New SIMD code for -a 1 -m 1460

pull/249/head
Jens Steube 8 years ago
parent c788ecdb80
commit eaaeac4aca

@ -5,6 +5,8 @@
#define _SHA256_
#define NEW_SIMD_CODE
#include "include/constants.h"
#include "include/kernel_vendor.h"
@ -16,9 +18,7 @@
#include "include/kernel_functions.c"
#include "OpenCL/types_ocl.c"
#include "OpenCL/common.c"
#define COMPARE_S "OpenCL/check_single_comp4.c"
#define COMPARE_M "OpenCL/check_multi_comp4.c"
#include "OpenCL/simd.c"
__constant u32 k_sha256[64] =
{
@ -40,33 +40,33 @@ __constant u32 k_sha256[64] =
SHA256C3c, SHA256C3d, SHA256C3e, SHA256C3f,
};
static void sha256_transform (const u32 w0[4], const u32 w1[4], const u32 w2[4], const u32 w3[4], u32 digest[8])
static void sha256_transform (const u32x w0[4], const u32x w1[4], const u32x w2[4], const u32x w3[4], u32x digest[8])
{
u32 a = digest[0];
u32 b = digest[1];
u32 c = digest[2];
u32 d = digest[3];
u32 e = digest[4];
u32 f = digest[5];
u32 g = digest[6];
u32 h = digest[7];
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 f = digest[5];
u32x g = digest[6];
u32x h = digest[7];
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];
#define ROUND_EXPAND() \
{ \
@ -126,7 +126,7 @@ static void sha256_transform (const u32 w0[4], const u32 w1[4], const u32 w2[4],
digest[7] += h;
}
static void hmac_sha256_pad (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u32 ipad[8], u32 opad[8])
static void hmac_sha256_pad (u32x w0[4], u32x w1[4], u32x w2[4], u32x w3[4], u32x ipad[8], u32x opad[8])
{
w0[0] = w0[0] ^ 0x36363636;
w0[1] = w0[1] ^ 0x36363636;
@ -185,7 +185,7 @@ static void hmac_sha256_pad (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u32 ipa
sha256_transform (w0, w1, w2, w3, opad);
}
static void hmac_sha256_run (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u32 ipad[8], u32 opad[8], u32 digest[8])
static void hmac_sha256_run (u32x w0[4], u32x w1[4], u32x w2[4], u32x w3[4], u32x ipad[8], u32x opad[8], u32x digest[8])
{
digest[0] = ipad[0];
digest[1] = ipad[1];
@ -243,93 +243,76 @@ __kernel void m01460_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
if (gid >= gid_max) return;
u32 wordl0[4];
wordl0[0] = pws[gid].i[ 0];
wordl0[1] = pws[gid].i[ 1];
wordl0[2] = pws[gid].i[ 2];
wordl0[3] = pws[gid].i[ 3];
u32 wordl1[4];
wordl1[0] = pws[gid].i[ 4];
wordl1[1] = pws[gid].i[ 5];
wordl1[2] = pws[gid].i[ 6];
wordl1[3] = pws[gid].i[ 7];
u32 wordl2[4];
u32 pw_buf0[4];
u32 pw_buf1[4];
wordl2[0] = 0;
wordl2[1] = 0;
wordl2[2] = 0;
wordl2[3] = 0;
u32 wordl3[4];
wordl3[0] = 0;
wordl3[1] = 0;
wordl3[2] = 0;
wordl3[3] = 0;
pw_buf0[0] = pws[gid].i[0];
pw_buf0[1] = pws[gid].i[1];
pw_buf0[2] = pws[gid].i[2];
pw_buf0[3] = pws[gid].i[3];
