From: Jens Steube Date: Sat, 6 Feb 2016 09:23:42 +0000 (+0100) Subject: Converted to new SIMD: -m 5400 -a 0 X-Git-Tag: v3.00-beta~334 X-Git-Url: https://www.flypig.org.uk/git/?a=commitdiff_plain;h=b3a43b47a12099b7534a627f40ede5953b32c29f;p=hashcat.git Converted to new SIMD: -m 5400 -a 0 --- diff --git a/OpenCL/m05400_a0.cl b/OpenCL/m05400_a0.cl index 659d04b..bc262ee 100644 --- a/OpenCL/m05400_a0.cl +++ b/OpenCL/m05400_a0.cl @@ -5,6 +5,8 @@ #define _SHA1_ +#define NEW_SIMD_CODE + #include "include/constants.h" #include "include/kernel_vendor.h" @@ -18,34 +20,32 @@ #include "OpenCL/common.c" #include "include/rp_kernel.h" #include "OpenCL/rp.c" +#include "OpenCL/simd.c" -#define COMPARE_S "OpenCL/check_single_comp4.c" -#define COMPARE_M "OpenCL/check_multi_comp4.c" - -static void sha1_transform (const u32 w0[4], const u32 w1[4], const u32 w2[4], const u32 w3[4], u32 digest[5]) +static void sha1_transform (const u32x w0[4], const u32x w1[4], const u32x w2[4], const u32x w3[4], u32x digest[5]) { - u32 A = digest[0]; - u32 B = digest[1]; - u32 C = digest[2]; - u32 D = digest[3]; - u32 E = digest[4]; - - u32 w0_t = w0[0]; - u32 w1_t = w0[1]; - u32 w2_t = w0[2]; - u32 w3_t = w0[3]; - u32 w4_t = w1[0]; - u32 w5_t = w1[1]; - u32 w6_t = w1[2]; - u32 w7_t = w1[3]; - u32 w8_t = w2[0]; - u32 w9_t = w2[1]; - u32 wa_t = w2[2]; - u32 wb_t = w2[3]; - u32 wc_t = w3[0]; - u32 wd_t = w3[1]; - u32 we_t = w3[2]; - u32 wf_t = w3[3]; + u32x A = digest[0]; + u32x B = digest[1]; + u32x C = digest[2]; + u32x D = digest[3]; + u32x E = digest[4]; + + u32x w0_t = w0[0]; + u32x w1_t = w0[1]; + u32x w2_t = w0[2]; + u32x w3_t = w0[3]; + u32x w4_t = w1[0]; + u32x w5_t = w1[1]; + u32x w6_t = w1[2]; + u32x w7_t = w1[3]; + u32x w8_t = w2[0]; + u32x w9_t = w2[1]; + u32x wa_t = w2[2]; + u32x wb_t = w2[3]; + u32x wc_t = w3[0]; + u32x wd_t = w3[1]; + u32x we_t = w3[2]; + u32x wf_t = w3[3]; #undef K #define K SHA1C00 @@ -150,7 +150,7 @@ static void sha1_transform (const u32 w0[4], const u32 w1[4], const u32 w2[4], c digest[4] += E; } -static void hmac_sha1_pad (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u32 ipad[5], u32 opad[5]) +static void hmac_sha1_pad (u32x w0[4], u32x w1[4], u32x w2[4], u32x w3[4], u32x ipad[5], u32x opad[5]) { w0[0] = w0[0] ^ 0x36363636; w0[1] = w0[1] ^ 0x36363636; @@ -203,7 +203,7 @@ static void hmac_sha1_pad (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u32 ipad[ sha1_transform (w0, w1, w2, w3, opad); } -static void hmac_sha1_run (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u32 ipad[5], u32 opad[5], u32 digest[5]) +static void hmac_sha1_run (u32x w0[4], u32x w1[4], u32x w2[4], u32x w3[4], u32x ipad[5], u32x opad[5], u32x digest[5]) { digest[0] = ipad[0]; digest[1] = ipad[1]; @@ -257,7 +257,7 @@ __kernel void m05400_m04 (__global pw_t *pws, __global kernel_rule_t * rules_bu for (u32 i = lid; i < 16; i += lsz) { - w_s[i] = swap32 (ikepsk_bufs[salt_pos].nr_buf[i]); + w_s[i] = swap32_S (ikepsk_bufs[salt_pos].nr_buf[i]); } barrier (CLK_LOCAL_MEM_FENCE); @@ -266,7 +266,7 @@ __kernel void m05400_m04 (__global pw_t *pws, __global kernel_rule_t * rules_bu for (u32 i = lid; i < 128; i += lsz) { - s_msg_buf[i] = swap32 (ikepsk_bufs[salt_pos].msg_buf[i]); + s_msg_buf[i] = swap32_S (ikepsk_bufs[salt_pos].msg_buf[i]); } barrier (CLK_LOCAL_MEM_FENCE); @@ -304,72 +304,49 @@ __kernel void m05400_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]; - - 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; + u32x w0[4] = { 0 }; + u32x w1[4] = { 0 }; + u32x w2[4] = { 0 }; + u32x w3[4] = { 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); /** * pads */ - u32 w0_t[4]; + u32x w0_t[4]; w0_t[0] = swap32 (w0[0]); w0_t[1] = swap32 (w0[1]); w0_t[2] = swap32 (w0[2]); w0_t[3] = swap32 (w0[3]); - u32 w1_t[4]; + u32x w1_t[4]; w1_t[0] = swap32 (w1[0]); w1_t[1] = swap32 (w1[1]); w1_t[2] = swap32 (w1[2]); w1_t[3] = swap32 (w1[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; - u32 ipad[5]; - u32 opad[5]; + u32x ipad[5]; + u32x opad[5]; hmac_sha1_pad (w0_t, w1_t, w2_t, w3_t, ipad, opad); @@ -390,7 +367,7 @@ __kernel void m05400_m04 (__global pw_t *pws, __global kernel_rule_t * rules_bu w3_t[2] = 0; w3_t[3] = (64 + nr_len) * 8; - u32 digest[5]; + u32x digest[5]; hmac_sha1_run (w0_t, w1_t, w2_t, w3_t, ipad, opad, digest); @@ -457,12 +434,7 @@ __kernel void m05400_m04 (__global pw_t *pws, __global kernel_rule_t * rules_bu hmac_sha1_run (w0_t, w1_t, w2_t, w3_t, ipad, opad, digest); - const u32 r0 = digest[3]; - const u32 r1 = digest[4]; - const u32 r2 = digest[2]; - const u32 r3 = digest[1]; - - #include COMPARE_M + COMPARE_M_SIMD (digest[3], digest[4], digest[2], digest[1]); } } @@ -492,7 +464,7 @@ __kernel void m05400_s04 (__global pw_t *pws, __global kernel_rule_t * rules_bu for (u32 i = lid; i < 16; i += lsz) { - w_s[i] = swap32 (ikepsk_bufs[salt_pos].nr_buf[i]); + w_s[i] = swap32_S (ikepsk_bufs[salt_pos].nr_buf[i]); } barrier (CLK_LOCAL_MEM_FENCE); @@ -501,7 +473,7 @@ __kernel void m05400_s04 (__global pw_t *pws, __global kernel_rule_t * rules_bu for (u32 i = lid; i < 128; i += lsz) { - s_msg_buf[i] = swap32 (ikepsk_bufs[salt_pos].msg_buf[i]); + s_msg_buf[i] = swap32_S (ikepsk_bufs[salt_pos].msg_buf[i]); } barrier (CLK_LOCAL_MEM_FENCE); @@ -551,72 +523,49 @@ __kernel void m05400_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]; - - 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]; + u32x w0[4] = { 0 }; + u32x w1[4] = { 0 }; + u32x w2[4] = { 0 }; + u32x w3[4] = { 0 }; - 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); /** * pads */ - u32 w0_t[4]; + u32x w0_t[4]; w0_t[0] = swap32 (w0[0]); w0_t[1] = swap32 (w0[1]); w0_t[2] = swap32 (w0[2]); w0_t[3] = swap32 (w0[3]); - u32 w1_t[4]; + u32x w1_t[4]; w1_t[0] = swap32 (w1[0]); w1_t[1] = swap32 (w1[1]); w1_t[2] = swap32 (w1[2]); w1_t[3] = swap32 (w1[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; - u32 ipad[5]; - u32 opad[5]; + u32x ipad[5]; + u32x opad[5]; hmac_sha1_pad (w0_t, w1_t, w2_t, w3_t, ipad, opad); @@ -637,7 +586,7 @@ __kernel void m05400_s04 (__global pw_t *pws, __global kernel_rule_t * rules_bu w3_t[2] = 0; w3_t[3] = (64 + nr_len) * 8; - u32 digest[5]; + u32x digest[5]; hmac_sha1_run (w0_t, w1_t, w2_t, w3_t, ipad, opad, digest); @@ -704,12 +653,7 @@ __kernel void m05400_s04 (__global pw_t *pws, __global kernel_rule_t * rules_bu hmac_sha1_run (w0_t, w1_t, w2_t, w3_t, ipad, opad, digest); - const u32 r0 = digest[3]; - const u32 r1 = digest[4]; - const u32 r2 = digest[2]; - const u32 r3 = digest[1]; - - #include COMPARE_S + COMPARE_S_SIMD (digest[3], digest[4], digest[2], digest[1]); } }