5_t = 0x80000000; w6_t = 0; w7_t = 0;