v25.7

Makefile.am

@@ -1,27 +1,49 @@
 
+if HAVE_APPLE
+# MacOS uses Homebrew to install needed packages but they aren't linked for
+# the jansson test in configure. Ignore the failed test & link them now,
+# different path for different CPU arch.
+
+if ARCH_ARM64
+  EXTRA_INCLUDES = -I/opt/homebrew/include
+  EXTRA_LIBS     = -L/opt/homebrew/lib
+else
+  EXTRA_INCLUDES = -I/usr/local/include
+  EXTRA_LIBS     = -L/usr/local/lib
+endif
+
+else
+
 if WANT_JANSSON
-JANSSON_INCLUDES= -I$(top_srcdir)/compat/jansson
+# Can't find jansson libraries, compile the included source code.
+  EXTRA_INCLUDES = -I$(top_srcdir)/compat/jansson
+  EXTRA_LIBS     = -L$(top_srcdir)/compat/jansson
 else
-JANSSON_INCLUDES=
+  EXTRA_INCLUDES =
+  EXTRA_LIBS     =
+endif
+
 endif
 
 EXTRA_DIST = example-cfg.json nomacro.pl
 
 SUBDIRS = compat
 
-ALL_INCLUDES	= @PTHREAD_FLAGS@ -fno-strict-aliasing $(JANSSON_INCLUDES) -I.
+ALL_INCLUDES = @PTHREAD_FLAGS@ -fno-strict-aliasing $(EXTRA_INCLUDES) -I.
 
 bin_PROGRAMS = cpuminer
 
 dist_man_MANS = cpuminer.1
 
 cpuminer_SOURCES = \
+  dummy.cpp \
   cpu-miner.c \
   util.c \
   api.c \
   sysinfos.c \
   algo-gate-api.c\
   malloc-huge.c \
+  simd-utils/simd-constants.c \
   algo/argon2d/argon2d-gate.c \
   algo/argon2d/blake2/blake2b.c \
   algo/argon2d/argon2d/argon2.c \
@@ -113,7 +135,6 @@ cpuminer_SOURCES = \
   algo/lyra2/phi2-4way.c \
   algo/lyra2/phi2.c \
   algo/m7m/m7m.c \
-  algo/m7m/magimath.cpp \
   algo/nist5/nist5-gate.c \
   algo/nist5/nist5-4way.c \
   algo/nist5/nist5.c \
@@ -166,9 +187,6 @@ cpuminer_SOURCES = \
   algo/shavite/sph-shavite-aesni.c \
   algo/shavite/shavite-hash-2way.c \
   algo/shavite/shavite-hash-4way.c \
-  algo/shavite/shavite.c \
-  algo/simd/nist.c \
-  algo/simd/vector.c \
   algo/simd/sph_simd.c \
   algo/simd/simd-hash-2way.c \
   algo/skein/sph_skein.c \
@@ -276,28 +294,28 @@ cpuminer_SOURCES = \
   algo/yespower/yespower-ref.c \
   algo/yespower/yespower-blake2b-ref.c
 
-disable_flags =
-
-if USE_ASM
-   cpuminer_SOURCES += asm/neoscrypt_asm.S
-else
-   disable_flags += -DNOASM
-endif
-
 if HAVE_WINDOWS
    cpuminer_SOURCES += compat/winansi.c
 endif
 
+if USE_ASM
+   disable_flags =
+   cpuminer_SOURCES += asm/neoscrypt_asm.S
+else
+   disable_flags = -DNOASM
+endif
+
 cpuminer_LDFLAGS = @LDFLAGS@
-cpuminer_LDADD	= @LIBCURL@ @JANSSON_LIBS@ @PTHREAD_LIBS@ @WS2_LIBS@  -lgmp
+cpuminer_LDADD	= $(EXTRA_LIBS) @LIBCURL@ -ljansson @PTHREAD_LIBS@ @WS2_LIBS@ -lgmp
 cpuminer_CPPFLAGS = @LIBCURL_CPPFLAGS@ $(ALL_INCLUDES)
 cpuminer_CFLAGS = -Wno-pointer-sign -Wno-pointer-to-int-cast $(disable_flags)
 
-if HAVE_WINDOWS
+if ARCH_ARM64
-cpuminer_CFLAGS += -Wl,--stack,10485760
+   cpuminer_CFLAGS += -flax-vector-conversions
 endif
 
 if HAVE_WINDOWS
+
 # use to profile an object
 # gprof_cflags = -pg -g3
 # cpuminer_LDFLAGS += -pg
@@ -311,5 +329,4 @@ cpuminer-neoscrypt.o: neoscrypt.c
 	@echo "CUSTOM ${@}: ${filter %.o,${^}} ${filter %.c,${^}}"
 	$(CC) $(common_ccflags) -g -O3 $(gprof_cflags) -MT $@ -MD -MP -c -o $@ $<
 
-
 endif

README.md

@@ -36,44 +36,28 @@ for compile instructions.
 Requirements
 ------------
 
-1. A x86_64 architecture CPU with a minimum of SSE2 support. This includes
+1. A 64 bit CPU supporting x86_64 (Intel or AMD) or aarch64 (ARM).
-Intel Core2 and newer and AMD equivalents. Further optimizations are available
+x86_64 requires SSE2, aarch64 requires armv8 & NEON.
-on some algoritms for CPUs with AES, AVX, AVX2, SHA, AVX512 and VAES.
-
-32 bit CPUs are not supported.
-Other CPU architectures such as ARM, Raspberry Pi, RISC-V, Xeon Phi, etc,
-are not supported.
 
 Mobile CPUs like laptop computers are not recommended because they aren't
 designed for extreme heat of operating at full load for extended periods of
 time.
 
-Older CPUs and ARM architecture may be supported by cpuminer-multi by TPruvot.
+2. 64 bit operating system including Linux, Windows, MacOS, or BSD.
-
+Android, IOS and alt OSs like Haiku & ReactOS are not supported.
-2. 64 bit Linux or Windows OS. Ubuntu and Fedora based distributions,
-including Mint and Centos, are known to work and have all dependencies
-in their repositories. Others may work but may require more effort. Older
-versions such as Centos 6 don't work due to missing features. 
-
-Windows 7 or newer is supported with mingw_w64 and msys or using the pre-built
-binaries. WindowsXP 64 bit is YMMV.
-
-FreeBSD is not actively tested but should work, YMMV.
-MacOS, OSx and Android are not supported.
 
 3. Stratum pool supporting stratum+tcp:// or stratum+ssl:// protocols or
-RPC getwork using http:// or https://.
+RPC getblocktemplate using http:// or https://.
-GBT is YMMV.
 
 Supported Algorithms
 --------------------
 
                           allium        Garlicoin
                           anime         Animecoin
-                          argon2        Argon2 coin (AR2)
+                          argon2d250    
-                          argon2d250    argon2d-crds, Credits (CRDS)
+                          argon2d500
-                          argon2d500    argon2d-dyn,  Dynamic (DYN)
+                          argon2d1000
-                          argon2d4096   argon2d-uis, Unitus, (UIS)
+                          argon2d4096
                           blake         Blake-256
                           blake2b       Blake2-512
                           blake2s       Blake2-256

RELEASE_NOTES

@@ -75,6 +75,110 @@ If not what makes it happen or not happen?
 Change Log
 ----------
 
+v25.7
+
+Fixed a bug calculating TTF longer than 1 year.
+Faster argon2d.
+Faster hamsi AVX512.
+Faster switfftx AVX2.
+Other small fixes and improvements.
+
+v25.6
+
+Added argon2d1000, argon2d16000 algos.
+Target specific AES optimizations improve shavite for ARM64 & x86_64.
+
+v25.5
+
+x86_64: Fixed an insidious bug in sha256 early rejection optimization for AVX2 & AVX512.
+x86_64: Faster sha256d, sha256dt for AVX2 & AVX512.
+Other small bug fixes.
+
+v25.4
+
+x86_64: improved handling of vector constants used for byte permutations.
+x86_64: removed hooks for cancelled AVX10-256.
+Minor bug fixes & improvements.
+More code cleanup.
+
+v25.3
+
+#442, #443: Fixed a regression in Makefile.am.
+Removed algo features log display.
+Some code cleanup.
+
+v25.2
+
+ARM: Fixed regression from v25.1 that could cause build fail.
+BSD: FreeBSD is now supported. Other BSDs may also work.
+MacOS: build with installed jansson library instead of compiling the included source code.
+Windows: remove "_WIN32_WINNT=0x0601" which was a downgrade on Win11.
+Changed build.sh shell from bash to sh.
+
+v25.1
+
+MacOS ARM64: m7m algo is now working.
+MacOS ARM64: can now be compiled with GCC.
+MacOS x86_64: is now working compiled with GCC.
+Fixed some minor bugs & removed some obsolete code.
+
+v24.8
+
+ARM: Apple MacOS on M series CPU is now supported compiled from source
+code, see Wiki for details.
+ARM: Fix incorrect compiler version display when using clang. 
+build.sh can now be used to compile all targets, arm_build.sh & build_msys2.sh
+have been removed.
+Windows: MSys2 build now enables CPU groups by default, prebuilt binaries
+continue to be compiled with CPU groups disabled.
+
+v24.7
+
+ARM: compile works for Windows using MSys2 & MingW, see wiki for details.
+
+v24.6
+
+ARM: Fixed scryptn2, x16*, broken in v24.2. 
+ARM: Small improvement to interleaving.
+Eliminated some potential compile errors in code that was dependent on 
+compiler optimisations.
+x86_64: improved support for AVX10 compilation, needs GCC-14 or higher.
+
+v24.5
+
+Fix MinGW compile error after MSys2 upgrade to GCC-14.2. 
+#427: GBT: Improved handling of new work.
+Removed shavite3 algo.
+
+v24.4
+
+x86_64: fixed a bug in ornot macro for AVX2 which broke some algos in v24.2.
+x86_64: fixed a bug in alignr macros for SSE2.
+ARM: CPU feature reporting enhancements.
+Some code cleanup.
+
+v24.3
+
+ARM: CPU feature detection and reporting is now working.
+ARM: Verthash is now working.
+ARM: Small speedup for yescrypt, yespower & argon2d.
+Code cleanup.
+
+v24.2
+
+x86_64: Fixed blakes2s for AVX2 & AVX512, x25x for AVX512, broken in v3.23.4.
+x86_64: Initial support for CPUs with AVX10, needs GCC-14.
+ARM NEON: Various code optimisations.
+
+v24.1
+
+#414: fix bug in merkle error handling.
+#416: change $nproc to $(nproc) in build scripts.
+#420: change some inline function definitions to static inline. 
+#413: Fix formatting error for share result log when using no-color.
+Faster 2 way interleaving.
+Cleanup sha256 architecture targetting.
+
 v23.15
 
 Fixed x11gost (sib) algo for all architectures, broken in v3.23.4.

algo-gate-api.c

@@ -184,7 +184,7 @@ int scanhash_4way_64in_32out( struct work *work, uint32_t max_nonce,
 
 #endif
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 //int scanhash_8way_64_64( struct work *work, uint32_t max_nonce,
 //                      uint64_t *hashes_done, struct thr_info *mythr )
@@ -263,8 +263,8 @@ static void init_algo_gate( algo_gate_t* gate )
    gate->build_block_header      = (void*)&std_build_block_header;
    gate->build_extraheader       = (void*)&std_build_extraheader;
    gate->set_work_data_endian    = (void*)&do_nothing;
-   gate->resync_threads          = (void*)&do_nothing;
+//   gate->resync_threads          = (void*)&do_nothing;
-   gate->do_this_thread          = (void*)&return_true;
+//   gate->do_this_thread          = (void*)&return_true;
    gate->longpoll_rpc_call       = (void*)&std_longpoll_rpc_call;
    gate->get_work_data_size      = (void*)&std_get_work_data_size;
    gate->optimizations           = EMPTY_SET;
@@ -295,8 +295,10 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
   {
     case ALGO_ALLIUM:       rc = register_allium_algo        ( gate ); break;
     case ALGO_ANIME:        rc = register_anime_algo         ( gate ); break;
-    case ALGO_ARGON2D250:   rc = register_argon2d_crds_algo  ( gate ); break;
+    case ALGO_ARGON2D250:   rc = register_argon2d250_algo    ( gate ); break;
-    case ALGO_ARGON2D500:   rc = register_argon2d_dyn_algo   ( gate ); break;
+    case ALGO_ARGON2D500:   rc = register_argon2d500_algo    ( gate ); break;
+    case ALGO_ARGON2D1000:  rc = register_argon2d1000_algo   ( gate ); break;
+    case ALGO_ARGON2D16000: rc = register_argon2d16000_algo  ( gate ); break;
     case ALGO_ARGON2D4096:  rc = register_argon2d4096_algo   ( gate ); break;
     case ALGO_AXIOM:        rc = register_axiom_algo         ( gate ); break;
     case ALGO_BLAKE:        rc = register_blake_algo         ( gate ); break;
@@ -340,7 +342,6 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
     case ALGO_SHA256T:      rc = register_sha256t_algo       ( gate ); break;
     case ALGO_SHA3D:        rc = register_sha3d_algo         ( gate ); break;
     case ALGO_SHA512256D:   rc = register_sha512256d_algo    ( gate ); break;
-    case ALGO_SHAVITE3:     rc = register_shavite_algo       ( gate ); break;
     case ALGO_SKEIN:        rc = register_skein_algo         ( gate ); break;
     case ALGO_SKEIN2:       rc = register_skein2_algo        ( gate ); break;
     case ALGO_SKUNK:        rc = register_skunk_algo         ( gate ); break;
@@ -417,8 +418,6 @@ void exec_hash_function( int algo, void *output, const void *pdata )
 const char* const algo_alias_map[][2] =
 {
 //   alias                proper
-  { "argon2d-dyn",       "argon2d500"     },
-  { "argon2d-uis",       "argon2d4096"    },
   { "bcd",               "x13bcd"         },
   { "bitcore",           "timetravel10"   },
   { "bitzeny",           "yescryptr8"     },

algo-gate-api.h

@@ -98,25 +98,27 @@ typedef  uint32_t set_t;
 #define AVX512_OPT       1 <<  6   // Skylake-X, Zen4 (AVX512[F,VL,DQ,BW])
 #define AES_OPT          1 <<  7   // Intel Westmere, AArch64
 #define VAES_OPT         1 <<  8   // Icelake, Zen3
-#define SHA_OPT          1 <<  9   // Zen1, Icelake, AArch64 
+#define SHA256_OPT       1 <<  9   // Zen1, Icelake, AArch64 
 #define SHA512_OPT       1 << 10   // Intel Arrow Lake, AArch64 
 #define NEON_OPT         1 << 11   // AArch64 
+#define AVX10_256        1 << 12
+#define AVX10_512        1 << 13
 
 // AVX10 does not have explicit algo features:
 //  AVX10_512 is compatible with AVX512 + VAES
 //  AVX10_256 is compatible with AVX2 + VAES
 
 // return set containing all elements from sets a & b
-inline set_t set_union ( set_t a, set_t b ) { return a | b; }
+static inline set_t set_union ( set_t a, set_t b ) { return a | b; }
 
 // return set contained common elements from sets a & b
-inline set_t set_intsec ( set_t a, set_t b) { return a & b; }
+static inline set_t set_intsec ( set_t a, set_t b) { return a & b; }
 
 // all elements in set a are included in set b
-inline bool set_incl ( set_t a, set_t b ) { return (a & b) == a; }
+static inline bool set_incl ( set_t a, set_t b ) { return (a & b) == a; }
 
 // no elements in set a are included in set b
-inline bool set_excl ( set_t a, set_t b ) { return (a & b) == 0; }
+static inline bool set_excl ( set_t a, set_t b ) { return (a & b) == 0; }
 
 typedef struct
 {
@@ -163,15 +165,18 @@ char* ( *malloc_txs_request )   ( struct work* );
 void ( *set_work_data_endian )  ( struct work* );
 
 // Diverge mining threads
-bool ( *do_this_thread )        ( int );
+//bool ( *do_this_thread )        ( int );
 
 // After do_this_thread
-void ( *resync_threads )        ( int, struct work* );
+//void ( *resync_threads )        ( int, struct work* );
 
 json_t* ( *longpoll_rpc_call )  ( CURL*, int*, char* );
 
+// Deprecated
 set_t optimizations;
+
 int  ( *get_work_data_size )     ();
+
 int  ntime_index;
 int  nbits_index;
 int  nonce_index;            // use with caution, see warning below
@@ -246,7 +251,7 @@ int scanhash_4way_64in_32out( struct work *work, uint32_t max_nonce,
 
 #endif
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 //int scanhash_8way_64in_64out( struct work *work, uint32_t max_nonce,
 //                      uint64_t *hashes_done, struct thr_info *mythr );
@@ -272,8 +277,6 @@ void std_get_new_work( struct work *work, struct work *g_work, int thr_id,
 
 void sha256d_gen_merkle_root( char *merkle_root, struct stratum_ctx *sctx );
 void sha256_gen_merkle_root ( char *merkle_root, struct stratum_ctx *sctx );
-// OpenSSL sha256 deprecated
-//void SHA256_gen_merkle_root ( char *merkle_root, struct stratum_ctx *sctx );
 
 bool std_le_work_decode( struct work *work );
 bool std_be_work_decode( struct work *work );

algo/argon2d/argon2d-gate.c

@@ -6,9 +6,39 @@ static const size_t INPUT_BYTES = 80;  // Lenth of a block header in bytes. Inpu
 static const size_t OUTPUT_BYTES = 32; // Length of output needed for a 256-bit hash
 static const unsigned int DEFAULT_ARGON2_FLAG = 2; //Same as ARGON2_DEFAULT_FLAGS
 
-// Credits
+// generic, works with most variations of argon2d
+int scanhash_argon2d( struct work *work, uint32_t max_nonce,
+                      uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t _ALIGN(64) edata[20];
+   uint32_t _ALIGN(64) hash[8];
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   const int thr_id = mythr->id;
+   const uint32_t first_nonce = (const uint32_t)pdata[19];
+   const uint32_t last_nonce = (const uint32_t)max_nonce;
+   uint32_t nonce = first_nonce;
+   const bool bench = opt_benchmark;
 
-void argon2d_crds_hash( void *output, const void *input )
+   v128_bswap32_80( edata, pdata );
+   do
+   {
+      edata[19] = nonce;
+      algo_gate.hash( hash, edata, thr_id );
+      if ( unlikely( valid_hash( hash, ptarget ) && !bench ) )
+      {
+          pdata[19] = bswap_32( nonce );
+          submit_solution( work, hash, mythr );
+      }
+      nonce++;
+  } while ( likely( nonce < last_nonce && !work_restart[thr_id].restart ) );
+
+   pdata[19] = nonce;
+   *hashes_done = pdata[19] - first_nonce;
+   return 0;
+}
+
+void argon2d250_hash( void *output, const void *input )
 {
 	argon2_context context;
 	context.out = (uint8_t *)output;
@@ -34,48 +64,15 @@ void argon2d_crds_hash( void *output, const void *input )
 	argon2_ctx( &context, Argon2_d );
 }
 
-int scanhash_argon2d_crds( struct work *work, uint32_t max_nonce,
+bool register_argon2d250_algo( algo_gate_t* gate )
-                      uint64_t *hashes_done, struct thr_info *mythr )
 {
-   uint32_t _ALIGN(64) edata[20];
+        gate->scanhash = (void*)&scanhash_argon2d;
-   uint32_t _ALIGN(64) hash[8];
+        gate->hash = (void*)&argon2d250_hash;
-   uint32_t *pdata = work->data;
-   uint32_t *ptarget = work->target;
-   int thr_id = mythr->id;  // thr_id arg is deprecated
-   const uint32_t first_nonce = pdata[19];
-   const uint32_t Htarg = ptarget[7];
-   uint32_t nonce = first_nonce;
-
-   swab32_array( edata, pdata, 20 );
-
-   do {
-      be32enc(&edata[19], nonce);
-      argon2d_crds_hash( hash, edata );
-      if ( hash[7] <= Htarg && fulltest( hash, ptarget ) && !opt_benchmark )
-      {
-          pdata[19] = nonce;
-          submit_solution( work, hash, mythr );
-      }
-      nonce++;
-   } while (nonce < max_nonce && !work_restart[thr_id].restart);
-
-   pdata[19] = nonce;
-   *hashes_done = pdata[19] - first_nonce + 1;
-   return 0;
-}
-
-bool register_argon2d_crds_algo( algo_gate_t* gate )
-{
-        gate->scanhash = (void*)&scanhash_argon2d_crds;
-        gate->hash = (void*)&argon2d_crds_hash;
-        gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT | NEON_OPT;
         opt_target_factor = 65536.0;
         return true;
 }
 
-// Dynamic
+void argon2d500_hash( void *output, const void *input )
-
-void argon2d_dyn_hash( void *output, const void *input )
 {
     argon2_context context;
     context.out = (uint8_t *)output;
@@ -101,48 +98,81 @@ void argon2d_dyn_hash( void *output, const void *input )
     argon2_ctx( &context, Argon2_d );
 }
 
-int scanhash_argon2d_dyn( struct work *work, uint32_t max_nonce,
+bool register_argon2d500_algo( algo_gate_t* gate )
-                      uint64_t *hashes_done, struct thr_info *mythr )
 {
-   uint32_t _ALIGN(64) edata[20];
+        gate->scanhash = (void*)&scanhash_argon2d;
-   uint32_t _ALIGN(64) hash[8];
+        gate->hash = (void*)&argon2d500_hash;
-   uint32_t *pdata = work->data;
-   uint32_t *ptarget = work->target;
-   const int thr_id = mythr->id; 
-   const uint32_t first_nonce = (const uint32_t)pdata[19];
-   const uint32_t last_nonce = (const uint32_t)max_nonce;
-   uint32_t nonce = first_nonce;
-   const bool bench = opt_benchmark;
-
-   v128_bswap32_80( edata, pdata );
-   do
-   {
-      edata[19] = nonce;
-      argon2d_dyn_hash( hash, edata );
-      if ( unlikely( valid_hash( (uint64_t*)hash, (uint64_t*)ptarget )
-           && !bench ) )
-      {
-          pdata[19] = bswap_32( nonce );;
-          submit_solution( work, hash, mythr );
-      }
-      nonce++;
-  } while ( likely( nonce < last_nonce && !work_restart[thr_id].restart ) );
-
-   pdata[19] = nonce;
-   *hashes_done = pdata[19] - first_nonce;
-   return 0;
-}
-
-bool register_argon2d_dyn_algo( algo_gate_t* gate )
-{
-        gate->scanhash = (void*)&scanhash_argon2d_dyn;
-        gate->hash = (void*)&argon2d_dyn_hash;
-        gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT | NEON_OPT;
         opt_target_factor = 65536.0;
         return true;
 }
 
-// Unitus
+void argon2d1000_hash( void *output, const void *input )
+{
+    argon2_context context;
+    context.out = (uint8_t *)output;
+    context.outlen = (uint32_t)OUTPUT_BYTES;
+    context.pwd = (uint8_t *)input;
+    context.pwdlen = (uint32_t)INPUT_BYTES;
+    context.salt = (uint8_t *)input; //salt = input
+    context.saltlen = (uint32_t)INPUT_BYTES;
+    context.secret = NULL;
+    context.secretlen = 0;
+    context.ad = NULL;
+    context.adlen = 0;
+    context.allocate_cbk = NULL;
+    context.free_cbk = NULL;
+    context.flags = DEFAULT_ARGON2_FLAG; // = ARGON2_DEFAULT_FLAGS
+    // main configurable Argon2 hash parameters
+    context.m_cost = 1000;  // Memory in KiB (1MB)
+    context.lanes = 8;     // Degree of Parallelism
+    context.threads = 1;   // Threads
+    context.t_cost = 2;    // Iterations
+    context.version = ARGON2_VERSION_10;
+
+    argon2_ctx( &context, Argon2_d );
+}
+
+bool register_argon2d1000_algo( algo_gate_t* gate )
+{
+        gate->scanhash = (void*)&scanhash_argon2d;
+        gate->hash = (void*)&argon2d1000_hash;
+        opt_target_factor = 65536.0;
+        return true;
+}
+
+void argon2d16000_hash( void *output, const void *input )
+{
+   argon2_context context;
+   context.out = (uint8_t *)output;
+   context.outlen = (uint32_t)OUTPUT_BYTES;
+   context.pwd = (uint8_t *)input;
+   context.pwdlen = (uint32_t)INPUT_BYTES;
+   context.salt = (uint8_t *)input; //salt = input
+   context.saltlen = (uint32_t)INPUT_BYTES;
+   context.secret = NULL;
+   context.secretlen = 0;
+   context.ad = NULL;
+   context.adlen = 0;
+   context.allocate_cbk = NULL;
+   context.free_cbk = NULL;
+   context.flags = DEFAULT_ARGON2_FLAG; // = ARGON2_DEFAULT_FLAGS
+   // main configurable Argon2 hash parameters
+   context.m_cost = 16000; // Memory in KiB (~16384KB)
+   context.lanes = 1;    // Degree of Parallelism
+   context.threads = 1;  // Threads
+   context.t_cost = 1;   // Iterations
+   context.version = ARGON2_VERSION_10;
+
+   argon2_ctx( &context, Argon2_d );
+}
+
+bool register_argon2d16000_algo( algo_gate_t* gate )
+{
+        gate->scanhash = (void*)&scanhash_argon2d;
+        gate->hash = (void*)&argon2d16000_hash;
+        opt_target_factor = 65536.0;
+        return true;
+}
 
 int scanhash_argon2d4096( struct work *work, uint32_t max_nonce,
                            uint64_t *hashes_done, struct thr_info *mythr )
@@ -154,7 +184,7 @@ int scanhash_argon2d4096( struct work *work, uint32_t max_nonce,
    const uint32_t first_nonce = pdata[19];
    const uint32_t last_nonce = (const uint32_t)max_nonce;
    uint32_t n = first_nonce;
-   const int thr_id = mythr->id;  // thr_id arg is deprecated
+   const int thr_id = mythr->id;  
    uint32_t t_cost = 1; // 1 iteration
    uint32_t m_cost = 4096; // use 4MB
    uint32_t parallelism = 1; // 1 thread, 2 lanes
@@ -182,7 +212,6 @@ int scanhash_argon2d4096( struct work *work, uint32_t max_nonce,
 bool register_argon2d4096_algo( algo_gate_t* gate )
 {
         gate->scanhash = (void*)&scanhash_argon2d4096;
-        gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT |NEON_OPT;
         opt_target_factor = 65536.0;
         return true;
 }

algo/argon2d/argon2d-gate.h

@@ -4,22 +4,27 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-// Credits: version = 0x10, m_cost = 250.
+int scanhash_argon2d( struct work *work, uint32_t max_nonce,
-bool register_argon2d_crds_algo( algo_gate_t* gate );
-
-void argon2d_crds_hash( void *state, const void *input );
-
-int scanhash_argon2d_crds( struct work *work, uint32_t max_nonce,
                     uint64_t *hashes_done, struct thr_info *mythr );
 
+// Credits: version = 0x10, m_cost = 250.
+bool register_argon2d250_algo( algo_gate_t* gate );
+
+void argon2d250_hash( void *state, const void *input );
+
 // Dynamic: version = 0x10, m_cost = 500.
-bool register_argon2d_dyn_algo( algo_gate_t* gate );
+bool register_argon2d500_algo( algo_gate_t* gate );
 
-void argon2d_dyn_hash( void *state, const void *input );
+void argon2d500_hash( void *state, const void *input );
 
-int scanhash_argon2d_dyn( struct work *work, uint32_t max_nonce,
+// Zero Dynamics Cash: version = 0x10, m_cost = 1000.
-                    uint64_t *hashes_done, struct thr_info *mythr );
+bool register_argon2d1000_algo( algo_gate_t* gate );
 
+void argon2d1000_hash( void *state, const void *input );
+
+bool register_argon2d16000_algo( algo_gate_t* gate );
+
+void argon2d16000_hash( void *state, const void *input );
 
 // Unitus: version = 0x13, m_cost = 4096.
 bool register_argon2d4096_algo( algo_gate_t* gate );

algo/argon2d/argon2d/opt.c

@@ -35,7 +35,7 @@
  * @pre all block pointers must be valid
  */
 
-#if defined(__AVX512F__)
+#if defined(SIMD512)
 
 static inline __m512i blamka( __m512i x, __m512i y )
 {
@@ -237,7 +237,7 @@ void fill_segment(const argon2_instance_t *instance,
     uint64_t pseudo_rand, ref_index, ref_lane;
     uint32_t prev_offset, curr_offset;
     uint32_t starting_index, i;
-#if defined(__AVX512F__)
+#if defined(SIMD512)
     __m512i state[ARGON2_512BIT_WORDS_IN_BLOCK];
 #elif defined(__AVX2__)
     __m256i state[ARGON2_HWORDS_IN_BLOCK];

algo/argon2d/blake2/blamka-round-opt.h

@@ -21,7 +21,7 @@
 #include "blake2-impl.h"
 #include "simd-utils.h"
 
-#if !defined(__AVX512F__)
+#if !defined(SIMD512)
 
 #if !defined(__AVX2__)
 
@@ -67,81 +67,59 @@ static BLAKE2_INLINE v128_t fBlaMka(v128_t x, v128_t y)
 #if defined(__SSSE3__)  || defined(__ARM_NEON)
 
 #define DIAGONALIZE( A0, B0, C0, D0, A1, B1, C1, D1 ) \
-    do {                                                                       \
+{ \
-        v128_t t0 = v128_alignr8(B1, B0, 8);                               \
+   v128_t t = v128_alignr8( B1, B0, 8 ); \
-        v128_t t1 = v128_alignr8(B0, B1, 8);                               \
+   B1 = v128_alignr8( B0, B1, 8 ); \
-        B0 = t0;                                                               \
+   B0 = t; \
-        B1 = t1;                                                               \
+   t = v128_alignr8( D1, D0, 8 ); \
-                                                                               \
+   D0 = v128_alignr8( D0, D1, 8 ); \
-        t0 = C0;                                                               \
+   D1 = t; \
-        C0 = C1;                                                               \
+}
-        C1 = t0;                                                               \
-                                                                               \
-        t0 = v128_alignr8(D1, D0, 8);                                       \
-        t1 = v128_alignr8(D0, D1, 8);                                       \
-        D0 = t1;                                                               \
-        D1 = t0;                                                               \
-    } while ((void)0, 0)
 
 #define UNDIAGONALIZE( A0, B0, C0, D0, A1, B1, C1, D1 ) \
-    do {                                                                       \
+{ \
-        v128_t t0 = v128_alignr8(B0, B1, 8);                               \
+    v128_t t = v128_alignr8( B0, B1, 8 ); \
-        v128_t t1 = v128_alignr8(B1, B0, 8);                               \
+    B1 = v128_alignr8( B1, B0, 8 ); \
-        B0 = t0;                                                               \
+    B0 = t; \
-        B1 = t1;                                                               \
+    t = v128_alignr8( D0, D1, 8 ); \
-                                                                               \
+    D0 = v128_alignr8( D1, D0, 8 ); \
-        t0 = C0;                                                               \
+    D1 = t; \
-        C0 = C1;                                                               \
+}
-        C1 = t0;                                                               \
-                                                                               \
-        t0 = v128_alignr8(D0, D1, 8);                                       \
-        t1 = v128_alignr8(D1, D0, 8);                                       \
-        D0 = t1;                                                               \
-        D1 = t0;                                                               \
-    } while ((void)0, 0)
 
 #else /* SSE2 */
 
 #define DIAGONALIZE( A0, B0, C0, D0, A1, B1, C1, D1 ) \
-    do {                                                                       \
+{ \
-        v128_t t0 = D0;                                                       \
+    v128_t t = D0; \
-        v128_t t1 = B0;                                                       \
+    D0 = v128_unpackhi64( D1, v128_unpacklo64( D0, D0 ) ); \
-        D0 = C0;                                                               \
+    D1 = v128_unpackhi64( t, v128_unpacklo64( D1, D1 ) ); \
-        C0 = C1;                                                               \
+    t = B0; \
-        C1 = D0;                                                               \
-        D0 = v128_unpackhi64(D1, v128_unpacklo64(t0, t0));               \
-        D1 = v128_unpackhi64(t0, v128_unpacklo64(D1, D1));               \
     B0 = v128_unpackhi64( B0, v128_unpacklo64( B1, B1 ) ); \
-        B1 = v128_unpackhi64(B1, v128_unpacklo64(t1, t1));               \
+    B1 = v128_unpackhi64( B1, v128_unpacklo64( t, t ) ); \
-    } while ((void)0, 0)
+}
 
 #define UNDIAGONALIZE( A0, B0, C0, D0, A1, B1, C1, D1 ) \
-    do {                                                                       \
+{ \
-        v128_t t0, t1;                                                        \
+    v128_t t = B0; \
-        t0 = C0;                                                               \
-        C0 = C1;                                                               \
-        C1 = t0;                                                               \
-        t0 = B0;                                                               \
-        t1 = D0;                                                               \
     B0 = v128_unpackhi64( B1, v128_unpacklo64( B0, B0 ) ); \
-        B1 = v128_unpackhi64(t0, v128_unpacklo64(B1, B1));               \
+    B1 = v128_unpackhi64( t, v128_unpacklo64( B1, B1 ) ); \
+    t = D0; \
     D0 = v128_unpackhi64( D0, v128_unpacklo64( D1, D1 ) ); \
-        D1 = v128_unpackhi64(D1, v128_unpacklo64(t1, t1));               \
+    D1 = v128_unpackhi64( D1, v128_unpacklo64( t, t ) ); \
-    } while ((void)0, 0)
+}
+
 #endif
 
 #define BLAKE2_ROUND( A0, A1, B0, B1, C0, C1, D0, D1 ) \
-    do {                                                                       \
+{ \
     G1( A0, B0, C0, D0, A1, B1, C1, D1 ); \
     G2( A0, B0, C0, D0, A1, B1, C1, D1 ); \
-                                                                               \
     DIAGONALIZE( A0, B0, C0, D0, A1, B1, C1, D1 ); \
-                                                                               \
+    G1( A0, B0, C1, D0, A1, B1, C0, D1 ); \
-        G1(A0, B0, C0, D0, A1, B1, C1, D1);                                    \
+    G2( A0, B0, C1, D0, A1, B1, C0, D1 ); \
-        G2(A0, B0, C0, D0, A1, B1, C1, D1);                                    \
-                                                                               \
     UNDIAGONALIZE( A0, B0, C0, D0, A1, B1, C1, D1 ); \
-    } while ((void)0, 0)
+}
+
 #else /* __AVX2__ */
 
 #include <immintrin.h>
@@ -211,7 +189,6 @@ static BLAKE2_INLINE v128_t fBlaMka(v128_t x, v128_t y)
         B0 = _mm256_permute4x64_epi64(B0, _MM_SHUFFLE(0, 3, 2, 1)); \
         C0 = _mm256_permute4x64_epi64(C0, _MM_SHUFFLE(1, 0, 3, 2)); \
         D0 = _mm256_permute4x64_epi64(D0, _MM_SHUFFLE(2, 1, 0, 3)); \
-        \
         B1 = _mm256_permute4x64_epi64(B1, _MM_SHUFFLE(0, 3, 2, 1)); \
         C1 = _mm256_permute4x64_epi64(C1, _MM_SHUFFLE(1, 0, 3, 2)); \
         D1 = _mm256_permute4x64_epi64(D1, _MM_SHUFFLE(2, 1, 0, 3)); \
@@ -219,17 +196,14 @@ static BLAKE2_INLINE v128_t fBlaMka(v128_t x, v128_t y)
 
 #define DIAGONALIZE_2(A0, A1, B0, B1, C0, C1, D0, D1) \
     do { \
-        __m256i tmp1 = _mm256_blend_epi32(B0, B1, 0xCC); \
+        __m256i tmp1 = _mm256_blend_epi32(B0, B1, 0x33); \
-        __m256i tmp2 = _mm256_blend_epi32(B0, B1, 0x33); \
+        __m256i tmp2 = _mm256_blend_epi32(B0, B1, 0xCC); \
-        B1 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
+        B0 = _mm256_shuffle_epi32( tmp1, 0x4e ); \
-        tmp1 = C0; \
+        B1 = _mm256_shuffle_epi32( tmp2, 0x4e ); \
-        B0 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
-        C0 = C1; \
-        tmp2 = _mm256_blend_epi32(D0, D1, 0x33); \
-        C1 = tmp1; \
         tmp1 = _mm256_blend_epi32(D0, D1, 0xCC); \
-        D1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
+        tmp2 = _mm256_blend_epi32(D0, D1, 0x33); \
-        D0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
+        D0 = _mm256_shuffle_epi32( tmp1, 0x4e ); \
+        D1 = _mm256_shuffle_epi32( tmp2, 0x4e ); \
     } while(0);
 
 #define UNDIAGONALIZE_1(A0, B0, C0, D0, A1, B1, C1, D1) \
@@ -237,7 +211,6 @@ static BLAKE2_INLINE v128_t fBlaMka(v128_t x, v128_t y)
         B0 = _mm256_permute4x64_epi64(B0, _MM_SHUFFLE(2, 1, 0, 3)); \
         C0 = _mm256_permute4x64_epi64(C0, _MM_SHUFFLE(1, 0, 3, 2)); \
         D0 = _mm256_permute4x64_epi64(D0, _MM_SHUFFLE(0, 3, 2, 1)); \
-        \
         B1 = _mm256_permute4x64_epi64(B1, _MM_SHUFFLE(2, 1, 0, 3)); \
         C1 = _mm256_permute4x64_epi64(C1, _MM_SHUFFLE(1, 0, 3, 2)); \
         D1 = _mm256_permute4x64_epi64(D1, _MM_SHUFFLE(0, 3, 2, 1)); \
@@ -247,27 +220,21 @@ static BLAKE2_INLINE v128_t fBlaMka(v128_t x, v128_t y)
     do { \
         __m256i tmp1 = _mm256_blend_epi32(B0, B1, 0xCC); \
         __m256i tmp2 = _mm256_blend_epi32(B0, B1, 0x33); \
-        B0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
+        B0 = _mm256_shuffle_epi32( tmp1, 0x4e ); \
-        tmp1 = C0; \
+        B1 = _mm256_shuffle_epi32( tmp2, 0x4e ); \
-        B1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
-        C0 = C1; \
         tmp2 = _mm256_blend_epi32(D0, D1, 0xCC); \
-        C1 = tmp1; \
         tmp1 = _mm256_blend_epi32(D0, D1, 0x33); \
-        D1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
+        D1 = _mm256_shuffle_epi32( tmp2, 0x4e ); \
-        D0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
+        D0 = _mm256_shuffle_epi32( tmp1, 0x4e ); \
     } while((void)0, 0);
 
 #define BLAKE2_ROUND_1(A0, A1, B0, B1, C0, C1, D0, D1) \
     do{ \
         G1_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
         G2_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
-        \
         DIAGONALIZE_1(A0, B0, C0, D0, A1, B1, C1, D1) \
-        \
         G1_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
         G2_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
-        \
         UNDIAGONALIZE_1(A0, B0, C0, D0, A1, B1, C1, D1) \
     } while((void)0, 0);
 
@@ -275,12 +242,9 @@ static BLAKE2_INLINE v128_t fBlaMka(v128_t x, v128_t y)
     do{ \
         G1_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
         G2_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
-        \
         DIAGONALIZE_2(A0, A1, B0, B1, C0, C1, D0, D1) \
-        \
+        G1_AVX2(A0, A1, B0, B1, C1, C0, D0, D1) \
-        G1_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
+        G2_AVX2(A0, A1, B0, B1, C1, C0, D0, D1) \
-        G2_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
-        \
         UNDIAGONALIZE_2(A0, A1, B0, B1, C0, C1, D0, D1) \
     } while((void)0, 0);
 
@@ -290,12 +254,73 @@ static BLAKE2_INLINE v128_t fBlaMka(v128_t x, v128_t y)
 
 #include <immintrin.h>
 
+/*
 static inline __m512i muladd(__m512i x, __m512i y)
 {
     __m512i z = _mm512_mul_epu32(x, y);
     return _mm512_add_epi64(_mm512_add_epi64(x, y), _mm512_add_epi64(z, z));
 }
+*/
 
+#define G1( A0, B0, C0, D0, A1, B1, C1, D1 ) \
+{ \
+  __m512i z0, z1; \
+  z0 = _mm512_mul_epu32( A0, B0 ); \
+  z1 = _mm512_mul_epu32( A1, B1 ); \
+  A0 = _mm512_add_epi64( A0, B0 ); \
+  A1 = _mm512_add_epi64( A1, B1 ); \
+  z0 = _mm512_add_epi64( z0, z0 ); \
+  z1 = _mm512_add_epi64( z1, z1 ); \
+  A0 = _mm512_add_epi64( A0, z0 ); \
+  A1 = _mm512_add_epi64( A1, z1 ); \
+  D0 = _mm512_xor_si512(D0, A0); \
+  D1 = _mm512_xor_si512(D1, A1); \
+  D0 = _mm512_ror_epi64(D0, 32); \
+  D1 = _mm512_ror_epi64(D1, 32); \
+  z0 = _mm512_mul_epu32( C0, D0 ); \
+  z1 = _mm512_mul_epu32( C1, D1 ); \
+  C0 = _mm512_add_epi64( C0, D0 ); \
+  C1 = _mm512_add_epi64( C1, D1 ); \
+  z0 = _mm512_add_epi64( z0, z0 ); \
+  z1 = _mm512_add_epi64( z1, z1 ); \
+  C0 = _mm512_add_epi64( C0, z0 ); \
+  C1 = _mm512_add_epi64( C1, z1 ); \
+  B0 = _mm512_xor_si512(B0, C0); \
+  B1 = _mm512_xor_si512(B1, C1); \
+  B0 = _mm512_ror_epi64(B0, 24); \
+  B1 = _mm512_ror_epi64(B1, 24); \
+}
+
+#define G2( A0, B0, C0, D0, A1, B1, C1, D1 ) \
+{ \
+  __m512i z0, z1; \
+  z0 = _mm512_mul_epu32( A0, B0 ); \
+  z1 = _mm512_mul_epu32( A1, B1 ); \
+  A0 = _mm512_add_epi64( A0, B0 ); \
+  A1 = _mm512_add_epi64( A1, B1 ); \
+  z0 = _mm512_add_epi64( z0, z0 ); \
+  z1 = _mm512_add_epi64( z1, z1 ); \
+  A0 = _mm512_add_epi64( A0, z0 ); \
+  A1 = _mm512_add_epi64( A1, z1 ); \
+  D0 = _mm512_xor_si512(D0, A0); \
+  D1 = _mm512_xor_si512(D1, A1); \
+  D0 = _mm512_ror_epi64(D0, 16); \
+  D1 = _mm512_ror_epi64(D1, 16); \
+  z0 = _mm512_mul_epu32( C0, D0 ); \
+  z1 = _mm512_mul_epu32( C1, D1 ); \
+  C0 = _mm512_add_epi64( C0, D0 ); \
+  C1 = _mm512_add_epi64( C1, D1 ); \
+  z0 = _mm512_add_epi64( z0, z0 ); \
+  z1 = _mm512_add_epi64( z1, z1 ); \
+  C0 = _mm512_add_epi64( C0, z0 ); \
+  C1 = _mm512_add_epi64( C1, z1 ); \
+  B0 = _mm512_xor_si512(B0, C0); \
+  B1 = _mm512_xor_si512(B1, C1); \
+  B0 = _mm512_ror_epi64(B0, 63); \
+  B1 = _mm512_ror_epi64(B1, 63); \
+}
+
+/*
 #define G1(A0, B0, C0, D0, A1, B1, C1, D1) \
     do { \
         A0 = muladd(A0, B0); \
@@ -316,7 +341,8 @@ static inline __m512i muladd(__m512i x, __m512i y)
         B0 = _mm512_ror_epi64(B0, 24); \
         B1 = _mm512_ror_epi64(B1, 24); \
     } while ((void)0, 0)
-
+*/
+/* 
 #define G2(A0, B0, C0, D0, A1, B1, C1, D1) \
     do { \
         A0 = muladd(A0, B0); \
@@ -337,15 +363,14 @@ static inline __m512i muladd(__m512i x, __m512i y)
         B0 = _mm512_ror_epi64(B0, 63); \
         B1 = _mm512_ror_epi64(B1, 63); \
     } while ((void)0, 0)
+*/
 
