v3.9.4

INSTALL_LINUX

@@ -42,9 +42,6 @@ openssl 1.1.0e or higher. Add one of the following, depending on the
 compiler version, to CFLAGS:
 "-march=native" or "-march=znver1" or "-msha".
 
-Due to poor AVX2 performance on Ryzen users should add -DRYZEN_ to CFLAGS
-to override multiway AVX2 on algos with sha256, and use SHA instead.
-
 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.

Makefile.am

@@ -131,6 +131,7 @@ cpuminer_SOURCES = \
   algo/lyra2/lyra2h-4way.c \
   algo/lyra2/allium-4way.c \
   algo/lyra2/allium.c \
+  algo/lyra2/phi2-4way.c \
   algo/lyra2/phi2.c \
   algo/m7m.c \
   algo/neoscrypt/neoscrypt.c \
@@ -147,6 +148,9 @@ cpuminer_SOURCES = \
   algo/quark/anime-gate.c \
   algo/quark/anime.c \
   algo/quark/anime-4way.c \
+  algo/quark/hmq1725-gate.c \
+  algo/quark/hmq1725-4way.c \
+  algo/quark/hmq1725.c \
   algo/qubit/qubit-gate.c \
   algo/qubit/qubit.c \
   algo/qubit/qubit-2way.c \
@@ -257,7 +261,6 @@ cpuminer_SOURCES = \
   algo/x17/xevan-gate.c \
   algo/x17/xevan.c \
   algo/x17/xevan-4way.c \
-  algo/x17/hmq1725.c \
   algo/x17/sonoa-gate.c \
   algo/x17/sonoa-4way.c \
   algo/x17/sonoa.c \

README.txt

@@ -29,7 +29,7 @@ cpuminer-sse2.exe      "-msse2"                  Core2, Nehalem
 cpuminer-aes-sse42.exe "-march=westmere"         Westmere
 cpuminer-avx.exe       "-march=corei7-avx"       Sandy-Ivybridge
 cpuminer-avx2.exe      "-march=core-avx2"        Haswell, Sky-Kaby-Coffeelake
-cpuminer-zen           "-march=znver1 -DRYZEN_"  Ryzen
+cpuminer-zen           "-march=znver1"           AMD Ryzen, Threadripper
 
 If you like this software feel free to donate:
 

RELEASE_NOTES

@@ -38,6 +38,25 @@ supported.
 Change Log
 ----------
 
+v3.9.4
+
+Faster AVX2 for lyra2v3, quark, anime.
+Fixed skein AVX2 regression (invalid shares since v3.9.0) and faster.
+Faster skein2 with 4way AVX2 enabled.
+Automatic SHA override on Ryzen CPUs, no need for -DRYZEN compile flag.
+Ongoing restructuring.
+
+v3.9.3.1
+
+Skipped v3.9.3 due to misidentification of v3.9.2.5 as v3.9.3.
+Fixed x16r algo 25% invalid share reject rate. The bug may have also
+affected other algos.
+
+v3.9.2.5
+
+Fixed 2 regressions: hodl AES detection, x16r invalid shares with AVX2.
+More restructuring.
+
 v3.9.2.4
 
 Yet another affinity fix. Hopefully the last one.

algo-gate-api.h

@@ -2,8 +2,7 @@
 #include <stdbool.h>
 #include <stdint.h>
 #include "miner.h"
-#include "avxdefs.h"
-#include "interleave.h"
+#include "simd-utils.h"
 
 /////////////////////////////
 ////

algo/argon2/argon2d/argon2d/core.c

@@ -112,7 +112,7 @@ int allocate_memory(const argon2_context *context, uint8_t **memory,
 void free_memory(const argon2_context *context, uint8_t *memory,
                  size_t num, size_t size) {
     size_t memory_size = num*size;
-    clear_internal_memory(memory, memory_size);
+//    clear_internal_memory(memory, memory_size);
     if (context->free_cbk) {
         (context->free_cbk)(memory, memory_size);
     } else {
@@ -137,7 +137,7 @@ void NOT_OPTIMIZED secure_wipe_memory(void *v, size_t n) {
 int FLAG_clear_internal_memory = 0;
 void clear_internal_memory(void *v, size_t n) {
   if (FLAG_clear_internal_memory && v) {
-    secure_wipe_memory(v, n);
+//    secure_wipe_memory(v, n);
   }
 }
 
@@ -559,7 +559,7 @@ void initial_hash(uint8_t *blockhash, argon2_context *context,
                        context->pwdlen);
 
         if (context->flags & ARGON2_FLAG_CLEAR_PASSWORD) {
-            secure_wipe_memory(context->pwd, context->pwdlen);
+//            secure_wipe_memory(context->pwd, context->pwdlen);
             context->pwdlen = 0;
         }
     }
@@ -580,7 +580,7 @@ void initial_hash(uint8_t *blockhash, argon2_context *context,
                        context->secretlen);
 
         if (context->flags & ARGON2_FLAG_CLEAR_SECRET) {
-            secure_wipe_memory(context->secret, context->secretlen);
+//            secure_wipe_memory(context->secret, context->secretlen);
             context->secretlen = 0;
         }
     }

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

@@ -45,7 +45,7 @@ extern "C"{
 
 #include <stddef.h>
 #include "algo/sha/sph_types.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #define SPH_SIZE_blake256   256
 

algo/blake/blake2s-hash-4way.h

@@ -16,7 +16,7 @@
 
 #if defined(__SSE4_2__)
 
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #include <stddef.h>
 #include <stdint.h>

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

@@ -43,7 +43,7 @@ extern "C"{
 #include <stddef.h>
 
 #include "algo/sha/sph_types.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #define SPH_SIZE_bmw256   256
 

algo/bmw/bmw256-hash-4way.c

@@ -537,6 +537,8 @@ bmw32_4way(bmw_4way_small_context *sc, const void *data, size_t len)
       }
    }
    sc->ptr = ptr;
+
+
    if ( h1 != sc->H )
         memcpy_128( sc->H, h1, 16 );
 }
@@ -571,6 +573,7 @@ bmw32_4way_close(bmw_4way_small_context *sc, unsigned ub, unsigned n,
 
    for ( u = 0; u < 16; u ++ )
       buf[u] = h2[u];
+
    compress_small( buf, (__m128i*)final_s, h1 );
 
    for (u = 0, v = 16 - out_size_w32; u < out_size_w32; u ++, v ++)
@@ -1041,22 +1044,22 @@ static const __m256i final_s8[16] =
 
 void bmw256_8way_init( bmw256_8way_context *ctx )
 {
-   ctx->H[ 0] = _mm256_set1_epi64x( IV256[ 0] );
-   ctx->H[ 1] = _mm256_set1_epi64x( IV256[ 1] );
-   ctx->H[ 2] = _mm256_set1_epi64x( IV256[ 2] );
-   ctx->H[ 3] = _mm256_set1_epi64x( IV256[ 3] );
-   ctx->H[ 4] = _mm256_set1_epi64x( IV256[ 4] );
-   ctx->H[ 5] = _mm256_set1_epi64x( IV256[ 5] );
-   ctx->H[ 6] = _mm256_set1_epi64x( IV256[ 6] );
-   ctx->H[ 7] = _mm256_set1_epi64x( IV256[ 7] );
-   ctx->H[ 8] = _mm256_set1_epi64x( IV256[ 8] );
-   ctx->H[ 9] = _mm256_set1_epi64x( IV256[ 9] );
-   ctx->H[10] = _mm256_set1_epi64x( IV256[10] );
-   ctx->H[11] = _mm256_set1_epi64x( IV256[11] );
-   ctx->H[12] = _mm256_set1_epi64x( IV256[12] );
-   ctx->H[13] = _mm256_set1_epi64x( IV256[13] );
-   ctx->H[14] = _mm256_set1_epi64x( IV256[14] );
-   ctx->H[15] = _mm256_set1_epi64x( IV256[15] );
+   ctx->H[ 0] = _mm256_set1_epi32( IV256[ 0] );
+   ctx->H[ 1] = _mm256_set1_epi32( IV256[ 1] );
+   ctx->H[ 2] = _mm256_set1_epi32( IV256[ 2] );
+   ctx->H[ 3] = _mm256_set1_epi32( IV256[ 3] );
+   ctx->H[ 4] = _mm256_set1_epi32( IV256[ 4] );
+   ctx->H[ 5] = _mm256_set1_epi32( IV256[ 5] );
+   ctx->H[ 6] = _mm256_set1_epi32( IV256[ 6] );
+   ctx->H[ 7] = _mm256_set1_epi32( IV256[ 7] );
+   ctx->H[ 8] = _mm256_set1_epi32( IV256[ 8] );
+   ctx->H[ 9] = _mm256_set1_epi32( IV256[ 9] );
+   ctx->H[10] = _mm256_set1_epi32( IV256[10] );
+   ctx->H[11] = _mm256_set1_epi32( IV256[11] );
+   ctx->H[12] = _mm256_set1_epi32( IV256[12] );
+   ctx->H[13] = _mm256_set1_epi32( IV256[13] );
+   ctx->H[14] = _mm256_set1_epi32( IV256[14] );
+   ctx->H[15] = _mm256_set1_epi32( IV256[15] );
    ctx->ptr       = 0;
    ctx->bit_count = 0;
 
@@ -1076,14 +1079,15 @@ void bmw256_8way( bmw256_8way_context *ctx, const void *data, size_t len )
    ptr = ctx->ptr;
    h1 = ctx->H;
    h2 = htmp;
+
    while ( len > 0 )
    {
       size_t clen;
       clen = buf_size - ptr;
       if ( clen > len )
          clen = len;
-      memcpy_256( buf + (ptr>>3), vdata, clen >> 3 );
-      vdata = vdata + (clen>>3);
+      memcpy_256( buf + (ptr>>2), vdata, clen >> 2 );
+      vdata = vdata + (clen>>2);
       len -= clen;
       ptr += clen;
       if ( ptr == buf_size )
@@ -1097,6 +1101,7 @@ void bmw256_8way( bmw256_8way_context *ctx, const void *data, size_t len )
       }
    }
    ctx->ptr = ptr;
+
    if ( h1 != ctx->H )
         memcpy_256( ctx->H, h1, 16 );
 }
@@ -1106,24 +1111,26 @@ void bmw256_8way_close( bmw256_8way_context *ctx, void *dst )
    __m256i *buf;
    __m256i h1[16], h2[16], *h;
    size_t ptr, u, v;
-//   unsigned z;
    const int buf_size = 64;  // bytes of one lane, compatible with len
 
    buf = ctx->buf;
    ptr = ctx->ptr;
-   buf[ ptr>>3 ] = _mm256_set1_epi32( 0x80 );
-   ptr += 8;
+   buf[ ptr>>2 ] = _mm256_set1_epi32( 0x80 );
+   ptr += 4;
    h = ctx->H;
 
-   if (  ptr > (buf_size - 8) )
+   if (  ptr > (buf_size - 4) )
    {
-      memset_zero_256( buf + (ptr>>3), (buf_size - ptr) >> 3 );
+      memset_zero_256( buf + (ptr>>2), (buf_size - ptr) >> 2 );
       compress_small_8way( buf, h, h1 );
       ptr = 0;
       h = h1;
    }
-   memset_zero_256( buf + (ptr>>3), (buf_size - 8 - ptr) >> 3 );
-   buf[ (buf_size - 8) >> 3 ] = _mm256_set1_epi64x( ctx->bit_count );
+   memset_zero_256( buf + (ptr>>2), (buf_size - 8 - ptr) >> 2 );
+   buf[ (buf_size - 8) >> 2 ] = _mm256_set1_epi32( ctx->bit_count );
+   buf[ (buf_size - 4) >> 2 ] = m256_zero;
+
+
    compress_small_8way( buf, h, h2 );
 
    for ( u = 0; u < 16; u ++ )

algo/cubehash/cube-hash-2way.h

@@ -4,7 +4,7 @@
 #if defined(__AVX2__)
 
 #include <stdint.h>
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 // 2x128, 2 way parallel SSE2
 

algo/cubehash/cubehash_sse2.c

@@ -13,7 +13,7 @@
 #include <stdbool.h>
 #include <unistd.h>
 #include <memory.h>
-#include "avxdefs.h"
+#include "simd-utils.h"
 #include <stdio.h>
 
 // The result of hashing 10 rounds of initial data which is params and 

algo/groestl/aes_ni/hash-groestl.c

@@ -12,7 +12,7 @@
 #include <memory.h>
 #include "hash-groestl.h"
 #include "miner.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #ifndef NO_AES_NI
 

algo/groestl/aes_ni/hash-groestl256.c

@@ -9,7 +9,7 @@
 #include <memory.h>
 #include "hash-groestl256.h"
 #include "miner.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #ifndef NO_AES_NI
 

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

@@ -40,7 +40,7 @@
 
 #if defined (__AVX2__)
 
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #ifdef __cplusplus
 extern "C"{

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

@@ -69,7 +69,7 @@ extern "C"{
 
 #include <stddef.h>
 #include "algo/sha/sph_types.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #define SPH_SIZE_haval256_5   256
 

algo/hodl/hodl-gate.c

@@ -156,7 +156,7 @@ int hodl_scanhash( int thr_id, struct work* work, uint32_t max_nonce,
 
 bool register_hodl_algo( algo_gate_t* gate )
 {
-#if defined(__AES__)
+#if !defined(__AES__)
   applog( LOG_ERR, "Only CPUs with AES are supported, use legacy version.");
   return false;
 #endif

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

@@ -44,7 +44,7 @@ extern "C"{
 
 #include <stddef.h>
 #include "algo/sha/sph_types.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #define SPH_SIZE_jh256   256
 

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

@@ -44,7 +44,7 @@ extern "C"{
 
 #include <stddef.h>
 #include "algo/sha/sph_types.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #define SPH_SIZE_keccak256   256
 

algo/luffa/luffa-hash-2way.c

@@ -24,7 +24,7 @@
 
 #if defined(__AVX2__)
 
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #define MASK _mm256_set_epi32( 0UL, 0UL, 0UL, 0xffffffffUL, \
                                0UL, 0UL, 0UL, 0xffffffffUL )

algo/luffa/luffa-hash-2way.h

@@ -24,7 +24,7 @@
 
 #include <immintrin.h>
 #include "algo/sha/sha3-defs.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 /* The length of digests*/
 #define DIGEST_BIT_LEN_224 224

algo/luffa/luffa_for_sse2.c

@@ -20,7 +20,7 @@
 
 #include <string.h>
 #include <emmintrin.h>
-#include "avxdefs.h"
+#include "simd-utils.h"
 #include "luffa_for_sse2.h"
 
 #define MULT2(a0,a1) do \

algo/lyra2/allium-4way.c

@@ -44,10 +44,11 @@ void allium_4way_hash( void *state, const void *input )
    blake256_4way( &ctx.blake, input + (64<<2), 16 );
    blake256_4way_close( &ctx.blake, vhash32 );
 
-   mm256_reinterleave_4x64( vhash64, vhash32, 256 );
+   mm256_rintrlv_4x32_4x64( vhash64, vhash32, 256 );
    keccak256_4way( &ctx.keccak, vhash64, 32 );
    keccak256_4way_close( &ctx.keccak, vhash64 );
-   mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
+
+   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
 
    LYRA2RE( hash0, 32, hash0, 32, hash0, 32, 1, 8, 8 );
    LYRA2RE( hash1, 32, hash1, 32, hash1, 32, 1, 8, 8 );
@@ -67,26 +68,23 @@ void allium_4way_hash( void *state, const void *input )
    LYRA2RE( hash2, 32, hash2, 32, hash2, 32, 1, 8, 8 );
    LYRA2RE( hash3, 32, hash3, 32, hash3, 32, 1, 8, 8 );
 
-   mm256_interleave_4x64( vhash64, hash0, hash1, hash2, hash3, 256 );
+   mm256_intrlv_4x64( vhash64, hash0, hash1, hash2, hash3, 256 );
+
    skein256_4way( &ctx.skein, vhash64, 32 );
    skein256_4way_close( &ctx.skein, vhash64 );
-   mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
 
-   update_and_final_groestl256( &ctx.groestl, hash0, hash0, 256 );
-   memcpy( &ctx.groestl, &allium_4way_ctx.groestl,
-           sizeof(hashState_groestl256) );
-   update_and_final_groestl256( &ctx.groestl, hash1, hash1, 256 );
-   memcpy( &ctx.groestl, &allium_4way_ctx.groestl,
-           sizeof(hashState_groestl256) );
-   update_and_final_groestl256( &ctx.groestl, hash2, hash2, 256 );
-   memcpy( &ctx.groestl, &allium_4way_ctx.groestl,
-           sizeof(hashState_groestl256) );
-   update_and_final_groestl256( &ctx.groestl, hash3, hash3, 256 );
+   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
 
-   memcpy( state,    hash0, 32 );
-   memcpy( state+32, hash1, 32 );
-   memcpy( state+64, hash2, 32 );
-   memcpy( state+96, hash3, 32 );
+   update_and_final_groestl256( &ctx.groestl, state, hash0, 256 );
+   memcpy( &ctx.groestl, &allium_4way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, state+32, hash1, 256 );
+   memcpy( &ctx.groestl, &allium_4way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, state+64, hash2, 256 );
+   memcpy( &ctx.groestl, &allium_4way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, state+96, hash3, 256 );
 }
 
 int scanhash_allium_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -94,7 +92,6 @@ int scanhash_allium_4way( int thr_id, struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[8*4] __attribute__ ((aligned (64)));
    uint32_t vdata[20*4] __attribute__ ((aligned (64)));
-   uint32_t _ALIGN(64) edata[20];
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t first_nonce = pdata[19];
@@ -106,13 +103,7 @@ int scanhash_allium_4way( int thr_id, struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ( (uint32_t*)ptarget )[7] = 0x0000ff;
 
-   casti_m128i( edata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
-   casti_m128i( edata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
-   casti_m128i( edata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
-   casti_m128i( edata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
-   casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
-
-   mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
+   mm128_bswap_intrlv80_4x32( vdata, pdata );
    blake256_4way_init( &allium_4way_ctx.blake );
    blake256_4way( &allium_4way_ctx.blake, vdata, 64 );
 
@@ -124,7 +115,7 @@ int scanhash_allium_4way( int thr_id, struct work *work, uint32_t max_nonce,
 
      for ( int lane = 0; lane < 4; lane++ ) if ( (hash+(lane<<3))[7] <= Htarg )
      {
-        if ( fulltest( hash+(lane<<3), ptarget ) )
+        if ( fulltest( hash+(lane<<3), ptarget ) && !opt_benchmark )
         {
            pdata[19] = n + lane;
            submit_solution( work, hash+(lane<<3), mythr, lane );

algo/lyra2/lyra2-gate.c

@@ -47,7 +47,9 @@ bool lyra2rev3_thread_init()
 
    int size = (int64_t)ROW_LEN_BYTES * 4; // nRows;
    l2v3_wholeMatrix = _mm_malloc( size, 64 );
-#if defined (LYRA2REV3_4WAY)
+#if defined (LYRA2REV3_8WAY)
+   init_lyra2rev3_8way_ctx();;
+#elif defined (LYRA2REV3_4WAY)
    init_lyra2rev3_4way_ctx();;
 #else
    init_lyra2rev3_ctx();
@@ -57,7 +59,10 @@ bool lyra2rev3_thread_init()
 
 bool register_lyra2rev3_algo( algo_gate_t* gate )
 {
-#if defined (LYRA2REV3_4WAY)
+#if defined (LYRA2REV3_8WAY)
+  gate->scanhash  = (void*)&scanhash_lyra2rev3_8way;
+  gate->hash      = (void*)&lyra2rev3_8way_hash;
+#elif defined (LYRA2REV3_4WAY)
   gate->scanhash  = (void*)&scanhash_lyra2rev3_4way;
   gate->hash      = (void*)&lyra2rev3_4way_hash;
 #else
@@ -203,13 +208,18 @@ void phi2_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
 
 bool register_phi2_algo( algo_gate_t* gate )
 {
-   init_phi2_ctx();
+//   init_phi2_ctx();
    gate->optimizations = SSE2_OPT | AES_OPT | SSE42_OPT | AVX2_OPT;
    gate->get_work_data_size = (void*)&phi2_get_work_data_size;
    gate->decode_extra_data  = (void*)&phi2_decode_extra_data;
    gate->build_extraheader  = (void*)&phi2_build_extraheader;
    gate->set_target         = (void*)&alt_set_target; 
    gate->get_max64          = (void*)&get_max64_0xffffLL;
+#if defined(PHI2_4WAY)
+   gate->scanhash           = (void*)&scanhash_phi2_4way;
+#else
+   init_phi2_ctx();
    gate->scanhash           = (void*)&scanhash_phi2;
+#endif
    return true;
 }

algo/lyra2/lyra2-gate.h

@@ -6,14 +6,24 @@
 #include "lyra2.h"
 
 #if defined(__AVX2__)
+  #define LYRA2REV3_8WAY
+#endif
+
+#if defined(__SSE2__)
   #define LYRA2REV3_4WAY
 #endif
 
 extern __thread uint64_t* l2v3_wholeMatrix;
 
 bool register_lyra2rev3_algo( algo_gate_t* gate );
+#if defined(LYRA2REV3_8WAY)
 
-#if defined(LYRA2REV3_4WAY)
+void lyra2rev3_8way_hash( void *state, const void *input );
+int scanhash_lyra2rev3_8way( int thr_id, struct work *work, uint32_t max_nonce,
+                             uint64_t *hashes_done, struct thr_info *mythr );
+bool init_lyra2rev3_8way_ctx();
+
+#elif defined(LYRA2REV3_4WAY)
 
 void lyra2rev3_4way_hash( void *state, const void *input );
 int scanhash_lyra2rev3_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -140,15 +150,29 @@ bool init_allium_ctx();
 
 /////////////////////////////////////////
 
+#if defined(__AVX2__) && defined(__AES__)
+//  #define PHI2_4WAY
+#endif
+
 bool phi2_has_roots;
 
 bool register_phi2_algo( algo_gate_t* gate );
+#if defined(PHI2_4WAY)
+
+void phi2_hash_4way( void *state, const void *input );
+int scanhash_phi2_4way( int thr_id, struct work *work, uint32_t max_nonce,
+                     uint64_t *hashes_done, struct thr_info *mythr );
+//void init_phi2_ctx();
+
+#else
 
 void phi2_hash( void *state, const void *input );
 int scanhash_phi2( int thr_id, struct work *work, uint32_t max_nonce,
                      uint64_t *hashes_done, struct thr_info *mythr );
 void init_phi2_ctx();
 
+#endif
+
 #endif  // LYRA2_GATE_H__
 
 

algo/lyra2/lyra2.c

@@ -566,7 +566,7 @@ int LYRA2RE( void *K, uint64_t kLen, const void *pwd, const uint64_t pwdlen,
 
 #if defined(__AVX2__)
    memset_zero_256( (__m256i*)wholeMatrix, i>>5 );
-#elif defined(__SSE4_2__)
+#elif defined(__SSE2__)
    memset_zero_128( (__m128i*)wholeMatrix, i>>4 );   
 #else
    memset( wholeMatrix, 0, i );

algo/lyra2/lyra2h-4way.c

@@ -36,17 +36,16 @@ void lyra2h_4way_hash( void *state, const void *input )
      blake256_4way( &ctx_blake, input + (64*4), 16 );
      blake256_4way_close( &ctx_blake, vhash );
 
-     mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
+     mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
 
-     LYRA2Z( lyra2h_4way_matrix, hash0, 32, hash0, 32, hash0, 32, 16, 16, 16 );
-     LYRA2Z( lyra2h_4way_matrix, hash1, 32, hash1, 32, hash1, 32, 16, 16, 16 );
-     LYRA2Z( lyra2h_4way_matrix, hash2, 32, hash2, 32, hash2, 32, 16, 16, 16 );
-     LYRA2Z( lyra2h_4way_matrix, hash3, 32, hash3, 32, hash3, 32, 16, 16, 16 );
-
-     memcpy( state,    hash0, 32 );
-     memcpy( state+32, hash1, 32 );
-     memcpy( state+64, hash2, 32 );
-     memcpy( state+96, hash3, 32 );
+     LYRA2Z( lyra2h_4way_matrix, state, 32, hash0, 32, hash0, 32,
+             16, 16, 16 );
+     LYRA2Z( lyra2h_4way_matrix, state+32, 32, hash1, 32, hash1,
+             32, 16, 16, 16 );
+     LYRA2Z( lyra2h_4way_matrix, state+64, 32, hash2, 32, hash2,
+             32, 16, 16, 16 );
+     LYRA2Z( lyra2h_4way_matrix, state+96, 32, hash3, 32, hash3,
+             32, 16, 16, 16 );
 }
 
 int scanhash_lyra2h_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -54,49 +53,36 @@ int scanhash_lyra2h_4way( int thr_id, struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[8*4] __attribute__ ((aligned (64)));
    uint32_t vdata[20*4] __attribute__ ((aligned (64)));
-   uint32_t _ALIGN(64) edata[20];
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t Htarg = ptarget[7];
    const uint32_t first_nonce = pdata[19];
    uint32_t n = first_nonce;
-   uint32_t *nonces = work->nonces;
-   int num_found = 0;
-   uint32_t *noncep= vdata + 76; // 19*4
+   __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
    /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
    if ( opt_benchmark )
       ptarget[7] = 0x0000ff;
 
-   for ( int i=0; i < 20; i++ )
-      be32enc( &edata[i], pdata[i] );
-
-   mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
-
+   mm128_bswap_intrlv80_4x32( vdata, pdata );
    lyra2h_4way_midstate( vdata );
 
    do {
-      be32enc( noncep,   n   );
-      be32enc( noncep+1, n+1 );
-      be32enc( noncep+2, n+2 );
-      be32enc( noncep+3, n+3 );
-
-      be32enc( &edata[19], n );
+     *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
       lyra2h_4way_hash( hash, vdata );
 
       for ( int i = 0; i < 4; i++ )
-      if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
+      if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget )
+           && !opt_benchmark )
       {
           pdata[19] = n+i;         
-          nonces[ num_found++ ] = n+i;
-          work_set_target_ratio( work, hash+(i<<3) );
+          submit_solution( work, hash+(i<<3), mythr, i );
       }
       n += 4;
-   } while ( (num_found == 0) && (n < max_nonce-4)
-                   && !work_restart[thr_id].restart);
+   } while (  (n < max_nonce-4) && !work_restart[thr_id].restart);
 
    *hashes_done = n - first_nonce + 1;
-   return num_found;
+   return 0;
 }
 
 #endif

algo/lyra2/lyra2re.c

@@ -6,7 +6,7 @@
 #include "algo/keccak/sph_keccak.h"
 #include "lyra2.h"
 #include "algo-gate-api.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 #if defined(__AES__)
   #include "algo/groestl/aes_ni/hash-groestl256.h"
 #endif

algo/lyra2/lyra2rev2-4way.c

@@ -42,10 +42,12 @@ void lyra2rev2_4way_hash( void *state, const void *input )
    blake256_4way( &ctx.blake, input + (64<<2), 16 );
    blake256_4way_close( &ctx.blake, vhash );
 
