diff --git a/OpenCL/m12600_a0.cl b/OpenCL/m12600_a0.cl index 22d2f51f4..2d16d83e9 100644 --- a/OpenCL/m12600_a0.cl +++ b/OpenCL/m12600_a0.cl @@ -5,6 +5,8 @@ #define _SHA256_SHA1_ +#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_upper8(i) (u32x) (l_bin2asc[(i)]) @@ -100,37 +100,14 @@ __kernel void m12600_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); @@ -138,31 +115,31 @@ __kernel void m12600_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, * sha1 */ - 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 = out_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 = out_len * 8; - u32 a = SHA1M_A; - u32 b = SHA1M_B; - u32 c = SHA1M_C; - u32 d = SHA1M_D; - u32 e = SHA1M_E; - u32 f = 0; - u32 g = 0; - u32 h = 0; + u32x a = SHA1M_A; + u32x b = SHA1M_B; + u32x c = SHA1M_C; + u32x d = SHA1M_D; + u32x e = SHA1M_E; + u32x f = 0; + u32x g = 0; + u32x h = 0; #undef K #define K SHA1C00 @@ -385,12 +362,7 @@ __kernel void m12600_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, 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); - const u32 r0 = d; - const u32 r1 = h; - const u32 r2 = c; - const u32 r3 = g; - - #include COMPARE_M + COMPARE_M_SIMD (d, h, c, g); } } @@ -482,37 +454,14 @@ __kernel void m12600_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); @@ -520,31 +469,31 @@ __kernel void m12600_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, * sha1 */ - 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 = out_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 = out_len * 8; - u32 a = SHA1M_A; - u32 b = SHA1M_B; - u32 c = SHA1M_C; - u32 d = SHA1M_D; - u32 e = SHA1M_E; - u32 f = 0; - u32 g = 0; - u32 h = 0; + u32x a = SHA1M_A; + u32x b = SHA1M_B; + u32x c = SHA1M_C; + u32x d = SHA1M_D; + u32x e = SHA1M_E; + u32x f = 0; + u32x g = 0; + u32x h = 0; #undef K #define K SHA1C00 @@ -648,7 +597,6 @@ __kernel void m12600_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, d += SHA1M_D; e += SHA1M_E; - /** * sha256 */ @@ -768,12 +716,7 @@ __kernel void m12600_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, 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); - const u32 r0 = d; - const u32 r1 = h; - const u32 r2 = c; - const u32 r3 = g; - - #include COMPARE_S + COMPARE_S_SIMD (d, h, c, g); } }