pw_buf1[0] = pws[gid].i[4];
pw_buf1[1] = pws[gid].i[5];
pw_buf1[2] = pws[gid].i[6];
pw_buf1[3] = pws[gid].i[7];
const u32 pw_l_len = pws[gid].pw_len;
if (combs_mode == COMBINATOR_MODE_BASE_RIGHT)
{
switch_buffer_by_offset_le (wordl0, wordl1, wordl2, wordl3, combs_buf[0].pw_len);
}
/**
* salt
*/
u32 salt_buf0[4];
salt_buf0[0] = salt_bufs[salt_pos].salt_buf[ 0];
salt_buf0[1] = salt_bufs[salt_pos].salt_buf[ 1];
salt_buf0[2] = salt_bufs[salt_pos].salt_buf[ 2];
salt_buf0[3] = salt_bufs[salt_pos].salt_buf[ 3];
u32 salt_buf1[4];
salt_buf1[0] = salt_bufs[salt_pos].salt_buf[ 4];
salt_buf1[1] = salt_bufs[salt_pos].salt_buf[ 5];
salt_buf1[2] = salt_bufs[salt_pos].salt_buf[ 6];
salt_buf1[3] = salt_bufs[salt_pos].salt_buf[ 7];
u32 salt_buf2[4];
u32 salt_buf3[4];
salt_buf0[0] = swap32_S (salt_bufs[salt_pos].salt_buf[ 0]);
salt_buf0[1] = swap32_S (salt_bufs[salt_pos].salt_buf[ 1]);
salt_buf0[2] = swap32_S (salt_bufs[salt_pos].salt_buf[ 2]);
salt_buf0[3] = swap32_S (salt_bufs[salt_pos].salt_buf[ 3]);
salt_buf1[0] = swap32_S (salt_bufs[salt_pos].salt_buf[ 4]);
salt_buf1[1] = swap32_S (salt_bufs[salt_pos].salt_buf[ 5]);
salt_buf1[2] = swap32_S (salt_bufs[salt_pos].salt_buf[ 6]);
salt_buf1[3] = swap32_S (salt_bufs[salt_pos].salt_buf[ 7]);
salt_buf2[0] = swap32_S (salt_bufs[salt_pos].salt_buf[ 8]);
salt_buf2[1] = swap32_S (salt_bufs[salt_pos].salt_buf[ 9]);
salt_buf2[2] = swap32_S (salt_bufs[salt_pos].salt_buf[10]);
salt_buf2[3] = swap32_S (salt_bufs[salt_pos].salt_buf[11]);
salt_buf3[0] = swap32_S (salt_bufs[salt_pos].salt_buf[12]);
salt_buf3[1] = swap32_S (salt_bufs[salt_pos].salt_buf[13]);
salt_buf3[2] = swap32_S (salt_bufs[salt_pos].salt_buf[14]);
salt_buf3[3] = swap32_S (salt_bufs[salt_pos].salt_buf[15]);
const u32 salt_len = salt_bufs[salt_pos].salt_len;
/**
* pads
*/
u32 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];
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];
w2_t[0] = 0;
w2_t[1] = 0;
w2_t[2] = 0;
w2_t[3] = 0;
u32 w3_t[4];
w3_t[0] = 0;
w3_t[1] = 0;
w3_t[2] = 0;
w3_t[3] = 0;
u32 ipad[8];
u32 opad[8];
u32x w0_t[4];
u32x w1_t[4];
u32x w2_t[4];
u32x w3_t[4];
w0_t[0] = salt_buf0[0];
w0_t[1] = salt_buf0[1];
w0_t[2] = salt_buf0[2];
w0_t[3] = salt_buf0[3];
w1_t[0] = salt_buf1[0];
w1_t[1] = salt_buf1[1];
w1_t[2] = salt_buf1[2];
w1_t[3] = salt_buf1[3];
w2_t[0] = salt_buf2[0];
w2_t[1] = salt_buf2[1];
w2_t[2] = salt_buf2[2];
w2_t[3] = salt_buf2[3];
w3_t[0] = salt_buf3[0];
w3_t[1] = salt_buf3[1];
w3_t[2] = salt_buf3[2];
w3_t[3] = salt_buf3[3];
u32x ipad[8];
u32x opad[8];
hmac_sha256_pad (w0_t, w1_t, w2_t, w3_t, ipad, opad);
@ -337,102 +320,99 @@ __kernel void m01460_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
* loop
*/