 #define DIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1) \
     do { \
         B0 = _mm512_permutex_epi64(B0, _MM_SHUFFLE(0, 3, 2, 1)); \
         B1 = _mm512_permutex_epi64(B1, _MM_SHUFFLE(0, 3, 2, 1)); \
-\
         C0 = _mm512_permutex_epi64(C0, _MM_SHUFFLE(1, 0, 3, 2)); \
         C1 = _mm512_permutex_epi64(C1, _MM_SHUFFLE(1, 0, 3, 2)); \
-\
         D0 = _mm512_permutex_epi64(D0, _MM_SHUFFLE(2, 1, 0, 3)); \
         D1 = _mm512_permutex_epi64(D1, _MM_SHUFFLE(2, 1, 0, 3)); \
     } while ((void)0, 0)
@@ -354,10 +379,8 @@ static inline __m512i muladd(__m512i x, __m512i y)
     do { \
         B0 = _mm512_permutex_epi64(B0, _MM_SHUFFLE(2, 1, 0, 3)); \
         B1 = _mm512_permutex_epi64(B1, _MM_SHUFFLE(2, 1, 0, 3)); \
-\
         C0 = _mm512_permutex_epi64(C0, _MM_SHUFFLE(1, 0, 3, 2)); \
         C1 = _mm512_permutex_epi64(C1, _MM_SHUFFLE(1, 0, 3, 2)); \
-\
         D0 = _mm512_permutex_epi64(D0, _MM_SHUFFLE(0, 3, 2, 1)); \
         D1 = _mm512_permutex_epi64(D1, _MM_SHUFFLE(0, 3, 2, 1)); \
     } while ((void)0, 0)
@@ -366,15 +389,17 @@ static inline __m512i muladd(__m512i x, __m512i y)
     do { \
         G1(A0, B0, C0, D0, A1, B1, C1, D1); \
         G2(A0, B0, C0, D0, A1, B1, C1, D1); \
-\
         DIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1); \
-\
         G1(A0, B0, C0, D0, A1, B1, C1, D1); \
         G2(A0, B0, C0, D0, A1, B1, C1, D1); \
-\
         UNDIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1); \
     } while ((void)0, 0)
 
+static const __m512i swap_q0  = { 0,1,  8,9,  2,3,  10,11 }; 
+static const __m512i swap_q1  = { 4,5, 12,13, 6,7,  14,15 };
+static const __m512i uswap_q0 = { 0,1,  4,5,  8,9,  12,13 };
+static const __m512i uswap_q1 = { 2,3,  6,7, 10,11, 14,15 };
+
 #define SWAP_HALVES(A0, A1) \
     do { \
         __m512i t; \
@@ -383,19 +408,36 @@ static inline __m512i muladd(__m512i x, __m512i y)
         A0 = t; \
     } while((void)0, 0)
 
+#define SWAP_QUARTERS(A0, A1) \
+{ \
+   __m512i t = _mm512_permutex2var_epi64( A0, swap_q0, A1 ); \
+   A1 = _mm512_permutex2var_epi64( A0, swap_q1, A1 ); \
+   A0 = t; \
+}   
+
+#define UNSWAP_QUARTERS(A0, A1) \
+{ \
+   __m512i t = _mm512_permutex2var_epi64( A0, uswap_q0, A1 ); \
+   A1 = _mm512_permutex2var_epi64( A0, uswap_q1, A1 ); \
+   A0 = t; \
+}   
+   
+/*
 #define SWAP_QUARTERS(A0, A1) \
     do { \
         SWAP_HALVES(A0, A1); \
         A0 = _mm512_shuffle_i64x2( A0, A0, 0xd8 ); \
         A1 = _mm512_shuffle_i64x2( A1, A1, 0xd8 ); \
     } while((void)0, 0)
-
+*/
+/*
 #define UNSWAP_QUARTERS(A0, A1) \
     do { \
         A0 = _mm512_shuffle_i64x2( A0, A0, 0xd8 ); \
         A1 = _mm512_shuffle_i64x2( A1, A1, 0xd8 ); \
         SWAP_HALVES(A0, A1); \
     } while((void)0, 0)
+*/
 
 #define BLAKE2_ROUND_1(A0, C0, B0, D0, A1, C1, B1, D1) \
     do { \

algo/blake/blake-4way.c

@@ -6,15 +6,15 @@
 
 #if defined (BLAKE_4WAY)
 
-blake256r14_4way_context blake_4w_ctx;
+blake256r14_4x32_context blake_4w_ctx;
 
 void blakehash_4way(void *state, const void *input)
 {
      uint32_t vhash[8*4] __attribute__ ((aligned (64)));
-     blake256r14_4way_context ctx;
+     blake256r14_4x32_context ctx;
      memcpy( &ctx, &blake_4w_ctx, sizeof ctx );
-     blake256r14_4way_update( &ctx, input + (64<<2), 16 );
+     blake256r14_4x32_update( &ctx, input + (64<<2), 16 );
-     blake256r14_4way_close( &ctx, vhash );
+     blake256r14_4x32_close( &ctx, vhash );
      dintrlv_4x32( state, state+32, state+64, state+96, vhash, 256 );
 }
 
@@ -35,8 +35,8 @@ int scanhash_blake_4way( struct work *work, uint32_t max_nonce,
       HTarget = 0x7f;
 
    v128_bswap32_intrlv80_4x32( vdata, pdata );
-   blake256r14_4way_init( &blake_4w_ctx );
+   blake256r14_4x32_init( &blake_4w_ctx );
-   blake256r14_4way_update( &blake_4w_ctx, vdata, 64 );
+   blake256r14_4x32_update( &blake_4w_ctx, vdata, 64 );
 
    do {
       *noncev = v128_bswap32( _mm_set_epi32( n+3, n+2, n+1, n ) );
@@ -61,15 +61,15 @@ int scanhash_blake_4way( struct work *work, uint32_t max_nonce,
 
 #if defined(BLAKE_8WAY)
 
-blake256r14_8way_context blake_8w_ctx;
+blake256r14_8x32_context blake_8w_ctx;
 
 void blakehash_8way( void *state, const void *input )
 {
      uint32_t vhash[8*8] __attribute__ ((aligned (64)));
-     blake256r14_8way_context ctx;
+     blake256r14_8x32_context ctx;
      memcpy( &ctx, &blake_8w_ctx, sizeof ctx );
-     blake256r14_8way( &ctx, input + (64<<3), 16 );
+     blake256r14_8x32( &ctx, input + (64<<3), 16 );
-     blake256r14_8way_close( &ctx, vhash );
+     blake256r14_8x32_close( &ctx, vhash );
      _dintrlv_8x32( state,     state+ 32, state+ 64, state+ 96,
                     state+128, state+160, state+192, state+224,
                     vhash, 256 );
@@ -93,8 +93,8 @@ int scanhash_blake_8way( struct work *work, uint32_t max_nonce,
 
    mm256_bswap32_intrlv80_8x32( vdata, pdata );
 
-   blake256r14_8way_init( &blake_8w_ctx );
+   blake256r14_8x32_init( &blake_8w_ctx );
-   blake256r14_8way( &blake_8w_ctx, vdata, 64 );
+   blake256r14_8x32( &blake_8w_ctx, vdata, 64 );
 
    do {
       *noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,

algo/blake/blake256-hash.c

@@ -423,33 +423,6 @@ void blake256_transform_le( uint32_t *H, const uint32_t *buf,
 		(state)->T1 = T1; \
 	} while (0)
 
-
-#if defined(__SSSE3__)
-
-#define BLAKE256_4X32_BLOCK_BSWAP32 \
-{ \
-   v128_t shuf_bswap32 = v128_set64( 0x0c0d0e0f08090a0b, \
-                                     0x0405060700010203 ); \
-   M0 = _mm_shuffle_epi8( buf[ 0], shuf_bswap32 ); \
-   M1 = _mm_shuffle_epi8( buf[ 1], shuf_bswap32 ); \
-   M2 = _mm_shuffle_epi8( buf[ 2], shuf_bswap32 ); \
-   M3 = _mm_shuffle_epi8( buf[ 3], shuf_bswap32 ); \
-   M4 = _mm_shuffle_epi8( buf[ 4], shuf_bswap32 ); \
-   M5 = _mm_shuffle_epi8( buf[ 5], shuf_bswap32 ); \
-   M6 = _mm_shuffle_epi8( buf[ 6], shuf_bswap32 ); \
-   M7 = _mm_shuffle_epi8( buf[ 7], shuf_bswap32 ); \
-   M8 = _mm_shuffle_epi8( buf[ 8], shuf_bswap32 ); \
-   M9 = _mm_shuffle_epi8( buf[ 9], shuf_bswap32 ); \
-   MA = _mm_shuffle_epi8( buf[10], shuf_bswap32 ); \
-   MB = _mm_shuffle_epi8( buf[11], shuf_bswap32 ); \
-   MC = _mm_shuffle_epi8( buf[12], shuf_bswap32 ); \
-   MD = _mm_shuffle_epi8( buf[13], shuf_bswap32 ); \
-   ME = _mm_shuffle_epi8( buf[14], shuf_bswap32 ); \
-   MF = _mm_shuffle_epi8( buf[15], shuf_bswap32 ); \
-}
-
-#else  // SSE2
-
 #define BLAKE256_4X32_BLOCK_BSWAP32 \
 { \
    M0 = v128_bswap32( buf[0] ); \
@@ -470,8 +443,6 @@ void blake256_transform_le( uint32_t *H, const uint32_t *buf,
    MF = v128_bswap32( buf[15] ); \
 }
 
-#endif  // SSSE3 else SSE2
-
 #define COMPRESS32_4X32( rounds ) \
 { \
    v128_t M0, M1, M2, M3, M4, M5, M6, M7; \
@@ -926,22 +897,6 @@ void blake256_4x32_final_rounds_le( void *final_hash, const void *midstate,
       ROUND_S_4X32_3;
    }
 
-#if defined(__SSSE3__)
-
-   const v128_t shuf_bswap32 =
-                      v128_set64( 0x0c0d0e0f08090a0b, 0x0405060700010203 );
-
-   H[0] = _mm_shuffle_epi8( v128_xor3( V8, V0, h[0] ), shuf_bswap32 );
-   H[1] = _mm_shuffle_epi8( v128_xor3( V9, V1, h[1] ), shuf_bswap32 );
-   H[2] = _mm_shuffle_epi8( v128_xor3( VA, V2, h[2] ), shuf_bswap32 );
-   H[3] = _mm_shuffle_epi8( v128_xor3( VB, V3, h[3] ), shuf_bswap32 );
-   H[4] = _mm_shuffle_epi8( v128_xor3( VC, V4, h[4] ), shuf_bswap32 );
-   H[5] = _mm_shuffle_epi8( v128_xor3( VD, V5, h[5] ), shuf_bswap32 );
-   H[6] = _mm_shuffle_epi8( v128_xor3( VE, V6, h[6] ), shuf_bswap32 );
-   H[7] = _mm_shuffle_epi8( v128_xor3( VF, V7, h[7] ), shuf_bswap32 );
-
-#else
-
    H[0] = v128_bswap32( v128_xor3( V8, V0, h[0] ) );
    H[1] = v128_bswap32( v128_xor3( V9, V1, h[1] ) );
    H[2] = v128_bswap32( v128_xor3( VA, V2, h[2] ) );
@@ -950,8 +905,6 @@ void blake256_4x32_final_rounds_le( void *final_hash, const void *midstate,
    H[5] = v128_bswap32( v128_xor3( VD, V5, h[5] ) );
    H[6] = v128_bswap32( v128_xor3( VE, V6, h[6] ) );
    H[7] = v128_bswap32( v128_xor3( VF, V7, h[7] ) );
-
-#endif
 }
 
 #if defined (__AVX2__)
@@ -1291,24 +1244,22 @@ do { \
    VD = v256_32( T0 ^ 0x299F31D0 ); \
    VE = v256_32( T1 ^ 0x082EFA98 ); \
    VF = v256_32( T1 ^ 0xEC4E6C89 ); \
-   const __m256i shuf_bswap32 = mm256_set2_64( \
+   M0 = mm256_bswap_32( * buf     ); \
-                               0x0c0d0e0f08090a0b, 0x0405060700010203 ); \
+   M1 = mm256_bswap_32( *(buf+ 1) ); \
-   M0 = _mm256_shuffle_epi8( * buf    , shuf_bswap32 ); \
+   M2 = mm256_bswap_32( *(buf+ 2) ); \
-   M1 = _mm256_shuffle_epi8( *(buf+ 1), shuf_bswap32 ); \
+   M3 = mm256_bswap_32( *(buf+ 3) ); \
-   M2 = _mm256_shuffle_epi8( *(buf+ 2), shuf_bswap32 ); \
+   M4 = mm256_bswap_32( *(buf+ 4) ); \
-   M3 = _mm256_shuffle_epi8( *(buf+ 3), shuf_bswap32 ); \
+   M5 = mm256_bswap_32( *(buf+ 5) ); \
-   M4 = _mm256_shuffle_epi8( *(buf+ 4), shuf_bswap32 ); \
+   M6 = mm256_bswap_32( *(buf+ 6) ); \
-   M5 = _mm256_shuffle_epi8( *(buf+ 5), shuf_bswap32 ); \
+   M7 = mm256_bswap_32( *(buf+ 7) ); \
-   M6 = _mm256_shuffle_epi8( *(buf+ 6), shuf_bswap32 ); \
+   M8 = mm256_bswap_32( *(buf+ 8) ); \
-   M7 = _mm256_shuffle_epi8( *(buf+ 7), shuf_bswap32 ); \
+   M9 = mm256_bswap_32( *(buf+ 9) ); \
-   M8 = _mm256_shuffle_epi8( *(buf+ 8), shuf_bswap32 ); \
+   MA = mm256_bswap_32( *(buf+10) ); \
-   M9 = _mm256_shuffle_epi8( *(buf+ 9), shuf_bswap32 ); \
+   MB = mm256_bswap_32( *(buf+11) ); \
-   MA = _mm256_shuffle_epi8( *(buf+10), shuf_bswap32 ); \
+   MC = mm256_bswap_32( *(buf+12) ); \
-   MB = _mm256_shuffle_epi8( *(buf+11), shuf_bswap32 ); \
+   MD = mm256_bswap_32( *(buf+13) ); \
-   MC = _mm256_shuffle_epi8( *(buf+12), shuf_bswap32 ); \
+   ME = mm256_bswap_32( *(buf+14) ); \
-   MD = _mm256_shuffle_epi8( *(buf+13), shuf_bswap32 ); \
+   MF = mm256_bswap_32( *(buf+15) ); \
-   ME = _mm256_shuffle_epi8( *(buf+14), shuf_bswap32 ); \
-   MF = _mm256_shuffle_epi8( *(buf+15), shuf_bswap32 ); \
    ROUND_S_8WAY(0); \
    ROUND_S_8WAY(1); \
    ROUND_S_8WAY(2); \
@@ -1401,7 +1352,7 @@ do { \
    H7 = mm256_xor3( VF, V7, H7 ); \
 }
 
-void blake256_8way_round0_prehash_le( void *midstate, const void *midhash,
+void blake256_8x32_round0_prehash_le( void *midstate, const void *midhash,
                                       void *data )
 {
    __m256i *M = (__m256i*)data;
@@ -1491,7 +1442,7 @@ void blake256_8way_round0_prehash_le( void *midstate, const void *midhash,
                          _mm256_xor_si256( v256_32( CSE ), M[15] ) );
 }
 
-void blake256_8way_final_rounds_le( void *final_hash, const void *midstate,
+void blake256_8x32_final_rounds_le( void *final_hash, const void *midstate,
                      const void *midhash, const void *data, const int rounds )
 {
    __m256i *H = (__m256i*)final_hash;
@@ -1596,22 +1547,19 @@ void blake256_8way_final_rounds_le( void *final_hash, const void *midstate,
       ROUND256_8WAY_3;
    }
 
-   const __m256i shuf_bswap32 =
+   H[0] = mm256_bswap_32( mm256_xor3( V8, V0, h[0] ) );
-                  mm256_set2_64( 0x0c0d0e0f08090a0b, 0x0405060700010203 );
+   H[1] = mm256_bswap_32( mm256_xor3( V9, V1, h[1] ) );
-
+   H[2] = mm256_bswap_32( mm256_xor3( VA, V2, h[2] ) );
-   H[0] = _mm256_shuffle_epi8( mm256_xor3( V8, V0, h[0] ), shuf_bswap32 );
+   H[3] = mm256_bswap_32( mm256_xor3( VB, V3, h[3] ) );
-   H[1] = _mm256_shuffle_epi8( mm256_xor3( V9, V1, h[1] ), shuf_bswap32 );
+   H[4] = mm256_bswap_32( mm256_xor3( VC, V4, h[4] ) );
-   H[2] = _mm256_shuffle_epi8( mm256_xor3( VA, V2, h[2] ), shuf_bswap32 );
+   H[5] = mm256_bswap_32( mm256_xor3( VD, V5, h[5] ) );
-   H[3] = _mm256_shuffle_epi8( mm256_xor3( VB, V3, h[3] ), shuf_bswap32 );
+   H[6] = mm256_bswap_32( mm256_xor3( VE, V6, h[6] ) );
-   H[4] = _mm256_shuffle_epi8( mm256_xor3( VC, V4, h[4] ), shuf_bswap32 );
+   H[7] = mm256_bswap_32( mm256_xor3( VF, V7, h[7] ) );
-   H[5] = _mm256_shuffle_epi8( mm256_xor3( VD, V5, h[5] ), shuf_bswap32 );
-   H[6] = _mm256_shuffle_epi8( mm256_xor3( VE, V6, h[6] ), shuf_bswap32 );
-   H[7] = _mm256_shuffle_epi8( mm256_xor3( VF, V7, h[7] ), shuf_bswap32 );
 }
 
 #endif
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 ///////////////////////////////////////
 //
@@ -1933,8 +1881,6 @@ do { \
    __m512i M8, M9, MA, MB, MC, MD, ME, MF; \
    __m512i V0, V1, V2, V3, V4, V5, V6, V7; \
    __m512i V8, V9, VA, VB, VC, VD, VE, VF; \
-   const __m512i shuf_bswap32 = mm512_bcast_m128( v128_set64( \
-                                 0x0c0d0e0f08090a0b, 0x0405060700010203 ) ); \
    V0 = H0; \
    V1 = H1; \
    V2 = H2; \
@@ -1951,22 +1897,22 @@ do { \
    VD = v512_32( T0 ^ 0x299F31D0 ); \
    VE = v512_32( T1 ^ 0x082EFA98 ); \
    VF = v512_32( T1 ^ 0xEC4E6C89 ); \
-   M0 = _mm512_shuffle_epi8( * buf    , shuf_bswap32 ); \
+   M0 = mm512_bswap_32( * buf     ); \
-   M1 = _mm512_shuffle_epi8( *(buf+ 1), shuf_bswap32 ); \
+   M1 = mm512_bswap_32( *(buf+ 1) ); \
-   M2 = _mm512_shuffle_epi8( *(buf+ 2), shuf_bswap32 ); \
+   M2 = mm512_bswap_32( *(buf+ 2) ); \
-   M3 = _mm512_shuffle_epi8( *(buf+ 3), shuf_bswap32 ); \
+   M3 = mm512_bswap_32( *(buf+ 3) ); \
-   M4 = _mm512_shuffle_epi8( *(buf+ 4), shuf_bswap32 ); \
+   M4 = mm512_bswap_32( *(buf+ 4) ); \
-   M5 = _mm512_shuffle_epi8( *(buf+ 5), shuf_bswap32 ); \
+   M5 = mm512_bswap_32( *(buf+ 5) ); \
-   M6 = _mm512_shuffle_epi8( *(buf+ 6), shuf_bswap32 ); \
+   M6 = mm512_bswap_32( *(buf+ 6) ); \
-   M7 = _mm512_shuffle_epi8( *(buf+ 7), shuf_bswap32 ); \
+   M7 = mm512_bswap_32( *(buf+ 7) ); \
-   M8 = _mm512_shuffle_epi8( *(buf+ 8), shuf_bswap32 ); \
+   M8 = mm512_bswap_32( *(buf+ 8) ); \
-   M9 = _mm512_shuffle_epi8( *(buf+ 9), shuf_bswap32 ); \
+   M9 = mm512_bswap_32( *(buf+ 9) ); \
-   MA = _mm512_shuffle_epi8( *(buf+10), shuf_bswap32 ); \
+   MA = mm512_bswap_32( *(buf+10) ); \
-   MB = _mm512_shuffle_epi8( *(buf+11), shuf_bswap32 ); \
+   MB = mm512_bswap_32( *(buf+11) ); \
-   MC = _mm512_shuffle_epi8( *(buf+12), shuf_bswap32 ); \
+   MC = mm512_bswap_32( *(buf+12) ); \
-   MD = _mm512_shuffle_epi8( *(buf+13), shuf_bswap32 ); \
+   MD = mm512_bswap_32( *(buf+13) ); \
-   ME = _mm512_shuffle_epi8( *(buf+14), shuf_bswap32 ); \
+   ME = mm512_bswap_32( *(buf+14) ); \
-   MF = _mm512_shuffle_epi8( *(buf+15), shuf_bswap32 ); \
+   MF = mm512_bswap_32( *(buf+15) ); \
    ROUND_S_16WAY(0); \
    ROUND_S_16WAY(1); \
    ROUND_S_16WAY(2); \
@@ -2063,7 +2009,7 @@ do { \
 // is constant for every nonce and only needs to be run once per job. The
 // second part is run for each nonce using the precalculated midstate and the
 // hash from the first block.
-void blake256_16way_round0_prehash_le( void *midstate, const void *midhash,
+void blake256_16x32_round0_prehash_le( void *midstate, const void *midhash,
                                        void *data )
 {
    __m512i *M = (__m512i*)data;
@@ -2157,7 +2103,7 @@ void blake256_16way_round0_prehash_le( void *midstate, const void *midhash,
 }
 
 // Dfault is 14 rounds, blakecoin & vanilla are 8.
-void blake256_16way_final_rounds_le( void *final_hash, const void *midstate,
+void blake256_16x32_final_rounds_le( void *final_hash, const void *midstate,
                      const void *midhash, const void *data, const int rounds )
 {
    __m512i *H = (__m512i*)final_hash;
@@ -2274,27 +2220,23 @@ void blake256_16way_final_rounds_le( void *final_hash, const void *midstate,
    }
 
    // Byte swap final hash
-   const __m512i shuf_bswap32 =  mm512_bcast_m128( v128_set64( 
+   H[0] = mm512_bswap_32( mm512_xor3( V8, V0, h[0] ) );
-                                 0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
+   H[1] = mm512_bswap_32( mm512_xor3( V9, V1, h[1] ) );
-   H[0] = _mm512_shuffle_epi8( mm512_xor3( V8, V0, h[0] ), shuf_bswap32 );
+   H[2] = mm512_bswap_32( mm512_xor3( VA, V2, h[2] ) );
-   H[1] = _mm512_shuffle_epi8( mm512_xor3( V9, V1, h[1] ), shuf_bswap32 );
+   H[3] = mm512_bswap_32( mm512_xor3( VB, V3, h[3] ) );
-   H[2] = _mm512_shuffle_epi8( mm512_xor3( VA, V2, h[2] ), shuf_bswap32 );
+   H[4] = mm512_bswap_32( mm512_xor3( VC, V4, h[4] ) );
-   H[3] = _mm512_shuffle_epi8( mm512_xor3( VB, V3, h[3] ), shuf_bswap32 );
+   H[5] = mm512_bswap_32( mm512_xor3( VD, V5, h[5] ) );
-   H[4] = _mm512_shuffle_epi8( mm512_xor3( VC, V4, h[4] ), shuf_bswap32 );
+   H[6] = mm512_bswap_32( mm512_xor3( VE, V6, h[6] ) );
-   H[5] = _mm512_shuffle_epi8( mm512_xor3( VD, V5, h[5] ), shuf_bswap32 );
+   H[7] = mm512_bswap_32( mm512_xor3( VF, V7, h[7] ) );
-   H[6] = _mm512_shuffle_epi8( mm512_xor3( VE, V6, h[6] ), shuf_bswap32 );
-   H[7] = _mm512_shuffle_epi8( mm512_xor3( VF, V7, h[7] ), shuf_bswap32 );
 }
 
 #endif
 
 // Blake-256 4 way
 
-static const uint32_t salt_zero_4x32_small[4] = { 0, 0, 0, 0 };
-
 static void
 blake32_4x32_init( blake_4x32_small_context *ctx, const uint32_t *iv,
-                   const uint32_t *salt, int rounds )
+                   int rounds )
 {
    casti_v128( ctx->H, 0 ) = v128_64( 0x6A09E6676A09E667 );
    casti_v128( ctx->H, 1 ) = v128_64( 0xBB67AE85BB67AE85 );
@@ -2404,11 +2346,10 @@ blake32_4x32_close( blake_4x32_small_context *ctx, unsigned ub, unsigned n,
 
 // Blake-256 8 way
 
-static const uint32_t salt_zero_8way_small[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
 
 static void
-blake32_8way_init( blake_8way_small_context *sc, const uint32_t *iv,
+blake32_8way_init( blake256_8x32_context *sc, const uint32_t *iv,
-                   const uint32_t *salt, int rounds )
+                   int rounds )
 {
    casti_m256i( sc->H, 0 ) = v256_64( 0x6A09E6676A09E667 );
    casti_m256i( sc->H, 1 ) = v256_64( 0xBB67AE85BB67AE85 );
@@ -2424,7 +2365,7 @@ blake32_8way_init( blake_8way_small_context *sc, const uint32_t *iv,
 }
 
 static void
-blake32_8way( blake_8way_small_context *sc, const void *data, size_t len )
+blake32_8way( blake256_8x32_context *sc, const void *data, size_t len )
 {
    __m256i *vdata = (__m256i*)data;
    __m256i *buf;
@@ -2466,7 +2407,7 @@ blake32_8way( blake_8way_small_context *sc, const void *data, size_t len )
 }
 
 static void
-blake32_8way_close( blake_8way_small_context *sc, unsigned ub, unsigned n,
+blake32_8way_close( blake256_8x32_context *sc, unsigned ub, unsigned n,
                     void *dst, size_t out_size_w32 )
 {
    __m256i buf[16];
@@ -2520,7 +2461,7 @@ blake32_8way_close( blake_8way_small_context *sc, unsigned ub, unsigned n,
 }
 
 static void
-blake32_8way_le( blake_8way_small_context *sc, const void *data, size_t len )
+blake32_8way_le( blake256_8x32_context *sc, const void *data, size_t len )
 {
    __m256i *vdata = (__m256i*)data;
    __m256i *buf;
@@ -2562,7 +2503,7 @@ blake32_8way_le( blake_8way_small_context *sc, const void *data, size_t len )
 }
 
 static void
-blake32_8way_close_le( blake_8way_small_context *sc, unsigned ub, unsigned n,
+blake32_8way_close_le( blake256_8x32_context *sc, unsigned ub, unsigned n,
                        void *dst, size_t out_size_w32 )
 {
    __m256i buf[16];
@@ -2617,13 +2558,13 @@ blake32_8way_close_le( blake_8way_small_context *sc, unsigned ub, unsigned n,
 #endif
 
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 //Blake-256 16 way AVX512
 
 static void
-blake32_16way_init( blake_16way_small_context *sc, const uint32_t *iv,
+blake32_16way_init( blake256_16x32_context *sc, const uint32_t *iv,
-                   const uint32_t *salt, int rounds )
+                    int rounds )
 {
    casti_m512i( sc->H, 0 ) = v512_64( 0x6A09E6676A09E667 );
    casti_m512i( sc->H, 1 ) = v512_64( 0xBB67AE85BB67AE85 );
@@ -2639,7 +2580,7 @@ blake32_16way_init( blake_16way_small_context *sc, const uint32_t *iv,
 }
 
 static void
-blake32_16way( blake_16way_small_context *sc, const void *data, size_t len )
+blake32_16way( blake256_16x32_context *sc, const void *data, size_t len )
 {
    __m512i *vdata = (__m512i*)data;
    __m512i *buf;
@@ -2679,7 +2620,7 @@ blake32_16way( blake_16way_small_context *sc, const void *data, size_t len )
    sc->ptr = ptr;
 }
 static void
-blake32_16way_close( blake_16way_small_context *sc, unsigned ub, unsigned n,
+blake32_16way_close( blake256_16x32_context *sc, unsigned ub, unsigned n,
                     void *dst, size_t out_size_w32 )
 {
    __m512i buf[16];
@@ -2733,7 +2674,7 @@ blake32_16way_close( blake_16way_small_context *sc, unsigned ub, unsigned n,
 }
 
 static void
-blake32_16way_le( blake_16way_small_context *sc, const void *data, size_t len )
+blake32_16way_le( blake256_16x32_context *sc, const void *data, size_t len )
 {
    __m512i *vdata = (__m512i*)data;
    __m512i *buf;
@@ -2776,7 +2717,7 @@ blake32_16way_le( blake_16way_small_context *sc, const void *data, size_t len )
 }
 
 static void
-blake32_16way_close_le( blake_16way_small_context *sc, unsigned ub, unsigned n,
+blake32_16way_close_le( blake256_16x32_context *sc, unsigned ub, unsigned n,
                     void *dst, size_t out_size_w32 )
 {
    __m512i buf[16];
@@ -2827,65 +2768,65 @@ blake32_16way_close_le( blake_16way_small_context *sc, unsigned ub, unsigned n,
 }
 
 void
-blake256_16way_init(void *cc)
+blake256_16x32_init(void *cc)
 {
-   blake32_16way_init( cc, IV256, salt_zero_8way_small, 14 );
+   blake32_16way_init( cc, IV256, 14 );
 }
 
 void
-blake256_16way_update(void *cc, const void *data, size_t len)
+blake256_16x32_update(void *cc, const void *data, size_t len)
 {
         blake32_16way(cc, data, len);
 }
 
 void
-blake256_16way_close(void *cc, void *dst)
+blake256_16x32_close(void *cc, void *dst)
 {
         blake32_16way_close(cc, 0, 0, dst, 8);
 }
 
 void
-blake256_16way_update_le(void *cc, const void *data, size_t len)
+blake256_16x32_update_le(void *cc, const void *data, size_t len)
 {
    blake32_16way_le(cc, data, len);
 }
 
 void
-blake256_16way_close_le(void *cc, void *dst)
+blake256_16x32_close_le(void *cc, void *dst)
 {
     blake32_16way_close_le(cc, 0, 0, dst, 8);
 }
 
 void blake256r14_16way_init(void *cc)
 {
-   blake32_16way_init( cc, IV256, salt_zero_8way_small, 14 );
+   blake32_16way_init( cc, IV256, 14 );
 }
 
 void
-blake256r14_16way_update(void *cc, const void *data, size_t len)
+blake256r14_16x32_update(void *cc, const void *data, size_t len)
 {
    blake32_16way(cc, data, len);
 }
 
 void
-blake256r14_16way_close(void *cc, void *dst)
+blake256r14_16x32_close(void *cc, void *dst)
 {
    blake32_16way_close(cc, 0, 0, dst, 8);
 }
 
 void blake256r8_16way_init(void *cc)
 {
-   blake32_16way_init( cc, IV256, salt_zero_8way_small, 8 );
+   blake32_16way_init( cc, IV256, 8 );
 }
 
 void
-blake256r8_16way_update(void *cc, const void *data, size_t len)
+blake256r8_16x32_update(void *cc, const void *data, size_t len)
 {
    blake32_16way(cc, data, len);
 }
 
 void
-blake256r8_16way_close(void *cc, void *dst)
+blake256r8_16x32_close(void *cc, void *dst)
 {
    blake32_16way_close(cc, 0, 0, dst, 8);
 }
@@ -2898,7 +2839,7 @@ blake256r8_16way_close(void *cc, void *dst)
 void
 blake256_4x32_init(void *ctx)
 {
-   blake32_4x32_init( ctx, IV256, salt_zero_4x32_small, 14 );
+   blake32_4x32_init( ctx, IV256, 14 );
 }
 
 void
@@ -2918,31 +2859,31 @@ blake256_4x32_close(void *ctx, void *dst)
 // Blake-256 8 way
 
 void
-blake256_8way_init(void *cc)
+blake256_8x32_init(void *cc)
 {
-   blake32_8way_init( cc, IV256, salt_zero_8way_small, 14 );
+   blake32_8way_init( cc, IV256, 14 );
 }
 
 void
-blake256_8way_update(void *cc, const void *data, size_t len)
+blake256_8x32_update(void *cc, const void *data, size_t len)
 {
         blake32_8way(cc, data, len);
 }
 
 void
-blake256_8way_close(void *cc, void *dst)
+blake256_8x32_close(void *cc, void *dst)
 {
         blake32_8way_close(cc, 0, 0, dst, 8);
 }
 
 void
-blake256_8way_update_le(void *cc, const void *data, size_t len)
+blake256_8x32_update_le(void *cc, const void *data, size_t len)
 {
         blake32_8way_le(cc, data, len);
 }
 
 void
-blake256_8way_close_le(void *cc, void *dst)
+blake256_8x32_close_le(void *cc, void *dst)
 {
         blake32_8way_close_le(cc, 0, 0, dst, 8);
 }
@@ -2952,7 +2893,7 @@ blake256_8way_close_le(void *cc, void *dst)
 // 14 rounds Blake, Decred
 void blake256r14_4x32_init(void *cc)
 {
-   blake32_4x32_init( cc, IV256, salt_zero_4x32_small, 14 );
+   blake32_4x32_init( cc, IV256, 14 );
 }
 
 void
@@ -2969,19 +2910,19 @@ blake256r14_4x32_close(void *cc, void *dst)
 
 #if defined(__AVX2__)
 
-void blake256r14_8way_init(void *cc)
+void blake256r14_8x32_init(void *cc)
 {
-   blake32_8way_init( cc, IV256, salt_zero_8way_small, 14 );
+   blake32_8way_init( cc, IV256, 14 );
 }
 
 void
-blake256r14_8way_update(void *cc, const void *data, size_t len)
+blake256r14_8x32_update(void *cc, const void *data, size_t len)
 {
    blake32_8way(cc, data, len);
 }
 
 void
-blake256r14_8way_close(void *cc, void *dst)
+blake256r14_8x32_close(void *cc, void *dst)
 {
    blake32_8way_close(cc, 0, 0, dst, 8);
 }
@@ -2991,7 +2932,7 @@ blake256r14_8way_close(void *cc, void *dst)
 // 8 rounds Blakecoin, Vanilla
 void blake256r8_4x32_init(void *cc)
 {
-   blake32_4x32_init( cc, IV256, salt_zero_4x32_small, 8 );
+   blake32_4x32_init( cc, IV256, 8 );
 }
 
 void
@@ -3008,19 +2949,19 @@ blake256r8_4x32_close(void *cc, void *dst)
 
 #if defined (__AVX2__)
 
-void blake256r8_8way_init(void *cc)
+void blake256r8_8x32_init(void *cc)
 {
-   blake32_8way_init( cc, IV256, salt_zero_8way_small, 8 );
+   blake32_8way_init( cc, IV256, 8 );
 }
 
 void
-blake256r8_8way_update(void *cc, const void *data, size_t len)
+blake256r8_8x32_update(void *cc, const void *data, size_t len)
 {
    blake32_8way(cc, data, len);
 }
 
 void
-blake256r8_8way_close(void *cc, void *dst)
+blake256r8_8x32_close(void *cc, void *dst)
 {
    blake32_8way_close(cc, 0, 0, dst, 8);
 }

algo/blake/blake256-hash.h

@@ -29,13 +29,6 @@ typedef struct
 
 void blake256_transform_le( uint32_t *H, const uint32_t *buf,
                             const uint32_t T0, const uint32_t T1, int rounds );
-/*
-void blake256_init( blake256_context *sc );
-void blake256_update( blake256_context *sc, const void *data, size_t len );
-void blake256_close( blake256_context *sc, void *dst );
-void blake256_full( blake256_context *sc, void *dst, const void *data,
-                    size_t len );
-*/
 
 //////////////////////////////////
 //
@@ -55,6 +48,10 @@ typedef blake_4x32_small_context blake256_4x32_context;
 void blake256_4x32_init(void *ctx);
 void blake256_4x32_update(void *ctx, const void *data, size_t len);
 void blake256_4x32_close(void *ctx, void *dst);
+void blake256_4x32_round0_prehash_le( void *midstate, const void *midhash,
+                                      void *data );
+void blake256_4x32_final_rounds_le( void *final_hash, const void *midstate,
+                     const void *midhash, const void *data, const int rounds );
 
 // 14 rounds
 typedef blake_4x32_small_context blake256r14_4x32_context;
@@ -68,29 +65,6 @@ void blake256r8_4x32_init(void *cc);
 void blake256r8_4x32_update(void *cc, const void *data, size_t len);
 void blake256r8_4x32_close(void *cc, void *dst);
 
-void blake256_4x32_round0_prehash_le( void *midstate, const void *midhash,
-                                      void *data );
-void blake256_4x32_final_rounds_le( void *final_hash, const void *midstate,
-                     const void *midhash, const void *data, const int rounds );
-
-#define blake_4way_small_context         blake256_4x32_context
-#define blake256_4way_context            blake256_4x32_context
-#define blake256_4way_init               blake256_4x32_init
-#define blake256_4way_update             blake256_4x32_update
-#define blake256_4way_close              blake256_4x32_close
-#define blake256_4way_update_le          blake256_4x32_update_le
-#define blake256_4way_close_le           blake256_4x32_close_le
-#define blake256_4way_round0_prehash_le  blake256_4x32_round0_prehash_le
-#define blake256_4way_final_rounds_le    blake256_4x32_final_rounds_le
-#define blake256r14_4way_context         blake256r14_4x32_context
-#define blake256r14_4way_init            blake256r14_4x32_init
-#define blake256r14_4way_update          blake256r14_4x32_update
-#define blake256r14_4way_close           blake256r14_4x32_close
-#define blake256r8_4way_context          blake256r14_4x32_context
-#define blake256r8_4way_init             blake256r14_4x32_init
-#define blake256r8_4way_update           blake256r14_4x32_update
-#define blake256r8_4way_close            blake256r14_4x32_close
-
 #ifdef __AVX2__
 
 //////////////////////////////
@@ -107,47 +81,30 @@ typedef struct
 } blake_8way_small_context;
 
 // Default 14 rounds
-typedef blake_8way_small_context blake256_8way_context;
+typedef blake_8way_small_context blake256_8x32_context;
-void blake256_8way_init(void *cc);
+void blake256_8x32_init(void *cc);
-void blake256_8way_update(void *cc, const void *data, size_t len);
+void blake256_8x32_update(void *cc, const void *data, size_t len);
-void blake256_8way_close(void *cc, void *dst);
+void blake256_8x32_close(void *cc, void *dst);
-void blake256_8way_update_le(void *cc, const void *data, size_t len);
+void blake256_8x32_update_le(void *cc, const void *data, size_t len);
-void blake256_8way_close_le(void *cc, void *dst);
+void blake256_8x32_close_le(void *cc, void *dst);
-void blake256_8way_round0_prehash_le( void *midstate, const void *midhash,
+void blake256_8x32_round0_prehash_le( void *midstate, const void *midhash,
                                       void *data );
-void blake256_8way_final_rounds_le( void *final_hash, const void *midstate,
+void blake256_8x32_final_rounds_le( void *final_hash, const void *midstate,
                     const void *midhash, const void *data, const int rounds );
 
 // 14 rounds, blake, decred
-typedef blake_8way_small_context blake256r14_8way_context;
+typedef blake_8way_small_context blake256r14_8x32_context;
-void blake256r14_8way_init(void *cc);
+void blake256r14_8x32_init(void *cc);
-void blake256r14_8way_update(void *cc, const void *data, size_t len);
+void blake256r14_8x32_update(void *cc, const void *data, size_t len);
-void blake256r14_8way_close(void *cc, void *dst);
+void blake256r14_8x32_close(void *cc, void *dst);
 
 // 8 rounds, blakecoin, vanilla
-typedef blake_8way_small_context blake256r8_8way_context;
+typedef blake_8way_small_context blake256r8_8x32_context;
-void blake256r8_8way_init(void *cc);
+void blake256r8_8x32_init(void *cc);
-void blake256r8_8way_update(void *cc, const void *data, size_t len);
+void blake256r8_8x32_update(void *cc, const void *data, size_t len);
-void blake256r8_8way_close(void *cc, void *dst);
+void blake256r8_8x32_close(void *cc, void *dst);
 
-#define blake_8x32_small_context      blake256_8way_context
+#if defined(SIMD512)
-#define blake_8x32_init               blake256_8way_init
-#define blake_8x32_update             blake256_8way_update
-#define blake_8x32_close              blake256_8way_close
-#define blake_8x32_update_le          blake256_8way_update_le
-#define blake_8x32_close_le           blake256_8way_close_le
-#define blake_8x32_round0_prehash_le  blake256_8way_round0_prehash
-#define blake_8x32_final_rounds_le    blake256_8way_final_rounds_le
-#define blake256r14_8x32_context      blake256r14_8way_context
-#define blake256r14_8x32_init         blake256r14_8way_init
-#define blake256r14_8x32_update       blake256r14_8way_update
-#define blake256r14_8x32_close        blake256r14_8way_close
-#define blake256r8_8x32_context       blake256r14_8way_context
-#define blake256r8_8x32_init          blake256r14_8way_init
-#define blake256r8_8x32_update        blake256r14_8way_update
-#define blake256r8_8x32_close         blake256r14_8way_close
-
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
 
 ///////////////////////////////////
 //
@@ -163,46 +120,29 @@ typedef struct
 } blake_16way_small_context __attribute__ ((aligned (128)));
 
 // Default 14 rounds
-typedef blake_16way_small_context blake256_16way_context;
+typedef blake_16way_small_context blake256_16x32_context;
-void blake256_16way_init(void *cc);
+void blake256_16x32_init(void *cc);
-void blake256_16way_update(void *cc, const void *data, size_t len);
+void blake256_16x32_update(void *cc, const void *data, size_t len);
-void blake256_16way_close(void *cc, void *dst);
+void blake256_16x32_close(void *cc, void *dst);
 // Expects data in little endian order, no byte swap needed
-void blake256_16way_update_le(void *cc, const void *data, size_t len);
+void blake256_16x32_update_le(void *cc, const void *data, size_t len);
-void blake256_16way_close_le(void *cc, void *dst);
+void blake256_16x32_close_le(void *cc, void *dst);
-void blake256_16way_round0_prehash_le( void *midstate, const void *midhash,
+void blake256_16x32_round0_prehash_le( void *midstate, const void *midhash,
                                        void *data );
-void blake256_16way_final_rounds_le( void *final_hash, const void *midstate,
+void blake256_16x32_final_rounds_le( void *final_hash, const void *midstate,
                      const void *midhash, const void *data, const int rounds );
 
 // 14 rounds, blake, decred
-typedef blake_16way_small_context blake256r14_16way_context;
+typedef blake_16way_small_context blake256r14_16x32_context;
-void blake256r14_16way_init(void *cc);
+void blake256r14_16x32_init(void *cc);
-void blake256r14_16way_update(void *cc, const void *data, size_t len);
+void blake256r14_16x32_update(void *cc, const void *data, size_t len);
-void blake256r14_16way_close(void *cc, void *dst);
+void blake256r14_16x32_close(void *cc, void *dst);
 
 // 8 rounds, blakecoin, vanilla
-typedef blake_16way_small_context blake256r8_16way_context;
+typedef blake_16way_small_context blake256r8_16x32_context;
-void blake256r8_16way_init(void *cc);
+void blake256r8_16x32_init(void *cc);
-void blake256r8_16way_update(void *cc, const void *data, size_t len);
+void blake256r8_16x32_update(void *cc, const void *data, size_t len);
-void blake256r8_16way_close(void *cc, void *dst);
+void blake256r8_16x32_close(void *cc, void *dst);
-
-#define blake_16x32_small_context      blake256_16way_context
-#define blake_16x32_init               blake256_16way_init
-#define blake_16x32_update             blake256_16way_update
-#define blake_16x32_close              blake256_16way_close
-#define blake_16x32_update_le          blake256_16way_update_le
-#define blake_16x32_close_le           blake256_16way_close_le
-#define blake_16x32_round0_prehash_le  blake256_16way_round0_prehash
-#define blake_16x32_final_rounds_le    blake256_16way_final_rounds_le
-#define blake256r14_16x32_context      blake256r14_16way_context
-#define blake256r14_16x32_init         blake256r14_16way_init
-#define blake256r14_16x32_update       blake256r14_16way_update
-#define blake256r14_16x32_close        blake256r14_16way_close
-#define blake256r8_16x32_context       blake256r8_16way_context
-#define blake256r8_16x32_init          blake256r8_16way_init
-#define blake256r8_16x32_update        blake256r8_16way_update
-#define blake256r8_16x32_close         blake256r8_16way_close
 
