From 4149ecf78f3ede8bb091185b7def0d9105daa6f8 Mon Sep 17 00:00:00 2001 From: Jens Steube Date: Mon, 18 Apr 2016 10:24:44 +0200 Subject: [PATCH] Cleanup -m 101xx kernels to latest standard --- OpenCL/m10100_a0.cl | 20 +- OpenCL/m10100_a1.cl | 463 +++++++++++++++++++++++--------------------- OpenCL/m10100_a3.cl | 36 ++-- 3 files changed, 273 insertions(+), 246 deletions(-) diff --git a/OpenCL/m10100_a0.cl b/OpenCL/m10100_a0.cl index 90f6c95d9..d1bd6e94a 100644 --- a/OpenCL/m10100_a0.cl +++ b/OpenCL/m10100_a0.cl @@ -56,14 +56,12 @@ __kernel void m10100_m04 (__global pw_t *pws, __global kernel_rule_t * rules_bu 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]; @@ -72,7 +70,7 @@ __kernel void m10100_m04 (__global pw_t *pws, __global kernel_rule_t * rules_bu const u32 pw_len = pws[gid].pw_len; /** - * base + * salt */ u64x v0p = SIPHASHM_0; @@ -98,6 +96,10 @@ __kernel void m10100_m04 (__global pw_t *pws, __global kernel_rule_t * rules_bu const u32x out_len = apply_rules_vect (pw_buf0, pw_buf1, pw_len, rules_buf, il_pos, w0, w1); + /** + * siphash + */ + switch (out_len / 8) { case 0: w0[1] |= out_len << 24; break; @@ -181,14 +183,12 @@ __kernel void m10100_s04 (__global pw_t *pws, __global kernel_rule_t * rules_bu 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]; @@ -197,7 +197,7 @@ __kernel void m10100_s04 (__global pw_t *pws, __global kernel_rule_t * rules_bu const u32 pw_len = pws[gid].pw_len; /** - * base + * salt */ u64x v0p = SIPHASHM_0; @@ -235,6 +235,10 @@ __kernel void m10100_s04 (__global pw_t *pws, __global kernel_rule_t * rules_bu const u32x out_len = apply_rules_vect (pw_buf0, pw_buf1, pw_len, rules_buf, il_pos, w0, w1); + /** + * siphash + */ + switch (out_len / 8) { case 0: w0[1] |= out_len << 24; break; diff --git a/OpenCL/m10100_a1.cl b/OpenCL/m10100_a1.cl index b522949af..28bffece7 100644 --- a/OpenCL/m10100_a1.cl +++ b/OpenCL/m10100_a1.cl @@ -5,6 +5,9 @@ #define _SIPHASH_ +//incompatible to simd +//#define NEW_SIMD_CODE + #include "include/constants.h" #include "include/kernel_vendor.h" @@ -16,15 +19,13 @@ #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" #define SIPROUND(v0,v1,v2,v3) \ (v0) += (v1); \ (v1) = rotl64 ((v1), 13); \ (v1) ^= (v0); \ - (v0) = as_ulong (as_uint2 ((v0)).s10); \ + (v0) = rotl64 ((v0), 32); \ (v2) += (v3); \ (v3) = rotl64 ((v3), 16); \ (v3) ^= (v2); \ @@ -34,7 +35,7 @@ (v2) += (v1); \ (v1) = rotl64 ((v1), 17); \ (v1) ^= (v2); \ - (v2) = as_ulong (as_uint2 ((v2)).s10); + (v2) = rotl64 ((v2), 32) __kernel void m10100_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) { @@ -52,49 +53,28 @@ __kernel void m10100_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, if (gid >= gid_max) return; - u32 wordl0[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]; + u32 pw_buf0[4]; + u32 pw_buf1[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) - { - switch_buffer_by_offset_le (wordl0, wordl1, wordl2, wordl3, combs_buf[0].pw_len); - } - /** - * base + * salt */ - u64 v0p = SIPHASHM_0; - u64 v1p = SIPHASHM_1; - u64 v2p = SIPHASHM_2; - u64 v3p = SIPHASHM_3; + u64x v0p = SIPHASHM_0; + u64x v1p = SIPHASHM_1; + u64x v2p = SIPHASHM_2; + u64x v3p = SIPHASHM_3; v0p ^= hl32_to_64 (salt_bufs[salt_pos].salt_buf[1], salt_bufs[salt_pos].salt_buf[0]); v1p ^= hl32_to_64 (salt_bufs[salt_pos].salt_buf[3], salt_bufs[salt_pos].salt_buf[2]); @@ -105,79 +85,110 @@ __kernel void m10100_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, * 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 u32 pw_len = pw_l_len + pw_r_len; - - u32 wordr0[4]; + const u32x pw_r_len = pwlenx_create_combt (combs_buf, il_pos); + + const u32x pw_len = pw_l_len + pw_r_len; + + /** + * concat password candidate + */ + + u32x wordl0[4] = { 0 }; + u32x wordl1[4] = { 0 }; + u32x wordl2[4] = { 0 }; + u32x wordl3[4] = { 0 }; + + 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]; + + u32x wordr0[4] = { 0 }; + u32x wordr1[4] = { 0 }; + u32x wordr2[4] = { 0 }; + u32x wordr3[4] = { 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); - 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]; + if (combs_mode == COMBINATOR_MODE_BASE_LEFT) + { + 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 wordr1[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]; + w1[0] = wordl1[0] | wordr1[0]; + w1[1] = wordl1[1] | wordr1[1]; + w1[2] = wordl1[2] | wordr1[2]; + w1[3] = wordl1[3] | wordr1[3]; + w2[0] = wordl2[0] | wordr2[0]; + w2[1] = wordl2[1] | wordr2[1]; + w2[2] = wordl2[2] | wordr2[2]; + w2[3] = wordl2[3] | wordr2[3]; + w3[0] = wordl3[0] | wordr3[0]; + w3[1] = wordl3[1] | wordr3[1]; + w3[2] = wordl3[2] | wordr3[2]; + w3[3] = wordl3[3] | wordr3[3]; + + /** + * siphash + */ + + switch (pw_len / 8) + { + case 0: w0[1] |= pw_len << 24; break; + case 1: w0[3] |= pw_len << 24; break; + case 2: w1[1] |= pw_len << 24; break; + case 3: w1[3] |= pw_len << 24; break; + } - 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]; + u64x v0 = v0p; + u64x v1 = v1p; + u64x v2 = v2p; + u64x v3 = v3p; - u32 wordr2[4]; + int i; + int j; - wordr2[0] = 0; - wordr2[1] = 0; - wordr2[2] = 0; - wordr2[3] = 0; + for (i = 0, j = 0; i <= pw_len && i < 16; i += 8, j += 2) + { + u64x m = hl32_to_64 (w0[j + 1], w0[j + 0]); - u32 wordr3[4]; + v3 ^= m; - wordr3[0] = 0; - wordr3[1] = 0; - wordr3[2] = 0; - wordr3[3] = 0; + SIPROUND (v0, v1, v2, v3); + SIPROUND (v0, v1, v2, v3); - if (combs_mode == COMBINATOR_MODE_BASE_LEFT) - { - switch_buffer_by_offset_le (wordr0, wordr1, wordr2, wordr3, pw_l_len); + v0 ^= m; } - u32 w[16]; - - w[ 0] = wordl0[0] | wordr0[0]; - w[ 1] = wordl0[1] | wordr0[1]; - w[ 2] = wordl0[2] | wordr0[2]; - w[ 3] = wordl0[3] | wordr0[3]; - w[ 4] = wordl1[0] | wordr1[0]; - w[ 5] = wordl1[1] | wordr1[1]; - w[ 6] = wordl1[2] | wordr1[2]; - w[ 7] = wordl1[3] | wordr1[3]; - w[ 8] = wordl2[0] | wordr2[0]; - w[ 9] = wordl2[1] | wordr2[1]; - w[10] = wordl2[2] | wordr2[2]; - w[11] = wordl2[3] | wordr2[3]; - w[12] = wordl3[0] | wordr3[0]; - w[13] = wordl3[1] | wordr3[1]; - w[14] = wordl3[2] | wordr3[2]; - w[15] = wordl3[3] | wordr3[3]; - - u64 *w_ptr = (u64 *) w; - - w_ptr[pw_len / 8] |= (u64) pw_len << 56; - - u64 v0 = v0p; - u64 v1 = v1p; - u64 v2 = v2p; - u64 v3 = v3p; - - int i; - int j; - - for (i = 0, j = 0; i <= pw_len; i += 8, j += 2) + for ( j = 0; i <= pw_len && i < 32; i += 8, j += 2) { - u64 m = hl32_to_64 (w[j + 1], w[j + 0]); + u64x m = hl32_to_64 (w1[j + 1], w1[j + 0]); v3 ^= m; @@ -194,17 +205,14 @@ __kernel void m10100_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, SIPROUND (v0, v1, v2, v3); SIPROUND (v0, v1, v2, v3); - const u64 v = v0 ^ v1 ^ v2 ^ v3; + const u64x v = v0 ^ v1 ^ v2 ^ v3; - const u32 a = l32_from_64 (v); - const u32 b = h32_from_64 (v); + const u32x a = l32_from_64 (v); + const u32x b = h32_from_64 (v); + const u32x c = 0; + const u32x d = 0; - const u32 r0 = a; - const u32 r1 = b; - const u32 r2 = 0; - const u32 r3 = 0; - - #include COMPARE_M + COMPARE_M_SIMD (a, b, c, d); } } @@ -232,40 +240,33 @@ __kernel void m10100_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, if (gid >= gid_max) return; - u32 wordl0[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 pw_buf0[4]; + u32 pw_buf1[4]; - u32 wordl1[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]; + 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]; - 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]; + const u32 pw_l_len = pws[gid].pw_len; - wordl3[0] = 0; - wordl3[1] = 0; - wordl3[2] = 0; - wordl3[3] = 0; + /** + * salt + */ - const u32 pw_l_len = pws[gid].pw_len; + u64x v0p = SIPHASHM_0; + u64x v1p = SIPHASHM_1; + u64x v2p = SIPHASHM_2; + u64x v3p = SIPHASHM_3; - if (combs_mode == COMBINATOR_MODE_BASE_RIGHT) - { - switch_buffer_by_offset_le (wordl0, wordl1, wordl2, wordl3, combs_buf[0].pw_len); - } + v0p ^= hl32_to_64 (salt_bufs[salt_pos].salt_buf[1], salt_bufs[salt_pos].salt_buf[0]); + v1p ^= hl32_to_64 (salt_bufs[salt_pos].salt_buf[3], salt_bufs[salt_pos].salt_buf[2]); + v2p ^= hl32_to_64 (salt_bufs[salt_pos].salt_buf[1], salt_bufs[salt_pos].salt_buf[0]); + v3p ^= hl32_to_64 (salt_bufs[salt_pos].salt_buf[3], salt_bufs[salt_pos].salt_buf[2]); /** * digest @@ -279,97 +280,114 @@ __kernel void m10100_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, digests_buf[digests_offset].digest_buf[DGST_R3] }; - /** - * base - */ - - u64 v0p = SIPHASHM_0; - u64 v1p = SIPHASHM_1; - u64 v2p = SIPHASHM_2; - u64 v3p = SIPHASHM_3; - - v0p ^= hl32_to_64 (salt_bufs[salt_pos].salt_buf[1], salt_bufs[salt_pos].salt_buf[0]); - v1p ^= hl32_to_64 (salt_bufs[salt_pos].salt_buf[3], salt_bufs[salt_pos].salt_buf[2]); - v2p ^= hl32_to_64 (salt_bufs[salt_pos].salt_buf[1], salt_bufs[salt_pos].salt_buf[0]); - v3p ^= hl32_to_64 (salt_bufs[salt_pos].salt_buf[3], salt_bufs[salt_pos].salt_buf[2]); - /** * 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 u32 pw_len = pw_l_len + pw_r_len; + const u32x pw_r_len = pwlenx_create_combt (combs_buf, il_pos); + + const u32x pw_len = pw_l_len + pw_r_len; + + /** + * concat password candidate + */ + + u32x wordl0[4] = { 0 }; + u32x wordl1[4] = { 0 }; + u32x wordl2[4] = { 0 }; + u32x wordl3[4] = { 0 }; + + 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]; + + u32x wordr0[4] = { 0 }; + u32x wordr1[4] = { 0 }; + u32x wordr2[4] = { 0 }; + u32x wordr3[4] = { 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); - u32 wordr0[4]; - - 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]; + if (combs_mode == COMBINATOR_MODE_BASE_LEFT) + { + 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 wordr1[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]; + w1[0] = wordl1[0] | wordr1[0]; + w1[1] = wordl1[1] | wordr1[1]; + w1[2] = wordl1[2] | wordr1[2]; + w1[3] = wordl1[3] | wordr1[3]; + w2[0] = wordl2[0] | wordr2[0]; + w2[1] = wordl2[1] | wordr2[1]; + w2[2] = wordl2[2] | wordr2[2]; + w2[3] = wordl2[3] | wordr2[3]; + w3[0] = wordl3[0] | wordr3[0]; + w3[1] = wordl3[1] | wordr3[1]; + w3[2] = wordl3[2] | wordr3[2]; + w3[3] = wordl3[3] | wordr3[3]; + + /** + * siphash + */ + + switch (pw_len / 8) + { + case 0: w0[1] |= pw_len << 