for (u32 il_pos = 0; il_pos < il_cnt; il_pos++)
for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
{
const u32 pw_r_len = combs_buf[il_pos].pw_len;
const u32 pw_len = pw_l_len + pw_r_len;
u32 wordr0[4];
wordr0[0] = combs_buf[il_pos].i[0];
wordr0[1] = combs_buf[il_pos].i[1];
wordr0[2] = combs_buf[il_pos].i[2];
wordr0[3] = combs_buf[il_pos].i[3];
u32 wordr1[4];
wordr1[0] = combs_buf[il_pos].i[4];
wordr1[1] = combs_buf[il_pos].i[5];
wordr1[2] = combs_buf[il_pos].i[6];
wordr1[3] = combs_buf[il_pos].i[7];
u32 wordr2[4];
wordr2[0] = 0;
wordr2[1] = 0;
wordr2[2] = 0;
wordr2[3] = 0;
u32 wordr3[4];
wordr3[0] = 0;
wordr3[1] = 0;
wordr3[2] = 0;
wordr3[3] = 0;
const u32x pw_r_len = pwlenx_create_combt (combs_buf, il_pos);
const u32x pw_len = pw_l_len + pw_r_len;
/**
* concat password candidate
*/
u32x wordl0[4] = { 0 };
u32x wordl1[4] = { 0 };
u32x wordl2[4] = { 0 };
u32x wordl3[4] = { 0 };
wordl0[0] = pw_buf0[0];
wordl0[1] = pw_buf0[1];
wordl0[2] = pw_buf0[2];
wordl0[3] = pw_buf0[3];
wordl1[0] = pw_buf1[0];
wordl1[1] = pw_buf1[1];
wordl1[2] = pw_buf1[2];
wordl1[3] = pw_buf1[3];
u32x wordr0[4] = { 0 };
u32x wordr1[4] = { 0 };
u32x wordr2[4] = { 0 };
u32x wordr3[4] = { 0 };
wordr0[0] = ix_create_combt (combs_buf, il_pos, 0);
wordr0[1] = ix_create_combt (combs_buf, il_pos, 1);
wordr0[2] = ix_create_combt (combs_buf, il_pos, 2);
wordr0[3] = ix_create_combt (combs_buf, il_pos, 3);
wordr1[0] = ix_create_combt (combs_buf, il_pos, 4);
wordr1[1] = ix_create_combt (combs_buf, il_pos, 5);
wordr1[2] = ix_create_combt (combs_buf, il_pos, 6);
wordr1[3] = ix_create_combt (combs_buf, il_pos, 7);
if (combs_mode == COMBINATOR_MODE_BASE_LEFT)
{
switch_buffer_by_offset_le (wordr0, wordr1, wordr2, wordr3, pw_l_len);
switch_buffer_by_offset_le_VV (wordr0, wordr1, wordr2, wordr3, pw_l_len);
}
else
{
switch_buffer_by_offset_le_VV (wordl0, wordl1, wordl2, wordl3, pw_r_len);
}
u32 w0[4];
u32x w0[4];
u32x w1[4];
u32x w2[4];
u32x w3[4];
w0[0] = wordl0[0] | wordr0[0];
w0[1] = wordl0[1] | wordr0[1];
w0[2] = wordl0[2] | wordr0[2];
w0[3] = wordl0[3] | wordr0[3];
u32 w1[4];
w1[0] = wordl1[0] | wordr1[0];
w1[1] = wordl1[1] | wordr1[1];
w1[2] = wordl1[2] | wordr1[2];
w1[3] = wordl1[3] | wordr1[3];
u32 w2[4];
w2[0] = wordl2[0] | wordr2[0];
w2[1] = wordl2[1] | wordr2[1];
w2[2] = wordl2[2] | wordr2[2];
w2[3] = wordl2[3] | wordr2[3];
u32 w3[4];
w3[0] = wordl3[0] | wordr3[0];
w3[1] = wordl3[1] | wordr3[1];
w3[2] = wordl3[2] | wordr3[2];
w3[3] = wordl3[3] | wordr3[3];
append_0x80_4x4_VV (w0, w1, w2, w3, pw_len);
w0[0] = swap32 (w0[0]);
w0[1] = swap32 (w0[1]);
w0[2] = swap32 (w0[2]);
w0[3] = swap32 (w0[3]);
w1[0] = swap32 (w1[0]);
w1[1] = swap32 (w1[1]);
w1[2] = swap32 (w1[2]);
w1[3] = swap32 (w1[3]);