-   mm256_reinterleave_4x64( vhash64, vhash, 256 );
+   mm256_rintrlv_4x32_4x64( vhash64, vhash, 256 );
+
    keccak256_4way( &ctx.keccak, vhash64, 32 );
    keccak256_4way_close( &ctx.keccak, vhash64 );
-   mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
+
+   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
 
    cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
    cubehashInit( &ctx.cube, 256, 16, 32 );
@@ -60,10 +62,12 @@ void lyra2rev2_4way_hash( void *state, const void *input )
    LYRA2REV2( l2v2_wholeMatrix, hash2, 32, hash2, 32, hash2, 32, 1, 4, 4 );
    LYRA2REV2( l2v2_wholeMatrix, hash3, 32, hash3, 32, hash3, 32, 1, 4, 4 );
 
-   mm256_interleave_4x64( vhash64, hash0, hash1, hash2, hash3, 256 );
+   mm256_intrlv_4x64( vhash64, hash0, hash1, hash2, hash3, 256 );
+
    skein256_4way( &ctx.skein, vhash64, 32 );
    skein256_4way_close( &ctx.skein, vhash64 );
-   mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
+
+   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
 
    cubehashInit( &ctx.cube, 256, 16, 32 );
    cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
@@ -74,11 +78,10 @@ void lyra2rev2_4way_hash( void *state, const void *input )
    cubehashInit( &ctx.cube, 256, 16, 32 );
    cubehashUpdateDigest( &ctx.cube, (byte*) hash3, (const byte*) hash3, 32 );
 
-   mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 256 );
-   bmw256_4way( &ctx.bmw, vhash, 32 );
-   bmw256_4way_close( &ctx.bmw, vhash );
+   mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 256 );
 
-   mm128_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
+   bmw256_4way( &ctx.bmw, vhash, 32 );
+   bmw256_4way_close( &ctx.bmw, state );
 }
 
 int scanhash_lyra2rev2_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -86,49 +89,44 @@ int scanhash_lyra2rev2_4way( int thr_id, struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[8*4] __attribute__ ((aligned (64)));
    uint32_t vdata[20*4] __attribute__ ((aligned (64)));
-   uint32_t _ALIGN(64) edata[20];
+   uint32_t *hash7 = &(hash[7<<2]);
+   uint32_t lane_hash[8];
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t first_nonce = pdata[19];
    uint32_t n = first_nonce;
    const uint32_t Htarg = ptarget[7];
-   uint32_t *nonces = work->nonces;
-   int num_found = 0;
-   uint32_t *noncep = vdata + 76; // 19*4
+   __m128i *noncev = (__m128i*)vdata + 19;   // aligned
    /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
    if ( opt_benchmark )
       ( (uint32_t*)ptarget )[7] = 0x0000ff;
 
-   swab32_array( edata, pdata, 20 );
-
-   mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
+   mm128_bswap_intrlv80_4x32( vdata, pdata );
 
    blake256_4way_init( &l2v2_4way_ctx.blake );
    blake256_4way( &l2v2_4way_ctx.blake, vdata, 64 );
 
-   do {
-      be32enc( noncep,   n   );
-      be32enc( noncep+1, n+1 );
-      be32enc( noncep+2, n+2 );
-      be32enc( noncep+3, n+3 );
+   do
+   {
+      *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
 
       lyra2rev2_4way_hash( hash, vdata );
       pdata[19] = n;
 
-      for ( int i = 0; i < 4; i++ )
-      if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
+      for ( int lane = 0; lane < 4; lane++ ) if ( hash7[lane] <= Htarg )
       {
-          pdata[19] = n+i;         
-          nonces[ num_found++ ] = n+i;
-          work_set_target_ratio( work, hash+(i<<3) );
+         mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
+         if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+         {
+            pdata[19] = n + lane;         
+            submit_solution( work, lane_hash, mythr, lane );
+         }
       }
       n += 4;
-   } while ( (num_found == 0) && (n < max_nonce-4)
-                   && !work_restart[thr_id].restart);
-
+   } while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
    *hashes_done = n - first_nonce + 1;
-   return num_found;
+   return 0;
 }
 
 #endif

algo/lyra2/lyra2rev3-4way.c

@@ -1,12 +1,138 @@
 #include "lyra2-gate.h"
 #include <memory.h>
 
-#if defined (LYRA2REV3_4WAY)	
-
 #include "algo/blake/blake-hash-4way.h"
 #include "algo/bmw/bmw-hash-4way.h"
 #include "algo/cubehash/cubehash_sse2.h" 
 
+
+#if defined (LYRA2REV3_8WAY)
+
+typedef struct {
+   blake256_8way_context     blake;
+   cubehashParam             cube;
+   bmw256_8way_context       bmw;
+} lyra2v3_8way_ctx_holder;
+
+static lyra2v3_8way_ctx_holder l2v3_8way_ctx;
+
+bool init_lyra2rev3_8way_ctx()
+{
+   blake256_8way_init( &l2v3_8way_ctx.blake );
+   cubehashInit( &l2v3_8way_ctx.cube, 256, 16, 32 );
+   bmw256_8way_init( &l2v3_8way_ctx.bmw );
+   return true;
+}
+
+void lyra2rev3_8way_hash( void *state, const void *input )
+{
+   uint32_t vhash[8*8] __attribute__ ((aligned (64)));
+   uint32_t hash0[8] __attribute__ ((aligned (64)));
+   uint32_t hash1[8] __attribute__ ((aligned (32)));
+   uint32_t hash2[8] __attribute__ ((aligned (32)));
+   uint32_t hash3[8] __attribute__ ((aligned (32)));
+   uint32_t hash4[8] __attribute__ ((aligned (32)));
+   uint32_t hash5[8] __attribute__ ((aligned (32)));
+   uint32_t hash6[8] __attribute__ ((aligned (32)));
+   uint32_t hash7[8] __attribute__ ((aligned (32)));
+   lyra2v3_8way_ctx_holder ctx __attribute__ ((aligned (64)));
+   memcpy( &ctx, &l2v3_8way_ctx, sizeof(l2v3_8way_ctx) );
+
+   blake256_8way( &ctx.blake, input, 80 );
+   blake256_8way_close( &ctx.blake, vhash );
+
+   mm256_dintrlv_8x32( hash0, hash1, hash2, hash3,
+                       hash4, hash5, hash6, hash7, vhash, 256 );
+
+   LYRA2REV3( l2v3_wholeMatrix, hash0, 32, hash0, 32, hash0, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash1, 32, hash1, 32, hash1, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash2, 32, hash2, 32, hash2, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash3, 32, hash3, 32, hash3, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash4, 32, hash4, 32, hash4, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash5, 32, hash5, 32, hash5, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash6, 32, hash6, 32, hash6, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash7, 32, hash7, 32, hash7, 32, 1, 4, 4 );
+
+   cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
+   cubehashInit( &ctx.cube, 256, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*) hash1, (const byte*) hash1, 32 );
+   cubehashInit( &ctx.cube, 256, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*) hash2, (const byte*) hash2, 32 );
+   cubehashInit( &ctx.cube, 256, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*) hash3, (const byte*) hash3, 32 );
+   cubehashInit( &ctx.cube, 256, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*) hash4, (const byte*) hash4, 32 );
+   cubehashInit( &ctx.cube, 256, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*) hash5, (const byte*) hash5, 32 );
+   cubehashInit( &ctx.cube, 256, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*) hash6, (const byte*) hash6, 32 );
+   cubehashInit( &ctx.cube, 256, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*) hash7, (const byte*) hash7, 32 );
+
+   LYRA2REV3( l2v3_wholeMatrix, hash0, 32, hash0, 32, hash0, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash1, 32, hash1, 32, hash1, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash2, 32, hash2, 32, hash2, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash3, 32, hash3, 32, hash3, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash4, 32, hash4, 32, hash4, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash5, 32, hash5, 32, hash5, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash6, 32, hash6, 32, hash6, 32, 1, 4, 4 );
+   LYRA2REV3( l2v3_wholeMatrix, hash7, 32, hash7, 32, hash7, 32, 1, 4, 4 );
+
+   mm256_intrlv_8x32( vhash, hash0, hash1, hash2, hash3,
+                             hash4, hash5, hash6, hash7, 256 );
+
+   bmw256_8way( &ctx.bmw, vhash, 32 );
+   bmw256_8way_close( &ctx.bmw, state );
+
+   }
+
+int scanhash_lyra2rev3_8way( int thr_id, struct work *work, uint32_t max_nonce,
+                             uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t hash[8*8] __attribute__ ((aligned (64)));
+   uint32_t vdata[20*8] __attribute__ ((aligned (64)));
+   uint32_t *hash7 = &(hash[7<<3]);
+   uint32_t lane_hash[8];
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   const uint32_t first_nonce = pdata[19];
+   uint32_t n = first_nonce;
+   const uint32_t Htarg = ptarget[7];
+   __m256i  *noncev = (__m256i*)vdata + 19;   // aligned
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
+
+   if ( opt_benchmark )
+      ( (uint32_t*)ptarget )[7] = 0x0000ff;
+
+   mm256_bswap_intrlv80_8x32( vdata, pdata );
+   do
+   {
+      *noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
+                                                  n+3, n+2, n+1, n ) );
+
+      lyra2rev3_8way_hash( hash, vdata );
+      pdata[19] = n;
+
+      for ( int lane = 0; lane < 8; lane++ ) if ( hash7[lane] <= Htarg )
+      {
+         mm256_extract_lane_8x32( lane_hash, hash, lane, 256 );
+         if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+         {
+              pdata[19] = n + lane;
+              submit_solution( work, lane_hash, mythr, lane );
+         }
+      }
+      n += 8;
+   } while ( (n < max_nonce-8) && !work_restart[thr_id].restart);
+   *hashes_done = n - first_nonce + 1;
+   return 0;
+}
+
+#endif
+
+#if defined (LYRA2REV3_4WAY)  
+
+
 typedef struct {
    blake256_4way_context     blake;
    cubehashParam             cube;
@@ -35,7 +161,7 @@ void lyra2rev3_4way_hash( void *state, const void *input )
 
    blake256_4way( &ctx.blake, input, 80 );
    blake256_4way_close( &ctx.blake, vhash );
-   mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
+   mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
 
    LYRA2REV3( l2v3_wholeMatrix, hash0, 32, hash0, 32, hash0, 32, 1, 4, 4 );
    LYRA2REV3( l2v3_wholeMatrix, hash1, 32, hash1, 32, hash1, 32, 1, 4, 4 );
@@ -55,10 +181,9 @@ void lyra2rev3_4way_hash( void *state, const void *input )
    LYRA2REV3( l2v3_wholeMatrix, hash2, 32, hash2, 32, hash2, 32, 1, 4, 4 );
    LYRA2REV3( l2v3_wholeMatrix, hash3, 32, hash3, 32, hash3, 32, 1, 4, 4 );
 
-   mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 256 );
+   mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 256 );
    bmw256_4way( &ctx.bmw, vhash, 32 );
    bmw256_4way_close( &ctx.bmw, state );
-
 }
 
 int scanhash_lyra2rev3_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -66,7 +191,6 @@ int scanhash_lyra2rev3_4way( int thr_id, struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[8*4] __attribute__ ((aligned (64)));
    uint32_t vdata[20*4] __attribute__ ((aligned (64)));
-   uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t *hash7 = &(hash[7<<2]);
    uint32_t lane_hash[8];
    uint32_t *pdata = work->data;
@@ -80,15 +204,7 @@ int scanhash_lyra2rev3_4way( int thr_id, struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ( (uint32_t*)ptarget )[7] = 0x0000ff;
 
-   // Need big endian data
-   casti_m128i( edata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
-   casti_m128i( edata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
-   casti_m128i( edata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
-   casti_m128i( edata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
-   casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
-
-   mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
-
+   mm128_bswap_intrlv80_4x32( vdata, pdata );
    do
    {
       *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
@@ -99,16 +215,14 @@ int scanhash_lyra2rev3_4way( int thr_id, struct work *work, uint32_t max_nonce,
       for ( int lane = 0; lane < 4; lane++ ) if ( hash7[lane] <= Htarg )
       {
          mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
-
-         if ( fulltest( lane_hash, ptarget ) )
+         if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
          {
               pdata[19] = n + lane;    
               submit_solution( work, lane_hash, mythr, lane );
-	 }
+	      }
       }
       n += 4;
    } while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
-
    *hashes_done = n - first_nonce + 1;
    return 0;
 }

algo/lyra2/lyra2z-4way.c

@@ -36,17 +36,12 @@ void lyra2z_4way_hash( void *state, const void *input )
      blake256_4way( &ctx_blake, input + (64*4), 16 );
      blake256_4way_close( &ctx_blake, vhash );
 
-     mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
+     mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
 
-     LYRA2Z( lyra2z_4way_matrix, hash0, 32, hash0, 32, hash0, 32, 8, 8, 8 );
-     LYRA2Z( lyra2z_4way_matrix, hash1, 32, hash1, 32, hash1, 32, 8, 8, 8 );
-     LYRA2Z( lyra2z_4way_matrix, hash2, 32, hash2, 32, hash2, 32, 8, 8, 8 );
-     LYRA2Z( lyra2z_4way_matrix, hash3, 32, hash3, 32, hash3, 32, 8, 8, 8 );
-
-     memcpy( state,    hash0, 32 );
-     memcpy( state+32, hash1, 32 );
-     memcpy( state+64, hash2, 32 );
-     memcpy( state+96, hash3, 32 );
+     LYRA2Z( lyra2z_4way_matrix, state   , 32, hash0, 32, hash0, 32, 8, 8, 8 );
+     LYRA2Z( lyra2z_4way_matrix, state+32, 32, hash1, 32, hash1, 32, 8, 8, 8 );
+     LYRA2Z( lyra2z_4way_matrix, state+64, 32, hash2, 32, hash2, 32, 8, 8, 8 );
+     LYRA2Z( lyra2z_4way_matrix, state+96, 32, hash3, 32, hash3, 32, 8, 8, 8 );
 }
 
 int scanhash_lyra2z_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -54,7 +49,6 @@ int scanhash_lyra2z_4way( int thr_id, struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[8*4] __attribute__ ((aligned (64)));
    uint32_t vdata[20*4] __attribute__ ((aligned (64)));
-   uint32_t _ALIGN(64) edata[20];
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t Htarg = ptarget[7];
@@ -66,13 +60,7 @@ int scanhash_lyra2z_4way( int thr_id, struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ptarget[7] = 0x0000ff;
 
-   casti_m128i( edata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
-   casti_m128i( edata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
-   casti_m128i( edata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
-   casti_m128i( edata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
-   casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
-   mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
-
+   mm128_bswap_intrlv80_4x32( vdata, pdata );
    lyra2z_4way_midstate( vdata );
 
    do {
@@ -82,16 +70,11 @@ int scanhash_lyra2z_4way( int thr_id, struct work *work, uint32_t max_nonce,
       pdata[19] = n;
 
       for ( int i = 0; i < 4; i++ )
-      if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
+      if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget )
+           && !opt_benchmark )
       {
           pdata[19] = n+i;         
-          work_set_target_ratio( work, hash+(i<<3) );
-          if ( submit_work( mythr, work ) )
-              applog( LOG_NOTICE, "Share %d submitted by thread %d, lane %d.",
-                             accepted_share_count + rejected_share_count + 1,
-                             thr_id, i );
-          else
-              applog( LOG_WARNING, "Failed to submit share." );
+          submit_solution( work, hash+(i<<3), mythr, i );
       }
       n += 4;
    } while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
@@ -136,8 +119,8 @@ void lyra2z_8way_hash( void *state, const void *input )
      blake256_8way( &ctx_blake, input + (64*8), 16 );
      blake256_8way_close( &ctx_blake, vhash );
 
-     mm256_deinterleave_8x32( hash0, hash1, hash2, hash3,
-                              hash4, hash5, hash6, hash7, vhash, 256 );
+     mm256_dintrlv_8x32( hash0, hash1, hash2, hash3,
+                         hash4, hash5, hash6, hash7, vhash, 256 );
 
      LYRA2Z( lyra2z_8way_matrix, hash0, 32, hash0, 32, hash0, 32, 8, 8, 8 );
      LYRA2Z( lyra2z_8way_matrix, hash1, 32, hash1, 32, hash1, 32, 8, 8, 8 );
@@ -163,7 +146,6 @@ int scanhash_lyra2z_8way( int thr_id, struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[8*8] __attribute__ ((aligned (64)));
    uint32_t vdata[20*8] __attribute__ ((aligned (64)));
-   uint32_t _ALIGN(64) edata[20];
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t Htarg = ptarget[7];
@@ -175,13 +157,7 @@ int scanhash_lyra2z_8way( int thr_id, struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ptarget[7] = 0x0000ff;
 
-   casti_m256i( edata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
-   casti_m256i( edata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
-   casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
-
-   mm256_interleave_8x32( vdata, edata, edata, edata, edata,
-                                 edata, edata, edata, edata, 640 );
-
+   mm256_bswap_intrlv80_8x32( vdata, pdata );
    lyra2z_8way_midstate( vdata );
 
    do {
@@ -191,7 +167,8 @@ int scanhash_lyra2z_8way( int thr_id, struct work *work, uint32_t max_nonce,
       pdata[19] = n;
 
       for ( int i = 0; i < 8; i++ )
-      if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
+      if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget )
+           && !opt_benchmark )
       {
           pdata[19] = n+i;         
           submit_solution( work, hash+(i<<3), mythr, i );

algo/lyra2/lyra2z.c

@@ -3,7 +3,7 @@
 #include "lyra2-gate.h"
 #include "lyra2.h"
 #include "algo/blake/sph_blake.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 __thread uint64_t* lyra2z_matrix;
 

algo/lyra2/lyra2z330.c

@@ -1,7 +1,7 @@
 #include <memory.h>
 #include "algo-gate-api.h"
 #include "lyra2.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 __thread uint64_t* lyra2z330_wholeMatrix;
 
@@ -30,14 +30,17 @@ int scanhash_lyra2z330( int thr_id, struct work *work, uint32_t max_nonce,
    if (opt_benchmark)
 	ptarget[7] = 0x0000ff;
 
-   for (int i=0; i < 19; i++)
-      be32enc(&endiandata[i], pdata[i]);
-        
+   casti_m128i( endiandata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
+   casti_m128i( endiandata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
+   casti_m128i( endiandata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
+   casti_m128i( endiandata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
+   casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
+   
    do
    {
       be32enc(&endiandata[19], nonce);
       lyra2z330_hash( hash, endiandata, work->height );
-      if ( hash[7] <= Htarg && fulltest(hash, ptarget) )
+      if ( hash[7] <= Htarg && fulltest(hash, ptarget) && !opt_benchmark )
       {
          work_set_target_ratio(work, hash);
          pdata[19] = nonce;

algo/lyra2/phi2-4way.c

@@ -0,0 +1,233 @@
+/**
+ * Phi-2 algo Implementation
+ */
+
+#include "lyra2-gate.h"
+
+#if defined(PHI2_4WAY)
+
+#include "algo/skein/skein-hash-4way.h"
+#include "algo/jh/jh-hash-4way.h"
+#include "algo/gost/sph_gost.h"
+#include "algo/cubehash/cubehash_sse2.h"
+#include "algo/echo/aes_ni/hash_api.h"
+
+typedef struct {
+     cubehashParam           cube;
+     jh512_4way_context      jh;
+     hashState_echo          echo;
+//     hashState_echo          echo2;
+     sph_gost512_context     gost;
+     skein512_4way_context   skein;
+} phi2_ctx_holder;
+/*
+phi2_ctx_holder phi2_ctx;
+
+void init_phi2_ctx()
+{
+   cubehashInit( &phi2_ctx.cube, 512, 16, 32 );
+   sph_jh512_init(&phi2_ctx.jh);
+   init_echo( &phi2_ctx.echo1, 512 );
+   init_echo( &phi2_ctx.echo2, 512 );
+   sph_gost512_init(&phi2_ctx.gost);
+   sph_skein512_init(&phi2_ctx.skein);
+};
+*/
+void phi2_hash_4way( void *state, const void *input )
+{
+   uint32_t hash[4][16] __attribute__ ((aligned (64)));
+   uint32_t hashA[4][16] __attribute__ ((aligned (64)));
+   uint32_t hashB[4][16] __attribute__ ((aligned (64)));
+   uint32_t vhash[4*16] __attribute__ ((aligned (64)));
+
+//   unsigned char _ALIGN(128) hash[64];
+//	unsigned char _ALIGN(128) hashA[64];
+//	unsigned char _ALIGN(128) hashB[64];
+
+   phi2_ctx_holder ctx __attribute__ ((aligned (64)));
+//  memcpy( &ctx, &phi2_ctx, sizeof(phi2_ctx) );
+
+   cubehashInit( &ctx.cube, 512, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*)hashB[0], (const byte*)input,
+                        phi2_has_roots ? 144 : 80 );
+   cubehashInit( &ctx.cube, 512, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*)hashB[1], (const byte*)input+144,
+                        phi2_has_roots ? 144 : 80 );
+   cubehashInit( &ctx.cube, 512, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*)hashB[2], (const byte*)input+288,
+                        phi2_has_roots ? 144 : 80 );
+   cubehashInit( &ctx.cube, 512, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*)hashB[3], (const byte*)input+432,
+                        phi2_has_roots ? 144 : 80 );
+
+	LYRA2RE( &hashA[0][0], 32, &hashB[0][0], 32, &hashB[0][0], 32, 1, 8, 8 );
+	LYRA2RE( &hashA[0][8], 32, &hashB[0][8], 32, &hashB[0][8], 32, 1, 8, 8 );
+   LYRA2RE( &hashA[1][0], 32, &hashB[1][0], 32, &hashB[1][0], 32, 1, 8, 8 );
+   LYRA2RE( &hashA[1][8], 32, &hashB[1][8], 32, &hashB[1][8], 32, 1, 8, 8 );
+   LYRA2RE( &hashA[2][0], 32, &hashB[2][0], 32, &hashB[2][0], 32, 1, 8, 8 );
+   LYRA2RE( &hashA[2][8], 32, &hashB[2][8], 32, &hashB[2][8], 32, 1, 8, 8 );
+   LYRA2RE( &hashA[3][0], 32, &hashB[3][0], 32, &hashB[3][0], 32, 1, 8, 8 );
+   LYRA2RE( &hashA[3][8], 32, &hashB[3][8], 32, &hashB[3][8], 32, 1, 8, 8 );
+
+   mm256_intrlv_4x64( vhash, hashA[0], hashA[1], hashA[2], hashA[3], 512 );
+
+   jh512_4way_init( &ctx.jh );
+   jh512_4way( &ctx.jh, vhash, 64 );
+   jh512_4way_close( &ctx.jh, vhash );
+
+   mm256_dintrlv_4x64( hash[0], hash[1], hash[2], hash[3], vhash, 512 );
+
+   if ( hash[0][0] & 1 )
+  	{
+      sph_gost512_init( &ctx.gost );
+      sph_gost512( &ctx.gost, (const void*)hash[0], 64 );
+	   sph_gost512_close( &ctx.gost, (void*)hash[0] );
+	}
+  	else
+  	{
+      init_echo( &ctx.echo, 512 );
+      update_final_echo ( &ctx.echo, (BitSequence *)hash[0],
+                          (const BitSequence *)hash[0], 512 );
+      init_echo( &ctx.echo, 512 );
+      update_final_echo ( &ctx.echo, (BitSequence *)hash[0],
+                          (const BitSequence *)hash[0], 512 );
+	}
+
+   if ( hash[1][0] & 1 )
+   {
+      sph_gost512_init( &ctx.gost );
+      sph_gost512( &ctx.gost, (const void*)hash[1], 64 );
+      sph_gost512_close( &ctx.gost, (void*)hash[1] );
+   }
+   else
+   {
+      init_echo( &ctx.echo, 512 );
+      update_final_echo ( &ctx.echo, (BitSequence *)hash[1],
+                          (const BitSequence *)hash[1], 512 );
+      init_echo( &ctx.echo, 512 );
+      update_final_echo ( &ctx.echo, (BitSequence *)hash[1],
+                          (const BitSequence *)hash[1], 512 );
+   }
+
+   if ( hash[2][0] & 1 )
+   {
+      sph_gost512_init( &ctx.gost );
+      sph_gost512( &ctx.gost, (const void*)hash[2], 64 );
+      sph_gost512_close( &ctx.gost, (void*)hash[2] );
+   }
+   else
+   {
+      init_echo( &ctx.echo, 512 );
+      update_final_echo ( &ctx.echo, (BitSequence *)hash[2],
+                          (const BitSequence *)hash[2], 512 );
+      init_echo( &ctx.echo, 512 );
+      update_final_echo ( &ctx.echo, (BitSequence *)hash[2],
+                          (const BitSequence *)hash[2], 512 );
+   }
+
+   if ( hash[3][0] & 1 )
+   {
+      sph_gost512_init( &ctx.gost );
+      sph_gost512( &ctx.gost, (const void*)hash[3], 64 );
+      sph_gost512_close( &ctx.gost, (void*)hash[3] );
+   }
+   else
+   {
+      init_echo( &ctx.echo, 512 );
+      update_final_echo ( &ctx.echo, (BitSequence *)hash[3],
+                          (const BitSequence *)hash[3], 512 );
+      init_echo( &ctx.echo, 512 );
+      update_final_echo ( &ctx.echo, (BitSequence *)hash[3],
+                          (const BitSequence *)hash[3], 512 );
+   }
+
+   mm256_intrlv_4x64( vhash, hash[0], hash[1], hash[2], hash[3], 512 );
+   
+   skein512_4way_init( &ctx.skein );
+	skein512_4way( &ctx.skein, vhash, 64 );
+	skein512_4way_close( &ctx.skein, vhash );
+
+   for (int i=0; i<4; i++)
+   {
+      ( (uint64_t*)vhash    )[i] ^= ( (uint64_t*)vhash    )[i+4];
+      ( (uint64_t*)vhash+ 8 )[i] ^= ( (uint64_t*)vhash+ 8 )[i+4];
+      ( (uint64_t*)vhash+16 )[i] ^= ( (uint64_t*)vhash+16 )[i+4];
+      ( (uint64_t*)vhash+24 )[i] ^= ( (uint64_t*)vhash+24 )[i+4];
+   }
+//   for ( int i = 0; i < 4; i++ )
+//      casti_m256i( vhash, i ) = _mm256_xor_si256( casti_m256i( vhash, i   ),
+//                                                  casti_m256i( vhash, i+4 ) );
+
+	memcpy( state, vhash, 128 );
+}
+
+int scanhash_phi2_4way( int thr_id, struct work *work, uint32_t max_nonce,
+	                     uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t _ALIGN(128) hash[8];
+   uint32_t _ALIGN(128) edata[36];
+   uint32_t vdata[4][36] __attribute__ ((aligned (64)));
+   uint32_t *hash7 = &(hash[25]);
+   uint32_t lane_hash[8];
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   const uint32_t Htarg = ptarget[7];
+   const uint32_t first_nonce = pdata[19];
+   uint32_t n = first_nonce;
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
+
+   if(opt_benchmark){
+   	ptarget[7] = 0x00ff;
+   }
+
+// Data is not interleaved, but hash is.
+// any non-zero data at index 20 or above sets roots true.
+// Split up the operations, bswap first, then set roots.
+
+   phi2_has_roots = false;
+   for ( int i=0; i < 36; i++ )
+   {
+   be32enc(&edata[i], pdata[i]);
+   if (i >= 20 && pdata[i]) phi2_has_roots = true;
+   }
+/*
+   casti_m256i( vdata[0], 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );   
+   casti_m256i( vdata[0], 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
+   casti_m256i( vdata[0], 2 ) = mm256_bswap_32( casti_m256i( pdata, 2 ) );
+   casti_m256i( vdata[0], 3 ) = mm256_bswap_32( casti_m256i( pdata, 3 ) );
+   casti_m128i( vdata[0], 8 ) = mm128_bswap_32( casti_m128i( pdata, 8 ) );
+   phi2_has_roots = mm128_anybits1( casti_m128i( vdata[0], 5 ) ) ||
+                    mm128_anybits1( casti_m128i( vdata[0], 6 ) ) ||
+                    mm128_anybits1( casti_m128i( vdata[0], 7 ) ) ||
+                    mm128_anybits1( casti_m128i( vdata[0], 8 ) );
+*/   
+
+   memcpy( vdata[0], edata, 144 );
+   memcpy( vdata[1], edata, 144 );
+   memcpy( vdata[2], edata, 144 );
+   memcpy( vdata[3], edata, 144 );
+
+   do {
+      be32enc( &vdata[0][19], n );
+      be32enc( &vdata[1][19], n+1 );
+      be32enc( &vdata[2][19], n+2 );
+      be32enc( &vdata[3][19], n+3 );
+
+      phi2_hash_4way( hash, vdata );
+
+      for ( int lane = 0; lane < 4; lane++ ) if (  hash7[ lane<<1 ] < Htarg )
+      {
+          mm256_extract_lane_4x64( lane_hash, hash, lane, 256 );
+          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+          {
+              pdata[19] = n + lane;
+              submit_solution( work, lane_hash, mythr, lane );
+          }
+       }
+       n += 4;
+    } while ( ( n < max_nonce - 4 ) && !work_restart[thr_id].restart );
+    *hashes_done = n - first_nonce + 1;
+    return 0;
+}
+   
+#endif  // PHI2_4WAY

algo/lyra2/phi2.c

@@ -50,11 +50,11 @@ void phi2_hash(void *state, const void *input)
 	unsigned char _ALIGN(128) hashA[64];
 	unsigned char _ALIGN(128) hashB[64];
 
