v3.7.4

Makefile.am

@@ -22,24 +22,6 @@ cpuminer_SOURCES = \
   api.c \
   sysinfos.c \
   algo-gate-api.c\
-  algo/groestl/sph_groestl.c \
-  algo/bmw/sph_bmw.c \
-  algo/shavite/sph_shavite.c \
-  algo/shavite/shavite.c \
-  algo/echo/sph_echo.c \
-  algo/heavy/sph_hefty1.c \
-  algo/luffa/sph_luffa.c \
-  algo/cubehash/sph_cubehash.c \
-  algo/simd/sph_simd.c \
-  algo/hamsi/sph_hamsi.c \
-  algo/fugue/sph_fugue.c \
-  algo/gost/sph_gost.c \
-  algo/jh/sph_jh.c \
-  algo/sha/sph_sha2.c \
-  algo/sha/sph_sha2big.c \
-  algo/shabal/sph_shabal.c \
-  algo/sm3/sm3.c \
-  algo/whirlpool/sph_whirlpool.c\
   crypto/blake2s.c \
   crypto/oaes_lib.c \
   crypto/c_keccak.c \
@@ -67,22 +49,34 @@ cpuminer_SOURCES = \
   algo/blake/blake2s.c \
   algo/blake/mod_blakecoin.c \
   algo/blake/blakecoin.c \
+  algo/blake/decred-gate.c \
   algo/blake/decred.c \
+  algo/blake/decred-4way.c \
+  algo/blake/pentablake-gate.c \
+  algo/blake/pentablake-4way.c \
   algo/blake/pentablake.c \
+  algo/bmw/sph_bmw.c \
   algo/bmw/bmw256.c \
-  algo/cubehash/sse2/cubehash_sse2.c\
   algo/cryptonight/cryptolight.c \
   algo/cryptonight/cryptonight-common.c\
   algo/cryptonight/cryptonight-aesni.c\
   algo/cryptonight/cryptonight.c\
+  algo/cubehash/sph_cubehash.c \
+  algo/cubehash/sse2/cubehash_sse2.c\
   algo/drop.c \
+  algo/echo/sph_echo.c \
   algo/echo/aes_ni/hash.c\
   algo/fresh.c \
+  algo/gost/sph_gost.c \
+  algo/groestl/sph_groestl.c \
   algo/groestl/groestl.c \
   algo/groestl/myr-groestl.c \
   algo/groestl/aes_ni/hash-groestl.c \
   algo/groestl/aes_ni/hash-groestl256.c \
+  algo/fugue/sph_fugue.c \
+  algo/hamsi/sph_hamsi.c \
   algo/haval/haval.c\
+  algo/heavy/sph_hefty1.c \
   algo/heavy/heavy.c \
   algo/heavy/bastion.c \
   algo/hmq1725.c \
@@ -91,6 +85,10 @@ cpuminer_SOURCES = \
   algo/hodl/hodl-wolf.c \
   algo/hodl/sha512_avx.c \
   algo/hodl/sha512_avx2.c \
+  algo/jh/sph_jh.c \
+  algo/jh/jh-hash-4way.c \
+  algo/jh/jha-gate.c \
+  algo/jh/jha-4way.c \
   algo/jh/jha.c \
   algo/keccak/sph_keccak.c \
   algo/keccak/keccak.c\
@@ -99,6 +97,7 @@ cpuminer_SOURCES = \
   algo/keccak/keccak-gate.c \
   algo/keccak/sse2/keccak.c \
   algo/lbry.c \
+  algo/luffa/sph_luffa.c \
   algo/luffa/luffa.c \
   algo/luffa/sse2/luffa_for_sse2.c \
   algo/lyra2/lyra2.c \
@@ -109,7 +108,9 @@ cpuminer_SOURCES = \
   algo/lyra2/lyra2z330.c \
   algo/m7m.c \
   algo/neoscrypt.c \
-  algo/nist5.c \
+  algo/nist5/nist5-gate.c \
+  algo/nist5/nist5-4way.c \
+  algo/nist5/nist5.c \
   algo/pluck.c \
   algo/polytimos/polytimos-gate.c \
   algo/polytimos/polytimos.c \
@@ -119,8 +120,14 @@ cpuminer_SOURCES = \
   algo/ripemd/sph_ripemd.c \
   algo/scrypt.c \
   algo/scryptjane/scrypt-jane.c \
+  algo/sha/sph_sha2.c \
+  algo/sha/sph_sha2big.c \
   algo/sha/sha2.c \
   algo/sha/sha256t.c \
+  algo/shabal/sph_shabal.c \
+  algo/shavite/sph_shavite.c \
+  algo/shavite/shavite.c \
+  algo/simd/sph_simd.c \
   algo/simd/sse2/nist.c \
   algo/simd/sse2/vector.c \
   algo/skein/sph_skein.c \
@@ -132,11 +139,18 @@ cpuminer_SOURCES = \
   algo/skein/skein2-4way.c \
   algo/skein/skein2-gate.c \
   algo/skunk.c \
+  algo/sm3/sm3.c \
   algo/tiger/sph_tiger.c \
   algo/timetravel.c \
   algo/timetravel10.c \
-  algo/tribus.c \
+  algo/tribus/tribus-gate.c \
+  algo/tribus/tribus.c \
+  algo/tribus/tribus-4way.c \
   algo/veltor.c \
+  algo/whirlpool/sph_whirlpool.c \
+  algo/whirlpool/whirlpool-hash-4way.c \
+  algo/whirlpool/whirlpool-gate.c \
+  algo/whirlpool/whirlpool-4way.c \
   algo/whirlpool/whirlpool.c \
   algo/whirlpool/whirlpoolx.c \
   algo/x11/phi1612.c \