 #endif  // AVX512
 #endif  // AVX2

algo/blake/blake2b-hash.c

@@ -226,7 +226,7 @@ static const uint8_t sigma[12][16] =
 #define Mx_(n)      Mx__(n)
 #define Mx__(n)     M ## n
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 #define B2B8W_G(a, b, c, d, x, y) \
 { \
@@ -240,7 +240,7 @@ static const uint8_t sigma[12][16] =
    v[b] = mm512_ror_64( _mm512_xor_si512( v[b], v[c] ), 63 ); \
 }
 
-static void blake2b_8way_compress( blake2b_8way_ctx *ctx, int last )
+static void blake2b_8x64_compress( blake2b_8x64_ctx *ctx, int last )
 {  
    __m512i v[16], m[16];
 
@@ -306,7 +306,7 @@ static void blake2b_8way_compress( blake2b_8way_ctx *ctx, int last )
    ctx->h[7] = mm512_xor3( ctx->h[7], v[7], v[15] );
 }
 
-int blake2b_8way_init( blake2b_8way_ctx *ctx )
+int blake2b_8x64_init( blake2b_8x64_ctx *ctx )
 {
    size_t i;
 
@@ -333,7 +333,7 @@ int blake2b_8way_init( blake2b_8way_ctx *ctx )
 }
 
 
-void blake2b_8way_update( blake2b_8way_ctx *ctx, const void *input,
+void blake2b_8x64_update( blake2b_8x64_ctx *ctx, const void *input,
                           size_t inlen )
 {
    __m512i* in =(__m512i*)input;
@@ -348,7 +348,7 @@ void blake2b_8way_update( blake2b_8way_ctx *ctx, const void *input,
          ctx->t[0] += ctx->c;
          if ( ctx->t[0] < ctx->c )
             ctx->t[1]++;
-         blake2b_8way_compress( ctx, 0 );
+         blake2b_8x64_compress( ctx, 0 );
          ctx->c = 0;
       }
       ctx->b[ c++ ] = in[i];
@@ -356,7 +356,7 @@ void blake2b_8way_update( blake2b_8way_ctx *ctx, const void *input,
    }
 }
 
-void blake2b_8way_final( blake2b_8way_ctx *ctx, void *out )
+void blake2b_8x64_final( blake2b_8x64_ctx *ctx, void *out )
 {
    size_t c;
    c = ctx->c >> 3;
@@ -371,7 +371,7 @@ void blake2b_8way_final( blake2b_8way_ctx *ctx, void *out )
       ctx->c += 8;
    }
 
-   blake2b_8way_compress( ctx, 1 );           // final block flag = 1
+   blake2b_8x64_compress( ctx, 1 );           // final block flag = 1
 
    casti_m512i( out, 0 ) = ctx->h[0];
    casti_m512i( out, 1 ) = ctx->h[1];
@@ -407,7 +407,7 @@ static const uint64_t blake2b_iv[8] = {
 };
 */
 
-static void blake2b_4way_compress( blake2b_4way_ctx *ctx, int last )
+static void blake2b_4x64_compress( blake2b_4x64_ctx *ctx, int last )
 {
 	__m256i v[16], m[16];
 
@@ -473,7 +473,7 @@ static void blake2b_4way_compress( blake2b_4way_ctx *ctx, int last )
    ctx->h[7] = _mm256_xor_si256( _mm256_xor_si256( ctx->h[7], v[7] ), v[15] );
 }
 
-int blake2b_4way_init( blake2b_4way_ctx *ctx ) 
+int blake2b_4x64_init( blake2b_4x64_ctx *ctx ) 
 {
 	size_t i;
 
@@ -499,7 +499,7 @@ int blake2b_4way_init( blake2b_4way_ctx *ctx )
 	return 0;
 }
 
-void blake2b_4way_update( blake2b_4way_ctx *ctx, const void *input,
+void blake2b_4x64_update( blake2b_4x64_ctx *ctx, const void *input,
                           size_t inlen ) 
 {
    __m256i* in =(__m256i*)input;
@@ -514,7 +514,7 @@ void blake2b_4way_update( blake2b_4way_ctx *ctx, const void *input,
 			ctx->t[0] += ctx->c;
 			if ( ctx->t[0] < ctx->c )
 				ctx->t[1]++;
-			blake2b_4way_compress( ctx, 0 );
+			blake2b_4x64_compress( ctx, 0 );
 			ctx->c = 0;
 		}
       ctx->b[ c++ ] = in[i];
@@ -522,7 +522,7 @@ void blake2b_4way_update( blake2b_4way_ctx *ctx, const void *input,
    }
 }
 
-void blake2b_4way_final( blake2b_4way_ctx *ctx, void *out )
+void blake2b_4x64_final( blake2b_4x64_ctx *ctx, void *out )
 {
 	size_t c;
    c = ctx->c >> 3;
@@ -537,7 +537,7 @@ void blake2b_4way_final( blake2b_4way_ctx *ctx, void *out )
       ctx->c += 8;
    }
 
-   blake2b_4way_compress( ctx, 1 );           // final block flag = 1
+   blake2b_4x64_compress( ctx, 1 );           // final block flag = 1
 
    casti_m256i( out, 0 ) = ctx->h[0];
    casti_m256i( out, 1 ) = ctx->h[1];

algo/blake/blake2b-hash.h

@@ -1,6 +1,6 @@
 #pragma once
-#ifndef __BLAKE2B_HASH_4WAY_H__
+#ifndef BLAKE2B_HASH_4WAY_H__
-#define __BLAKE2B_HASH_4WAY_H__
+#define BLAKE2B_HASH_4WAY_H__
 
 #include "simd-utils.h"
 #include <stddef.h>
@@ -14,8 +14,7 @@
 #define ALIGN(x) __attribute__((aligned(x)))
 #endif
 
-
+#if defined(SIMD512)
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
 
 typedef struct ALIGN( 64 ) {
    __m512i b[16]; // input buffer
@@ -23,12 +22,12 @@ typedef struct ALIGN( 64 ) {
    uint64_t t[2];  // total number of bytes
    size_t c;       // pointer for b[]
    size_t outlen;  // digest size
-} blake2b_8way_ctx;
+} blake2b_8x64_ctx;
 
-int blake2b_8way_init( blake2b_8way_ctx *ctx );
+int blake2b_8x64_init( blake2b_8x64_ctx *ctx );
-void blake2b_8way_update( blake2b_8way_ctx *ctx, const void *input,
+void blake2b_8x64_update( blake2b_8x64_ctx *ctx, const void *input,
                           size_t inlen );
-void blake2b_8way_final( blake2b_8way_ctx *ctx, void *out );
+void blake2b_8x64_final( blake2b_8x64_ctx *ctx, void *out );
 
 #endif
 
@@ -41,12 +40,12 @@ typedef struct ALIGN( 64 ) {
 	uint64_t t[2];  // total number of bytes
 	size_t c;       // pointer for b[]
 	size_t outlen;  // digest size
-} blake2b_4way_ctx;
+} blake2b_4x64_ctx;
 
-int blake2b_4way_init( blake2b_4way_ctx *ctx );
+int blake2b_4x64_init( blake2b_4x64_ctx *ctx );
-void blake2b_4way_update( blake2b_4way_ctx *ctx, const void *input,
+void blake2b_4x64_update( blake2b_4x64_ctx *ctx, const void *input,
                           size_t inlen );
-void blake2b_4way_final( blake2b_4way_ctx *ctx, void *out );
+void blake2b_4x64_final( blake2b_4x64_ctx *ctx, void *out );
 
 #endif
 

algo/blake/blake2b.c

@@ -3,7 +3,7 @@
 #include <stdint.h>
 #include "blake2b-hash.h"
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define BLAKE2B_8WAY
 #elif defined(__AVX2__)
   #define BLAKE2B_4WAY
@@ -17,7 +17,7 @@ int scanhash_blake2b_8way( struct work *work, uint32_t max_nonce,
    uint32_t hash[8*8] __attribute__ ((aligned (128)));;
    uint32_t vdata[20*8] __attribute__ ((aligned (64)));;
    uint32_t lane_hash[8] __attribute__ ((aligned (64)));
-   blake2b_8way_ctx ctx __attribute__ ((aligned (64)));
+   blake2b_8x64_ctx ctx __attribute__ ((aligned (64)));
    uint32_t *hash7 = &(hash[49]);   // 3*16+1
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
@@ -35,9 +35,9 @@ int scanhash_blake2b_8way( struct work *work, uint32_t max_nonce,
                 _mm512_set_epi32( n+7, 0, n+6, 0, n+5, 0, n+4, 0,
                                   n+3, 0, n+2, 0, n+1, 0, n  , 0 ) ), *noncev );
 
-      blake2b_8way_init( &ctx );
+      blake2b_8x64_init( &ctx );
-      blake2b_8way_update( &ctx, vdata, 80 );
+      blake2b_8x64_update( &ctx, vdata, 80 );
-      blake2b_8way_final( &ctx, hash );
+      blake2b_8x64_final( &ctx, hash );
 
       for ( int lane = 0; lane < 8; lane++ )
       if ( hash7[ lane<<1 ] <= Htarg )
@@ -61,10 +61,10 @@ int scanhash_blake2b_8way( struct work *work, uint32_t max_nonce,
 // Function not used, code inlined.
 void blake2b_4way_hash(void *output, const void *input)
 {
-    blake2b_4way_ctx ctx;
+    blake2b_4x64_ctx ctx;
-    blake2b_4way_init( &ctx );
+    blake2b_4x64_init( &ctx );
-    blake2b_4way_update( &ctx, input, 80 );
+    blake2b_4x64_update( &ctx, input, 80 );
-    blake2b_4way_final( &ctx, output );
+    blake2b_4x64_final( &ctx, output );
 }
 
 int scanhash_blake2b_4way( struct work *work, uint32_t max_nonce,
@@ -73,7 +73,7 @@ int scanhash_blake2b_4way( struct work *work, uint32_t max_nonce,
 	uint32_t hash[8*4] __attribute__ ((aligned (64)));;
    uint32_t vdata[20*4] __attribute__ ((aligned (32)));;
    uint32_t lane_hash[8] __attribute__ ((aligned (32)));
-   blake2b_4way_ctx ctx __attribute__ ((aligned (32)));
+   blake2b_4x64_ctx ctx __attribute__ ((aligned (32)));
    uint32_t *hash7 = &(hash[25]);   // 3*8+1
 	uint32_t *pdata = work->data;
 	uint32_t *ptarget = work->target;
@@ -90,9 +90,9 @@ int scanhash_blake2b_4way( struct work *work, uint32_t max_nonce,
       *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
                 _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
 
-      blake2b_4way_init( &ctx ); 
+      blake2b_4x64_init( &ctx ); 
-      blake2b_4way_update( &ctx, vdata, 80 );
+      blake2b_4x64_update( &ctx, vdata, 80 );
-      blake2b_4way_final( &ctx, hash );
+      blake2b_4x64_final( &ctx, hash );
 
       for ( int lane = 0; lane < 4; lane++ )
       if ( hash7[ lane<<1 ] <= Htarg )

algo/blake/blake2s-hash.c

@@ -497,7 +497,7 @@ int blake2s_8way_full_blocks( blake2s_8way_state *S, void *out,
 
 #endif // __AVX2__
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 // Blake2s-256 16 way
 

algo/blake/blake2s-hash.h

@@ -11,8 +11,8 @@
  * this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
  */
 //#pragma once
-#ifndef __BLAKE2S_HASH_4WAY_H__
+#ifndef BLAKE2S_HASH_4WAY_H__
-#define __BLAKE2S_HASH_4WAY_H__ 1
+#define BLAKE2S_HASH_4WAY_H__ 1
 
 #if defined(__SSE2__) || defined(__ARM_NEON)
 
@@ -61,13 +61,18 @@ int blake2s_4way_final( blake2s_4way_state *S, void *out, uint8_t outlen );
 int blake2s_4way_full_blocks( blake2s_4way_state *S, void *out,
                               const void *input, uint64_t inlen );
 
+#define blake2s_4x32_state       blake2s_4way_state
+#define blake2s_4x32_init        blake2s_4way_init
+#define blake2s_4x32_update      blake2s_4way_update
+#define blake2s_4x32_final       blake2s_4way_final
+#define blake2s_4x32_full_blocks blake2s_4way_full_blocks
 
 #if defined(__AVX2__)
 
 typedef struct ALIGN( 64 ) __blake2s_8way_state
 {
    __m256i h[8];
-   uint8_t  buf[ 32 * 8 ];
+   uint8_t  buf[ 64 * 8 ];
    uint32_t t[2];
    uint32_t f[2];
    size_t   buflen;
@@ -81,14 +86,20 @@ int blake2s_8way_final( blake2s_8way_state *S, void *out, uint8_t outlen );
 int blake2s_8way_full_blocks( blake2s_8way_state *S, void *out,
                               const void *input, uint64_t inlen );
 
+#define blake2s_8x32_state       blake2s_8way_state
+#define blake2s_8x32_init        blake2s_8way_init
+#define blake2s_8x32_update      blake2s_8way_update
+#define blake2s_8x32_final       blake2s_8way_final
+#define blake2s_8x32_full_blocks blake2s_8way_full_blocks
+
 #endif
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 typedef struct ALIGN( 64 ) __blake2s_16way_state
 {
    __m512i h[8];
-   uint8_t  buf[ 32 * 16 ];
+   uint8_t  buf[ 64 * 16 ];
    uint32_t t[2];
    uint32_t f[2];
    size_t   buflen;
@@ -100,6 +111,11 @@ int blake2s_16way_update( blake2s_16way_state *S, const void *in,
                          uint64_t inlen );
 int blake2s_16way_final( blake2s_16way_state *S, void *out, uint8_t outlen );
 
+#define blake2s_16x32_state       blake2s_16way_state
+#define blake2s_16x32_init        blake2s_16way_init
+#define blake2s_16x32_update      blake2s_16way_update
+#define blake2s_16x32_final       blake2s_16way_final
+
 #endif
 
 #if 0

algo/blake/blake2s.c

@@ -3,7 +3,7 @@
 #include <string.h>
 #include <stdint.h>
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define BLAKE2S_16WAY
 #elif defined(__AVX2__)
   #define BLAKE2S_8WAY

algo/blake/blake512-hash.c

@@ -349,16 +349,16 @@ void blake512_transform( uint64_t *H, const uint64_t *buf, const uint64_t T0,
    Va = v128_add64( Va, v128_add64( Vb, \
                             v128_set64( CBx( r, Sd ) ^ Mx( r, Sc ), \
                                         CBx( r, Sb ) ^ Mx( r, Sa ) ) ) ); \
-   Vd = v128_ror64( v128_xor( Vd, Va ), 32 ); \
+   Vd = v128_ror64xor( Vd, Va, 32 ); \
    Vc = v128_add64( Vc, Vd ); \
-   Vb = v128_ror64( v128_xor( Vb, Vc ), 25 ); \
+   Vb = v128_ror64xor( Vb, Vc, 25 ); \
 \
    Va = v128_add64( Va, v128_add64( Vb, \
                             v128_set64( CBx( r, Sc ) ^ Mx( r, Sd ), \
                                         CBx( r, Sa ) ^ Mx( r, Sb ) ) ) ); \
-   Vd = v128_ror64( v128_xor( Vd, Va ), 16 ); \
+   Vd = v128_ror64xor( Vd, Va, 16 ); \
    Vc = v128_add64( Vc, Vd ); \
-   Vb = v128_ror64( v128_xor( Vb, Vc ), 11 ); \
+   Vb = v128_ror64xor( Vb, Vc, 11 ); \
 }
 
 #define BLAKE512_ROUND( R ) \
@@ -559,7 +559,7 @@ void blake512_full( blake512_context *sc, void *dst, const void *data,
 
 #if defined(__AVX2__)
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 ////////////////////////////////////
 //
@@ -617,24 +617,22 @@ void blake512_full( blake512_context *sc, void *dst, const void *data,
   VD = v512_64( CB5 ^ T0 ); \
   VE = v512_64( CB6 ^ T1 ); \
   VF = v512_64( CB7 ^ T1 ); \
-  const __m512i shuf_bswap64 = mm512_bcast_m128( v128_set64( \
+  M0 = mm512_bswap_64( *(buf+ 0) ); \
-                                   0x08090a0b0c0d0e0f, 0x0001020304050607 ) ); \
+  M1 = mm512_bswap_64( *(buf+ 1) ); \
-  M0 = _mm512_shuffle_epi8( *(buf+ 0), shuf_bswap64 ); \
+  M2 = mm512_bswap_64( *(buf+ 2) ); \
-  M1 = _mm512_shuffle_epi8( *(buf+ 1), shuf_bswap64 ); \
+  M3 = mm512_bswap_64( *(buf+ 3) ); \
-  M2 = _mm512_shuffle_epi8( *(buf+ 2), shuf_bswap64 ); \
+  M4 = mm512_bswap_64( *(buf+ 4) ); \
-  M3 = _mm512_shuffle_epi8( *(buf+ 3), shuf_bswap64 ); \
+  M5 = mm512_bswap_64( *(buf+ 5) ); \
-  M4 = _mm512_shuffle_epi8( *(buf+ 4), shuf_bswap64 ); \
+  M6 = mm512_bswap_64( *(buf+ 6) ); \
-  M5 = _mm512_shuffle_epi8( *(buf+ 5), shuf_bswap64 ); \
+  M7 = mm512_bswap_64( *(buf+ 7) ); \
-  M6 = _mm512_shuffle_epi8( *(buf+ 6), shuf_bswap64 ); \
+  M8 = mm512_bswap_64( *(buf+ 8) ); \
-  M7 = _mm512_shuffle_epi8( *(buf+ 7), shuf_bswap64 ); \
+  M9 = mm512_bswap_64( *(buf+ 9) ); \
-  M8 = _mm512_shuffle_epi8( *(buf+ 8), shuf_bswap64 ); \
+  MA = mm512_bswap_64( *(buf+10) ); \
-  M9 = _mm512_shuffle_epi8( *(buf+ 9), shuf_bswap64 ); \
+  MB = mm512_bswap_64( *(buf+11) ); \
-  MA = _mm512_shuffle_epi8( *(buf+10), shuf_bswap64 ); \
+  MC = mm512_bswap_64( *(buf+12) ); \
-  MB = _mm512_shuffle_epi8( *(buf+11), shuf_bswap64 ); \
+  MD = mm512_bswap_64( *(buf+13) ); \
-  MC = _mm512_shuffle_epi8( *(buf+12), shuf_bswap64 ); \
+  ME = mm512_bswap_64( *(buf+14) ); \
-  MD = _mm512_shuffle_epi8( *(buf+13), shuf_bswap64 ); \
+  MF = mm512_bswap_64( *(buf+15) ); \
-  ME = _mm512_shuffle_epi8( *(buf+14), shuf_bswap64 ); \
-  MF = _mm512_shuffle_epi8( *(buf+15), shuf_bswap64 ); \
   ROUND_B_8WAY(0); \
   ROUND_B_8WAY(1); \
   ROUND_B_8WAY(2); \
@@ -661,7 +659,7 @@ void blake512_full( blake512_context *sc, void *dst, const void *data,
   H7 = mm512_xor3( VF, V7, H7 ); \
 }
 
-void blake512_8way_compress( blake_8way_big_context *sc )
+void blake512_8x64_compress( blake_8x64_big_context *sc )
 { 
   __m512i M0, M1, M2, M3, M4, M5, M6, M7;
   __m512i M8, M9, MA, MB, MC, MD, ME, MF;
@@ -685,25 +683,22 @@ void blake512_8way_compress( blake_8way_big_context *sc )
   VE = v512_64( CB6 ^ sc->T1 );
   VF = v512_64( CB7 ^ sc->T1 );
 
-  const __m512i shuf_bswap64 = mm512_bcast_m128( v128_set64( 
+  M0 = mm512_bswap_64( sc->buf[ 0] );
-                                   0x08090a0b0c0d0e0f, 0x0001020304050607 ) );
+  M1 = mm512_bswap_64( sc->buf[ 1] );
-
+  M2 = mm512_bswap_64( sc->buf[ 2] );
-  M0 = _mm512_shuffle_epi8( sc->buf[ 0], shuf_bswap64 );
+  M3 = mm512_bswap_64( sc->buf[ 3] );
-  M1 = _mm512_shuffle_epi8( sc->buf[ 1], shuf_bswap64 );
+  M4 = mm512_bswap_64( sc->buf[ 4] );
-  M2 = _mm512_shuffle_epi8( sc->buf[ 2], shuf_bswap64 );
+  M5 = mm512_bswap_64( sc->buf[ 5] );
-  M3 = _mm512_shuffle_epi8( sc->buf[ 3], shuf_bswap64 );
+  M6 = mm512_bswap_64( sc->buf[ 6] );
-  M4 = _mm512_shuffle_epi8( sc->buf[ 4], shuf_bswap64 );
+  M7 = mm512_bswap_64( sc->buf[ 7] );
-  M5 = _mm512_shuffle_epi8( sc->buf[ 5], shuf_bswap64 );
+  M8 = mm512_bswap_64( sc->buf[ 8] );
-  M6 = _mm512_shuffle_epi8( sc->buf[ 6], shuf_bswap64 );
+  M9 = mm512_bswap_64( sc->buf[ 9] );
-  M7 = _mm512_shuffle_epi8( sc->buf[ 7], shuf_bswap64 );
+  MA = mm512_bswap_64( sc->buf[10] );
-  M8 = _mm512_shuffle_epi8( sc->buf[ 8], shuf_bswap64 );
+  MB = mm512_bswap_64( sc->buf[11] );
-  M9 = _mm512_shuffle_epi8( sc->buf[ 9], shuf_bswap64 );
+  MC = mm512_bswap_64( sc->buf[12] );
-  MA = _mm512_shuffle_epi8( sc->buf[10], shuf_bswap64 );
+  MD = mm512_bswap_64( sc->buf[13] );
-  MB = _mm512_shuffle_epi8( sc->buf[11], shuf_bswap64 );
+  ME = mm512_bswap_64( sc->buf[14] );
-  MC = _mm512_shuffle_epi8( sc->buf[12], shuf_bswap64 );
+  MF = mm512_bswap_64( sc->buf[15] );
-  MD = _mm512_shuffle_epi8( sc->buf[13], shuf_bswap64 );
-  ME = _mm512_shuffle_epi8( sc->buf[14], shuf_bswap64 );
-  MF = _mm512_shuffle_epi8( sc->buf[15], shuf_bswap64 );
 
   ROUND_B_8WAY(0);
   ROUND_B_8WAY(1);
@@ -733,7 +728,7 @@ void blake512_8way_compress( blake_8way_big_context *sc )
 }
 
 // won't be used after prehash implemented
-void blake512_8way_compress_le( blake_8x64_big_context *sc )
+void blake512_8x64_compress_le( blake_8x64_big_context *sc )
 {
   __m512i M0, M1, M2, M3, M4, M5, M6, M7;
   __m512i M8, M9, MA, MB, MC, MD, ME, MF;
@@ -1177,7 +1172,7 @@ void blake512_8x64_full( blake_8x64_big_context *sc, void * dst,
    {
       if ( ( sc->T0 = sc->T0 + 1024 ) < 1024 )
             sc->T1 = sc->T1 + 1;
-      blake512_8way_compress( sc );
+      blake512_8x64_compress( sc );
       sc->ptr = 0;
    }
 
@@ -1213,7 +1208,7 @@ void blake512_8x64_full( blake_8x64_big_context *sc, void * dst,
    if ( ( sc->T0 = sc->T0 + 1024 ) < 1024 )
        sc->T1 = sc->T1 + 1;
 
-   blake512_8way_compress( sc );
+   blake512_8x64_compress( sc );
    
    mm512_block_bswap_64( (__m512i*)dst, sc->H );
 }
@@ -1244,7 +1239,7 @@ void blake512_8x64_full_le( blake_8x64_big_context *sc, void * dst,
    {
       if ( ( sc->T0 = sc->T0 + 1024 ) < 1024 )
             sc->T1 = sc->T1 + 1;
-      blake512_8way_compress_le( sc );
+      blake512_8x64_compress_le( sc );
       sc->ptr = 0;
    }
 
@@ -1280,7 +1275,7 @@ void blake512_8x64_full_le( blake_8x64_big_context *sc, void * dst,
    if ( ( sc->T0 = sc->T0 + 1024 ) < 1024 )
        sc->T1 = sc->T1 + 1;
 
-   blake512_8way_compress_le( sc );
+   blake512_8x64_compress_le( sc );
 
    mm512_block_bswap_64( (__m512i*)dst, sc->H );
 }
@@ -1355,24 +1350,22 @@ blake512_8x64_close(void *cc, void *dst)
   VD = v256_64( CB5 ^ T0 ); \
   VE = v256_64( CB6 ^ T1 ); \
   VF = v256_64( CB7 ^ T1 ); \
-  const __m256i shuf_bswap64 = mm256_bcast_m128( v128_set64( \
+  M0 = mm256_bswap_64( *(buf+ 0) ); \
-                             0x08090a0b0c0d0e0f, 0x0001020304050607 ) ); \
+  M1 = mm256_bswap_64( *(buf+ 1) ); \
-  M0 = _mm256_shuffle_epi8( *(buf+ 0), shuf_bswap64 ); \
+  M2 = mm256_bswap_64( *(buf+ 2) ); \
-  M1 = _mm256_shuffle_epi8( *(buf+ 1), shuf_bswap64 ); \
+  M3 = mm256_bswap_64( *(buf+ 3) ); \
-  M2 = _mm256_shuffle_epi8( *(buf+ 2), shuf_bswap64 ); \
+  M4 = mm256_bswap_64( *(buf+ 4) ); \
-  M3 = _mm256_shuffle_epi8( *(buf+ 3), shuf_bswap64 ); \
+  M5 = mm256_bswap_64( *(buf+ 5) ); \
-  M4 = _mm256_shuffle_epi8( *(buf+ 4), shuf_bswap64 ); \
+  M6 = mm256_bswap_64( *(buf+ 6) ); \
-  M5 = _mm256_shuffle_epi8( *(buf+ 5), shuf_bswap64 ); \
+  M7 = mm256_bswap_64( *(buf+ 7) ); \
-  M6 = _mm256_shuffle_epi8( *(buf+ 6), shuf_bswap64 ); \
+  M8 = mm256_bswap_64( *(buf+ 8) ); \
-  M7 = _mm256_shuffle_epi8( *(buf+ 7), shuf_bswap64 ); \
+  M9 = mm256_bswap_64( *(buf+ 9) ); \
-  M8 = _mm256_shuffle_epi8( *(buf+ 8), shuf_bswap64 ); \
+  MA = mm256_bswap_64( *(buf+10) ); \
-  M9 = _mm256_shuffle_epi8( *(buf+ 9), shuf_bswap64 ); \
+  MB = mm256_bswap_64( *(buf+11) ); \
-  MA = _mm256_shuffle_epi8( *(buf+10), shuf_bswap64 ); \
+  MC = mm256_bswap_64( *(buf+12) ); \
-  MB = _mm256_shuffle_epi8( *(buf+11), shuf_bswap64 ); \
+  MD = mm256_bswap_64( *(buf+13) ); \
-  MC = _mm256_shuffle_epi8( *(buf+12), shuf_bswap64 ); \
+  ME = mm256_bswap_64( *(buf+14) ); \
-  MD = _mm256_shuffle_epi8( *(buf+13), shuf_bswap64 ); \
+  MF = mm256_bswap_64( *(buf+15) ); \
-  ME = _mm256_shuffle_epi8( *(buf+14), shuf_bswap64 ); \
-  MF = _mm256_shuffle_epi8( *(buf+15), shuf_bswap64 ); \
   ROUND_B_4WAY(0); \
   ROUND_B_4WAY(1); \
   ROUND_B_4WAY(2); \
@@ -1400,7 +1393,7 @@ blake512_8x64_close(void *cc, void *dst)
 }
 
 
-void blake512_4way_compress( blake_4x64_big_context *sc )
+void blake512_4x64_compress( blake_4x64_big_context *sc )
 {
   __m256i M0, M1, M2, M3, M4, M5, M6, M7;
   __m256i M8, M9, MA, MB, MC, MD, ME, MF;
@@ -1423,25 +1416,23 @@ void blake512_4way_compress( blake_4x64_big_context *sc )
   VD = v256_64( CB5 ^ sc->T0 );
   VE = v256_64( CB6 ^ sc->T1 );
   VF = v256_64( CB7 ^ sc->T1 );
-  const __m256i shuf_bswap64 = mm256_bcast_m128( v128_set64(
-                                    0x08090a0b0c0d0e0f, 0x0001020304050607 ) );
 
-  M0 = _mm256_shuffle_epi8( sc->buf[ 0], shuf_bswap64 );
+  M0 = mm256_bswap_64( sc->buf[ 0] );
-  M1 = _mm256_shuffle_epi8( sc->buf[ 1], shuf_bswap64 );
+  M1 = mm256_bswap_64( sc->buf[ 1] );
-  M2 = _mm256_shuffle_epi8( sc->buf[ 2], shuf_bswap64 );
+  M2 = mm256_bswap_64( sc->buf[ 2] );
-  M3 = _mm256_shuffle_epi8( sc->buf[ 3], shuf_bswap64 );
+  M3 = mm256_bswap_64( sc->buf[ 3] );
-  M4 = _mm256_shuffle_epi8( sc->buf[ 4], shuf_bswap64 );
+  M4 = mm256_bswap_64( sc->buf[ 4] );
-  M5 = _mm256_shuffle_epi8( sc->buf[ 5], shuf_bswap64 );
+  M5 = mm256_bswap_64( sc->buf[ 5] );
-  M6 = _mm256_shuffle_epi8( sc->buf[ 6], shuf_bswap64 );
+  M6 = mm256_bswap_64( sc->buf[ 6] );
-  M7 = _mm256_shuffle_epi8( sc->buf[ 7], shuf_bswap64 );
+  M7 = mm256_bswap_64( sc->buf[ 7] );
-  M8 = _mm256_shuffle_epi8( sc->buf[ 8], shuf_bswap64 );
+  M8 = mm256_bswap_64( sc->buf[ 8] );
-  M9 = _mm256_shuffle_epi8( sc->buf[ 9], shuf_bswap64 );
+  M9 = mm256_bswap_64( sc->buf[ 9] );
-  MA = _mm256_shuffle_epi8( sc->buf[10], shuf_bswap64 );
+  MA = mm256_bswap_64( sc->buf[10] );
-  MB = _mm256_shuffle_epi8( sc->buf[11], shuf_bswap64 );
+  MB = mm256_bswap_64( sc->buf[11] );
-  MC = _mm256_shuffle_epi8( sc->buf[12], shuf_bswap64 );
+  MC = mm256_bswap_64( sc->buf[12] );
-  MD = _mm256_shuffle_epi8( sc->buf[13], shuf_bswap64 );
+  MD = mm256_bswap_64( sc->buf[13] );
-  ME = _mm256_shuffle_epi8( sc->buf[14], shuf_bswap64 );
+  ME = mm256_bswap_64( sc->buf[14] );
-  MF = _mm256_shuffle_epi8( sc->buf[15], shuf_bswap64 );
+  MF = mm256_bswap_64( sc->buf[15] );
 
   ROUND_B_4WAY(0);
   ROUND_B_4WAY(1);
@@ -1470,7 +1461,7 @@ void blake512_4way_compress( blake_4x64_big_context *sc )
   sc->H[7] = mm256_xor3( VF, V7, sc->H[7] );
 }
 
-void blake512_4x64_prehash_le( blake_4x64_big_context *sc, __m256i *midstate,
+void blake512_4x64_prehash_le( blake512_4x64_context *sc, __m256i *midstate,
                                const void *data )
 {
    __m256i V0, V1, V2, V3, V4, V5, V6, V7;
@@ -1562,7 +1553,7 @@ void blake512_4x64_prehash_le( blake_4x64_big_context *sc, __m256i *midstate,
    midstate[15] = VF;
 }
 
-void blake512_4x64_final_le( blake_4x64_big_context *sc, void *hash,
+void blake512_4x64_final_le( blake512_4x64_context *sc, void *hash,
                              const __m256i nonce, const __m256i *midstate )
 {
    __m256i M0, M1, M2, M3, M4, M5, M6, M7;
@@ -1685,7 +1676,7 @@ void blake512_4x64_final_le( blake_4x64_big_context *sc, void *hash,
 }
 
 
-void blake512_4x64_init( blake_4x64_big_context *sc )
+void blake512_4x64_init( blake512_4x64_context *sc )
 {
    casti_m256i( sc->H, 0 ) = v256_64( 0x6A09E667F3BCC908 );
    casti_m256i( sc->H, 1 ) = v256_64( 0xBB67AE8584CAA73B );
@@ -1798,7 +1789,7 @@ blake64_4way_close( blake_4x64_big_context *sc, void *dst )
 }
 
 // init, update & close
-void blake512_4x64_full( blake_4x64_big_context *sc, void * dst,
+void blake512_4x64_full( blake512_4x64_context *sc, void * dst,
                          const void *data, size_t len )
 {
 
@@ -1824,7 +1815,7 @@ void blake512_4x64_full( blake_4x64_big_context *sc, void * dst,
    {
       if ( ( sc->T0 = sc->T0 + 1024 ) < 1024 )
          sc->T1 =  sc->T1 + 1;
-      blake512_4way_compress( sc );
+      blake512_4x64_compress( sc );
       sc->ptr = 0;
    }
 
@@ -1859,7 +1850,7 @@ void blake512_4x64_full( blake_4x64_big_context *sc, void * dst,
    if ( ( sc->T0 = sc->T0 + 1024 ) < 1024 )
        sc->T1 = sc->T1 + 1;
 
-   blake512_4way_compress( sc );
+   blake512_4x64_compress( sc );
 
    mm256_block_bswap_64( (__m256i*)dst, sc->H );
 }
@@ -1887,13 +1878,13 @@ blake512_4x64_close(void *cc, void *dst)
 #define GB_2X64( m0, m1, c0, c1, a, b, c, d ) \
 { \
    a = v128_add64( v128_add64( v128_xor( v128_64( c1 ), m0 ), b ), a ); \
-   d = v128_ror64( v128_xor( d, a ), 32 ); \
+   d = v128_ror64xor( d, a, 32 ); \
    c = v128_add64( c, d ); \
-   b = v128_ror64( v128_xor( b, c ), 25 ); \
+   b = v128_ror64xor( b, c, 25 ); \
    a = v128_add64( v128_add64( v128_xor( v128_64( c0 ), m1 ), b ), a ); \
-   d = v128_ror64( v128_xor( d, a ), 16 ); \
+   d = v128_ror64xor( d, a, 16 ); \
    c = v128_add64( c, d ); \
-   b = v128_ror64( v128_xor( b, c ), 11 ); \
+   b = v128_ror64xor( b, c, 11 ); \
 }
 
 #define ROUND_B_2X64(r) \
@@ -1934,29 +1925,6 @@ void blake512_2x64_compress( blake_2x64_big_context *sc )
   VE = v128_64( CB6 ^ sc->T1 );
   VF = v128_64( CB7 ^ sc->T1 );
 
-#if defined(__SSSE3__)
-
-  const v128u64_t shuf_bswap64 = v128_set64(
-                                 0x08090a0b0c0d0e0f, 0x0001020304050607 );
-  M0 = v128_shuffle8( sc->buf[ 0], shuf_bswap64 );
-  M1 = v128_shuffle8( sc->buf[ 1], shuf_bswap64 );
-  M2 = v128_shuffle8( sc->buf[ 2], shuf_bswap64 );
-  M3 = v128_shuffle8( sc->buf[ 3], shuf_bswap64 );
-  M4 = v128_shuffle8( sc->buf[ 4], shuf_bswap64 );
-  M5 = v128_shuffle8( sc->buf[ 5], shuf_bswap64 );
-  M6 = v128_shuffle8( sc->buf[ 6], shuf_bswap64 );
-  M7 = v128_shuffle8( sc->buf[ 7], shuf_bswap64 );
-  M8 = v128_shuffle8( sc->buf[ 8], shuf_bswap64 );
-  M9 = v128_shuffle8( sc->buf[ 9], shuf_bswap64 );
-  MA = v128_shuffle8( sc->buf[10], shuf_bswap64 );
-  MB = v128_shuffle8( sc->buf[11], shuf_bswap64 );
-  MC = v128_shuffle8( sc->buf[12], shuf_bswap64 );
-  MD = v128_shuffle8( sc->buf[13], shuf_bswap64 );
-  ME = v128_shuffle8( sc->buf[14], shuf_bswap64 );
-  MF = v128_shuffle8( sc->buf[15], shuf_bswap64 );
-
-#else  // SSE2 & NEON
-
   M0 = v128_bswap64( sc->buf[ 0] );
   M1 = v128_bswap64( sc->buf[ 1] );
   M2 = v128_bswap64( sc->buf[ 2] );
@@ -1974,8 +1942,6 @@ void blake512_2x64_compress( blake_2x64_big_context *sc )
   ME = v128_bswap64( sc->buf[14] );
   MF = v128_bswap64( sc->buf[15] );
   
-#endif
-
   ROUND_B_2X64(0);
   ROUND_B_2X64(1);
   ROUND_B_2X64(2);
@@ -2054,9 +2020,9 @@ void blake512_2x64_prehash_part1_le( blake_2x64_big_context *sc,
    // G4 skip nonce
    V0 = v128_add64( v128_add64( v128_xor( v128_64( CB9 ), sc->buf[ 8] ), V5 ),
                                           V0 );
-   VF = v128_ror64( v128_xor( VF, V0 ), 32 );
+   VF = v128_ror64xor( VF, V0, 32 );
    VA = v128_add64( VA, VF );
-   V5 = v128_ror64( v128_xor( V5, VA ), 25 );
+   V5 = v128_ror64xor( V5, VA, 25 );
    V0 = v128_add64( V0, V5 );
 
    GB_2X64( sc->buf[10], sc->buf[11], CBA, CBB, V1, V6, VB, VC );
@@ -2137,9 +2103,9 @@ void blake512_2x64_prehash_part2_le( blake_2x64_big_context *sc, void *hash,
 
    // finish round 0, with the nonce now available 
    V0 = v128_add64( V0, v128_xor( v128_64( CB8 ), M9 ) );
-   VF = v128_ror64( v128_xor( VF, V0 ), 16 );
+   VF = v128_ror64xor( VF, V0, 16 );
    VA = v128_add64( VA, VF );
-   V5 = v128_ror64( v128_xor( V5, VA ), 11 );
+   V5 = v128_ror64xor( V5, VA, 11 );
 
    // Round 1
    // G0
@@ -2147,34 +2113,34 @@ void blake512_2x64_prehash_part2_le( blake_2x64_big_context *sc, void *hash,
 
    // G1
    V1 = v128_add64( V1, V5 );
-   VD = v128_ror64( v128_xor( VD, V1 ), 32 );
+   VD = v128_ror64xor( VD, V1, 32 );
    V9 = v128_add64( V9, VD );
-   V5 = v128_ror64( v128_xor( V5, V9 ), 25 );
+   V5 = v128_ror64xor( V5, V9, 25 );
    V1 = v128_add64( V1, v128_add64( v128_xor( v128_64( CBx(1,2) ), Mx(1,3) ),
                                               V5 ) );
-   VD = v128_ror64( v128_xor( VD, V1 ), 16 );
+   VD = v128_ror64xor( VD, V1, 16 );
    V9 = v128_add64( V9, VD );
-   V5 = v128_ror64( v128_xor( V5, V9 ), 11 );
+   V5 = v128_ror64xor( V5, V9, 11 );
 
    // G2
    V2 = v128_add64( V2, v128_xor( v128_64( CBF ), M9 ) );
-   VE = v128_ror64( v128_xor( VE, V2 ), 32 );
+   VE = v128_ror64xor( VE, V2, 32 );
    VA = v128_add64( VA, VE );
-   V6 = v128_ror64( v128_xor( V6, VA ), 25 );
+   V6 = v128_ror64xor( V6, VA, 25 );
    V2 = v128_add64( V2, v128_add64( v128_xor( v128_64( CB9 ), MF ), V6 ) );
-   VE = v128_ror64( v128_xor( VE, V2 ), 16 );
+   VE = v128_ror64xor( VE, V2, 16 );
    VA = v128_add64( VA, VE );
-   V6 = v128_ror64( v128_xor( V6, VA ), 11 );
+   V6 = v128_ror64xor( V6, VA, 11 );
 
    // G3
-   VF = v128_ror64( v128_xor( VF, V3 ), 32 );
+   VF = v128_ror64xor( VF, V3, 32 );
    VB = v128_add64( VB, VF );
-   V7 = v128_ror64( v128_xor( V7, VB ), 25 );
+   V7 = v128_ror64xor( V7, VB, 25 );
    V3 = v128_add64( V3, v128_add64( v128_xor( v128_64( CBx(1, 6) ), Mx(1, 7) ),
                                               V7 ) );
-   VF = v128_ror64( v128_xor( VF, V3 ), 16 );
+   VF = v128_ror64xor( VF, V3, 16 );
    VB = v128_add64( VB, VF );
-   V7 = v128_ror64( v128_xor( V7, VB ), 11 );
+   V7 = v128_ror64xor( V7, VB, 11 );
 
    // G4, G5, G6, G7
    GB_2X64(Mx(1, 8), Mx(1, 9), CBx(1, 8), CBx(1, 9), V0, V5, VA, VF);

algo/blake/blake512-hash.h

@@ -92,7 +92,7 @@ void blake512_4x64_final_le( blake_4x64_big_context *sc, void *hash,
 #define blake512_4way_prehash_le  blake512_4x64_prehash_le
 #define blake512_4way_final_le    blake512_4x64_final_le
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 ////////////////////////////
 //

algo/blake/blakecoin-4way.c

@@ -54,10 +54,10 @@ int scanhash_blakecoin_16way( struct work *work, uint32_t max_nonce,
                                n+ 7, n+ 6, n+ 5, n+ 4, n+ 3, n+ 2, n +1, n );
 
    // Partialy prehash second block without touching nonces in block_buf[3].
-   blake256_16way_round0_prehash_le( midstate_vars, block0_hash, block_buf );
+   blake256_16x32_round0_prehash_le( midstate_vars, block0_hash, block_buf );
 
    do {
-      blake256_16way_final_rounds_le( hash32, midstate_vars, block0_hash,
+      blake256_16x32_final_rounds_le( hash32, midstate_vars, block0_hash,
                                       block_buf, rounds );
       for ( int lane = 0; lane < 16; lane++ )
       if ( unlikely( hash32_d7[ lane ] <= targ32_d7 ) )
@@ -123,10 +123,10 @@ int scanhash_blakecoin_8way( struct work *work, uint32_t max_nonce,
    block_buf[3] = _mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n );
 
    // Partialy prehash second block without touching nonces in block_buf[3].
-   blake256_8way_round0_prehash_le( midstate_vars, block0_hash, block_buf );
+   blake256_8x32_round0_prehash_le( midstate_vars, block0_hash, block_buf );
 
    do {
-      blake256_8way_final_rounds_le( hash32, midstate_vars, block0_hash,
+      blake256_8x32_final_rounds_le( hash32, midstate_vars, block0_hash,
                                      block_buf, rounds );
       for ( int lane = 0; lane < 8; lane++ )
       if ( unlikely( hash32_d7[ lane ] <= targ32_d7 ) )
@@ -148,16 +148,16 @@ int scanhash_blakecoin_8way( struct work *work, uint32_t max_nonce,
    
 #elif defined (BLAKECOIN_4WAY)
 
-blake256r8_4way_context blakecoin_4w_ctx;
+blake256r8_4x32_context blakecoin_4w_ctx;
 
 void blakecoin_4way_hash(void *state, const void *input)
 {
      uint32_t vhash[8*4] __attribute__ ((aligned (64)));
-     blake256r8_4way_context ctx;
+     blake256r8_4x32_context ctx;
 
      memcpy( &ctx, &blakecoin_4w_ctx, sizeof ctx );
-     blake256r8_4way_update( &ctx, input + (64<<2), 16 );
+     blake256r8_4x32_update( &ctx, input + (64<<2), 16 );
-     blake256r8_4way_close( &ctx, vhash );
+     blake256r8_4x32_close( &ctx, vhash );
 
      dintrlv_4x32( state, state+32, state+64, state+96, vhash, 256 );
 }
@@ -178,8 +178,8 @@ int scanhash_blakecoin_4way( struct work *work, uint32_t max_nonce,
       HTarget = 0x7f;
 
    v128_bswap32_intrlv80_4x32( vdata, pdata );
-   blake256r8_4way_init( &blakecoin_4w_ctx );
+   blake256r8_4x32_init( &blakecoin_4w_ctx );
-   blake256r8_4way_update( &blakecoin_4w_ctx, vdata, 64 );
+   blake256r8_4x32_update( &blakecoin_4w_ctx, vdata, 64 );
 
    do {
       *noncev = v128_bswap32( _mm_set_epi32( n+3, n+2, n+1, n ) );

algo/blake/blakecoin-gate.h

@@ -4,7 +4,7 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define BLAKECOIN_16WAY
 #elif defined(__AVX2__)
   #define BLAKECOIN_8WAY

algo/blake/pentablake-4way.c

@@ -16,28 +16,27 @@ extern void pentablakehash_4way( 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)));
-     blake512_4way_context ctx;
+     blake512_4x64_context ctx;
 
