v3.7.10

algo/blake/blakecoin-4way.c

@@ -0,0 +1,106 @@
+#include "blakecoin-gate.h"
+
+#if defined (__AVX__)
+
+#include "blake-hash-4way.h"
+#include <string.h>
+#include <stdint.h>
+#include <memory.h>
+
+blake256r8_4way_context blakecoin_ctx;
+
+void blakecoin_4way_hash(void *state, const void *input)
+{
+     uint32_t vhash[8*4] __attribute__ ((aligned (64)));
+     blake256r8_4way_context ctx;
+     memcpy( &ctx, &blakecoin_ctx, sizeof ctx );
+     blake256r8_4way( &ctx, input + (64<<2), 16 );
+     blake256r8_4way_close( &ctx, vhash );
+     mm_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
+}
+
+int scanhash_blakecoin_4way( int thr_id, struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done )
+{
+   uint32_t vdata[20*4] __attribute__ ((aligned (64)));
+   uint32_t hash[8*4] __attribute__ ((aligned (32)));
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   const uint32_t first_nonce = pdata[19];
+   uint32_t HTarget = ptarget[7];
+   uint32_t _ALIGN(32) edata[20];
+   uint32_t n = first_nonce;
+   uint32_t *nonces = work->nonces;
+   bool *found = work->nfound;
+   int num_found = 0;
+
+   if (opt_benchmark)
+      HTarget = 0x7f;
+
+   // we need big endian data...
+   swab32_array( edata, pdata, 20 );
+
+   mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
+
+   blake256r8_4way_init( &blakecoin_ctx );
+   blake256r8_4way( &blakecoin_ctx, vdata, 64 );
+
+   uint32_t *noncep = vdata + 76;   // 19*4
+   do {
+      found[0] = found[1] = found[2] = found[3] = false;
+      be32enc( noncep,    n   );
+      be32enc( noncep +1, n+1 );
+      be32enc( noncep +2, n+2 );
+      be32enc( noncep +3, n+3 );
+
+      blakecoin_4way_hash( hash, vdata );
+      pdata[19] = n;
+
+      if (  hash[7] <= HTarget && fulltest( hash, ptarget ) )
+      {
+          found[0] = true;
+          num_found++;
+          nonces[0] = n;
+          work_set_target_ratio( work, hash );
+      }
+      if ( (hash+8)[7] <= HTarget && fulltest( hash+8, ptarget ) ) 
+      {
+          found[1] = true;
+          num_found++;
+          nonces[1] = n+1;
+          work_set_target_ratio( work, hash+8 );
+      }
+      if ( (hash+16)[7] <= HTarget && fulltest( hash+16, ptarget ) )
+      {
+           found[2] = true;
+           num_found++;
+           nonces[2] = n+2;
+           work_set_target_ratio( work, hash+16 );
+      }
+      if ( (hash+24)[7] <= HTarget && fulltest( hash+24, ptarget ) )
+      {
+           found[3] = true;
+           num_found++;
+           nonces[3] = n+3;
+           work_set_target_ratio( work, hash+24 );
+      }
+      n += 4;
+
+   } while ( (num_found == 0) && (n < max_nonce) 
+             && !work_restart[thr_id].restart );
+
+   *hashes_done = n - first_nonce + 1;
+
+   // workaround to prevent flood of hash reports when nonce range exhasuted
+   // and thread is spinning waiting for new work
+   if ( ( n >= max_nonce ) && ( *hashes_done < 10 ) )
+   {
+      *hashes_done = 0;
+      sleep(1);
+   }
+
+   return num_found;
+}
+
+#endif
+