-        phi2_ctx_holder ctx __attribute__ ((aligned (64)));
-        memcpy( &ctx, &phi2_ctx, sizeof(phi2_ctx) );
+  phi2_ctx_holder ctx __attribute__ ((aligned (64)));
+  memcpy( &ctx, &phi2_ctx, sizeof(phi2_ctx) );
 
-        cubehashUpdateDigest( &ctx.cube, (byte*)hashB, (const byte*)input,
-		              phi2_has_roots ? 144 : 80 );
+  cubehashUpdateDigest( &ctx.cube, (byte*)hashB, (const byte*)input,
+                        phi2_has_roots ? 144 : 80 );
 
 	LYRA2RE( &hashA[ 0], 32, &hashB[ 0], 32, &hashB[ 0], 32, 1, 8, 8 );
 	LYRA2RE( &hashA[32], 32, &hashB[32], 32, &hashB[32], 32, 1, 8, 8 );
@@ -63,17 +63,17 @@ void phi2_hash(void *state, const void *input)
 	sph_jh512_close( &ctx.jh, (void*)hash );
 
 	if ( hash[0] & 1 )
-       	{
-           sph_gost512( &ctx.gost, (const void*)hash, 64 );
+  	{
+      sph_gost512( &ctx.gost, (const void*)hash, 64 );
 	   sph_gost512_close( &ctx.gost, (void*)hash );
 	}
-       	else
-       	{
+  	else
+  	{
 #if defined(__AES__)
-           update_final_echo ( &ctx.echo1, (BitSequence *)hash,
-                               (const BitSequence *)hash, 512 );
-           update_final_echo ( &ctx.echo2, (BitSequence *)hash,
-                               (const BitSequence *)hash, 512 );
+      update_final_echo ( &ctx.echo1, (BitSequence *)hash,
+                          (const BitSequence *)hash, 512 );
+      update_final_echo ( &ctx.echo2, (BitSequence *)hash,
+                          (const BitSequence *)hash, 512 );
 #else
 	   sph_echo512( &ctx.echo1, (const void*)hash, 64 );
 	   sph_echo512_close( &ctx.echo1, (void*)hash );

algo/lyra2/sponge.c

@@ -51,7 +51,7 @@ inline void initState( uint64_t State[/*16*/] )
   state[3] = _mm256_set_epi64x( blake2b_IV[7], blake2b_IV[6],
                                 blake2b_IV[5], blake2b_IV[4] );
 
-#elif defined (__SSE4_2__)
+#elif defined (__SSE2__)
 
   __m128i* state = (__m128i*)State;
 
@@ -137,7 +137,7 @@ inline void squeeze( uint64_t *State, byte *Out, unsigned int len )
     //Squeezes remaining bytes
     memcpy_256( out, state, ( len_m256i % BLOCK_LEN_M256I ) );
 
-#elif defined (__SSE4_2__)
+#elif defined (__SSE2__)
 
     const int len_m128i = len / 16;
     const int fullBlocks = len_m128i / BLOCK_LEN_M128I;
@@ -205,7 +205,7 @@ inline void absorbBlock( uint64_t *State, const uint64_t *In )
     _mm256_store_si256( (__m256i*)State + 2, state2 );
     _mm256_store_si256( (__m256i*)State + 3, state3 );
 
-#elif defined (__SSE4_2__)
+#elif defined (__SSE2__)
 
     __m128i* state = (__m128i*)State;
     __m128i* in    = (__m128i*)In;
@@ -273,7 +273,7 @@ inline void absorbBlockBlake2Safe( uint64_t *State, const uint64_t *In )
     _mm256_store_si256( (__m256i*)State + 2, state2 );
     _mm256_store_si256( (__m256i*)State + 3, state3 );
 
-#elif defined (__SSE4_2__)
+#elif defined (__SSE2__)
 
     __m128i* state = (__m128i*)State;
     __m128i* in    = (__m128i*)In;
@@ -355,7 +355,7 @@ inline void reducedSqueezeRow0( uint64_t* State, uint64_t* rowOut,
     _mm256_store_si256( (__m256i*)State + 2, state2 );
     _mm256_store_si256( (__m256i*)State + 3, state3 );
 
-#elif defined (__SSE4_2__)
+#elif defined (__SSE2__)
 
     __m128i* state = (__m128i*)State;
     __m128i  state0 = _mm_load_si128(  state    );
@@ -494,7 +494,7 @@ inline void reducedDuplexRow1( uint64_t *State, uint64_t *rowIn,
     _mm256_store_si256( (__m256i*)State + 2, state2 );
     _mm256_store_si256( (__m256i*)State + 3, state3 );
 
-#elif defined (__SSE4_2__)
+#elif defined (__SSE2__)
 
     __m128i* state = (__m128i*)State;
     __m128i  state0 = _mm_load_si128(  state    );
@@ -694,7 +694,7 @@ inline void reducedDuplexRowSetup( uint64_t *State, uint64_t *rowIn,
     _mm256_store_si256( (__m256i*)State + 2, state2 );
     _mm256_store_si256( (__m256i*)State + 3, state3 );
 
-#elif defined (__SSE4_2__)
+#elif defined (__SSE2__)
 
     __m128i* in    = (__m128i*)rowIn;
     __m128i* inout = (__m128i*)rowInOut;
@@ -713,9 +713,9 @@ inline void reducedDuplexRowSetup( uint64_t *State, uint64_t *rowIn,
     __m128i* state = (__m128i*)State;
 
     // For the last round in this function not optimized for AVX
-    uint64_t* ptrWordIn = rowIn;        //In Lyra2: pointer to prev
-    uint64_t* ptrWordInOut = rowInOut;  //In Lyra2: pointer to row*
-    uint64_t* ptrWordOut = rowOut + (nCols-1)*BLOCK_LEN_INT64; //In Lyra2: pointer to row
+//    uint64_t* ptrWordIn = rowIn;        //In Lyra2: pointer to prev
+//    uint64_t* ptrWordInOut = rowInOut;  //In Lyra2: pointer to row*
+//    uint64_t* ptrWordOut = rowOut + (nCols-1)*BLOCK_LEN_INT64; //In Lyra2: pointer to row
 
     for ( i = 0; i < nCols; i++ )
     {
@@ -750,6 +750,28 @@ inline void reducedDuplexRowSetup( uint64_t *State, uint64_t *rowIn,
         out[4] = _mm_xor_si128( state[4], in[4] );
         out[5] = _mm_xor_si128( state[5], in[5] );
 
+
+       __m128i t0, t1;
+       t0 = _mm_srli_si128( state[0], 8 );
+       t1 = _mm_srli_si128( state[1], 8 );
+       inout[0] = _mm_xor_si128( inout[0],
+                              _mm_or_si128( _mm_slli_si128( state[0], 8 ),
+                                            _mm_srli_si128( state[5], 8 ) ) );
+       inout[1] = _mm_xor_si128( inout[1],
+                        _mm_or_si128( _mm_slli_si128( state[1], 8 ), t0 ) );
+       t0 = _mm_srli_si128( state[2], 8 );
+       inout[2] = _mm_xor_si128( inout[2],
+                        _mm_or_si128( _mm_slli_si128( state[2], 8 ), t1 ) );
+       t1 = _mm_srli_si128( state[3], 8 );
+       inout[3] = _mm_xor_si128( inout[3],
+                        _mm_or_si128( _mm_slli_si128( state[3], 8 ), t0 ) );
+       t0 = _mm_srli_si128( state[4], 8 );
+       inout[4] = _mm_xor_si128( inout[4],
+                        _mm_or_si128( _mm_slli_si128( state[4], 8 ), t1 ) );
+       inout[5] = _mm_xor_si128( inout[5],
+                        _mm_or_si128( _mm_slli_si128( state[5], 8 ), t0 ) );
+
+/*
         ptrWordInOut[0]  ^= State[11];
         ptrWordInOut[1]  ^= State[0];
         ptrWordInOut[2]  ^= State[1];
@@ -768,7 +790,7 @@ inline void reducedDuplexRowSetup( uint64_t *State, uint64_t *rowIn,
         ptrWordIn += BLOCK_LEN_INT64;
         //Output: goes to previous column
         ptrWordOut -= BLOCK_LEN_INT64;
-
+*/
         inout += BLOCK_LEN_M128I;
         in    += BLOCK_LEN_M128I;
         out   -= BLOCK_LEN_M128I;
@@ -930,7 +952,7 @@ inline void reducedDuplexRow( uint64_t *State, uint64_t *rowIn,
    _mm256_store_si256( (__m256i*)State + 2, state2 );
    _mm256_store_si256( (__m256i*)State + 3, state3 );
 
-#elif defined(__SSE4_2__)
+#elif defined (__SSE2__)
 
     __m128i* state = (__m128i*)State;
     __m128i* in    = (__m128i*)rowIn;

algo/lyra2/sponge.h

@@ -23,7 +23,7 @@
 #define SPONGE_H_
 
 #include <stdint.h>
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #if defined(__GNUC__)
 #define ALIGN __attribute__ ((aligned(32)))
@@ -59,7 +59,7 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
 // returns void, updates all args
 #define G_4X64(a,b,c,d) \
    a = _mm256_add_epi64( a, b ); \
-   d = mm256_ror_64( _mm256_xor_si256( d, a), 32 ); \
+   d = mm256_ror_64( _mm256_xor_si256( d, a ), 32 ); \
    c = _mm256_add_epi64( c, d ); \
    b = mm256_ror_64( _mm256_xor_si256( b, c ), 24 ); \
    a = _mm256_add_epi64( a, b ); \

algo/m7m.c

@@ -144,8 +144,8 @@ void init_m7m_ctx()
 #define NM7M 5
 #define SW_DIVS 5
 #define M7_MIDSTATE_LEN 76
-int scanhash_m7m_hash( int thr_id, struct work* work,
-                       uint64_t max_nonce, unsigned long *hashes_done )
+int scanhash_m7m_hash( int thr_id, struct work* work, uint64_t max_nonce,
+                       unsigned long *hashes_done, struct thr_info *mythr )
 {
     uint32_t *pdata = work->data;
     uint32_t *ptarget = work->target;
@@ -154,6 +154,7 @@ int scanhash_m7m_hash( int thr_id, struct work* work,
     uint32_t hash[8] __attribute__((aligned(64)));
     uint8_t bhash[7][64] __attribute__((aligned(64)));
     uint32_t n = pdata[19] - 1;
+    /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
     uint32_t usw_, mpzscale;
     const uint32_t first_nonce = pdata[19];
     char data_str[161], hash_str[65], target_str[65];

algo/pluck.c

@@ -445,7 +445,7 @@ void pluck_hash(uint32_t *hash, const uint32_t *data, uchar *hashbuffer, const i
 }
 
 int scanhash_pluck(int thr_id, struct work *work, uint32_t max_nonce,
-        uint64_t *hashes_done  )
+        uint64_t *hashes_done, struct thr_info *mythr  )
 {
         uint32_t *pdata = work->data;
         uint32_t *ptarget = work->target;
@@ -454,6 +454,8 @@ int scanhash_pluck(int thr_id, struct work *work, uint32_t max_nonce,
 	const uint32_t first_nonce = pdata[19];
 	volatile uint8_t *restart = &(work_restart[thr_id].restart);
 	uint32_t n = first_nonce;
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
+
 
 	if (opt_benchmark)
 		((uint32_t*)ptarget)[7] = 0x0ffff;

algo/quark/anime-4way.c

@@ -48,8 +48,8 @@ void anime_4way_hash( void *state, const void *input )
     __m256i* vhA = (__m256i*)vhashA;
     __m256i* vhB = (__m256i*)vhashB;
     __m256i vh_mask;
+    const uint32_t mask = 8;
     const __m256i bit3_mask = _mm256_set1_epi64x( 8 );
-    int i;
     anime_4way_ctx_holder ctx;
     memcpy( &ctx, &anime_4way_ctx, sizeof(anime_4way_ctx) );
 
@@ -62,27 +62,44 @@ void anime_4way_hash( void *state, const void *input )
     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ),
                                   m256_zero );
 
-    mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
-    update_and_final_groestl( &ctx.groestl, (char*)hash0,
-                                            (char*)hash0, 512 );
-    reinit_groestl( &ctx.groestl );
-    update_and_final_groestl( &ctx.groestl, (char*)hash1,
-                                            (char*)hash1, 512 );
-    reinit_groestl( &ctx.groestl );
-    update_and_final_groestl( &ctx.groestl, (char*)hash2,
-                                            (char*)hash2, 512 );
-    reinit_groestl( &ctx.groestl );
-    update_and_final_groestl( &ctx.groestl, (char*)hash3,
-                                            (char*)hash3, 512 );
-    mm256_interleave_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+    mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
-    skein512_4way( &ctx.skein, vhash, 64 );
-    skein512_4way_close( &ctx.skein, vhashB );
+    if ( hash0[0] & mask )
+    {
+       update_and_final_groestl( &ctx.groestl, (char*)hash0,
+                                               (char*)hash0, 512 );
+    }
+    if ( hash1[0] & mask )
+    {
+       reinit_groestl( &ctx.groestl );
+       update_and_final_groestl( &ctx.groestl, (char*)hash1,
+                                               (char*)hash1, 512 );
+    }
+    if ( hash2[0] & mask )
+    {
+       reinit_groestl( &ctx.groestl );
+       update_and_final_groestl( &ctx.groestl, (char*)hash2,
+                                               (char*)hash2, 512 );
+    }
+    if ( hash3[0] & mask )
+    {
+       reinit_groestl( &ctx.groestl );
+       update_and_final_groestl( &ctx.groestl, (char*)hash3,
+                                               (char*)hash3, 512 );
+    }
 
-    for ( i = 0; i < 8; i++ )
-       vh[i] = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );
+    mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+
+    if ( mm256_anybits0( vh_mask ) )
+    {
+       skein512_4way( &ctx.skein, vhash, 64 );
+       skein512_4way_close( &ctx.skein, vhashB );
+    }
+
+    mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
+
+    mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
-    mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
     reinit_groestl( &ctx.groestl );
     update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
     reinit_groestl( &ctx.groestl );
@@ -91,7 +108,8 @@ void anime_4way_hash( void *state, const void *input )
     update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
     reinit_groestl( &ctx.groestl );
     update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
-    mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+
+    mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
     jh512_4way( &ctx.jh, vhash, 64 );
     jh512_4way_close( &ctx.jh, vhash );
@@ -99,16 +117,20 @@ void anime_4way_hash( void *state, const void *input )
     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ),
                                   m256_zero );
 
+    if ( mm256_anybits1( vh_mask ) )
+    {
        blake512_4way_init( &ctx.blake );
        blake512_4way( &ctx.blake, vhash, 64 );
        blake512_4way_close( &ctx.blake, vhashA );
-
+    }
+    if ( mm256_anybits0( vh_mask ) )
+    {
        bmw512_4way_init( &ctx.bmw );
        bmw512_4way( &ctx.bmw, vhash, 64 );
        bmw512_4way_close( &ctx.bmw, vhashB );
+    }
 
-    for ( i = 0; i < 8; i++ )
-       vh[i] = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );
+    mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
     keccak512_4way( &ctx.keccak, vhash, 64 );
     keccak512_4way_close( &ctx.keccak, vhash );
@@ -120,33 +142,35 @@ void anime_4way_hash( void *state, const void *input )
     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ),
                                   m256_zero );
 
-    keccak512_4way_init( &ctx.keccak );
-    keccak512_4way( &ctx.keccak, vhash, 64 );
-    keccak512_4way_close( &ctx.keccak, vhashA );
+    if ( mm256_anybits1( vh_mask ) )
+    {
+       keccak512_4way_init( &ctx.keccak );
+       keccak512_4way( &ctx.keccak, vhash, 64 );
+       keccak512_4way_close( &ctx.keccak, vhashA );
+    }
+    if ( mm256_anybits0( vh_mask ) )
+    {
+       jh512_4way_init( &ctx.jh );
+       jh512_4way( &ctx.jh, vhash, 64 );
+       jh512_4way_close( &ctx.jh, vhashB );
+    }
 
-    jh512_4way_init( &ctx.jh );
-    jh512_4way( &ctx.jh, vhash, 64 );
-    jh512_4way_close( &ctx.jh, vhashB );
+    mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
-    for ( i = 0; i < 8; i++ )
-       vh[i] = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );
-
-    mm256_deinterleave_4x64( state, state+32, state+64, state+96, vhash, 256 );
+    mm256_dintrlv_4x64( state, state+32, state+64, state+96, vhash, 256 );
 }
 
 int scanhash_anime_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                         uint64_t *hashes_done)
+                         uint64_t *hashes_done, struct thr_info *mythr )
 {
     uint32_t hash[4*8] __attribute__ ((aligned (64)));
     uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-    uint32_t endiandata[20] __attribute__((aligned(64)));
     uint32_t *pdata = work->data;
     uint32_t *ptarget = work->target;
     uint32_t n = pdata[19];
     const uint32_t first_nonce = pdata[19];
-    uint32_t *nonces = work->nonces;
-    int num_found = 0;
-    uint32_t *noncep = vdata + 73;   // 9*8 + 1
+    __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
+    /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
     const uint32_t Htarg = ptarget[7];
     uint64_t htmax[] = {
                 0,
@@ -165,10 +189,7 @@ int scanhash_anime_4way( int thr_id, struct work *work, uint32_t max_nonce,
                 0
         };
 
-    swab32_array( endiandata, pdata, 20 );
-
-    uint64_t *edata = (uint64_t*)endiandata;
-    mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+    mm256_bswap_intrlv80_4x64( vdata, pdata );
 
     for (int m=0; m < 6; m++)
        if (Htarg <= htmax[m])
@@ -177,30 +198,26 @@ int scanhash_anime_4way( int thr_id, struct work *work, uint32_t max_nonce,
 
           do
           {
-             be32enc( noncep,   n   );
-             be32enc( noncep+2, n+1 );
-             be32enc( noncep+4, n+2 );
-             be32enc( noncep+6, n+3 );
+             *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
+                _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
 
              anime_4way_hash( hash, vdata );
              pdata[19] = n;
 
              for ( int i = 0; i < 4; i++ )
              if ( ( ( (hash+(i<<3))[7] & mask ) == 0 )
-                && fulltest( hash+(i<<3), ptarget ) )
+                && fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
              {
                 pdata[19] = n+i;
-                nonces[ num_found++ ] = n+i;
-                work_set_target_ratio( work, hash+(i<<3) );
+                submit_solution( work, hash+(i<<3), mythr, i );
              }
              n += 4;
-          } while ( ( num_found == 0 ) && ( n < max_nonce )
-              && !work_restart[thr_id].restart );
+          } while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
           break;
        }
 
     *hashes_done = n - first_nonce + 1;
-    return num_found;
+    return 0;
 }
 
 #endif

algo/quark/anime-gate.h

@@ -13,19 +13,15 @@ bool register_anime_algo( algo_gate_t* gate );
 #if defined(ANIME_4WAY)
 
 void anime_4way_hash( void *state, const void *input );
-
 int scanhash_anime_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                         uint64_t *hashes_done );
-
+                         uint64_t *hashes_done, struct thr_info *mythr );
 void init_anime_4way_ctx();
 
 #endif
 
 void anime_hash( void *state, const void *input );
-
 int scanhash_anime( int thr_id, struct work *work, uint32_t max_nonce,
-                    uint64_t *hashes_done );
-
+                    uint64_t *hashes_done, struct thr_info *mythr );
 void init_anime_ctx();
 
 #endif

algo/quark/anime.c

@@ -46,20 +46,6 @@ void init_anime_ctx()
 void anime_hash( void *state, const void *input )
 {
     unsigned char hash[128] __attribute__ ((aligned (32)));
-/*
-    uint64_t hash0[8] __attribute__ ((aligned (64)));
-    uint64_t hash1[8] __attribute__ ((aligned (64)));
-    uint64_t hash2[8] __attribute__ ((aligned (64)));
-    uint64_t hash3[8] __attribute__ ((aligned (64)));
-    uint64_t vhash[8*4] __attribute__ ((aligned (64)));
-    uint64_t vhashA[8*4] __attribute__ ((aligned (64)));
-    uint64_t vhashB[8*4] __attribute__ ((aligned (64)));
-    __m256i* vh  = (__m256i*)vhash;
-    __m256i* vhA = (__m256i*)vhashA;
-    __m256i* vhB = (__m256i*)vhashB;
-    __m256i vh_mask;
-    __m256i bit3_mask; bit3_mask = _mm256_set1_epi64x( 8 );
-*/
     uint32_t mask = 8;
     anime_ctx_holder ctx;
     memcpy( &ctx, &anime_ctx, sizeof(anime_ctx) );
@@ -134,7 +120,7 @@ void anime_hash( void *state, const void *input )
 }
 
 int scanhash_anime( int thr_id, struct work *work, uint32_t max_nonce,
-                         uint64_t *hashes_done)
+                         uint64_t *hashes_done, struct thr_info *mythr)
 {
     uint32_t hash[8] __attribute__ ((aligned (64)));
     uint32_t endiandata[20] __attribute__((aligned(64)));
@@ -142,6 +128,7 @@ int scanhash_anime( int thr_id, struct work *work, uint32_t max_nonce,
     uint32_t *ptarget = work->target;
     uint32_t n = pdata[19];
     const uint32_t first_nonce = pdata[19];
+    /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
     const uint32_t Htarg = ptarget[7];
     uint64_t htmax[] = {
                 0,