w2[0] = swap32 (w2[0]);
w2[1] = swap32 (w2[1]);
w2[2] = swap32 (w2[2]);
w2[3] = swap32 (w2[3]);
w3[0] = swap32 (w3[0]);
w3[1] = swap32 (w3[1]);
w3[2] = 0;
w3[3] = 0;
append_0x80_4x4 (w0, w1, w2, w3, pw_len);
w0_t[0] = swap32 (w0[0]);
w0_t[1] = swap32 (w0[1]);
w0_t[2] = swap32 (w0[2]);
w0_t[3] = swap32 (w0[3]);
w1_t[0] = swap32 (w1[0]);
w1_t[1] = swap32 (w1[1]);
w1_t[2] = swap32 (w1[2]);
w1_t[3] = swap32 (w1[3]);
w2_t[0] = swap32 (w2[0]);
w2_t[1] = swap32 (w2[1]);
w2_t[2] = swap32 (w2[2]);
w2_t[3] = swap32 (w2[3]);
w3_t[0] = swap32 (w3[0]);
w3_t[1] = swap32 (w3[1]);
w3_t[2] = 0;
w3_t[3] = (64 + pw_len) * 8;
u32 digest[8];
hmac_sha256_run (w0_t, w1_t, w2_t, w3_t, ipad, opad, digest);
const u32 r0 = digest[3];
const u32 r1 = digest[7];
const u32 r2 = digest[2];
const u32 r3 = digest[6];
#include COMPARE_M
w3[3] = (64 + pw_len) * 8;
u32x digest[8];
hmac_sha256_run (w0, w1, w2, w3, ipad, opad, digest);
COMPARE_M_SIMD (digest[3], digest[7], digest[2], digest[6]);
}
}
@ -460,93 +440,76 @@ __kernel void m01460_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
if (gid >= gid_max) return;
u32 wordl0[4];
wordl0[0] = pws[gid].i[ 0];
wordl0[1] = pws[gid].i[ 1];
wordl0[2] = pws[gid].i[ 2];
wordl0[3] = pws[gid].i[ 3];
u32 wordl1[4];
wordl1[0] = pws[gid].i[ 4];
wordl1[1] = pws[gid].i[ 5];
wordl1[2] = pws[gid].i[ 6];
wordl1[3] = pws[gid].i[ 7];
u32 wordl2[4];
wordl2[0] = 0;
wordl2[1] = 0;
wordl2[2] = 0;
wordl2[3] = 0;
u32 pw_buf0[4];
u32 pw_buf1[4];
u32 wordl3[4];
wordl3[0] = 0;
wordl3[1] = 0;
wordl3[2] = 0;
wordl3[3] = 0;
pw_buf0[0] = pws[gid].i[0];
pw_buf0[1] = pws[gid].i[1];
pw_buf0[2] = pws[gid].i[2];
pw_buf0[3] = pws[gid].i[3];
pw_buf1[0] = pws[gid].i[4];
pw_buf1[1] = pws[gid].i[5];
pw_buf1[2] = pws[gid].i[6];
pw_buf1[3] = pws[gid].i[7];
const u32 pw_l_len = pws[gid].pw_len;
if (combs_mode == COMBINATOR_MODE_BASE_RIGHT)
{
switch_buffer_by_offset_le (wordl0, wordl1, wordl2, wordl3, combs_buf[0].pw_len);
}
/**
* salt
*/
u32 salt_buf0[4];
salt_buf0[0] = salt_bufs[salt_pos].salt_buf[ 0];
salt_buf0[1] = salt_bufs[salt_pos].salt_buf[ 1];
salt_buf0[2] = salt_bufs[salt_pos].salt_buf[ 2];
salt_buf0[3] = salt_bufs[salt_pos].salt_buf[ 3];
u32 salt_buf1[4];
salt_buf1[0] = salt_bufs[salt_pos].salt_buf[ 4];
salt_buf1[1] = salt_bufs[salt_pos].salt_buf[ 5];
salt_buf1[2] = salt_bufs[salt_pos].salt_buf[ 6];
salt_buf1[3] = salt_bufs[salt_pos].salt_buf[ 7];
u32 salt_buf2[4];
u32 salt_buf3[4];
salt_buf0[0] = swap32_S (salt_bufs[salt_pos].salt_buf[ 0]);
salt_buf0[1] = swap32_S (salt_bufs[salt_pos].salt_buf[ 1]);
salt_buf0[2] = swap32_S (salt_bufs[salt_pos].salt_buf[ 2]);
salt_buf0[3] = swap32_S (salt_bufs[salt_pos].salt_buf[ 3]);
