-#include "include/constants.h"
-#include "include/kernel_vendor.h"
-
-#define DGST_R0 0
-#define DGST_R1 3
-#define DGST_R2 2
-#define DGST_R3 1
-
-#include "include/kernel_functions.c"
-#include "OpenCL/types_ocl.c"
-#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"
-
-__kernel void m00010_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 rules_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+#define NEW_SIMD_CODE
+
+#include "inc_vendor.cl"
+#include "inc_hash_constants.h"
+#include "inc_hash_functions.cl"
+#include "inc_types.cl"
+#include "inc_common.cl"
+#include "inc_rp.h"
+#include "inc_rp.cl"
+#include "inc_simd.cl"
+
+__kernel void m00010_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
- MD5_STEP (MD5_H1, a, b, c, d, w1[1], MD5C20, MD5S20);
- MD5_STEP (MD5_H2, d, a, b, c, w2[0], MD5C21, MD5S21);
- MD5_STEP (MD5_H1, c, d, a, b, w2[3], MD5C22, MD5S22);
- MD5_STEP (MD5_H2, b, c, d, a, w3[2], MD5C23, MD5S23);
- MD5_STEP (MD5_H1, a, b, c, d, w0[1], MD5C24, MD5S20);
- MD5_STEP (MD5_H2, d, a, b, c, w1[0], MD5C25, MD5S21);
- MD5_STEP (MD5_H1, c, d, a, b, w1[3], MD5C26, MD5S22);
- MD5_STEP (MD5_H2, b, c, d, a, w2[2], MD5C27, MD5S23);
- MD5_STEP (MD5_H1, a, b, c, d, w3[1], MD5C28, MD5S20);
- MD5_STEP (MD5_H2, d, a, b, c, w0[0], MD5C29, MD5S21);
- MD5_STEP (MD5_H1, c, d, a, b, w0[3], MD5C2a, MD5S22);
- MD5_STEP (MD5_H2, b, c, d, a, w1[2], MD5C2b, MD5S23);
- MD5_STEP (MD5_H1, a, b, c, d, w2[1], MD5C2c, MD5S20);
- MD5_STEP (MD5_H2, d, a, b, c, w3[0], MD5C2d, MD5S21);
- MD5_STEP (MD5_H1, c, d, a, b, w3[3], MD5C2e, MD5S22);
- MD5_STEP (MD5_H2, b, c, d, a, w0[2], 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);
-__kernel void m00010_m08 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 rules_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+__kernel void m00010_m08 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
-__kernel void m00010_m16 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 rules_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+__kernel void m00010_m16 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
-__kernel void m00010_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 rules_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+__kernel void m00010_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
- MD5_STEP (MD5_H1, a, b, c, d, w1[1], MD5C20, MD5S20);
- MD5_STEP (MD5_H2, d, a, b, c, w2[0], MD5C21, MD5S21);
- MD5_STEP (MD5_H1, c, d, a, b, w2[3], MD5C22, MD5S22);
- MD5_STEP (MD5_H2, b, c, d, a, w3[2], MD5C23, MD5S23);
- MD5_STEP (MD5_H1, a, b, c, d, w0[1], MD5C24, MD5S20);
- MD5_STEP (MD5_H2, d, a, b, c, w1[0], MD5C25, MD5S21);
- MD5_STEP (MD5_H1, c, d, a, b, w1[3], MD5C26, MD5S22);
- MD5_STEP (MD5_H2, b, c, d, a, w2[2], MD5C27, MD5S23);
- MD5_STEP (MD5_H1, a, b, c, d, w3[1], MD5C28, MD5S20);
- MD5_STEP (MD5_H2, d, a, b, c, w0[0], MD5C29, MD5S21);
- MD5_STEP (MD5_H1, c, d, a, b, w0[3], MD5C2a, MD5S22);
- MD5_STEP (MD5_H2, b, c, d, a, w1[2], MD5C2b, MD5S23);
- MD5_STEP (MD5_H1, a, b, c, d, w2[1], MD5C2c, MD5S20);
- MD5_STEP (MD5_H2, d, a, b, c, w3[0], MD5C2d, MD5S21);
- MD5_STEP (MD5_H1, c, d, a, b, w3[3], MD5C2e, MD5S22);
- MD5_STEP (MD5_H2, b, c, d, a, w0[2], 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 , 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);
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);
-__kernel void m00010_s08 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 rules_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+__kernel void m00010_s08 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
-__kernel void m00010_s16 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 rules_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+__kernel void m00010_s16 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)