From fbd46f61be18239e81b2628b7bf30e7be1bf0859 Mon Sep 17 00:00:00 2001 From: Jens Steube Date: Fri, 5 Feb 2016 18:01:37 +0100 Subject: [PATCH] Converted to new SIMD: -m 1760 -a 0 --- OpenCL/m01760_a0.cl | 260 ++++++++++++++++++-------------------------- 1 file changed, 106 insertions(+), 154 deletions(-) diff --git a/OpenCL/m01760_a0.cl b/OpenCL/m01760_a0.cl index cf23b3e92..4c441292d 100644 --- a/OpenCL/m01760_a0.cl +++ b/OpenCL/m01760_a0.cl @@ -5,6 +5,8 @@ #define _SHA512_ +#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" __constant u64 k_sha512[80] = { @@ -46,33 +46,33 @@ __constant u64 k_sha512[80] = SHA512C4c, SHA512C4d, SHA512C4e, SHA512C4f, }; -static void sha512_transform (const u64 w0[4], const u64 w1[4], const u64 w2[4], const u64 w3[4], u64 digest[8]) +static void sha512_transform (const u64x w0[4], const u64x w1[4], const u64x w2[4], const u64x w3[4], u64x digest[8]) { - u64 w0_t = w0[0]; - u64 w1_t = w0[1]; - u64 w2_t = w0[2]; - u64 w3_t = w0[3]; - u64 w4_t = w1[0]; - u64 w5_t = w1[1]; - u64 w6_t = w1[2]; - u64 w7_t = w1[3]; - u64 w8_t = w2[0]; - u64 w9_t = w2[1]; - u64 wa_t = w2[2]; - u64 wb_t = w2[3]; - u64 wc_t = w3[0]; - u64 wd_t = w3[1]; - u64 we_t = w3[2]; - u64 wf_t = w3[3]; + u64x w0_t = w0[0]; + u64x w1_t = w0[1]; + u64x w2_t = w0[2]; + u64x w3_t = w0[3]; + u64x w4_t = w1[0]; + u64x w5_t = w1[1]; + u64x w6_t = w1[2]; + u64x w7_t = w1[3]; + u64x w8_t = w2[0]; + u64x w9_t = w2[1]; + u64x wa_t = w2[2]; + u64x wb_t = w2[3]; + u64x wc_t = w3[0]; + u64x wd_t = w3[1]; + u64x we_t = w3[2]; + u64x wf_t = w3[3]; - u64 a = digest[0]; - u64 b = digest[1]; - u64 c = digest[2]; - u64 d = digest[3]; - u64 e = digest[4]; - u64 f = digest[5]; - u64 g = digest[6]; - u64 h = digest[7]; + u64x a = digest[0]; + u64x b = digest[1]; + u64x c = digest[2]; + u64x d = digest[3]; + u64x e = digest[4]; + u64x f = digest[5]; + u64x g = digest[6]; + u64x h = digest[7]; #define ROUND_EXPAND() \ { \ @@ -132,29 +132,29 @@ static void sha512_transform (const u64 w0[4], const u64 w1[4], const u64 w2[4], digest[7] += h; } -static void hmac_sha512_pad (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u64 ipad[8], u64 opad[8]) +static void hmac_sha512_pad (u32x w0[4], u32x w1[4], u32x w2[4], u32x w3[4], u64x ipad[8], u64x opad[8]) { - u64 w0_t[4]; - u64 w1_t[4]; - u64 w2_t[4]; - u64 w3_t[4]; + u64x w0_t[4]; + u64x w1_t[4]; + u64x w2_t[4]; + u64x