salt_buf1[0] = swap32_S (salt_bufs[salt_pos].salt_buf[ 4]);
salt_buf1[1] = swap32_S (salt_bufs[salt_pos].salt_buf[ 5]);
salt_buf1[2] = swap32_S (salt_bufs[salt_pos].salt_buf[ 6]);
salt_buf1[3] = swap32_S (salt_bufs[salt_pos].salt_buf[ 7]);
salt_buf2[0] = swap32_S (salt_bufs[salt_pos].salt_buf[ 8]);
salt_buf2[1] = swap32_S (salt_bufs[salt_pos].salt_buf[ 9]);
salt_buf2[2] = swap32_S (salt_bufs[salt_pos].salt_buf[10]);
salt_buf2[3] = swap32_S (salt_bufs[salt_pos].salt_buf[11]);
salt_buf3[0] = swap32_S (salt_bufs[salt_pos].salt_buf[12]);
salt_buf3[1] = swap32_S (salt_bufs[salt_pos].salt_buf[13]);
salt_buf3[2] = swap32_S (salt_bufs[salt_pos].salt_buf[14]);
salt_buf3[3] = swap32_S (salt_bufs[salt_pos].salt_buf[15]);
const u32 salt_len = salt_bufs[salt_pos].salt_len;
/**
* pads
*/
u32 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];
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];
w2_t[0] = 0;
w2_t[1] = 0;
w2_t[2] = 0;
w2_t[3] = 0;
u32 w3_t[4];
w3_t[0] = 0;
w3_t[1] = 0;
w3_t[2] = 0;
w3_t[3] = 0;
u32 ipad[8];
u32 opad[8];
u32x w0_t[4];
u32x w1_t[4];
u32x w2_t[4];
u32x w3_t[4];
w0_t[0] = salt_buf0[0];
w0_t[1] = salt_buf0[1];
w0_t[2] = salt_buf0[2];
w0_t[3] = salt_buf0[3];
w1_t[0] = salt_buf1[0];
w1_t[1] = salt_buf1[1];
w1_t[2] = salt_buf1[2];
w1_t[3] = salt_buf1[3];
w2_t[0] = salt_buf2[0];
w2_t[1] = salt_buf2[1];
w2_t[2] = salt_buf2[2];
w2_t[3] = salt_buf2[3];
w3_t[0] = salt_buf3[0];
w3_t[1] = salt_buf3[1];
w3_t[2] = salt_buf3[2];
w3_t[3] = salt_buf3[3];
u32x ipad[8];
u32x opad[8];
hmac_sha256_pad (w0_t, w1_t, w2_t, w3_t, ipad, opad);
@ -566,102 +529,99 @@ __kernel void m01460_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
* loop
*/
for (u32 il_pos = 0; il_pos < il_cnt; il_pos++)
for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
{
const u32 pw_r_len = combs_buf[il_pos].pw_len;
const u32 pw_len = pw_l_len + pw_r_len;
u32 wordr0[4];
wordr0[0] = combs_buf[il_pos].i[0];
wordr0[1] = combs_buf[il_pos].i[1];
wordr0[2] = combs_buf[il_pos].i[2];
wordr0[3] = combs_buf[il_pos].i[3];
u32 wordr1[4];
wordr1[0] = combs_buf[il_pos].i[4];
wordr1[1] = combs_buf[il_pos].i[5];
wordr1[2] = combs_buf[il_pos].i[6];
wordr1[3] = combs_buf[il_pos].i[7];
u32 wordr2[4];
wordr2[0] = 0;
wordr2[1] = 0;
wordr2[2] = 0;
wordr2[3] = 0;
u32 wordr3[4];
wordr3[0] = 0;
wordr3[1] = 0;
wordr3[2] = 0;
wordr3[3] = 0;
const u32x pw_r_len = pwlenx_create_combt (combs_buf, il_pos);
const u32x pw_len = pw_l_len + pw_r_len;
/**
* concat password candidate
*/
u32x wordl0[4] = { 0 };
u32x wordl1[4] = { 0 };
u32x wordl2[4] = { 0 };
u32x wordl3[4] = { 0 };
wordl0[0] = pw_buf0[0];
wordl0[1] = pw_buf0[1];
wordl0[2] = pw_buf0[2];
wordl0[3] = pw_buf0[3];
wordl1[0] = pw_buf1[0];