+     blake512_4x64_init( &ctx );
+     blake512_4x64_update( &ctx, input, 80 );
+     blake512_4x64_close( &ctx, vhash );
 
-     blake512_4way_init( &ctx );
+     blake512_4x64_init( &ctx );
-     blake512_4way_update( &ctx, input, 80 );
+     blake512_4x64_update( &ctx, vhash, 64 );
-     blake512_4way_close( &ctx, vhash );
+     blake512_4x64_close( &ctx, vhash );
 
-     blake512_4way_init( &ctx );
+     blake512_4x64_init( &ctx );
-     blake512_4way_update( &ctx, vhash, 64 );
+     blake512_4x64_update( &ctx, vhash, 64 );
-     blake512_4way_close( &ctx, vhash );
+     blake512_4x64_close( &ctx, vhash );
 
-     blake512_4way_init( &ctx );
+     blake512_4x64_init( &ctx );
-     blake512_4way_update( &ctx, vhash, 64 );
+     blake512_4x64_update( &ctx, vhash, 64 );
-     blake512_4way_close( &ctx, vhash );
+     blake512_4x64_close( &ctx, vhash );
 
-     blake512_4way_init( &ctx );
+     blake512_4x64_init( &ctx );
-     blake512_4way_update( &ctx, vhash, 64 );
+     blake512_4x64_update( &ctx, vhash, 64 );
-     blake512_4way_close( &ctx, vhash );
+     blake512_4x64_close( &ctx, vhash );
-
-     blake512_4way_init( &ctx );
-     blake512_4way_update( &ctx, vhash, 64 );
-     blake512_4way_close( &ctx, vhash );
 
      memcpy( output,    hash0, 32 );
      memcpy( output+32, hash1, 32 );

algo/blake/sph-blake2s.c

@@ -227,7 +227,7 @@ int blake2s_compress( blake2s_state *S, const uint8_t block[64] )
 	v[14] = S->f[0] ^ blake2s_IV[6];
 	v[15] = S->f[1] ^ blake2s_IV[7];
 
-#if defined(__SSE2__)
+#if defined(__SSE2__) || defined(__ARM_NEON)
 
    v128_t *V = (v128_t*)v;
 
@@ -263,19 +263,6 @@ int blake2s_compress( blake2s_state *S, const uint8_t block[64] )
    V[3] = v128_swap64( V[3] ); \
    V[2] = v128_shufll32( V[2] )
 
-   BLAKE2S_ROUND(0);
-   BLAKE2S_ROUND(1);
-   BLAKE2S_ROUND(2);
-   BLAKE2S_ROUND(3);
-   BLAKE2S_ROUND(4);
-   BLAKE2S_ROUND(5);
-   BLAKE2S_ROUND(6);
-   BLAKE2S_ROUND(7);
-   BLAKE2S_ROUND(8);
-   BLAKE2S_ROUND(9);
-   
-#undef BLAKE2S_ROUND
-
 #else
 
 #define G(r,i,a,b,c,d) \
@@ -290,7 +277,7 @@ int blake2s_compress( blake2s_state *S, const uint8_t block[64] )
 		b = SPH_ROTR32(b ^ c, 7); \
 	} while(0)
 
-#define ROUND(r)  \
+#define BLAKE2S_ROUND(r)  \
 	do { \
 		G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
 		G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
@@ -302,24 +289,25 @@ int blake2s_compress( blake2s_state *S, const uint8_t block[64] )
 		G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
 	} while(0)
 
-   ROUND( 0 );
-	ROUND( 1 );
-	ROUND( 2 );
-	ROUND( 3 );
-	ROUND( 4 );
-	ROUND( 5 );
-	ROUND( 6 );
-	ROUND( 7 );
-	ROUND( 8 );
-	ROUND( 9 );
-
 #endif
 
+   BLAKE2S_ROUND(0);
+   BLAKE2S_ROUND(1);
+   BLAKE2S_ROUND(2);
+   BLAKE2S_ROUND(3);
+   BLAKE2S_ROUND(4);
+   BLAKE2S_ROUND(5);
+   BLAKE2S_ROUND(6);
+   BLAKE2S_ROUND(7);
+   BLAKE2S_ROUND(8);
+   BLAKE2S_ROUND(9);
+   
+
 	for( size_t i = 0; i < 8; ++i )
 		S->h[i] = S->h[i] ^ v[i] ^ v[i + 8];
 
 #undef G
-#undef ROUND
+#undef BLAKE2S_ROUND
 	return 0;
 }
 

algo/blake/sph_blake2b.c

@@ -101,15 +101,15 @@
 { \
    Va = v128_add64( Va, v128_add64( Vb, \
                  v128_set64( m[ sigmaR[ Sc ] ], m[ sigmaR[ Sa ] ] ) ) ); \
-   Vd = v128_ror64( v128_xor( Vd, Va ), 32 ); \
+   Vd = v128_ror64xor( Vd, Va, 32 ); \
    Vc = v128_add64( Vc, Vd ); \
-   Vb = v128_ror64( v128_xor( Vb, Vc ), 24 ); \
+   Vb = v128_ror64xor( Vb, Vc, 24 ); \
 \
    Va = v128_add64( Va, v128_add64( Vb, \
                  v128_set64( m[ sigmaR[ Sd ] ], m[ sigmaR[ Sb ] ] ) ) ); \
-   Vd = v128_ror64( v128_xor( Vd, Va ), 16 ); \
+   Vd = v128_ror64xor( Vd, Va, 16 ); \
    Vc = v128_add64( Vc, Vd ); \
-   Vb = v128_ror64( v128_xor( Vb, Vc ), 63 ); \
+   Vb = v128_ror64xor( Vb, Vc, 63 ); \
 }
 
 #define BLAKE2B_ROUND( R ) \

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

@@ -39,16 +39,14 @@
 #include <stddef.h>
 #include "simd-utils.h"
 
-#define SPH_SIZE_bmw256   256
-
-#define SPH_SIZE_bmw512   512
-
 // BMW-256 4 way 32
 
+#if defined(__SSE2__) || defined(__ARM_NEON)
+
 typedef struct
 {
-   v128_t buf[64];
+   v128u32_t buf[64];
-   v128_t H[16];
+   v128u32_t H[16];
    size_t ptr;
    uint32_t bit_count;  // assume bit_count fits in 32 bits
 } bmw_4way_small_context;
@@ -58,13 +56,19 @@ typedef bmw_4way_small_context bmw256_4way_context;
 void bmw256_4way_init( bmw256_4way_context *ctx );
 
 void bmw256_4way_update(void *cc, const void *data, size_t len);
-#define bmw256_4way bmw256_4way_update
 
 void bmw256_4way_close(void *cc, void *dst);
 
 void bmw256_4way_addbits_and_close(
         void *cc, unsigned ub, unsigned n, void *dst);
 
+#define bmw256_4x32_context bmw256_4way_context
+#define bmw256_4x32_init    bmw256_4way_init
+#define bmw256_4x32_update  bmw256_4way_update
+#define bmw256_4x32_close   bmw256_4way_close
+
+#endif
+
 #if defined(__AVX2__)
 
 // BMW-256 8 way 32
@@ -85,9 +89,14 @@ void bmw256_8way_update( bmw256_8way_context *ctx, const void *data,
 #define bmw256_8way bmw256_8way_update
 void bmw256_8way_close( bmw256_8way_context *ctx, void *dst );
 
+#define bmw256_8x32_context bmw256_8way_context
+#define bmw256_8x32_init    bmw256_8way_init
+#define bmw256_8x32_update  bmw256_8way_update
+#define bmw256_8x32_close   bmw256_8way_close
+
 #endif
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 // BMW-256 16 way 32
 
@@ -106,6 +115,11 @@ void bmw256_16way_update( bmw256_16way_context *ctx, const void *data,
                           size_t len );
 void bmw256_16way_close( bmw256_16way_context *ctx, void *dst );
 
+#define bmw256_16x32_context bmw256_16way_context
+#define bmw256_16x32_init    bmw256_16way_init
+#define bmw256_16x32_update  bmw256_16way_update
+#define bmw256_16x32_close   bmw256_16way_close
+
 #endif
 
 // BMW-512 2 way 64
@@ -157,7 +171,7 @@ void bmw512_4way_addbits_and_close(
 
 #endif  // __AVX2__
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 // BMW-512 64 bit 8 way
 typedef struct

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

@@ -45,7 +45,7 @@ extern "C"{
 
 #define LPAR   (
 
-#if defined(__SSE2__)
+#if defined(__SSE2__) || defined(__ARM_NEON)
 
 // BMW-256 4 way 32
 /*
@@ -284,9 +284,9 @@ static const uint32_t IV256[] = {
                      v128_xor( M[13], H[13] ) ) )
 
 
-void compress_small( const v128u64_t *M, const v128u64_t H[16], v128u64_t dH[16] )
+void compress_small( const v128u32_t *M, const v128u32_t H[16], v128u32_t dH[16] )
 {
-   v128u64_t qt[32], xl, xh; \
+   v128u32_t qt[32], xl, xh; \
 
    qt[ 0] = v128_add32( ss0( Ws0 ), H[ 1] );
    qt[ 1] = v128_add32( ss1( Ws1 ), H[ 2] );
@@ -428,49 +428,25 @@ static const uint32_t final_s[16][4] =
    { 0xaaaaaaae, 0xaaaaaaae, 0xaaaaaaae, 0xaaaaaaae },
    { 0xaaaaaaaf, 0xaaaaaaaf, 0xaaaaaaaf, 0xaaaaaaaf }
 };
-/*
+
-static const v128u64_t final_s[16] =
-{
-   { 0xaaaaaaa0aaaaaaa0, 0xaaaaaaa0aaaaaaa0 },
-   { 0xaaaaaaa1aaaaaaa1, 0xaaaaaaa1aaaaaaa1 },
-   { 0xaaaaaaa2aaaaaaa2, 0xaaaaaaa2aaaaaaa2 },
-   { 0xaaaaaaa3aaaaaaa3, 0xaaaaaaa3aaaaaaa3 },
-   { 0xaaaaaaa4aaaaaaa4, 0xaaaaaaa4aaaaaaa4 },
-   { 0xaaaaaaa5aaaaaaa5, 0xaaaaaaa5aaaaaaa5 },
-   { 0xaaaaaaa6aaaaaaa6, 0xaaaaaaa6aaaaaaa6 },
-   { 0xaaaaaaa7aaaaaaa7, 0xaaaaaaa7aaaaaaa7 },
-   { 0xaaaaaaa8aaaaaaa8, 0xaaaaaaa8aaaaaaa8 },
-   { 0xaaaaaaa9aaaaaaa9, 0xaaaaaaa9aaaaaaa9 },
-   { 0xaaaaaaaaaaaaaaaa, 0xaaaaaaaaaaaaaaaa },
-   { 0xaaaaaaabaaaaaaab, 0xaaaaaaabaaaaaaab },
-   { 0xaaaaaaacaaaaaaac, 0xaaaaaaacaaaaaaac },
-   { 0xaaaaaaadaaaaaaad, 0xaaaaaaadaaaaaaad },
-   { 0xaaaaaaaeaaaaaaae, 0xaaaaaaaeaaaaaaae },
-   { 0xaaaaaaafaaaaaaaf, 0xaaaaaaafaaaaaaaf }
-};
-*/
 void bmw256_4way_init( bmw256_4way_context *ctx )
 {
-   ctx->H[ 0] = v128_64( 0x4041424340414243 );
+   ctx->H[ 0] = v128_32( 0x40414243 );
-   ctx->H[ 1] = v128_64( 0x4445464744454647 );
+   ctx->H[ 1] = v128_32( 0x44454647 );
-   ctx->H[ 2] = v128_64( 0x48494A4B48494A4B );
+   ctx->H[ 2] = v128_32( 0x48494A4B );
-   ctx->H[ 3] = v128_64( 0x4C4D4E4F4C4D4E4F );
+   ctx->H[ 3] = v128_32( 0x4C4D4E4F );
-   ctx->H[ 4] = v128_64( 0x5051525350515253 );
+   ctx->H[ 4] = v128_32( 0x50515253 );
-   ctx->H[ 5] = v128_64( 0x5455565754555657 );
+   ctx->H[ 5] = v128_32( 0x54555657 );
-   ctx->H[ 6] = v128_64( 0x58595A5B58595A5B );
+   ctx->H[ 6] = v128_32( 0x58595A5B );
-   ctx->H[ 7] = v128_64( 0x5C5D5E5F5C5D5E5F );
+   ctx->H[ 7] = v128_32( 0x5C5D5E5F );
-   ctx->H[ 8] = v128_64( 0x6061626360616263 );
+   ctx->H[ 8] = v128_32( 0x60616263 );
-   ctx->H[ 9] = v128_64( 0x6465666764656667 );
+   ctx->H[ 9] = v128_32( 0x64656667 );
-   ctx->H[10] = v128_64( 0x68696A6B68696A6B );
+   ctx->H[10] = v128_32( 0x68696A6B );
-   ctx->H[11] = v128_64( 0x6C6D6E6F6C6D6E6F );
+   ctx->H[11] = v128_32( 0x6C6D6E6F );
-   ctx->H[12] = v128_64( 0x7071727370717273 );
+   ctx->H[12] = v128_32( 0x70717273 );
-   ctx->H[13] = v128_64( 0x7475767774757677 );
+   ctx->H[13] = v128_32( 0x74757677 );
-   ctx->H[14] = v128_64( 0x78797A7B78797A7B );
+   ctx->H[14] = v128_32( 0x78797A7B );
-   ctx->H[15] = v128_64( 0x7C7D7E7F7C7D7E7F );
+   ctx->H[15] = v128_32( 0x7C7D7E7F );
-
-
-//   for ( int i = 0; i < 16; i++ )
-//      sc->H[i] = v128_32( iv[i] );
    ctx->ptr = 0;
    ctx->bit_count = 0;
 }
@@ -478,10 +454,10 @@ void bmw256_4way_init( bmw256_4way_context *ctx )
 static void
 bmw32_4way(bmw_4way_small_context *sc, const void *data, size_t len)
 {
-   v128u64_t *vdata = (v128u64_t*)data;
+   v128u32_t *vdata = (v128u32_t*)data;
-   v128u64_t *buf;
+   v128u32_t *buf;
-   v128u64_t htmp[16];
+   v128u32_t htmp[16];
-   v128u64_t *h1, *h2;
+   v128u32_t *h1, *h2;
    size_t ptr;
    const int buf_size = 64;  // bytes of one lane, compatible with len
 
@@ -503,7 +479,7 @@ bmw32_4way(bmw_4way_small_context *sc, const void *data, size_t len)
       ptr += clen;
       if ( ptr == buf_size )
       {
-         v128u64_t *ht;
+         v128u32_t *ht;
          compress_small( buf, h1, h2 );
          ht = h1;
          h1 = h2;
@@ -521,14 +497,14 @@ static void
 bmw32_4way_close(bmw_4way_small_context *sc, unsigned ub, unsigned n,
 	void *dst, size_t out_size_w32)
 {
-   v128u64_t *buf;
+   v128u32_t *buf;
-   v128u64_t h1[16], h2[16], *h;
+   v128u32_t h1[16], h2[16], *h;
    size_t ptr, u, v;
    const int buf_size = 64;  // bytes of one lane, compatible with len
 
    buf = sc->buf;
    ptr = sc->ptr;
-   buf[ ptr>>2 ] = v128_64( 0x0000008000000080 );
+   buf[ ptr>>2 ] = v128_32( 0x00000080 );
    ptr += 4;
    h = sc->H;
 
@@ -548,7 +524,7 @@ 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, (v128u64_t*)final_s, h1 );
+   compress_small( buf, (v128u32_t*)final_s, h1 );
 
    for (u = 0, v = 16 - out_size_w32; u < out_size_w32; u ++, v ++)
       casti_v128( dst, u ) = h1[v];
@@ -1057,7 +1033,7 @@ void bmw256_8way_close( bmw256_8way_context *ctx, void *dst )
 
 #endif // __AVX2__
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 // BMW-256 16 way 32
 

algo/bmw/bmw512-gate.h

@@ -4,7 +4,7 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define BMW512_8WAY 1
 #elif defined(__AVX2__)
   #define BMW512_4WAY 1

algo/bmw/bmw512-hash-4way.c

@@ -683,8 +683,9 @@ void compress_big( const __m256i *M, const __m256i H[16], __m256i dH[16] )
    mj[14] = mm256_rol_64( M[14], 15 );
    mj[15] = mm256_rol_64( M[15], 16 );
 
-   __m256i K = _mm256_set1_epi64x( 16 * 0x0555555555555555ULL );
+   __m256i K = _mm256_set1_epi64x( 0x5555555555555550ULL );
-   const __m256i Kincr = _mm256_set1_epi64x( 0x0555555555555555ULL );
+   static const __m256i Kincr = { 0x0555555555555555ULL, 0x0555555555555555ULL,
+                                  0x0555555555555555ULL, 0x0555555555555555ULL };
 
    qt[16] = add_elt_b( mj[ 0], mj[ 3], mj[10], H[ 7], K );
    K = _mm256_add_epi64( K, Kincr );
@@ -950,7 +951,7 @@ bmw512_4way_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
 
 #endif  // __AVX2__
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 // BMW-512 8 WAY
 
@@ -1094,7 +1095,7 @@ void compress_big_8way( const __m512i *M, const __m512i H[16],
                         __m512i dH[16] )
 {
    __m512i qt[32], xl, xh;
-   __m512i mh[16];
+   __m512i mh[16], mj[16];
    int i;
 
    for ( i = 0; i < 16; i++ )
@@ -1117,8 +1118,6 @@ void compress_big_8way( const __m512i *M, const __m512i H[16],
    qt[14] = _mm512_add_epi64( s8b4( W8b14), H[15] );
    qt[15] = _mm512_add_epi64( s8b0( W8b15), H[ 0] );
 
-   __m512i mj[16];
- 
    mj[ 0] = mm512_rol_64( M[ 0],  1 );
    mj[ 1] = mm512_rol_64( M[ 1],  2 );
    mj[ 2] = mm512_rol_64( M[ 2],  3 );
@@ -1136,8 +1135,11 @@ void compress_big_8way( const __m512i *M, const __m512i H[16],
    mj[14] = mm512_rol_64( M[14], 15 );
    mj[15] = mm512_rol_64( M[15], 16 );
 
-   __m512i K = _mm512_set1_epi64( 16 * 0x0555555555555555ULL );
+   __m512i K = _mm512_set1_epi64( 0x5555555555555550ULL );
-   const __m512i Kincr = _mm512_set1_epi64( 0x0555555555555555ULL );
+   static const __m512i Kincr = { 0x0555555555555555ULL, 0x0555555555555555ULL,
+                                  0x0555555555555555ULL, 0x0555555555555555ULL,
+                                  0x0555555555555555ULL, 0x0555555555555555ULL,
+                                  0x0555555555555555ULL, 0x0555555555555555ULL };
 
    qt[16] = add_elt_b8( mj[ 0], mj[ 3], mj[10], H[ 7], K );
    K = _mm512_add_epi64( K, Kincr );

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

@@ -26,7 +26,7 @@ static const uint64_t IV512[] =
 0xA5A70E75D65C8A2B, 0xBC796576B1C62456, 0xE7989AF11921C8F7, 0xD43E3B447795D246
 };
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 // 4 way 128 is handy to avoid reinterleaving in many algos.
 // If reinterleaving is necessary it may be more efficient to use

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

@@ -6,7 +6,7 @@
 
 #if defined(__AVX2__)
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 struct _cube_4way_context
 {

algo/cubehash/cubehash_sse2.c

@@ -13,7 +13,7 @@ static void transform( cubehashParam *sp )
     int r;
     const int rounds = sp->rounds;
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
     register __m512i x0, x1;
 
@@ -39,7 +39,7 @@ static void transform( cubehashParam *sp )
 
 #elif defined(__AVX2__)
 
-    register __m256i x0, x1, x2, x3, y0, y1;
+    register __m256i x0, x1, x2, x3, t0;
 
     x0 = _mm256_load_si256( (__m256i*)sp->x     );
     x1 = _mm256_load_si256( (__m256i*)sp->x + 1 );   
@@ -50,10 +50,10 @@ static void transform( cubehashParam *sp )
     { 
         x2 = _mm256_add_epi32( x0, x2 );
         x3 = _mm256_add_epi32( x1, x3 );
-        y0 = mm256_rol_32( x1, 7 );
+        t0 = mm256_rol_32( x1, 7 );
-        y1 = mm256_rol_32( x0, 7 );
+        x1 = mm256_rol_32( x0, 7 );
-        x0 = _mm256_xor_si256( y0, x2 );
+        x0 = _mm256_xor_si256( t0, x2 );
-        x1 = _mm256_xor_si256( y1, x3 );
+        x1 = _mm256_xor_si256( x1, x3 );
         x2 = mm256_swap128_64( x2 );
         x3 = mm256_swap128_64( x3 );
         x2 = _mm256_add_epi32( x0, x2 );
@@ -75,7 +75,7 @@ static void transform( cubehashParam *sp )
 
 #else   // AVX, SSE2, NEON
 
-    v128_t x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3;
+    v128_t x0, x1, x2, x3, x4, x5, x6, x7, t0, t1;
 
     x0 = casti_v128( sp->x, 0 );
     x1 = casti_v128( sp->x, 1 );
@@ -92,16 +92,12 @@ static void transform( cubehashParam *sp )
        x5 = v128_add32( x1, x5 );
        x6 = v128_add32( x2, x6 );
        x7 = v128_add32( x3, x7 );
-       y0 = x2;
+       t0 = v128_rol32( x2, 7 );
-       y1 = x3;
+       t1 = v128_rol32( x3, 7 );
-       y2 = x0;
+       x2 = v128_rol32( x0, 7 );
-       y3 = x1;
+       x3 = v128_rol32( x1, 7 );
-       x0 = v128_rol32( y0, 7 );
+       x0 = v128_xor( t0, x4 );
-       x1 = v128_rol32( y1, 7 );
+       x1 = v128_xor( t1, x5 );
-       x2 = v128_rol32( y2, 7 );
-       x3 = v128_rol32( y3, 7 );
-       x0 = v128_xor( x0, x4 );
-       x1 = v128_xor( x1, x5 );
        x2 = v128_xor( x2, x6 );
        x3 = v128_xor( x3, x7 );
        x4 = v128_swap64( x4 );
@@ -112,17 +108,13 @@ static void transform( cubehashParam *sp )
        x5 = v128_add32( x1, x5 );
        x6 = v128_add32( x2, x6 );
        x7 = v128_add32( x3, x7 );
-       y0 = x1;
+       t0 = v128_rol32( x1, 11 );
-       y1 = x0;
+       x1 = v128_rol32( x0, 11 );
-       y2 = x3;
+       t1 = v128_rol32( x3, 11 );
-       y3 = x2;
+       x3 = v128_rol32( x2, 11 );
-       x0 = v128_rol32( y0, 11 );
+       x0 = v128_xor( t0, x4 );
-       x1 = v128_rol32( y1, 11 );
-       x2 = v128_rol32( y2, 11 );
-       x3 = v128_rol32( y3, 11 );
-	    x0 = v128_xor( x0, x4 );
        x1 = v128_xor( x1, x5 );
-	    x2 = v128_xor( x2, x6 );
+       x2 = v128_xor( t1, x6 );
        x3 = v128_xor( x3, x7 );
        x4 = v128_swap64_32( x4 );
 	    x5 = v128_swap64_32( x5 );

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

@@ -11,7 +11,7 @@ static const unsigned int mul2ipt[] __attribute__ ((aligned (64))) =
 };
 */
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 #define ECHO_SUBBYTES4(state, j) \
    state[0][j] = _mm512_aesenc_epi128( state[0][j], k1 ); \

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

@@ -5,7 +5,7 @@
 
 #include "simd-utils.h"
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 typedef struct
 {

algo/gost/sph_gost.c

@@ -696,7 +696,7 @@ static void AddModulo512(const void *a,const void *b,void *c)
 
 static void AddXor512(const void *a,const void *b,void *c)
 {
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
    casti_m512i( c, 0 ) = _mm512_xor_si512( casti_m512i( a, 0 ),
                                            casti_m512i( b, 0 ) );
 #elif defined(__AVX2__)

algo/groestl/aes_ni/groestl-intr-aes.h

@@ -103,7 +103,7 @@ static const v128u32_t gr_mask __attribute__ ((aligned (16))) =
    This implementation costs 7.7 c/b giving total speed on SNB: 10.7c/b.
    K. Matusiewicz, 2011/05/29 */
 
-#if defined(__AVX512VL__)
+#if defined(VL256)
 
 #define MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
   /* t_i = a_i + a_{i+1} */\

algo/groestl/aes_ni/groestl256-intr-aes.h

@@ -95,7 +95,7 @@ static const v128u32_t gr_mask __attribute__ ((aligned (16))) =
    This implementation costs 7.7 c/b giving total speed on SNB: 10.7c/b.
    K. Matusiewicz, 2011/05/29 */
 
-#if defined(__AVX512VL__)
+#if defined(VL256)
 
 #define MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
   /* t_i = a_i + a_{i+1} */\

algo/groestl/groestl-gate.h

@@ -4,7 +4,7 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(__VAES__) && defined(SIMD512)
   #define GROESTL_4WAY_VAES 1
 #endif
 

algo/groestl/groestl256-hash-4way.c

@@ -17,7 +17,7 @@
 
 #if defined(__AVX2__) && defined(__VAES__)
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 
 int groestl256_4way_init( groestl256_4way_context* ctx, uint64_t hashlen )

algo/groestl/groestl256-hash-4way.h

@@ -43,7 +43,7 @@
 
 #define SIZE256 (SIZE_512/16)
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 typedef struct {
   __attribute__ ((aligned (128))) __m512i chaining[SIZE256];

algo/groestl/groestl256-intr-4way.h

@@ -42,7 +42,7 @@ static const __m128i round_const_l7[] __attribute__ ((aligned (64))) =
    { 0x0000000000000000, 0x8696a6b6c6d6e6f6 }
 };
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 static const __m512i TRANSP_MASK = { 0x0d0509010c040800, 0x0f070b030e060a02,
                                      0x1d1519111c141810, 0x1f171b131e161a12,

algo/groestl/groestl512-hash-4way.c

@@ -17,7 +17,7 @@
 
 #if defined(__AVX2__) && defined(__VAES__)
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 int groestl512_4way_init( groestl512_4way_context* ctx, uint64_t hashlen )
 {

algo/groestl/groestl512-hash-4way.h

@@ -33,7 +33,7 @@
 
 #define SIZE512 (SIZE_1024/16)
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 typedef struct {
   __attribute__ ((aligned (128))) __m512i chaining[SIZE512];

algo/groestl/groestl512-intr-4way.h

@@ -50,7 +50,7 @@ static const __m128i round_const_q[] __attribute__ ((aligned (64))) =
    { 0x8292a2b2c2d2e2f2, 0x0212223242526272 }
 };
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 static const __m512i TRANSP_MASK = { 0x0d0509010c040800, 0x0f070b030e060a02,
                                      0x1d1519111c141810, 0x1f171b131e161a12,
@@ -239,7 +239,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
   { \
     /* AddRoundConstant P1024 */\
     xmm8 = _mm512_xor_si512( xmm8, mm512_bcast_m128( \
-             casti_m128i( round_const_p, round_counter ) ) ); \
+             casti_v128u32( round_const_p, round_counter ) ) ); \
     /* ShiftBytes P1024 + pre-AESENCLAST */\
     xmm8  = _mm512_shuffle_epi8( xmm8,  SUBSH_MASK0 ); \
     xmm9  = _mm512_shuffle_epi8( xmm9,  SUBSH_MASK1 );\
@@ -254,7 +254,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
     \
      /* AddRoundConstant P1024 */\
     xmm0 = _mm512_xor_si512( xmm0, mm512_bcast_m128( \
-             casti_m128i( round_const_p, round_counter+1 ) ) ); \
+             casti_v128u32( round_const_p, round_counter+1 ) ) ); \
     /* ShiftBytes P1024 + pre-AESENCLAST */\
     xmm0 = _mm512_shuffle_epi8( xmm0, SUBSH_MASK0 );\
     xmm1 = _mm512_shuffle_epi8( xmm1, SUBSH_MASK1 );\
@@ -283,7 +283,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
     xmm13 = _mm512_xor_si512( xmm13, xmm1 );\
     xmm14 = _mm512_xor_si512( xmm14, xmm1 );\
     xmm15 = _mm512_xor_si512( xmm15, mm512_bcast_m128( \
-                 casti_m128i( round_const_q, round_counter ) ) ); \
+                 casti_v128u32( round_const_q, round_counter ) ) ); \
     /* ShiftBytes Q1024 + pre-AESENCLAST */\
     xmm8  = _mm512_shuffle_epi8( xmm8,  SUBSH_MASK1 );\
     xmm9  = _mm512_shuffle_epi8( xmm9,  SUBSH_MASK3 );\
@@ -306,7 +306,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
     xmm5 = _mm512_xor_si512( xmm5, xmm9 );\
     xmm6 = _mm512_xor_si512( xmm6, xmm9 );\
     xmm7 = _mm512_xor_si512( xmm7, mm512_bcast_m128( \
-             casti_m128i( round_const_q, round_counter+1 ) ) ); \
+             casti_v128u32( round_const_q, round_counter+1 ) ) ); \
     /* ShiftBytes Q1024 + pre-AESENCLAST */\
     xmm0 = _mm512_shuffle_epi8( xmm0, SUBSH_MASK1 );\
     xmm1 = _mm512_shuffle_epi8( xmm1, SUBSH_MASK3 );\
@@ -812,7 +812,7 @@ static const __m256i SUBSH_MASK7_2WAY =
   { \
     /* AddRoundConstant P1024 */\
     xmm8 = _mm256_xor_si256( xmm8, mm256_bcast_m128( \
-             casti_m128i( round_const_p, round_counter ) ) ); \
+             casti_v128u32( round_const_p, round_counter ) ) ); \
     /* ShiftBytes P1024 + pre-AESENCLAST */\
     xmm8  = _mm256_shuffle_epi8( xmm8,  SUBSH_MASK0_2WAY ); \
     xmm9  = _mm256_shuffle_epi8( xmm9,  SUBSH_MASK1_2WAY );\
@@ -827,7 +827,7 @@ static const __m256i SUBSH_MASK7_2WAY =
     \
      /* AddRoundConstant P1024 */\
     xmm0 = _mm256_xor_si256( xmm0, mm256_bcast_m128( \
-             casti_m128i( round_const_p, round_counter+1 ) ) ); \
+             casti_v128u32( round_const_p, round_counter+1 ) ) ); \
     /* ShiftBytes P1024 + pre-AESENCLAST */\
     xmm0 = _mm256_shuffle_epi8( xmm0, SUBSH_MASK0_2WAY );\
     xmm1 = _mm256_shuffle_epi8( xmm1, SUBSH_MASK1_2WAY );\
@@ -856,7 +856,7 @@ static const __m256i SUBSH_MASK7_2WAY =
     xmm13 = _mm256_xor_si256( xmm13, xmm1 );\
     xmm14 = _mm256_xor_si256( xmm14, xmm1 );\
     xmm15 = _mm256_xor_si256( xmm15, mm256_bcast_m128( \
-                 casti_m128i( round_const_q, round_counter ) ) ); \
+                 casti_v128u32( round_const_q, round_counter ) ) ); \
     /* ShiftBytes Q1024 + pre-AESENCLAST */\
     xmm8  = _mm256_shuffle_epi8( xmm8,  SUBSH_MASK1_2WAY );\
     xmm9  = _mm256_shuffle_epi8( xmm9,  SUBSH_MASK3_2WAY );\
@@ -879,7 +879,7 @@ static const __m256i SUBSH_MASK7_2WAY =
     xmm5 = _mm256_xor_si256( xmm5, xmm9 );\
     xmm6 = _mm256_xor_si256( xmm6, xmm9 );\
     xmm7 = _mm256_xor_si256( xmm7, mm256_bcast_m128( \
-             casti_m128i( round_const_q, round_counter+1 ) ) ); \
+             casti_v128u32( round_const_q, round_counter+1 ) ) ); \
     /* ShiftBytes Q1024 + pre-AESENCLAST */\
     xmm0 = _mm256_shuffle_epi8( xmm0, SUBSH_MASK1_2WAY );\
     xmm1 = _mm256_shuffle_epi8( xmm1, SUBSH_MASK3_2WAY );\

algo/groestl/myrgr-4way.c

@@ -17,7 +17,7 @@ typedef struct {
 #else
    hashState_groestl       groestl;
 #endif
-   sha256_8way_context     sha;
+   sha256_8x32_context     sha;
 } myrgr_8way_ctx_holder;
 
 myrgr_8way_ctx_holder myrgr_8way_ctx;
@@ -29,7 +29,7 @@ void init_myrgr_8way_ctx()
 #else
      init_groestl( &myrgr_8way_ctx.groestl, 64 );
 #endif
-     sha256_8way_init( &myrgr_8way_ctx.sha );
+     sha256_8x32_init( &myrgr_8way_ctx.sha );
 }
 
 void myriad_8way_hash( void *output, const void *input )
@@ -96,8 +96,8 @@ void myriad_8way_hash( void *output, const void *input )
      intrlv_8x32_512( vhash, hash0, hash1, hash2, hash3, hash4, hash5,
                        hash6, hash7 );
      
-     sha256_8way_update( &ctx.sha, vhash, 64 );
+     sha256_8x32_update( &ctx.sha, vhash, 64 );
-     sha256_8way_close( &ctx.sha, output );
+     sha256_8x32_close( &ctx.sha, output );
 }
 
 int scanhash_myriad_8way( struct work *work, uint32_t max_nonce,
@@ -156,7 +156,7 @@ int scanhash_myriad_8way( struct work *work, uint32_t max_nonce,
 
 typedef struct {
     hashState_groestl       groestl;
-    sha256_4way_context     sha;
+    sha256_4x32_context     sha;
 } myrgr_4way_ctx_holder;
 
 myrgr_4way_ctx_holder myrgr_4way_ctx;
@@ -164,7 +164,7 @@ myrgr_4way_ctx_holder myrgr_4way_ctx;
 void init_myrgr_4way_ctx()
 {
      init_groestl (&myrgr_4way_ctx.groestl, 64 );
-     sha256_4way_init( &myrgr_4way_ctx.sha );
+     sha256_4x32_init( &myrgr_4way_ctx.sha );
 }
 
 void myriad_4way_hash( void *output, const void *input )
@@ -189,8 +189,8 @@ void myriad_4way_hash( void *output, const void *input )
 
      intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
 
-     sha256_4way_update( &ctx.sha, vhash, 64 );
+     sha256_4x32_update( &ctx.sha, vhash, 64 );
-     sha256_4way_close( &ctx.sha, output );
+     sha256_4x32_close( &ctx.sha, output );
 }
 
 int scanhash_myriad_4way( struct work *work, uint32_t max_nonce,

algo/groestl/myrgr-gate.c

@@ -16,7 +16,7 @@ bool register_myriad_algo( algo_gate_t* gate )
   init_myrgr_ctx();
   gate->scanhash  = (void*)&scanhash_myriad;
   gate->hash      = (void*)&myriad_hash;
-  gate->optimizations = AES_OPT | SSE2_OPT | AVX2_OPT | SHA_OPT | VAES_OPT;
+  gate->optimizations = AES_OPT | SSE2_OPT | AVX2_OPT | SHA256_OPT | VAES_OPT;
 #endif
   return true;
 };

algo/groestl/myrgr-gate.h

@@ -4,7 +4,7 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(__VAES__) && defined(SIMD512)
   #define MYRGR_8WAY 1
 #elif defined(__AVX2__) && defined(__AES__) && !defined(__SHA__)
   #define MYRGR_4WAY 1

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

@@ -382,12 +382,12 @@ static const uint32_t T512[64][16] __attribute__ ((aligned (32))) =
 #define S1F   MF
 
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 // Hamsi 8 way AVX512 
 
 // Intel docs say _mm512_movepi64_mask & _mm512_cmplt_epi64_mask have same
-// timig. However, when tested hashing X13 on i9-9940x using cmplt with zero
+// timing. However, testing hashing X13 on i9-9940x using cmplt with zero
 // had a 3% faster overall hashrate than than using movepi. 
 
 #define INPUT_BIG8 \
@@ -418,13 +418,11 @@ static const uint32_t T512[64][16] __attribute__ ((aligned (32))) =
   tb = mm512_xoror( b, d, a ); \
   a = _mm512_xor_si512( a, c ); \
   b = mm512_xoror( td, tb, a ); \
-  td = mm512_xorand( a, td, tb ); \
+  d = _mm512_ternarylogic_epi64( a, td, tb, 0x87 );/* not( xorand( a, td, tb ) ); */ \
   a = c; \
-  c = mm512_xor3( tb, b, td ); \
+  c = _mm512_ternarylogic_epi64( tb, b, d, 0x69 ); /* not( xor3( tb, b, d ) ); */ \
-  d = mm512_not( td ); \
 }
 
-
 /*
 #define SBOX8( a, b, c, d ) \
 do { \
@@ -505,32 +503,28 @@ do { \
   SBOX8( s2, s6, sA, sE ); /* ( m1, c3, m5, c7 ) */ \
   SBOX8( s3, s7, sB, sF ); /* ( c1, m3, c5, m7 ) */ \
   s4 = mm512_swap64_32( s4 ); \
-  s5 = mm512_swap64_32( s5 ); \
+  t0 = _mm512_mask_shuffle_epi32( s4, 0xaaaa, s5, 0xb1 ); \
   sD = mm512_swap64_32( sD ); \
-  sE = mm512_swap64_32( sE ); \
+  t1 = _mm512_mask_shuffle_epi32( sD, 0xaaaa, sE, 0xb1 ); \
-  t0 = _mm512_mask_blend_epi32( 0xaaaa, s4, s5 ); \
-  t1 = _mm512_mask_blend_epi32( 0xaaaa, sD, sE ); \
   L8( s0, t0, s9, t1 ); \
   s6 = mm512_swap64_32( s6 ); \
   sF = mm512_swap64_32( sF ); \
-  t2 = _mm512_mask_blend_epi32( 0xaaaa, s5, s6 ); \
+  t2 = _mm512_mask_shuffle_epi32( s6, 0x5555, s5, 0xb1 ); \
-  t3 = _mm512_mask_blend_epi32( 0xaaaa, sE, sF ); \
+  t3 = _mm512_mask_shuffle_epi32( sF, 0x5555, sE, 0xb1 ); \
   L8( s1, t2, sA, t3 ); \
   s5 = _mm512_mask_blend_epi32( 0x5555, t0, t2 ); \
   sE = _mm512_mask_blend_epi32( 0x5555, t1, t3 ); \
 \
-  s7 = mm512_swap64_32( s7 ); \
+  t4 = _mm512_mask_shuffle_epi32( s6, 0xaaaa, s7, 0xb1 ); \
-  sC = mm512_swap64_32( sC ); \
+  t5 = _mm512_mask_shuffle_epi32( sF, 0xaaaa, sC, 0xb1 ); \
-  t4 = _mm512_mask_blend_epi32( 0xaaaa, s6, s7 ); \
-  t5 = _mm512_mask_blend_epi32( 0xaaaa, sF, sC ); \
   L8( s2, t4, sB, t5 ); \
   s6 = _mm512_mask_blend_epi32( 0x5555, t2, t4 ); \
   sF = _mm512_mask_blend_epi32( 0x5555, t3, t5 ); \
   s6 = mm512_swap64_32( s6 ); \
   sF = mm512_swap64_32( sF ); \
 \
-  t2 = _mm512_mask_blend_epi32( 0xaaaa, s7, s4 ); \
+  t2 = _mm512_mask_shuffle_epi32( s4, 0x5555, s7, 0xb1 ); \
-  t3 = _mm512_mask_blend_epi32( 0xaaaa, sC, sD ); \
+  t3 = _mm512_mask_shuffle_epi32( sD, 0x5555, sC, 0xb1 ); \
   L8( s3, t2, s8, t3 ); \
   s7 = _mm512_mask_blend_epi32( 0x5555, t4, t2 ); \
   s4 = _mm512_mask_blend_epi32( 0xaaaa, t0, t2 ); \
@@ -539,21 +533,20 @@ do { \
   s7 = mm512_swap64_32( s7 ); \
   sC = mm512_swap64_32( sC ); \
 \
-  t0 = _mm512_mask_blend_epi32( 0xaaaa, s0, mm512_swap64_32( s8 ) ); \
+  t0 = _mm512_mask_shuffle_epi32( s0, 0xaaaa, s8, 0xb1 ); \
   t1 = _mm512_mask_blend_epi32( 0xaaaa, s1, s9 ); \
-  t2 = _mm512_mask_blend_epi32( 0xaaaa, mm512_swap64_32( s2 ), sA ); \
+  t2 = _mm512_mask_shuffle_epi32( sA, 0x5555, s2, 0xb1 ); \
   t3 = _mm512_mask_blend_epi32( 0x5555, s3, sB ); \
   t3 = mm512_swap64_32( t3 ); \
   L8( t0, t1, t2, t3 ); \
-  t3 = mm512_swap64_32( t3 ); \
   s0 = _mm512_mask_blend_epi32( 0x5555, s0, t0 ); \
-  s8 = _mm512_mask_blend_epi32( 0x5555, s8, mm512_swap64_32( t0 ) ); \
+  s8 = _mm512_mask_shuffle_epi32( s8, 0x5555, t0, 0xb1 ); \
   s1 = _mm512_mask_blend_epi32( 0x5555, s1, t1 ); \
   s9 = _mm512_mask_blend_epi32( 0xaaaa, s9, t1 ); \
-  s2 = _mm512_mask_blend_epi32( 0xaaaa, s2, mm512_swap64_32( t2 ) ); \
+  s2 = _mm512_mask_shuffle_epi32( s2, 0xaaaa, t2, 0xb1 ); \
   sA = _mm512_mask_blend_epi32( 0xaaaa, sA, t2 ); \
-  s3 = _mm512_mask_blend_epi32( 0xaaaa, s3, t3 ); \
+  s3 = _mm512_mask_shuffle_epi32( s3, 0xaaaa, t3, 0xb1 ); \
-  sB = _mm512_mask_blend_epi32( 0x5555, sB, t3 ); \
+  sB = _mm512_mask_shuffle_epi32( sB, 0x5555, t3, 0xb1 ); \
 \
   t0 = _mm512_mask_blend_epi32( 0xaaaa, s4, sC ); \
   t1 = _mm512_mask_blend_epi32( 0xaaaa, s5, sD ); \
@@ -1061,7 +1054,7 @@ void hamsi_8way_big( hamsi_8way_big_context *sc, __m512i *buf, size_t num )
    WRITE_STATE_BIG8( sc );
 }
 
-void hamsi_8way_big_final( hamsi_8way_big_context *sc, __m512i *buf )
+void hamsi_8way_big_final( hamsi512_8x64_context *sc, __m512i *buf )
 {
    __m512i m0, m1, m2, m3, m4, m5, m6, m7;
 
@@ -1073,7 +1066,7 @@ void hamsi_8way_big_final( hamsi_8way_big_context *sc, __m512i *buf )
    WRITE_STATE_BIG8( sc );
 }
 
-void hamsi512_8way_init( hamsi_8way_big_context *sc )
+void hamsi512_8x64_init( hamsi512_8x64_context *sc )
 {
    sc->partial_len = 0;
    sc->count_high = sc->count_low = 0;
@@ -1089,7 +1082,7 @@ void hamsi512_8way_init( hamsi_8way_big_context *sc )
    sc->h[7] = v512_64( iv[7] );
    }
 
-void hamsi512_8way_update( hamsi_8way_big_context *sc, const void *data,
+void hamsi512_8x64_update( hamsi512_8x64_context *sc, const void *data,
                            size_t len )
 {
    __m512i *vdata = (__m512i*)data;
@@ -1101,7 +1094,7 @@ void hamsi512_8way_update( hamsi_8way_big_context *sc, const void *data,
    sc->partial_len = len;
 }
 
-void hamsi512_8way_close( hamsi_8way_big_context *sc, void *dst )
+void hamsi512_8x64_close( hamsi512_8x64_context *sc, void *dst )
 {
    __m512i pad[1];
    uint32_t ch, cl;
@@ -1122,7 +1115,7 @@ void hamsi512_8way_close( hamsi_8way_big_context *sc, void *dst )
 
