v3.11.0

JayDDee · Jan 3, 2020 · 3572cb5 · 3572cb5
1 parent 241bc26
commit 3572cb5
Show file tree

Hide file tree

Showing 118 changed files with 7,024 additions and 1,569 deletions.
diff --git a/Makefile.am b/Makefile.am
@@ -87,8 +87,11 @@ cpuminer_SOURCES = \
   algo/echo/echo-hash-4way.c \
   algo/echo/aes_ni/hash.c\
   algo/gost/sph_gost.c \
+  algo/groestl/groestl-gate.c \
+  algo/groestl/groestl512-hash-4way.c \
   algo/groestl/sph_groestl.c \
   algo/groestl/groestl.c \
+  algo/groestl/groestl-4way.c \
   algo/groestl/myrgr-gate.c \
   algo/groestl/myrgr-4way.c \
   algo/groestl/myr-groestl.c \
@@ -188,6 +191,7 @@ cpuminer_SOURCES = \
   algo/shavite/sph_shavite.c \
   algo/shavite/sph-shavite-aesni.c \
   algo/shavite/shavite-hash-2way.c \
+  algo/shavite/shavite-hash-4way.c \
   algo/shavite/shavite.c \
   algo/simd/sph_simd.c \
   algo/simd/nist.c \

diff --git a/RELEASE_NOTES b/RELEASE_NOTES
@@ -21,7 +21,7 @@ required.
 Compile Instructions
 --------------------
 
-See INSTALL_LINUX or INSTALL_WINDOWS fror compile instruuctions
+See INSTALL_LINUX or INSTALL_WINDOWS for compile instruuctions
 
 Requirements
 ------------
@@ -35,13 +35,26 @@ not supported. FreeBSD YMMV.
 Change Log
 ----------
 
+v3.11.0
+
+Fixed x25x AVX512 lane 4 invalid shares.
+
+AVX512 for hex, phi2.
+
+VAES optimzation for Intel Icelake CPUs for most algos recently optimized
+with AVX512, source code only.
+
+v3.10.7
+
+AVX512 for x25x, lbry, x13bcd (bcd).
+
 v3.10.6
 
 Added support for SSL stratum: stratum+tcps://
 
 Added job id reporting again, but leaner, suppressed with --quiet.
 
-AVX512 for x21s, x22i, lyra2z, allium
+AVX512 for x21s, x22i, lyra2z, allium.
 
 Fixed share overflow warnings mining lbry with Ryzen (SHA).
 

diff --git a/algo-gate-api.c b/algo-gate-api.c
@@ -317,6 +317,7 @@ const char* const algo_alias_map[][2] =
   { "argon2d-crds",      "argon2d250"   },
   { "argon2d-dyn",       "argon2d500"   },
   { "argon2d-uis",       "argon2d4096"  },
+  { "bcd",               "x13bcd"       },
   { "bitcore",           "timetravel10" },
   { "bitzeny",           "yescryptr8"   },
   { "blake256r8",        "blakecoin"    },

diff --git a/algo/blake/blake-hash-4way.h b/algo/blake/blake-hash-4way.h
@@ -104,7 +104,7 @@ typedef struct {
 typedef blake_8way_small_context blake256_8way_context;
 void blake256_8way_init(void *cc);
 void blake256_8way_update(void *cc, const void *data, size_t len);
-#define blake256_8way blake256_8way_update
+//#define blake256_8way blake256_8way_update
 void blake256_8way_close(void *cc, void *dst);
 
 // 14 rounds, blake, decred

diff --git a/algo/blake/blake256-hash-4way.c b/algo/blake/blake256-hash-4way.c
@@ -842,7 +842,8 @@ blake32_4way_init( blake_4way_small_context *ctx, const uint32_t *iv,
 }
 
 static void
-blake32_4way( blake_4way_small_context *ctx, const void *data, size_t len )
+blake32_4way( blake_4way_small_context *ctx, const void *data,
+              size_t len )
 {
    __m128i *buf = (__m128i*)ctx->buf;
    size_t  bptr = ctx->ptr<<2;
@@ -1237,7 +1238,7 @@ blake256_4way_init(void *ctx)
 }
 
 void
-blake256_4way(void *ctx, const void *data, size_t len)
+blake256_4way_update(void *ctx, const void *data, size_t len)
 {
 	blake32_4way(ctx, data, len);
 }

diff --git a/algo/blake/blake2s-hash-4way.h b/algo/blake/blake2s-hash-4way.h
@@ -14,7 +14,6 @@
 #ifndef __BLAKE2S_HASH_4WAY_H__
 #define __BLAKE2S_HASH_4WAY_H__ 1
 
-//#if defined(__SSE4_2__)
 #if defined(__SSE2__)
 
 #include "simd-utils.h"
@@ -132,6 +131,6 @@ int blake2s_16way_final( blake2s_16way_state *S, void *out, uint8_t outlen );
 }
 #endif
 