w3_t[4]; - w0_t[0] = hl32_to_64 (w0[0], w0[1]) ^ 0x3636363636363636; - w0_t[1] = hl32_to_64 (w0[2], w0[3]) ^ 0x3636363636363636; - w0_t[2] = hl32_to_64 (w1[0], w1[1]) ^ 0x3636363636363636; - w0_t[3] = hl32_to_64 (w1[2], w1[3]) ^ 0x3636363636363636; - w1_t[0] = hl32_to_64 (w2[0], w2[1]) ^ 0x3636363636363636; - w1_t[1] = hl32_to_64 (w2[2], w2[3]) ^ 0x3636363636363636; - w1_t[2] = hl32_to_64 (w3[0], w3[1]) ^ 0x3636363636363636; - w1_t[3] = hl32_to_64 (w3[2], w3[3]) ^ 0x3636363636363636; - w2_t[0] = 0 ^ 0x3636363636363636; - w2_t[1] = 0 ^ 0x3636363636363636; - w2_t[2] = 0 ^ 0x3636363636363636; - w2_t[3] = 0 ^ 0x3636363636363636; - w3_t[0] = 0 ^ 0x3636363636363636; - w3_t[1] = 0 ^ 0x3636363636363636; - w3_t[2] = 0 ^ 0x3636363636363636; - w3_t[3] = 0 ^ 0x3636363636363636; + w0_t[0] = hl32_to_64 (w0[0], w0[1]) ^ (u64x) 0x3636363636363636; + w0_t[1] = hl32_to_64 (w0[2], w0[3]) ^ (u64x) 0x3636363636363636; + w0_t[2] = hl32_to_64 (w1[0], w1[1]) ^ (u64x) 0x3636363636363636; + w0_t[3] = hl32_to_64 (w1[2], w1[3]) ^ (u64x) 0x3636363636363636; + w1_t[0] = hl32_to_64 (w2[0], w2[1]) ^ (u64x) 0x3636363636363636; + w1_t[1] = hl32_to_64 (w2[2], w2[3]) ^ (u64x) 0x3636363636363636; + w1_t[2] = hl32_to_64 (w3[0], w3[1]) ^ (u64x) 0x3636363636363636; + w1_t[3] = hl32_to_64 (w3[2], w3[3]) ^ (u64x) 0x3636363636363636; + w2_t[0] = (u64x) 0x3636363636363636; + w2_t[1] = (u64x) 0x3636363636363636; + w2_t[2] = (u64x) 0x3636363636363636; + w2_t[3] = (u64x) 0x3636363636363636; + w3_t[0] = (u64x) 0x3636363636363636; + w3_t[1] = (u64x) 0x3636363636363636; + w3_t[2] = (u64x) 0x3636363636363636; + w3_t[3] = (u64x) 0x3636363636363636; ipad[0] = SHA512M_A; ipad[1] = SHA512M_B; @@ -167,22 +167,22 @@ static void hmac_sha512_pad (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u64 ipa sha512_transform (w0_t, w1_t, w2_t, w3_t, ipad); - w0_t[0] = hl32_to_64 (w0[0], w0[1]) ^ 0x5c5c5c5c5c5c5c5c; - w0_t[1] = hl32_to_64 (w0[2], w0[3]) ^ 0x5c5c5c5c5c5c5c5c; - w0_t[2] = hl32_to_64 (w1[0], w1[1]) ^ 0x5c5c5c5c5c5c5c5c; - w0_t[3] = hl32_to_64 (w1[2], w1[3]) ^ 0x5c5c5c5c5c5c5c5c; - w1_t[0] = hl32_to_64 (w2[0], w2[1]) ^ 0x5c5c5c5c5c5c5c5c; - w1_t[1] = hl32_to_64 (w2[2], w2[3]) ^ 0x5c5c5c5c5c5c5c5c; - w1_t[2] = hl32_to_64 (w3[0], w3[1]) ^ 0x5c5c5c5c5c5c5c5c; - w1_t[3] = hl32_to_64 (w3[2], w3[3]) ^ 0x5c5c5c5c5c5c5c5c; - w2_t[0] = 0 ^ 0x5c5c5c5c5c5c5c5c; - w2_t[1] = 0 ^ 0x5c5c5c5c5c5c5c5c; - w2_t[2] = 