algo/quark/hmq1725-4way.c

@@ -0,0 +1,618 @@
+#include "hmq1725-gate.h"
+
+#if defined(HMQ1725_4WAY)
+
+#include <string.h>
+#include <stdint.h>
+#include "algo/blake/blake-hash-4way.h"
+#include "algo/bmw/bmw-hash-4way.h"
+#include "algo/groestl/aes_ni/hash-groestl.h"
+#include "algo/skein/skein-hash-4way.h"
+#include "algo/jh/jh-hash-4way.h"
+#include "algo/keccak/keccak-hash-4way.h"
+#include "algo/luffa/luffa_for_sse2.h"
+#include "algo/cubehash/cubehash_sse2.h"
+#include "algo/simd/nist.h"
+#include "algo/shavite/sph_shavite.h"
+#include "algo/simd/simd-hash-2way.h"
+#include "algo/echo/aes_ni/hash_api.h"
+#include "algo/hamsi/hamsi-hash-4way.h"
+#include "algo/fugue/sph_fugue.h"
+#include "algo/shabal/shabal-hash-4way.h"
+#include "algo/whirlpool/sph_whirlpool.h"
+#include "algo/haval/haval-hash-4way.h"
+#include "algo/sha/sha2-hash-4way.h"
+
+union _hmq1725_4way_context_overlay
+{
+    blake512_4way_context   blake;
+    bmw512_4way_context     bmw;
+    hashState_groestl       groestl;
+    skein512_4way_context   skein;
+    jh512_4way_context      jh;
+    keccak512_4way_context  keccak;
+    hashState_luffa         luffa;
+    cubehashParam           cube;
+    sph_shavite512_context  shavite;
+    hashState_sd            simd;
+    hashState_echo          echo;
+    hamsi512_4way_context   hamsi;
+    sph_fugue512_context    fugue;
+    shabal512_4way_context  shabal;
+    sph_whirlpool_context   whirlpool;
+    sha512_4way_context     sha512;
+    haval256_5_4way_context haval;
+};
+typedef union _hmq1725_4way_context_overlay hmq1725_4way_context_overlay;
+
+extern void hmq1725_4way_hash(void *state, const void *input)
+{
+// why so big? only really need 8, haval thing uses 16.
+     uint32_t hash0 [32]    __attribute__ ((aligned (64)));
+     uint32_t hash1 [32]    __attribute__ ((aligned (64)));
+     uint32_t hash2 [32]    __attribute__ ((aligned (64)));
+     uint32_t hash3 [32]    __attribute__ ((aligned (64)));
+     uint32_t vhash [32<<2] __attribute__ ((aligned (64)));
+     uint32_t vhashA[32<<2] __attribute__ ((aligned (64)));
+     uint32_t vhashB[32<<2] __attribute__ ((aligned (64)));
+     hmq1725_4way_context_overlay ctx __attribute__ ((aligned (64)));
+     __m256i vh_mask;     
+     const __m256i vmask = _mm256_set1_epi64x( 24 );
+     const uint32_t mask = 24;
+     __m256i* vh  = (__m256i*)vhash;
+     __m256i* vhA = (__m256i*)vhashA;
+     __m256i* vhB = (__m256i*)vhashB;
+
+     bmw512_4way_init( &ctx.bmw );
+     bmw512_4way( &ctx.bmw, input, 80 );
+     bmw512_4way_close( &ctx.bmw, vhash );
+
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+
+     sph_whirlpool_init( &ctx.whirlpool );
+     sph_whirlpool( &ctx.whirlpool, hash0, 64 );
+     sph_whirlpool_close( &ctx.whirlpool, hash0 );
+     sph_whirlpool_init( &ctx.whirlpool );
+     sph_whirlpool( &ctx.whirlpool, hash1, 64 );
+     sph_whirlpool_close( &ctx.whirlpool, hash1 );
+     sph_whirlpool_init( &ctx.whirlpool );
+     sph_whirlpool( &ctx.whirlpool, hash2, 64 );
+     sph_whirlpool_close( &ctx.whirlpool, hash2 );
+     sph_whirlpool_init( &ctx.whirlpool );
+     sph_whirlpool( &ctx.whirlpool, hash3, 64 );
+     sph_whirlpool_close( &ctx.whirlpool, hash3 );
+
+// first fork, A is groestl serial, B is skein parallel.
+
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+
+     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
+                                   m256_zero );
+
+// A
+
+//     if ( hash0[0] & mask )
+//     {
+       init_groestl( &ctx.groestl, 64 );
+       update_and_final_groestl( &ctx.groestl, (char*)hash0,
+                                               (char*)hash0, 512 );
+//     }
+//     if ( hash1[0] & mask )
+//     {
+       init_groestl( &ctx.groestl, 64 );
+       update_and_final_groestl( &ctx.groestl, (char*)hash1,
+                                               (char*)hash1, 512 );
+//     }
+//     if ( hash2[0] & mask )
+//     {
+       init_groestl( &ctx.groestl, 64 );
+       update_and_final_groestl( &ctx.groestl, (char*)hash2,
+                                               (char*)hash2, 512 );
+//     }
+//     if ( hash3[0] & mask )
+//     {
+       init_groestl( &ctx.groestl, 64 );
+       update_and_final_groestl( &ctx.groestl, (char*)hash3,
+                                               (char*)hash3, 512 );
+//     }
+
+     mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+
+// B
+
+//     if ( mm256_any_clr_256( vh_mask ) )
+//     {
+       skein512_4way_init( &ctx.skein );
+       skein512_4way( &ctx.skein, vhash, 64 );
+       skein512_4way_close( &ctx.skein, vhashB );
+//     }
+
+     mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
+
+     jh512_4way_init( &ctx.jh );
+     jh512_4way( &ctx.jh, vhash, 64 );
+     jh512_4way_close( &ctx.jh, vhash );
+
+     keccak512_4way_init( &ctx.keccak );
+     keccak512_4way( &ctx.keccak, vhash, 64 );
+     keccak512_4way_close( &ctx.keccak, vhash );
+
+// second fork, A = blake parallel, B= bmw parallel.
+    
+     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
+                                   m256_zero );
+
+//     if ( mm256_any_set_256( vh_mask ) )
+//     {
+       blake512_4way_init( &ctx.blake );
+       blake512_4way( &ctx.blake, vhash, 64 );
+       blake512_4way_close( &ctx.blake, vhashA );
+//     }
+
+//     if ( mm256_any_clr_256( vh_mask ) )
+//     {
+       bmw512_4way_init( &ctx.bmw );
+       bmw512_4way( &ctx.bmw, vhash, 64 );
+       bmw512_4way_close( &ctx.bmw, vhashB );
+//     }
+
+     mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
+    
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+
+     init_luffa( &ctx.luffa, 512 );
+     update_and_final_luffa( &ctx.luffa, (BitSequence*)hash0,
+                                   (const BitSequence*)hash0, 64 );
+     init_luffa( &ctx.luffa, 512 );
+     update_and_final_luffa( &ctx.luffa, (BitSequence*)hash1,
+                                   (const BitSequence*)hash1, 64 );
+     init_luffa( &ctx.luffa, 512 );
+     update_and_final_luffa( &ctx.luffa, (BitSequence*)hash2,
+                                   (const BitSequence*)hash2, 64 );
+     init_luffa( &ctx.luffa, 512 );
+     update_and_final_luffa( &ctx.luffa, (BitSequence*)hash3,
+                                   (const BitSequence*)hash3, 64 );
+
+     cubehashInit( &ctx.cube, 512, 16, 32 );
+     cubehashUpdateDigest( &ctx.cube, (BitSequence *)hash0,
+                                (const BitSequence *)hash0, 64 );
+     cubehashInit( &ctx.cube, 512, 16, 32 );
+     cubehashUpdateDigest( &ctx.cube, (BitSequence *)hash1,
+                                (const BitSequence *)hash1, 64 );
+     cubehashInit( &ctx.cube, 512, 16, 32 );
+     cubehashUpdateDigest( &ctx.cube, (BitSequence *)hash2,
+                                (const BitSequence *)hash2, 64 );
+     cubehashInit( &ctx.cube, 512, 16, 32 );
+     cubehashUpdateDigest( &ctx.cube, (BitSequence *)hash3,
+                                (const BitSequence *)hash3, 64 );
+
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+
+// A= keccak parallel, B= jh parallel
+    
+     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
+                                  m256_zero );
+
+//     if ( mm256_any_set_256( vh_mask ) )
+//     {
+        keccak512_4way_init( &ctx.keccak );
+        keccak512_4way( &ctx.keccak, vhash, 64 );
+        keccak512_4way_close( &ctx.keccak, vhashA );
+//     }
+
+//     if ( mm256_any_clr_256( vh_mask ) )
+//     {
+        jh512_4way_init( &ctx.jh );
+        jh512_4way( &ctx.jh, vhash, 64 );
+        jh512_4way_close( &ctx.jh, vhashB );
+//     }
+
+     mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
+
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+
+     sph_shavite512_init( &ctx.shavite );
+     sph_shavite512 ( &ctx.shavite, hash0, 64 );
+     sph_shavite512_close( &ctx.shavite, hash0 );
+     sph_shavite512_init( &ctx.shavite );
+     sph_shavite512 ( &ctx.shavite, hash1, 64 );
+     sph_shavite512_close( &ctx.shavite, hash1 );
+     sph_shavite512_init( &ctx.shavite );
+     sph_shavite512 ( &ctx.shavite, hash2, 64 );
+     sph_shavite512_close( &ctx.shavite, hash2 );
+     sph_shavite512_init( &ctx.shavite );
+     sph_shavite512 ( &ctx.shavite, hash3, 64 );
+     sph_shavite512_close( &ctx.shavite, hash3 );
+
+     init_sd( &ctx.simd, 512 );
+     update_final_sd( &ctx.simd, (BitSequence *)hash0,
+                           (const BitSequence *)hash0, 512 );
+     init_sd( &ctx.simd, 512 );
+     update_final_sd( &ctx.simd, (BitSequence *)hash1,
+                           (const BitSequence *)hash1, 512 );
+     init_sd( &ctx.simd, 512 );
+     update_final_sd( &ctx.simd, (BitSequence *)hash2,
+                           (const BitSequence *)hash2, 512 );
+     init_sd( &ctx.simd, 512 );
+     update_final_sd( &ctx.simd, (BitSequence *)hash3,
+                           (const BitSequence *)hash3, 512 );
+
+// A is whirlpool serial, B is haval parallel.
+    
+
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+
+     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
+                                   m256_zero );
+     // A
+    
+//     if ( hash0[0] & mask )
+//     {
+        sph_whirlpool_init( &ctx.whirlpool );
+        sph_whirlpool( &ctx.whirlpool, hash0, 64 );
+        sph_whirlpool_close( &ctx.whirlpool, hash0 );
+//     }
+//     if ( hash1[0] & mask )
+//     {
+        sph_whirlpool_init( &ctx.whirlpool );
+        sph_whirlpool( &ctx.whirlpool, hash1, 64 );
+        sph_whirlpool_close( &ctx.whirlpool, hash1 );
+//     }
+//     if ( hash2[0] & mask )
+//     {
+        sph_whirlpool_init( &ctx.whirlpool );
+        sph_whirlpool( &ctx.whirlpool, hash2, 64 );
+        sph_whirlpool_close( &ctx.whirlpool, hash2 );
+//     }
+//     if ( hash3[0] & mask )
+//     {
+        sph_whirlpool_init( &ctx.whirlpool );
+        sph_whirlpool( &ctx.whirlpool, hash3, 64 );
+        sph_whirlpool_close( &ctx.whirlpool, hash3 );
+//     }
+
+     mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+
+// B
+
+//     if ( mm256_any_clr_256( vh_mask ) )
+//     {
+        haval256_5_4way_init( &ctx.haval );
+        haval256_5_4way( &ctx.haval, vhash, 64 );
+        haval256_5_4way_close( &ctx.haval, vhashB );
+        memset( &vhashB[8<<2], 0, 32<<2);
+//     }
+
+     mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
+
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+    
+     init_echo( &ctx.echo, 512 );
+     update_final_echo( &ctx.echo, (BitSequence *)hash0,
+                             (const BitSequence *)hash0, 512 );
+     init_echo( &ctx.echo, 512 );
+     update_final_echo( &ctx.echo, (BitSequence *)hash1,
+                             (const BitSequence *)hash1, 512 );
+     init_echo( &ctx.echo, 512 );
+     update_final_echo( &ctx.echo, (BitSequence *)hash2,
+                             (const BitSequence *)hash2, 512 );
+     init_echo( &ctx.echo, 512 );
+     update_final_echo( &ctx.echo, (BitSequence *)hash3,
+                             (const BitSequence *)hash3, 512 );
+
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     
+     blake512_4way_init( &ctx.blake );
+     blake512_4way( &ctx.blake, vhash, 64 );
+     blake512_4way_close( &ctx.blake, vhash );
+
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+
+// shavite & luffa, both serial, select individually.
+
+   if ( hash0[0] & mask )
+   {
+      sph_shavite512_init( &ctx.shavite );
+      sph_shavite512( &ctx.shavite, hash0, 64 ); //
+      sph_shavite512_close( &ctx.shavite, hash0 ); //8
+   }
+   else
+   {
+      init_luffa( &ctx.luffa, 512 );
+      update_and_final_luffa( &ctx.luffa, (BitSequence *)hash0,
+                                    (const BitSequence *)hash0, 64 );
+   }
+
+   if ( hash1[0] & mask )
+   {
+      sph_shavite512_init( &ctx.shavite );
+      sph_shavite512( &ctx.shavite, hash1, 64 ); //
+      sph_shavite512_close( &ctx.shavite, hash1 ); //8
+   }
+   else
+   {
+      init_luffa( &ctx.luffa, 512 );
+      update_and_final_luffa( &ctx.luffa, (BitSequence *)hash1,
+                                    (const BitSequence *)hash1, 64 );
+   }
+
+   if ( hash2[0] & mask )
+   {
+      sph_shavite512_init( &ctx.shavite );
+      sph_shavite512( &ctx.shavite, hash2, 64 ); //
+      sph_shavite512_close( &ctx.shavite, hash2 ); //8
+   }
+   else
+   {
+      init_luffa( &ctx.luffa, 512 );
+      update_and_final_luffa( &ctx.luffa, (BitSequence *)hash2,
+                                    (const BitSequence *)hash2, 64 );
+   }
+
+   if ( hash3[0] & mask )
+   {
+      sph_shavite512_init( &ctx.shavite );
+      sph_shavite512( &ctx.shavite, hash3, 64 ); //
+      sph_shavite512_close( &ctx.shavite, hash3 ); //8
+   }
+   else
+   {
+      init_luffa( &ctx.luffa, 512 );
+      update_and_final_luffa( &ctx.luffa, (BitSequence *)hash3,
+                                    (const BitSequence *)hash3, 64 );
+   }
+
+   mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+
+   hamsi512_4way_init( &ctx.hamsi );
+   hamsi512_4way( &ctx.hamsi, vhash, 64 );
+   hamsi512_4way_close( &ctx.hamsi, vhash );
+
+   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+
+   sph_fugue512_init( &ctx.fugue );
+   sph_fugue512( &ctx.fugue, hash0, 64 );
+   sph_fugue512_close( &ctx.fugue, hash0 );
+   sph_fugue512_init( &ctx.fugue );
+   sph_fugue512( &ctx.fugue, hash1, 64 );
+   sph_fugue512_close( &ctx.fugue, hash1 );
+   sph_fugue512_init( &ctx.fugue );
+   sph_fugue512( &ctx.fugue, hash2, 64 );
+   sph_fugue512_close( &ctx.fugue, hash2 );
+   sph_fugue512_init( &ctx.fugue );
+   sph_fugue512( &ctx.fugue, hash3, 64 );
+   sph_fugue512_close( &ctx.fugue, hash3 );
+
+
+//  A echo, B sd both serial
+   
+   if ( hash0[0] & mask ) //4
+   {
+       init_echo( &ctx.echo, 512 );
+       update_final_echo( &ctx.echo, (BitSequence *)hash0,
+                               (const BitSequence *)hash0, 512 );
+   }
+   else
+   {
+       init_sd( &ctx.simd, 512 );
+       update_final_sd( &ctx.simd, (BitSequence *)hash0,
+                             (const BitSequence *)hash0, 512 );
+   }
+
+   if ( hash1[0] & mask ) //4
+   {
+       init_echo( &ctx.echo, 512 );
+       update_final_echo( &ctx.echo, (BitSequence *)hash1,
+                               (const BitSequence *)hash1, 512 );
+   }
+   else
+   {
+       init_sd( &ctx.simd, 512 );
+       update_final_sd( &ctx.simd, (BitSequence *)hash1,
+                             (const BitSequence *)hash1, 512 );
+   }
+
+   if ( hash2[0] & mask ) //4
+   {
+       init_echo( &ctx.echo, 512 );
+       update_final_echo( &ctx.echo, (BitSequence *)hash2,
+                               (const BitSequence *)hash2, 512 );
+   }
+   else
+   {
+       init_sd( &ctx.simd, 512 );
+       update_final_sd( &ctx.simd, (BitSequence *)hash2,
+                             (const BitSequence *)hash2, 512 );
+   }
+
+   if ( hash3[0] & mask ) //4
+   {
+       init_echo( &ctx.echo, 512 );
+       update_final_echo( &ctx.echo, (BitSequence *)hash3,
+                               (const BitSequence *)hash3, 512 );
+   }
+   else
+   {
+       init_sd( &ctx.simd, 512 );
+       update_final_sd( &ctx.simd, (BitSequence *)hash3,
+                             (const BitSequence *)hash3, 512 );
+   }
+
+   mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
+
+   shabal512_4way_init( &ctx.shabal );
+   shabal512_4way( &ctx.shabal, vhash, 64 );
+   shabal512_4way_close( &ctx.shabal, vhash );
+
+   mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
+
+   sph_whirlpool_init( &ctx.whirlpool );
+   sph_whirlpool( &ctx.whirlpool, hash0, 64 );
+   sph_whirlpool_close( &ctx.whirlpool, hash0 );
+   sph_whirlpool_init( &ctx.whirlpool );
+   sph_whirlpool( &ctx.whirlpool, hash1, 64 );
+   sph_whirlpool_close( &ctx.whirlpool, hash1 );
+   sph_whirlpool_init( &ctx.whirlpool );
+   sph_whirlpool( &ctx.whirlpool, hash2, 64 );
+   sph_whirlpool_close( &ctx.whirlpool, hash2 );
+   sph_whirlpool_init( &ctx.whirlpool );
+   sph_whirlpool( &ctx.whirlpool, hash3, 64 );
+   sph_whirlpool_close( &ctx.whirlpool, hash3 );
+
+// A = fugue serial, B = sha512 prarallel
+   
+   mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+
+   vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
+                                 m256_zero );
+
+//   if ( hash0[0] & mask ) 
+//   {
+      sph_fugue512_init( &ctx.fugue );
+      sph_fugue512( &ctx.fugue, hash0, 64 );
+      sph_fugue512_close( &ctx.fugue, hash0 );
+//   }
+//   if ( hash1[0] & mask ) 
+//   {
+      sph_fugue512_init( &ctx.fugue );
+      sph_fugue512( &ctx.fugue, hash1, 64 );
+      sph_fugue512_close( &ctx.fugue, hash1 );
+//   }
+//   if ( hash2[0] & mask ) 
+//   {
+      sph_fugue512_init( &ctx.fugue );
+      sph_fugue512( &ctx.fugue, hash2, 64 );
+      sph_fugue512_close( &ctx.fugue, hash2 );
+//   }
+//   if ( hash3[0] & mask ) 
+//   {
+      sph_fugue512_init( &ctx.fugue );
+      sph_fugue512( &ctx.fugue, hash3, 64 );
+      sph_fugue512_close( &ctx.fugue, hash3 );
+//   }
+
+   mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+
+//   if ( mm256_any_clr_256( vh_mask ) )
+//   {
+      sha512_4way_init( &ctx.sha512 );
+      sha512_4way( &ctx.sha512, vhash, 64 );
+      sha512_4way_close( &ctx.sha512, vhashB );
+//   }
+
+   mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
+
+   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+
+   init_groestl( &ctx.groestl, 64 );
+   update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
+   init_groestl( &ctx.groestl, 64 );
+   update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
+   init_groestl( &ctx.groestl, 64 );
+   update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
+   init_groestl( &ctx.groestl, 64 );
+   update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
+
+   mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+
+   sha512_4way_init( &ctx.sha512 ); 
+   sha512_4way( &ctx.sha512, vhash, 64 );
+   sha512_4way_close( &ctx.sha512, vhash ); 
+
+// A = haval parallel, B = Whirlpool serial
+
+   vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
+                                 m256_zero );
+
+   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     
+//   if ( mm256_any_set_256( vh_mask ) ) //4
+//   {
+      haval256_5_4way_init( &ctx.haval );
+      haval256_5_4way( &ctx.haval, vhash, 64 );
+      haval256_5_4way_close( &ctx.haval, vhashA );
+      memset( &vhashA[8<<2], 0, 32<<2 );
+//   }
+
+//   if ( !( hash0[0] & mask ) )
+//   {
+      sph_whirlpool_init( &ctx.whirlpool );
+      sph_whirlpool( &ctx.whirlpool, hash0, 64 );
+      sph_whirlpool_close( &ctx.whirlpool, hash0 );
+//   }
+//   if ( !( hash2[0] & mask ) )
+//   {
+      sph_whirlpool_init( &ctx.whirlpool );
+      sph_whirlpool( &ctx.whirlpool, hash1, 64 );
+      sph_whirlpool_close( &ctx.whirlpool, hash1 );
+//   }
+//   if ( !( hash2[0] & mask ) )
+//   {
+      sph_whirlpool_init( &ctx.whirlpool );
+      sph_whirlpool( &ctx.whirlpool, hash2, 64 );
+      sph_whirlpool_close( &ctx.whirlpool, hash2 );
+//   }
+//   if ( !( hash3[0] & mask ) )
+//   {
+      sph_whirlpool_init( &ctx.whirlpool );
+      sph_whirlpool( &ctx.whirlpool, hash3, 64 );
+      sph_whirlpool_close( &ctx.whirlpool, hash3 );
+//   }
+
+   mm256_intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, 512 );
+
+   mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
+
+   bmw512_4way_init( &ctx.bmw );
+   bmw512_4way( &ctx.bmw, vhash, 64 );
+   bmw512_4way_close( &ctx.bmw, vhash );
+
+ 	memcpy(state, vhash, 32<<2 );
+}
+
+int scanhash_hmq1725_4way( int thr_id, struct work *work, uint32_t max_nonce,
+                           uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t hash[4*8] __attribute__ ((aligned (64)));
+//   uint32_t *hash7 = &(hash[7<<2]);
+//   uint32_t lane_hash[8];
+   uint32_t vdata[24*4] __attribute__ ((aligned (64)));
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   uint32_t n = pdata[19] - 1;
+   const uint32_t first_nonce = pdata[19];
+   __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
+   const uint32_t Htarg = ptarget[7];
+   uint64_t htmax[] = {          0,        0xF,       0xFF,
+                             0xFFF,     0xFFFF, 0x10000000  };
+   uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
+                        0xFFFFF000, 0xFFFF0000,          0  };
+
+   mm256_bswap_intrlv80_4x64( vdata, pdata );
+   for ( int m = 0; m < 6; m++ ) if ( Htarg <= htmax[m] )
+   {
+      uint32_t mask = masks[ m ];
+      do
+      {
+         *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
+                 _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
+         hmq1725_4way_hash( hash, vdata );
+         for ( int i = 0; i < 4; i++ )
+         if ( ( (hash+(i<<3))[7] & mask ) == 0 )
+         {
+            if ( fulltest( (hash+(i<<3)), ptarget ) && !opt_benchmark )
+            {
+               pdata[19] = n + i;
+               submit_solution( work, (hash+(i<<3)), mythr, i );
+            }
+         }
+	      n += 4;
+      } while ( ( n < max_nonce-4 ) && !work_restart[thr_id].restart );	
+	   break;
+	}
+	*hashes_done = n - first_nonce + 1;
+	return 0;
+}
+
+#endif // HMQ1725_4WAY

algo/quark/hmq1725-gate.c

@@ -0,0 +1,17 @@
+#include "hmq1725-gate.h"
+
+bool register_hmq1725_algo( algo_gate_t* gate )
+{
+#if defined(HMQ1725_4WAY)
+  gate->scanhash  = (void*)&scanhash_hmq1725_4way;
+  gate->hash      = (void*)&hmq1725_4way_hash;
+#else
+  init_hmq1725_ctx();
+  gate->scanhash  = (void*)&scanhash_hmq1725;
+  gate->hash      = (void*)&hmq1725hash;
+#endif
+  gate->set_target       = (void*)&scrypt_set_target;
+  gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT;
+  return true;
+};
+

algo/quark/hmq1725-gate.h

@@ -0,0 +1,28 @@
+#ifndef HMQ1725_GATE_H__
+#define HMQ1725_GATE_H__ 1
+
+#include "algo-gate-api.h"
+#include <stdint.h>
+
+#if defined(__AVX2__) && defined(__AES__)
+//  #define HMQ1725_4WAY
+#endif
+
+bool register_hmq1725_algo( algo_gate_t* gate );
+
+#if defined(HMQ1725_4WAY)
+
+void hmq1725_4way_hash( void *state, const void *input );
+int scanhash_hmq1725_4way( int thr_id, struct work *work, uint32_t max_nonce,
+                           uint64_t *hashes_done, struct thr_info *mythr );
+
+#else
+
+void hmq1725hash( void *state, const void *input );
+int scanhash_hmq1725( int thr_id, struct work *work, uint32_t max_nonce,
+                      uint64_t *hashes_done, struct thr_info *mythr );
+void init_hmq1725_ctx();
+
+#endif
+
+#endif  // HMQ1725_GATE_H__

algo/quark/hmq1725.c

@@ -1,4 +1,4 @@
-#include "algo-gate-api.h"
+#include "hmq1725-gate.h"
 #include <string.h>
 #include <stdint.h>
 #include "algo/blake/sph_blake.h"
@@ -298,19 +298,22 @@ extern void hmq1725hash(void *state, const void *input)
 	memcpy(state, hashA, 32);
 }
 
-int scanhash_hmq1725( int thr_id, struct work *work, int32_t max_nonce,
-                      uint64_t *hashes_done )
+int scanhash_hmq1725( int thr_id, struct work *work, uint32_t max_nonce,
+                      uint64_t *hashes_done, struct thr_info *mythr )
 {
-        uint32_t endiandata[32] __attribute__((aligned(64)));
+//        uint32_t endiandata[32] __attribute__((aligned(64)));
+        uint32_t endiandata[20] __attribute__((aligned(64)));
         uint32_t hash64[8] __attribute__((aligned(64)));
         uint32_t *pdata = work->data;
         uint32_t *ptarget = work->target;
 	uint32_t n = pdata[19] - 1;
 	const uint32_t first_nonce = pdata[19];
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 	//const uint32_t Htarg = ptarget[7];
 
 	//we need bigendian data...
-        for (int k = 0; k < 32; k++)
+//        for (int k = 0; k < 32; k++)
+        for (int k = 0; k < 20; k++)
                 be32enc(&endiandata[k], pdata[k]);
 
         hmq_bmw512_midstate( endiandata );
@@ -406,14 +409,14 @@ int scanhash_hmq1725( int thr_id, struct work *work, int32_t max_nonce,
 	pdata[19] = n;
 	return 0;
 }
-
+/*
 bool register_hmq1725_algo( algo_gate_t* gate )
 {
   init_hmq1725_ctx();
-  gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | SHA_OPT;
+  gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT;
   gate->set_target       = (void*)&scrypt_set_target;
   gate->scanhash         = (void*)&scanhash_hmq1725;
   gate->hash             = (void*)&hmq1725hash;
   return true;
 };
-
+*/

algo/quark/quark-4way.c

@@ -48,9 +48,10 @@ void quark_4way_hash( void *state, const void *input )
     __m256i* vhA = (__m256i*)vhashA;
     __m256i* vhB = (__m256i*)vhashB;
     __m256i vh_mask;
-    __m256i bit3_mask; bit3_mask = _mm256_set1_epi64x( 8 );
-    int i;
     quark_4way_ctx_holder ctx;
+    const __m256i bit3_mask = _mm256_set1_epi64x( 8 );
+    const uint32_t mask = 8;
+
     memcpy( &ctx, &quark_4way_ctx, sizeof(quark_4way_ctx) );
 
     blake512_4way( &ctx.blake, input, 80 );
@@ -62,27 +63,44 @@ void quark_4way_hash( void *state, const void *input )
     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ),
                                   m256_zero );
 
-       mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+    mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+
+    if ( hash0[0] & mask )
+    {
        update_and_final_groestl( &ctx.groestl, (char*)hash0,
                                                (char*)hash0, 512 );
+    }
+    if ( hash1[0] & mask )
+    {
        reinit_groestl( &ctx.groestl );
        update_and_final_groestl( &ctx.groestl, (char*)hash1,
                                                (char*)hash1, 512 );
+    }
+    if ( hash2[0] & mask )
+    {   
        reinit_groestl( &ctx.groestl );
        update_and_final_groestl( &ctx.groestl, (char*)hash2,
                                                (char*)hash2, 512 );
+    }
+    if ( hash3[0] & mask )
+    {   
        reinit_groestl( &ctx.groestl );
        update_and_final_groestl( &ctx.groestl, (char*)hash3,
                                                (char*)hash3, 512 );
-       mm256_interleave_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+    }
 