 // Hamsi 4 way AVX2
 
-#if defined(__AVX512VL__)
+#if defined(VL256)
 
 #define INPUT_BIG \
 do { \
@@ -1160,6 +1153,94 @@ do { \
 
 #else
 
+#define INPUT_BIG_sub( db_i ) \
+{ \
+     const __m256i dm = _mm256_cmpgt_epi64( zero, db_i ); \
+     m0 = _mm256_xor_si256( m0, _mm256_and_si256( dm, v256_64( tp[0] ) ) ); \
+     m1 = _mm256_xor_si256( m1, _mm256_and_si256( dm, v256_64( tp[1] ) ) ); \
+     m2 = _mm256_xor_si256( m2, _mm256_and_si256( dm, v256_64( tp[2] ) ) ); \
+     m3 = _mm256_xor_si256( m3, _mm256_and_si256( dm, v256_64( tp[3] ) ) ); \
+     m4 = _mm256_xor_si256( m4, _mm256_and_si256( dm, v256_64( tp[4] ) ) ); \
+     m5 = _mm256_xor_si256( m5, _mm256_and_si256( dm, v256_64( tp[5] ) ) ); \
+     m6 = _mm256_xor_si256( m6, _mm256_and_si256( dm, v256_64( tp[6] ) ) ); \
+     m7 = _mm256_xor_si256( m7, _mm256_and_si256( dm, v256_64( tp[7] ) ) ); \
+     tp += 8; \
+}
+
+#define INPUT_BIG \
+{ \
+  const __m256i db = *buf; \
+  const __m256i zero = m256_zero; \
+  const uint64_t *tp = (const uint64_t*)T512;  \
+  m0 = m1 = m2 = m3 = m4 = m5 = m6 = m7 = zero; \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,63 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,62 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,61 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,60 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,59 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,58 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,57 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,56 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,55 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,54 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,53 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,52 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,51 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,50 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,49 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,48 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,47 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,46 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,45 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,44 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,43 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,42 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,41 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,40 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,39 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,38 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,37 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,36 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,35 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,34 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,33 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,32 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,31 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,30 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,29 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,28 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,27 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,26 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,25 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,24 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,23 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,22 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,21 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,20 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,19 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,18 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,17 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,16 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,15 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,14 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,13 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,12 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,11 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db,10 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db, 9 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db, 8 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db, 7 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db, 6 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db, 5 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db, 4 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db, 3 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db, 2 ) ); \
+  INPUT_BIG_sub( _mm256_slli_epi64( db, 1 ) ); \
+  INPUT_BIG_sub( db ); \
+}
+
+#if 0
+// dependent on the compiler unrolling the loop
 #define INPUT_BIG \
 do { \
   __m256i db = *buf; \
@@ -1180,8 +1261,9 @@ do { \
      tp += 8; \
   } \
 } while (0)
+#endif
 
-// v3 no ternary logic, 15 instructions, 9 TL equivalent instructions
+// v3, 15 instructions
 #define SBOX( a, b, c, d ) \
 { \
   __m256i tb, td; \
@@ -1199,7 +1281,7 @@ do { \
 #endif
 
 /*
-/ v2, 16 instructions, 10 TL equivalent instructions
+/ v2, 16 instructions
 #define SBOX( a, b, c, d ) \
 { \
   __m256i t = mm256_xorand( d, a, c ); \
@@ -1219,7 +1301,7 @@ do { \
 do { \
    a = mm256_rol_32( a, 13 ); \
    c = mm256_rol_32( c,  3 ); \
-   b = mm256_xor3( a, b, c ); \
+   b = mm256_xor3( b, a, c ); \
    d = mm256_xor3( d, c, _mm256_slli_epi32( a, 3 ) ); \
    b = mm256_rol_32( b, 1 ); \
    d = mm256_rol_32( d, 7 ); \
@@ -1501,7 +1583,7 @@ do { /* order is important */ \
    sc->h[14] = CE; \
    sc->h[15] = CF;
 
-#if defined(__AVX512VL__)
+#if defined(VL256)
 
 #define INPUT_8X32 \
 { \
@@ -1857,7 +1939,7 @@ void hamsi512_8x32_full( hamsi512_8x32_context *sc, void * dst,
 
 ////////////
 
-void hamsi_big( hamsi_4way_big_context *sc, __m256i *buf, size_t num )
+void hamsi_big( hamsi512_4x64_context *sc, __m256i *buf, size_t num )
 {
    DECL_STATE_BIG
    uint32_t tmp;
@@ -1881,7 +1963,7 @@ void hamsi_big( hamsi_4way_big_context *sc, __m256i *buf, size_t num )
    WRITE_STATE_BIG( sc );
 }
 
-void hamsi_big_final( hamsi_4way_big_context *sc, __m256i *buf )
+void hamsi_big_final( hamsi512_4x64_context *sc, __m256i *buf )
 {
    __m256i m0, m1, m2, m3, m4, m5, m6, m7;
    DECL_STATE_BIG
@@ -1892,7 +1974,7 @@ void hamsi_big_final( hamsi_4way_big_context *sc, __m256i *buf )
    WRITE_STATE_BIG( sc );
 }
 
-void hamsi512_4way_init( hamsi_4way_big_context *sc )
+void hamsi512_4x64_init( hamsi512_4x64_context *sc )
 {
    sc->partial_len = 0;
    sc->count_high = sc->count_low = 0;
@@ -1907,7 +1989,7 @@ void hamsi512_4way_init( hamsi_4way_big_context *sc )
    sc->h[7] = v256_64( iv[7] );
 }
 
-void hamsi512_4way_update( hamsi_4way_big_context *sc, const void *data,
+void hamsi512_4x64_update( hamsi512_4x64_context *sc, const void *data,
       size_t len )
 {
    __m256i *vdata = (__m256i*)data;
@@ -1919,7 +2001,7 @@ void hamsi512_4way_update( hamsi_4way_big_context *sc, const void *data,
    sc->partial_len = len;
 }
 
-void hamsi512_4way_close( hamsi_4way_big_context *sc, void *dst )
+void hamsi512_4x64_close( hamsi512_4x64_context *sc, void *dst )
 {
    __m256i pad[1];
    uint32_t ch, cl;
@@ -1961,6 +2043,94 @@ void hamsi512_4way_close( hamsi_4way_big_context *sc, void *dst )
    sc->h[6] = c6; \
    sc->h[7] = c7;
 
+#define INPUT_2x64_sub( db_i ) \
+{ \
+     const v128u64_t dm = v128_cmpgt64( zero, db_i ); \
+     m0 = v128_xor( m0, v128_and( dm, v128_64( tp[0] ) ) ); \
+     m1 = v128_xor( m1, v128_and( dm, v128_64( tp[1] ) ) ); \
+     m2 = v128_xor( m2, v128_and( dm, v128_64( tp[2] ) ) ); \
+     m3 = v128_xor( m3, v128_and( dm, v128_64( tp[3] ) ) ); \
+     m4 = v128_xor( m4, v128_and( dm, v128_64( tp[4] ) ) ); \
+     m5 = v128_xor( m5, v128_and( dm, v128_64( tp[5] ) ) ); \
+     m6 = v128_xor( m6, v128_and( dm, v128_64( tp[6] ) ) ); \
+     m7 = v128_xor( m7, v128_and( dm, v128_64( tp[7] ) ) ); \
+     tp += 8; \
+}
+
+#define INPUT_2x64 \
+{ \
+  const v128u64_t db = *buf; \
+  const v128u64_t zero = v128_zero; \
+  const uint64_t *tp = (const uint64_t*)T512;  \
+  m0 = m1 = m2 = m3 = m4 = m5 = m6 = m7 = zero; \
+  INPUT_2x64_sub( v128_sl64( db,63 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,62 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,61 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,60 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,59 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,58 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,57 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,56 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,55 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,54 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,53 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,52 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,51 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,50 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,49 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,48 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,47 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,46 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,45 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,44 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,43 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,42 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,41 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,40 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,39 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,38 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,37 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,36 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,35 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,34 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,33 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,32 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,31 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,30 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,29 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,28 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,27 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,26 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,25 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,24 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,23 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,22 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,21 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,20 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,19 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,18 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,17 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,16 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,15 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,14 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,13 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,12 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,11 ) ); \
+  INPUT_2x64_sub( v128_sl64( db,10 ) ); \
+  INPUT_2x64_sub( v128_sl64( db, 9 ) ); \
+  INPUT_2x64_sub( v128_sl64( db, 8 ) ); \
+  INPUT_2x64_sub( v128_sl64( db, 7 ) ); \
+  INPUT_2x64_sub( v128_sl64( db, 6 ) ); \
+  INPUT_2x64_sub( v128_sl64( db, 5 ) ); \
+  INPUT_2x64_sub( v128_sl64( db, 4 ) ); \
+  INPUT_2x64_sub( v128_sl64( db, 3 ) ); \
+  INPUT_2x64_sub( v128_sl64( db, 2 ) ); \
+  INPUT_2x64_sub( v128_sl64( db, 1 ) ); \
+  INPUT_2x64_sub( db ); \
+}
+
+#if 0
+// Dependent on the compiler unrolling the loop.
 #define INPUT_2x64 \
 { \
   v128u64_t db = *buf; \
@@ -1981,6 +2151,7 @@ void hamsi512_4way_close( hamsi_4way_big_context *sc, void *dst )
      tp += 8; \
   } \
 }
+#endif
 
 // v3 no ternary logic, 15 instructions, 9 TL equivalent instructions
 #define SBOX_2x64( a, b, c, d ) \
@@ -2001,7 +2172,7 @@ void hamsi512_4way_close( hamsi_4way_big_context *sc, void *dst )
 { \
    a = v128_rol32( a, 13 ); \
    c = v128_rol32( c,  3 ); \
-   b = v128_xor3( a, b, c ); \
+   b = v128_xor3( c, a, b ); \
    d = v128_xor3( d, c, v128_sl32( a, 3 ) ); \
    b = v128_rol32( b, 1 ); \
    d = v128_rol32( d, 7 ); \

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

@@ -72,17 +72,17 @@ typedef struct
    size_t partial_len;
    uint32_t count_high, count_low;
 } hamsi_4way_big_context;
-typedef hamsi_4way_big_context hamsi512_4way_context;
+typedef hamsi_4way_big_context hamsi512_4x64_context;
 
-void hamsi512_4way_init( hamsi512_4way_context *sc );
+void hamsi512_4x64_init( hamsi512_4x64_context *sc );
-void hamsi512_4way_update( hamsi512_4way_context *sc, const void *data,
+void hamsi512_4x64_update( hamsi512_4x64_context *sc, const void *data,
       size_t len );
-void hamsi512_4way_close( hamsi512_4way_context *sc, void *dst );
+void hamsi512_4x64_close( hamsi512_4x64_context *sc, void *dst );
 
-#define hamsi512_4x64_context   hamsi512_4way_context
+#define hamsi512_4way_context   hamsi512_4x64_context
-#define hamsi512_4x64_init      hamsi512_4way_init
+#define hamsi512_4way_init      hamsi512_4x64_init
-#define hamsi512_4x64_update    hamsi512_4way_update
+#define hamsi512_4way_update    hamsi512_4x64_update
-#define hamsi512_4x64_close     hamsi512_4way_close
+#define hamsi512_4way_close     hamsi512_4x64_close
 
 // Hamsi-512 8x32
 
@@ -104,7 +104,7 @@ void hamsi512_8x32_full( hamsi512_8x32_context *sc, void *dst, const void *data,
 
 #endif
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 // Hamsi-512 8x64
 
@@ -115,17 +115,17 @@ typedef struct
    size_t partial_len;
    uint32_t count_high, count_low;
 } hamsi_8way_big_context;
-typedef hamsi_8way_big_context hamsi512_8way_context;
+typedef hamsi_8way_big_context hamsi512_8x64_context;
 
-void hamsi512_8way_init( hamsi512_8way_context *sc );
+void hamsi512_8x64_init( hamsi512_8x64_context *sc );
-void hamsi512_8way_update( hamsi512_8way_context *sc, const void *data,
+void hamsi512_8x64_update( hamsi512_8x64_context *sc, const void *data,
                            size_t len );
-void hamsi512_8way_close( hamsi512_8way_context *sc, void *dst );
+void hamsi512_8x64_close( hamsi512_8x64_context *sc, void *dst );
 
-#define hamsi512_8x64_context   hamsi512_8way_context
+#define hamsi512_8way_context   hamsi512_8x64_context
-#define hamsi512_8x64_init      hamsi512_8way_init
+#define hamsi512_8way_init      hamsi512_8x64_init
-#define hamsi512_8x64_update    hamsi512_8way_update
+#define hamsi512_8way_update    hamsi512_8x64_update
-#define hamsi512_8x64_close     hamsi512_8way_close
+#define hamsi512_8way_close     hamsi512_8x64_close
 
 // Hamsi-512 16x32
 

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

@@ -53,7 +53,7 @@ extern "C"{
 #define SPH_SMALL_FOOTPRINT_HAVAL   1
 //#endif
 
-#if defined(__AVX512VL__)
+#if defined(VL256)
 
 // ( ~( a ^ b ) ) & c
 #define v128_andnotxor( a, b, c ) \
@@ -583,7 +583,7 @@ do { \
 
 // Haval-256 8 way 32 bit avx2
 
-#if defined (__AVX512VL__)
+#if defined (VL256)
 
 // ( ~( a ^ b ) ) & c
 #define mm256_andnotxor( a, b, c ) \
@@ -882,7 +882,7 @@ do { \
 
 #endif // AVX2
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512) 
 
 // ( ~( a ^ b ) ) & c
 #define mm512_andnotxor( a, b, c ) \

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

@@ -82,12 +82,15 @@ typedef struct {
 typedef haval_4way_context haval256_5_4way_context;
 
 void haval256_5_4way_init( void *cc );
-
 void haval256_5_4way_update( void *cc, const void *data, size_t len );
 //#define haval256_5_4way haval256_5_4way_update
-
 void haval256_5_4way_close( void *cc, void *dst );
 
+#define haval256_4x32_context    haval256_5_4way_context
+#define haval256_4x32_init       haval256_5_4way_init
+#define haval256_4x32_update     haval256_5_4way_update
+#define haval256_4x32_close      haval256_5_4way_close
+
 #if defined(__AVX2__)
 
 typedef struct {
@@ -100,14 +103,17 @@ typedef struct {
 typedef haval_8way_context haval256_5_8way_context;
 
 void haval256_5_8way_init( void *cc );
-
 void haval256_5_8way_update( void *cc, const void *data, size_t len );
-
 void haval256_5_8way_close( void *cc, void *dst );
 
+#define haval256_8x32_context    haval256_5_8way_context
+#define haval256_8x32_init       haval256_5_8way_init
+#define haval256_8x32_update     haval256_5_8way_update
+#define haval256_8x32_close      haval256_5_8way_close
+
 #endif // AVX2
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 typedef struct {
    __m512i buf[32];
@@ -119,11 +125,14 @@ typedef struct {
 typedef haval_16way_context haval256_5_16way_context;
 
 void haval256_5_16way_init( void *cc );
-
 void haval256_5_16way_update( void *cc, const void *data, size_t len );
-
 void haval256_5_16way_close( void *cc, void *dst );
 
+#define haval256_16x32_context    haval256_5_16way_context
+#define haval256_16x32_init       haval256_5_16way_init
+#define haval256_16x32_update     haval256_5_16way_update
+#define haval256_16x32_close      haval256_5_16way_close
+
 #endif // AVX512
 
 #ifdef __cplusplus

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

@@ -204,7 +204,7 @@ static const uint64_t IV512[] =
       (state)->H[15] = h7l; \
    } while (0)
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 #define Sb_8W(x0, x1, x2, x3, c) \
 { \
@@ -364,8 +364,7 @@ static const uint64_t IV512[] =
 
 #if defined(__AVX2__)
 
-#if defined(__AVX512VL__)
+#if defined(VL256)
-//TODO enable for AVX10_256, not used with AVX512VL
 
 #define notxorandnot( a, b, c ) \
    _mm256_ternarylogic_epi64( a, b, c, 0x2d )
@@ -522,7 +521,7 @@ static const uint64_t IV512[] =
 
 #endif   // AVX2
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 void jh256_8x64_init( jh_8x64_context *sc )
 {

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

@@ -55,7 +55,7 @@
  * <code>memcpy()</code>).
  */
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 typedef struct
 {

algo/keccak/keccak-gate.h

@@ -4,7 +4,7 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define KECCAK_8WAY 1
 #elif defined(__AVX2__)
   #define KECCAK_4WAY 1
@@ -12,7 +12,7 @@
   #define KECCAK_2WAY 1
 #endif
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define SHA3D_8WAY 1
 #elif defined(__AVX2__)
   #define SHA3D_4WAY 1

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

@@ -57,7 +57,7 @@ static const uint64_t RC[] = {
 
 #define DO(x)   x
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 #define INPUT_BUF(size)   do { \
     size_t j; \
@@ -161,29 +161,25 @@ keccak64_8way_core( keccak64_ctx_m512i *kc, const void *data, size_t len,
 static void keccak64_8way_close( keccak64_ctx_m512i *kc, void *dst,
                                  size_t byte_len, size_t lim )
 {
-    unsigned eb;
+    __m512i tmp[lim + 1] __attribute__ ((aligned (64)));
-    union {
-       __m512i tmp[lim + 1];
-       uint64_t dummy;   /* for alignment */
-    } u;
     size_t j;
     size_t m512_len = byte_len >> 3;
+    const unsigned eb = hard_coded_eb;
 
-    eb = hard_coded_eb;
     if ( kc->ptr == (lim - 8) )
     {
         const uint64_t t = eb | 0x8000000000000000;
-        u.tmp[0] = _mm512_set1_epi64( t );
+        tmp[0] = _mm512_set1_epi64( t );
         j = 8;
     }
     else
     {
         j = lim - kc->ptr;
-        u.tmp[0] = _mm512_set1_epi64( eb );
+        tmp[0] = _mm512_set1_epi64( eb );
-        memset_zero_512( u.tmp + 1, (j>>3) - 2 );
+        memset_zero_512( tmp + 1, (j>>3) - 2 );
-        u.tmp[ (j>>3) - 1] = _mm512_set1_epi64( 0x8000000000000000 );
+        tmp[ (j>>3) - 1] = _mm512_set1_epi64( 0x8000000000000000 );
     }
-    keccak64_8way_core( kc, u.tmp, j, lim );
+    keccak64_8way_core( kc, tmp, j, lim );
     /* Finalize the "lane complement" */
     NOT64( kc->w[ 1], kc->w[ 1] );
     NOT64( kc->w[ 2], kc->w[ 2] );
@@ -194,7 +190,7 @@ static void keccak64_8way_close( keccak64_ctx_m512i *kc, void *dst,
     memcpy_512( dst, kc->w, m512_len );
 }
 
-void keccak256_8way_init( void *kc )
+void keccak256_8x64_init( void *kc )
 {
    keccak64_8way_init( kc, 256 );
 }
@@ -361,29 +357,25 @@ keccak64_core( keccak64_ctx_m256i *kc, const void *data, size_t len,
 static void keccak64_close( keccak64_ctx_m256i *kc, void *dst, size_t byte_len,
             size_t lim )
 {
-    unsigned eb;
+    __m256i tmp[lim + 1] __attribute__ ((aligned (32)));
-    union {
-       __m256i tmp[lim + 1];
-       uint64_t dummy;   /* for alignment */
-    } u;
     size_t j;
     size_t m256_len = byte_len >> 3;
+    const unsigned eb = hard_coded_eb;
 
-    eb = hard_coded_eb;
     if ( kc->ptr == (lim - 8) )
     {
         const uint64_t t = eb | 0x8000000000000000;
-        u.tmp[0] = _mm256_set1_epi64x( t );
+        tmp[0] = _mm256_set1_epi64x( t );
         j = 8;
     }
     else
     {
         j = lim - kc->ptr;
-        u.tmp[0] = _mm256_set1_epi64x( eb );
+        tmp[0] = _mm256_set1_epi64x( eb );
-        memset_zero_256( u.tmp + 1, (j>>3) - 2 );
+        memset_zero_256( tmp + 1, (j>>3) - 2 );
-        u.tmp[ (j>>3) - 1] = _mm256_set1_epi64x( 0x8000000000000000 );
+        tmp[ (j>>3) - 1] = _mm256_set1_epi64x( 0x8000000000000000 );
     }
-    keccak64_core( kc, u.tmp, j, lim );
+    keccak64_core( kc, tmp, j, lim );
     /* Finalize the "lane complement" */
     NOT64( kc->w[ 1], kc->w[ 1] );
     NOT64( kc->w[ 2], kc->w[ 2] );

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

@@ -4,7 +4,7 @@
 #include <stddef.h>
 #include "simd-utils.h"
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 typedef struct
 {

algo/keccak/sha3d-4way.c

@@ -9,7 +9,7 @@
 void sha3d_hash_8way(void *state, const void *input)
 {
     uint32_t buffer[16*8] __attribute__ ((aligned (128)));
-    keccak256_8way_context ctx;
+    keccak256_8x64_context ctx;
 
     keccak256_8x64_init( &ctx );
     keccak256_8x64_update( &ctx, input, 80 );
@@ -69,7 +69,7 @@ int scanhash_sha3d_8way( struct work *work, uint32_t max_nonce,
 void sha3d_hash_4way(void *state, const void *input)
 {
     uint32_t buffer[16*4] __attribute__ ((aligned (64)));
-    keccak256_4way_context ctx;
+    keccak256_4x64_context ctx;
 
     keccak256_4x64_init( &ctx );
     keccak256_4x64_update( &ctx, input, 80 );

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

@@ -59,7 +59,7 @@ static const uint32_t CNS_INIT[128] __attribute((aligned(64))) = {
 };
 
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 #define cns4w(i)  mm512_bcast_m128( ( (__m128i*)CNS_INIT)[i] )
 
@@ -80,14 +80,14 @@ static const uint32_t CNS_INIT[128] __attribute((aligned(64))) = {
     __m512i t = a0; \
     a0 = mm512_xoror( a3, a0, a1 ); \
     a2 = _mm512_xor_si512( a2, a3 ); \
-    a1 = _mm512_ternarylogic_epi64( a1, a3, t, 0x87 ); /* a1 xnor (a3 & t) */ \
+    a1 = _mm512_ternarylogic_epi64( a1, a3, t, 0x87 ); /* a1 nxor (a3 & t) */ \
     a3 = mm512_xorand( a2, a3, t ); \
     a2 = mm512_xorand( a1, a2, a0); \
     a1 = _mm512_or_si512( a1, a3 ); \
     a3 = _mm512_xor_si512( a3, a2 ); \
     t  = _mm512_xor_si512( t, a1 ); \
     a2 = _mm512_and_si512( a2, a1 ); \
-    a1 = mm512_xnor( a1, a0 ); \
+    a1 = mm512_nxor( a1, a0 ); \
     a0 = t; \
 }
 
@@ -273,8 +273,6 @@ void finalization512_4way( luffa_4way_context *state, uint32 *b )
     uint32_t hash[8*4] __attribute((aligned(128)));
     __m512i* chainv = state->chainv;
     __m512i t[2];
-    const __m512i shuff_bswap32 = mm512_bcast_m128( _mm_set_epi64x( 
-                                  0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
 
     /*---- blank round with m=0 ----*/
     rnd512_4way( state, NULL );
@@ -289,10 +287,8 @@ void finalization512_4way( luffa_4way_context *state, uint32 *b )
     _mm512_store_si512( (__m512i*)&hash[ 0], t[0] );
     _mm512_store_si512( (__m512i*)&hash[16], t[1] );
 
-    casti_m512i( b,0 ) = _mm512_shuffle_epi8(
+    casti_m512i( b,0 ) = mm512_bswap_32( casti_m512i( hash,0 ) );
-                                  casti_m512i( hash,0 ), shuff_bswap32 );
+    casti_m512i( b,1 ) = mm512_bswap_32( casti_m512i( hash,1 ) );
-    casti_m512i( b,1 ) = _mm512_shuffle_epi8(
-                                  casti_m512i( hash,1 ), shuff_bswap32 );
 
     rnd512_4way( state, NULL );
 
@@ -306,10 +302,8 @@ void finalization512_4way( luffa_4way_context *state, uint32 *b )
     _mm512_store_si512( (__m512i*)&hash[ 0], t[0] );
     _mm512_store_si512( (__m512i*)&hash[16], t[1] );
 
-    casti_m512i( b,2 ) = _mm512_shuffle_epi8(
+    casti_m512i( b,2 ) = mm512_bswap_32( casti_m512i( hash,0 ) );
-                                  casti_m512i( hash,0 ), shuff_bswap32 );
+    casti_m512i( b,3 ) = mm512_bswap_32( casti_m512i( hash,1 ) );
-    casti_m512i( b,3 ) = _mm512_shuffle_epi8(
-                                  casti_m512i( hash,1 ), shuff_bswap32 );
 }
 
 int luffa_4way_init( luffa_4way_context *state, int hashbitlen )
@@ -349,16 +343,14 @@ int luffa_4way_update( luffa_4way_context *state, const void *data,
     __m512i msg[2];
     int i;
     int blocks = (int)len >> 5;
-    const __m512i shuff_bswap32 = mm512_bcast_m128( _mm_set_epi64x(  
-                                   0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
 
     state->rembytes = (int)len & 0x1F;
 
     // full blocks
     for ( i = 0; i < blocks; i++, vdata+=2 )
     {
-       msg[0] = _mm512_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[0] = mm512_bswap_32( vdata[ 0 ] );
-       msg[1] = _mm512_shuffle_epi8( vdata[ 1 ], shuff_bswap32 );
+       msg[1] = mm512_bswap_32( vdata[ 1 ] );
        rnd512_4way( state, msg );
     }
 
@@ -367,7 +359,7 @@ int luffa_4way_update( luffa_4way_context *state, const void *data,
     if ( state->rembytes  )
     {
       // remaining data bytes
-      buffer[0] = _mm512_shuffle_epi8( vdata[0], shuff_bswap32 );
+      buffer[0] = mm512_bswap_32( vdata[0] );
       buffer[1] = mm512_bcast128lo_64( 0x0000000080000000 );
     }
     return 0;
@@ -434,16 +426,14 @@ int luffa512_4way_full( luffa_4way_context *state, void *output,
     __m512i msg[2];
     int i;
     const int blocks = (int)( inlen >> 5 );
-    const __m512i shuff_bswap32 = mm512_bcast_m128( _mm_set_epi64x( 
-                                   0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
 
     state->rembytes = inlen & 0x1F;
 
     // full blocks
     for ( i = 0; i < blocks; i++, vdata+=2 )
     {
-       msg[0] = _mm512_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[0] = mm512_bswap_32( vdata[ 0 ] );
-       msg[1] = _mm512_shuffle_epi8( vdata[ 1 ], shuff_bswap32 );
+       msg[1] = mm512_bswap_32( vdata[ 1 ] );
        rnd512_4way( state, msg );
     }
 
@@ -451,7 +441,7 @@ int luffa512_4way_full( luffa_4way_context *state, void *output,
     if ( state->rembytes  )
     {
        // padding of partial block
-       msg[0] = _mm512_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[0] = mm512_bswap_32( vdata[ 0 ] );
        msg[1] = mm512_bcast128lo_64( 0x0000000080000000 );
        rnd512_4way( state, msg );
     }
@@ -479,16 +469,14 @@ int luffa_4way_update_close( luffa_4way_context *state,
     __m512i msg[2];
     int i;
     const int blocks = (int)( inlen >> 5 );
-    const __m512i shuff_bswap32 = mm512_bcast_m128( _mm_set_epi64x( 
-                                   0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
 
     state->rembytes = inlen & 0x1F;
 
     // full blocks
     for ( i = 0; i < blocks; i++, vdata+=2 )
     {
-       msg[0] = _mm512_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[0] = mm512_bswap_32( vdata[ 0 ] );
-       msg[1] = _mm512_shuffle_epi8( vdata[ 1 ], shuff_bswap32 );
+       msg[1] = mm512_bswap_32( vdata[ 1 ] );
        rnd512_4way( state, msg );
     }
 
@@ -496,7 +484,7 @@ int luffa_4way_update_close( luffa_4way_context *state,
     if ( state->rembytes  )
     {
        // padding of partial block
-       msg[0] = _mm512_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[0] = mm512_bswap_32( vdata[ 0 ] );
        msg[1] = mm512_bcast128lo_64( 0x0000000080000000 );
        rnd512_4way( state, msg );
     }
@@ -524,8 +512,7 @@ int luffa_4way_update_close( luffa_4way_context *state,
     a = _mm256_xor_si256( a, c0 ); \
     b = _mm256_xor_si256( b, c1 );
 
-//TODO Enable for AVX10_256, not used with AVX512 or AVX10_512
+#if defined(VL256) 
-#if defined(__AVX512VL__) 
 
 #define MULT2( a0, a1 ) \
 { \
@@ -540,14 +527,14 @@ int luffa_4way_update_close( luffa_4way_context *state,
     __m256i t = a0; \
     a0 = mm256_xoror( a3, a0, a1 ); \
     a2 = _mm256_xor_si256( a2, a3 ); \
-    a1 = _mm256_ternarylogic_epi64( a1, a3, t, 0x87 ); /* a1 xnor (a3 & t) */ \
+    a1 = _mm256_ternarylogic_epi64( a1, a3, t, 0x87 ); /* a1 nxor (a3 & t) */ \
     a3 = mm256_xorand( a2, a3, t ); \
     a2 = mm256_xorand( a1, a2, a0); \
     a1 = _mm256_or_si256( a1, a3 ); \
     a3 = _mm256_xor_si256( a3, a2 ); \
     t  = _mm256_xor_si256( t, a1 ); \
     a2 = _mm256_and_si256( a2, a1 ); \
-    a1 = mm256_xnor( a1, a0 ); \
+    a1 = mm256_nxor( a1, a0 ); \
     a0 = t; \
 }
 
@@ -776,8 +763,6 @@ void finalization512_2way( luffa_2way_context *state, uint32 *b )
     uint32 hash[8*2] __attribute((aligned(64)));
     __m256i* chainv = state->chainv;
     __m256i t0, t1;
-    const __m256i shuff_bswap32 = mm256_set2_64( 0x0c0d0e0f08090a0b,
-                                                 0x0405060700010203 );
     /*---- blank round with m=0 ----*/
     rnd512_2way( state, NULL );
 
@@ -792,10 +777,8 @@ void finalization512_2way( luffa_2way_context *state, uint32 *b )
     _mm256_store_si256( (__m256i*)&hash[0], t0 );
     _mm256_store_si256( (__m256i*)&hash[8], t1 );
 
-    casti_m256i( b, 0 ) = _mm256_shuffle_epi8(
+    casti_m256i( b, 0 ) = mm256_bswap_32( casti_m256i( hash, 0 ) );
-                                  casti_m256i( hash, 0 ), shuff_bswap32 );
+    casti_m256i( b, 1 ) = mm256_bswap_32( casti_m256i( hash, 1 ) );
-    casti_m256i( b, 1 ) = _mm256_shuffle_epi8( 
-                                  casti_m256i( hash, 1 ), shuff_bswap32 );
 
     rnd512_2way( state, NULL );
 
@@ -810,10 +793,8 @@ void finalization512_2way( luffa_2way_context *state, uint32 *b )
     _mm256_store_si256( (__m256i*)&hash[0], t0 );
     _mm256_store_si256( (__m256i*)&hash[8], t1 );
 
-    casti_m256i( b, 2 ) = _mm256_shuffle_epi8( 
+    casti_m256i( b, 2 ) = mm256_bswap_32( casti_m256i( hash, 0 ) );
-                                  casti_m256i( hash, 0 ), shuff_bswap32 );
+    casti_m256i( b, 3 ) = mm256_bswap_32( casti_m256i( hash, 1 ) );
-    casti_m256i( b, 3 ) = _mm256_shuffle_epi8( 
-                                  casti_m256i( hash, 1 ), shuff_bswap32 );
 }
 
 int luffa_2way_init( luffa_2way_context *state, int hashbitlen )
@@ -848,15 +829,13 @@ int luffa_2way_update( luffa_2way_context *state, const void *data,
     __m256i msg[2];
     int i;
     int blocks = (int)len >> 5;
-    const __m256i shuff_bswap32 = mm256_set2_64( 0x0c0d0e0f08090a0b,
-                                                 0x0405060700010203 );
     state-> rembytes = (int)len & 0x1F;
 
     // full blocks
     for ( i = 0; i < blocks; i++, vdata+=2 )
     {
-       msg[0] = _mm256_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[0] = mm256_bswap_32( vdata[ 0 ] );
-       msg[1] = _mm256_shuffle_epi8( vdata[ 1 ], shuff_bswap32 );
+       msg[1] = mm256_bswap_32( vdata[ 1 ] );
        rnd512_2way( state, msg );
     }
 
@@ -865,7 +844,7 @@ int luffa_2way_update( luffa_2way_context *state, const void *data,
     if ( state->rembytes  )
     {
       // remaining data bytes
-      buffer[0] = _mm256_shuffle_epi8( vdata[0], shuff_bswap32 );
+      buffer[0] = mm256_bswap_32( vdata[0] );
       buffer[1] = mm256_bcast128lo_64( 0x0000000080000000 );
     }
     return 0;
@@ -917,16 +896,14 @@ int luffa512_2way_full( luffa_2way_context *state, void *output,
     __m256i msg[2];
     int i;
     const int blocks = (int)( inlen >> 5 );
-    const __m256i shuff_bswap32 = mm256_set2_64( 0x0c0d0e0f08090a0b,
-                                                 0x0405060700010203 );
 
     state->rembytes = inlen & 0x1F;
 
     // full blocks
     for ( i = 0; i < blocks; i++, vdata+=2 )
     {
-       msg[0] = _mm256_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[0] = mm256_bswap_32( vdata[ 0 ] );
-       msg[1] = _mm256_shuffle_epi8( vdata[ 1 ], shuff_bswap32 );
+       msg[1] = mm256_bswap_32( vdata[ 1 ] );
        rnd512_2way( state, msg );
     }
 
@@ -934,7 +911,7 @@ int luffa512_2way_full( luffa_2way_context *state, void *output,
     if ( state->rembytes  )
     {
        // padding of partial block
-       msg[0] = _mm256_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[0] = mm256_bswap_32( vdata[ 0 ] );
        msg[1] = mm256_bcast128lo_64( 0x0000000080000000 );
        rnd512_2way( state, msg );
     }
@@ -962,16 +939,14 @@ int luffa_2way_update_close( luffa_2way_context *state,
     __m256i msg[2];
     int i;
     const int blocks = (int)( inlen >> 5 );
-    const __m256i shuff_bswap32 = mm256_set2_64( 0x0c0d0e0f08090a0b,
-                                                 0x0405060700010203 );
 
     state->rembytes = inlen & 0x1F;
 
     // full blocks
     for ( i = 0; i < blocks; i++, vdata+=2 )
     {
-       msg[0] = _mm256_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[0] = mm256_bswap_32( vdata[ 0 ] );
-       msg[1] = _mm256_shuffle_epi8( vdata[ 1 ], shuff_bswap32 );
+       msg[1] = mm256_bswap_32( vdata[ 1 ] );
        rnd512_2way( state, msg );
     }
 
@@ -979,7 +954,7 @@ int luffa_2way_update_close( luffa_2way_context *state,
     if ( state->rembytes  )
     {
        // padding of partial block
-       msg[0] = _mm256_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[0] = mm256_bswap_32( vdata[ 0 ] );
        msg[1] = mm256_bcast128lo_64( 0x0000000080000000 );
        rnd512_2way( state, msg );
     }

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

@@ -51,7 +51,7 @@
 #define LIMIT_512 128
 /*********************************/
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 typedef struct {
     uint32_t buffer[8*4];

algo/luffa/luffa_for_sse2.c

@@ -28,8 +28,7 @@
     a = v128_xor( a, c0 ); \
     b = v128_xor( b, c1 ); \
 
-#if defined(__AVX512VL__)
+#if defined(VL256)
-//TODO enable for AVX10_512 AVX10_256
 
 #define MULT2( a0, a1 ) \
 { \
@@ -48,47 +47,40 @@
   a1 = _mm_alignr_epi8( b, a1, 4 ); \
 }
 
-#elif defined(__ARM_NEON)
+
+#elif defined(__ARM_NEON) || defined(__SSE2__)
 
 // { a1_0, 0, a1_0, a1_0 }
 #define MULT2( a0, a1 ) \
 { \
-  v128_t b = v128_xor( a0, v128_and( vdupq_laneq_u32( a1, 0 ), MASK ) ); \
+  v128_t b = v128_xor( a0, v128_and( v128_bcast32( a1 ), MASK ) ); \
   a0 = v128_alignr32( a1, b, 1 ); \
   a1 = v128_alignr32( b, a1, 1 ); \
 }
 
-#else   // assume SSE2
+#else
-
+  #warning __FILE__ ":" __LINE__ " Unknown or unsupported CPU architecture."
-#define MULT2( a0, a1 ) \
-{ \
-  v128_t b = v128_xor( a0, v128_and( _mm_shuffle_epi32( a1, 0 ), MASK ) ); \
-  a0 = v128_or( _mm_srli_si128(  b, 4 ), _mm_slli_si128( a1, 12 ) ); \
-  a1 = v128_or( _mm_srli_si128( a1, 4 ), _mm_slli_si128(  b, 12 ) ); \
-} 
-
 #endif
 
-#if defined(__AVX512VL__)
+#if defined(VL256)
-//TODO enable for AVX10_512 AVX10_256
 
 #define SUBCRUMB( a0, a1, a2, a3 ) \
 { \
     v128_t t = a0; \
     a0 = v128_xoror( a3, a0, a1 ); \
     a2 = v128_xor( a2, a3 ); \
-    a1 = _mm_ternarylogic_epi64( a1, a3, t, 0x87 ); /* a1 xnor (a3 & t) */ \
+    a1 = _mm_ternarylogic_epi64( a1, a3, t, 0x87 ); /* ~a1 ^ (a3 & t) */ \
     a3 = v128_xorand( a2, a3, t ); \
     a2 = v128_xorand( a1, a2, a0 ); \
     a1 = v128_or( a1, a3 ); \
     a3 = v128_xor( a3, a2 ); \
     t  = v128_xor( t, a1 ); \
     a2 = v128_and( a2, a1 ); \
-    a1 = v128_xnor( a1, a0 ); \
+    a1 = v128_nxor( a1, a0 ); \
     a0 = t; \
 }
 
-#else
+#elif defined(__ARM_NEON) || defined(__SSE2__)
 
 #define SUBCRUMB( a0, a1, a2, a3 ) \
 { \

algo/luffa/luffa_for_sse2.h

@@ -68,4 +68,4 @@ int update_and_final_luffa( hashState_luffa *state, void* output,
 
 int luffa_full( hashState_luffa *state, void* output, int hashbitlen,
                                    const void* data, size_t inlen );
-#endif   // LUFFA_FOR_SSE2_H___
+#endif   // LUFFA_FOR_SSE2_H__

algo/lyra2/allium-4way.c

@@ -15,7 +15,7 @@
  #include "algo/groestl/sph_groestl.h"
 #endif
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define ALLIUM_16WAY 1
 #elif defined(__AVX2__)
   #define ALLIUM_8WAY 1
@@ -26,9 +26,9 @@
 #if defined (ALLIUM_16WAY)  
 
 typedef union {
-   keccak256_8way_context    keccak;
+   keccak256_8x64_context    keccak;
    cube_4way_2buf_context    cube;
-   skein256_8way_context     skein;
+   skein256_8x64_context     skein;
 #if defined(__VAES__)
    groestl256_4way_context   groestl;
 #else
@@ -60,7 +60,7 @@ static void allium_16way_hash( void *state, const void *midstate_vars,
    uint32_t hash15[8] __attribute__ ((aligned (32)));
    allium_16way_ctx_holder ctx __attribute__ ((aligned (64)));
 
-   blake256_16way_final_rounds_le( vhash, midstate_vars, midhash, block, 14 );
+   blake256_16x32_final_rounds_le( vhash, midstate_vars, midhash, block, 14 );
 
    dintrlv_16x32( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
                   hash8, hash9, hash10, hash11, hash12, hash13, hash14, hash15,
@@ -70,12 +70,12 @@ static void allium_16way_hash( void *state, const void *midstate_vars,
    intrlv_8x64( vhashB, hash8, hash9, hash10, hash11, hash12, hash13, hash14,
                 hash15, 256 );
    
-   keccak256_8way_init( &ctx.keccak );
+   keccak256_8x64_init( &ctx.keccak );
-   keccak256_8way_update( &ctx.keccak, vhashA, 32 );
+   keccak256_8x64_update( &ctx.keccak, vhashA, 32 );
-   keccak256_8way_close( &ctx.keccak, vhashA);
+   keccak256_8x64_close( &ctx.keccak, vhashA);
-   keccak256_8way_init( &ctx.keccak );
+   keccak256_8x64_init( &ctx.keccak );
-   keccak256_8way_update( &ctx.keccak, vhashB, 32 );
+   keccak256_8x64_update( &ctx.keccak, vhashB, 32 );
-   keccak256_8way_close( &ctx.keccak, vhashB);
+   keccak256_8x64_close( &ctx.keccak, vhashB);
 
    dintrlv_8x64( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
                  vhashA, 256 );
@@ -153,12 +153,12 @@ static void allium_16way_hash( void *state, const void *midstate_vars,
    intrlv_8x64( vhashB, hash8, hash9, hash10, hash11, hash12, hash13, hash14,
                 hash15, 256 );
 
-   skein256_8way_init( &ctx.skein );
+   skein256_8x64_init( &ctx.skein );
-   skein256_8way_update( &ctx.skein, vhashA, 32 );
+   skein256_8x64_update( &ctx.skein, vhashA, 32 );
-   skein256_8way_close( &ctx.skein, vhashA );
+   skein256_8x64_close( &ctx.skein, vhashA );
-   skein256_8way_init( &ctx.skein );
+   skein256_8x64_init( &ctx.skein );
-   skein256_8way_update( &ctx.skein, vhashB, 32 );
+   skein256_8x64_update( &ctx.skein, vhashB, 32 );
-   skein256_8way_close( &ctx.skein, vhashB );
+   skein256_8x64_close( &ctx.skein, vhashB );
 
    dintrlv_8x64( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
                  vhashA, 256 );
@@ -251,7 +251,7 @@ int scanhash_allium_16way( struct work *work, uint32_t max_nonce,
                                n+ 7, n+ 6, n+ 5, n+ 4, n+ 3, n+ 2, n+ 1, n );
 
    // Partialy prehash second block without touching nonces in block_buf[3].
-   blake256_16way_round0_prehash_le( midstate_vars, block0_hash, block_buf );
+   blake256_16x32_round0_prehash_le( midstate_vars, block0_hash, block_buf );
 
    do {
      allium_16way_hash( hash, midstate_vars, block0_hash, block_buf );
@@ -273,9 +273,9 @@ int scanhash_allium_16way( struct work *work, uint32_t max_nonce,
 #elif defined (ALLIUM_8WAY)  
 
 typedef union {
-   keccak256_4way_context    keccak;
+   keccak256_4x64_context    keccak;
    cube_2way_context         cube;
-   skein256_4way_context     skein;
+   skein256_4x64_context     skein;
 #if defined(__VAES__)
    groestl256_2way_context   groestl;
 #else
@@ -298,19 +298,19 @@ static void allium_8way_hash( void *hash, const void *midstate_vars,
    uint64_t *hash7 = (uint64_t*)hash+28;
    allium_8way_ctx_holder ctx __attribute__ ((aligned (64))); 
 
-   blake256_8way_final_rounds_le( vhashA, midstate_vars, midhash, block, 14 );
+   blake256_8x32_final_rounds_le( vhashA, midstate_vars, midhash, block, 14 );
 
    dintrlv_8x32( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
                  vhashA, 256 );
    intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 256 );
    intrlv_4x64( vhashB, hash4, hash5, hash6, hash7, 256 );
 
-   keccak256_4way_init( &ctx.keccak );
+   keccak256_4x64_init( &ctx.keccak );
-   keccak256_4way_update( &ctx.keccak, vhashA, 32 );
+   keccak256_4x64_update( &ctx.keccak, vhashA, 32 );
-   keccak256_4way_close( &ctx.keccak, vhashA );
+   keccak256_4x64_close( &ctx.keccak, vhashA );
-   keccak256_4way_init( &ctx.keccak );
+   keccak256_4x64_init( &ctx.keccak );
-   keccak256_4way_update( &ctx.keccak, vhashB, 32 );
+   keccak256_4x64_update( &ctx.keccak, vhashB, 32 );
-   keccak256_4way_close( &ctx.keccak, vhashB );
+   keccak256_4x64_close( &ctx.keccak, vhashB );
 
    dintrlv_4x64( hash0, hash1, hash2, hash3, vhashA, 256 );
    dintrlv_4x64( hash4, hash5, hash6, hash7, vhashB, 256 );
@@ -350,12 +350,12 @@ static void allium_8way_hash( void *hash, const void *midstate_vars,
    intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 256 );
    intrlv_4x64( vhashB, hash4, hash5, hash6, hash7, 256 );
 