0 ^ 0x5c5c5c5c5c5c5c5c; - w2_t[3] = 0 ^ 0x5c5c5c5c5c5c5c5c; - w3_t[0] = 0 ^ 0x5c5c5c5c5c5c5c5c; - w3_t[1] = 0 ^ 0x5c5c5c5c5c5c5c5c; - w3_t[2] = 0 ^ 0x5c5c5c5c5c5c5c5c; - w3_t[3] = 0 ^ 0x5c5c5c5c5c5c5c5c; + w0_t[0] = hl32_to_64 (w0[0], w0[1]) ^ (u64x) 0x5c5c5c5c5c5c5c5c; + w0_t[1] = hl32_to_64 (w0[2], w0[3]) ^ (u64x) 0x5c5c5c5c5c5c5c5c; + w0_t[2] = hl32_to_64 (w1[0], w1[1]) ^ (u64x) 0x5c5c5c5c5c5c5c5c; + w0_t[3] = hl32_to_64 (w1[2], w1[3]) ^ (u64x) 0x5c5c5c5c5c5c5c5c; + w1_t[0] = hl32_to_64 (w2[0], w2[1]) ^ (u64x) 0x5c5c5c5c5c5c5c5c; + w1_t[1] = hl32_to_64 (w2[2], w2[3]) ^ (u64x) 0x5c5c5c5c5c5c5c5c; + w1_t[2] = hl32_to_64 (w3[0], w3[1]) ^ (u64x) 0x5c5c5c5c5c5c5c5c; + w1_t[3] = hl32_to_64 (w3[2], w3[3]) ^ (u64x) 0x5c5c5c5c5c5c5c5c; + w2_t[0] = (u64x) 0x5c5c5c5c5c5c5c5c; + w2_t[1] = (u64x) 0x5c5c5c5c5c5c5c5c; + w2_t[2] = (u64x) 0x5c5c5c5c5c5c5c5c; + w2_t[3] = (u64x) 0x5c5c5c5c5c5c5c5c; + w3_t[0] = (u64x) 0x5c5c5c5c5c5c5c5c; + w3_t[1] = (u64x) 0x5c5c5c5c5c5c5c5c; + w3_t[2] = (u64x) 0x5c5c5c5c5c5c5c5c; + w3_t[3] = (u64x) 0x5c5c5c5c5c5c5c5c; opad[0] = SHA512M_A; opad[1] = SHA512M_B; @@ -196,12 +196,12 @@ static void hmac_sha512_pad (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u64 ipa sha512_transform (w0_t, w1_t, w2_t, w3_t, opad); } -static void hmac_sha512_run (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u64 ipad[8], u64 opad[8], u64 digest[8]) +static void hmac_sha512_run (u32x w0[4], u32x w1[4], u32x w2[4], u32x w3[4], u64x ipad[8], u64x opad[8], u64x digest[8]) { - u64 w0_t[4]; - u64 w1_t[4]; - u64 w2_t[4]; - u64 w3_t[4]; + u64x w0_t[4]; + u64x w1_t[4]; + u64x w2_t[4]; + u64x w3_t[4]; w0_t[0] = hl32_to_64 (w0[0], w0[1]); w0_t[1] = hl32_to_64 (w0[2], w0[3]); @@ -314,36 +314,36 @@ __kernel void m01760_m04 (__global pw_t *pws, __global kernel_rule_t * rules_bu * pads */ - u32 w0_t[4]; + u32x w0_t[4]; w0_t[0] = swap32 (salt_buf0[0]); w0_t[1] = swap32 (salt_buf0[1]); w0_t[2] = swap32 (salt_buf0[2]); w0_t[3] = swap32 (salt_buf0[3]); - u32 w1_t[4]; + u32x w1_t[4]; w1_t[0] = swap32 (salt_buf1[0]); w1_t[1] = swap32 (salt_buf1[1]); w1_t[2] = swap32 (salt_buf1[2]); w1_t[3] = swap32 (salt_buf1[3]); - u32 w2_t[4]; + u32x w2_t[4]; w2_t[0] = 0; w2_t[1] = 0; w2_t[2] = 0; w2_t[3] = 0; - u32 w3_t[4]; + u32x w3_t[4]; w3_t[0] = 0; w3_t[1] = 0; w3_t[2] = 0; w3_t[3] = 0; - u64 ipad[8]; - u64 opad[8]; + u64x ipad[8]; + u64x