+    mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+
+    if ( mm256_anybits0( vh_mask ) )   
+    {
        skein512_4way( &ctx.skein, vhash, 64 );
        skein512_4way_close( &ctx.skein, vhashB );
+    }
 
-    for ( i = 0; i < 8; i++ )
-       vh[i] = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );
+    mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
+
+    mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
-    mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
     reinit_groestl( &ctx.groestl );
     update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
     reinit_groestl( &ctx.groestl );
@@ -91,7 +109,8 @@ void quark_4way_hash( void *state, const void *input )
     update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
     reinit_groestl( &ctx.groestl );
     update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
-    mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+
+    mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
     jh512_4way( &ctx.jh, vhash, 64 );
     jh512_4way_close( &ctx.jh, vhash );
@@ -99,16 +118,21 @@ void quark_4way_hash( void *state, const void *input )
     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ),
                                   m256_zero );
 
+    if ( mm256_anybits1( vh_mask ) )
+    {
        blake512_4way_init( &ctx.blake );
        blake512_4way( &ctx.blake, vhash, 64 );
        blake512_4way_close( &ctx.blake, vhashA );
+    }
 
+    if ( mm256_anybits0( vh_mask ) )
+    {
        bmw512_4way_init( &ctx.bmw );
        bmw512_4way( &ctx.bmw, vhash, 64 );
        bmw512_4way_close( &ctx.bmw, vhashB );
+    }
 
-    for ( i = 0; i < 8; i++ )
-       vh[i] = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );
+    mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
     keccak512_4way( &ctx.keccak, vhash, 64 );
     keccak512_4way_close( &ctx.keccak, vhash );
@@ -120,63 +144,65 @@ void quark_4way_hash( void *state, const void *input )
     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ),
                                   m256_zero );
 
+    if ( mm256_anybits1( vh_mask ) )
+    {
        keccak512_4way_init( &ctx.keccak );
        keccak512_4way( &ctx.keccak, vhash, 64 );
        keccak512_4way_close( &ctx.keccak, vhashA );
+    }
 
+    if ( mm256_anybits0( vh_mask ) )
+    {
        jh512_4way_init( &ctx.jh );
        jh512_4way( &ctx.jh, vhash, 64 );
        jh512_4way_close( &ctx.jh, vhashB );
+    }
 
-    for ( i = 0; i < 8; i++ )
-       vh[i] = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );
-
-    mm256_deinterleave_4x64( state, state+32, state+64, state+96, vhash, 256 );
+    // Final blend, directly to state, only need 32 bytes.
+    casti_m256i( state, 0 ) = _mm256_blendv_epi8( vhA[0], vhB[0], vh_mask );
+    casti_m256i( state, 1 ) = _mm256_blendv_epi8( vhA[1], vhB[1], vh_mask );
+    casti_m256i( state, 2 ) = _mm256_blendv_epi8( vhA[2], vhB[2], vh_mask );
+    casti_m256i( state, 3 ) = _mm256_blendv_epi8( vhA[3], vhB[3], vh_mask );
 }
 
 int scanhash_quark_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                         uint64_t *hashes_done)
+                         uint64_t *hashes_done, struct thr_info *mythr )
 {
     uint32_t hash[4*8] __attribute__ ((aligned (64)));
     uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-    uint32_t endiandata[20] __attribute__((aligned(64)));
+    uint32_t lane_hash[8] __attribute__ ((aligned (64)));
+    uint32_t *hash7 = &(hash[25]);
     uint32_t *pdata = work->data;
     uint32_t *ptarget = work->target;
     uint32_t n = pdata[19];
     const uint32_t first_nonce = pdata[19];
-    uint32_t *nonces = work->nonces;
-    int num_found = 0;
-    uint32_t *noncep = vdata + 73;   // 9*8 + 1
-
-    swab32_array( endiandata, pdata, 20 );
-
-    uint64_t *edata = (uint64_t*)endiandata;
-    mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+    __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
+    /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
+    mm256_bswap_intrlv80_4x64( vdata, pdata );
     do
     {
-       be32enc( noncep,   n   );
-       be32enc( noncep+2, n+1 );
-       be32enc( noncep+4, n+2 );
-       be32enc( noncep+6, n+3 );
+       *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
+                _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
 
        quark_4way_hash( hash, vdata );
        pdata[19] = n;
 
        for ( int i = 0; i < 4; i++ )
-       if ( ( ( (hash+(i<<3))[7] & 0xFFFFFF00 ) == 0 )
-            && fulltest( hash+(i<<3), ptarget ) )
+       if ( ( hash7[ i<<1 ] & 0xFFFFFF00 ) == 0 )
        {
-          pdata[19] = n+i;
-          nonces[ num_found++ ] = n+i;
-          work_set_target_ratio( work, hash+(i<<3) );
+          mm256_extract_lane_4x64( lane_hash, hash, i, 256 );
+          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark  )
+          {
+            pdata[19] = n+i;
+            submit_solution( work, lane_hash, mythr, i );
+          }
        }
        n += 4;
-    } while ( ( num_found == 0 ) && ( n < max_nonce )
-              && !work_restart[thr_id].restart );
+    } while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
 
     *hashes_done = n - first_nonce + 1;
-    return num_found;
+    return 0;
 }
 
 #endif

algo/quark/quark-gate.h

@@ -13,19 +13,15 @@ bool register_quark_algo( algo_gate_t* gate );
 #if defined(QUARK_4WAY)
 
 void quark_4way_hash( void *state, const void *input );
-
 int scanhash_quark_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                         uint64_t *hashes_done );
-
+                         uint64_t *hashes_done, struct thr_info *mythr );
 void init_quark_4way_ctx();
 
 #endif
 
 void quark_hash( void *state, const void *input );
-
 int scanhash_quark( int thr_id, struct work *work, uint32_t max_nonce,
-                    uint64_t *hashes_done );
-
+                    uint64_t *hashes_done, struct thr_info *mythr );
 void init_quark_ctx();
 
 #endif

algo/quark/quark.c

@@ -173,16 +173,17 @@ void quark_hash(void *state, const void *input)
 }
 
 int scanhash_quark( int thr_id, struct work *work, uint32_t max_nonce,
-                    uint64_t *hashes_done)
+                    uint64_t *hashes_done, struct thr_info *mythr )
 {
-        uint32_t endiandata[20] __attribute__((aligned(64)));
-        uint32_t hash64[8] __attribute__((aligned(32)));
-        uint32_t *pdata = work->data;
-        uint32_t *ptarget = work->target;
+   uint32_t endiandata[20] __attribute__((aligned(64)));
+   uint32_t hash64[8] __attribute__((aligned(32)));
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
 	uint32_t n = pdata[19] - 1;
 	const uint32_t first_nonce = pdata[19];
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
-        swab32_array( endiandata, pdata, 20 );
+   swab32_array( endiandata, pdata, 20 );
 
 	do {
 		pdata[19] = ++n;

algo/qubit/deep-2way.c

@@ -64,7 +64,7 @@ void deep_2way_hash( void *output, const void *input )
 }
 
 int scanhash_deep_2way( int thr_id, struct work *work,uint32_t max_nonce,
-                         uint64_t *hashes_done )
+                         uint64_t *hashes_done, struct thr_info *mythr )
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
@@ -73,17 +73,17 @@ int scanhash_deep_2way( int thr_id, struct work *work,uint32_t max_nonce,
      uint32_t *ptarget = work->target;
      uint32_t n = pdata[19];
      const uint32_t first_nonce = pdata[19];
-     uint32_t *nonces = work->nonces;
-     int num_found = 0;
      uint32_t *noncep = vdata + 32+3;   // 4*8 + 3
+     /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
      const uint32_t Htarg = ptarget[7];
      uint64_t htmax[] = {          0,        0xF,       0xFF,
                                0xFFF,     0xFFFF, 0x10000000  };
      uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
-     // big endian encode 0..18 uint32_t, 64 bits at a time
-     swab32_array( endiandata, pdata, 20 );
+     casti_m256i( endiandata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
+     casti_m256i( endiandata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
+     casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
 
      uint64_t *edata = (uint64_t*)endiandata;
      mm256_interleave_2x128( (uint64_t*)vdata, edata, edata, 640 );
@@ -102,23 +102,24 @@ int scanhash_deep_2way( int thr_id, struct work *work,uint32_t max_nonce,
             deep_2way_hash( hash, vdata );
             pdata[19] = n;
 
-            if ( !( hash[7] & mask ) && fulltest( hash, ptarget) )
+            if ( !( hash[7] & mask ) )
+            if ( fulltest( hash, ptarget) && !opt_benchmark )
             {
-               nonces[ num_found++ ] = n;
-               work_set_target_ratio( work, hash );
+                pdata[19] = n;
+                submit_solution( work, hash, mythr, 0 );
             }
-            if ( !( (hash+8)[7] & mask ) && fulltest( hash+8, ptarget) )
+            if ( !( (hash+8)[7] & mask ) )
+            if ( fulltest( hash+8, ptarget) && !opt_benchmark )
             {
-               nonces[ num_found++ ] = n+1;
-               work_set_target_ratio( work, hash+8 );
+               pdata[19] = n+1;
+               submit_solution( work, hash+8, mythr, 1 );
             }
             n += 2;
-         } while ( ( num_found == 0 ) && ( n < max_nonce )
-                   && !work_restart[thr_id].restart );
+         } while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
          break;
      }
      *hashes_done = n - first_nonce + 1;
-     return num_found;
+     return 0;
 }
 
 #endif

algo/qubit/deep-gate.h

@@ -13,19 +13,15 @@ bool register_deep_algo( algo_gate_t* gate );
 #if defined(DEEP_2WAY)
 
 void deep_2way_hash( void *state, const void *input );
-
 int scanhash_deep_2way( int thr_id, struct work *work, uint32_t max_nonce,
-                         uint64_t *hashes_done );
-
+                         uint64_t *hashes_done, struct thr_info *mythr );
 void init_deep_2way_ctx();
 
 #endif
 
 void deep_hash( void *state, const void *input );
-
 int scanhash_deep( int thr_id, struct work *work, uint32_t max_nonce,
-                    uint64_t *hashes_done );
-
+                    uint64_t *hashes_done, struct thr_info *mythr );
 void init_deep_ctx();
 
 #endif

algo/qubit/deep.c

@@ -72,14 +72,15 @@ void deep_hash(void *output, const void *input)
 }
 
 int scanhash_deep( int thr_id, struct work *work, uint32_t max_nonce,
-                   uint64_t *hashes_done)
+                   uint64_t *hashes_done, struct thr_info *mythr )
 {
-        uint32_t endiandata[20] __attribute__((aligned(64)));
-        uint32_t hash64[8] __attribute__((aligned(32)));
-        uint32_t *pdata = work->data;
-        uint32_t *ptarget = work->target;
+   uint32_t endiandata[20] __attribute__((aligned(64)));
+   uint32_t hash64[8] __attribute__((aligned(32)));
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
 	uint32_t n = pdata[19] - 1;
 	const uint32_t first_nonce = pdata[19];
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 	const uint32_t Htarg = ptarget[7];
 
 	uint64_t htmax[] = { 0, 0xF, 0xFF,  0xFFF, 0xFFFF, 0x10000000 };

algo/qubit/qubit-2way.c

@@ -70,7 +70,7 @@ void qubit_2way_hash( void *output, const void *input )
 }
 
 int scanhash_qubit_2way( int thr_id, struct work *work,uint32_t max_nonce,
-                         uint64_t *hashes_done )
+                         uint64_t *hashes_done, struct thr_info *mythr )
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
@@ -79,17 +79,17 @@ int scanhash_qubit_2way( int thr_id, struct work *work,uint32_t max_nonce,
      uint32_t *ptarget = work->target;
      uint32_t n = pdata[19];
      const uint32_t first_nonce = pdata[19];
-     uint32_t *nonces = work->nonces;
-     int num_found = 0;
      uint32_t *noncep = vdata + 32+3;   // 4*8 + 3
+     /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
      const uint32_t Htarg = ptarget[7];
      uint64_t htmax[] = {          0,        0xF,       0xFF,
                                0xFFF,     0xFFFF, 0x10000000  };
      uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
-     // big endian encode 0..18 uint32_t, 64 bits at a time
-     swab32_array( endiandata, pdata, 20 );
+     casti_m256i( endiandata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
+     casti_m256i( endiandata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
+     casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
 
      uint64_t *edata = (uint64_t*)endiandata;
      mm256_interleave_2x128( (uint64_t*)vdata, edata, edata, 640 );
@@ -107,25 +107,24 @@ int scanhash_qubit_2way( int thr_id, struct work *work,uint32_t max_nonce,
             qubit_2way_hash( hash, vdata );
             pdata[19] = n;
 
-
-            if ( !( hash[7] & mask ) && fulltest( hash, ptarget) )
+            if ( !( hash[7] & mask ) )
+            if ( fulltest( hash, ptarget) && !opt_benchmark )
             {
-               nonces[ num_found++ ] = n;
-               work_set_target_ratio( work, hash );
+                pdata[19] = n;
+                submit_solution( work, hash, mythr, 0 );
             }
-            if ( !( (hash+8)[7] & mask ) && fulltest( hash+8, ptarget) )
+            if ( !( (hash+8)[7] & mask ) )
+            if ( fulltest( hash+8, ptarget) && !opt_benchmark )
             {
                pdata[19] = n+1;
-               nonces[ num_found++ ] = n+1;
-               work_set_target_ratio( work, hash+8 );
+               submit_solution( work, hash+8, mythr, 1 );
             }
             n += 2;
-         } while ( ( num_found == 0 ) && ( n < max_nonce )
-                   && !work_restart[thr_id].restart );
+         } while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
          break;
      }
      *hashes_done = n - first_nonce + 1;
-     return num_found;
+     return 0;
 }
 
 #endif

algo/qubit/qubit-gate.h

@@ -13,19 +13,15 @@ bool register_qubit_algo( algo_gate_t* gate );
 #if defined(QUBIT_2WAY)
 
 void qubit_2way_hash( void *state, const void *input );
-
 int scanhash_qubit_2way( int thr_id, struct work *work, uint32_t max_nonce,
-                         uint64_t *hashes_done );
-
+                         uint64_t *hashes_done, struct thr_info *mythr );
 void init_qubit_2way_ctx();
 
 #endif
 
 void qubit_hash( void *state, const void *input );
-
 int scanhash_qubit( int thr_id, struct work *work, uint32_t max_nonce,
-                    uint64_t *hashes_done );
-
+                    uint64_t *hashes_done, struct thr_info *mythr );
 void init_qubit_ctx();
 
 #endif

algo/qubit/qubit.c

@@ -83,15 +83,16 @@ void qubit_hash(void *output, const void *input)
         memcpy(output, hash, 32);
 }
 
-int scanhash_qubit(int thr_id, struct work *work,
-		uint32_t max_nonce, uint64_t *hashes_done)
+int scanhash_qubit( int thr_id, struct work *work,	uint32_t max_nonce,
+                    uint64_t *hashes_done, struct thr_info *mythr )
 {
-        uint32_t endiandata[20] __attribute__((aligned(64)));
-        uint32_t hash64[8] __attribute__((aligned(64)));
-        uint32_t *pdata = work->data;
-        uint32_t *ptarget = work->target;
+   uint32_t endiandata[20] __attribute__((aligned(64)));
+   uint32_t hash64[8] __attribute__((aligned(64)));
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
 	uint32_t n = pdata[19] - 1;
 	const uint32_t first_nonce = pdata[19];
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 	const uint32_t Htarg = ptarget[7];
 
 	uint64_t htmax[] = { 0, 0xF, 0xFF,  0xFFF, 0xFFFF, 0x10000000 };

algo/ripemd/lbry-4way.c

@@ -40,9 +40,9 @@ void lbry_8way_hash( void* output, const void* input )
    sha256_8way_close( &ctx_sha256, vhashA );
 
    // reinterleave to do sha512 4-way 64 bit twice.
-   mm256_deinterleave_8x32( h0, h1, h2, h3, h4, h5, h6, h7, vhashA, 256 );
-   mm256_interleave_4x64( vhashA, h0, h1, h2, h3, 256 );
-   mm256_interleave_4x64( vhashB, h4, h5, h6, h7, 256 );
+   mm256_dintrlv_8x32( h0, h1, h2, h3, h4, h5, h6, h7, vhashA, 256 );
+   mm256_intrlv_4x64( vhashA, h0, h1, h2, h3, 256 );
+   mm256_intrlv_4x64( vhashB, h4, h5, h6, h7, 256 );
 
    sha512_4way_init( &ctx_sha512 );
    sha512_4way( &ctx_sha512, vhashA, 32 );
@@ -53,9 +53,9 @@ void lbry_8way_hash( void* output, const void* input )
    sha512_4way_close( &ctx_sha512, vhashB );
 
    // back to 8-way 32 bit
-   mm256_deinterleave_4x64( h0, h1, h2, h3, vhashA, 512 );
-   mm256_deinterleave_4x64( h4, h5, h6, h7, vhashB, 512 );
-   mm256_interleave_8x32( vhashA, h0, h1, h2, h3, h4, h5, h6, h7, 512 );
+   mm256_dintrlv_4x64( h0, h1, h2, h3, vhashA, 512 );
+   mm256_dintrlv_4x64( h4, h5, h6, h7, vhashB, 512 );
+   mm256_intrlv_8x32( vhashA, h0, h1, h2, h3, h4, h5, h6, h7, 512 );
 
    ripemd160_8way_init( &ctx_ripemd );
    ripemd160_8way( &ctx_ripemd, vhashA, 32 );
@@ -72,27 +72,24 @@ void lbry_8way_hash( void* output, const void* input )
 
    sha256_8way_init( &ctx_sha256 );
    sha256_8way( &ctx_sha256, vhashA, 32 );
-   sha256_8way_close( &ctx_sha256, vhashA );
-
-   mm256_deinterleave_8x32( output,     output+ 32, output+ 64, output+ 96,
-                            output+128, output+160, output+192, output+224,
-                            vhashA, 256 );
+   sha256_8way_close( &ctx_sha256, output );
 }
 
 int scanhash_lbry_8way( int thr_id, struct work *work, uint32_t max_nonce,
-                        uint64_t *hashes_done)
+                        uint64_t *hashes_done, struct thr_info *mythr )
 {
    uint32_t hash[8*8] __attribute__ ((aligned (64)));
    uint32_t vdata[32*8] __attribute__ ((aligned (64)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (32)));
+   uint32_t *hash7 = &(hash[7<<3]);
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    uint32_t n = pdata[27];
    const uint32_t first_nonce = pdata[27];
    const uint32_t Htarg = ptarget[7];
    uint32_t edata[32] __attribute__ ((aligned (64)));
-   uint32_t *nonces = work->nonces;
-   int num_found = 0;
-   uint32_t *noncep = vdata + 216; // 27*8
+   __m256i  *noncev = (__m256i*)vdata + 27;   // aligned
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
    uint64_t htmax[] = {          0,        0xF,       0xFF,
                              0xFFF,     0xFFFF, 0x10000000 };
@@ -100,9 +97,12 @@ int scanhash_lbry_8way( int thr_id, struct work *work, uint32_t max_nonce,
                         0xFFFFF000, 0xFFFF0000,          0 };
 
    // we need bigendian data...
-   swab32_array( edata, pdata, 32 );
-   mm256_interleave_8x32( vdata, edata, edata, edata, edata,
-                                 edata, edata, edata, edata, 1024 );
+   casti_m256i( edata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
+   casti_m256i( edata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
+   casti_m256i( edata, 2 ) = mm256_bswap_32( casti_m256i( pdata, 2 ) );
+   casti_m256i( edata, 3 ) = mm256_bswap_32( casti_m256i( pdata, 3 ) );
+   mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
+                             edata, edata, edata, edata, 1024 );
    sha256_8way_init( &sha256_8w_mid );
    sha256_8way( &sha256_8w_mid, vdata, LBRY_MIDSTATE );
 
@@ -111,136 +111,26 @@ int scanhash_lbry_8way( int thr_id, struct work *work, uint32_t max_nonce,
       uint32_t mask = masks[m];
       do
       {
-         be32enc( noncep,   n   );
-         be32enc( noncep+1, n+1 );
-         be32enc( noncep+2, n+2 );
-         be32enc( noncep+3, n+3 );
-         be32enc( noncep+4, n+4 );
-         be32enc( noncep+5, n+5 );
-         be32enc( noncep+6, n+6 );
-         be32enc( noncep+7, n+7 );
-
+        *noncev = mm256_bswap_32( _mm256_set_epi32(
+                                          n+7,n+6,n+5,n+4,n+3,n+2,n+1,n ) );
          lbry_8way_hash( hash, vdata );
 
-         for ( int i = 0; i < 8; i++ )
-         if ( !( (hash+(i<<3))[7] & mask ) && fulltest( hash+(i<<3), ptarget ) )
+         for ( int i = 0; i < 8; i++ )  if ( !( hash7[ i ] & mask ) )
          {
-            pdata[27] = n+i;
-            nonces[ num_found++ ] = n+i;
-            work_set_target_ratio( work, hash+(i<<3) );
+            // deinterleave hash for lane
+            mm256_extract_lane_8x32( lane_hash, hash, i, 256 );
+            if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+            {
+              pdata[27] = n + i;
+              submit_solution( work, lane_hash, mythr, i );
+            }
          }
-         n+=8;
-      } while ( ( num_found == 0 ) && ( n < max_nonce )
-                   && !work_restart[thr_id].restart );
+         n += 8;
+      } while ( (n < max_nonce-10) && !work_restart[thr_id].restart );
       break;
    }
-
-   *hashes_done = n - first_nonce;
-   return num_found;
-}
-
-#elif defined(LBRY_4WAY)
-
-static __thread sha256_4way_context sha256_mid;
-
-void lbry_4way_hash( void* output, const void* input )
-{
-   sha256_4way_context     ctx_sha256 __attribute__ ((aligned (64)));
-   sha512_4way_context     ctx_sha512;
-   ripemd160_4way_context  ctx_ripemd;
-   uint32_t _ALIGN(64) vhashA[16<<2];
-   uint32_t _ALIGN(64) vhashB[16<<2];
-   uint32_t _ALIGN(64) vhashC[16<<2];
-
-   memcpy( &ctx_sha256, &sha256_mid, sizeof(ctx_sha256) );
-   sha256_4way( &ctx_sha256, input + (LBRY_MIDSTATE<<2), LBRY_TAIL );
-   sha256_4way_close( &ctx_sha256, vhashA );
-
-   sha256_4way_init( &ctx_sha256 );
-   sha256_4way( &ctx_sha256, vhashA, 32 );
-   sha256_4way_close( &ctx_sha256, vhashA );
-
-   // sha512 64 bit data, 64 byte output
-   mm256_reinterleave_4x64( vhashB, vhashA, 256 );
-   sha512_4way_init( &ctx_sha512 );
-   sha512_4way( &ctx_sha512, vhashB, 32 );
-   sha512_4way_close( &ctx_sha512, vhashB );
-   mm256_reinterleave_4x32( vhashA, vhashB, 512 );
-
-   ripemd160_4way_init( &ctx_ripemd );
-   ripemd160_4way( &ctx_ripemd, vhashA, 32 );
-   ripemd160_4way_close( &ctx_ripemd, vhashB );
-
-   ripemd160_4way_init( &ctx_ripemd );
-   ripemd160_4way( &ctx_ripemd, vhashA+(8<<2), 32 );
-   ripemd160_4way_close( &ctx_ripemd, vhashC );
-
-   sha256_4way_init( &ctx_sha256 );
-   sha256_4way( &ctx_sha256, vhashB, 20 );
-   sha256_4way( &ctx_sha256, vhashC, 20 );
-   sha256_4way_close( &ctx_sha256, vhashA );
-
-   sha256_4way_init( &ctx_sha256 );
-   sha256_4way( &ctx_sha256, vhashA, 32 );
-   sha256_4way_close( &ctx_sha256, vhashA );
-
-   mm128_deinterleave_4x32( output, output+32, output+64, output+96,
-		            vhashA, 256 );
-}
-
-int scanhash_lbry_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                        uint64_t *hashes_done)
-{
-   uint32_t hash[4*8] __attribute__ ((aligned (64)));
-   uint32_t vdata[32*4] __attribute__ ((aligned (64)));
-   uint32_t *pdata = work->data;
-   uint32_t *ptarget = work->target;
-   uint32_t n = pdata[27];
-   const uint32_t first_nonce = pdata[27];
-   const uint32_t Htarg = ptarget[7];
-   uint32_t edata[32] __attribute__ ((aligned (64)));
-   uint32_t *nonces = work->nonces;
-   int num_found = 0;
-   uint32_t *noncep = vdata + 108; // 27*4
-
-   uint64_t htmax[] = {          0,        0xF,       0xFF,
-                             0xFFF,     0xFFFF, 0x10000000 };
-   uint32_t masks[] = {	0xFFFFFFFF, 0xFFFFFFF0,	0xFFFFFF00,
-                        0xFFFFF000, 0xFFFF0000,          0 };
-
-   // we need bigendian data...
-   swab32_array( edata, pdata, 32 );
-   mm128_interleave_4x32( vdata, edata, edata, edata, edata, 1024 );
-   sha256_4way_init( &sha256_mid );
-   sha256_4way( &sha256_mid, vdata, LBRY_MIDSTATE );
-
-   for ( int m = 0; m < sizeof(masks); m++ ) if ( Htarg <= htmax[m] )
-   {
-      uint32_t mask = masks[m];
-      do
-      {
-         be32enc( noncep,   n   );
-         be32enc( noncep+1, n+1 );
-         be32enc( noncep+2, n+2 );
-         be32enc( noncep+3, n+3 );
-
-         lbry_4way_hash( hash, vdata );
-
-         for ( int i = 0; i < 4; i++ )
-         if ( !( (hash+(i<<3))[7] & mask ) && fulltest( hash+(i<<3), ptarget ) )
-         {
-            pdata[27] = n+i;
-            nonces[ num_found++ ] = n+i;
-            work_set_target_ratio( work, hash+(i<<3) );
-         }
-         n+=4;
-      } while ( ( num_found == 0 ) && ( n < max_nonce )
-                   && !work_restart[thr_id].restart );
-      break;
-   }
-
-   *hashes_done = n - first_nonce;
-   return num_found;
+   *hashes_done = n - first_nonce + 1;
+   return 0;
 }
 
 #endif

algo/ripemd/lbry-gate.h

@@ -4,12 +4,10 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-// Overide multi way on ryzen, SHA is better.
-#if !defined(RYZEN_)
-// need sha512 2 way AVX x2 or 1 way scalar x4 to support 4way AVX.
-#if defined(__AVX2__)
+#if !defined(__SHA__)
+ #if defined(__AVX2__)
   #define LBRY_8WAY
-#endif
+ #endif
 #endif
 
