First performance optimization for SM3 hash

OpenCL/m31100_a3-optimized.cl

@@ -148,6 +148,8 @@ DECLSPEC void m31100s (PRIVATE_AS u32 *w, const u32 pw_len, KERN_ATTR_FUNC_VECTO
     digests_buf[DIGESTS_OFFSET_HOST].digest_buf[DGST_R3]
   };
 
+  const u32 d_rev = hc_rotr32_S (search[0], 9);
+
   /**
    * loop
    */
@@ -250,6 +252,9 @@ DECLSPEC void m31100s (PRIVATE_AS u32 *w, const u32 pw_len, KERN_ATTR_FUNC_VECTO
     we_t = SM3_EXPAND(we_t, w5_t, wb_t, w1_t, w8_t); SM3_ROUND2(c, d, a, b, g, h, e, f, SM3_T58, wa_t, wa_t ^ we_t);
     wf_t = SM3_EXPAND(wf_t, w6_t, wc_t, w2_t, w9_t); SM3_ROUND2(b, c, d, a, f, g, h, e, SM3_T59, wb_t, wb_t ^ wf_t);
     w0_t = SM3_EXPAND(w0_t, w7_t, wd_t, w3_t, wa_t); SM3_ROUND2(a, b, c, d, e, f, g, h, SM3_T60, wc_t, wc_t ^ w0_t);
+
+    if (MATCHES_NONE_VS (d, d_rev)) continue;
+
     w1_t = SM3_EXPAND(w1_t, w8_t, we_t, w4_t, wb_t); SM3_ROUND2(d, a, b, c, h, e, f, g, SM3_T61, wd_t, wd_t ^ w1_t);
     w2_t = SM3_EXPAND(w2_t, w9_t, wf_t, w5_t, wc_t); SM3_ROUND2(c, d, a, b, g, h, e, f, SM3_T62, we_t, we_t ^ w2_t);
     w3_t = SM3_EXPAND(w3_t, wa_t, w0_t, w6_t, wd_t); SM3_ROUND2(b, c, d, a, f, g, h, e, SM3_T63, wf_t, wf_t ^ w3_t);