opad[8]; hmac_sha512_pad (w0_t, w1_t, w2_t, w3_t, ipad, opad); @@ -351,37 +351,14 @@ __kernel void m01760_m04 (__global pw_t *pws, __global kernel_rule_t * rules_bu * 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); @@ -402,17 +379,16 @@ __kernel void m01760_m04 (__global pw_t *pws, __global kernel_rule_t * rules_bu w3_t[2] = 0; w3_t[3] = (128 + out_len) * 8; - u64 digest[8]; + u64x digest[8]; hmac_sha512_run (w0_t, w1_t, w2_t, w3_t, ipad, opad, digest); + const u32x r0 = l32_from_64 (digest[7]); + const u32x r1 = h32_from_64 (digest[7]); + const u32x r2 = l32_from_64 (digest[3]); + const u32x r3 = h32_from_64 (digest[3]); - const u32 r0 = l32_from_64 (digest[7]); - const u32 r1 = h32_from_64 (digest[7]); - const u32 r2 = l32_from_64 (digest[3]); - const u32 r3 = h32_from_64 (digest[3]); - - #include COMPARE_M + COMPARE_M_SIMD (r0, r1, r2, r3); } } @@ -478,36 +454,36 @@ __kernel void m01760_s04 (__global pw_t *pws, __global kernel_rule_t * rules_bu * pads */ - u32 w0_t[4]; + u32x w0_t[4]; w0_t[0] = swap32 (salt_buf0[0]); w0_t[1] = swap32 (salt_buf0[1]); w0_t[2] = swap32 (salt_buf0[2]); w0_t[3] = swap32 (salt_buf0[3]); - u32 w1_t[4]; + u32x w1_t[4]; w1_t[0] = swap32 (salt_buf1[0]); w1_t[1] = swap32 (salt_buf1[1]); w1_t[2] = swap32 (salt_buf1[2]); w1_t[3] = swap32 (salt_buf1[3]); - u32 w2_t[4]; + u32x w2_t[4]; w2_t[0] = 0; w2_t[1] = 0; w2_t[2] = 0; w2_t[3] = 0; - u32 w3_t[4]; + u32x w3_t[4]; w3_t[0] = 0; w3_t[1] = 0; w3_t[2] = 0; w3_t[3] = 0; - u64 ipad[8]; - u64 opad[8]; + u64x ipad[8]; + u64x opad[8]; hmac_sha512_pad (w0_t, w1_t, w2_t, w3_t, ipad, opad); @@ -527,37 +503,14 @@ __kernel void m01760_s04 (__global pw_t *pws, __global kernel_rule_t * rules_bu * 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); @@ -578,17 +531,16 @@ __kernel void m01760_s04 (__global pw_t *pws, __global kernel_rule_t * rules_bu w3_t[2] = 0; w3_t[3] = (128 + out_len) * 8; - u64 digest[8]; + u64x digest[8]; hmac_sha512_run (w0_t, w1_t, w2_t, w3_t, ipad, opad, digest); + const u32x r0 = l32_from_64 (digest[7]); + const u32x r1 = h32_from_64 (digest[7]); + const u32x r2 = l32_from_64 (digest[3]); + const u32x r3 = h32_from_64 (digest[3]); - const u32 r0 = l32_from_64 (digest[7]); - const u32 r1 = h32_from_64 (digest[7]); - const u32 r2 = l32_from_64 (digest[3]); - const u32 r3 = h32_from_64 (digest[3]); - - #include COMPARE_S + COMPARE_S_SIMD (r0, r1, r2, r3); } }