 #define LBRY_NTIME_INDEX 25
@@ -24,17 +22,18 @@ bool register_lbry_algo( algo_gate_t* gate );
 
 void lbry_8way_hash( void *state, const void *input );
 int scanhash_lbry_8way( int thr_id, struct work *work, uint32_t max_nonce,
-                         uint64_t *hashes_done );
-
+                         uint64_t *hashes_done, struct thr_info *mythr );
+/*
 #elif defined(LBRY_4WAY)
 
 void lbry_4way_hash( void *state, const void *input );
 int scanhash_lbry_4way( int thr_id, struct work *work, uint32_t max_nonce,
                          uint64_t *hashes_done );
+*/
 #else
 
 void lbry_hash( void *state, const void *input );
 int scanhash_lbry( int thr_id, struct work *work, uint32_t max_nonce,
-                    uint64_t *hashes_done );
+                    uint64_t *hashes_done, struct thr_info *mythr );
 #endif
 #endif

algo/ripemd/lbry.c

@@ -48,13 +48,14 @@ void lbry_hash(void* output, const void* input)
 }
 
 int scanhash_lbry( int thr_id, struct work *work, uint32_t max_nonce,
-                   uint64_t *hashes_done)
+                   uint64_t *hashes_done, struct thr_info *mythr)
 {
   uint32_t *pdata = work->data;
   uint32_t *ptarget = work->target;
 	uint32_t n = pdata[27] - 1;
 	const uint32_t first_nonce = pdata[27];
 	const uint32_t Htarg = ptarget[7];
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
 	uint32_t hash64[8] __attribute__((aligned(64)));
 	uint32_t endiandata[32] __attribute__ ((aligned (64)));

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

@@ -6,7 +6,7 @@
 
 #if defined(__SSE4_2__)
 
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 typedef struct
 {

algo/scrypt.c

@@ -696,7 +696,7 @@ static void scrypt_1024_1_1_256_24way(const uint32_t *input,
 #endif /* HAVE_SCRYPT_6WAY */
 
 extern int scanhash_scrypt( int thr_id, struct work *work, uint32_t max_nonce,
-                            uint64_t *hashes_done )
+                            uint64_t *hashes_done, struct thr_info *mythr )
 {
         uint32_t *pdata = work->data;
         uint32_t *ptarget = work->target;
@@ -704,6 +704,7 @@ extern int scanhash_scrypt( int thr_id, struct work *work, uint32_t max_nonce,
 	uint32_t midstate[8];
 	uint32_t n = pdata[19] - 1;
 	const uint32_t Htarg = ptarget[7];
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 	int throughput = scrypt_best_throughput();
 	int i;
 	

algo/sha/sha2-hash-4way.h

@@ -42,7 +42,7 @@
 
 #include <stddef.h>
 #include "sph_types.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #if defined(__SSE2__)
 //#if defined(__SSE4_2__)

algo/sha/sha256_hash_11way.c

@@ -1,3 +1,4 @@
+#if 0
 
 #include <stddef.h>
 #include <string.h>
@@ -65,7 +66,7 @@ static const uint32_t K256[64] =
    _mm_xor_si64( _mm_xor_si64( \
        mm64_ror_32(x,2), mm64_ror_32(x,13) ), _mm_srli_pi32(x,22) )
 
-#define BSG2_0z(x)  ( ror_32(x,2) ^ ror_32(x,13)  ^ ((x)>>22) )
+#define BSG2_0z(x)  ( u32_ror_32(x,2) ^ u32_ror_32(x,13)  ^ ((x)>>22) )
 
 #define BSG2_1x(x) \
    _mm256_xor_si256( _mm256_xor_si256( \
@@ -75,7 +76,7 @@ static const uint32_t K256[64] =
    _mm_xor_si64( _mm_xor_si64( \
        mm64_ror_32(x,6), mm64_ror_32(x,11) ), _mm_srli_pi32(x,25) )
 
-#define BSG2_1z(x)   ( ror_32(x,6) ^ ror_32(x,11) ^ ((x)>>25) )
+#define BSG2_1z(x)   ( u32_ror_32(x,6) ^ u32_ror_32(x,11) ^ ((x)>>25) )
 
 #define SSG2_0x(x) \
    _mm256_xor_si256( _mm256_xor_si256( \
@@ -85,7 +86,7 @@ static const uint32_t K256[64] =
    _mm_xor_si64( _mm_xor_si64( \
        mm64_ror_32(x,7), mm64_ror_32(x,18) ), _mm_srli_pi32(x,3) )
 
-#define SSG2_0z(x)  (( ror_32(x,7) ^ ror_32(x,18) ) ^ ((x)>>3) )
+#define SSG2_0z(x)  (( u32_ror_32(x,7) ^ u32_ror_32(x,18) ) ^ ((x)>>3) )
 
 #define SSG2_1x(x) \
    _mm256_xor_si256( _mm256_xor_si256( \
@@ -95,7 +96,7 @@ static const uint32_t K256[64] =
    _mm_xor_si64( _mm_xor_si64( \
        mm64_ror_32(x,17), mm64_ror_32(x,19) ), _mm_srli_pi32(x,10) )
 
-#define SSG2_1z(x)   ( ror_32(x,17) ^ ror_32(x,19)  ^ ((x)>>10) )
+#define SSG2_1z(x)   ( u32_ror_32(x,17) ^ u32_ror_32(x,19)  ^ ((x)>>10) )
 
 #define SHA2x_MEXP( a, b, c, d ) \
      _mm256_add_epi32( _mm256_add_epi32( _mm256_add_epi32( \
@@ -449,7 +450,7 @@ void sha256_11way_update( sha256_11way_context *ctx, const void *datax,
       if ( clen > len )
          clen = len;
       memcpy_256( ctx->bufx + (ptr>>2), vdatax + (ptr>>2), clen>>2 );
-      memcpy_64 ( ctx->bufy + (ptr>>2), vdatay + (ptr>>2), clen>>2 );
+      memcpy_m64( ctx->bufy + (ptr>>2), vdatay + (ptr>>2), clen>>2 );
       memcpy    ( ctx->bufz +  ptr,     idataz +  ptr,     clen    );
       ptr += clen;
       len -= clen;
@@ -486,19 +487,19 @@ void sha256_11way_close( sha256_11way_context *ctx, void *dstx, void *dsty,
     if ( ptr > pad )
     {
          memset_zero_256( ctx->bufx + (ptr>>2), (buf_size - ptr) >> 2 );
-         memset_zero_64(  ctx->bufy + (ptr>>2), (buf_size - ptr) >> 2 );
+         memset_zero_m64( ctx->bufy + (ptr>>2), (buf_size - ptr) >> 2 );
          memset(      ctx->bufz + (ptr>>2), 0,  (buf_size - ptr) >> 2 );
          sha256_11way_round( ctx->bufx, ctx->valx,
 			     ctx->bufy, ctx->valy,
 			     ctx->bufz, ctx->valz );
          memset_zero_256( ctx->bufx, pad >> 2 );
-         memset_zero_64(  ctx->bufy, pad >> 2 );
+         memset_zero_m64(  ctx->bufy, pad >> 2 );
          memset(      ctx->bufz, 0,  pad >> 2 );
     }
     else
     {
         memset_zero_256( ctx->bufx + (ptr>>2),    (pad - ptr) >> 2 );
-        memset_zero_64(  ctx->bufy + (ptr>>2),    (pad - ptr) >> 2 );
+        memset_zero_m64(  ctx->bufy + (ptr>>2),    (pad - ptr) >> 2 );
         memset(          ctx->bufz + (ptr>>2), 0, (pad - ptr) >> 2 );
     }
 
@@ -534,3 +535,4 @@ void sha256_11way_close( sha256_11way_context *ctx, void *dstx, void *dsty,
 }
 
 #endif
+#endif   // 0

algo/sha/sha256q-4way.c

@@ -36,7 +36,6 @@ int scanhash_sha256q_8way( int thr_id, struct work *work, uint32_t max_nonce,
 {
    uint32_t vdata[20*8] __attribute__ ((aligned (64)));
    uint32_t hash[8*8] __attribute__ ((aligned (32)));
-   uint32_t edata[20] __attribute__ ((aligned (32)));;
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t Htarg = ptarget[7];
@@ -59,12 +58,7 @@ int scanhash_sha256q_8way( int thr_id, struct work *work, uint32_t max_nonce,
                                         0 };
 
    // Need big endian data
-   casti_m256i( edata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
-   casti_m256i( edata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
-   casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
-
-   mm256_interleave_8x32( vdata, edata, edata, edata, edata,
-                                 edata, edata, edata, edata, 640 );
+   mm256_bswap_intrlv80_8x32( vdata, pdata );
    sha256_8way_init( &sha256_ctx8 );
    sha256_8way( &sha256_ctx8, vdata, 64 );
 
@@ -73,11 +67,10 @@ int scanhash_sha256q_8way( int thr_id, struct work *work, uint32_t max_nonce,
       uint32_t mask = masks[m];
       do
       {
-        *noncev = mm256_bswap_32(
-		 _mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n ) );
-
-	 pdata[19] = n;
+         *noncev = mm256_bswap_32(
+		            _mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n ) );
 
+	      pdata[19] = n;
          sha256q_8way_hash( hash, vdata );
 
          uint32_t *hash7 = &(hash[7<<3]); 
@@ -86,27 +79,19 @@ int scanhash_sha256q_8way( int thr_id, struct work *work, uint32_t max_nonce,
          if ( !( hash7[ lane ] & mask ) )
          { 
             // deinterleave hash for lane
-	    uint32_t lane_hash[8];
-	    mm256_extract_lane_8x32( lane_hash, hash, lane, 256 );
+	         uint32_t lane_hash[8];
+	         mm256_extract_lane_8x32( lane_hash, hash, lane, 256 );
 
-	    if ( fulltest( lane_hash, ptarget ) )
+	         if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
             {
-	      pdata[19] = n + lane;
-              work_set_target_ratio( work, lane_hash );
-              if ( submit_work( mythr, work ) )
-                applog( LOG_NOTICE, "Share %d submitted by thread %d, lane %d.",
-                             accepted_share_count + rejected_share_count + 1,
-                             thr_id, lane );
-              else
-                applog( LOG_WARNING, "Failed to submit share." );
-	    }
-	 }
+	           pdata[19] = n + lane;
+              submit_solution( work, lane_hash, mythr, lane );
+            }
+	      }
          n += 8;
-
       } while ( (n < max_nonce-10) && !work_restart[thr_id].restart );
       break;
    }
-    
    *hashes_done = n - first_nonce + 1;
    return 0;
 }
@@ -146,7 +131,6 @@ int scanhash_sha256q_4way( int thr_id, struct work *work, uint32_t max_nonce,
    uint32_t hash[8*4] __attribute__ ((aligned (32)));
    uint32_t *hash7 = &(hash[7<<2]);
    uint32_t lane_hash[8];
-   uint32_t edata[20] __attribute__ ((aligned (32)));;
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t Htarg = ptarget[7];
@@ -168,13 +152,7 @@ int scanhash_sha256q_4way( int thr_id, struct work *work, uint32_t max_nonce,
                                0xFFFF0000,
                                         0 };
 
-   casti_m128i( edata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
-   casti_m128i( edata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
-   casti_m128i( edata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
-   casti_m128i( edata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
-   casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
-
-   mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
+   mm128_bswap_intrlv80_4x32( vdata, pdata );
    sha256_4way_init( &sha256_ctx4 );
    sha256_4way( &sha256_ctx4, vdata, 64 );
 
@@ -183,7 +161,7 @@ int scanhash_sha256q_4way( int thr_id, struct work *work, uint32_t max_nonce,
       uint32_t mask = masks[m];
       do {
          *noncev = mm128_bswap_32( _mm_set_epi32( n+3,n+2,n+1,n ) );
-	 pdata[19] = n;
+         pdata[19] = n;
 
          sha256q_4way_hash( hash, vdata );
 
@@ -192,25 +170,16 @@ int scanhash_sha256q_4way( int thr_id, struct work *work, uint32_t max_nonce,
          {
             mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
 
-            if ( fulltest( lane_hash, ptarget ) )
+            if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
             {
               pdata[19] = n + lane;
-              work_set_target_ratio( work, lane_hash );
-              if ( submit_work( mythr, work ) )
-                applog( LOG_NOTICE, "Share %d submitted by thread %d, lane %d.",
-                             accepted_share_count + rejected_share_count + 1,
-                             thr_id, lane );
-              else
-                applog( LOG_WARNING, "Failed to submit share." );
+              submit_solution( work, lane_hash, mythr, lane );
             }
          }
-
-	 n += 4;
-
+         n += 4;
       } while ( (n < max_nonce - 4) && !work_restart[thr_id].restart );
       break;
    }
-
    *hashes_done = n - first_nonce + 1;
    return 0;
 }

algo/sha/sha256t-4way.c

@@ -72,7 +72,7 @@ int scanhash_sha256t_11way( int thr_id, struct work *work, uint32_t max_nonce,
    casti_m256i( dataz, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
    casti_m128i( dataz, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
 
-   mm256_interleave_8x32( datax, dataz, dataz, dataz, dataz,
+   mm256_intrlv_8x32( datax, dataz, dataz, dataz, dataz,
                                  dataz, dataz, dataz, dataz, 640 );
    mm64_interleave_2x32( datay, dataz, dataz, 640 );
 
@@ -156,15 +156,15 @@ void sha256t_8way_hash( void* output, const void* input )
    sha256_8way_init( &ctx );
    sha256_8way( &ctx, vhash, 32 );
    sha256_8way_close( &ctx, output );
-
 }
 
 int scanhash_sha256t_8way( int thr_id, struct work *work, uint32_t max_nonce,
                            uint64_t *hashes_done, struct thr_info *mythr )
 {
-   uint32_t vdata[20*8] __attribute__ ((aligned (64)));
-   uint32_t hash[8*8] __attribute__ ((aligned (32)));
-   uint32_t edata[20] __attribute__ ((aligned (32)));;
+   uint32_t vdata[20*8]  __attribute__ ((aligned (64)));
+   uint32_t hash[8*8]    __attribute__ ((aligned (32)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (32)));
+   uint32_t *hash7 = &(hash[7<<3]);
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t Htarg = ptarget[7];
@@ -187,12 +187,7 @@ int scanhash_sha256t_8way( int thr_id, struct work *work, uint32_t max_nonce,
                                         0 };
 
    // Need big endian data
-   casti_m256i( edata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
-   casti_m256i( edata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
-   casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
-
-   mm256_interleave_8x32( vdata, edata, edata, edata, edata,
-                                 edata, edata, edata, edata, 640 );
+   mm256_bswap_intrlv80_8x32( vdata, pdata );
    sha256_8way_init( &sha256_ctx8 );
    sha256_8way( &sha256_ctx8, vdata, 64 );
 
@@ -201,29 +196,22 @@ int scanhash_sha256t_8way( int thr_id, struct work *work, uint32_t max_nonce,
       uint32_t mask = masks[m];
       do
       {
-        *noncev = mm256_bswap_32(
-                 _mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n ) );
+        *noncev = mm256_bswap_32( _mm256_set_epi32(
+                                          n+7,n+6,n+5,n+4,n+3,n+2,n+1,n ) );
          pdata[19] = n;
-
          sha256t_8way_hash( hash, vdata );
-
-         uint32_t *hash7 = &(hash[7<<3]);
-
          for ( int lane = 0; lane < 8; lane++ )
          if ( !( hash7[ lane ] & mask ) )
          {
             // deinterleave hash for lane
-            uint32_t lane_hash[8] __attribute__ ((aligned (64)));
             mm256_extract_lane_8x32( lane_hash, hash, lane, 256 );
-
-            if ( fulltest( lane_hash, ptarget ) )
+            if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
             {
               pdata[19] = n + lane;
               submit_solution( work, lane_hash, mythr, lane );
-	    }
+	         }
          }
          n += 8;
-
       } while ( (n < max_nonce-10) && !work_restart[thr_id].restart );
       break;
    }
@@ -253,7 +241,6 @@ void sha256t_4way_hash( void* output, const void* input )
    sha256_4way_init( &ctx );
    sha256_4way( &ctx, vhash, 32 );
    sha256_4way_close( &ctx, output );
-
 }
 
 int scanhash_sha256t_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -262,7 +249,6 @@ int scanhash_sha256t_4way( int thr_id, struct work *work, uint32_t max_nonce,
    uint32_t vdata[20*4] __attribute__ ((aligned (64)));
    uint32_t hash[8*4] __attribute__ ((aligned (32)));
    uint32_t lane_hash[8] __attribute__ ((aligned (64)));
-   uint32_t edata[20] __attribute__ ((aligned (32)));;
    uint32_t *hash7 = &(hash[7<<2]);
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
@@ -278,20 +264,14 @@ int scanhash_sha256t_4way( int thr_id, struct work *work, uint32_t max_nonce,
                                    0xFFF,
                                   0xFFFF,
                               0x10000000 };
-   const uint32_t masks[] = {  0xFFFFFFFF,
-                               0xFFFFFFF0,
-                               0xFFFFFF00,
-                               0xFFFFF000,
-                               0xFFFF0000,
-                                        0 };
+   const uint32_t masks[] = { 0xFFFFFFFF,
+                              0xFFFFFFF0,
+                              0xFFFFFF00,
+                              0xFFFFF000,
+                              0xFFFF0000,
+                                       0 };
 
-   casti_m128i( edata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
-   casti_m128i( edata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
-   casti_m128i( edata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
-   casti_m128i( edata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
-   casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
-
-   mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
+   mm128_bswap_intrlv80_4x32( vdata, pdata );
    sha256_4way_init( &sha256_ctx4 );
    sha256_4way( &sha256_ctx4, vdata, 64 );
 
@@ -300,7 +280,7 @@ int scanhash_sha256t_4way( int thr_id, struct work *work, uint32_t max_nonce,
       uint32_t mask = masks[m];
       do {
          *noncev = mm128_bswap_32( _mm_set_epi32( n+3,n+2,n+1,n ) );
-	 pdata[19] = n;
+         pdata[19] = n;
 
          sha256t_4way_hash( hash, vdata );
 
@@ -308,15 +288,13 @@ int scanhash_sha256t_4way( int thr_id, struct work *work, uint32_t max_nonce,
          if ( !( hash7[ lane ] & mask ) )
          {
             mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
-
-            if ( fulltest( lane_hash, ptarget ) )
+            if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
             {
               pdata[19] = n + lane;
               submit_solution( work, lane_hash, mythr, lane );
-	    }
-         }
-	 n += 4;
-
+	         }
+         } 
+         n += 4;
       } while ( (n < max_nonce - 4) && !work_restart[thr_id].restart );
       break;
    }

algo/sha/sha256t-gate.c

@@ -2,11 +2,7 @@
 
 bool register_sha256t_algo( algo_gate_t* gate )
 {
-#if defined(SHA256T_11WAY)
-    gate->optimizations = SSE2_OPT | AVX2_OPT | SHA_OPT;
-    gate->scanhash   = (void*)&scanhash_sha256t_11way;
-    gate->hash       = (void*)&sha256t_11way_hash;
-#elif defined(SHA256T_8WAY)
+#if defined(SHA256T_8WAY)
     gate->optimizations = SSE2_OPT | AVX2_OPT | SHA_OPT;
     gate->scanhash   = (void*)&scanhash_sha256t_8way;
     gate->hash       = (void*)&sha256t_8way_hash;
@@ -25,11 +21,7 @@ gate->optimizations = SHA_OPT;
 
 bool register_sha256q_algo( algo_gate_t* gate )
 {
-#if defined(SHA256T_8WAY)
-    gate->optimizations = SSE2_OPT | AVX2_OPT | SHA_OPT;
-    gate->scanhash   = (void*)&scanhash_sha256q_8way;
-    gate->hash       = (void*)&sha256q_8way_hash;
-#elif defined(SHA256T_4WAY)
+#if defined(SHA256T_4WAY)
     gate->optimizations = SSE2_OPT | AVX2_OPT | SHA_OPT;
     gate->scanhash   = (void*)&scanhash_sha256q_4way;
     gate->hash       = (void*)&sha256q_4way_hash;

algo/sha/sha256t-gate.h

@@ -5,30 +5,17 @@
 #include "algo-gate-api.h"
 
 // Override multi way on ryzen, SHA is better.
-#if !defined(RYZEN_)
-#if defined(__SSE2__)
-  #define SHA256T_4WAY
-#endif
-#if defined(__AVX2__)
+#if !defined(__SHA__)
+ #if defined(__AVX2__)
   #define SHA256T_8WAY
-//  #define SHA256T_11WAY
-#endif
+#elif defined(__SSE2__)
+  #define SHA256T_4WAY
+ #endif
 #endif
 
 bool register_sha256t_algo( algo_gate_t* gate );
 bool register_sha256q_algo( algo_gate_t* gate );
 
-#if defined(SHA256T_11WAY)
-
-void sha256t_11way_hash( void *outx, void *outy, void *outz, const void *inpx,
-	                 const void *inpy, const void *inpz );
-int scanhash_sha256t_11way( int thr_id, struct work *work, uint32_t max_nonce,
-                            uint64_t *hashes_done, struct thr_info *mythr );
-//void sha256q_8way_hash( void *output, const void *input );
-//int scanhash_sha256q_11way( int thr_id, struct work *work, uint32_t max_nonce,
-//                            uint64_t *hashes_done, struct thr_info *mythr );
-#endif
-
 #if defined(SHA256T_8WAY)
 
 void sha256t_8way_hash( void *output, const void *input );

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

@@ -40,7 +40,7 @@
 
 #include <stddef.h>
 #include "algo/sha/sph_types.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #ifdef __cplusplus
 extern "C"{

algo/shavite/shavite-hash-2way.h

@@ -3,7 +3,7 @@
 
 #if defined(__AVX2__)
   
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 typedef struct {
         unsigned char buf[128<<1];

algo/shavite/sph-shavite-aesni.c

@@ -36,7 +36,7 @@
 #ifdef __AES__
 
 #include "sph_shavite.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #ifdef __cplusplus
 extern "C"{

algo/simd/simd-hash-2way.c

@@ -6,6 +6,12 @@
 
 #if defined (__AVX2__)
 
+union _m256_v16 {
+  uint16_t u16[16];
+  __m256i v256;
+};
+typedef union _m256_v16 m256_v16;
+
 // imported from simd_iv.h
 
 uint32_t SIMD_IV_512[] = { 0x0ba16b95, 0x72f999ad, 0x9fecc2ae, 0xba3264fc,

algo/simd/simd-hash-2way.h

@@ -5,7 +5,7 @@
 
 #if defined(__AVX2__)
 
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 typedef struct {
   uint32_t A[ 32*2 ] __attribute__((aligned(64)));

algo/skein/skein-4way.c

@@ -2,7 +2,11 @@
 #include <string.h>
 #include <stdint.h>
 #include "skein-hash-4way.h"
-#include "algo/sha/sha2-hash-4way.h"
+#if defined(__SHA__)
+  #include <openssl/sha.h>
+#else
+  #include "algo/sha/sha2-hash-4way.h"
+#endif
 
 #if defined (SKEIN_4WAY)
 
@@ -11,53 +15,69 @@ void skeinhash_4way( void *state, const void *input )
      uint64_t vhash64[8*4] __attribute__ ((aligned (64)));
      uint32_t vhash32[16*4] __attribute__ ((aligned (64)));
      skein512_4way_context ctx_skein;
+#if defined(__SHA__)
+     uint32_t hash0[16] __attribute__ ((aligned (64)));
+     uint32_t hash1[16] __attribute__ ((aligned (64)));
+     uint32_t hash2[16] __attribute__ ((aligned (64)));
+     uint32_t hash3[16] __attribute__ ((aligned (64)));
+     SHA256_CTX           ctx_sha256;
+#else
      sha256_4way_context ctx_sha256;
+#endif
 
      skein512_4way_init( &ctx_skein );
      skein512_4way( &ctx_skein, input, 80 );
      skein512_4way_close( &ctx_skein, vhash64 );
 
-     mm256_reinterleave_4x32( vhash32, vhash64, 512 );
+#if defined(__SHA__)      
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 512 );
+
+     SHA256_Init( &ctx_sha256 );
+     SHA256_Update( &ctx_sha256, (unsigned char*)hash0, 64 );
+     SHA256_Final( (unsigned char*)hash0, &ctx_sha256 );
+
+     SHA256_Init( &ctx_sha256 );
+     SHA256_Update( &ctx_sha256, (unsigned char*)hash1, 64 );
+     SHA256_Final( (unsigned char*)hash1, &ctx_sha256 );
+
+     SHA256_Init( &ctx_sha256 );
+     SHA256_Update( &ctx_sha256, (unsigned char*)hash2, 64 );
+     SHA256_Final( (unsigned char*)hash2, &ctx_sha256 );
+
+     SHA256_Init( &ctx_sha256 );
+     SHA256_Update( &ctx_sha256, (unsigned char*)hash3, 64 );
+     SHA256_Final( (unsigned char*)hash3, &ctx_sha256 );
+
+     mm128_intrlv_4x32( state, hash0, hash1, hash2, hash3, 256 );
+#else
+     mm256_rintrlv_4x64_4x32( vhash32, vhash64, 512 );
 
      sha256_4way_init( &ctx_sha256 );
      sha256_4way( &ctx_sha256, vhash32, 64 );
      sha256_4way_close( &ctx_sha256, state );
-
-     mm128_deinterleave_4x32( state, state+32, state+64, state+96,
-		              vhash32, 256 );
+#endif
 }
 
 int scanhash_skein_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                    uint64_t *hashes_done )
+                         uint64_t *hashes_done, struct thr_info *mythr )
 {
     uint32_t vdata[20*4] __attribute__ ((aligned (64)));
     uint32_t hash[8*4] __attribute__ ((aligned (64)));
     uint32_t lane_hash[8];
     uint32_t *hash7 = &(hash[7<<2]);
-    uint32_t edata[20] __attribute__ ((aligned (64)));
     uint32_t *pdata = work->data;
     uint32_t *ptarget = work->target;
     const uint32_t Htarg = ptarget[7];
     const uint32_t first_nonce = pdata[19];
     uint32_t n = first_nonce;
-    // hash is returned deinterleaved
-    uint32_t *nonces = work->nonces;
-    int num_found = 0;
-
-// data is 80 bytes, 20 u32 or 4 u64.
-	
-    swab32_array( edata, pdata, 20 );
- 
-    mm256_interleave_4x64( vdata, edata, edata, edata, edata, 640 );
-
-    uint32_t *noncep = vdata + 73;   // 9*8 + 1
+    __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
+    /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
+   mm256_bswap_intrlv80_4x64( vdata, pdata );
    do
    {
-       be32enc( noncep,   n   );
-       be32enc( noncep+2, n+1 );
-       be32enc( noncep+4, n+2 );
-       be32enc( noncep+6, n+3 );
+       *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
+                _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
 
        skeinhash_4way( hash, vdata );
 
