#define _SHA1_
+#define NEW_SIMD_CODE
+
#include "include/constants.h"
#include "include/kernel_vendor.h"
#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"
// no unicode yet
* 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);
/**
* append salt
*/
- u32 s0[4];
+ u32x s0[4];
s0[0] = salt_buf0[0];
s0[1] = salt_buf0[1];
s0[2] = salt_buf0[2];
s0[3] = salt_buf0[3];
- u32 s1[4];
+ u32x s1[4];
s1[0] = salt_buf1[0];
s1[1] = salt_buf1[1];
s1[2] = salt_buf1[2];
s1[3] = salt_buf1[3];
- u32 s2[4];
+ u32x s2[4];
s2[0] = 0;
s2[1] = 0;
s2[2] = 0;
s2[3] = 0;
- u32 s3[4];
+ u32x s3[4];
s3[0] = 0;
s3[1] = 0;
const u32 out_salt_len = (out_len * 2) + salt_len;
- u32 w0_t[4];
- u32 w1_t[4];
- u32 w2_t[4];
- u32 w3_t[4];
+ u32x w0_t[4];
+ u32x w1_t[4];
+ u32x w2_t[4];
+ u32x w3_t[4];
make_unicode (w0, w0_t, w1_t);
make_unicode (w1, w2_t, w3_t);
//w3_t[2] = swap32 (w3_t[2]);
//w3_t[3] = swap32 (w3_t[3]);
- u32 a = SHA1M_A;
- u32 b = SHA1M_B;
- u32 c = SHA1M_C;
- u32 d = SHA1M_D;
- u32 e = SHA1M_E;
+ u32x a = SHA1M_A;
+ u32x b = SHA1M_B;
+ u32x c = SHA1M_C;
+ u32x d = SHA1M_D;
+ u32x e = SHA1M_E;
#undef K
#define K SHA1C00
w3_t[2] = rotl32 ((w2_t[3] ^ w1_t[2] ^ w0_t[0] ^ w3_t[2]), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, w3_t[2]);
w3_t[3] = rotl32 ((w3_t[0] ^ w1_t[3] ^ w0_t[1] ^ w3_t[3]), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, w3_t[3]);
-
- const u32 r0 = d;
- const u32 r1 = e;
- const u32 r2 = c;
- const u32 r3 = b;
-
- #include COMPARE_M
+ COMPARE_M_SIMD (d, e, c, b);
}
}
* reverse
*/
- const u32 e_rev = rotl32 (search[1], 2u);
+ const u32 e_rev = rotl32_S (search[1], 2u);
/**
* 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 salt
*/
- u32 s0[4];
+ u32x s0[4];
s0[0] = salt_buf0[0];
s0[1] = salt_buf0[1];
s0[2] = salt_buf0[2];
s0[3] = salt_buf0[3];
- u32 s1[4];
+ u32x s1[4];
s1[0] = salt_buf1[0];
s1[1] = salt_buf1[1];
s1[2] = salt_buf1[2];
s1[3] = salt_buf1[3];
- u32 s2[4];
+ u32x s2[4];
s2[0] = 0;
s2[1] = 0;
s2[2] = 0;
s2[3] = 0;
- u32 s3[4];
+ u32x s3[4];
s3[0] = 0;
s3[1] = 0;
const u32 out_salt_len = (out_len * 2) + salt_len;
- u32 w0_t[4];
- u32 w1_t[4];
- u32 w2_t[4];
- u32 w3_t[4];
+ u32x w0_t[4];
+ u32x w1_t[4];
+ u32x w2_t[4];
+ u32x w3_t[4];
make_unicode (w0, w0_t, w1_t);
make_unicode (w1, w2_t, w3_t);
//w3_t[2] = swap32 (w3_t[2]);
//w3_t[3] = swap32 (w3_t[3]);
- u32 a = SHA1M_A;
- u32 b = SHA1M_B;
- u32 c = SHA1M_C;
- u32 d = SHA1M_D;
- u32 e = SHA1M_E;
+ u32x a = SHA1M_A;
+ u32x b = SHA1M_B;
+ u32x c = SHA1M_C;
+ u32x d = SHA1M_D;
+ u32x e = SHA1M_E;
#undef K
#define K SHA1C00
w2_t[1] = rotl32 ((w1_t[2] ^ w0_t[1] ^ w2_t[3] ^ w2_t[1]), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, w2_t[1]);
w2_t[2] = rotl32 ((w1_t[3] ^ w0_t[2] ^ w3_t[0] ^ w2_t[2]), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, w2_t[2]);
w2_t[3] = rotl32 ((w2_t[0] ^ w0_t[3] ^ w3_t[1] ^ w2_t[3]), 1u); SHA1_STEP (SHA1_F1, a, b, c, d, e, w2_t[3]);
- w3_t[0] = rotl32 ((w2_t[1] ^ w1_t[0] ^ w3_t[2] ^ w3_t[0]), 1u); SHA1_STEP (SHA1_F1, e, a, b, c, d, w3_t[0]);
- if (allx (e != e_rev)) continue;
+ if (MATCHES_NONE_VS (e, e_rev)) continue;
+ w3_t[0] = rotl32 ((w2_t[1] ^ w1_t[0] ^ w3_t[2] ^ w3_t[0]), 1u); SHA1_STEP (SHA1_F1, e, a, b, c, d, w3_t[0]);
w3_t[1] = rotl32 ((w2_t[2] ^ w1_t[1] ^ w3_t[3] ^ w3_t[1]), 1u); SHA1_STEP (SHA1_F1, d, e, a, b, c, w3_t[1]);
w3_t[2] = rotl32 ((w2_t[3] ^ w1_t[2] ^ w0_t[0] ^ w3_t[2]), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, w3_t[2]);
w3_t[3] = rotl32 ((w3_t[0] ^ w1_t[3] ^ w0_t[1] ^ w3_t[3]), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, w3_t[3]);
-
- const u32 r0 = d;
- const u32 r1 = e;
- const u32 r2 = c;
- const u32 r3 = b;
-
- #include COMPARE_S
+ COMPARE_S_SIMD (d, e, c, b);
}
}