From 1f44c1c98f9859722502a5cd7cfc69d619a53c42 Mon Sep 17 00:00:00 2001
From: Jens Steube <jens.steube@gmail.com>
Date: Wed, 29 Dec 2021 00:00:49 +0100
Subject: [PATCH] Add -a0 and -a3 kernel for ha256($salt.sha256_bin($pass))

---
 OpenCL/m21420_a1-pure.cl | 213 ++++++++++++++++++++++++++++++++++
 OpenCL/m21420_a3-pure.cl | 239 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 452 insertions(+)
 create mode 100644 OpenCL/m21420_a1-pure.cl
 create mode 100644 OpenCL/m21420_a3-pure.cl

diff --git a/OpenCL/m21420_a1-pure.cl b/OpenCL/m21420_a1-pure.cl
new file mode 100644
index 000000000..654a2afc1
--- /dev/null
+++ b/OpenCL/m21420_a1-pure.cl
@@ -0,0 +1,213 @@
+/**
+ * Author......: See docs/credits.txt
+ * License.....: MIT
+ */
+
+//#define NEW_SIMD_CODE
+
+#ifdef KERNEL_STATIC
+#include "inc_vendor.h"
+#include "inc_types.h"
+#include "inc_platform.cl"
+#include "inc_common.cl"
+#include "inc_scalar.cl"
+#include "inc_hash_sha256.cl"
+#endif
+
+KERNEL_FQ void m21420_mxx (KERN_ATTR_BASIC ())
+{
+  /**
+   * modifier
+   */
+
+  const u64 lid = get_local_id (0);
+  const u64 gid = get_global_id (0);
+  const u64 lsz = get_local_size (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * base
+   */
+
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+
+  u32 s[64] = { 0 };
+
+  const u32 salt_len = salt_bufs[SALT_POS].salt_len;
+
+  for (int i = 0, idx = 0; i < salt_len; i += 4, idx += 1)
+  {
+    s[idx] = hc_swap32_S (salt_bufs[SALT_POS].salt_buf[idx]);
+  }
+
+  sha256_ctx_t ctx1;
+
+  sha256_init (&ctx1);
+
+  sha256_update_global_swap (&ctx1, pws[gid].i, pws[gid].pw_len);
+
+  /**
+   * loop
+   */
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos++)
+  {
+    sha256_ctx_t ctx0 = ctx1;
+
+    sha256_update_global_swap (&ctx0, combs_buf[il_pos].i, combs_buf[il_pos].pw_len);
+
+    sha256_final (&ctx0);
+
+    const u32 a = ctx0.h[0];
+    const u32 b = ctx0.h[1];
+    const u32 c = ctx0.h[2];
+    const u32 d = ctx0.h[3];
+    const u32 e = ctx0.h[4];
+    const u32 f = ctx0.h[5];
+    const u32 g = ctx0.h[6];
+    const u32 h = ctx0.h[7];
+
+    sha256_ctx_t ctx;
+
+    sha256_init (&ctx);
+
+    sha256_update (&ctx, s, salt_len);
+
+    w0[0] = a;
+    w0[1] = b;
+    w0[2] = c;
+    w0[3] = d;
+    w1[0] = e;
+    w1[1] = f;
+    w1[2] = g;
+    w1[3] = h;
+    w2[0] = 0;
+    w2[1] = 0;
+    w2[2] = 0;
+    w2[3] = 0;
+    w3[0] = 0;
+    w3[1] = 0;
+    w3[2] = 0;
+    w3[3] = 0;
+
+    sha256_update_64 (&ctx, w0, w1, w2, w3, 32);
+
+    sha256_final (&ctx);
+
+    const u32 r0 = ctx.h[DGST_R0];
+    const u32 r1 = ctx.h[DGST_R1];
+    const u32 r2 = ctx.h[DGST_R2];
+    const u32 r3 = ctx.h[DGST_R3];
+
+    COMPARE_M_SCALAR (r0, r1, r2, r3);
+  }
+}
+
+KERNEL_FQ void m21420_sxx (KERN_ATTR_BASIC ())
+{
+  /**
+   * modifier
+   */
+
+  const u64 lid = get_local_id (0);
+  const u64 gid = get_global_id (0);
+  const u64 lsz = get_local_size (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * digest
+   */
+
+  const u32 search[4] =
+  {
+    digests_buf[DIGESTS_OFFSET].digest_buf[DGST_R0],
+    digests_buf[DIGESTS_OFFSET].digest_buf[DGST_R1],
+    digests_buf[DIGESTS_OFFSET].digest_buf[DGST_R2],