@@ -68,16 +88,14 @@ int scanhash_skein_4way( int thr_id, struct work *work, uint32_t max_nonce,
           if ( fulltest( lane_hash, ptarget ) )
           {
              pdata[19] = n + lane;
-             nonces[ num_found++ ] = n + lane;
-             work_set_target_ratio( work, lane_hash );
+             submit_solution( work, lane_hash, mythr, lane );
           }
        }
        n += 4;
-    } while ( (num_found == 0) && (n < max_nonce)
-               && !work_restart[thr_id].restart );
+    } while ( (n < max_nonce) && !work_restart[thr_id].restart );
 
     *hashes_done = n - first_nonce + 1;
-    return num_found;
+    return 0;
 }
 
 #endif

algo/skein/skein-gate.h

@@ -3,24 +3,21 @@
 #include <stdint.h>
 #include "algo-gate-api.h"
 
-// Override multi way on ryzen, SHA is better.
-#if !defined(RYZEN_)
 #if defined(__AVX2__)
   #define SKEIN_4WAY
 #endif
-#endif
 
 #if defined(SKEIN_4WAY)
 
 void skeinhash_4way( void *output, const void *input );
 
 int scanhash_skein_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                         uint64_t *hashes_done );
+                         uint64_t *hashes_done, struct thr_info *mythr );
 #endif
 
 void skeinhash( void *output, const void *input );
 
 int scanhash_skein( int thr_id, struct work *work, uint32_t max_nonce,
-                    uint64_t *hashes_done );
+                    uint64_t *hashes_done, struct thr_info *mythr );
 
 #endif

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

@@ -49,7 +49,7 @@ extern "C"{
 
 #include <stddef.h>
 #include "algo/sha/sph_types.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 // Output size in bits
 #define SPH_SIZE_skein256   256

algo/skein/skein.c

@@ -21,8 +21,8 @@ void skeinhash(void *state, const void *input)
      memcpy(state, hash, 32);
 }
 
-int scanhash_skein(int thr_id, struct work *work,
-	uint32_t max_nonce, uint64_t *hashes_done)
+int scanhash_skein( int thr_id, struct work *work, uint32_t max_nonce,
+                    uint64_t *hashes_done, struct thr_info *mythr )
 {
         uint32_t *pdata = work->data;
         uint32_t *ptarget = work->target;
@@ -31,7 +31,8 @@ int scanhash_skein(int thr_id, struct work *work,
 	const uint32_t Htarg = ptarget[7];
 	const uint32_t first_nonce = pdata[19];
 	uint32_t n = first_nonce;
-	
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
+
         swab32_array( endiandata, pdata, 20 );
 
 	do {

algo/skein/skein2-4way.c

@@ -20,55 +20,43 @@ void skein2hash_4way( void *output, const void *input )
 }
 
 int scanhash_skein2_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                          uint64_t *hashes_done )
+                          uint64_t *hashes_done, struct thr_info *mythr )
 {
     uint32_t hash[8*4] __attribute__ ((aligned (64)));
     uint32_t *hash7 = &(hash[25]);
     uint32_t vdata[20*4] __attribute__ ((aligned (64)));
-    uint32_t endiandata[20] __attribute__ ((aligned (64)));
-    uint64_t *edata = (uint64_t*)endiandata;
     uint32_t *pdata = work->data;
     uint32_t *ptarget = work->target;
     const uint32_t Htarg = ptarget[7];
     const uint32_t first_nonce = pdata[19];
     uint32_t n = first_nonce;
-    uint32_t *nonces = work->nonces;
-    int num_found = 0;
-
-    swab32_array( endiandata, pdata, 20 );
-
-    mm256_interleave_4x64( vdata, edata, edata, edata, edata, 640 );
-
-    uint32_t *noncep = vdata + 73;   // 9*8 + 1
+    __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
+    /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
+    mm256_bswap_intrlv80_4x64( vdata, pdata );
     do 
     {
-       be32enc( noncep,   n   );
-       be32enc( noncep+2, n+1 );
-       be32enc( noncep+4, n+2 );
-       be32enc( noncep+6, n+3 );
+       *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
+                _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
 
-       skein2hash( hash, vdata );
+       skein2hash_4way( hash, vdata );
 
        for ( int lane = 0; lane < 4; lane++ )
-       if ( hash7[ lane ] <= Htarg )
+       if ( hash7[ lane<<1 ] <= Htarg )
        {
-          // deinterleave hash for lane
           uint32_t lane_hash[8];
           mm256_extract_lane_4x64( lane_hash, hash, lane, 256 );
-          if ( fulltest( lane_hash, ptarget ) )
+          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
           {
              pdata[19] = n + lane;
-             nonces[ num_found++ ] = n + lane;
-             work_set_target_ratio( work, lane_hash );
+             submit_solution( work, lane_hash, mythr, lane );
           }
        }
        n += 4;
-    } while ( (num_found == 0) && (n < max_nonce)
-             &&  !work_restart[thr_id].restart );
+    } while ( (n < max_nonce) && !work_restart[thr_id].restart );
 
     *hashes_done = n - first_nonce + 1;
-    return num_found;
+    return 0;
 }
 
 #endif

algo/skein/skein2-gate.c

@@ -10,10 +10,9 @@ int64_t skein2_get_max64 ()
 bool register_skein2_algo( algo_gate_t* gate )
 {
   gate->optimizations = AVX2_OPT;
-#if defined (FOUR_WAY) && defined (__AVX2__)
+#if defined (SKEIN2_4WAY)
   gate->scanhash  = (void*)&scanhash_skein2_4way;
   gate->hash      = (void*)&skein2hash_4way;
-  four_way_not_tested();
 #else
   gate->scanhash  = (void*)&scanhash_skein2;
   gate->hash      = (void*)&skein2hash;

algo/skein/skein2-gate.h

@@ -10,11 +10,11 @@
 #if defined(SKEIN2_4WAY)
 void skein2hash_4way( void *output, const void *input );
 int scanhash_skein2_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                          uint64_t* hashes_done );
+                          uint64_t* hashes_done, struct thr_info *mythr );
 #endif
 
 void skein2hash( void *output, const void *input );
 int scanhash_skein2( int thr_id, struct work *work, uint32_t max_nonce,
-                     uint64_t *hashes_done );
+                     uint64_t *hashes_done, struct thr_info *mythr );
 #endif
 

algo/skein/skein2.c

@@ -34,8 +34,8 @@ void skein2hash(void *output, const void *input)
 
 }
 
-int scanhash_skein2(int thr_id, struct work *work,
-	uint32_t max_nonce, uint64_t *hashes_done)
+int scanhash_skein2( int thr_id, struct work *work,	uint32_t max_nonce,
+                     uint64_t *hashes_done, struct thr_info *mythr )
 {
         uint32_t *pdata = work->data;
         uint32_t *ptarget = work->target;
@@ -44,6 +44,7 @@ int scanhash_skein2(int thr_id, struct work *work,
 	const uint32_t Htarg = ptarget[7];
 	const uint32_t first_nonce = pdata[19];
 	uint32_t n = first_nonce;
+    /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
         swab32_array( endiandata, pdata, 20 );
 

algo/sm3/sm3-hash-4way.h

@@ -59,7 +59,7 @@
 #include <sys/types.h>
 #include <stdint.h>
 #include <string.h>
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 #ifdef __cplusplus
 extern "C" {

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

@@ -52,7 +52,7 @@
 
 #include <stddef.h>
 #include "algo/sha/sph_types.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 
 /**
  * Output size (in bits) for WHIRLPOOL.

algo/x14/axiom.c

@@ -44,12 +44,13 @@ void axiomhash(void *output, const void *input)
 }
 
 int scanhash_axiom(int thr_id, struct work *work,
-	uint32_t max_nonce, uint64_t *hashes_done)
+	uint32_t max_nonce, uint64_t *hashes_done, struct thr_info *mythr )
 {
         uint32_t *pdata = work->data;
         uint32_t *ptarget = work->target;
 	uint32_t _ALIGN(64) hash64[8];
 	uint32_t _ALIGN(64) endiandata[20];
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
 	const uint32_t Htarg = ptarget[7];
 	const uint32_t first_nonce = pdata[19];

algo/x14/polytimos-4way.c

@@ -13,26 +13,16 @@
 #include "algo/gost/sph_gost.h"
 #include "algo/echo/aes_ni/hash_api.h"
 
-typedef struct {
+union _poly_4way_context_overlay
+{
    skein512_4way_context   skein;
    shabal512_4way_context  shabal;
    hashState_echo          echo;
    luffa_2way_context      luffa;
    sph_fugue512_context    fugue;
    sph_gost512_context     gost;
-} poly_4way_ctx_holder;
-
-poly_4way_ctx_holder poly_4way_ctx;
-
-void init_polytimos_4way_ctx()
-{
-    skein512_4way_init( &poly_4way_ctx.skein );
-    shabal512_4way_init( &poly_4way_ctx.shabal );
-    init_echo( &poly_4way_ctx.echo, 512  );
-    luffa_2way_init( &poly_4way_ctx.luffa, 512 );
-    sph_fugue512_init( &poly_4way_ctx.fugue );
-    sph_gost512_init( &poly_4way_ctx.gost );
-}
+};
+typedef union _poly_4way_context_overlay poly_4way_context_overlay;
 
 void polytimos_4way_hash( void *output, const void *input )
 {
@@ -41,51 +31,57 @@ void polytimos_4way_hash( void *output, const void *input )
      uint64_t hash2[8] __attribute__ ((aligned (64)));
      uint64_t hash3[8] __attribute__ ((aligned (64)));
      uint64_t vhash[8*4] __attribute__ ((aligned (64)));
-     poly_4way_ctx_holder ctx __attribute__ ((aligned (64)));
-     memcpy( &ctx, &poly_4way_ctx, sizeof(poly_4way_ctx) );
+     poly_4way_context_overlay ctx;
 
+     skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, input, 80 );
      skein512_4way_close( &ctx.skein, vhash );
 
      // Need to convert from 64 bit interleaved to 32 bit interleaved.
      uint32_t vhash32[16*4];
-     mm256_reinterleave_4x32( vhash32, vhash, 512 );
+     mm256_rintrlv_4x64_4x32( vhash32, vhash, 512 );
+     shabal512_4way_init( &ctx.shabal );
      shabal512_4way( &ctx.shabal, vhash32, 64 );
      shabal512_4way_close( &ctx.shabal, vhash32 );
-     mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash32, 512 );
+     mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash32, 512 );
 
+     init_echo( &ctx.echo, 512 );
      update_final_echo ( &ctx.echo, (BitSequence *)hash0,
                          (const BitSequence *)hash0, 512 );
-     memcpy( &ctx.echo, &poly_4way_ctx.echo, sizeof(hashState_echo) );
+     init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash1,
                        (const BitSequence *) hash1, 512 );
-     memcpy( &ctx.echo, &poly_4way_ctx.echo, sizeof(hashState_echo) );
+     init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash2,
                        (const BitSequence *) hash2, 512 );
-     memcpy( &ctx.echo, &poly_4way_ctx.echo, sizeof(hashState_echo) );
+     init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_interleave_2x128( vhash, hash0, hash1, 512 );
-     luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
-     mm256_interleave_2x128( vhash, hash2, hash3, 512 );
+     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
+     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     luffa_2way_init( &ctx.luffa, 512 );
+     luffa_2way_init( &ctx.luffa, 512 );
+     luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
+     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
 
+     sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
      sph_fugue512_close( &ctx.fugue, hash0 );
-     memcpy( &ctx.fugue, &poly_4way_ctx.fugue, sizeof(sph_fugue512_context) );
+     sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash1, 64 );
      sph_fugue512_close( &ctx.fugue, hash1 );
-     memcpy( &ctx.fugue, &poly_4way_ctx.fugue, sizeof(sph_fugue512_context) );
+     sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash2, 64 );
      sph_fugue512_close( &ctx.fugue, hash2 );
-     memcpy( &ctx.fugue, &poly_4way_ctx.fugue, sizeof(sph_fugue512_context) );
+     sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash3, 64 );
      sph_fugue512_close( &ctx.fugue, hash3 );
 
+     sph_gost512_init( &ctx.gost );
      sph_gost512( &ctx.gost, hash0, 64 );
      sph_gost512_close( &ctx.gost, hash0 );
      sph_gost512_init( &ctx.gost );
@@ -104,51 +100,43 @@ void polytimos_4way_hash( void *output, const void *input )
      memcpy( output+96, hash3, 32 );
 }
 
-int scanhash_polytimos_4way( int thr_id, struct work *work, uint32_t max_nonce,                              uint64_t *hashes_done )
+int scanhash_polytimos_4way( int thr_id, struct work *work, uint32_t max_nonce,
+                             uint64_t *hashes_done, struct thr_info *mythr )
 {
    uint32_t hash[4*8] __attribute__ ((aligned (64)));
    uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-   uint32_t endiandata[20] __attribute__((aligned(64)));
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t first_nonce = pdata[19];
    uint32_t n = first_nonce;
-   uint32_t *nonces = work->nonces;
-   int num_found = 0;
-   uint32_t *noncep = vdata + 73;   // 9*8 + 1
+     __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
    const uint32_t Htarg = ptarget[7];
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
    volatile uint8_t *restart = &(work_restart[thr_id].restart);
 
    if ( opt_benchmark )
       ptarget[7] = 0x0cff;
 
-   for ( int i=0; i < 19; i++ )
-      be32enc( &endiandata[i], pdata[i] );
-
-   uint64_t *edata = (uint64_t*)endiandata;
-   mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+   mm256_bswap_intrlv80_4x64( vdata, pdata );
    do {
-      be32enc( noncep,   n   );
-      be32enc( noncep+2, n+1 );
-      be32enc( noncep+4, n+2 );
-      be32enc( noncep+6, n+3 );
+      *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
+                 _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
 
       polytimos_4way_hash(hash, vdata);
       pdata[19] = n;
 
-      for ( int i = 0; i < 4; i++ )
-      if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
+      for ( int i = 0; i < 4; i++ ) if ( (hash+(i<<3))[7] <= Htarg )
+      if( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
       {
          pdata[19] = n+i;
-         nonces[ num_found++ ] = n+i;
-         work_set_target_ratio( work, hash+(i<<3) );
+         submit_solution( work, hash+(i<<3), mythr, i );
       }
       n += 4;
 
-   } while ( ( num_found == 0 ) && ( n < max_nonce ) && !(*restart));
+   } while ( ( n < max_nonce-4 ) && !(*restart));
 
    *hashes_done = n - first_nonce + 1;
-   return num_found;
+   return 0;
 }
 
 #endif

algo/x14/polytimos-gate.c

@@ -4,7 +4,6 @@ bool register_polytimos_algo( algo_gate_t* gate )
 {
   gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT;
 #ifdef POLYTIMOS_4WAY
-  init_polytimos_4way_ctx();
   gate->scanhash  = (void*)&scanhash_polytimos_4way;
   gate->hash      = (void*)&polytimos_4way_hash;
 #else

algo/x14/polytimos-gate.h

@@ -13,19 +13,14 @@ bool register_polytimos_algo( algo_gate_t* gate );
 #if defined(POLYTIMOS_4WAY)
 
 void polytimos_4way_hash( void *state, const void *input );
-
 int scanhash_polytimos_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                       uint64_t *hashes_done );
-
-void init_polytimos_4way_ctx();
+                       uint64_t *hashes_done, struct thr_info *mythr );
 
 #endif
 
 void polytimos_hash( void *state, const void *input );
-
 int scanhash_polytimos( int thr_id, struct work *work, uint32_t max_nonce,
-                  uint64_t *hashes_done );
-
+                  uint64_t *hashes_done, struct thr_info *mythr );
 void init_polytimos_ctx();
 
 #endif

algo/x14/polytimos.c

@@ -76,13 +76,14 @@ void polytimos_hash(void *output, const void *input)
 	memcpy(output, hashA, 32);
 }
 
-int scanhash_polytimos(int thr_id, struct work *work, uint32_t max_nonce, uint64_t *hashes_done)
+int scanhash_polytimos( int thr_id, struct work *work, uint32_t max_nonce,
+                        uint64_t *hashes_done, struct thr_info *mythr )
 {
 	uint32_t _ALIGN(128) hash[8];
 	uint32_t _ALIGN(128) endiandata[20];
 	uint32_t *pdata = work->data;
 	uint32_t *ptarget = work->target;
-
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 	const uint32_t Htarg = ptarget[7];
 	const uint32_t first_nonce = pdata[19];
 	uint32_t nonce = first_nonce;

algo/x14/veltor-4way.c

@@ -40,7 +40,7 @@ void veltor_4way_hash( void *output, const void *input )
 
      skein512_4way( &ctx.skein, input, 80 );
      skein512_4way_close( &ctx.skein, vhash );
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_shavite512( &ctx.shavite, hash0, 64 );
      sph_shavite512_close( &ctx.shavite, hash0 );
@@ -54,10 +54,10 @@ void veltor_4way_hash( void *output, const void *input )
      sph_shavite512( &ctx.shavite, hash3, 64 );
      sph_shavite512_close( &ctx.shavite, hash3 );
 
-     mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
      shabal512_4way( &ctx.shabal, vhash, 64 );
      shabal512_4way_close( &ctx.shabal, vhash );
-     mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_gost512( &ctx.gost, hash0, 64 );
      sph_gost512_close( &ctx.gost, hash0 );
@@ -78,7 +78,7 @@ void veltor_4way_hash( void *output, const void *input )
 }
 
 int scanhash_veltor_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                          uint64_t *hashes_done )
+                          uint64_t *hashes_done, struct thr_info *mythr )
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
@@ -91,6 +91,7 @@ int scanhash_veltor_4way( int thr_id, struct work *work, uint32_t max_nonce,
      uint32_t *nonces = work->nonces;
      int num_found = 0;
      uint32_t *noncep = vdata + 73;   // 9*8 + 1
+     /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
      volatile uint8_t *restart = &(work_restart[thr_id].restart);
 
      if ( opt_benchmark )
@@ -101,7 +102,7 @@ int scanhash_veltor_4way( int thr_id, struct work *work, uint32_t max_nonce,
      }
 
      uint64_t *edata = (uint64_t*)endiandata;
-     mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+     mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
      do
      {
          be32enc( noncep,   n   );

algo/x14/veltor-gate.h

@@ -15,7 +15,7 @@ bool register_veltor_algo( algo_gate_t* gate );
 void veltor_4way_hash( void *state, const void *input );
 
 int scanhash_veltor_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                       uint64_t *hashes_done );
+                       uint64_t *hashes_done, struct thr_info *mythr );
 
 void init_veltor_4way_ctx();
 
@@ -24,7 +24,7 @@ void init_veltor_4way_ctx();
 void veltor_hash( void *state, const void *input );
 
 int scanhash_veltor( int thr_id, struct work *work, uint32_t max_nonce,
-                  uint64_t *hashes_done );
+                  uint64_t *hashes_done, struct thr_info *mythr );
 
 void init_veltor_ctx();
 

algo/x14/veltor.c

@@ -61,12 +61,14 @@ void veltor_hash(void *output, const void *input)
 	memcpy(output, hashB, 32);
 }
 
-int scanhash_veltor(int thr_id, struct work *work, uint32_t max_nonce, uint64_t *hashes_done)
+int scanhash_veltor( int thr_id, struct work *work, uint32_t max_nonce,
+                     uint64_t *hashes_done, struct thr_info *mythr )
 {
 	uint32_t _ALIGN(128) hash[8];
 	uint32_t _ALIGN(128) endiandata[20];
 	uint32_t *pdata = work->data;
 	uint32_t *ptarget = work->target;
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
 	const uint32_t Htarg = ptarget[7];
 	const uint32_t first_nonce = pdata[19];

algo/x14/x14-4way.c

@@ -78,7 +78,7 @@ void x14_4way_hash( void *state, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serial
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 3 Groestl
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -90,7 +90,7 @@ void x14_4way_hash( void *state, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
      // Parallel 4way
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      // 4 Skein
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -105,16 +105,16 @@ void x14_4way_hash( void *state, const void *input )
      keccak512_4way_close( &ctx.keccak, vhash );
 
      // Serial
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 7 Luffa
-     mm256_interleave_2x128( vhash, hash0, hash1, 512 );
+     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
-     mm256_interleave_2x128( vhash, hash2, hash3, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
+     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
+     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 8 Cubehash
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
@@ -142,13 +142,13 @@ void x14_4way_hash( void *state, const void *input )
      sph_shavite512_close( &ctx.shavite, hash3 );
 
      // 10 Simd
-     mm256_interleave_2x128( vhash, hash0, hash1, 512 );
+     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
-     mm256_interleave_2x128( vhash, hash2, hash3, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
+     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
+     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 11 Echo
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -164,10 +164,10 @@ void x14_4way_hash( void *state, const void *input )
                        (const BitSequence *) hash3, 512 );
 
      // 12 Hamsi parallel 4way 32 bit
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 13 Fugue serial
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -183,14 +183,14 @@ void x14_4way_hash( void *state, const void *input )
      sph_fugue512_close( &ctx.fugue, hash3 );
 
      // 14 Shabal, parallel 32 bit
-     mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
      shabal512_4way( &ctx.shabal, vhash, 64 );
      shabal512_4way_close( &ctx.shabal, state );
 
 }
 
 int scanhash_x14_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                       uint64_t *hashes_done )
+                       uint64_t *hashes_done, struct thr_info *mythr )
 {
      uint32_t hash[4*16] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
@@ -203,6 +203,7 @@ int scanhash_x14_4way( int thr_id, struct work *work, uint32_t max_nonce,
      int num_found = 0;
      uint32_t *noncep = vdata + 73;   // 9*8 + 1
      const uint32_t Htarg = ptarget[7];
+     /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
      uint64_t htmax[] = {          0,        0xF,       0xFF,
                                0xFFF,     0xFFFF, 0x10000000  };
      uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
@@ -212,7 +213,7 @@ int scanhash_x14_4way( int thr_id, struct work *work, uint32_t max_nonce,
      swab32_array( endiandata, pdata, 20 );
 
      uint64_t *edata = (uint64_t*)endiandata;
-     mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+     mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
 
      for ( int m=0; m < 6; m++ )
        if ( Htarg <= htmax[m] )

algo/x14/x14-gate.h

@@ -13,19 +13,15 @@ bool register_x14_algo( algo_gate_t* gate );
 #if defined(X14_4WAY)
 
 void x14_4way_hash( void *state, const void *input );
-
 int scanhash_x14_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                       uint64_t *hashes_done );
-
+                       uint64_t *hashes_done, struct thr_info *mythr );
 void init_x14_4way_ctx();
 
 #endif
 
 void x14hash( void *state, const void *input );
-
 int scanhash_x14( int thr_id, struct work *work, uint32_t max_nonce,
-                  uint64_t *hashes_done );
-
+                  uint64_t *hashes_done, struct thr_info *mythr );
 void init_x14_ctx();
 
 #endif

algo/x14/x14.c

@@ -180,8 +180,8 @@ void x14hash(void *output, const void *input)
 	memcpy(output, hash, 32);
 }
 
-int scanhash_x14(int thr_id, struct work *work,
-				uint32_t max_nonce, uint64_t *hashes_done)
+int scanhash_x14( int thr_id, struct work *work, uint32_t max_nonce,
+                  uint64_t *hashes_done, struct thr_info *mythr )
 {
         uint32_t endiandata[20] __attribute__((aligned(64)));
         uint32_t hash64[8] __attribute__((aligned(64)));
@@ -190,6 +190,7 @@ int scanhash_x14(int thr_id, struct work *work,
 	uint32_t n = pdata[19] - 1;
 	const uint32_t first_nonce = pdata[19];
 	const uint32_t Htarg = ptarget[7];
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
 	uint64_t htmax[] = {
 		0,

algo/x15/x15-4way.c

@@ -81,7 +81,7 @@ void x15_4way_hash( void *state, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serial
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 3 Groestl
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -93,7 +93,7 @@ void x15_4way_hash( void *state, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
      // Parallel 4way
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      // 4 Skein
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -108,16 +108,16 @@ void x15_4way_hash( void *state, const void *input )
      keccak512_4way_close( &ctx.keccak, vhash );
 
      // Serial to the end
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 7 Luffa
-     mm256_interleave_2x128( vhash, hash0, hash1, 512 );
+     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
-     mm256_interleave_2x128( vhash, hash2, hash3, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
+     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
+     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 8 Cubehash
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
@@ -145,13 +145,13 @@ void x15_4way_hash( void *state, const void *input )
      sph_shavite512_close( &ctx.shavite, hash3 );
 
      // 10 Simd
-     mm256_interleave_2x128( vhash, hash0, hash1, 512 );
+     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
-     mm256_interleave_2x128( vhash, hash2, hash3, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
+     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
+     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 11 Echo
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -167,10 +167,10 @@ void x15_4way_hash( void *state, const void *input )
                        (const BitSequence *) hash3, 512 );
 
      // 12 Hamsi parallel 4way 32 bit
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 13 Fugue
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -186,10 +186,10 @@ void x15_4way_hash( void *state, const void *input )
      sph_fugue512_close( &ctx.fugue, hash3 );
 
      // 14 Shabal, parallel 32 bit
-     mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
      shabal512_4way( &ctx.shabal, vhash, 64 );
      shabal512_4way_close( &ctx.shabal, vhash );
-     mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
        
      // 15 Whirlpool
      sph_whirlpool( &ctx.whirlpool, hash0, 64 );
@@ -214,7 +214,7 @@ void x15_4way_hash( void *state, const void *input )
 }
 
 int scanhash_x15_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                       uint64_t *hashes_done )
+                       uint64_t *hashes_done, struct thr_info *mythr )
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
@@ -227,6 +227,7 @@ int scanhash_x15_4way( int thr_id, struct work *work, uint32_t max_nonce,
      int num_found = 0;
      uint32_t *noncep = vdata + 73;   // 9*8 + 1
      const uint32_t Htarg = ptarget[7];
+     /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
      uint64_t htmax[] = {          0,        0xF,       0xFF,
                                0xFFF,     0xFFFF, 0x10000000  };
      uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
@@ -236,7 +237,7 @@ int scanhash_x15_4way( int thr_id, struct work *work, uint32_t max_nonce,
      swab32_array( endiandata, pdata, 20 );
 
      uint64_t *edata = (uint64_t*)endiandata;
-     mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+     mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
 
      for ( int m=0; m < 6; m++ )
        if ( Htarg <= htmax[m] )

algo/x15/x15-gate.h

@@ -13,19 +13,15 @@ bool register_x15_algo( algo_gate_t* gate );
 #if defined(X15_4WAY)
 
 void x15_4way_hash( void *state, const void *input );
-
 int scanhash_x15_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                       uint64_t *hashes_done );
-
+                       uint64_t *hashes_done, struct thr_info *mythr );
 void init_x15_4way_ctx();
 