-   skein256_4way_init( &ctx.skein );
+   skein256_4x64_init( &ctx.skein );
-   skein256_4way_update( &ctx.skein, vhashA, 32 );
+   skein256_4x64_update( &ctx.skein, vhashA, 32 );
-   skein256_4way_close( &ctx.skein, vhashA );
+   skein256_4x64_close( &ctx.skein, vhashA );
-   skein256_4way_init( &ctx.skein );
+   skein256_4x64_init( &ctx.skein );
-   skein256_4way_update( &ctx.skein, vhashB, 32 );
+   skein256_4x64_update( &ctx.skein, vhashB, 32 );
-   skein256_4way_close( &ctx.skein, vhashB );
+   skein256_4x64_close( &ctx.skein, vhashB );
 
 #if defined(__VAES__)
 
@@ -433,7 +433,7 @@ int scanhash_allium_8way( struct work *work, uint32_t max_nonce,
                                      n+ 3, n+ 2, n+ 1, n );
 
    // Partialy prehash second block without touching nonces
-   blake256_8way_round0_prehash_le( midstate_vars, block0_hash, block_buf );
+   blake256_8x32_round0_prehash_le( midstate_vars, block0_hash, block_buf );
 
    do {
      allium_8way_hash( hash, midstate_vars, block0_hash, block_buf );
@@ -483,7 +483,7 @@ static void allium_4way_hash( void *hash, const void *midstate_vars,
    uint64_t *hash3 = (uint64_t*)hash+12;
    allium_4way_ctx_holder ctx __attribute__ ((aligned (64)));
 
-   blake256_4way_final_rounds_le( vhashA, midstate_vars, midhash, block, 14 );
+   blake256_4x32_final_rounds_le( vhashA, midstate_vars, midhash, block, 14 );
    dintrlv_4x32( hash0, hash1, hash2, hash3, vhashA, 256 );
 
    intrlv_2x64( vhashA, hash0, hash1, 256 );
@@ -588,7 +588,7 @@ int scanhash_allium_4way( struct work *work, uint32_t max_nonce,
    block_buf[15] = v128_32( 640 );
 
       // Partialy prehash second block without touching nonces
-   blake256_4way_round0_prehash_le( midstate_vars, block0_hash, block_buf );
+   blake256_4x32_round0_prehash_le( midstate_vars, block0_hash, block_buf );
 
    do {
      allium_4way_hash( hash, midstate_vars, block0_hash, block_buf );
@@ -616,7 +616,6 @@ int scanhash_allium_4way( struct work *work, uint32_t max_nonce,
 //
 //  1 way
 
-
 typedef struct 
 {
         blake256_context        blake;

algo/lyra2/lyra2-gate.h

@@ -5,7 +5,7 @@
 #include <stdint.h>
 #include "lyra2.h"
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define LYRA2REV3_16WAY 1
 #elif defined(__AVX2__)
   #define LYRA2REV3_8WAY 1
@@ -49,7 +49,7 @@ bool init_lyra2rev3_ctx();
 
 //////////////////////////////////
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define LYRA2REV2_16WAY 1
 #elif defined(__AVX2__)
   #define LYRA2REV2_8WAY 1
@@ -108,7 +108,7 @@ bool lyra2h_thread_init();
 
 /////////////////////////////////////////
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define PHI2_8WAY 1
 #elif defined(__AVX2__) && defined(__AES__)
   #define PHI2_4WAY 1

algo/lyra2/lyra2-hash-2way.c

@@ -41,7 +41,7 @@
 //  lyra2z330, lyra2h, 
 
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 /**
  * Executes Lyra2 based on the G function from Blake2b. This version supports salts and passwords

algo/lyra2/lyra2.h

@@ -59,7 +59,7 @@ int LYRA2Z( uint64_t*, void *K, uint64_t kLen, const void *pwd,
 
 int LYRA2(void *K, int64_t kLen, const void *pwd, int32_t pwdlen, const void *salt, int32_t saltlen, int64_t timeCost, const int16_t nRows, const int16_t nCols);
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 int LYRA2RE_2WAY( void *K, uint64_t kLen, const void *pwd, uint64_t pwdlen,
                   uint64_t timeCost, uint64_t nRows, uint64_t nCols );

algo/lyra2/lyra2h-4way.c

@@ -14,12 +14,12 @@ bool lyra2h_4way_thread_init()
  return ( lyra2h_4way_matrix = mm_malloc( LYRA2H_MATRIX_SIZE, 64 ) );
 }
 
-static __thread blake256_4way_context l2h_4way_blake_mid;
+static __thread blake256_4x32_context l2h_4way_blake_mid;
 
 void lyra2h_4way_midstate( const void* input )
 {
-       blake256_4way_init( &l2h_4way_blake_mid );
+       blake256_4x32_init( &l2h_4way_blake_mid );
-       blake256_4way_update( &l2h_4way_blake_mid, input, 64 );
+       blake256_4x32_update( &l2h_4way_blake_mid, input, 64 );
 }
 
 void lyra2h_4way_hash( void *state, const void *input )
@@ -29,11 +29,11 @@ void lyra2h_4way_hash( void *state, const void *input )
      uint32_t hash2[8] __attribute__ ((aligned (64)));
      uint32_t hash3[8] __attribute__ ((aligned (64)));
      uint32_t vhash[8*4] __attribute__ ((aligned (64)));
-     blake256_4way_context ctx_blake __attribute__ ((aligned (64)));
+     blake256_4x32_context ctx_blake __attribute__ ((aligned (64)));
 
      memcpy( &ctx_blake, &l2h_4way_blake_mid, sizeof l2h_4way_blake_mid );
-     blake256_4way_update( &ctx_blake, input + (64*4), 16 );
+     blake256_4x32_update( &ctx_blake, input + (64*4), 16 );
-     blake256_4way_close( &ctx_blake, vhash );
+     blake256_4x32_close( &ctx_blake, vhash );
 
      dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
 

algo/lyra2/lyra2rev2-4way.c

@@ -7,25 +7,24 @@
 #include "algo/cubehash/cubehash_sse2.h" 
 #include "algo/cubehash/cube-hash-2way.h"
 
-
 #if defined (LYRA2REV2_16WAY)
 
 typedef struct {
-   blake256_16way_context    blake;
+   blake256_16x32_context    blake;
-   keccak256_8way_context    keccak;
+   keccak256_8x64_context    keccak;
    cubehashParam             cube;
-   skein256_8way_context     skein;
+   skein256_8x64_context     skein;
-   bmw256_16way_context      bmw;
+   bmw256_16x32_context      bmw;
 } lyra2v2_16way_ctx_holder __attribute__ ((aligned (64)));
 
 static lyra2v2_16way_ctx_holder l2v2_16way_ctx;
 
 bool init_lyra2rev2_16way_ctx()
 {
-   keccak256_8way_init( &l2v2_16way_ctx.keccak );
+   keccak256_8x64_init( &l2v2_16way_ctx.keccak );
    cubehashInit( &l2v2_16way_ctx.cube, 256, 16, 32 );
-   skein256_8way_init( &l2v2_16way_ctx.skein );
+   skein256_8x64_init( &l2v2_16way_ctx.skein );
-   bmw256_16way_init( &l2v2_16way_ctx.bmw );
+   bmw256_16x32_init( &l2v2_16way_ctx.bmw );
    return true;
 }
 
@@ -51,8 +50,8 @@ void lyra2rev2_16way_hash( void *state, const void *input )
    lyra2v2_16way_ctx_holder ctx __attribute__ ((aligned (64)));
    memcpy( &ctx, &l2v2_16way_ctx, sizeof(l2v2_16way_ctx) );
 
-   blake256_16way_update( &ctx.blake, input + (64<<4), 16 );
+   blake256_16x32_update( &ctx.blake, input + (64<<4), 16 );
-   blake256_16way_close( &ctx.blake, vhash );
+   blake256_16x32_close( &ctx.blake, vhash );
 
    dintrlv_16x32( hash0,  hash1,  hash2,  hash3,
                   hash4,  hash5,  hash6,  hash7,
@@ -62,17 +61,17 @@ void lyra2rev2_16way_hash( void *state, const void *input )
    intrlv_8x64( vhash, hash0, hash1, hash2, hash3,
                        hash4, hash5, hash6, hash7, 256 );
 
-   keccak256_8way_update( &ctx.keccak, vhash, 32 );
+   keccak256_8x64_update( &ctx.keccak, vhash, 32 );
-   keccak256_8way_close( &ctx.keccak, vhash );
+   keccak256_8x64_close( &ctx.keccak, vhash );
 
    dintrlv_8x64( hash0, hash1, hash2, hash3,
                  hash4, hash5, hash6, hash7, vhash, 256 );
    intrlv_8x64( vhash, hash8,  hash9,  hash10, hash11,
                        hash12, hash13, hash14, hash15, 256 );
 
-   keccak256_8way_init( &ctx.keccak );
+   keccak256_8x64_init( &ctx.keccak );
-   keccak256_8way_update( &ctx.keccak, vhash, 32 );
+   keccak256_8x64_update( &ctx.keccak, vhash, 32 );
-   keccak256_8way_close( &ctx.keccak, vhash );
+   keccak256_8x64_close( &ctx.keccak, vhash );
 
    dintrlv_8x64( hash8,  hash9,  hash10,  hash11,
                  hash12, hash13, hash14, hash15, vhash, 256 );
@@ -122,22 +121,21 @@ void lyra2rev2_16way_hash( void *state, const void *input )
 
    intrlv_8x64( vhash, hash0, hash1, hash2, hash3,
                        hash4, hash5, hash6, hash7, 256 );
-   skein256_8way_update( &ctx.skein, vhash, 32 );
+   skein256_8x64_update( &ctx.skein, vhash, 32 );
-   skein256_8way_close( &ctx.skein, vhash );
+   skein256_8x64_close( &ctx.skein, vhash );
 
    dintrlv_8x64( hash0, hash1, hash2, hash3,
                  hash4, hash5, hash6, hash7, vhash, 256 );
    intrlv_8x64( vhash, hash8,  hash9,  hash10, hash11, hash12,
                        hash13, hash14, hash15, 256 );
 
-   skein256_8way_init( &ctx.skein );
+   skein256_8x64_init( &ctx.skein );
-   skein256_8way_update( &ctx.skein, vhash, 32 );
+   skein256_8x64_update( &ctx.skein, vhash, 32 );
-   skein256_8way_close( &ctx.skein, vhash );
+   skein256_8x64_close( &ctx.skein, vhash );
 
    dintrlv_8x64( hash8,  hash9,  hash10, hash11,
                  hash12, hash13, hash14, hash15, vhash, 256 );
    
-   
    cubehash_full( &ctx.cube, (byte*) hash0,  256, (const byte*) hash0, 32 );
    cubehash_full( &ctx.cube, (byte*) hash1,  256, (const byte*) hash1, 32 );
    cubehash_full( &ctx.cube, (byte*) hash2,  256, (const byte*) hash2, 32 );
@@ -160,8 +158,8 @@ void lyra2rev2_16way_hash( void *state, const void *input )
                         hash8,  hash9,  hash10, hash11,
                         hash12, hash13, hash14, hash15, 256 );
 
-   bmw256_16way_update( &ctx.bmw, vhash, 32 );
+   bmw256_16x32_update( &ctx.bmw, vhash, 32 );
-   bmw256_16way_close( &ctx.bmw, state );
+   bmw256_16x32_close( &ctx.bmw, state );
 }
 
 int scanhash_lyra2rev2_16way( struct work *work, const uint32_t max_nonce,
@@ -186,8 +184,8 @@ int scanhash_lyra2rev2_16way( struct work *work, const uint32_t max_nonce,
    mm512_bswap32_intrlv80_16x32( vdata, pdata );
    *noncev = _mm512_set_epi32( n+15, n+14, n+13, n+12, n+11, n+10, n+ 9, n+ 8,
                                n+ 7, n+ 6, n+ 5, n+ 4, n+ 3, n+ 2, n+ 1, n );
-   blake256_16way_init( &l2v2_16way_ctx.blake );
+   blake256_16x32_init( &l2v2_16way_ctx.blake );
-   blake256_16way_update( &l2v2_16way_ctx.blake, vdata, 64 );
+   blake256_16x32_update( &l2v2_16way_ctx.blake, vdata, 64 );
 
    do
    {
@@ -214,21 +212,21 @@ int scanhash_lyra2rev2_16way( struct work *work, const uint32_t max_nonce,
 #elif defined (LYRA2REV2_8WAY)
 
 typedef struct {
-   blake256_8way_context     blake;
+   blake256_8x32_context     blake;
-   keccak256_4way_context    keccak;
+   keccak256_4x64_context    keccak;
    cubehashParam             cube;
-   skein256_4way_context     skein;
+   skein256_4x64_context     skein;
-   bmw256_8way_context       bmw;
+   bmw256_8x32_context       bmw;
 } lyra2v2_8way_ctx_holder __attribute__ ((aligned (64)));
 
 static lyra2v2_8way_ctx_holder l2v2_8way_ctx;
 
 bool init_lyra2rev2_8way_ctx()
 {
-   keccak256_4way_init( &l2v2_8way_ctx.keccak );
+   keccak256_4x64_init( &l2v2_8way_ctx.keccak );
    cubehashInit( &l2v2_8way_ctx.cube, 256, 16, 32 );
-   skein256_4way_init( &l2v2_8way_ctx.skein );
+   skein256_4x64_init( &l2v2_8way_ctx.skein );
-   bmw256_8way_init( &l2v2_8way_ctx.bmw );
+   bmw256_8x32_init( &l2v2_8way_ctx.bmw );
    return true;
 }
 
@@ -246,20 +244,20 @@ void lyra2rev2_8way_hash( void *state, const void *input )
    lyra2v2_8way_ctx_holder ctx __attribute__ ((aligned (64)));
    memcpy( &ctx, &l2v2_8way_ctx, sizeof(l2v2_8way_ctx) );
 
-   blake256_8way_update( &ctx.blake, input + (64<<3), 16 );
+   blake256_8x32_update( &ctx.blake, input + (64<<3), 16 );
-   blake256_8way_close( &ctx.blake, vhash );
+   blake256_8x32_close( &ctx.blake, vhash );
 
    dintrlv_8x32( hash0, hash1, hash2, hash3,
                  hash4, hash5, hash6, hash7, vhash, 256 );
 
    intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 256 );
-   keccak256_4way_update( &ctx.keccak, vhash, 32 );
+   keccak256_4x64_update( &ctx.keccak, vhash, 32 );
-   keccak256_4way_close( &ctx.keccak, vhash );
+   keccak256_4x64_close( &ctx.keccak, vhash );
    dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 256 );
    intrlv_4x64( vhash, hash4, hash5, hash6, hash7, 256 );
-   keccak256_4way_init( &ctx.keccak );
+   keccak256_4x64_init( &ctx.keccak );
-   keccak256_4way_update( &ctx.keccak, vhash, 32 );
+   keccak256_4x64_update( &ctx.keccak, vhash, 32 );
-   keccak256_4way_close( &ctx.keccak, vhash );
+   keccak256_4x64_close( &ctx.keccak, vhash );
    dintrlv_4x64( hash4, hash5, hash6, hash7, vhash, 256 );
 
    cubehash_full( &ctx.cube, (byte*) hash0, 256, (const byte*) hash0, 32 );
@@ -282,13 +280,13 @@ void lyra2rev2_8way_hash( void *state, const void *input )
    LYRA2REV2( l2v2_wholeMatrix, hash7, 32, hash7, 32, hash7, 32, 1, 4, 4 );
    
    intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 256 );
-   skein256_4way_update( &ctx.skein, vhash, 32 );
+   skein256_4x64_update( &ctx.skein, vhash, 32 );
-   skein256_4way_close( &ctx.skein, vhash );
+   skein256_4x64_close( &ctx.skein, vhash );
    dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 256 );
    intrlv_4x64( vhash, hash4, hash5, hash6, hash7, 256 );
-   skein256_4way_init( &ctx.skein );
+   skein256_4x64_init( &ctx.skein );
-   skein256_4way_update( &ctx.skein, vhash, 32 );
+   skein256_4x64_update( &ctx.skein, vhash, 32 );
-   skein256_4way_close( &ctx.skein, vhash );
+   skein256_4x64_close( &ctx.skein, vhash );
    dintrlv_4x64( hash4, hash5, hash6, hash7, vhash, 256 );
 
    cubehash_full( &ctx.cube, (byte*) hash0, 256, (const byte*) hash0, 32 );
@@ -303,8 +301,8 @@ void lyra2rev2_8way_hash( void *state, const void *input )
    intrlv_8x32( vhash, hash0, hash1, hash2, hash3,
                        hash4, hash5, hash6, hash7, 256 );
 
-   bmw256_8way_update( &ctx.bmw, vhash, 32 );
+   bmw256_8x32_update( &ctx.bmw, vhash, 32 );
-   bmw256_8way_close( &ctx.bmw, state );
+   bmw256_8x32_close( &ctx.bmw, state );
 }
 
 int scanhash_lyra2rev2_8way( struct work *work, const uint32_t max_nonce,
@@ -328,8 +326,8 @@ int scanhash_lyra2rev2_8way( struct work *work, const uint32_t max_nonce,
 
    mm256_bswap32_intrlv80_8x32( vdata, pdata );
    *noncev = _mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n );
-   blake256_8way_init( &l2v2_8way_ctx.blake );
+   blake256_8x32_init( &l2v2_8way_ctx.blake );
-   blake256_8way_update( &l2v2_8way_ctx.blake, vdata, 64 );
+   blake256_8x32_update( &l2v2_8way_ctx.blake, vdata, 64 );
 
    do
    {
@@ -356,21 +354,21 @@ int scanhash_lyra2rev2_8way( struct work *work, const uint32_t max_nonce,
 #elif defined (LYRA2REV2_4WAY)
 
 typedef struct {
-   blake256_4way_context     blake;
+   blake256_4x32_context     blake;
-   keccak256_4way_context    keccak;
+   keccak256_4x64_context    keccak;
    cubehashParam             cube;
-   skein256_4way_context     skein;
+   skein256_4x64_context     skein;
-   bmw256_4way_context          bmw;
+   bmw256_4x32_context          bmw;
 } lyra2v2_4way_ctx_holder;
 
 static lyra2v2_4way_ctx_holder l2v2_4way_ctx;
 
 bool init_lyra2rev2_4way_ctx()
 {
-   keccak256_4way_init( &l2v2_4way_ctx.keccak );
+   keccak256_4x64_init( &l2v2_4way_ctx.keccak );
    cubehashInit( &l2v2_4way_ctx.cube, 256, 16, 32 );
-   skein256_4way_init( &l2v2_4way_ctx.skein );
+   skein256_4x64_init( &l2v2_4way_ctx.skein );
-   bmw256_4way_init( &l2v2_4way_ctx.bmw );
+   bmw256_4x32_init( &l2v2_4way_ctx.bmw );
    return true;
 }
 
@@ -385,13 +383,13 @@ void lyra2rev2_4way_hash( void *state, const void *input )
    lyra2v2_4way_ctx_holder ctx __attribute__ ((aligned (64))); 
    memcpy( &ctx, &l2v2_4way_ctx, sizeof(l2v2_4way_ctx) );
 
-   blake256_4way_update( &ctx.blake, input + (64<<2), 16 );
+   blake256_4x32_update( &ctx.blake, input + (64<<2), 16 );
-   blake256_4way_close( &ctx.blake, vhash );
+   blake256_4x32_close( &ctx.blake, vhash );
 
    rintrlv_4x32_4x64( vhash64, vhash, 256 );
 
-   keccak256_4way_update( &ctx.keccak, vhash64, 32 );
+   keccak256_4x64_update( &ctx.keccak, vhash64, 32 );
-   keccak256_4way_close( &ctx.keccak, vhash64 );
+   keccak256_4x64_close( &ctx.keccak, vhash64 );
 
    dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
 
@@ -410,8 +408,8 @@ void lyra2rev2_4way_hash( void *state, const void *input )
 
    intrlv_4x64( vhash64, hash0, hash1, hash2, hash3, 256 );
 
-   skein256_4way_update( &ctx.skein, vhash64, 32 );
+   skein256_4x64_update( &ctx.skein, vhash64, 32 );
-   skein256_4way_close( &ctx.skein, vhash64 );
+   skein256_4x64_close( &ctx.skein, vhash64 );
 
    dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
 
@@ -426,8 +424,8 @@ void lyra2rev2_4way_hash( void *state, const void *input )
 
    intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 256 );
 
-   bmw256_4way_update( &ctx.bmw, vhash, 32 );
+   bmw256_4x32_update( &ctx.bmw, vhash, 32 );
-   bmw256_4way_close( &ctx.bmw, state );
+   bmw256_4x32_close( &ctx.bmw, state );
 }
 
 int scanhash_lyra2rev2_4way( struct work *work, uint32_t max_nonce,
@@ -451,8 +449,8 @@ int scanhash_lyra2rev2_4way( struct work *work, uint32_t max_nonce,
 
    v128_bswap32_intrlv80_4x32( vdata, pdata );
 
-   blake256_4way_init( &l2v2_4way_ctx.blake );
+   blake256_4x32_init( &l2v2_4way_ctx.blake );
-   blake256_4way_update( &l2v2_4way_ctx.blake, vdata, 64 );
+   blake256_4x32_update( &l2v2_4way_ctx.blake, vdata, 64 );
 
    do
    {

algo/lyra2/lyra2rev3-4way.c

@@ -9,18 +9,18 @@
 #if defined (LYRA2REV3_16WAY)
 
 typedef struct {
-   blake256_16way_context     blake;
+   blake256_16x32_context     blake;
    cube_4way_context          cube;
-   bmw256_16way_context       bmw;
+   bmw256_16x32_context       bmw;
 } lyra2v3_16way_ctx_holder;
 
 static __thread lyra2v3_16way_ctx_holder l2v3_16way_ctx;
 
 bool init_lyra2rev3_16way_ctx()
 {
-   blake256_16way_init( &l2v3_16way_ctx.blake );
+   blake256_16x32_init( &l2v3_16way_ctx.blake );
    cube_4way_init( &l2v3_16way_ctx.cube, 256, 16, 32 );
-   bmw256_16way_init( &l2v3_16way_ctx.bmw );
+   bmw256_16x32_init( &l2v3_16way_ctx.bmw );
    return true;
 }
 
@@ -46,8 +46,8 @@ void lyra2rev3_16way_hash( void *state, const void *input )
    lyra2v3_16way_ctx_holder ctx __attribute__ ((aligned (64)));
    memcpy( &ctx, &l2v3_16way_ctx, sizeof(l2v3_16way_ctx) );
 
-   blake256_16way_update( &ctx.blake, input + (64*16), 16 );
+   blake256_16x32_update( &ctx.blake, input + (64*16), 16 );
-   blake256_16way_close( &ctx.blake, vhash );
+   blake256_16x32_close( &ctx.blake, vhash );
 
    dintrlv_16x32( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
            hash8, hash9, hash10, hash11 ,hash12, hash13, hash14, hash15,
@@ -120,8 +120,8 @@ void lyra2rev3_16way_hash( void *state, const void *input )
              hash7, hash8, hash9, hash10, hash11, hash12, hash13, hash14,
              hash15, 256 );
 
-   bmw256_16way_update( &ctx.bmw, vhash, 32 );
+   bmw256_16x32_update( &ctx.bmw, vhash, 32 );
-   bmw256_16way_close( &ctx.bmw, state );
+   bmw256_16x32_close( &ctx.bmw, state );
 }
 
 
@@ -145,8 +145,8 @@ int scanhash_lyra2rev3_16way( struct work *work, const uint32_t max_nonce,
 
    mm512_bswap32_intrlv80_16x32( vdata, pdata );
 
-   blake256_16way_init( &l2v3_16way_ctx.blake );
+   blake256_16x32_init( &l2v3_16way_ctx.blake );
-   blake256_16way_update( &l2v3_16way_ctx.blake, vdata, 64 );
+   blake256_16x32_update( &l2v3_16way_ctx.blake, vdata, 64 );
 
    do
    {
@@ -178,18 +178,18 @@ int scanhash_lyra2rev3_16way( struct work *work, const uint32_t max_nonce,
 #elif defined (LYRA2REV3_8WAY)
 
 typedef struct {
-   blake256_8way_context     blake;
+   blake256_8x32_context     blake;
    cubehashParam             cube;
-   bmw256_8way_context       bmw;
+   bmw256_8x32_context       bmw;
 } lyra2v3_8way_ctx_holder;
 
 static __thread lyra2v3_8way_ctx_holder l2v3_8way_ctx;
 
 bool init_lyra2rev3_8way_ctx()
 {
-   blake256_8way_init( &l2v3_8way_ctx.blake );
+   blake256_8x32_init( &l2v3_8way_ctx.blake );
    cubehashInit( &l2v3_8way_ctx.cube, 256, 16, 32 );
-   bmw256_8way_init( &l2v3_8way_ctx.bmw );
+   bmw256_8x32_init( &l2v3_8way_ctx.bmw );
    return true;
 }
 
@@ -207,8 +207,8 @@ void lyra2rev3_8way_hash( void *state, const void *input )
    lyra2v3_8way_ctx_holder ctx __attribute__ ((aligned (64)));
    memcpy( &ctx, &l2v3_8way_ctx, sizeof(l2v3_8way_ctx) );
 
-   blake256_8way_update( &ctx.blake, input + (64*8), 16 );
+   blake256_8x32_update( &ctx.blake, input + (64*8), 16 );
-   blake256_8way_close( &ctx.blake, vhash );
+   blake256_8x32_close( &ctx.blake, vhash );
 
    dintrlv_8x32( hash0, hash1, hash2, hash3,
                        hash4, hash5, hash6, hash7, vhash, 256 );
@@ -243,8 +243,8 @@ void lyra2rev3_8way_hash( void *state, const void *input )
    intrlv_8x32( vhash, hash0, hash1, hash2, hash3,
                              hash4, hash5, hash6, hash7, 256 );
 
-   bmw256_8way_update( &ctx.bmw, vhash, 32 );
+   bmw256_8x32_update( &ctx.bmw, vhash, 32 );
-   bmw256_8way_close( &ctx.bmw, state );
+   bmw256_8x32_close( &ctx.bmw, state );
 
    }
 
@@ -269,8 +269,8 @@ int scanhash_lyra2rev3_8way( struct work *work, const uint32_t max_nonce,
 
    mm256_bswap32_intrlv80_8x32( vdata, pdata );
    *noncev = _mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n );
-   blake256_8way_init( &l2v3_8way_ctx.blake );
+   blake256_8x32_init( &l2v3_8way_ctx.blake );
-   blake256_8way_update( &l2v3_8way_ctx.blake, vdata, 64 );
+   blake256_8x32_update( &l2v3_8way_ctx.blake, vdata, 64 );
 
    do
    {
@@ -300,19 +300,18 @@ int scanhash_lyra2rev3_8way( struct work *work, const uint32_t max_nonce,
 #if defined (LYRA2REV3_4WAY)  
 
 typedef struct {
-   blake256_4way_context     blake;
+   blake256_4x32_context     blake;
    cubehashParam             cube;
-   bmw256_4way_context       bmw;
+   bmw256_4x32_context       bmw;
 } lyra2v3_4way_ctx_holder;
 
-//static lyra2v3_4way_ctx_holder l2v3_4way_ctx;
 static __thread lyra2v3_4way_ctx_holder l2v3_4way_ctx;
 
 bool init_lyra2rev3_4way_ctx()
 {
-   blake256_4way_init( &l2v3_4way_ctx.blake );
+   blake256_4x32_init( &l2v3_4way_ctx.blake );
    cubehashInit( &l2v3_4way_ctx.cube, 256, 16, 32 );
-   bmw256_4way_init( &l2v3_4way_ctx.bmw );
+   bmw256_4x32_init( &l2v3_4way_ctx.bmw );
    return true;
 }
 
@@ -326,8 +325,8 @@ void lyra2rev3_4way_hash( void *state, const void *input )
    lyra2v3_4way_ctx_holder ctx __attribute__ ((aligned (64))); 
    memcpy( &ctx, &l2v3_4way_ctx, sizeof(l2v3_4way_ctx) );
 
-   blake256_4way_update( &ctx.blake, input + (64*4), 16 );
+   blake256_4x32_update( &ctx.blake, input + (64*4), 16 );
-   blake256_4way_close( &ctx.blake, vhash );
+   blake256_4x32_close( &ctx.blake, vhash );
    dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
 
    LYRA2REV3( l2v3_wholeMatrix, hash0, 32, hash0, 32, hash0, 32, 1, 4, 4 );
@@ -349,8 +348,8 @@ void lyra2rev3_4way_hash( void *state, const void *input )
    LYRA2REV3( l2v3_wholeMatrix, hash3, 32, hash3, 32, hash3, 32, 1, 4, 4 );
 
    intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 256 );
-   bmw256_4way_update( &ctx.bmw, vhash, 32 );
+   bmw256_4x32_update( &ctx.bmw, vhash, 32 );
-   bmw256_4way_close( &ctx.bmw, state );
+   bmw256_4x32_close( &ctx.bmw, state );
 }
 
 int scanhash_lyra2rev3_4way( struct work *work, const uint32_t max_nonce,
@@ -374,8 +373,8 @@ int scanhash_lyra2rev3_4way( struct work *work, const uint32_t max_nonce,
    v128_bswap32_intrlv80_4x32( vdata, pdata );
    *noncev = _mm_set_epi32( n+3, n+2, n+1, n );
 
-   blake256_4way_init( &l2v3_4way_ctx.blake );
+   blake256_4x32_init( &l2v3_4way_ctx.blake );
-   blake256_4way_update( &l2v3_4way_ctx.blake, vdata, 64 );
+   blake256_4x32_update( &l2v3_4way_ctx.blake, vdata, 64 );
 
    do
    {

algo/lyra2/lyra2z-4way.c

@@ -3,7 +3,7 @@
 #include "lyra2.h"
 #include "algo/blake/blake256-hash.h"
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define LYRA2Z_16WAY 1
 #elif defined(__AVX2__)
   #define LYRA2Z_8WAY 1
@@ -45,7 +45,7 @@ static void lyra2z_16way_hash( void *state, const void *midstate_vars,
     uint32_t hash14[8] __attribute__ ((aligned (32)));
     uint32_t hash15[8] __attribute__ ((aligned (32)));
 
-    blake256_16way_final_rounds_le( vhash, midstate_vars, midhash, block, 14 );
+    blake256_16x32_final_rounds_le( vhash, midstate_vars, midhash, block, 14 );
 
     dintrlv_16x32( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
               hash8, hash9, hash10, hash11 ,hash12, hash13, hash14, hash15,
@@ -139,7 +139,7 @@ int scanhash_lyra2z_16way( struct work *work, uint32_t max_nonce,
                                n+ 7, n+ 6, n+ 5, n+ 4, n+ 3, n+ 2, n +1, n );
 
    // Partialy prehash second block without touching nonces in block_buf[3].
-   blake256_16way_round0_prehash_le( midstate_vars, block0_hash, block_buf );
+   blake256_16x32_round0_prehash_le( midstate_vars, block0_hash, block_buf );
 
    do {
      lyra2z_16way_hash( hash, midstate_vars, block0_hash, block_buf );
@@ -180,7 +180,7 @@ static void lyra2z_8way_hash( void *state, const void *midstate_vars,
      uint32_t hash7[8] __attribute__ ((aligned (32)));
      uint32_t vhash[8*8] __attribute__ ((aligned (64)));
 
-     blake256_8way_final_rounds_le( vhash, midstate_vars, midhash, block, 14 );
+     blake256_8x32_final_rounds_le( vhash, midstate_vars, midhash, block, 14 );
 
      dintrlv_8x32( hash0, hash1, hash2, hash3,
                    hash4, hash5, hash6, hash7, vhash, 256 );
@@ -246,7 +246,7 @@ int scanhash_lyra2z_8way( struct work *work, uint32_t max_nonce,
             _mm256_set_epi32( n+ 7, n+ 6, n+ 5, n+ 4, n+ 3, n+ 2, n +1, n );
 
    // Partialy prehash second block without touching nonces
-   blake256_8way_round0_prehash_le( midstate_vars, block0_hash, block_buf );
+   blake256_8x32_round0_prehash_le( midstate_vars, block0_hash, block_buf );
 
    do {
      lyra2z_8way_hash( hash, midstate_vars, block0_hash, block_buf );
@@ -279,12 +279,12 @@ bool lyra2z_4way_thread_init()
  return ( lyra2z_4way_matrix = mm_malloc( LYRA2Z_MATRIX_SIZE, 64 ) );
 }
 
-static __thread blake256_4way_context l2z_4way_blake_mid;
+static __thread blake256_4x32_context l2z_4way_blake_mid;
 
 void lyra2z_4way_midstate( const void* input )
 {
-       blake256_4way_init( &l2z_4way_blake_mid );
+       blake256_4x32_init( &l2z_4way_blake_mid );
-       blake256_4way_update( &l2z_4way_blake_mid, input, 64 );
+       blake256_4x32_update( &l2z_4way_blake_mid, input, 64 );
 }
 
 void lyra2z_4way_hash( void *hash, const void *midstate_vars,
@@ -295,15 +295,8 @@ void lyra2z_4way_hash( void *hash, const void *midstate_vars,
      uint32_t hash2[8] __attribute__ ((aligned (64)));
      uint32_t hash3[8] __attribute__ ((aligned (64)));
      uint32_t vhash[8*4] __attribute__ ((aligned (64)));
-//     blake256_4way_context ctx_blake __attribute__ ((aligned (64)));
 
-     blake256_4way_final_rounds_le( vhash, midstate_vars, midhash, block, 14 );
+     blake256_4x32_final_rounds_le( vhash, midstate_vars, midhash, block, 14 );
-
-/*
-     memcpy( &ctx_blake, &l2z_4way_blake_mid, sizeof l2z_4way_blake_mid );
-     blake256_4way_update( &ctx_blake, input + (64*4), 16 );
-     blake256_4way_close( &ctx_blake, vhash );
-*/
 
      dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
 
@@ -357,7 +350,7 @@ int scanhash_lyra2z_4way( struct work *work, uint32_t max_nonce,
    block_buf[15] = v128_32( 640 );
 
    // Partialy prehash second block without touching nonces
-   blake256_4way_round0_prehash_le( midstate_vars, block0_hash, block_buf );
+   blake256_4x32_round0_prehash_le( midstate_vars, block0_hash, block_buf );
 
    do {
       lyra2z_4way_hash( hash, midstate_vars, block0_hash, block_buf );
@@ -454,11 +447,9 @@ bool register_lyra2z_algo( algo_gate_t* gate )
 #if defined(LYRA2Z_16WAY)
   gate->miner_thread_init = (void*)&lyra2z_16way_thread_init;
   gate->scanhash          = (void*)&scanhash_lyra2z_16way;
-//  gate->hash       = (void*)&lyra2z_16way_hash;
 #elif defined(LYRA2Z_8WAY)
   gate->miner_thread_init = (void*)&lyra2z_8way_thread_init;
   gate->scanhash          = (void*)&scanhash_lyra2z_8way;
-//  gate->hash       = (void*)&lyra2z_8way_hash;
 #elif defined(LYRA2Z_4WAY)
   gate->miner_thread_init = (void*)&lyra2z_4way_thread_init;
   gate->scanhash          = (void*)&scanhash_lyra2z_4way;

algo/lyra2/phi2-4way.c

@@ -4,7 +4,7 @@
 #include "algo/gost/sph_gost.h"
 #include "algo/cubehash/cubehash_sse2.h"
 #include "lyra2.h"
-#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(__VAES__) && defined(SIMD512)
   #include "algo/echo/echo-hash-4way.h"
 #elif defined(__AES__)
   #include "algo/echo/aes_ni/hash_api.h"

algo/lyra2/sponge-2way.c

@@ -27,7 +27,7 @@
 #include "lyra2.h"
 #include "simd-utils.h"
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 inline void squeeze_2way( uint64_t *State, byte *Out, unsigned int len )
 {

algo/lyra2/sponge.h

@@ -43,9 +43,9 @@ static const uint64_t blake2b_IV[8] =
   0x1f83d9abfb41bd6bULL, 0x5be0cd19137e2179ULL
 };
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
-#define G2W_4X64(a,b,c,d) \
+#define G2W(a,b,c,d) \
    a = _mm512_add_epi64( a, b ); \
    d = _mm512_ror_epi64( _mm512_xor_si512( d, a ), 32 ); \
    c = _mm512_add_epi64( c, d ); \
@@ -56,27 +56,15 @@ static const uint64_t blake2b_IV[8] =
    b = _mm512_ror_epi64( _mm512_xor_si512( b, c ), 63 );
 
 #define LYRA_ROUND_2WAY_AVX512( s0, s1, s2, s3 ) \
-   G2W_4X64( s0, s1, s2, s3 ); \
+   G2W( s0, s1, s2, s3 ); \
    s0 = mm512_shufll256_64( s0 ); \
    s3 = mm512_swap256_128( s3 ); \
    s2 = mm512_shuflr256_64( s2 ); \
-   G2W_4X64( s0, s1, s2, s3 ); \
+   G2W( s0, s1, s2, s3 ); \
    s0 = mm512_shuflr256_64( s0 ); \
    s3 = mm512_swap256_128( s3 ); \
    s2 = mm512_shufll256_64( s2 ); 
 
-/*
-#define LYRA_ROUND_2WAY_AVX512( s0, s1, s2, s3 ) \
-   G2W_4X64( s0, s1, s2, s3 ); \
-   s3 = mm512_shufll256_64( s3 ); \
-   s1 = mm512_shuflr256_64( s1); \
-   s2 = mm512_swap256_128( s2 ); \
-   G2W_4X64( s0, s1, s2, s3 ); \
-   s3 = mm512_shuflr256_64( s3 ); \
-   s1 = mm512_shufll256_64( s1 ); \
-   s2 = mm512_swap256_128( s2 ); 
-*/
-
 #define LYRA_12_ROUNDS_2WAY_AVX512( s0, s1, s2, s3 ) \
    LYRA_ROUND_2WAY_AVX512( s0, s1, s2, s3 ) \
    LYRA_ROUND_2WAY_AVX512( s0, s1, s2, s3 ) \
@@ -95,7 +83,7 @@ static const uint64_t blake2b_IV[8] =
 
 #if defined(__AVX2__)
 
-#define G_4X64(a,b,c,d) \
+#define G_AVX2(a,b,c,d) \
    a = _mm256_add_epi64( a, b ); \
    d = mm256_ror_64( _mm256_xor_si256( d, a ), 32 ); \
    c = _mm256_add_epi64( c, d ); \
@@ -107,27 +95,15 @@ static const uint64_t blake2b_IV[8] =
 
 // Pivot about s1 instead of s0 reduces latency.
 #define LYRA_ROUND_AVX2( s0, s1, s2, s3 ) \
-   G_4X64( s0, s1, s2, s3 ); \
+   G_AVX2( s0, s1, s2, s3 ); \
    s0 = mm256_shufll_64( s0 ); \
    s3 = mm256_swap_128( s3 ); \
    s2 = mm256_shuflr_64( s2 ); \
-   G_4X64( s0, s1, s2, s3 ); \
+   G_AVX2( s0, s1, s2, s3 ); \
    s0 = mm256_shuflr_64( s0 ); \
    s3 = mm256_swap_128( s3 ); \
    s2 = mm256_shufll_64( s2 );
 
-/*
-#define LYRA_ROUND_AVX2( s0, s1, s2, s3 ) \
-   G_4X64( s0, s1, s2, s3 ); \
-   s3 = mm256_shufll_64( s3 ); \
-   s1 = mm256_shuflr_64( s1); \
-   s2 = mm256_swap_128( s2 ); \
-   G_4X64( s0, s1, s2, s3 ); \
-   s3 = mm256_shuflr_64( s3 ); \
-   s1 = mm256_shufll_64( s1 ); \
-   s2 = mm256_swap_128( s2 );
-*/
-
 #define LYRA_12_ROUNDS_AVX2( s0, s1, s2, s3 ) \
    LYRA_ROUND_AVX2( s0, s1, s2, s3 ) \
    LYRA_ROUND_AVX2( s0, s1, s2, s3 ) \
@@ -148,29 +124,29 @@ static const uint64_t blake2b_IV[8] =
 
 // process 2 columns in parallel
 // returns void, all args updated
-#define G_2X64(a,b,c,d) \
+#define G_128(a,b,c,d) \
    a = v128_add64( a, b ); \
-   d = v128_ror64( v128_xor( d, a), 32 ); \
+   d = v128_ror64xor( d, a, 32 ); \
    c = v128_add64( c, d ); \
-   b = v128_ror64( v128_xor( b, c ), 24 ); \
+   b = v128_ror64xor( b, c, 24 ); \
    a = v128_add64( a, b ); \
-   d = v128_ror64( v128_xor( d, a ), 16 ); \
+   d = v128_ror64xor( d, a, 16 ); \
    c = v128_add64( c, d ); \
-   b = v128_ror64( v128_xor( b, c ), 63 );
+   b = v128_ror64xor( b, c, 63 );
 
 #define LYRA_ROUND_AVX(s0,s1,s2,s3,s4,s5,s6,s7) \
 { \
    v128u64_t t; \
-   G_2X64( s0, s2, s4, s6 ); \
+   G_128( s0, s2, s4, s6 ); \
-   G_2X64( s1, s3, s5, s7 ); \
+   G_128( s1, s3, s5, s7 ); \
    t =  v128_alignr64( s7, s6, 1 ); \
    s6 = v128_alignr64( s6, s7, 1 ); \
    s7 = t; \
    t =  v128_alignr64( s2, s3, 1 ); \
    s2 = v128_alignr64( s3, s2, 1 ); \
    s3 = t; \
-   G_2X64( s0, s2, s5, s6 ); \
+   G_128( s0, s2, s5, s6 ); \
-   G_2X64( s1, s3, s4, s7 ); \
+   G_128( s1, s3, s4, s7 ); \
    t =  v128_alignr64( s6, s7, 1 ); \
    s6 = v128_alignr64( s7, s6, 1 ); \
    s7 = t; \
@@ -195,10 +171,6 @@ static const uint64_t blake2b_IV[8] =
 
 #endif // AVX2 else SSE2
 
-static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
-    return ( w >> c ) | ( w << ( 64 - c ) );
-}
-
 #define G( r, i, a, b, c, d ) \
 { \
     a = a + b; \
@@ -222,7 +194,7 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
     G( r, 7, v[ 3], v[ 4], v[ 9], v[14] );
 
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 union _ovly_512
 {

algo/m7m/m7m.c

@@ -1,8 +1,6 @@
 #include "cpuminer-config.h"
 #include "algo-gate-api.h"
 
-#if !defined(__APPLE__)
-
 #include <gmp.h>
 #include <stdbool.h>
 #include <stdlib.h>
@@ -21,7 +19,7 @@
 #define EPS1 DBL_EPSILON
 #define EPS2 3.0e-11
 
-inline double exp_n( double xt )
+static inline double exp_n( double xt )
 {
     if ( xt < -700.0 )
         return 0;
@@ -33,7 +31,8 @@ inline double exp_n( double xt )
         return exp( xt );
 }
 
-inline double exp_n2( double x1, double x2 )
+/*
+static inline double exp_n2( double x1, double x2 )
 {
     double p1 = -700., p2 = -37., p3 = -0.8e-8, p4 = 0.8e-8,
            p5 = 37., p6 = 700.;
@@ -53,6 +52,7 @@ inline double exp_n2( double x1, double x2 )
     else if ( xt > p6 - 1.e-200 )
         return 0.;
 }
+*/
 
 double swit2_( double wvnmb )
 {
@@ -298,15 +298,9 @@ int scanhash_m7m_hash( struct work* work, uint64_t max_nonce,
     return 0;
 }
 