-#endif  // __SSE4_2__
+#endif  // __SSE2__
 
 #endif
diff --git a/algo/bmw/bmw512-4way.c b/algo/bmw/bmw512-4way.c
@@ -41,7 +41,6 @@ int scanhash_bmw512_8way( struct work *work, uint32_t max_nonce,
 
       for ( int lane = 0; lane < 8; lane++ )
       if ( unlikely( hash7[ lane<<1 ] < Htarg ) )
-//      if ( ( ( hash7[ lane<<1 ] & 0xFFFFFF00 ) == 0 ) )
       {
           extr_lane_8x64( lane_hash, hash, lane, 256 );
           if ( fulltest( lane_hash, ptarget ) )
@@ -66,7 +65,7 @@ void bmw512hash_4way(void *state, const void *input)
 {
     bmw512_4way_context ctx;
     bmw512_4way_init( &ctx );
-    bmw512_4way( &ctx, input, 80 );
+    bmw512_4way_update( &ctx, input, 80 );
     bmw512_4way_close( &ctx, state );
 }
 

diff --git a/algo/cubehash/cubehash_sse2.c b/algo/cubehash/cubehash_sse2.c
@@ -21,7 +21,27 @@ static void transform( cubehashParam *sp )
     int r;
     const int rounds = sp->rounds;
 
-#ifdef __AVX2__
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+    register __m512i x0, x1;
+
+    x0 = _mm512_load_si512( (__m512i*)sp->x     );
+    x1 = _mm512_load_si512( (__m512i*)sp->x + 1 );
+
+    for ( r = 0; r < rounds; ++r )
+    { 
+        x1 = _mm512_add_epi32( x0, x1 );
+        x0 = _mm512_xor_si512( mm512_rol_32( mm512_swap_256( x0 ), 7 ), x1 );
+        x1 = _mm512_add_epi32( x0, mm512_swap128_64( x1 ) );
+        x0 = _mm512_xor_si512( mm512_rol_32(
+                                         mm512_swap256_128( x0 ), 11 ), x1 );
+        x1 = mm512_swap64_32( x1 );
+    }
+
+    _mm512_store_si512( (__m512i*)sp->x,     x0 );
+    _mm512_store_si512( (__m512i*)sp->x + 1, x1 );
+
+#elif defined(__AVX2__)
 
     register __m256i x0, x1, x2, x3, y0, y1;
 

diff --git a/algo/echo/aes_ni/hash.c b/algo/echo/aes_ni/hash.c
@@ -186,7 +186,7 @@ void Compress(hashState_echo *ctx, const unsigned char *pmsg, unsigned int uBloc
 	{
 	   for(i = 0; i < 4; i++)
 	   {
-		_state[i][j] = _mm_loadu_si128((__m128i*)pmsg + 4 * (j - (ctx->uHashSize / 256)) + i);
+		_state[i][j] = _mm_load_si128((__m128i*)pmsg + 4 * (j - (ctx->uHashSize / 256)) + i);
 	   }
 	}
 
@@ -390,13 +390,13 @@ HashReturn final_echo(hashState_echo *state, BitSequence *hashval)
 	}
 
 	// Store the hash value
-	_mm_storeu_si128((__m128i*)hashval + 0, state->state[0][0]);
-	_mm_storeu_si128((__m128i*)hashval + 1, state->state[1][0]);
+	_mm_store_si128((__m128i*)hashval + 0, state->state[0][0]);
+	_mm_store_si128((__m128i*)hashval + 1, state->state[1][0]);
 
 	if(state->uHashSize == 512)
 	{
-		_mm_storeu_si128((__m128i*)hashval + 2, state->state[2][0]);
-		_mm_storeu_si128((__m128i*)hashval + 3, state->state[3][0]);
+		_mm_store_si128((__m128i*)hashval + 2, state->state[2][0]);
+		_mm_store_si128((__m128i*)hashval + 3, state->state[3][0]);
 	}
 
 	return SUCCESS;
@@ -513,13 +513,13 @@ HashReturn update_final_echo( hashState_echo *state, BitSequence *hashval,
    }
 
    // Store the hash value
-   _mm_storeu_si128( (__m128i*)hashval + 0, state->state[0][0] );
-   _mm_storeu_si128( (__m128i*)hashval + 1, state->state[1][0] );
+   _mm_store_si128( (__m128i*)hashval + 0, state->state[0][0] );
+   _mm_store_si128( (__m128i*)hashval + 1, state->state[1][0] );
 
    if( state->uHashSize == 512 )
    {
-        _mm_storeu_si128( (__m128i*)hashval + 2, state->state[2][0] );
-        _mm_storeu_si128( (__m128i*)hashval + 3, state->state[3][0] );
+        _mm_store_si128( (__m128i*)hashval + 2, state->state[2][0] );
+        _mm_store_si128( (__m128i*)hashval + 3, state->state[3][0] );
 
    }
    return SUCCESS;