README.txt

@@ -17,13 +17,17 @@ supported by cpuminer-opt due to an incompatible implementation of SSE2 on
 these CPUs. Some algos may crash the miner with an invalid instruction.
 Users are recommended to use an unoptimized miner such as cpuminer-multi.
 
-Exe name                  Compile opts       Arch name
-
-cpuminer-sse2.exe         -march=core2,      Core2   
-cpuminer-sse42.exe        -march=corei7,     Nehalem
-cpuminer-aes-sse42.exe    -maes -msse4.2     Westmere
-cpuminer-aes-avx.exe      -march=corei7-avx, Sandybridge, Ivybridge
-cpuminer-aes-avx2.exe     -march=core-avx2,  Haswell, Broadwell, Skylake, Kabylake
+Exe name                  Compile opts         Arch name
 
+cpuminer-sse2.exe         -march=core2         Core2   
+cpuminer-sse42.exe        -march=corei7        Nehalem
+cpuminer-aes-sse42.exe    -maes -msse4.2"      Westmere
+cpuminer-aes-avx.exe      -march=corei7-avx"   Sandybridge, Ivybridge
+cpuminer-aes-avx2.exe     "-march=core-avx2"   Haswell, Broadwell, Skylake, Kabylake
+cpuminer-4way.exe         "-march=core-avx2 -DFOUR_WAY"
 
+4way requires a CPU with AES and AVX2. It is still under development and
+only a few algos are supported. See change log in RELEASE_NOTES in source
+package for supported algos.
 
+There is no binary support available for SHA on AMD Ryzen CPUs.

RELEASE_NOTES

@@ -67,6 +67,11 @@ have been due to AVX and AVX2 optimizations added to that version.
 Additional improvements are expected on Ryzen with openssl 1.1.
 "-march-znver1" or "-msha".
 
+Additional instructions for static compilalation can be found here:
+https://lxadm.com/Static_compilation_of_cpuminer
+Static builds should only considered in a homogeneous HW and SW environment.
+Local builds will always have the best performance and compatibility.
+
 Extract cpuminer source.
 
 tar xvzf cpuminer-opt-x.y.z.tar.gz
@@ -96,6 +101,11 @@ Start mining.
 