wordl1[1] = pw_buf1[1];
wordl1[2] = pw_buf1[2];
wordl1[3] = pw_buf1[3];
u32x wordr0[4] = { 0 };
u32x wordr1[4] = { 0 };
u32x wordr2[4] = { 0 };
u32x wordr3[4] = { 0 };
wordr0[0] = ix_create_combt (combs_buf, il_pos, 0);
wordr0[1] = ix_create_combt (combs_buf, il_pos, 1);
wordr0[2] = ix_create_combt (combs_buf, il_pos, 2);
wordr0[3] = ix_create_combt (combs_buf, il_pos, 3);
wordr1[0] = ix_create_combt (combs_buf, il_pos, 4);
wordr1[1] = ix_create_combt (combs_buf, il_pos, 5);
wordr1[2] = ix_create_combt (combs_buf, il_pos, 6);
wordr1[3] = ix_create_combt (combs_buf, il_pos, 7);
if (combs_mode == COMBINATOR_MODE_BASE_LEFT)
{
switch_buffer_by_offset_le (wordr0, wordr1, wordr2, wordr3, pw_l_len);
switch_buffer_by_offset_le_VV (wordr0, wordr1, wordr2, wordr3, pw_l_len);
}
else
{
switch_buffer_by_offset_le_VV (wordl0, wordl1, wordl2, wordl3, pw_r_len);
}
u32 w0[4];
u32x w0[4];
u32x w1[4];
u32x w2[4];
u32x w3[4];
w0[0] = wordl0[0] | wordr0[0];
w0[1] = wordl0[1] | wordr0[1];
w0[2] = wordl0[2] | wordr0[2];
w0[3] = wordl0[3] | wordr0[3];
u32 w1[4];
w1[0] = wordl1[0] | wordr1[0];
w1[1] = wordl1[1] | wordr1[1];
w1[2] = wordl1[2] | wordr1[2];
w1[3] = wordl1[3] | wordr1[3];
u32 w2[4];
w2[0] = wordl2[0] | wordr2[0];
w2[1] = wordl2[1] | wordr2[1];
w2[2] = wordl2[2] | wordr2[2];
w2[3] = wordl2[3] | wordr2[3];
u32 w3[4];
w3[0] = wordl3[0] | wordr3[0];
w3[1] = wordl3[1] | wordr3[1];
w3[2] = wordl3[2] | wordr3[2];
w3[3] = wordl3[3] | wordr3[3];
append_0x80_4x4_VV (w0, w1, w2, w3, pw_len);
w0[0] = swap32 (w0[0]);
w0[1] = swap32 (w0[1]);
w0[2] = swap32 (w0[2]);
w0[3] = swap32 (w0[3]);
w1[0] = swap32 (w1[0]);
w1[1] = swap32 (w1[1]);
w1[2] = swap32 (w1[2]);
w1[3] = swap32 (w1[3]);
w2[0] = swap32 (w2[0]);
w2[1] = swap32 (w2[1]);
w2[2] = swap32 (w2[2]);
w2[3] = swap32 (w2[3]);
w3[0] = swap32 (w3[0]);
w3[1] = swap32 (w3[1]);
w3[2] = 0;
w3[3] = 0;
append_0x80_4x4 (w0, w1, w2, w3, pw_len);
w0_t[0] = swap32 (w0[0]);
w0_t[1] = swap32 (w0[1]);
w0_t[2] = swap32 (w0[2]);
w0_t[3] = swap32 (w0[3]);
w1_t[0] = swap32 (w1[0]);
w1_t[1] = swap32 (w1[1]);
w1_t[2] = swap32 (w1[2]);
w1_t[3] = swap32 (w1[3]);
w2_t[0] = swap32 (w2[0]);
w2_t[1] = swap32 (w2[1]);
w2_t[2] = swap32 (w2[2]);
w2_t[3] = swap32 (w2[3]);
w3_t[0] = swap32 (w3[0]);
w3_t[1] = swap32 (w3[1]);
w3_t[2] = 0;
w3_t[3] = (64 + pw_len) * 8;
u32 digest[8];
hmac_sha256_run (w0_t, w1_t, w2_t, w3_t, ipad, opad, digest);
const u32 r0 = digest[3];
const u32 r1 = digest[7];
const u32 r2 = digest[2];
const u32 r3 = digest[6];
#include COMPARE_S
w3[3] = (64 + pw_len) * 8;
u32x digest[8];
hmac_sha256_run (w0, w1, w2, w3, ipad, opad, digest);
COMPARE_S_SIMD (digest[3], digest[7], digest[2], digest[6]);
}
}

Loading…
Cancel
Save