-#endif   // not apple
-
 bool register_m7m_algo( algo_gate_t *gate )
 {
-#if defined(__APPLE__)
+  gate->optimizations = SHA256_OPT;
-  applog( LOG_ERR, "M7M algo is not supported on MacOS");
-  return false;
-#else  
-  gate->optimizations = SHA_OPT;
   init_m7m_ctx();
   gate->scanhash              = (void*)&scanhash_m7m_hash;
   gate->build_stratum_request = (void*)&std_be_build_stratum_request;
@@ -315,6 +309,5 @@ bool register_m7m_algo( algo_gate_t *gate )
   gate->set_work_data_endian  = (void*)&set_work_data_big_endian;
   opt_target_factor = 65536.0;
   return true;
-#endif
 }
 

algo/m7m/magimath.cpp

@@ -1,75 +0,0 @@
-// Copyright (c) 2014 The Magi developers
-// Distributed under the MIT/X11 software license, see the accompanying
-// file COPYING or http://www.opensource.org/licenses/mit-license.php.
-
-#include <iostream>
-#include <cfloat>
-#include <limits>
-#include <math.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <stdint.h>
-
-#include "magimath.h"
-
-#define EPS1 (std::numeric_limits<double>::epsilon())
-#define EPS2 3.0e-11
-
-static void gauleg(double x1, double x2, double x[], double w[], const int n)
-{
-	int m,j,i;
-	double z1, z, xm, xl, pp, p3, p2, p1;
-	m=(n+1)/2;
-	xm=0.5*(x2+x1);
-	xl=0.5*(x2-x1);
-	for (i=1;i<=m;i++) {
-		z=cos(3.141592654*(i-0.25)/(n+0.5));
-		do {
-			p1=1.0;
-			p2=0.0;
-			for (j=1;j<=n;j++) {
-				p3=p2;
-				p2=p1;
-				p1=((2.0*j-1.0)*z*p2-(j-1.0)*p3)/j;
-			}
-			pp=n*(z*p1-p2)/(z*z-1.0);
-			z1=z;
-			z=z1-p1/pp;
-		} while (fabs(z-z1) > EPS2);
-		x[i]=xm-xl*z;
-		x[n+1-i]=xm+xl*z;
-		w[i]=2.0*xl/((1.0-z*z)*pp*pp);
-		w[n+1-i]=w[i];
-	}
-}
-
-static double GaussianQuad_N(double func(const double), const double a2, const double b2, const int NptGQ)
-{
-	double s=0.0;
-#ifdef _MSC_VER
-#define SW_DIVS 23
-	double x[SW_DIVS+1], w[SW_DIVS+1];
-#else
-	double x[NptGQ+1], w[NptGQ+1];
-#endif
-
-	gauleg(a2, b2, x, w, NptGQ);
-
-	for (int j=1; j<=NptGQ; j++) {
-		s += w[j]*func(x[j]);
-	}
-
-	return s;
-}
-
-static double swit_(double wvnmb)
-{
-	return pow( (5.55243*(exp_n(-0.3*wvnmb/15.762) - exp_n(-0.6*wvnmb/15.762)))*wvnmb, 0.5)
-		/ 1034.66 * pow(sin(wvnmb/65.), 2.);
-}
-
-uint32_t sw_(int nnounce, int divs)
-{
-	double wmax = ((sqrt((double)(nnounce))*(1.+EPS1))/450+100);
-	return ((uint32_t)(GaussianQuad_N(swit_, 0., wmax, divs)*(1.+EPS1)*1.e6));
-}

algo/m7m/magimath.h

@@ -1,54 +0,0 @@
-// Copyright (c) 2014 The Magi developers
-// Distributed under the MIT/X11 software license, see the accompanying
-// file COPYING or http://www.opensource.org/licenses/mit-license.php.
-#ifndef MAGI_MATH_H
-#define MAGI_MATH_H
-
-#include <math.h>
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-uint32_t sw_(int nnounce, int divs);
-
-#ifdef __cplusplus
-}
-#endif
-
-
-inline double exp_n(double xt)
-{
-	double p1 = -700.0, p3 = -0.8e-8, p4 = 0.8e-8, p6 = 700.0;
-	if(xt < p1)
-		return 0;
-	else if(xt > p6)
-		return 1e200;
-	else if(xt > p3 && xt < p4)
-		return (1.0 + xt);
-	else
-		return exp(xt);
-}
-
-// 1 / (1 + exp(x1-x2))
-inline double exp_n2(double x1, double x2)
-{
-	double p1 = -700., p2 = -37., p3 = -0.8e-8, p4 = 0.8e-8, p5 = 37., p6 = 700.;
-	double xt = x1 - x2;
-	if (xt < p1+1.e-200)
-		return 1.;
-	else if (xt > p1 && xt < p2 + 1.e-200)
-		return ( 1. - exp(xt) );
-	else if (xt > p2 && xt < p3 + 1.e-200)
-		return ( 1. / (1. + exp(xt)) );
-	else if (xt > p3 && xt < p4)
-		return ( 1. / (2. + xt) );
-	else if (xt > p4 - 1.e-200 && xt < p5)
-		return ( exp(-xt) / (1. + exp(-xt)) );
-	else if (xt > p5 - 1.e-200 && xt < p6)
-		return ( exp(-xt) );
-	else //if (xt > p6 - 1.e-200)
-		return 0.;
-}
-
-#endif

algo/nist5/nist5-gate.h

@@ -4,7 +4,7 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define NIST5_8WAY 1
 #elif defined(__AVX2__) && defined(__AES__)
   #define NIST5_4WAY 1

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

@@ -71,8 +71,7 @@ do { \
 } while (0)
 
 #define GAMMA_4W(n0, n1, n2, n4)   \
-   (g ## n0 = v128_xor( a ## n0, \
+   (g ## n0 = v128_xor( a ## n0, v128_ornot( a ## n2, a ## n1 ) ) )
-                             v128_or( a ## n1, v128_not( a ## n2 ) ) ) )
 
 #define PI_ALL_4W   do { \
       a0  = g0; \
@@ -312,7 +311,7 @@ do { \
       BUPDATE1_8W( 7, 1 ); \
 } while (0)
 
-#if defined(__AVX512VL__)
+#if defined(VL256)
 
 #define GAMMA_8W(n0, n1, n2, n4)   \
    ( g ## n0 = _mm256_ternarylogic_epi32( a ## n0, a ## n2, a ## n1, 0x4b ) )  

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

@@ -18,11 +18,14 @@ typedef struct {
 } panama_4way_context __attribute__ ((aligned (64)));
 
 void panama_4way_init( void *cc );
-
 void panama_4way_update( void *cc, const void *data, size_t len );
-
 void panama_4way_close( void *cc, void *dst );
 
+#define panama_4x32_context panama_4way_context
+#define panama_4x32_init    panama_4way_init
+#define panama_4x32_update  panama_4way_update
+#define panama_4x32_close   panama_4way_close
+
 #if defined(__AVX2__)
 
 typedef struct {
@@ -34,10 +37,13 @@ typedef struct {
 } panama_8way_context __attribute__ ((aligned (128)));
 
 void panama_8way_init( void *cc );
-
 void panama_8way_update( void *cc, const void *data, size_t len );
-
 void panama_8way_close( void *cc, void *dst );
 
+#define panama_8x32_context panama_8way_context
+#define panama_8x32_init    panama_8way_init
+#define panama_8x32_update  panama_8way_update
+#define panama_8x32_close   panama_8way_close
+
 #endif
 #endif

algo/quark/anime-gate.h

@@ -4,7 +4,7 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define ANIME_8WAY 1
 #elif defined(__AVX2__) && defined(__AES__)
   #define ANIME_4WAY 1

algo/quark/hmq1725-4way.c

@@ -11,7 +11,6 @@
 #include "algo/luffa/luffa-hash-2way.h"
 #include "algo/cubehash/cube-hash-2way.h"
 #include "algo/cubehash/cubehash_sse2.h"
-#include "algo/simd/nist.h"
 #include "algo/shavite/sph_shavite.h"
 #include "algo/shavite/shavite-hash-2way.h"
 #include "algo/simd/simd-hash-2way.h"
@@ -32,20 +31,20 @@
 
 union _hmq1725_8way_context_overlay
 {
-    blake512_8way_context   blake;
+    blake512_8x64_context   blake;
-    bmw512_8way_context     bmw;
+    bmw512_8x64_context     bmw;
-    skein512_8way_context   skein;
+    skein512_8x64_context   skein;
-    jh512_8way_context      jh;
+    jh512_8x64_context      jh;
-    keccak512_8way_context  keccak;
+    keccak512_8x64_context  keccak;
     luffa_4way_context      luffa;
     cube_4way_context       cube;
     simd_4way_context       simd;
-    hamsi512_8way_context   hamsi;
+    hamsi512_8x64_context   hamsi;
     hashState_fugue         fugue;
-    shabal512_8way_context  shabal;
+    shabal512_8x32_context  shabal;
     sph_whirlpool_context   whirlpool;
-    sha512_8way_context     sha512;
+    sha512_8x64_context     sha512;
-    haval256_5_8way_context haval;
+    haval256_8x32_context   haval;
 #if defined(__VAES__)
     groestl512_4way_context groestl;
     shavite512_4way_context shavite;
@@ -82,7 +81,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
    __m512i* vhB = (__m512i*)vhashB;
    __m512i* vhC = (__m512i*)vhashC;
 
-   bmw512_8way_full( &ctx.bmw, vhash, input, 80 );
+   bmw512_8x64_full( &ctx.bmw, vhash, input, 80 );
 
    dintrlv_8x64_512( hash0, hash1, hash2, hash3,
                      hash4, hash5, hash6, hash7, vhash );
@@ -142,26 +141,26 @@ extern void hmq1725_8way_hash(void *state, const void *input)
 
    // B
    if ( likely( vh_mask & 0xff ) )
-       skein512_8way_full( &ctx.skein, vhashB, vhash, 64 );
+       skein512_8x64_full( &ctx.skein, vhashB, vhash, 64 );
 
    mm512_blend_hash_8x64( vh, vhC, vhB, vh_mask );
 
-   jh512_8way_init( &ctx.jh );
+   jh512_8x64_init( &ctx.jh );
-   jh512_8way_update( &ctx.jh, vhash, 64 );
+   jh512_8x64_update( &ctx.jh, vhash, 64 );
-   jh512_8way_close( &ctx.jh, vhash );
+   jh512_8x64_close( &ctx.jh, vhash );
 
-   keccak512_8way_init( &ctx.keccak );
+   keccak512_8x64_init( &ctx.keccak );
-   keccak512_8way_update( &ctx.keccak, vhash, 64 );
+   keccak512_8x64_update( &ctx.keccak, vhash, 64 );
-   keccak512_8way_close( &ctx.keccak, vhash );
+   keccak512_8x64_close( &ctx.keccak, vhash );
 
    vh_mask = _mm512_testn_epi64_mask( vh[0], vmask );
 
    // A
    if ( ( vh_mask & 0xff ) != 0xff )
-       blake512_8way_full( &ctx.blake, vhashA, vhash, 64 );
+       blake512_8x64_full( &ctx.blake, vhashA, vhash, 64 );
    // B
    if ( vh_mask & 0xff )
-       bmw512_8way_full( &ctx.bmw, vhashB, vhash, 64 );
+       bmw512_8x64_full( &ctx.bmw, vhashB, vhash, 64 );
 
    mm512_blend_hash_8x64( vh, vhA, vhB, vh_mask );
    rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
@@ -177,16 +176,16 @@ extern void hmq1725_8way_hash(void *state, const void *input)
 
    if ( likely( ( vh_mask & 0xff ) != 0xff ) )
    {
-      keccak512_8way_init( &ctx.keccak );
+      keccak512_8x64_init( &ctx.keccak );
-      keccak512_8way_update( &ctx.keccak, vhash, 64 );
+      keccak512_8x64_update( &ctx.keccak, vhash, 64 );
-      keccak512_8way_close( &ctx.keccak, vhashA );
+      keccak512_8x64_close( &ctx.keccak, vhashA );
    }
 
    if ( likely( vh_mask & 0xff ) )
    {
-      jh512_8way_init( &ctx.jh );
+      jh512_8x64_init( &ctx.jh );
-      jh512_8way_update( &ctx.jh, vhash, 64 );
+      jh512_8x64_update( &ctx.jh, vhash, 64 );
-      jh512_8way_close( &ctx.jh, vhashB );
+      jh512_8x64_close( &ctx.jh, vhashB );
    }
 
    mm512_blend_hash_8x64( vh, vhA, vhB, vh_mask );
@@ -252,9 +251,9 @@ extern void hmq1725_8way_hash(void *state, const void *input)
    // B
    if ( likely( vh_mask & 0xff ) )
    {
-      haval256_5_8way_init( &ctx.haval );
+      haval256_8x32_init( &ctx.haval );
-      haval256_5_8way_update( &ctx.haval, vhash, 64 );
+      haval256_8x32_update( &ctx.haval, vhash, 64 );
-      haval256_5_8way_close( &ctx.haval, vhash );
+      haval256_8x32_close( &ctx.haval, vhash );
       memset( &vhash[8<<3], 0, 32<<3 );
       rintrlv_8x32_8x64( vhashB, vhash, 512 );
    }
@@ -297,7 +296,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
 
 #endif
 
-   blake512_8way_full( &ctx.blake, vhash, vhash, 64 );
+   blake512_8x64_full( &ctx.blake, vhash, vhash, 64 );
 
    vh_mask = _mm512_testn_epi64_mask( vh[0], vmask );
 
@@ -352,9 +351,9 @@ extern void hmq1725_8way_hash(void *state, const void *input)
 
    mm512_blend_hash_8x64( vh, vhC, vhB, vh_mask );
 
-   hamsi512_8way_init( &ctx.hamsi );
+   hamsi512_8x64_init( &ctx.hamsi );
-   hamsi512_8way_update( &ctx.hamsi, vhash, 64 );
+   hamsi512_8x64_update( &ctx.hamsi, vhash, 64 );
-   hamsi512_8way_close( &ctx.hamsi, vhash );
+   hamsi512_8x64_close( &ctx.hamsi, vhash );
 
    dintrlv_8x64_512( hash0, hash1, hash2, hash3,
                      hash4, hash5, hash6, hash7, vhash );
@@ -430,9 +429,9 @@ extern void hmq1725_8way_hash(void *state, const void *input)
 
    rintrlv_8x64_8x32( vhashA, vhash, 512 );
 
-   shabal512_8way_init( &ctx.shabal );
+   shabal512_8x32_init( &ctx.shabal );
-   shabal512_8way_update( &ctx.shabal, vhashA, 64 );
+   shabal512_8x32_update( &ctx.shabal, vhashA, 64 );
-   shabal512_8way_close( &ctx.shabal, vhash );
+   shabal512_8x32_close( &ctx.shabal, vhash );
 
    dintrlv_8x32_512( hash0, hash1, hash2, hash3,
                      hash4, hash5, hash6, hash7, vhash );
@@ -475,9 +474,9 @@ extern void hmq1725_8way_hash(void *state, const void *input)
    // B
    if ( likely( vh_mask & 0xff ) )
    {
-      sha512_8way_init( &ctx.sha512 );
+      sha512_8x64_init( &ctx.sha512 );
-      sha512_8way_update( &ctx.sha512, vhash, 64 );
+      sha512_8x64_update( &ctx.sha512, vhash, 64 );
-      sha512_8way_close( &ctx.sha512, vhashB );
+      sha512_8x64_close( &ctx.sha512, vhashB );
    }
 
    mm512_blend_hash_8x64( vh, vhA, vhB, vh_mask );
@@ -510,9 +509,9 @@ extern void hmq1725_8way_hash(void *state, const void *input)
    
 #endif
 
-   sha512_8way_init( &ctx.sha512 );
+   sha512_8x64_init( &ctx.sha512 );
-   sha512_8way_update( &ctx.sha512, vhash, 64 );
+   sha512_8x64_update( &ctx.sha512, vhash, 64 );
-   sha512_8way_close( &ctx.sha512, vhash );
+   sha512_8x64_close( &ctx.sha512, vhash );
 
    vh_mask = _mm512_testn_epi64_mask( vh[0], vmask );
    dintrlv_8x64_512( hash0, hash1, hash2, hash3,
@@ -523,9 +522,9 @@ extern void hmq1725_8way_hash(void *state, const void *input)
    {
       intrlv_8x32_512( vhash, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
                          hash7 );
-      haval256_5_8way_init( &ctx.haval );
+      haval256_8x32_init( &ctx.haval );
-      haval256_5_8way_update( &ctx.haval, vhash, 64 );
+      haval256_8x32_update( &ctx.haval, vhash, 64 );
-      haval256_5_8way_close( &ctx.haval, vhash );
+      haval256_8x32_close( &ctx.haval, vhash );
       memset( &vhash[8<<3], 0, 32<<3 );
       rintrlv_8x32_8x64( vhashA, vhash, 512 );
    }
@@ -552,9 +551,9 @@ extern void hmq1725_8way_hash(void *state, const void *input)
                     hash7 );
    mm512_blend_hash_8x64( vh, vhA, vhB, vh_mask );
 
-   bmw512_8way_init( &ctx.bmw );
+   bmw512_8x64_init( &ctx.bmw );
-   bmw512_8way_update( &ctx.bmw, vhash, 64 );
+   bmw512_8x64_update( &ctx.bmw, vhash, 64 );
-   bmw512_8way_close( &ctx.bmw, state );
+   bmw512_8x64_close( &ctx.bmw, state );
 }
 
 int scanhash_hmq1725_8way( struct work *work, uint32_t max_nonce,
@@ -606,27 +605,27 @@ int scanhash_hmq1725_8way( struct work *work, uint32_t max_nonce,
 
 union _hmq1725_4way_context_overlay
 {
-    blake512_4way_context   blake;
+    blake512_4x64_context   blake;
-    bmw512_4way_context     bmw;
+    bmw512_4x64_context     bmw;
     hashState_groestl       groestl;
-    skein512_4way_context   skein;
+    skein512_4x64_context   skein;
-    jh512_4way_context      jh;
+    jh512_4x64_context      jh;
-    keccak512_4way_context  keccak;
+    keccak512_4x64_context  keccak;
     hashState_luffa         luffa;
     luffa_2way_context      luffa2;
     cubehashParam           cube;
     cube_2way_context       cube2;
     sph_shavite512_context  shavite;
-    hashState_sd            sd;
+    simd512_context         simd;
     shavite512_2way_context shavite2;
-    simd_2way_context       simd;
+    simd_2way_context       simd_2way;
     hashState_echo          echo;
-    hamsi512_4way_context   hamsi;
+    hamsi512_4x64_context   hamsi;
     hashState_fugue         fugue;
-    shabal512_4way_context  shabal;
+    shabal512_4x32_context  shabal;
     sph_whirlpool_context   whirlpool;
-    sha512_4way_context     sha512;
+    sha512_4x64_context     sha512;
-    haval256_5_4way_context haval;
+    haval256_4x32_context haval;
 #if defined(__VAES__)
     groestl512_2way_context groestl2;
     echo_2way_context       echo2;
@@ -653,9 +652,9 @@ extern void hmq1725_4way_hash(void *state, const void *input)
    __m256i* vhA = (__m256i*)vhashA;
    __m256i* vhB = (__m256i*)vhashB;
 
-   bmw512_4way_init( &ctx.bmw );
+   bmw512_4x64_init( &ctx.bmw );
-   bmw512_4way_update( &ctx.bmw, input, 80 );
+   bmw512_4x64_update( &ctx.bmw, input, 80 );
-   bmw512_4way_close( &ctx.bmw, vhash );
+   bmw512_4x64_close( &ctx.bmw, vhash );
 
    dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
@@ -687,17 +686,17 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 // B
 
     if ( h_mask & 0xffffffff )
-       skein512_4way_full( &ctx.skein, vhashB, vhash, 64 );
+       skein512_4x64_full( &ctx.skein, vhashB, vhash, 64 );
 
     mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
-    jh512_4way_init( &ctx.jh );
+    jh512_4x64_init( &ctx.jh );
-    jh512_4way_update( &ctx.jh, vhash, 64 );
+    jh512_4x64_update( &ctx.jh, vhash, 64 );
-    jh512_4way_close( &ctx.jh, vhash );
+    jh512_4x64_close( &ctx.jh, vhash );
 
-    keccak512_4way_init( &ctx.keccak );
+    keccak512_4x64_init( &ctx.keccak );
-    keccak512_4way_update( &ctx.keccak, vhash, 64 );
+    keccak512_4x64_update( &ctx.keccak, vhash, 64 );
-    keccak512_4way_close( &ctx.keccak, vhash );
+    keccak512_4x64_close( &ctx.keccak, vhash );
 
 // second fork, A = blake parallel, B= bmw parallel.
     
@@ -705,13 +704,13 @@ extern void hmq1725_4way_hash(void *state, const void *input)
     h_mask = _mm256_movemask_epi8( vh_mask );
 
     if ( ( h_mask & 0xffffffff ) != 0xffffffff )
-       blake512_4way_full( &ctx.blake, vhashA, vhash, 64 );
+       blake512_4x64_full( &ctx.blake, vhashA, vhash, 64 );
 
     if ( h_mask & 0xffffffff )
     {
-       bmw512_4way_init( &ctx.bmw );
+       bmw512_4x64_init( &ctx.bmw );
-       bmw512_4way_update( &ctx.bmw, vhash, 64 );
+       bmw512_4x64_update( &ctx.bmw, vhash, 64 );
-       bmw512_4way_close( &ctx.bmw, vhashB );
+       bmw512_4x64_close( &ctx.bmw, vhashB );
     }
 
     mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
@@ -734,16 +733,16 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
     if ( ( h_mask & 0xffffffff ) != 0xffffffff )
     {
-        keccak512_4way_init( &ctx.keccak );
+        keccak512_4x64_init( &ctx.keccak );
-        keccak512_4way_update( &ctx.keccak, vhash, 64 );
+        keccak512_4x64_update( &ctx.keccak, vhash, 64 );
-        keccak512_4way_close( &ctx.keccak, vhashA );
+        keccak512_4x64_close( &ctx.keccak, vhashA );
     }
 
     if ( h_mask & 0xffffffff )
     {
-        jh512_4way_init( &ctx.jh );
+        jh512_4x64_init( &ctx.jh );
-        jh512_4way_update( &ctx.jh, vhash, 64 );
+        jh512_4x64_update( &ctx.jh, vhash, 64 );
-        jh512_4way_close( &ctx.jh, vhashB );
+        jh512_4x64_close( &ctx.jh, vhashB );
     }
 
     mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
@@ -753,8 +752,8 @@ extern void hmq1725_4way_hash(void *state, const void *input)
     shavite512_2way_full( &ctx.shavite2, vhashA, vhashA, 64 );
     shavite512_2way_full( &ctx.shavite2, vhashB, vhashB, 64 );
 
-    simd512_2way_full( &ctx.simd, vhashA, vhashA, 64 );
+    simd512_2way_full( &ctx.simd_2way, vhashA, vhashA, 64 );
-    simd512_2way_full( &ctx.simd, vhashB, vhashB, 64 );
+    simd512_2way_full( &ctx.simd_2way, vhashB, vhashB, 64 );
 
     rintrlv_2x128_4x64( vhash, vhashA, vhashB, 512 );     
 
@@ -779,9 +778,9 @@ extern void hmq1725_4way_hash(void *state, const void *input)
     // B
     if ( h_mask & 0xffffffff )
     {
-       haval256_5_4way_init( &ctx.haval );
+       haval256_4x32_init( &ctx.haval );
-       haval256_5_4way_update( &ctx.haval, vhash, 64 );
+       haval256_4x32_update( &ctx.haval, vhash, 64 );
-       haval256_5_4way_close( &ctx.haval, vhash );
+       haval256_4x32_close( &ctx.haval, vhash );
        memset( &vhash[8<<2], 0, 32<<2 );
        rintrlv_4x32_4x64( vhashB, vhash, 512 );
     }
@@ -814,7 +813,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
 #endif
 
-    blake512_4way_full( &ctx.blake, vhash, vhash, 64 );
+    blake512_4x64_full( &ctx.blake, vhash, vhash, 64 );
 
     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
@@ -846,9 +845,9 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
-    hamsi512_4way_init( &ctx.hamsi );
+    hamsi512_4x64_init( &ctx.hamsi );
-    hamsi512_4way_update( &ctx.hamsi, vhash, 64 );
+    hamsi512_4x64_update( &ctx.hamsi, vhash, 64 );
-    hamsi512_4way_close( &ctx.hamsi, vhash );
+    hamsi512_4x64_close( &ctx.hamsi, vhash );
 
     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
@@ -869,47 +868,31 @@ extern void hmq1725_4way_hash(void *state, const void *input)
        echo_full( &ctx.echo, (BitSequence *)hash0, 512,
                        (const BitSequence *)hash0, 64 );
     else
-    {
+       simd512_ctx( &ctx.simd, hash0, hash0, 64 );
-       init_sd( &ctx.sd, 512 );
-       update_final_sd( &ctx.sd, (BitSequence *)hash0,
-                           (const BitSequence *)hash0, 512 );
-    }
 
    if ( hash1[0] & mask ) //4
        echo_full( &ctx.echo, (BitSequence *)hash1, 512,
                        (const BitSequence *)hash1, 64 );
    else
-   {
+       simd512_ctx( &ctx.simd, hash1, hash1, 64 );
-       init_sd( &ctx.sd, 512 );
-       update_final_sd( &ctx.sd, (BitSequence *)hash1,
-                           (const BitSequence *)hash1, 512 );
-   }
 
    if ( hash2[0] & mask ) //4
        echo_full( &ctx.echo, (BitSequence *)hash2, 512,
                        (const BitSequence *)hash2, 64 );
    else
-   {
+       simd512_ctx( &ctx.simd, hash2, hash2, 64 );
-       init_sd( &ctx.sd, 512 );
-       update_final_sd( &ctx.sd, (BitSequence *)hash2,
-                           (const BitSequence *)hash2, 512 );
-   }
 
    if ( hash3[0] & mask ) //4
        echo_full( &ctx.echo, (BitSequence *)hash3, 512,
                        (const BitSequence *)hash3, 64 );
    else
-   {
+       simd512_ctx( &ctx.simd, hash3, hash3, 64 );
-       init_sd( &ctx.sd, 512 );
-       update_final_sd( &ctx.sd, (BitSequence *)hash3,
-                           (const BitSequence *)hash3, 512 );
-   }
 
    intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
 
-   shabal512_4way_init( &ctx.shabal );
+   shabal512_4x32_init( &ctx.shabal );
-   shabal512_4way_update( &ctx.shabal, vhash, 64 );
+   shabal512_4x32_update( &ctx.shabal, vhash, 64 );
-   shabal512_4way_close( &ctx.shabal, vhash );
+   shabal512_4x32_close( &ctx.shabal, vhash );
 
    dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
 
@@ -938,9 +921,9 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
    if ( h_mask & 0xffffffff )
    {
-      sha512_4way_init( &ctx.sha512 );
+      sha512_4x64_init( &ctx.sha512 );
-      sha512_4way_update( &ctx.sha512, vhash, 64 );
+      sha512_4x64_update( &ctx.sha512, vhash, 64 );
-      sha512_4way_close( &ctx.sha512, vhashB );
+      sha512_4x64_close( &ctx.sha512, vhashB );
    }
 
    mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
@@ -967,9 +950,9 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
 #endif
 
-   sha512_4way_init( &ctx.sha512 ); 
+   sha512_4x64_init( &ctx.sha512 ); 
-   sha512_4way_update( &ctx.sha512, vhash, 64 );
+   sha512_4x64_update( &ctx.sha512, vhash, 64 );
-   sha512_4way_close( &ctx.sha512, vhash ); 
+   sha512_4x64_close( &ctx.sha512, vhash ); 
 
 // A = haval parallel, B = Whirlpool serial
 
@@ -981,9 +964,9 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
    if ( ( h_mask & 0xffffffff ) != 0xffffffff )
    {
-      haval256_5_4way_init( &ctx.haval );
+      haval256_4x32_init( &ctx.haval );
-      haval256_5_4way_update( &ctx.haval, vhash, 64 );
+      haval256_4x32_update( &ctx.haval, vhash, 64 );
-      haval256_5_4way_close( &ctx.haval, vhash );
+      haval256_4x32_close( &ctx.haval, vhash );
       memset( &vhash[8<<2], 0, 32<<2 );
       rintrlv_4x32_4x64( vhashA, vhash, 512 );
    }
@@ -1001,9 +984,9 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
    mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
-   bmw512_4way_init( &ctx.bmw );
+   bmw512_4x64_init( &ctx.bmw );
-   bmw512_4way_update( &ctx.bmw, vhash, 64 );
+   bmw512_4x64_update( &ctx.bmw, vhash, 64 );
-   bmw512_4way_close( &ctx.bmw, state );
+   bmw512_4x64_close( &ctx.bmw, state );
 }
 
 int scanhash_hmq1725_4way( struct work *work, uint32_t max_nonce,

algo/quark/hmq1725-gate.h

@@ -4,7 +4,7 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define HMQ1725_8WAY 1
 #elif defined(__AVX2__) && defined(__AES__)
   #define HMQ1725_4WAY 1

algo/quark/quark-gate.h

@@ -4,7 +4,7 @@
 #include "algo-gate-api.h"
 #include <stdint.h>
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define QUARK_8WAY 1
 #elif defined(__AVX2__) && defined(__AES__)
   #define QUARK_4WAY 1

algo/qubit/qubit-gate.h

@@ -5,7 +5,7 @@
 #include <stdint.h>
 
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define QUBIT_4WAY 1
 #elif defined(__AVX2__) && defined(__AES__)
   #define QUBIT_2WAY 1

algo/ripemd/lbry-4way.c

@@ -13,7 +13,7 @@
 
 #if defined(LBRY_16WAY)
 
-static __thread sha256_16way_context sha256_16w_mid;
+static __thread sha256_16x32_context sha256_16w_mid;
 
 void lbry_16way_hash( void* output, const void* input )
 {
@@ -36,17 +36,17 @@ void lbry_16way_hash( void* output, const void* input )
    uint32_t _ALIGN(64) h13[32];
    uint32_t _ALIGN(64) h14[32];
    uint32_t _ALIGN(64) h15[32];
-   sha256_16way_context    ctx_sha256 __attribute__ ((aligned (64)));
+   sha256_16x32_context    ctx_sha256 __attribute__ ((aligned (64)));
-   sha512_8way_context     ctx_sha512;
+   sha512_8x64_context     ctx_sha512;
-   ripemd160_16way_context ctx_ripemd;
+   ripemd160_16x32_context ctx_ripemd;
 
    memcpy( &ctx_sha256, &sha256_16w_mid, sizeof(ctx_sha256) );
-   sha256_16way_update( &ctx_sha256, input + (LBRY_MIDSTATE<<4), LBRY_TAIL );
+   sha256_16x32_update( &ctx_sha256, input + (LBRY_MIDSTATE<<4), LBRY_TAIL );
-   sha256_16way_close( &ctx_sha256, vhashA );
+   sha256_16x32_close( &ctx_sha256, vhashA );
 
-   sha256_16way_init( &ctx_sha256 );
+   sha256_16x32_init( &ctx_sha256 );
-   sha256_16way_update( &ctx_sha256, vhashA, 32 );
+   sha256_16x32_update( &ctx_sha256, vhashA, 32 );
-   sha256_16way_close( &ctx_sha256, vhashA );
+   sha256_16x32_close( &ctx_sha256, vhashA );
 
    // reinterleave to do sha512 4-way 64 bit twice.
    dintrlv_16x32( h0, h1, h2, h3, h4, h5, h6, h7,
@@ -54,13 +54,13 @@ void lbry_16way_hash( void* output, const void* input )
    intrlv_8x64( vhashA, h0, h1, h2, h3, h4, h5, h6, h7, 256 );
    intrlv_8x64( vhashB, h8, h9, h10, h11, h12, h13, h14, h15, 256 );
 
-   sha512_8way_init( &ctx_sha512 );
+   sha512_8x64_init( &ctx_sha512 );
-   sha512_8way_update( &ctx_sha512, vhashA, 32 );
+   sha512_8x64_update( &ctx_sha512, vhashA, 32 );
-   sha512_8way_close( &ctx_sha512, vhashA );
+   sha512_8x64_close( &ctx_sha512, vhashA );
 
-   sha512_8way_init( &ctx_sha512 );
+   sha512_8x64_init( &ctx_sha512 );
-   sha512_8way_update( &ctx_sha512, vhashB, 32 );
+   sha512_8x64_update( &ctx_sha512, vhashB, 32 );
-   sha512_8way_close( &ctx_sha512, vhashB );
+   sha512_8x64_close( &ctx_sha512, vhashB );
 
    // back to 8-way 32 bit
    dintrlv_8x64( h0, h1, h2, h3, h4, h5, h6, h7, vhashA, 512 );
@@ -68,22 +68,22 @@ void lbry_16way_hash( void* output, const void* input )
    intrlv_16x32( vhashA, h0, h1, h2, h3, h4, h5, h6, h7,
                          h8, h9, h10, h11, h12, h13, h14, h15, 512 );
 
-   ripemd160_16way_init( &ctx_ripemd );
+   ripemd160_16x32_init( &ctx_ripemd );
-   ripemd160_16way_update( &ctx_ripemd, vhashA, 32 );
+   ripemd160_16x32_update( &ctx_ripemd, vhashA, 32 );
-   ripemd160_16way_close( &ctx_ripemd, vhashB );
+   ripemd160_16x32_close( &ctx_ripemd, vhashB );
 
-   ripemd160_16way_init( &ctx_ripemd );
+   ripemd160_16x32_init( &ctx_ripemd );
-   ripemd160_16way_update( &ctx_ripemd, vhashA+(8<<4), 32 );
+   ripemd160_16x32_update( &ctx_ripemd, vhashA+(8<<4), 32 );
-   ripemd160_16way_close( &ctx_ripemd, vhashC );
+   ripemd160_16x32_close( &ctx_ripemd, vhashC );
 
-   sha256_16way_init( &ctx_sha256 );
+   sha256_16x32_init( &ctx_sha256 );
-   sha256_16way_update( &ctx_sha256, vhashB, 20 );
+   sha256_16x32_update( &ctx_sha256, vhashB, 20 );
-   sha256_16way_update( &ctx_sha256, vhashC, 20 );
+   sha256_16x32_update( &ctx_sha256, vhashC, 20 );
-   sha256_16way_close( &ctx_sha256, vhashA );
+   sha256_16x32_close( &ctx_sha256, vhashA );
 
-   sha256_16way_init( &ctx_sha256 );
+   sha256_16x32_init( &ctx_sha256 );
-   sha256_16way_update( &ctx_sha256, vhashA, 32 );
+   sha256_16x32_update( &ctx_sha256, vhashA, 32 );
-   sha256_16way_close( &ctx_sha256, output );
+   sha256_16x32_close( &ctx_sha256, output );
 }
 
 int scanhash_lbry_16way( struct work *work, uint32_t max_nonce,
@@ -104,19 +104,19 @@ int scanhash_lbry_16way( struct work *work, uint32_t max_nonce,
    int thr_id = mythr->id;  // thr_id arg is deprecated
 
    // we need bigendian data...
-   casti_m128i( edata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
+   casti_v128u32( edata, 0 ) = v128_bswap32( casti_v128u32( pdata, 0 ) );
-   casti_m128i( edata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
+   casti_v128u32( edata, 1 ) = v128_bswap32( casti_v128u32( pdata, 1 ) );
-   casti_m128i( edata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
+   casti_v128u32( edata, 2 ) = v128_bswap32( casti_v128u32( pdata, 2 ) );
-   casti_m128i( edata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
+   casti_v128u32( edata, 3 ) = v128_bswap32( casti_v128u32( pdata, 3 ) );
-   casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
+   casti_v128u32( edata, 4 ) = v128_bswap32( casti_v128u32( pdata, 4 ) );
-   casti_m128i( edata, 5 ) = mm128_bswap_32( casti_m128i( pdata, 5 ) );
+   casti_v128u32( edata, 5 ) = v128_bswap32( casti_v128u32( pdata, 5 ) );
-   casti_m128i( edata, 6 ) = mm128_bswap_32( casti_m128i( pdata, 6 ) );
+   casti_v128u32( edata, 6 ) = v128_bswap32( casti_v128u32( pdata, 6 ) );
-   casti_m128i( edata, 7 ) = mm128_bswap_32( casti_m128i( pdata, 7 ) );
+   casti_v128u32( edata, 7 ) = v128_bswap32( casti_v128u32( pdata, 7 ) );
    intrlv_16x32( vdata, edata, edata, edata, edata, edata, edata, edata,
         edata, edata, edata, edata, edata, edata, edata, edata, edata, 1024 );
 
-   sha256_16way_init( &sha256_16w_mid );
+   sha256_16x32_init( &sha256_16w_mid );
-   sha256_16way_update( &sha256_16w_mid, vdata, LBRY_MIDSTATE );
+   sha256_16x32_update( &sha256_16w_mid, vdata, LBRY_MIDSTATE );
 
    do
    {
@@ -144,7 +144,7 @@ int scanhash_lbry_16way( struct work *work, uint32_t max_nonce,
 
 #elif defined(LBRY_8WAY)
 
-static __thread sha256_8way_context sha256_8w_mid;
+static __thread sha256_8x32_context sha256_8w_mid;
 
 void lbry_8way_hash( void* output, const void* input )
 {
@@ -159,52 +159,52 @@ void lbry_8way_hash( void* output, const void* input )
    uint32_t _ALIGN(32) h5[32];
    uint32_t _ALIGN(32) h6[32];
    uint32_t _ALIGN(32) h7[32];
-   sha256_8way_context     ctx_sha256 __attribute__ ((aligned (64)));
+   sha256_8x32_context     ctx_sha256 __attribute__ ((aligned (64)));
-   sha512_4way_context     ctx_sha512;
+   sha512_4x64_context     ctx_sha512;
-   ripemd160_8way_context  ctx_ripemd;
+   ripemd160_8x32_context  ctx_ripemd;
 
    memcpy( &ctx_sha256, &sha256_8w_mid, sizeof(ctx_sha256) );
-   sha256_8way_update( &ctx_sha256, input + (LBRY_MIDSTATE<<3), LBRY_TAIL );
+   sha256_8x32_update( &ctx_sha256, input + (LBRY_MIDSTATE<<3), LBRY_TAIL );
-   sha256_8way_close( &ctx_sha256, vhashA );
+   sha256_8x32_close( &ctx_sha256, vhashA );
 
-   sha256_8way_init( &ctx_sha256 );
+   sha256_8x32_init( &ctx_sha256 );
-   sha256_8way_update( &ctx_sha256, vhashA, 32 );
+   sha256_8x32_update( &ctx_sha256, vhashA, 32 );
-   sha256_8way_close( &ctx_sha256, vhashA );
+   sha256_8x32_close( &ctx_sha256, vhashA );
 
    // reinterleave to do sha512 4-way 64 bit twice.
    dintrlv_8x32( h0, h1, h2, h3, h4, h5, h6, h7, vhashA, 256 );
    intrlv_4x64( vhashA, h0, h1, h2, h3, 256 );
    intrlv_4x64( vhashB, h4, h5, h6, h7, 256 );
 
-   sha512_4way_init( &ctx_sha512 );
+   sha512_4x64_init( &ctx_sha512 );
-   sha512_4way_update( &ctx_sha512, vhashA, 32 );
+   sha512_4x64_update( &ctx_sha512, vhashA, 32 );
-   sha512_4way_close( &ctx_sha512, vhashA );
+   sha512_4x64_close( &ctx_sha512, vhashA );
 
-   sha512_4way_init( &ctx_sha512 );
+   sha512_4x64_init( &ctx_sha512 );
-   sha512_4way_update( &ctx_sha512, vhashB, 32 );
+   sha512_4x64_update( &ctx_sha512, vhashB, 32 );
-   sha512_4way_close( &ctx_sha512, vhashB );
+   sha512_4x64_close( &ctx_sha512, vhashB );
 
    // back to 8-way 32 bit
    dintrlv_4x64( h0, h1, h2, h3, vhashA, 512 );
    dintrlv_4x64( h4, h5, h6, h7, vhashB, 512 );
    intrlv_8x32( vhashA, h0, h1, h2, h3, h4, h5, h6, h7, 512 );
 
-   ripemd160_8way_init( &ctx_ripemd );
+   ripemd160_8x32_init( &ctx_ripemd );
-   ripemd160_8way_update( &ctx_ripemd, vhashA, 32 );
+   ripemd160_8x32_update( &ctx_ripemd, vhashA, 32 );
-   ripemd160_8way_close( &ctx_ripemd, vhashB );
+   ripemd160_8x32_close( &ctx_ripemd, vhashB );
 
-   ripemd160_8way_init( &ctx_ripemd );
+   ripemd160_8x32_init( &ctx_ripemd );
-   ripemd160_8way_update( &ctx_ripemd, vhashA+(8<<3), 32 );
+   ripemd160_8x32_update( &ctx_ripemd, vhashA+(8<<3), 32 );
-   ripemd160_8way_close( &ctx_ripemd, vhashC );
+   ripemd160_8x32_close( &ctx_ripemd, vhashC );
 
-   sha256_8way_init( &ctx_sha256 );
+   sha256_8x32_init( &ctx_sha256 );
-   sha256_8way_update( &ctx_sha256, vhashB, 20 );
+   sha256_8x32_update( &ctx_sha256, vhashB, 20 );
-   sha256_8way_update( &ctx_sha256, vhashC, 20 );
+   sha256_8x32_update( &ctx_sha256, vhashC, 20 );
-   sha256_8way_close( &ctx_sha256, vhashA );
+   sha256_8x32_close( &ctx_sha256, vhashA );
 
-   sha256_8way_init( &ctx_sha256 );
+   sha256_8x32_init( &ctx_sha256 );
-   sha256_8way_update( &ctx_sha256, vhashA, 32 );
+   sha256_8x32_update( &ctx_sha256, vhashA, 32 );
-   sha256_8way_close( &ctx_sha256, output );
+   sha256_8x32_close( &ctx_sha256, output );
 }
 
 int scanhash_lbry_8way( struct work *work, uint32_t max_nonce,
@@ -224,19 +224,19 @@ int scanhash_lbry_8way( struct work *work, uint32_t max_nonce,
    int thr_id = mythr->id;  // thr_id arg is deprecated
 
    // we need bigendian data...
-   casti_m128i( edata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
+   casti_v128u32( edata, 0 ) = v128_bswap32( casti_v128u32( pdata, 0 ) );
-   casti_m128i( edata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
+   casti_v128u32( edata, 1 ) = v128_bswap32( casti_v128u32( pdata, 1 ) );
-   casti_m128i( edata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
+   casti_v128u32( edata, 2 ) = v128_bswap32( casti_v128u32( pdata, 2 ) );
-   casti_m128i( edata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
+   casti_v128u32( edata, 3 ) = v128_bswap32( casti_v128u32( pdata, 3 ) );
-   casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
+   casti_v128u32( edata, 4 ) = v128_bswap32( casti_v128u32( pdata, 4 ) );
-   casti_m128i( edata, 5 ) = mm128_bswap_32( casti_m128i( pdata, 5 ) );
+   casti_v128u32( edata, 5 ) = v128_bswap32( casti_v128u32( pdata, 5 ) );
-   casti_m128i( edata, 6 ) = mm128_bswap_32( casti_m128i( pdata, 6 ) );
+   casti_v128u32( edata, 6 ) = v128_bswap32( casti_v128u32( pdata, 6 ) );
-   casti_m128i( edata, 7 ) = mm128_bswap_32( casti_m128i( pdata, 7 ) );
+   casti_v128u32( edata, 7 ) = v128_bswap32( casti_v128u32( pdata, 7 ) );
    intrlv_8x32( vdata, edata, edata, edata, edata,
                        edata, edata, edata, edata, 1024 );
 
-   sha256_8way_init( &sha256_8w_mid );
+   sha256_8x32_init( &sha256_8w_mid );
-   sha256_8way_update( &sha256_8w_mid, vdata, LBRY_MIDSTATE );
+   sha256_8x32_update( &sha256_8w_mid, vdata, LBRY_MIDSTATE );
 
    do
    {

algo/ripemd/lbry-gate.c

@@ -51,7 +51,6 @@ int lbry_get_work_data_size() { return LBRY_WORK_DATA_SIZE; }
 
 bool register_lbry_algo( algo_gate_t* gate )
 {
-//  gate->optimizations = AVX2_OPT | AVX512_OPT | SHA_OPT;
 #if defined (LBRY_16WAY)
   gate->scanhash              = (void*)&scanhash_lbry_16way;
   gate->hash                  = (void*)&lbry_16way_hash;
@@ -67,7 +66,7 @@ bool register_lbry_algo( algo_gate_t* gate )
 #else 
   gate->scanhash              = (void*)&scanhash_lbry;
   gate->hash                  = (void*)&lbry_hash;
-  gate->optimizations = AVX2_OPT | AVX512_OPT | SHA_OPT;
+  gate->optimizations = AVX2_OPT | AVX512_OPT | SHA256_OPT;
 #endif
   gate->build_stratum_request = (void*)&lbry_le_build_stratum_request;
   gate->build_extraheader     = (void*)&lbry_build_extraheader;

algo/ripemd/lbry-gate.h

@@ -5,7 +5,7 @@
 #include <stdint.h>
 
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define LBRY_16WAY 1
 #elif defined(__AVX2__)
   #define LBRY_8WAY 1