 ./cpuminer -a algo -o url -u username -p password
 
+Windows
+
+The following in how the Windows binary releases are built. It's old and
+not very good but it works, for me anyway.
+
 Building on Windows prerequisites:
 
 msys
@@ -154,6 +164,12 @@ Support for even older x86_64 without AES_NI or SSE2 is not availble.
 Change Log
 ----------
 
+v3.7.4
+
+Removed unnecessary build options.
+
+Added 4way support for tribus and nist5.
+
 v3.7.3
 
 Added polytimos algo.

algo/blake/blake-4way.c

@@ -1,11 +1,11 @@
-#include "algo-gate-api.h"
+#include "blake-gate.h"
 #include "sph_blake.h"
 #include "blake-hash-4way.h"
 #include <string.h>
 #include <stdint.h>
 #include <memory.h>
 
-#if defined (__AVX__)
+#if defined (BLAKE_4WAY)
 
 void blakehash_4way(void *state, const void *input)
 {
@@ -41,7 +41,7 @@ int scanhash_blake_4way( int thr_id, struct work *work, uint32_t max_nonce,
    uint32_t n = first_nonce;
    uint32_t *nonces = work->nonces;
    bool *found = work->nfound;
-   int num_found;
+   int num_found = 0;
 
 //   if (opt_benchmark)
 //      HTarget = 0x7f;
@@ -55,7 +55,6 @@ int scanhash_blake_4way( int thr_id, struct work *work, uint32_t max_nonce,
    uint32_t *noncep = vdata + 76;   // 19*4
    do {
       found[0] = found[1] = found[2] = found[3] = false;
-      num_found = 0;
       be32enc( noncep,    n   );
       be32enc( noncep +2, n+1 );
       be32enc( noncep +4, n+2 );

algo/blake/blake-gate.c

@@ -8,11 +8,11 @@ int64_t blake_get_max64 ()
 bool register_blake_algo( algo_gate_t* gate )
 {
   gate->get_max64 = (void*)&blake_get_max64;
-#if defined (__AVX2__) && defined (FOUR_WAY)
+//#if defined (__AVX2__) && defined (FOUR_WAY)
 //   gate->optimizations = SSE2_OPT | AVX_OPT | AVX2_OPT;
 //  gate->scanhash  = (void*)&scanhash_blake_8way;
 //  gate->hash      = (void*)&blakehash_8way;
-#elif defined(__AVX__) && defined (FOUR_WAY)
+#if defined(BLAKE_4WAY)
   gate->optimizations = SSE2_OPT | AVX_OPT;
   gate->scanhash  = (void*)&scanhash_blake_4way;
   gate->hash      = (void*)&blakehash_4way;

algo/blake/blake-gate.h

@@ -4,13 +4,11 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-#if defined (__AVX2__) 
-//void blakehash_84way(void *state, const void *input);
-//int scanhash_blake_8way( int thr_id, struct work *work, uint32_t max_nonce,
-//                         uint64_t *hashes_done );
+#if defined(FOUR_WAY) && defined(__AVX__)
+  #define BLAKE_4WAY
 #endif
 
-#if defined (__AVX__)
+#if defined (BLAKE_4WAY)
 void blakehash_4way(void *state, const void *input);
 int scanhash_blake_4way( int thr_id, struct work *work, uint32_t max_nonce,
                          uint64_t *hashes_done );

algo/blake/blake-hash-4way.c

@@ -346,7 +346,7 @@ static const sph_u32 CS[16] = {
 #define CBF   SPH_C64(0x636920D871574E69)
 
 #if SPH_COMPACT_BLAKE_64
-
+// not used
 static const sph_u64 CB[16] = {
 	SPH_C64(0x243F6A8885A308D3), SPH_C64(0x13198A2E03707344),
 	SPH_C64(0xA4093822299F31D0), SPH_C64(0x082EFA98EC4E6C89),
@@ -428,7 +428,7 @@ do { \
 } while (0)
 
 #if SPH_COMPACT_BLAKE_64
-
+// not used
 #define ROUND_B_4WAY(r)   do { \
 	GB_4WAY(M[sigma[r][0x0]], M[sigma[r][0x1]], \
 		CB[sigma[r][0x0]], CB[sigma[r][0x1]], V0, V4, V8, VC); \
@@ -449,7 +449,7 @@ do { \
 } while (0)
 
 #else
-
+//current_impl
 #define ROUND_B_4WAY(r)   do { \
 	GB_4WAY(Mx(r, 0), Mx(r, 1), CBx(r, 0), CBx(r, 1), V0, V4, V8, VC); \
 	GB_4WAY(Mx(r, 2), Mx(r, 3), CBx(r, 2), CBx(r, 3), V1, V5, V9, VD); \
@@ -696,6 +696,7 @@ do { \
 