+    digests_buf[DIGESTS_OFFSET].digest_buf[DGST_R3]
+  };
+
+  /**
+   * base
+   */
+
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+
+  const u32 salt_len = salt_bufs[SALT_POS].salt_len;
+
+  u32 s[64] = { 0 };
+
+  for (int i = 0, idx = 0; i < salt_len; i += 4, idx += 1)
+  {
+    s[idx] = hc_swap32_S (salt_bufs[SALT_POS].salt_buf[idx]);
+  }
+
+  sha256_ctx_t ctx1;
+
+  sha256_init (&ctx1);
+
+  sha256_update_global_swap (&ctx1, pws[gid].i, pws[gid].pw_len);
+
+  /**
+   * loop
+   */
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos++)
+  {
+    sha256_ctx_t ctx0 = ctx1;
+
+    sha256_update_global_swap (&ctx0, combs_buf[il_pos].i, combs_buf[il_pos].pw_len);
+
+    sha256_final (&ctx0);
+
+    const u32 a = ctx0.h[0];
+    const u32 b = ctx0.h[1];
+    const u32 c = ctx0.h[2];
+    const u32 d = ctx0.h[3];
+    const u32 e = ctx0.h[4];
+    const u32 f = ctx0.h[5];
+    const u32 g = ctx0.h[6];
+    const u32 h = ctx0.h[7];
+
+    sha256_ctx_t ctx;
+
+    sha256_init (&ctx);
+
+    sha256_update (&ctx, s, salt_len);
+
+    w0[0] = a;
+    w0[1] = b;
+    w0[2] = c;
+    w0[3] = d;
+    w1[0] = e;
+    w1[1] = f;
+    w1[2] = g;
+    w1[3] = h;
+    w2[0] = 0;
+    w2[1] = 0;
+    w2[2] = 0;
+    w2[3] = 0;
+    w3[0] = 0;
+    w3[1] = 0;
+    w3[2] = 0;
+    w3[3] = 0;
+
+    sha256_update_64 (&ctx, w0, w1, w2, w3, 32);
+
+    sha256_final (&ctx);
+
+    const u32 r0 = ctx.h[DGST_R0];
+    const u32 r1 = ctx.h[DGST_R1];
+    const u32 r2 = ctx.h[DGST_R2];
+    const u32 r3 = ctx.h[DGST_R3];
+
+    COMPARE_S_SCALAR (r0, r1, r2, r3);
+  }
+}
diff --git a/OpenCL/m21420_a3-pure.cl b/OpenCL/m21420_a3-pure.cl
new file mode 100644
index 000000000..fa8d86f4a
--- /dev/null
+++ b/OpenCL/m21420_a3-pure.cl
@@ -0,0 +1,239 @@
+/**
+ * Author......: See docs/credits.txt
+ * License.....: MIT
+ */
+
+#define NEW_SIMD_CODE
+
+#ifdef KERNEL_STATIC
+#include "inc_vendor.h"
+#include "inc_types.h"
+#include "inc_platform.cl"
+#include "inc_common.cl"
+#include "inc_simd.cl"
+#include "inc_hash_sha256.cl"
+#endif
+
+KERNEL_FQ void m21420_mxx (KERN_ATTR_VECTOR ())
+{
+  /**
+   * modifier
+   */
+
+  const u64 gid = get_global_id (0);
+  const u64 lid = get_local_id (0);
+  const u64 lsz = get_local_size (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * base
+   */
+
+  u32x _w0[4];
+  u32x _w1[4];
+  u32x _w2[4];
+  u32x _w3[4];
+
+  const u32 pw_len = pws[gid].pw_len;
+
+  u32x w[64] = { 0 };
+
+  for (u32 i = 0, idx = 0; i < pw_len; i += 4, idx += 1)
+  {
+    w[idx] = pws[gid].i[idx];
+  }
+
+  const u32 salt_len = salt_bufs[SALT_POS].salt_len;
+
+  u32x s[64] = { 0 };
+
+  for (int i = 0, idx = 0; i < salt_len; i += 4, idx += 1)
+  {
+    s[idx] = hc_swap32_S (salt_bufs[SALT_POS].salt_buf[idx]);
+  }
+
+  /**
+   * loop
+   */
+
+  u32x w0l = w[0];
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
+  {
+    const u32x w0r = words_buf_r[il_pos / VECT_SIZE];
+
+    const u32x w0 = w0l | w0r;
+
+    w[0] = w0;
+
+    sha256_ctx_vector_t ctx0;
+
+    sha256_init_vector (&ctx0);
+
+    sha256_update_vector (&ctx0, w, pw_len);
+
+    sha256_final_vector (&ctx0);
+
+    const u32x a = ctx0.h[0];
+    const u32x b = ctx0.h[1];
