From ef3afbab9d500d4625ecb7386c259cc013b996fd Mon Sep 17 00:00:00 2001 From: Jens Steube Date: Fri, 5 Feb 2016 21:20:42 +0100 Subject: [PATCH] Converted to new SIMD: -m 2710 -a 0 --- OpenCL/m02710_a0.cl | 479 +++++++++++++++++++------------------------- 1 file changed, 211 insertions(+), 268 deletions(-) diff --git a/OpenCL/m02710_a0.cl b/OpenCL/m02710_a0.cl index ad00e5503..a0002bbe6 100644 --- a/OpenCL/m02710_a0.cl +++ b/OpenCL/m02710_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" #if VECT_SIZE == 1 #define uint_to_hex_lower8(i) (u32x) (l_bin2asc[(i)]) @@ -104,46 +104,23 @@ __kernel void m02710_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); append_0x80_2x4 (w0, w1, out_len); w3[2] = out_len * 8; - 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); @@ -218,109 +195,110 @@ __kernel void m02710_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, c += MD5M_C; d += MD5M_D; - const u32 w0_t = uint_to_hex_lower8 ((a >> 0) & 255) << 0 - | uint_to_hex_lower8 ((a >> 8) & 255) << 16; - const u32 w1_t = uint_to_hex_lower8 ((a >> 16) & 255) << 0 - | uint_to_hex_lower8 ((a >> 24) & 255) << 16; - const u32 w2_t = uint_to_hex_lower8 ((b >> 0) & 255) << 0 - | uint_to_hex_lower8 ((b >> 8) & 255) << 16; - const u32 w3_t = uint_to_hex_lower8 ((b >> 16) & 255) << 0 - | uint_to_hex_lower8 ((b >> 24) & 255) << 16; - const u32 w4_t = uint_to_hex_lower8 ((c >> 0) & 255) << 0 - | uint_to_hex_lower8 ((c >> 8) & 255) << 16; - const u32 w5_t = uint_to_hex_lower8 ((c >> 16) & 255) << 0 - | uint_to_hex_lower8 ((c >> 24) & 255) << 16; - const u32 w6_t = uint_to_hex_lower8 ((d >> 0) & 255) << 0 - | uint_to_hex_lower8 ((d >> 8) & 255) << 16; - const u32 w7_t = uint_to_hex_lower8 ((d >> 16) & 255) << 0 - | uint_to_hex_lower8 ((d >> 24) & 255) << 16; + w0[0] = uint_to_hex_lower8 ((a >> 0) & 255) << 0 + | uint_to_hex_lower8 ((a >> 8) & 255) << 16; + w0[1] = uint_to_hex_lower8 ((a >> 16) & 255) << 0 + | uint_to_hex_lower8 ((a >> 24) & 255) << 16; + w0[2] = uint_to_hex_lower8 ((b >> 0) & 255) << 0 + | uint_to_hex_lower8 ((b >> 8) & 255) << 16; + w0[3] = uint_to_hex_lower8 ((b >> 16) & 255) << 0 + | uint_to_hex_lower8 ((b >> 24) & 255) << 16; + w1[0] = uint_to_hex_lower8 ((c >> 0) & 255) << 0 + | uint_to_hex_lower8 ((c >> 8) & 255) << 16; + w1[1] = uint_to_hex_lower8 ((c >> 16) & 255) << 0 + | uint_to_hex_lower8 ((c >> 24) & 255) << 16; + w1[2] = uint_to_hex_lower8 ((d >> 0) & 255) << 0 + | uint_to_hex_lower8 ((d >> 8) & 255) << 16; + w1[3] = uint_to_hex_lower8 ((d >> 16) & 255) << 0 + | uint_to_hex_lower8 ((d >> 24) & 255) << 16; - const