24; break; + case 1: w0[3] |= pw_len << 24; break; + case 2: w1[1] |= pw_len << 24; break; + case 3: w1[3] |= pw_len << 24; break; + } - 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]; + u64x v0 = v0p; + u64x v1 = v1p; + u64x v2 = v2p; + u64x v3 = v3p; - u32 wordr2[4]; + int i; + int j; - wordr2[0] = 0; - wordr2[1] = 0; - wordr2[2] = 0; - wordr2[3] = 0; + for (i = 0, j = 0; i <= pw_len && i < 16; i += 8, j += 2) + { + u64x m = hl32_to_64 (w0[j + 1], w0[j + 0]); - u32 wordr3[4]; + v3 ^= m; - wordr3[0] = 0; - wordr3[1] = 0; - wordr3[2] = 0; - wordr3[3] = 0; + SIPROUND (v0, v1, v2, v3); + SIPROUND (v0, v1, v2, v3); - if (combs_mode == COMBINATOR_MODE_BASE_LEFT) - { - switch_buffer_by_offset_le (wordr0, wordr1, wordr2, wordr3, pw_l_len); + v0 ^= m; } - u32 w[16]; - - w[ 0] = wordl0[0] | wordr0[0]; - w[ 1] = wordl0[1] | wordr0[1]; - w[ 2] = wordl0[2] | wordr0[2]; - w[ 3] = wordl0[3] | wordr0[3]; - w[ 4] = wordl1[0] | wordr1[0]; - w[ 5] = wordl1[1] | wordr1[1]; - w[ 6] = wordl1[2] | wordr1[2]; - w[ 7] = wordl1[3] | wordr1[3]; - w[ 8] = wordl2[0] | wordr2[0]; - w[ 9] = wordl2[1] | wordr2[1]; - w[10] = wordl2[2] | wordr2[2]; - w[11] = wordl2[3] | wordr2[3]; - w[12] = wordl3[0] | wordr3[0]; - w[13] = wordl3[1] | wordr3[1]; - w[14] = wordl3[2] | wordr3[2]; - w[15] = wordl3[3] | wordr3[3]; - - u64 *w_ptr = (u64 *) w; - - w_ptr[pw_len / 8] |= (u64) pw_len << 56; - - u64 v0 = v0p; - u64 v1 = v1p; - u64 v2 = v2p; - u64 v3 = v3p; - - int i; - int j; - - for (i = 0, j = 0; i <= pw_len; i += 8, j += 2) + for ( j = 0; i <= pw_len && i < 32; i += 8, j += 2) { - u64 m = hl32_to_64 (w[j + 1], w[j + 0]); + u64x m = hl32_to_64 (w1[j + 1], w1[j + 0]); v3 ^= m; @@ -386,17 +404,14 @@ __kernel void m10100_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, SIPROUND (v0, v1, v2, v3); SIPROUND (v0, v1, v2, v3); - const u64 v = v0 ^ v1 ^ v2 ^ v3; - - const u32 a = l32_from_64 (v); - const u32 b = h32_from_64 (v); + const u64x v = v0 ^ v1 ^ v2 ^ v3; - const u32 r0 = a; - const u32 r1 = b; - const u32 r2 = 0; - const u32 r3 = 0; + const u32x a = l32_from_64 (v); + const u32x b = h32_from_64 (v); + const u32x c = 0; + const u32x d = 0; - #include COMPARE_S + COMPARE_S_SIMD (a, b, c, d); } } diff --git a/OpenCL/m10100_a3.cl b/OpenCL/m10100_a3.cl index 9a6df918d..f4a8db3c3 100644 --- a/OpenCL/m10100_a3.cl +++ b/OpenCL/m10100_a3.cl @@ -46,7 +46,7 @@ static void m10100m (u32 w[16], const u32 pw_len, __global pw_t *pws, __global k const u32 lid = get_local_id (0); /** - * base + * salt */ u64 v0p = SIPHASHM_0; @@ -75,6 +75,10 @@ static void m10100m (u32 w[16], const u32 pw_len, __global pw_t *pws, __global k const u32x w0 = w0l | w0r; + /** + * siphash + */ + u64x v0 = v0p; u64x v1 = v1p; u64x v2 = v2p; @@ -132,19 +136,7 @@ static void m10100s (u32 w[16], const u32 pw_len, __global pw_t *pws, __global k 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] - }; - - /** - * base + * salt */ u64 v0p = SIPHASHM_0; @@ -161,6 +153,18 @@ static void m10100s (u32 w[16], const u32 pw_len, __global pw_t *pws, __global k w_ptr[pw_len / 8] |= (u64) pw_len << 56; + /** + * 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 */ @@ -173,6 +177,10 @@ static void m10100s (u32 w[16], const u32 pw_len, __global pw_t *pws, __global k const u32x w0 = w0l | w0r; + /** + * siphash + */ + u64x v0 = v0p; u64x v1 = v1p; u64x v2 = v2p;