+    const u32x c = ctx0.h[2];
+    const u32x d = ctx0.h[3];
+    const u32x e = ctx0.h[4];
+    const u32x f = ctx0.h[5];
+    const u32x g = ctx0.h[6];
+    const u32x h = ctx0.h[7];
+
+    sha256_ctx_vector_t ctx;
+
+    sha256_init_vector (&ctx);
+
+    sha256_update_vector (&ctx, s, salt_len);
+
+    _w0[0] = a;
+    _w0[1] = b;
+    _w0[2] = c;
+    _w0[3] = d;
+    _w1[0] = e;
+    _w1[1] = f;
+    _w1[2] = g;
+    _w1[3] = h;
+    _w2[0] = 0;
+    _w2[1] = 0;
+    _w2[2] = 0;
+    _w2[3] = 0;
+    _w3[0] = 0;
+    _w3[1] = 0;
+    _w3[2] = 0;
+    _w3[3] = 0;
+
+    sha256_update_vector_64 (&ctx, _w0, _w1, _w2, _w3, 32);
+
+    sha256_final_vector (&ctx);
+
+    const u32x r0 = ctx.h[DGST_R0];
+    const u32x r1 = ctx.h[DGST_R1];
+    const u32x r2 = ctx.h[DGST_R2];
+    const u32x r3 = ctx.h[DGST_R3];
+
+    COMPARE_M_SIMD (r0, r1, r2, r3);
+  }
+}
+
+KERNEL_FQ void m21420_sxx (KERN_ATTR_VECTOR ())
+{
+  /**
+   * modifier
+   */
+
+  const u64 gid = get_global_id (0);
+  const u64 lid = get_local_id (0);
+  const u64 lsz = get_local_size (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * digest
+   */
+
+  const u32 search[4] =
+  {
+    digests_buf[DIGESTS_OFFSET].digest_buf[DGST_R0],
+    digests_buf[DIGESTS_OFFSET].digest_buf[DGST_R1],
+    digests_buf[DIGESTS_OFFSET].digest_buf[DGST_R2],
+    digests_buf[DIGESTS_OFFSET].digest_buf[DGST_R3]
+  };
+
+  /**
+   * base
+   */
+
+  u32x _w0[4];
+  u32x _w1[4];
+  u32x _w2[4];
+  u32x _w3[4];
+
+  const u32 pw_len = pws[gid].pw_len;
+
+  u32x w[64] = { 0 };
+
+  for (u32 i = 0, idx = 0; i < pw_len; i += 4, idx += 1)
+  {
+    w[idx] = pws[gid].i[idx];
+  }
+
+  const u32 salt_len = salt_bufs[SALT_POS].salt_len;
+
+  u32x s[64] = { 0 };
+
+  for (int i = 0, idx = 0; i < salt_len; i += 4, idx += 1)
+  {
+    s[idx] = hc_swap32_S (salt_bufs[SALT_POS].salt_buf[idx]);
+  }
+
+  /**
+   * loop
+   */
+
+  u32x w0l = w[0];
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
+  {
+    const u32x w0r = words_buf_r[il_pos / VECT_SIZE];
+
+    const u32x w0 = w0l | w0r;
+
+    w[0] = w0;
+
+    sha256_ctx_vector_t ctx0;
+
+    sha256_init_vector (&ctx0);
+
+    sha256_update_vector (&ctx0, w, pw_len);
+
+    sha256_final_vector (&ctx0);
+
+    const u32x a = ctx0.h[0];
+    const u32x b = ctx0.h[1];
+    const u32x c = ctx0.h[2];
+    const u32x d = ctx0.h[3];
+    const u32x e = ctx0.h[4];
+    const u32x f = ctx0.h[5];
+    const u32x g = ctx0.h[6];
+    const u32x h = ctx0.h[7];
+
+    sha256_ctx_vector_t ctx;
+
+    sha256_init_vector (&ctx);
+
+    sha256_update_vector (&ctx, s, salt_len);
+
+    _w0[0] = a;
+    _w0[1] = b;
+    _w0[2] = c;
+    _w0[3] = d;
+    _w1[0] = e;
+    _w1[1] = f;
+    _w1[2] = g;
+    _w1[3] = h;
+    _w2[0] = 0;
+    _w2[1] = 0;
+    _w2[2] = 0;
+    _w2[3] = 0;
+    _w3[0] = 0;
+    _w3[1] = 0;
+    _w3[2] = 0;
+    _w3[3] = 0;
+
+    sha256_update_vector_64 (&ctx, _w0, _w1, _w2, _w3, 32);
+
+    sha256_final_vector (&ctx);
+
+    const u32x r0 = ctx.h[DGST_R0];
+    const u32x r1 = ctx.h[DGST_R1];
+    const u32x r2 = ctx.h[DGST_R2];
+    const u32x r3 = ctx.h[DGST_R3];
+
+    COMPARE_S_SIMD (r0, r1, r2, r3);
+  }
+}