 #endif
 
 void x15hash( void *state, const void *input );
-
 int scanhash_x15( int thr_id, struct work *work, uint32_t max_nonce,
-                  uint64_t *hashes_done );
-
+                  uint64_t *hashes_done, struct thr_info *mythr );
 void init_x15_ctx();
 
 #endif

algo/x15/x15.c

@@ -186,8 +186,8 @@ void x15hash(void *output, const void *input)
 	memcpy(output, hashB, 32);
 }
 
-int scanhash_x15(int thr_id, struct work *work,
-                    uint32_t max_nonce, uint64_t *hashes_done)
+int scanhash_x15( int thr_id, struct work *work, uint32_t max_nonce,
+                  uint64_t *hashes_done, struct thr_info *mythr )
 {
         uint32_t endiandata[20] __attribute__((aligned(64)));
         uint32_t hash64[8] __attribute__((aligned(64)));
@@ -196,6 +196,7 @@ int scanhash_x15(int thr_id, struct work *work,
 	uint32_t n = pdata[19] - 1;
 	const uint32_t first_nonce = pdata[19];
 	const uint32_t Htarg = ptarget[7];
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
 
 	uint64_t htmax[] = {
 		0,

algo/x16/x16r-4way.c

@@ -32,8 +32,8 @@
 static __thread uint32_t s_ntime = UINT32_MAX;
 static __thread char hashOrder[X16R_HASH_FUNC_COUNT + 1] = { 0 };
 
-
-typedef struct {
+union _x16r_4way_context_overlay
+{
     blake512_4way_context   blake;
     bmw512_4way_context     bmw;
     hashState_echo          echo;
@@ -50,16 +50,8 @@ typedef struct {
     shabal512_4way_context  shabal;
     sph_whirlpool_context   whirlpool;
     sha512_4way_context     sha512;
-} x16r_4way_ctx_holder;
-
-x16r_4way_ctx_holder x16r_4way_ctx __attribute__ ((aligned (64)));
-
-// Cube needs one full init so fast reinits can be done in the hash loop.
-void init_x16r_4way_ctx()
-{
-   cubehashInit( &x16r_4way_ctx.cube, 512, 16, 32 );
 };
-
+typedef union _x16r_4way_context_overlay x16r_4way_context_overlay;
 
 void x16r_4way_hash( void* output, const void* input )
 {
@@ -68,14 +60,14 @@ void x16r_4way_hash( void* output, const void* input )
    uint32_t hash2[24] __attribute__ ((aligned (64)));
    uint32_t hash3[24] __attribute__ ((aligned (64)));
    uint32_t vhash[24*4] __attribute__ ((aligned (64)));
-   x16r_4way_ctx_holder ctx;
+   x16r_4way_context_overlay ctx;
    void *in0 = (void*) hash0;
    void *in1 = (void*) hash1;
    void *in2 = (void*) hash2;
    void *in3 = (void*) hash3;
    int size = 80;
 
-   mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, input, 640 );
+   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, input, 640 );
  
    if ( s_ntime == UINT32_MAX )
    {
@@ -104,11 +96,11 @@ void x16r_4way_hash( void* output, const void* input )
                blake512_4way( &ctx.blake, input, size );
             else
             {
-               mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
+               mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
                blake512_4way( &ctx.blake, vhash, size );
             }
             blake512_4way_close( &ctx.blake, vhash );
-            mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case BMW:
             bmw512_4way_init( &ctx.bmw );
@@ -116,11 +108,11 @@ void x16r_4way_hash( void* output, const void* input )
                bmw512_4way( &ctx.bmw, input, size );
             else
             {
-               mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
+               mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
                bmw512_4way( &ctx.bmw, vhash, size );
             }
             bmw512_4way_close( &ctx.bmw, vhash );
-            mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case GROESTL:
                init_groestl( &ctx.groestl, 64 );
@@ -142,11 +134,11 @@ void x16r_4way_hash( void* output, const void* input )
                skein512_4way( &ctx.skein, input, size );
             else
             {
-               mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
+               mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
                skein512_4way( &ctx.skein, vhash, size );
             }
             skein512_4way_close( &ctx.skein, vhash );
-            mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case JH:
             jh512_4way_init( &ctx.jh );
@@ -154,11 +146,11 @@ void x16r_4way_hash( void* output, const void* input )
                jh512_4way( &ctx.jh, input, size );
             else
             {
-               mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
+               mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
                jh512_4way( &ctx.jh, vhash, size );
             }
             jh512_4way_close( &ctx.jh, vhash );
-            mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case KECCAK:
             keccak512_4way_init( &ctx.keccak );
@@ -166,21 +158,21 @@ void x16r_4way_hash( void* output, const void* input )
                keccak512_4way( &ctx.keccak, input, size );
             else
             {
-               mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
+               mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
                keccak512_4way( &ctx.keccak, vhash, size );
             }
             keccak512_4way_close( &ctx.keccak, vhash );
-            mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case LUFFA:
-            mm256_interleave_2x128( vhash, in0, in1, size<<3 );
+            mm256_intrlv_2x128( vhash, in0, in1, size<<3 );
             luffa_2way_init( &ctx.luffa, 512 );
             luffa_2way_update_close( &ctx.luffa, vhash, vhash, size );
-            mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
-            mm256_interleave_2x128( vhash, in2, in3, size<<3 );
+            mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
+            mm256_intrlv_2x128( vhash, in2, in3, size<<3 );
             luffa_2way_init( &ctx.luffa, 512 );
             luffa_2way_update_close( &ctx.luffa, vhash, vhash, size);
-            mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
+            mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
          break;
          case CUBEHASH:
             cubehashInit( &ctx.cube, 512, 16, 32 );
@@ -194,7 +186,7 @@ void x16r_4way_hash( void* output, const void* input )
                                   (const byte*)in2, size );
             cubehashInit( &ctx.cube, 512, 16, 32 );
             cubehashUpdateDigest( &ctx.cube, (byte*) hash3,
-                                        (const byte*)in3, size );
+                                  (const byte*)in3, size );
          break;
          case SHAVITE:
             sph_shavite512_init( &ctx.shavite );
@@ -211,14 +203,14 @@ void x16r_4way_hash( void* output, const void* input )
             sph_shavite512_close( &ctx.shavite, hash3 );
          break;
          case SIMD:
-            mm256_interleave_2x128( vhash, in0, in1, size<<3 );
+            mm256_intrlv_2x128( vhash, in0, in1, size<<3 );
             simd_2way_init( &ctx.simd, 512 );
             simd_2way_update_close( &ctx.simd, vhash, vhash, size<<3 );
-            mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
-            mm256_interleave_2x128( vhash, in2, in3, size<<3 );
+            mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
+            mm256_intrlv_2x128( vhash, in2, in3, size<<3 );
             simd_2way_init( &ctx.simd, 512 );
             simd_2way_update_close( &ctx.simd, vhash, vhash, size<<3 );
-            mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
+            mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
          break;
          case ECHO:
              init_echo( &ctx.echo, 512 );
@@ -235,11 +227,11 @@ void x16r_4way_hash( void* output, const void* input )
                                 (const BitSequence*)in3, size<<3 );
          break;
          case HAMSI:
-             mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
+             mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
              hamsi512_4way_init( &ctx.hamsi );
              hamsi512_4way( &ctx.hamsi, vhash, size );
              hamsi512_4way_close( &ctx.hamsi, vhash );
-             mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+             mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case FUGUE:
              sph_fugue512_init( &ctx.fugue );
@@ -256,11 +248,11 @@ void x16r_4way_hash( void* output, const void* input )
              sph_fugue512_close( &ctx.fugue, hash3 );
          break;
          case SHABAL:
-             mm128_interleave_4x32( vhash, in0, in1, in2, in3, size<<3 );
+             mm128_intrlv_4x32( vhash, in0, in1, in2, in3, size<<3 );
              shabal512_4way_init( &ctx.shabal );
              shabal512_4way( &ctx.shabal, vhash, size );
              shabal512_4way_close( &ctx.shabal, vhash );
-             mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
+             mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case WHIRLPOOL:
              sph_whirlpool_init( &ctx.whirlpool );
@@ -277,11 +269,11 @@ void x16r_4way_hash( void* output, const void* input )
              sph_whirlpool_close( &ctx.whirlpool, hash3 );
          break;
          case SHA_512:
-             mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
+             mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
              sha512_4way_init( &ctx.sha512 );
              sha512_4way( &ctx.sha512, vhash, size );
              sha512_4way_close( &ctx.sha512, vhash );
-             mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+             mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
       }
       size = 64;
@@ -293,7 +285,7 @@ void x16r_4way_hash( void* output, const void* input )
 }
 
 int scanhash_x16r_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                        uint64_t *hashes_done )
+                        uint64_t *hashes_done, struct thr_info *mythr)
 {
    uint32_t hash[4*16] __attribute__ ((aligned (64)));
    uint32_t vdata[24*4] __attribute__ ((aligned (64)));
@@ -303,15 +295,14 @@ int scanhash_x16r_4way( int thr_id, struct work *work, uint32_t max_nonce,
    const uint32_t Htarg = ptarget[7];
    const uint32_t first_nonce = pdata[19];
    uint32_t n = first_nonce;
-   uint32_t *nonces = work->nonces;
-   int num_found = 0;
-   uint32_t *noncep = vdata + 73;   // 9*8 + 1
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
+    __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
    volatile uint8_t *restart = &(work_restart[thr_id].restart);
 
-   for ( int k=0; k < 19; k++ )
-      be32enc( &endiandata[k], pdata[k] );
+   casti_m256i( endiandata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
+   casti_m256i( endiandata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
+   casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
 
-//   if ( s_ntime != pdata[17] )
    if ( s_ntime != endiandata[17] )
    {
       uint32_t ntime = swab32(pdata[17]);
@@ -325,30 +316,27 @@ int scanhash_x16r_4way( int thr_id, struct work *work, uint32_t max_nonce,
       ptarget[7] = 0x0cff;
 
    uint64_t *edata = (uint64_t*)endiandata;
-   mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+   mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
 
    do
    {
-      be32enc( noncep,   n   );
-      be32enc( noncep+2, n+1 );
-      be32enc( noncep+4, n+2 );
-      be32enc( noncep+6, n+3 );
+      *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
+               _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
 
       x16r_4way_hash( hash, vdata );
       pdata[19] = n;
 
-      for ( int i = 0; i < 4; i++ )
-      if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
+      for ( int i = 0; i < 4; i++ )  if ( (hash+(i<<3))[7] <= Htarg )
+      if( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
       {
          pdata[19] = n+i;
-         nonces[ num_found++ ] = n+i;
-         work_set_target_ratio( work, hash+(i<<3) );
+         submit_solution( work, hash+(i<<3), mythr, i );
       }
       n += 4;
-   } while ( ( num_found == 0 ) && ( n < max_nonce ) && !(*restart) );
+   } while ( (  n < max_nonce ) && !(*restart) );
 
    *hashes_done = n - first_nonce + 1;
-   return num_found;
+   return 0;
 }
 
 #endif

algo/x16/x16r-gate.c

@@ -35,11 +35,9 @@ void x16s_getAlgoString( const uint8_t* prevblock, char *output )
 bool register_x16r_algo( algo_gate_t* gate )
 {
 #if defined (X16R_4WAY)
-  init_x16r_4way_ctx();
   gate->scanhash  = (void*)&scanhash_x16r_4way;
   gate->hash      = (void*)&x16r_4way_hash;
 #else
-  init_x16r_ctx();
   gate->scanhash  = (void*)&scanhash_x16r;
   gate->hash      = (void*)&x16r_hash;
 #endif
@@ -52,11 +50,9 @@ bool register_x16r_algo( algo_gate_t* gate )
 bool register_x16s_algo( algo_gate_t* gate )
 {
 #if defined (X16R_4WAY)
-  init_x16r_4way_ctx();
   gate->scanhash  = (void*)&scanhash_x16r_4way;
   gate->hash      = (void*)&x16r_4way_hash;
 #else
-  init_x16r_ctx();
   gate->scanhash  = (void*)&scanhash_x16r;
   gate->hash      = (void*)&x16r_hash;
 #endif

algo/x16/x16r-gate.h

@@ -2,7 +2,7 @@
 #define X16R_GATE_H__ 1
 
 #include "algo-gate-api.h"
-#include "avxdefs.h"
+#include "simd-utils.h"
 #include <stdint.h>
 
 #if defined(__AVX2__) && defined(__AES__)
@@ -39,20 +39,14 @@ bool register_x16s_algo( algo_gate_t* gate );
 #if defined(X16R_4WAY)
 
 void x16r_4way_hash( void *state, const void *input );
-
 int scanhash_x16r_4way( int thr_id, struct work *work, uint32_t max_nonce,
-                        uint64_t *hashes_done );
-
-void init_x16r_4way_ctx();
+                        uint64_t *hashes_done, struct thr_info *mythr );
 
 #endif
 
 void x16r_hash( void *state, const void *input );
-
 int scanhash_x16r( int thr_id, struct work *work, uint32_t max_nonce,
-                   uint64_t *hashes_done );
-
-void init_x16r_ctx();
+                   uint64_t *hashes_done, struct thr_info *mythr );
 
 #endif
 

algo/x16/x16r.c

@@ -33,7 +33,8 @@
 static __thread uint32_t s_ntime = UINT32_MAX;
 static __thread char hashOrder[X16R_HASH_FUNC_COUNT + 1] = { 0 };
 
-typedef struct {
+union _x16r_context_overlay
+{
 #if defined(__AES__)
         hashState_echo          echo;
         hashState_groestl       groestl;
@@ -55,19 +56,13 @@ typedef struct {
         sph_shabal512_context   shabal;
         sph_whirlpool_context   whirlpool;
         SHA512_CTX              sha512;
-} x16r_ctx_holder;
-
-x16r_ctx_holder x16r_ctx __attribute__ ((aligned (64)));
-
-void init_x16r_ctx()
-{
-   cubehashInit( &x16r_ctx.cube, 512, 16, 32 );
 };
+typedef union _x16r_context_overlay x16r_context_overlay;
 
 void x16r_hash( void* output, const void* input )
 {
    uint32_t _ALIGN(128) hash[16];
-   x16r_ctx_holder ctx;
+   x16r_context_overlay ctx;
    void *in = (void*) input;
    int size = 80;
 
@@ -126,7 +121,7 @@ void x16r_hash( void* output, const void* input )
                                     (const BitSequence*)in, size );
          break;
          case CUBEHASH:
-            memcpy( &ctx.cube, &x16r_ctx.cube, sizeof(cubehashParam) );
+            cubehashInit( &ctx.cube, 512, 16, 32 );
             cubehashUpdateDigest( &ctx.cube, (byte*) hash,
                                   (const byte*)in, size );
          break;
@@ -184,7 +179,7 @@ void x16r_hash( void* output, const void* input )
 }
 
 int scanhash_x16r( int thr_id, struct work *work, uint32_t max_nonce,
-                   uint64_t *hashes_done )
+                   uint64_t *hashes_done, struct thr_info *mythr )
 {
    uint32_t _ALIGN(128) hash32[8];
    uint32_t _ALIGN(128) endiandata[20];
@@ -192,16 +187,16 @@ int scanhash_x16r( int thr_id, struct work *work, uint32_t max_nonce,
    uint32_t *ptarget = work->target;
    const uint32_t Htarg = ptarget[7];
    const uint32_t first_nonce = pdata[19];
+   /* int */ thr_id = mythr->id;  // thr_id arg is deprecated
    uint32_t nonce = first_nonce;
    volatile uint8_t *restart = &(work_restart[thr_id].restart);
 
-   for ( int k=0; k < 19; k++ )
-      be32enc( &endiandata[k], pdata[k] );
+   casti_m128i( endiandata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
+   casti_m128i( endiandata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
+   casti_m128i( endiandata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
+   casti_m128i( endiandata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
+   casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
 
-// This code is suspicious. s_ntime is saved after byteswapping pdata[17]
-// but is tested vs unswapped pdata[17]. This should result in calling
-// getAlgoString every pass, but that doesn't seem to be the case.
-// It appears to be working correctly as is.
    if ( s_ntime != pdata[17] )
    {
       uint32_t ntime = swab32(pdata[17]);

algo/x17/sonoa-4way.c

@@ -47,30 +47,6 @@ union _sonoa_4way_context_overlay
 };
 
 typedef union _sonoa_4way_context_overlay sonoa_4way_context_overlay;
-/*
-sonoa_4way_ctx_holder sonoa_4way_ctx __attribute__ ((aligned (64)));
-
-void init_sonoa_4way_ctx()
-{
-     blake512_4way_init( &sonoa_4way_ctx.blake );
-     bmw512_4way_init( &sonoa_4way_ctx.bmw );
-     init_groestl( &sonoa_4way_ctx.groestl, 64 );
-     skein512_4way_init( &sonoa_4way_ctx.skein );
-     jh512_4way_init( &sonoa_4way_ctx.jh );
-     keccak512_4way_init( &sonoa_4way_ctx.keccak );
-     luffa_2way_init( &sonoa_4way_ctx.luffa, 512 );
-     cube_2way_init( &sonoa_4way_ctx.cube, 512, 16, 32 );
-     shavite512_2way_init( &sonoa_4way_ctx.shavite );
-     simd_2way_init( &sonoa_4way_ctx.simd, 512 );
-     init_echo( &sonoa_4way_ctx.echo, 512 );
-     hamsi512_4way_init( &sonoa_4way_ctx.hamsi );
-     sph_fugue512_init( &sonoa_4way_ctx.fugue );
-     shabal512_4way_init( &sonoa_4way_ctx.shabal );
-     sph_whirlpool_init( &sonoa_4way_ctx.whirlpool );
-     sha512_4way_init( &sonoa_4way_ctx.sha512 );
-     haval256_5_4way_init( &sonoa_4way_ctx.haval );
-};
-*/
 
 void sonoa_4way_hash( void *state, const void *input )
 {
@@ -82,8 +58,6 @@ void sonoa_4way_hash( void *state, const void *input )
      uint64_t vhashA[8*4] __attribute__ ((aligned (64)));
      uint64_t vhashB[8*4] __attribute__ ((aligned (64)));
      sonoa_4way_context_overlay ctx;
-//     sonoa_4way_ctx_holder ctx __attribute__ ((aligned (64)));
-//        memcpy( &ctx, &sonoa_4way_ctx, sizeof(sonoa_4way_ctx) );
 
 // 1
 
@@ -95,7 +69,7 @@ void sonoa_4way_hash( void *state, const void *input )
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -106,7 +80,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -120,7 +94,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -142,8 +116,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
-     mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
+     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -160,13 +134,13 @@ void sonoa_4way_hash( void *state, const void *input )
 
 // 2
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      bmw512_4way_init( &ctx.bmw );
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -177,7 +151,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -191,7 +165,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -213,8 +187,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
-     mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
+     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -229,7 +203,7 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
@@ -241,7 +215,7 @@ void sonoa_4way_hash( void *state, const void *input )
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -252,7 +226,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -266,7 +240,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -288,8 +262,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
-     mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
+     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -304,13 +278,13 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -326,13 +300,13 @@ void sonoa_4way_hash( void *state, const void *input )
      sph_fugue512_close( &ctx.fugue, hash3 );
 
 // 4
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      bmw512_4way_init( &ctx.bmw );
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -343,7 +317,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -357,7 +331,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -379,8 +353,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
-     mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
+     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -395,13 +369,13 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -416,19 +390,19 @@ void sonoa_4way_hash( void *state, const void *input )
      sph_fugue512( &ctx.fugue, hash3, 64 );
      sph_fugue512_close( &ctx.fugue, hash3 );
 
-     mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
 
      shabal512_4way_init( &ctx.shabal );
      shabal512_4way( &ctx.shabal, vhash, 64 );
      shabal512_4way_close( &ctx.shabal, vhash );
 
-     mm256_reinterleave_4x32_4x64( vhashB, vhash, 512 ); 
+     mm256_rintrlv_4x32_4x64( vhashB, vhash, 512 ); 
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhashB, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -443,8 +417,8 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_interleave_2x128( vhashA, hash0, hash1, 512 );
-     mm256_interleave_2x128( vhashB, hash2, hash3, 512 );
+     mm256_intrlv_2x128( vhashA, hash0, hash1, 512 );
+     mm256_intrlv_2x128( vhashB, hash2, hash3, 512 );
 
      shavite512_2way_init( &ctx.shavite );
      shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
@@ -452,19 +426,19 @@ void sonoa_4way_hash( void *state, const void *input )
      shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
 
 // 5
-     mm256_reinterleave_2x128_4x64( vhash, vhashA, vhashB, 512 );
+     mm256_rintrlv_2x128_4x64( vhash, vhashA, vhashB, 512 );
 
      bmw512_4way_init( &ctx.bmw );
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_reinterleave_4x64_4x32( vhashB, vhash,  512 );
+     mm256_rintrlv_4x64_4x32( vhashB, vhash,  512 );
 
      shabal512_4way_init( &ctx.shabal );
      shabal512_4way( &ctx.shabal, vhashB, 64 );
      shabal512_4way_close( &ctx.shabal, vhash );
 
-     mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -475,7 +449,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -489,7 +463,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -511,8 +485,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
-     mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
+     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -527,13 +501,13 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -548,13 +522,13 @@ void sonoa_4way_hash( void *state, const void *input )
      sph_fugue512( &ctx.fugue, hash3, 64 );
      sph_fugue512_close( &ctx.fugue, hash3 );
 
-     mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
 
      shabal512_4way_init( &ctx.shabal );
      shabal512_4way( &ctx.shabal, vhash, 64 );
      shabal512_4way_close( &ctx.shabal, vhash );
 
-     mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_whirlpool_init( &ctx.whirlpool );
      sph_whirlpool( &ctx.whirlpool, hash0, 64 );
@@ -571,13 +545,13 @@ void sonoa_4way_hash( void *state, const void *input )
 
 // 6
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
      
      bmw512_4way_init( &ctx.bmw );
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -588,7 +562,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -602,7 +576,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -624,8 +598,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
-     mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
+     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -640,13 +614,13 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -661,13 +635,13 @@ void sonoa_4way_hash( void *state, const void *input )
      sph_fugue512( &ctx.fugue, hash3, 64 );
      sph_fugue512_close( &ctx.fugue, hash3 );
 
-     mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
 
      shabal512_4way_init( &ctx.shabal );
      shabal512_4way( &ctx.shabal, vhash, 64 );
      shabal512_4way_close( &ctx.shabal, vhash );
 
-     mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_whirlpool_init( &ctx.whirlpool );
      sph_whirlpool( &ctx.whirlpool, hash0, 64 );
@@ -682,13 +656,13 @@ void sonoa_4way_hash( void *state, const void *input )
      sph_whirlpool( &ctx.whirlpool, hash3, 64 );
      sph_whirlpool_close( &ctx.whirlpool, hash3 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      sha512_4way_init( &ctx.sha512 );
      sha512_4way( &ctx.sha512, vhash, 64 );
      sha512_4way_close( &ctx.sha512, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_whirlpool_init( &ctx.whirlpool );
      sph_whirlpool( &ctx.whirlpool, hash0, 64 );
@@ -705,13 +679,13 @@ void sonoa_4way_hash( void *state, const void *input )
 
 // 7
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      bmw512_4way_init( &ctx.bmw );
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -722,7 +696,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -736,7 +710,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_reinterleave_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -758,8 +732,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_deinterleave_2x128( hash0, hash1, vhashA, 512 );
-     mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
+     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
+     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -774,13 +748,13 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -795,13 +769,13 @@ void sonoa_4way_hash( void *state, const void *input )
      sph_fugue512( &ctx.fugue, hash3, 64 );
      sph_fugue512_close( &ctx.fugue, hash3 );
 
-     mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
 
      shabal512_4way_init( &ctx.shabal );
      shabal512_4way( &ctx.shabal, vhash, 64 );
      shabal512_4way_close( &ctx.shabal, vhash );
 
-     mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
+     mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_whirlpool_init( &ctx.whirlpool );
      sph_whirlpool( &ctx.whirlpool, hash0, 64 );
@@ -816,13 +790,13 @@ void sonoa_4way_hash( void *state, const void *input )
      sph_whirlpool( &ctx.whirlpool, hash3, 64 );
      sph_whirlpool_close( &ctx.whirlpool, hash3 );
 
-     mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      sha512_4way_init( &ctx.sha512 );
      sha512_4way( &ctx.sha512, vhash, 64 );
      sha512_4way_close( &ctx.sha512, vhash );
 
-     mm256_reinterleave_4x64_4x32( vhashB, vhash,  512 );
+     mm256_rintrlv_4x64_4x32( vhashB, vhash,  512 );
 
      haval256_5_4way_init( &ctx.haval );
      haval256_5_4way( &ctx.haval, vhashB, 64 );
@@ -836,7 +810,6 @@ int scanhash_sonoa_4way( int thr_id, struct work *work, uint32_t max_nonce,
      uint32_t *hash7 = &(hash[7<<2]);
      uint32_t lane_hash[8];
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-     uint32_t endiandata[20] __attribute__((aligned(64)));
      uint32_t *pdata = work->data;
      uint32_t *ptarget = work->target;
      uint32_t n = pdata[19];
@@ -850,19 +823,13 @@ int scanhash_sonoa_4way( int thr_id, struct work *work, uint32_t max_nonce,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
      // Need big endian data
-     casti_m256i( endiandata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
-     casti_m256i( endiandata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
-     casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
-
-     uint64_t *edata = (uint64_t*)endiandata;
-     mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
-
+     mm256_bswap_intrlv80_4x64( vdata, pdata );
      for ( int m=0; m < 6; m++ ) if ( Htarg <= htmax[m] )
      {
         uint32_t mask = masks[m];
         do
         {
-           *noncev = mm256_interleave_blend_32( mm256_bswap_32(
+           *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
                              _mm256_set_epi32( n+3, 0,n+2, 0,n+1, 0, n, 0 ) ),
                                                 *noncev );
            sonoa_4way_hash( hash, vdata );
@@ -871,17 +838,10 @@ int scanhash_sonoa_4way( int thr_id, struct work *work, uint32_t max_nonce,
            if ( ( ( hash7[ lane ] & mask ) == 0 ) )
            {
               mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
-              if ( fulltest( lane_hash, ptarget ) )
+              if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
               {
                  pdata[19] = n + lane;
-                 work_set_target_ratio( work, lane_hash );
-                 if ( submit_work( mythr, work ) )
-                    applog( LOG_NOTICE,
-                             "Share %d submitted by thread %d, lane %d.",
-                             accepted_share_count + rejected_share_count + 1,
-                             thr_id, lane );
-                 else
-                    applog( LOG_WARNING, "Failed to submit share." );
+                 submit_solution( work, lane_hash, mythr, lane );
               }
            }
            n += 4;