 #if SPH_COMPACT_BLAKE_64
 
+// not used
 #define COMPRESS64_4WAY   do { \
 	__m256i M[16]; \
 	__m256i V0, V1, V2, V3, V4, V5, V6, V7; \
@@ -759,6 +760,8 @@ do { \
 
 #else
 
+//current impl
+
 #define COMPRESS64_4WAY   do { \
      __m256i M0, M1, M2, M3, M4, M5, M6, M7; \
      __m256i M8, M9, MA, MB, MC, MD, ME, MF; \
@@ -986,7 +989,7 @@ blake64_4way( blake_4way_big_context *sc, const void *data, size_t len)
    size_t ptr;
    DECL_STATE64_4WAY
 
-   const int buf_size = 64;  //  sizeof/8 
+   const int buf_size = 128;  //  sizeof/8 
 
    buf = sc->buf;
    ptr = sc->ptr;
@@ -1037,7 +1040,7 @@ blake64_4way_close( blake_4way_big_context *sc,
    __m256i *out;
 
    ptr = sc->ptr;
-   bit_len = ((unsigned)ptr << 3) + n;
+   bit_len = ((unsigned)ptr << 3);
    z = 0x80 >> n;
    zz = ((ub & -z) | z) & 0xFF;
    u.buf[ptr>>3] = _mm256_set_epi64x( zz, zz, zz, zz );
@@ -1057,9 +1060,9 @@ blake64_4way_close( blake_4way_big_context *sc,
    {
         sc->T0 -= 1024 - bit_len;
    }
-   if ( ptr <= (96 >> 3) )
+   if ( ptr <= 104 )
    {
-       memset_zero_m256i( u.buf + (ptr>>3) + 1, (96-ptr) >> 3 );
+       memset_zero_m256i( u.buf + (ptr>>3) + 1, (104-ptr) >> 3 );
        if ( out_size_w64 == 8 )
           u.buf[(104>>3)] = _mm256_or_si256( u.buf[(104>>3)],
                                     _mm256_set_epi64x( 0x0100000000000000,
@@ -1070,11 +1073,13 @@ blake64_4way_close( blake_4way_big_context *sc,
                                     _mm256_set_epi64x( th, th, th, th ) );
        *(u.buf+(120>>3)) = mm256_byteswap_epi64(
                                     _mm256_set_epi64x( tl, tl, tl, tl ) );
+
        blake64_4way( sc, u.buf + (ptr>>3), 128 - ptr );
    }
    else
   {
-       memset_zero_m256i( u.buf + (ptr>>3) + 1, (127 - ptr) >> 3 );
+       memset_zero_m256i( u.buf + (ptr>>3) + 1, (120 - ptr) >> 3 );
+
        blake64_4way( sc, u.buf + (ptr>>3), 128 - ptr );
        sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00);
        sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFF);
@@ -1089,6 +1094,7 @@ blake64_4way_close( blake_4way_big_context *sc,
                                     _mm256_set_epi64x( th, th, th, th ) );
        *(u.buf+(120>>3)) = mm256_byteswap_epi64(
                                     _mm256_set_epi64x( tl, tl, tl, tl ) );
+
        blake64_4way( sc, u.buf, 128 );
    }
    out = (__m256i*)dst;

algo/blake/blake-hash-4way.h

@@ -62,9 +62,9 @@ extern "C"{
 #ifdef __AVX__
 typedef struct {
         __m128i buf[16] __attribute__ ((aligned (64)));
-	size_t ptr;
         __m128i H[8];
         __m128i S[4];    
+        size_t ptr;
 	sph_u32 T0, T1;
 } blake_4way_small_context;
 
@@ -82,13 +82,13 @@ void blake256_4way_addbits_and_close(
 
 typedef struct {
         __m256i buf[16] __attribute__ ((aligned (64)));
-	size_t ptr;
         __m256i H[8];
         __m256i S[4];   
+        size_t ptr;
 	sph_u64 T0, T1;
 } blake_4way_big_context;
 
-typedef blake_4way_big_context blake512_avx2_context;
+typedef blake_4way_big_context blake512_4way_context;
 
 void blake512_4way_init(void *cc);
 void blake512_4way(void *cc, const void *data, size_t len);