u32 w8_t = s[0]; - const u32 w9_t = s[1]; - const u32 wa_t = s[2]; - const u32 wb_t = s[3]; - const u32 wc_t = s[4]; - const u32 wd_t = s[5]; - const u32 we_t = s[6]; - const u32 wf_t = s[7]; + w2[0] = s[0]; + w2[1] = s[1]; + w2[2] = s[2]; + w2[3] = s[3]; + + w3[0] = s[4]; + w3[1] = s[5]; + w3[2] = s[6]; + w3[3] = s[7]; a = MD5M_A; b = MD5M_B; c = MD5M_C; d = MD5M_D; - MD5_STEP (MD5_Fo, a, b, c, d, w0_t, MD5C00, MD5S00); - MD5_STEP (MD5_Fo, d, a, b, c, w1_t, MD5C01, MD5S01); - MD5_STEP (MD5_Fo, c, d, a, b, w2_t, MD5C02, MD5S02); - MD5_STEP (MD5_Fo, b, c, d, a, w3_t, MD5C03, MD5S03); - MD5_STEP (MD5_Fo, a, b, c, d, w4_t, MD5C04, MD5S00); - MD5_STEP (MD5_Fo, d, a, b, c, w5_t, MD5C05, MD5S01); - MD5_STEP (MD5_Fo, c, d, a, b, w6_t, MD5C06, MD5S02); - MD5_STEP (MD5_Fo, b, c, d, a, w7_t, MD5C07, MD5S03); - MD5_STEP (MD5_Fo, a, b, c, d, w8_t, MD5C08, MD5S00); - MD5_STEP (MD5_Fo, d, a, b, c, w9_t, MD5C09, MD5S01); - MD5_STEP (MD5_Fo, c, d, a, b, wa_t, MD5C0a, MD5S02); - MD5_STEP (MD5_Fo, b, c, d, a, wb_t, MD5C0b, MD5S03); - MD5_STEP (MD5_Fo, a, b, c, d, wc_t, MD5C0c, MD5S00); - MD5_STEP (MD5_Fo, d, a, b, c, wd_t, MD5C0d, MD5S01); - MD5_STEP (MD5_Fo, c, d, a, b, we_t, MD5C0e, MD5S02); - MD5_STEP (MD5_Fo, b, c, d, a, wf_t, MD5C0f, MD5S03); + MD5_STEP (MD5_Fo, a, b, c, d, w0[0], MD5C00, MD5S00); + MD5_STEP (MD5_Fo, d, a, b, c, w0[1], MD5C01, MD5S01); + MD5_STEP (MD5_Fo, c, d, a, b, w0[2], MD5C02, MD5S02); + MD5_STEP (MD5_Fo, b, c, d, a, w0[3], MD5C03, MD5S03); + MD5_STEP (MD5_Fo, a, b, c, d, w1[0], MD5C04, MD5S00); + MD5_STEP (MD5_Fo, d, a, b, c, w1[1], MD5C05, MD5S01); + MD5_STEP (MD5_Fo, c, d, a, b, w1[2], MD5C06, MD5S02); + MD5_STEP (MD5_Fo, b, c, d, a, w1[3], MD5C07, MD5S03); + MD5_STEP (MD5_Fo, a, b, c, d, w2[0], MD5C08, MD5S00); + MD5_STEP (MD5_Fo, d, a, b, c, w2[1], MD5C09, MD5S01); + MD5_STEP (MD5_Fo, c, d, a, b, w2[2], MD5C0a, MD5S02); + MD5_STEP (MD5_Fo, b, c, d, a, w2[3], MD5C0b, MD5S03); + MD5_STEP (MD5_Fo, a, b, c, d, w3[0], MD5C0c, MD5S00); + MD5_STEP (MD5_Fo, d, a, b, c, w3[1], MD5C0d, MD5S01); + MD5_STEP (MD5_Fo, c, d, a, b, w3[2], MD5C0e, MD5S02); + MD5_STEP (MD5_Fo, b, c, d, a, w3[3], MD5C0f, MD5S03); - MD5_STEP (MD5_Go, a, b, c, d, w1_t, MD5C10, MD5S10); - MD5_STEP (MD5_Go, d, a, b, c, w6_t, MD5C11, MD5S11); - MD5_STEP (MD5_Go, c, d, a, b, wb_t, MD5C12, MD5S12); - MD5_STEP (MD5_Go, b, c, d, a, w0_t, MD5C13, MD5S13); - MD5_STEP (MD5_Go, a, b, c, d, w5_t, MD5C14, MD5S10); - MD5_STEP (MD5_Go, d, a, b, c, wa_t, MD5C15, MD5S11); - MD5_STEP (MD5_Go, c, d, a, b, wf_t, MD5C16, MD5S12); - MD5_STEP (MD5_Go, b, c, d, a, w4_t, MD5C17, MD5S13); - MD5_STEP (MD5_Go, a, b, c, d, w9_t, MD5C18, MD5S10); - MD5_STEP (MD5_Go, d, a, b, c, we_t, MD5C19, MD5S11); - MD5_STEP (MD5_Go, c, d, a, b, w3_t, MD5C1a, MD5S12); - MD5_STEP (MD5_Go, b, c, d, a, w8_t, MD5C1b, MD5S13); - MD5_STEP (MD5_Go, a, b, c, d, wd_t, MD5C1c, MD5S10); - MD5_STEP (MD5_Go, d, a, b, c, w2_t, MD5C1d, MD5S11); - MD5_STEP (MD5_Go, c, d, a, b, w7_t, MD5C1e, MD5S12); - MD5_STEP (MD5_Go, b, c, d, a, wc_t, MD5C1f, MD5S13); + MD5_STEP (MD5_Go, a, b, c, d, w0[1], MD5C10, MD5S10); + MD5_STEP (MD5_Go, d, a, b, c, w1[2], MD5C11, MD5S11); + MD5_STEP (MD5_Go, c, d, a, b, w2[3], MD5C12, MD5S12); + MD5_STEP (MD5_Go, b, c, d, a, w0[0], MD5C13, MD5S13); + MD5_STEP (MD5_Go, a, b, c, d, w1[1], MD5C14, MD5S10); + MD5_STEP (MD5_Go, d, a, b, c, w2[2], MD5C15, MD5S11); + MD5_STEP (MD5_Go, c, d, a, b, w3[3], MD5C16, MD5S12); + MD5_STEP (MD5_Go, b, c, d, a, w1[0], MD5C17, MD5S13); + MD5_STEP (MD5_Go, a, b, c, d, w2[1], MD5C18, MD5S10); + MD5_STEP (MD5_Go, d, a, b, c, w3[2], MD5C19, MD5S11); + MD5_STEP (MD5_Go, c, d, a, b, w0[3], MD5C1a, MD5S12); + MD5_STEP (MD5_Go, b, c, d, a, w2[0], MD5C1b, MD5S13); + MD5_STEP (MD5_Go, a, b, c, d, w3[1], MD5C1c, MD5S10); + MD5_STEP (MD5_Go, d, a, b, c, w0[2], MD5C1d, MD5S11); + MD5_STEP (MD5_Go, c, d, a, b, w1[3], MD5C1e, MD5S12); + MD5_STEP (MD5_Go, b, c, d, a, w3[0], MD5C1f, MD5S13); - MD5_STEP (MD5_H , a, b, c, d, w5_t, MD5C20, MD5S20); - MD5_STEP (MD5_H , d, a, b, c, w8_t, MD5C21, MD5S21); - MD5_STEP (MD5_H , c, d, a, b, wb_t, MD5C22, MD5S22); - MD5_STEP (MD5_H , b, c, d, a, we_t, MD5C23, MD5S23); - MD5_STEP (MD5_H , a, b, c, d, w1_t, MD5C24, MD5S20); - MD5_STEP (MD5_H , d, a, b, c, w4_t, MD5C25, MD5S21); - MD5_STEP (MD5_H , c, d, a, b, w7_t, MD5C26, MD5S22); - MD5_STEP (MD5_H , b, c, d, a, wa_t, MD5C27, MD5S23); - MD5_STEP (MD5_H , a, b, c, d, wd_t, MD5C28, MD5S20); - MD5_STEP (MD5_H , d, a, b, c, w0_t, MD5C29, MD5S21); - MD5_STEP (MD5_H , c, d, a, b, w3_t, MD5C2a, MD5S22); - MD5_STEP (MD5_H , b, c, d, a, w6_t, MD5C2b, MD5S23); - MD5_STEP (MD5_H , a, b, c, d, w9_t, MD5C2c, MD5S20); - MD5_STEP (MD5_H , d, a, b, c, wc_t, MD5C2d, MD5S21); - MD5_STEP (MD5_H , c, d, a, b, wf_t, MD5C2e, MD5S22); - MD5_STEP (MD5_H , b, c, d, a, w2_t, MD5C2f, MD5S23); + MD5_STEP (MD5_H , a, b, c, d, w1[1], MD5C20, MD5S20); + MD5_STEP (MD5_H , d, a, b, c, w2[0], MD5C21, MD5S21); + MD5_STEP (MD5_H , c, d, a, b, w2[3], MD5C22, MD5S22); + MD5_STEP (MD5_H , b, c, d, a, w3[2], MD5C23, MD5S23); + MD5_STEP (MD5_H , a, b, c, d, w0[1], MD5C24, MD5S20); + MD5_STEP (MD5_H , d, a, b, c, w1[0], MD5C25, MD5S21); + MD5_STEP (MD5_H , c, d, a, b, w1[3], MD5C26, MD5S22); + MD5_STEP (MD5_H , b, c, d, a, w2[2], MD5C27, MD5S23); + MD5_STEP (MD5_H , a, b, c, d, w3[1], MD5C28, MD5S20); + MD5_STEP (MD5_H , d, a, b, c, w0[0], MD5C29, MD5S21); + MD5_STEP (MD5_H , c, d, a, b, w0[3], MD5C2a, MD5S22); + MD5_STEP (MD5_H , b, c, d, a, w1[2], MD5C2b, MD5S23); + MD5_STEP (MD5_H , a, b, c, d, w2[1], MD5C2c, MD5S20); + MD5_STEP (MD5_H , d, a, b, c, w3[0], MD5C2d, MD5S21); + MD5_STEP (MD5_H , c, d, a, b, w3[3], MD5C2e, MD5S22); + MD5_STEP (MD5_H , b, c, d, a, w0[2], MD5C2f, MD5S23); - MD5_STEP (MD5_I , a, b, c, d, w0_t, MD5C30, MD5S30); - MD5_STEP (MD5_I , d, a, b, c, w7_t, MD5C31, MD5S31); - MD5_STEP (MD5_I , c, d, a, b, we_t, MD5C32, MD5S32); - MD5_STEP (MD5_I , b, c, d, a, w5_t, MD5C33, MD5S33); - MD5_STEP (MD5_I , a, b, c, d, wc_t, MD5C34, MD5S30); - MD5_STEP (MD5_I , d, a, b, c, w3_t, MD5C35, MD5S31); - MD5_STEP (MD5_I , c, d, a, b, wa_t, MD5C36, MD5S32); - MD5_STEP (MD5_I , b, c, d, a, w1_t, MD5C37, MD5S33); - MD5_STEP (MD5_I , a, b, c, d, w8_t, MD5C38, MD5S30); - MD5_STEP (MD5_I , d, a, b, c, wf_t, MD5C39, MD5S31); - MD5_STEP (MD5_I , c, d, a, b, w6_t, MD5C3a, MD5S32); - MD5_STEP (MD5_I , b, c, d, a, wd_t, MD5C3b, MD5S33); - MD5_STEP (MD5_I , a, b, c, d, w4_t, MD5C3c, MD5S30); - MD5_STEP (MD5_I , d, a, b, c, wb_t, MD5C3d, MD5S31); - MD5_STEP (MD5_I , c, d, a, b, w2_t, MD5C3e, MD5S32); - MD5_STEP (MD5_I , b, c, d, a, w9_t, MD5C3f, MD5S33); + MD5_STEP (MD5_I , a, b, c, d, w0[0], MD5C30, MD5S30); + MD5_STEP (MD5_I , d, a, b, c, w1[3], MD5C31, MD5S31); + MD5_STEP (MD5_I , c, d, a, b, w3[2], MD5C32, MD5S32); + MD5_STEP (MD5_I , b, c, d, a, w1[1], MD5C33, MD5S33); + MD5_STEP (MD5_I , a, b, c, d, w3[0], MD5C34, MD5S30); + MD5_STEP (MD5_I , d, a, b, c, w0[3], MD5C35, MD5S31); + MD5_STEP (MD5_I , c, d, a, b, w2[2], MD5C36, MD5S32); + MD5_STEP (MD5_I , b, c, d, a, w0[1], MD5C37, MD5S33); + MD5_STEP (MD5_I , a, b, c, d, w2[0], MD5C38, MD5S30); + MD5_STEP (MD5_I , d, a, b, c, w3[3], MD5C39, MD5S31); + MD5_STEP (MD5_I , c, d, a, b, w1[2], MD5C3a, MD5S32); + MD5_STEP (MD5_I , b, c, d, a, w3[1], MD5C3b, MD5S33); + MD5_STEP (MD5_I , a, b, c, d, w1[0], MD5C3c, MD5S30); + 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 r_a = a + MD5M_A; - const u32 r_b = b + MD5M_B; - const u32 r_c = c + MD5M_C; - const u32 r_d = d + MD5M_D; + const u32x r_a = a + MD5M_A; + const u32x r_b = b + MD5M_B; + const u32x r_c = c + MD5M_C; + const u32x r_d = d + MD5M_D; a = r_a; b = r_b; @@ -400,12 +378,7 @@ __kernel void m02710_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, c += r_c; d += r_d; - 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); } } @@ -501,46 +474,23 @@ __kernel void m02710_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); append_0x80_2x4 (w0, w1, out_len); w3[2] = out_len * 8; - 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); @@ -615,109 +565,110 @@ __kernel void m02710_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, c += MD5M_C; d += MD5M_D; - const u32 w0_t = uint_to_hex_lower8 ((a >> 0) & 255) << 0 - | uint_to_hex_lower8 ((a >> 8) & 255) << 16; - const u32 w1_t = uint_to_hex_lower8 ((a >> 16) & 255) << 0 - | uint_to_hex_lower8 ((a >> 24) & 255) << 16; - const u32 w2_t = uint_to_hex_lower8 ((b >> 0) & 255) << 0 - | uint_to_hex_lower8 ((b >> 8) & 255) << 16; - const u32 w3_t = uint_to_hex_lower8 ((b >> 16) & 255) << 0 - | uint_to_hex_lower8 ((b >> 24) & 255) << 16; - const u32 w4_t = uint_to_hex_lower8 ((c >> 0) & 255) << 0 - | uint_to_hex_lower8 ((c >> 8) & 255) << 16; - const u32 w5_t = uint_to_hex_lower8 ((c >> 16) & 255) << 0 - | uint_to_hex_lower8 ((c >> 24) & 255) << 16; - const u32 w6_t = uint_to_hex_lower8 ((d >> 0) & 255) << 0 - | uint_to_hex_lower8 ((d >> 8) & 255) << 16; - const u32 w7_t = uint_to_hex_lower8 ((d >> 16) & 255) << 0 - | uint_to_hex_lower8 ((d >> 24) & 255) << 16; + w0[0] = uint_to_hex_lower8 ((a >> 0) & 255) << 0 + | uint_to_hex_lower8 ((a >> 8) & 255) << 16; + w0[1] = uint_to_hex_lower8 ((a >> 16) & 255) << 0 + | uint_to_hex_lower8 ((a >> 24) & 255) << 16; + w0[2] = uint_to_hex_lower8 ((b >> 0) & 255) << 0 + | uint_to_hex_lower8 ((b >> 8) & 255) << 16; + w0[3] = uint_to_hex_lower8 ((b >> 16) & 255) << 0 + | uint_to_hex_lower8 ((b >> 24) & 255) << 16; + w1[0] = uint_to_hex_lower8 ((c >> 0) & 255) << 0 + | uint_to_hex_lower8 ((c >> 8) & 255) << 16; + w1[1] = uint_to_hex_lower8 ((c >> 16) & 255) << 0 + | uint_to_hex_lower8 ((c >> 24) & 255) << 16; + w1[2] = uint_to_hex_lower8 ((d >> 0) & 255) << 0 + | uint_to_hex_lower8 ((d >> 8) & 255) << 16; + w1[3] = uint_to_hex_lower8 ((d >> 16) & 255) << 0 + | uint_to_hex_lower8 ((d >> 24) & 255) << 16; - const u32 w8_t = s[0]; - const u32 w9_t = s[1]; - const u32 wa_t = s[2]; - const u32 wb_t = s[3]; - const u32 wc_t = s[4]; - const u32 wd_t = s[5]; - const u32 we_t = s[6]; - const u32 wf_t = s[7]; + w2[0] = s[0]; + w2[1] = s[1]; + w2[2] = s[2]; + w2[3] = s[3]; + + w3[0] = s[4]; + w3[1] = s[5]; + w3[2] = s[6]; + w3[3] = s[7]; a = MD5M_A; b = MD5M_B; c = MD5M_C; d = MD5M_D; - MD5_STEP (MD5_Fo, a, b, c, d, w0_t, MD5C00, MD5S00); - MD5_STEP (MD5_Fo, d, a, b, c, w1_t, MD5C01, MD5S01); - MD5_STEP (MD5_Fo, c, d, a, b, w2_t, MD5C02, MD5S02); - MD5_STEP (MD5_Fo, b, c, d, a, w3_t, MD5C03, MD5S03); - MD5_STEP (MD5_Fo, a, b, c, d, w4_t, MD5C04, MD5S00); - MD5_STEP (MD5_Fo, d, a, b, c, w5_t, MD5C05, MD5S01); - MD5_STEP (MD5_Fo, c, d, a, b, w6_t, MD5C06, MD5S02); - MD5_STEP (MD5_Fo, b, c, d, a, w7_t, MD5C07, MD5S03); - MD5_STEP (MD5_Fo, a, b, c, d, w8_t, MD5C08, MD5S00); - MD5_STEP (MD5_Fo, d, a, b, c, w9_t, MD5C09, MD5S01); - MD5_STEP (MD5_Fo, c, d, a, b, wa_t, MD5C0a, MD5S02); - MD5_STEP (MD5_Fo, b, c, d, a, wb_t, MD5C0b, MD5S03); - MD5_STEP (MD5_Fo, a, b, c, d, wc_t, MD5C0c, MD5S00); - MD5_STEP (MD5_Fo, d, a, b, c, wd_t, MD5C0d, MD5S01); - MD5_STEP (MD5_Fo, c, d, a, b, we_t, MD5C0e, MD5S02); - MD5_STEP (MD5_Fo, b, c, d, a, wf_t, MD5C0f, MD5S03); + MD5_STEP (MD5_Fo, a, b, c, d, w0[0], MD5C00, MD5S00); + MD5_STEP (MD5_Fo, d, a, b, c, w0[1], MD5C01, MD5S01); + MD5_STEP (MD5_Fo, c, d, a, b, w0[2], MD5C02, MD5S02); + MD5_STEP (MD5_Fo, b, c, d, a, w0[3], MD5C03, MD5S03); + MD5_STEP (MD5_Fo, a, b, c, d, w1[0], MD5C04, MD5S00); + MD5_STEP (MD5_Fo, d, a, b, c, w1[1], MD5C05, MD5S01); + MD5_STEP (MD5_Fo, c, d, a, b, w1[2], MD5C06, MD5S02); + MD5_STEP (MD5_Fo, b, c, d, a, w1[3], MD5C07, MD5S03); + MD5_STEP (MD5_Fo, a, b, c, d, w2[0], MD5C08, MD5S00); + MD5_STEP (MD5_Fo, d, a, b, c, w2[1], MD5C09, MD5S01); + MD5_STEP (MD5_Fo, c, d, a, b, w2[2], MD5C0a, MD5S02); + MD5_STEP (MD5_Fo, b, c, d, a, w2[3], MD5C0b, MD5S03); + MD5_STEP (MD5_Fo, a, b, c, d, w3[0], MD5C0c, MD5S00); + MD5_STEP (MD5_Fo, d, a, b, c, w3[1], MD5C0d, MD5S01); + MD5_STEP (MD5_Fo, c, d, a, b, w3[2], MD5C0e, MD5S02); + MD5_STEP (MD5_Fo, b, c, d, a, w3[3], MD5C0f, MD5S03); - MD5_STEP (MD5_Go, a, b, c, d, w1_t, MD5C10, MD5S10); - MD5_STEP (MD5_Go, d, a, b, c, w6_t, MD5C11, MD5S11); - MD5_STEP (MD5_Go, c, d, a, b, wb_t, MD5C12, MD5S12); - MD5_STEP (MD5_Go, b, c, d, a, w0_t, MD5C13, MD5S13); - MD5_STEP (MD5_Go, a, b, c, d, w5_t, MD5C14, MD5S10); - MD5_STEP (MD5_Go, d, a, b, c, wa_t, MD5C15, MD5S11); - MD5_STEP (MD5_Go, c, d, a, b, wf_t, MD5C16, MD5S12); - MD5_STEP (MD5_Go, b, c, d, a, w4_t, MD5C17, MD5S13); - MD5_STEP (MD5_Go, a, b, c, d, w9_t, MD5C18, MD5S10); - MD5_STEP (MD5_Go, d, a, b, c, we_t, MD5C19, MD5S11); - MD5_STEP (MD5_Go, c, d, a, b, w3_t, MD5C1a, MD5S12); - MD5_STEP (MD5_Go, b, c, d, a, w8_t, MD5C1b, MD5S13); - MD5_STEP (MD5_Go, a, b, c, d, wd_t, MD5C1c, MD5S10); - MD5_STEP (MD5_Go, d, a, b, c, w2_t, MD5C1d, MD5S11); - MD5_STEP (MD5_Go, c, d, a, b, w7_t, MD5C1e, MD5S12); - MD5_STEP (MD5_Go, b, c, d, a, wc_t, MD5C1f, MD5S13); + MD5_STEP (MD5_Go, a, b, c, d, w0[1], MD5C10, MD5S10); + MD5_STEP (MD5_Go, d, a, b, c, w1[2], MD5C11, MD5S11); + MD5_STEP (MD5_Go, c, d, a, b, w2[3], MD5C12, MD5S12); + MD5_STEP (MD5_Go, b, c, d, a, w0[0], MD5C13, MD5S13); + MD5_STEP (MD5_Go, a, b, c, d, w1[1], MD5C14, MD5S10); + MD5_STEP (MD5_Go, d, a, b, c, w2[2], MD5C15, MD5S11); + MD5_STEP (MD5_Go, c, d, a, b, w3[3], MD5C16, MD5S12); + MD5_STEP (MD5_Go, b, c, d, a, w1[0], MD5C17, MD5S13); + MD5_STEP (MD5_Go, a, b, c, d, w2[1], MD5C18, MD5S10); + MD5_STEP (MD5_Go, d, a, b, c, w3[2], MD5C19, MD5S11); + MD5_STEP (MD5_Go, c, d, a, b, w0[3], MD5C1a, MD5S12); + MD5_STEP (MD5_Go, b, c, d, a, w2[0], MD5C1b, MD5S13); + MD5_STEP (MD5_Go, a, b, c, d, w3[1], MD5C1c, MD5S10); + MD5_STEP (MD5_Go, d, a, b, c, w0[2], MD5C1d, MD5S11); + MD5_STEP (MD5_Go, c, d, a, b, w1[3], MD5C1e, MD5S12); + MD5_STEP (MD5_Go, b, c, d, a, w3[0], MD5C1f, MD5S13); - MD5_STEP (MD5_H , a, b, c, d, w5_t, MD5C20, MD5S20); - MD5_STEP (MD5_H , d, a, b, c, w8_t, MD5C21, MD5S21); - MD5_STEP (MD5_H , c, d, a, b, wb_t, MD5C22, MD5S22); - MD5_STEP (MD5_H , b, c, d, a, we_t, MD5C23, MD5S23); - MD5_STEP (MD5_H , a, b, c, d, w1_t, MD5C24, MD5S20); - MD5_STEP (MD5_H , d, a, b, c, w4_t, MD5C25, MD5S21); - MD5_STEP (MD5_H , c, d, a, b, w7_t, MD5C26, MD5S22); - MD5_STEP (MD5_H , b, c, d, a, wa_t, MD5C27, MD5S23); - MD5_STEP (MD5_H , a, b, c, d, wd_t, MD5C28, MD5S20); - MD5_STEP (MD5_H , d, a, b, c, w0_t, MD5C29, MD5S21); - MD5_STEP (MD5_H , c, d, a, b, w3_t, MD5C2a, MD5S22); - MD5_STEP (MD5_H , b, c, d, a, w6_t, MD5C2b, MD5S23); - MD5_STEP (MD5_H , a, b, c, d, w9_t, MD5C2c, MD5S20); - MD5_STEP (MD5_H , d, a, b, c, wc_t, MD5C2d, MD5S21); - MD5_STEP (MD5_H , c, d, a, b, wf_t, MD5C2e, MD5S22); - MD5_STEP (MD5_H , b, c, d, a, w2_t, MD5C2f, MD5S23); + MD5_STEP (MD5_H , a, b, c, d, w1[1], MD5C20, MD5S20); + MD5_STEP (MD5_H , d, a, b, c, w2[0], MD5C21, MD5S21); + MD5_STEP (MD5_H , c, d, a, b, w2[3], MD5C22, MD5S22); + MD5_STEP (MD5_H , b, c, d, a, w3[2], MD5C23, MD5S23); + MD5_STEP (MD5_H , a, b, c, d, w0[1], MD5C24, MD5S20); + MD5_STEP (MD5_H , d, a, b, c, w1[0], MD5C25, MD5S21); + MD5_STEP (MD5_H , c, d, a, b, w1[3], MD5C26, MD5S22); + MD5_STEP (MD5_H , b, c, d, a, w2[2], MD5C27, MD5S23); + MD5_STEP (MD5_H , a, b, c, d, w3[1], MD5C28, MD5S20); + MD5_STEP (MD5_H , d, a, b, c, w0[0], MD5C29, MD5S21); + MD5_STEP (MD5_H , c, d, a, b, w0[3], MD5C2a, MD5S22); + MD5_STEP (MD5_H , b, c, d, a, w1[2], MD5C2b, MD5S23); + MD5_STEP (MD5_H , a, b, c, d, w2[1], MD5C2c, MD5S20); + MD5_STEP (MD5_H , d, a, b, c, w3[0], MD5C2d, MD5S21); + MD5_STEP (MD5_H , c, d, a, b, w3[3], MD5C2e, MD5S22); + MD5_STEP (MD5_H , b, c, d, a, w0[2], MD5C2f, MD5S23); - MD5_STEP (MD5_I , a, b, c, d, w0_t, MD5C30, MD5S30); - MD5_STEP (MD5_I , d, a, b, c, w7_t, MD5C31, MD5S31); - MD5_STEP (MD5_I , c, d, a, b, we_t, MD5C32, MD5S32); - MD5_STEP (MD5_I , b, c, d, a, w5_t, MD5C33, MD5S33); - MD5_STEP (MD5_I , a, b, c, d, wc_t, MD5C34, MD5S30); - MD5_STEP (MD5_I , d, a, b, c, w3_t, MD5C35, MD5S31); - MD5_STEP (MD5_I , c, d, a, b, wa_t, MD5C36, MD5S32); - MD5_STEP (MD5_I , b, c, d, a, w1_t, MD5C37, MD5S33); - MD5_STEP (MD5_I , a, b, c, d, w8_t, MD5C38, MD5S30); - MD5_STEP (MD5_I , d, a, b, c, wf_t, MD5C39, MD5S31); - MD5_STEP (MD5_I , c, d, a, b, w6_t, MD5C3a, MD5S32); - MD5_STEP (MD5_I , b, c, d, a, wd_t, MD5C3b, MD5S33); - MD5_STEP (MD5_I , a, b, c, d, w4_t, MD5C3c, MD5S30); - MD5_STEP (MD5_I , d, a, b, c, wb_t, MD5C3d, MD5S31); - MD5_STEP (MD5_I , c, d, a, b, w2_t, MD5C3e, MD5S32); - MD5_STEP (MD5_I , b, c, d, a, w9_t, MD5C3f, MD5S33); + MD5_STEP (MD5_I , a, b, c, d, w0[0], MD5C30, MD5S30); + MD5_STEP (MD5_I , d, a, b, c, w1[3], MD5C31, MD5S31); + MD5_STEP (MD5_I , c, d, a, b, w3[2], MD5C32, MD5S32); + MD5_STEP (MD5_I , b, c, d, a, w1[1], MD5C33, MD5S33); + MD5_STEP (MD5_I , a, b, c, d, w3[0], MD5C34, MD5S30); + MD5_STEP (MD5_I , d, a, b, c, w0[3], MD5C35, MD5S31); + MD5_STEP (MD5_I , c, d, a, b, w2[2], MD5C36, MD5S32); + MD5_STEP (MD5_I , b, c, d, a, w0[1], MD5C37, MD5S33); + MD5_STEP (MD5_I , a, b, c, d, w2[0], MD5C38, MD5S30); + MD5_STEP (MD5_I , d, a, b, c, w3[3], MD5C39, MD5S31); + MD5_STEP (MD5_I , c, d, a, b, w1[2], MD5C3a, MD5S32); + MD5_STEP (MD5_I , b, c, d, a, w3[1], MD5C3b, MD5S33); + MD5_STEP (MD5_I , a, b, c, d, w1[0], MD5C3c, MD5S30); + 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 r_a = a + MD5M_A; - const u32 r_b = b + MD5M_B; - const u32 r_c = c + MD5M_C; - const u32 r_d = d + MD5M_D; + const u32x r_a = a + MD5M_A; + const u32x r_b = b + MD5M_B; + const u32x r_c = c + MD5M_C; + const u32x r_d = d + MD5M_D; a = r_a; b = r_b; @@ -788,9 +739,6 @@ __kernel void m02710_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, MD5_STEP0(MD5_I , c, d, a, b, MD5C3a, MD5S32); MD5_STEP0(MD5_I , b, c, d, a, MD5C3b, MD5S33); MD5_STEP0(MD5_I , a, b, c, d, MD5C3c, MD5S30); - - if (allx ((a + r_a) != search[0])) continue; - MD5_STEP0(MD5_I , d, a, b, c, MD5C3d, MD5S31); MD5_STEP0(MD5_I , c, d, a, b, MD5C3e, MD5S32); MD5_STEP0(MD5_I , b, c, d, a, MD5C3f, MD5S33); @@ -800,12 +748,7 @@ __kernel void m02710_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, c += r_c; d += r_d; - 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); } }