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

@@ -35,13 +35,13 @@ static const uint32_t IV[5] =
    _mm_xor_si128( _mm_and_si128( _mm_xor_si128( y, z ), x ), z )
 
 #define F3(x, y, z) \
-   _mm_xor_si128( _mm_or_si128( x, v128_not( y ) ), z )
+   _mm_xor_si128( v128_ornot( y, x ), z )
 
 #define F4(x, y, z) \
    _mm_xor_si128( _mm_and_si128( _mm_xor_si128( x, y ), z ), y )
 
 #define F5(x, y, z) \
-   _mm_xor_si128( x, _mm_or_si128( y, v128_not( z ) ) )
+   _mm_xor_si128( x, v128_ornot( z, y ) )
 
 #define RR(a, b, c, d, e, f, s, r, k) \
 do{ \
@@ -57,7 +57,7 @@ do{ \
 #define ROUND2(a, b, c, d, e, f, s, r, k)  \
 	RR(a ## 2, b ## 2, c ## 2, d ## 2, e ## 2, f, s, r, K2 ## k)
 
-static void ripemd160_4way_round( ripemd160_4way_context *sc )
+static void ripemd160_4x32_round( ripemd160_4x32_context *sc )
 {
    const __m128i *in = (__m128i*)sc->buf;
    __m128i *h  = (__m128i*)sc->val;
@@ -249,7 +249,7 @@ static void ripemd160_4way_round( ripemd160_4way_context *sc )
    h[0] = tmp;
 }
 
-void ripemd160_4way_init( ripemd160_4way_context *sc )
+void ripemd160_4x32_init( ripemd160_4x32_context *sc )
 {
    sc->val[0] = _mm_set1_epi64x( 0x6745230167452301 );
    sc->val[1] = _mm_set1_epi64x( 0xEFCDAB89EFCDAB89 );
@@ -259,7 +259,7 @@ void ripemd160_4way_init( ripemd160_4way_context *sc )
    sc->count_high = sc->count_low = 0;
 }
 
-void ripemd160_4way_update( ripemd160_4way_context *sc, const void *data,
+void ripemd160_4x32_update( ripemd160_4x32_context *sc, const void *data,
                             size_t len )
 {
    __m128i *vdata = (__m128i*)data;
@@ -281,7 +281,7 @@ void ripemd160_4way_update( ripemd160_4way_context *sc, const void *data,
       len -= clen;
       if ( ptr == block_size )
       {
-         ripemd160_4way_round( sc );
+         ripemd160_4x32_round( sc );
          ptr = 0;
       }
       clow = sc->count_low;
@@ -292,7 +292,7 @@ void ripemd160_4way_update( ripemd160_4way_context *sc, const void *data,
    }
 }
 
-void ripemd160_4way_close( ripemd160_4way_context  *sc, void *dst )
+void ripemd160_4x32_close( ripemd160_4x32_context  *sc, void *dst )
 {
    unsigned ptr, u;
    uint32_t low, high;
@@ -306,7 +306,7 @@ void ripemd160_4way_close( ripemd160_4way_context  *sc, void *dst )
    if ( ptr > pad )
    {
        memset_zero_128( sc->buf + (ptr>>2), (block_size - ptr) >> 2 );
-       ripemd160_4way_round( sc );
+       ripemd160_4x32_round( sc );
        memset_zero_128( sc->buf, pad>>2 );
    }
    else
@@ -317,9 +317,9 @@ void ripemd160_4way_close( ripemd160_4way_context  *sc, void *dst )
     low = low << 3;
     sc->buf[  pad>>2      ] = _mm_set1_epi32( low  );
     sc->buf[ (pad>>2) + 1 ] = _mm_set1_epi32( high );
-    ripemd160_4way_round( sc );
+    ripemd160_4x32_round( sc );
     for (u = 0; u < 5; u ++)
-        casti_m128i( dst, u ) = sc->val[u];
+        casti_v128u32( dst, u ) = sc->val[u];
 }
 
 #endif
@@ -335,13 +335,13 @@ void ripemd160_4way_close( ripemd160_4way_context  *sc, void *dst )
    _mm256_xor_si256( _mm256_and_si256( _mm256_xor_si256( y, z ), x ), z )
 
 #define F8W_3(x, y, z) \
-   _mm256_xor_si256( _mm256_or_si256( x, mm256_not( y ) ), z )
+   _mm256_xor_si256( mm256_ornot( y, x ), z )
 
 #define F8W_4(x, y, z) \
    _mm256_xor_si256( _mm256_and_si256( _mm256_xor_si256( x, y ), z ), y )
 
 #define F8W_5(x, y, z) \
-   _mm256_xor_si256( x, _mm256_or_si256( y, mm256_not( z ) ) )
+   _mm256_xor_si256( x, mm256_ornot( z, y ) )
 
 #define RR_8W(a, b, c, d, e, f, s, r, k) \
 do{ \
@@ -357,7 +357,7 @@ do{ \
 #define ROUND2_8W(a, b, c, d, e, f, s, r, k)  \
         RR_8W(a ## 2, b ## 2, c ## 2, d ## 2, e ## 2, f, s, r, K2 ## k)
 
-static void ripemd160_8way_round( ripemd160_8way_context *sc )
+static void ripemd160_8x32_round( ripemd160_8x32_context *sc )
 {
    const __m256i *in = (__m256i*)sc->buf;
    __m256i *h  = (__m256i*)sc->val;
@@ -550,7 +550,7 @@ static void ripemd160_8way_round( ripemd160_8way_context *sc )
 }
 
 
-void ripemd160_8way_init( ripemd160_8way_context *sc )
+void ripemd160_8x32_init( ripemd160_8x32_context *sc )
 {
    sc->val[0] = _mm256_set1_epi64x( 0x6745230167452301 );
    sc->val[1] = _mm256_set1_epi64x( 0xEFCDAB89EFCDAB89 );
@@ -560,7 +560,7 @@ void ripemd160_8way_init( ripemd160_8way_context *sc )
    sc->count_high = sc->count_low = 0;
 }
 
-void ripemd160_8way_update( ripemd160_8way_context *sc, const void *data,
+void ripemd160_8x32_update( ripemd160_8x32_context *sc, const void *data,
                             size_t len )
 {
    __m256i *vdata = (__m256i*)data;
@@ -582,7 +582,7 @@ void ripemd160_8way_update( ripemd160_8way_context *sc, const void *data,
       len -= clen;
       if ( ptr == block_size )
       {
-         ripemd160_8way_round( sc );
+         ripemd160_8x32_round( sc );
          ptr = 0;
       }
       clow = sc->count_low;
@@ -593,7 +593,7 @@ void ripemd160_8way_update( ripemd160_8way_context *sc, const void *data,
    }
 }
 
-void ripemd160_8way_close( ripemd160_8way_context  *sc, void *dst )
+void ripemd160_8x32_close( ripemd160_8x32_context  *sc, void *dst )
 {
    unsigned ptr, u;
    uint32_t low, high;
@@ -607,7 +607,7 @@ void ripemd160_8way_close( ripemd160_8way_context  *sc, void *dst )
    if ( ptr > pad )
    {
        memset_zero_256( sc->buf + (ptr>>2), (block_size - ptr) >> 2 );
-       ripemd160_8way_round( sc );
+       ripemd160_8x32_round( sc );
        memset_zero_256( sc->buf, pad>>2 );
    }
    else
@@ -618,18 +618,17 @@ void ripemd160_8way_close( ripemd160_8way_context  *sc, void *dst )
     low = low << 3;
     sc->buf[  pad>>2      ] = _mm256_set1_epi32( low  );
     sc->buf[ (pad>>2) + 1 ] = _mm256_set1_epi32( high );
-    ripemd160_8way_round( sc );
+    ripemd160_8x32_round( sc );
     for (u = 0; u < 5; u ++)
         casti_m256i( dst, u ) = sc->val[u];
 }
 
 #endif // __AVX2__
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 //  RIPEMD-160 16 way
 
-
 #define F16W_1(x, y, z) \
    _mm512_xor_si512( _mm512_xor_si512( x, y ), z )
 
@@ -659,7 +658,7 @@ do{ \
 #define ROUND2_16W(a, b, c, d, e, f, s, r, k)  \
         RR_16W(a ## 2, b ## 2, c ## 2, d ## 2, e ## 2, f, s, r, K2 ## k)
 
-static void ripemd160_16way_round( ripemd160_16way_context *sc )
+static void ripemd160_16x32_round( ripemd160_16x32_context *sc )
 {
    const __m512i *in = (__m512i*)sc->buf;
    __m512i *h  = (__m512i*)sc->val;
@@ -851,7 +850,7 @@ static void ripemd160_16way_round( ripemd160_16way_context *sc )
    h[0] = tmp;
 }
 
-void ripemd160_16way_init( ripemd160_16way_context *sc )
+void ripemd160_16x32_init( ripemd160_16x32_context *sc )
 {
    sc->val[0] = _mm512_set1_epi64( 0x6745230167452301 );
    sc->val[1] = _mm512_set1_epi64( 0xEFCDAB89EFCDAB89 );
@@ -861,7 +860,7 @@ void ripemd160_16way_init( ripemd160_16way_context *sc )
    sc->count_high = sc->count_low = 0;
 }
 
-void ripemd160_16way_update( ripemd160_16way_context *sc, const void *data,
+void ripemd160_16x32_update( ripemd160_16x32_context *sc, const void *data,
                       size_t len )
 {
    __m512i *vdata = (__m512i*)data;
@@ -883,7 +882,7 @@ void ripemd160_16way_update( ripemd160_16way_context *sc, const void *data,
       len -= clen;
       if ( ptr == block_size )
       {
-         ripemd160_16way_round( sc );
+         ripemd160_16x32_round( sc );
          ptr = 0;
       }
       clow = sc->count_low;
@@ -894,7 +893,7 @@ void ripemd160_16way_update( ripemd160_16way_context *sc, const void *data,
    }
 }
 
-void ripemd160_16way_close( ripemd160_16way_context  *sc, void *dst )
+void ripemd160_16x32_close( ripemd160_16x32_context  *sc, void *dst )
 {
    unsigned ptr, u;
    uint32_t low, high;
@@ -908,7 +907,7 @@ void ripemd160_16way_close( ripemd160_16way_context  *sc, void *dst )
    if ( ptr > pad )
    {
        memset_zero_512( sc->buf + (ptr>>2), (block_size - ptr) >> 2 );
-       ripemd160_16way_round( sc );
+       ripemd160_16x32_round( sc );
        memset_zero_512( sc->buf, pad>>2 );
    }
    else
@@ -919,7 +918,7 @@ void ripemd160_16way_close( ripemd160_16way_context  *sc, void *dst )
     low = low << 3;
     sc->buf[  pad>>2      ] = _mm512_set1_epi32( low  );
     sc->buf[ (pad>>2) + 1 ] = _mm512_set1_epi32( high );
-    ripemd160_16way_round( sc );
+    ripemd160_16x32_round( sc );
     for (u = 0; u < 5; u ++)
         casti_m512i( dst, u ) = sc->val[u];
 }

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

@@ -12,12 +12,12 @@ typedef struct
    __m128i buf[64>>2];
    __m128i val[5];
    uint32_t count_high, count_low;
-} __attribute__ ((aligned (64))) ripemd160_4way_context;
+} __attribute__ ((aligned (64))) ripemd160_4x32_context;
 
-void ripemd160_4way_init( ripemd160_4way_context *sc );
+void ripemd160_4x32_init( ripemd160_4x32_context *sc );
-void ripemd160_4way_update( ripemd160_4way_context *sc, const void *data,
+void ripemd160_4x32_update( ripemd160_4x32_context *sc, const void *data,
                             size_t len );
-void ripemd160_4way_close( ripemd160_4way_context *sc, void *dst );
+void ripemd160_4x32_close( ripemd160_4x32_context *sc, void *dst );
 
 #if defined (__AVX2__)
 
@@ -26,26 +26,26 @@ typedef struct
    __m256i buf[64>>2];
    __m256i val[5];
    uint32_t count_high, count_low;
-} __attribute__ ((aligned (128))) ripemd160_8way_context;
+} __attribute__ ((aligned (128))) ripemd160_8x32_context;
 
-void ripemd160_8way_init( ripemd160_8way_context *sc );
+void ripemd160_8x32_init( ripemd160_8x32_context *sc );
-void ripemd160_8way_update( ripemd160_8way_context *sc, const void *data,
+void ripemd160_8x32_update( ripemd160_8x32_context *sc, const void *data,
                             size_t len );
-void ripemd160_8way_close( ripemd160_8way_context *sc, void *dst );
+void ripemd160_8x32_close( ripemd160_8x32_context *sc, void *dst );
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 typedef struct
 {
    __m512i buf[64>>2];
    __m512i val[5];
    uint32_t count_high, count_low;
-} __attribute__ ((aligned (128))) ripemd160_16way_context;
+} __attribute__ ((aligned (128))) ripemd160_16x32_context;
 
-void ripemd160_16way_init( ripemd160_16way_context *sc );
+void ripemd160_16x32_init( ripemd160_16x32_context *sc );
-void ripemd160_16way_update( ripemd160_16way_context *sc, const void *data,
+void ripemd160_16x32_update( ripemd160_16x32_context *sc, const void *data,
                       size_t len );
-void ripemd160_16way_close( ripemd160_16way_context *sc, void *dst );
+void ripemd160_16x32_close( ripemd160_16x32_context *sc, void *dst );
 
 #endif // AVX512
 #endif // __AVX2__

algo/scrypt/neoscrypt.c

@@ -46,7 +46,7 @@
 #endif
 
 #ifdef __GNUC__
-#if defined(NOASM) || defined(__arm__) || defined(__aarch64__)
+#if defined(NOASM) || defined(__arm__) || defined(__aarch64__) || defined(__APPLE__)
 #define ASM 0
 #else
 #define ASM 1
@@ -597,6 +597,45 @@ static void blake2s_compress(blake2s_state *S, const void *buf) {
     v[13] = S->t[1] ^ blake2s_IV[5];
     v[14] = S->f[0] ^ blake2s_IV[6];
     v[15] = S->f[1] ^ blake2s_IV[7];
+
+#if defined(__SSE2__) || defined(__ARM_NEON)
+
+   v128_t *V = (v128_t*)v;
+
+#define ROUND( r ) \
+   V[0] = v128_add32( V[0], v128_add32( V[1], v128_set32( \
+                  m[blake2s_sigma[r][ 6]], m[blake2s_sigma[r][ 4]], \
+                  m[blake2s_sigma[r][ 2]], m[blake2s_sigma[r][ 0]] ) ) ); \
+   V[3] = v128_ror32( v128_xor( V[3], V[0] ), 16 ); \
+   V[2] = v128_add32( V[2], V[3] ); \
+   V[1] = v128_ror32( v128_xor( V[1], V[2] ), 12 ); \
+   V[0] = v128_add32( V[0], v128_add32( V[1], v128_set32( \
+                   m[blake2s_sigma[r][ 7]], m[blake2s_sigma[r][ 5]], \
+                   m[blake2s_sigma[r][ 3]], m[blake2s_sigma[r][ 1]] ) ) ); \
+   V[3] = v128_ror32( v128_xor( V[3], V[0] ), 8 ); \
+   V[2] = v128_add32( V[2], V[3] ); \
+   V[1] = v128_ror32( v128_xor( V[1], V[2] ), 7 ); \
+   V[0] = v128_shufll32( V[0] ); \
+   V[3] = v128_swap64( V[3] ); \
+   V[2] = v128_shuflr32( V[2] ); \
+   V[0] = v128_add32( V[0], v128_add32( V[1], v128_set32( \
+                    m[blake2s_sigma[r][12]], m[blake2s_sigma[r][10]], \
+                    m[blake2s_sigma[r][ 8]], m[blake2s_sigma[r][14]] ) ) ); \
+   V[3] = v128_ror32( v128_xor( V[3], V[0] ), 16 ); \
+   V[2] = v128_add32( V[2], V[3] ); \
+   V[1] = v128_ror32( v128_xor( V[1], V[2] ), 12 ); \
+   V[0] = v128_add32( V[0], v128_add32( V[1], v128_set32( \
+                    m[blake2s_sigma[r][13]], m[blake2s_sigma[r][11]], \
+                    m[blake2s_sigma[r][ 9]], m[blake2s_sigma[r][15]] ) ) ); \
+   V[3] = v128_ror32( v128_xor( V[3], V[0] ), 8 ); \
+   V[2] = v128_add32( V[2], V[3] ); \
+   V[1] = v128_ror32( v128_xor( V[1], V[2] ), 7 ); \
+   V[0] = v128_shuflr32( V[0] ); \
+   V[3] = v128_swap64( V[3] ); \
+   V[2] = v128_shufll32( V[2] )
+
+#else
+
 #define G(r,i,a,b,c,d) \
   do { \
     a = a + b + m[blake2s_sigma[r][2*i+0]]; \
@@ -619,6 +658,9 @@ static void blake2s_compress(blake2s_state *S, const void *buf) {
     G(r, 6, v[ 2], v[ 7], v[ 8], v[13]); \
     G(r, 7, v[ 3], v[ 4], v[ 9], v[14]); \
   } while(0)
+
+#endif
+
     ROUND(0);
     ROUND(1);
     ROUND(2);

algo/scrypt/scrypt-core-4way.c

@@ -745,7 +745,7 @@ do{ \
    SALSA_2ROUNDS; SALSA_2ROUNDS; SALSA_2ROUNDS; SALSA_2ROUNDS;
 
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 // Tested OK but very slow
 // 16 way parallel, requires 16x32 interleaving
@@ -2074,7 +2074,7 @@ void scrypt_core_4way( v128_t *X, v128_t *V, const uint32_t N )
          v128_ovly v;    
          for ( int l = 0; l < 4; l++ )
             v.u32[l] = ( *(vptr[l] +i ) ) .u32[l];
-         X[i] = v128_xor( X[i], v.m128 );
+         X[i] = v128_xor( X[i], v.v128 );
       }
 
       xor_salsa8_4way( &X[ 0], &X[16] );
@@ -2211,10 +2211,10 @@ static void salsa8_simd128( uint32_t *b, const uint32_t * const c)
    // X2 is shuffled left 2 (swap_64)    { xd, x8, x7, x2 }
    // X3 is shuffled left 3 (ror_1x32)   { xc, xb, x6, x1 }
 
-   y[0].m128 = X0;
+   y[0].v128 = X0;
-   y[1].m128 = X1;
+   y[1].v128 = X1;
-   y[2].m128 = X2;
+   y[2].v128 = X2;
-   y[3].m128 = X3;
+   y[3].v128 = X3;
 
    z[0].u32[0] = y[0].u32[0];
    z[0].u32[3] = y[1].u32[0];
@@ -2236,10 +2236,10 @@ static void salsa8_simd128( uint32_t *b, const uint32_t * const c)
    z[3].u32[1] = y[2].u32[3];
    z[3].u32[0] = y[3].u32[3];
 
-   B[0] = v128_add32( B[0], z[0].m128 );
+   B[0] = v128_add32( B[0], z[0].v128 );
-   B[1] = v128_add32( B[1], z[1].m128 );
+   B[1] = v128_add32( B[1], z[1].v128 );
-   B[2] = v128_add32( B[2], z[2].m128 );
+   B[2] = v128_add32( B[2], z[2].v128 );
-   B[3] = v128_add32( B[3], z[3].m128 );
+   B[3] = v128_add32( B[3], z[3].v128 );
 
 #endif
 
@@ -2404,14 +2404,14 @@ static inline void salsa_simd128_unshuffle_2buf( uint32_t* xa, uint32_t* xb )
 /*
    v128_ovly ya[4], za[4], yb[4], zb[4];
 
-   ya[0].m128 = XA[0];
+   ya[0].v128 = XA[0];
-   yb[0].m128 = XB[0];
+   yb[0].v128 = XB[0];
-   ya[1].m128 = XA[1];
+   ya[1].v128 = XA[1];
-   yb[1].m128 = XB[1];
+   yb[1].v128 = XB[1];
-   ya[2].m128 = XA[2];
+   ya[2].v128 = XA[2];
-   yb[2].m128 = XB[2];
+   yb[2].v128 = XB[2];
-   ya[3].m128 = XA[3];
+   ya[3].v128 = XA[3];
-   yb[3].m128 = XB[3];
+   yb[3].v128 = XB[3];
 
    za[0].u32[0] = ya[0].u32[0];
    zb[0].u32[0] = yb[0].u32[0];
@@ -2449,14 +2449,14 @@ static inline void salsa_simd128_unshuffle_2buf( uint32_t* xa, uint32_t* xb )
    za[3].u32[3] = ya[0].u32[3];
    zb[3].u32[3] = yb[0].u32[3];
 
-   XA[0] = za[0].m128;
+   XA[0] = za[0].v128;
-   XB[0] = zb[0].m128;
+   XB[0] = zb[0].v128;
-   XA[1] = za[1].m128;
+   XA[1] = za[1].v128;
-   XB[1] = zb[1].m128;
+   XB[1] = zb[1].v128;
-   XA[2] = za[2].m128;
+   XA[2] = za[2].v128;
-   XB[2] = zb[2].m128;
+   XB[2] = zb[2].v128;
-   XA[3] = za[3].m128;
+   XA[3] = za[3].v128;
-   XB[3] = zb[3].m128;
+   XB[3] = zb[3].v128;
 */
 }
 
@@ -2487,7 +2487,7 @@ static void salsa8_simd128_2buf( uint32_t * const ba, uint32_t * const bb,
    XA3 = BA[3] = v128_xor( BA[3], CA[3] );
    XB3 = BB[3] = v128_xor( BB[3], CB[3] );
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
              
    SALSA_8ROUNDS_SIMD128_2BUF;
 
@@ -2770,18 +2770,18 @@ static inline void salsa_simd128_unshuffle_3buf( uint32_t* xa, uint32_t* xb,
 /*  
    v128_ovly ya[4], za[4], yb[4], zb[4], yc[4], zc[4];
 
-   ya[0].m128 = XA[0];
+   ya[0].v128 = XA[0];
-   yb[0].m128 = XB[0];
+   yb[0].v128 = XB[0];
-   yc[0].m128 = XC[0];
+   yc[0].v128 = XC[0];
-   ya[1].m128 = XA[1];
+   ya[1].v128 = XA[1];
-   yb[1].m128 = XB[1];
+   yb[1].v128 = XB[1];
-   yc[1].m128 = XC[1];
+   yc[1].v128 = XC[1];
-   ya[2].m128 = XA[2];
+   ya[2].v128 = XA[2];
-   yb[2].m128 = XB[2];
+   yb[2].v128 = XB[2];
-   yc[2].m128 = XC[2];
+   yc[2].v128 = XC[2];
-   ya[3].m128 = XA[3];
+   ya[3].v128 = XA[3];
-   yb[3].m128 = XB[3];
+   yb[3].v128 = XB[3];
-   yc[3].m128 = XC[3];
+   yc[3].v128 = XC[3];
 
    za[0].u32[0] = ya[0].u32[0];
    zb[0].u32[0] = yb[0].u32[0];
@@ -2835,18 +2835,18 @@ static inline void salsa_simd128_unshuffle_3buf( uint32_t* xa, uint32_t* xb,
    zb[3].u32[3] = yb[0].u32[3];
    zc[3].u32[3] = yc[0].u32[3];
 
-   XA[0] = za[0].m128;
+   XA[0] = za[0].v128;
-   XB[0] = zb[0].m128;
+   XB[0] = zb[0].v128;
-   XC[0] = zc[0].m128;
+   XC[0] = zc[0].v128;
-   XA[1] = za[1].m128;
+   XA[1] = za[1].v128;
-   XB[1] = zb[1].m128;
+   XB[1] = zb[1].v128;
-   XC[1] = zc[1].m128;
+   XC[1] = zc[1].v128;
-   XA[2] = za[2].m128;
+   XA[2] = za[2].v128;
-   XB[2] = zb[2].m128;
+   XB[2] = zb[2].v128;
-   XC[2] = zc[2].m128;
+   XC[2] = zc[2].v128;
-   XA[3] = za[3].m128;
+   XA[3] = za[3].v128;
-   XB[3] = zb[3].m128;
+   XB[3] = zb[3].v128;
-   XC[3] = zc[3].m128;
+   XC[3] = zc[3].v128;
 */
 }   
 
@@ -2886,7 +2886,7 @@ static void salsa8_simd128_3buf( uint32_t *ba, uint32_t *bb, uint32_t *bc,
    XB3 = BB[3] = v128_xor( BB[3], CB[3] );
    XC3 = BC[3] = v128_xor( BC[3], CC[3] );
       
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
    
    SALSA_8ROUNDS_SIMD128_3BUF;
 
@@ -3049,7 +3049,7 @@ static void xor_salsa8(uint32_t * const B, const uint32_t * const C)
             xf = (B[15] ^= C[15]);
 
    
-   #define ROL32( a, c )    ror32( a, c )
+   #define ROL32( a, c )    rol32( a, c )
    #define ADD32( a, b )    ( (a)+(b) )
    #define XOR( a, b )      ( (a)^(b) )
 

algo/scrypt/scrypt-core-4way.h

@@ -5,7 +5,7 @@
 #include <stdlib.h>
 #include <stdint.h>
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 void scrypt_core_16way( __m512i *X, __m512i *V, const uint32_t N );
 

algo/scrypt/scrypt.c

@@ -35,7 +35,7 @@
 //#include <mm_malloc.h>
 #include "malloc-huge.h"
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define SCRYPT_THROUGHPUT 16
 #elif defined(__SHA__) || defined(__ARM_FEATURE_SHA2)
   #define SCRYPT_THROUGHPUT 2
@@ -274,9 +274,6 @@ static inline void PBKDF2_SHA256_128_32_SHA_2BUF( uint32_t *tstate0,
 
 #endif   // SHA
 
-
-
-
 static const uint32_t keypad_4way[ 4*12 ] __attribute((aligned(32))) = 
 {
 	0x80000000, 0x80000000, 0x80000000, 0x80000000,
@@ -339,7 +336,7 @@ static const uint32_t _ALIGN(16) finalblk_4way[4 * 16] = {
 };
 */
 
-static inline void sha256_4way_init_state( void *state )
+static inline void sha256_4x32_init_state( void *state )
 {
    casti_v128( state, 0 ) = v128_32( 0x6A09E667 );
    casti_v128( state, 1 ) = v128_32( 0xBB67AE85 );
@@ -362,21 +359,21 @@ static inline void HMAC_SHA256_80_init_4way( const uint32_t *key,
 	memcpy( pad, key + 4*16, 4*16 );
 	memcpy( pad + 4*4, keypad_4way, 4*48 );
 
-   sha256_4way_transform_le( (v128_t*)ihash, (v128_t*)pad,
+   sha256_4x32_transform_le( (v128_t*)ihash, (v128_t*)pad,
                              (const v128_t*)tstate );
 
-   sha256_4way_init_state( tstate );
+   sha256_4x32_init_state( tstate );
 
 	for ( i = 0; i < 4*8; i++ )  pad[i] = ihash[i] ^ 0x5c5c5c5c;
 	for ( ; i < 4*16; i++ )      pad[i] = 0x5c5c5c5c;
 
-   sha256_4way_transform_le( (v128_t*)ostate, (v128_t*)pad,
+   sha256_4x32_transform_le( (v128_t*)ostate, (v128_t*)pad,
                              (const v128_t*)tstate );
    
    for ( i = 0; i < 4*8; i++ )  pad[i] = ihash[i] ^ 0x36363636;
 	for ( ; i < 4*16; i++ )      pad[i] = 0x36363636;
 
-   sha256_4way_transform_le( (v128_t*)tstate, (v128_t*)pad,
+   sha256_4x32_transform_le( (v128_t*)tstate, (v128_t*)pad,
                              (const v128_t*)tstate );
 }
 
@@ -389,7 +386,7 @@ static inline void PBKDF2_SHA256_80_128_4way( const uint32_t *tstate,
 	uint32_t _ALIGN(16) obuf[4 * 16];
 	int i, j;
 
-   sha256_4way_transform_le( (v128_t*)istate, (v128_t*)salt,
+   sha256_4x32_transform_le( (v128_t*)istate, (v128_t*)salt,
                              (const v128_t*)tstate );
 	
 	memcpy(ibuf, salt + 4 * 16, 4 * 16);
@@ -403,10 +400,10 @@ static inline void PBKDF2_SHA256_80_128_4way( const uint32_t *tstate,
 		ibuf[4 * 4 + 2] = i + 1;
 		ibuf[4 * 4 + 3] = i + 1;
 
-      sha256_4way_transform_le( (v128_t*)obuf, (v128_t*)ibuf,
+      sha256_4x32_transform_le( (v128_t*)obuf, (v128_t*)ibuf,
                                 (const v128_t*)istate );
       
-      sha256_4way_transform_le( (v128_t*)ostate2, (v128_t*)obuf,
+      sha256_4x32_transform_le( (v128_t*)ostate2, (v128_t*)obuf,
                                 (const v128_t*)ostate );
 
       for ( j = 0; j < 4 * 8; j++ )
@@ -421,9 +418,9 @@ static inline void PBKDF2_SHA256_128_32_4way( uint32_t *tstate,
 	uint32_t _ALIGN(64) buf[4 * 16];
 	int i;
 	
-   sha256_4way_transform_be( (v128_t*)tstate, (v128_t*)salt,
+   sha256_4x32_transform_be( (v128_t*)tstate, (v128_t*)salt,
                        (const v128_t*)tstate );
-   sha256_4way_transform_be( (v128_t*)tstate, (v128_t*)( salt + 4*16),
+   sha256_4x32_transform_be( (v128_t*)tstate, (v128_t*)( salt + 4*16),
                        (const v128_t*)tstate );
 
    final[ 0] = v128_32( 0x00000001 );
@@ -434,20 +431,20 @@ static inline void PBKDF2_SHA256_128_32_4way( uint32_t *tstate,
              = v128_xor( final[ 0], final[ 0] ); //_mm_setzero_si128();
    final[15] = v128_32 ( 0x00000620 );
 
-   sha256_4way_transform_le( (v128_t*)tstate, (v128_t*)final,
+   sha256_4x32_transform_le( (v128_t*)tstate, (v128_t*)final,
                        (const v128_t*)tstate );
    
    memcpy(buf, tstate, 4 * 32);
 	memcpy(buf + 4 * 8, outerpad_4way, 4 * 32);
 
-   sha256_4way_transform_le( (v128_t*)ostate, (v128_t*)buf,
+   sha256_4x32_transform_le( (v128_t*)ostate, (v128_t*)buf,
                              (const v128_t*)ostate );
 
    for ( i = 0; i < 4 * 8; i++ )
 		output[i] = bswap_32( ostate[i] );
 }
 
-#ifdef HAVE_SHA256_8WAY
+#if defined(__AVX2__)
 
 /*
 static const uint32_t _ALIGN(32) finalblk_8way[8 * 16] = {
@@ -470,7 +467,7 @@ static const uint32_t _ALIGN(32) finalblk_8way[8 * 16] = {
 };
 */
 
-static inline void sha256_8way_init_state( void *state )
+static inline void sha256_8x32_init_state( void *state )
 {
    casti_m256i( state, 0 ) = _mm256_set1_epi32( 0x6A09E667 );
    casti_m256i( state, 1 ) = _mm256_set1_epi32( 0xBB67AE85 );
@@ -494,21 +491,21 @@ static inline void HMAC_SHA256_80_init_8way( const uint32_t *key,
 	memset( pad + 8*5, 0x00, 8*40 );
 	for ( i = 0; i < 8; i++ )    pad[ 8*15 + i ] = 0x00000280;
 
-   sha256_8way_transform_le( (__m256i*)ihash, (__m256i*)pad,
+   sha256_8x32_transform_le( (__m256i*)ihash, (__m256i*)pad,
                              (const __m256i*)tstate );
 
-   sha256_8way_init_state( tstate );
+   sha256_8x32_init_state( tstate );
 
    for ( i = 0; i < 8*8; i++ )   pad[i] = ihash[i] ^ 0x5c5c5c5c;
 	for ( ; i < 8*16; i++ )       pad[i] = 0x5c5c5c5c;
 
-   sha256_8way_transform_le( (__m256i*)ostate, (__m256i*)pad,
+   sha256_8x32_transform_le( (__m256i*)ostate, (__m256i*)pad,
                              (const __m256i*)tstate );
 
    for ( i = 0; i < 8*8; i++ )   pad[i] = ihash[i] ^ 0x36363636;
 	for ( ; i < 8*16; i++ )       pad[i] = 0x36363636;
 
-   sha256_8way_transform_le( (__m256i*)tstate, (__m256i*)pad,
+   sha256_8x32_transform_le( (__m256i*)tstate, (__m256i*)pad,
                              (const __m256i*)tstate );
 }
 
@@ -521,7 +518,7 @@ static inline void PBKDF2_SHA256_80_128_8way( const uint32_t *tstate,
 	uint32_t _ALIGN(32) obuf[8 * 16];
 	int i, j;
 	
-   sha256_8way_transform_le( (__m256i*)istate, (__m256i*)salt,
+   sha256_8x32_transform_le( (__m256i*)istate, (__m256i*)salt,
                              (const __m256i*)tstate );
 
 	memcpy( ibuf, salt + 8*16, 8*16 );
@@ -544,10 +541,10 @@ static inline void PBKDF2_SHA256_80_128_8way( const uint32_t *tstate,
 		ibuf[8 * 4 + 6] = i + 1;
 		ibuf[8 * 4 + 7] = i + 1;
 
-      sha256_8way_transform_le( (__m256i*)obuf, (__m256i*)ibuf,
+      sha256_8x32_transform_le( (__m256i*)obuf, (__m256i*)ibuf,
                                 (const __m256i*)istate );
 
-      sha256_8way_transform_le( (__m256i*)ostate2, (__m256i*)obuf,
+      sha256_8x32_transform_le( (__m256i*)ostate2, (__m256i*)obuf,
                                 (const __m256i*)ostate );
 
       for ( j = 0; j < 8*8; j++ )
@@ -562,9 +559,9 @@ static inline void PBKDF2_SHA256_128_32_8way( uint32_t *tstate,
    uint32_t _ALIGN(128) buf[ 8*16 ];
 	int i;
 	
-   sha256_8way_transform_be( (__m256i*)tstate, (__m256i*)salt,
+   sha256_8x32_transform_be( (__m256i*)tstate, (__m256i*)salt,
                              (const __m256i*)tstate );
-   sha256_8way_transform_be( (__m256i*)tstate, (__m256i*)( salt + 8*16),
+   sha256_8x32_transform_be( (__m256i*)tstate, (__m256i*)( salt + 8*16),
                              (const __m256i*)tstate );
    
    final[ 0] = _mm256_set1_epi32( 0x00000001 );
@@ -575,7 +572,7 @@ static inline void PBKDF2_SHA256_128_32_8way( uint32_t *tstate,
              = _mm256_setzero_si256();
    final[15] = _mm256_set1_epi32 ( 0x00000620 );
 
-   sha256_8way_transform_le( (__m256i*)tstate, final,
+   sha256_8x32_transform_le( (__m256i*)tstate, final,
                              (const __m256i*)tstate );
 
 	memcpy( buf, tstate, 8*32 );
@@ -583,18 +580,18 @@ static inline void PBKDF2_SHA256_128_32_8way( uint32_t *tstate,
 	memset( buf + 8*9, 0x00, 8*24 );
 	for ( i = 0; i < 8; i++ )     buf[ 8*15 + i ] = 0x00000300;
 
-   sha256_8way_transform_le( (__m256i*)ostate, (__m256i*)buf,
+   sha256_8x32_transform_le( (__m256i*)ostate, (__m256i*)buf,
                              (const __m256i*)ostate );
 
 	for (i = 0; i < 8 * 8; i++)
 		output[i] = bswap_32(ostate[i]);
 }
 
-#endif /* HAVE_SHA256_8WAY */
+#endif //AVX2
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
-static inline void sha256_16way_init_state( void *state )
+static inline void sha256_16x32_init_state( void *state )
 {
    casti_m512i( state, 0 ) = _mm512_set1_epi32( 0x6A09E667 );
    casti_m512i( state, 1 ) = _mm512_set1_epi32( 0xBB67AE85 );
@@ -618,21 +615,21 @@ static inline void HMAC_SHA256_80_init_16way( const uint32_t *key,
    memset( pad + 16*5, 0x00, 16*40 );
    for ( i = 0; i < 16; i++ )       pad[ 16*15 + i ] = 0x00000280;
 
-   sha256_16way_transform_le( (__m512i*)ihash, (__m512i*)pad,
+   sha256_16x32_transform_le( (__m512i*)ihash, (__m512i*)pad,
                               (const __m512i*)tstate );
 
-   sha256_16way_init_state( tstate );
+   sha256_16x32_init_state( tstate );
 
    for ( i = 0; i < 16*8; i++ )    pad[i] = ihash[i] ^ 0x5c5c5c5c;
    for ( ; i < 16*16; i++ )        pad[i] = 0x5c5c5c5c;
 
-   sha256_16way_transform_le( (__m512i*)ostate, (__m512i*)pad,
+   sha256_16x32_transform_le( (__m512i*)ostate, (__m512i*)pad,
                               (const __m512i*)tstate );
 
    for ( i = 0; i < 16*8; i++ )   pad[i] = ihash[i] ^ 0x36363636;
    for ( ; i < 16*16; i++ )       pad[i] = 0x36363636;
  
-   sha256_16way_transform_le( (__m512i*)tstate, (__m512i*)pad,
+   sha256_16x32_transform_le( (__m512i*)tstate, (__m512i*)pad,
                               (const __m512i*)tstate );
 }
 
@@ -645,7 +642,7 @@ static inline void PBKDF2_SHA256_80_128_16way( const uint32_t *tstate,
    uint32_t _ALIGN(128) ostate2[ 16*8 ];
    int i, j;
 
-   sha256_16way_transform_le( (__m512i*)istate, (__m512i*)salt,
+   sha256_16x32_transform_le( (__m512i*)istate, (__m512i*)salt,
                               (const __m512i*)tstate );
 
    memcpy( ibuf, salt + 16*16, 16*16 );
@@ -676,10 +673,10 @@ static inline void PBKDF2_SHA256_80_128_16way( const uint32_t *tstate,
       ibuf[ 16*4 + 14 ] = i + 1;
       ibuf[ 16*4 + 15 ] = i + 1;
 
-      sha256_16way_transform_le( (__m512i*)obuf, (__m512i*)ibuf,
+      sha256_16x32_transform_le( (__m512i*)obuf, (__m512i*)ibuf,
                                  (const __m512i*)istate );
 
-      sha256_16way_transform_le( (__m512i*)ostate2, (__m512i*)obuf,
+      sha256_16x32_transform_le( (__m512i*)ostate2, (__m512i*)obuf,
                                  (const __m512i*)ostate );
 
       for ( j = 0; j < 16*8; j++ )
@@ -694,9 +691,9 @@ static inline void PBKDF2_SHA256_128_32_16way( uint32_t *tstate,
    uint32_t _ALIGN(128) buf[ 16*16 ];
    int i;
 
-   sha256_16way_transform_be( (__m512i*)tstate, (__m512i*)salt,
+   sha256_16x32_transform_be( (__m512i*)tstate, (__m512i*)salt,
                              (const __m512i*)tstate );
-   sha256_16way_transform_be( (__m512i*)tstate, (__m512i*)( salt + 16*16),
+   sha256_16x32_transform_be( (__m512i*)tstate, (__m512i*)( salt + 16*16),
                              (const __m512i*)tstate );
 
    final[ 0] = _mm512_set1_epi32( 0x00000001 );
@@ -707,7 +704,7 @@ static inline void PBKDF2_SHA256_128_32_16way( uint32_t *tstate,
              = _mm512_setzero_si512();
    final[15] = _mm512_set1_epi32 ( 0x00000620 );
 
-   sha256_16way_transform_le( (__m512i*)tstate, final,
+   sha256_16x32_transform_le( (__m512i*)tstate, final,
                              (const __m512i*)tstate );
 
    memcpy( buf, tstate, 16*32 );
@@ -715,7 +712,7 @@ static inline void PBKDF2_SHA256_128_32_16way( uint32_t *tstate,
    memset( buf + 16*9, 0x00, 16*24 );
    for ( i = 0; i < 16; i++ )      buf[ 16*15 + i ] = 0x00000300;
 
-   sha256_16way_transform_le( (__m512i*)ostate, (__m512i*)buf,
+   sha256_16x32_transform_le( (__m512i*)ostate, (__m512i*)buf,
                              (const __m512i*)ostate );
 
    for ( i = 0; i < 16*8; i++ )
@@ -724,25 +721,10 @@ static inline void PBKDF2_SHA256_128_32_16way( uint32_t *tstate,
 
 #endif // AVX512
 
-#define SCRYPT_MAX_WAYS 12
-#define HAVE_SCRYPT_3WAY 1
-void scrypt_core(uint32_t *X, uint32_t *V, int N);
-void scrypt_core_3way(uint32_t *X, uint32_t *V, int N);
-
-#if defined(__AVX2__)
-#undef SCRYPT_MAX_WAYS
-#define SCRYPT_MAX_WAYS 24
-#define HAVE_SCRYPT_6WAY 1
-void scrypt_core_6way(uint32_t *X, uint32_t *V, int N);
-#endif
-
-#ifndef SCRYPT_MAX_WAYS
-#define SCRYPT_MAX_WAYS 1
-#endif
-
 #include "scrypt-core-4way.h"
 
-/*
+#if ( SCRYPT_THROUGHPUT == 1 )
+   
 static bool scrypt_N_1_1_256( const uint32_t *input, uint32_t *output,
                               uint32_t *midstate, int N, int thr_id )
 {
@@ -752,15 +734,12 @@ static bool scrypt_N_1_1_256( const uint32_t *input, uint32_t *output,
 	memcpy(tstate, midstate, 32);
 	HMAC_SHA256_80_init(input, tstate, ostate);
 	PBKDF2_SHA256_80_128(tstate, ostate, input, X);
-
+   scrypt_core_1way( X, scratchbuf, N ); 
-   scrypt_core_simd128( X, scratchbuf, N );  // woring
-//   scrypt_core_1way( X, V, N );  // working
-//   scrypt_core(X, V, N);
-
 	PBKDF2_SHA256_128_32(tstate, ostate, X, output);
    return true;
 }
-*/
+
+#endif
 
 #if ( SCRYPT_THROUGHPUT == 8 )
 
@@ -1201,20 +1180,6 @@ static int scrypt_N_1_1_256_16way( const uint32_t *input, uint32_t *output,
    if ( work_restart[thrid].restart ) return 0;
    scrypt_core_simd128_2buf( X+448, V, N );
 ********************/
-/*
-   scrypt_core_3way( X,     V, N );
-   if ( work_restart[thrid].restart ) return 0;
-   scrypt_core_3way( X+ 96, V, N );
-   if ( work_restart[thrid].restart ) return 0;
-   scrypt_core_simd128_2buf( X+192, V, N );
-   if ( work_restart[thrid].restart ) return 0;
-   scrypt_core_3way( X+256, V, N );
-   if ( work_restart[thrid].restart ) return 0;
-   scrypt_core_3way( X+352, V, N );
-   if ( work_restart[thrid].restart ) return 0;
-   scrypt_core_simd128_2buf( X+448, V, N );
-*/
-
 
    if ( work_restart[thrid].restart ) return 0;
 
@@ -1321,8 +1286,7 @@ static int scrypt_N_1_1_256_4way_sha( const uint32_t *input, uint32_t *output,
    return 1;
 }
 
-#else  
+#elif defined(__SSE2__) || defined(__ARM_NEON)  
-// SSE2
 
 static int scrypt_N_1_1_256_4way( const uint32_t *input,	uint32_t *output,
            uint32_t *midstate, int N, int thrid )
@@ -1481,7 +1445,7 @@ bool scrypt_miner_thread_init( int thr_id )
 bool register_scrypt_algo( algo_gate_t* gate )
 {
 #if defined(__SHA__) || defined(__ARM_FEATURE_SHA2)
-   gate->optimizations = SSE2_OPT | SHA_OPT | NEON_OPT;
+   gate->optimizations = SSE2_OPT | SSE42_OPT | AVX_OPT | SHA256_OPT | NEON_OPT;
 #else
    gate->optimizations = SSE2_OPT | SSE42_OPT | AVX_OPT | AVX2_OPT | AVX512_OPT | NEON_OPT;
 #endif
@@ -1491,31 +1455,31 @@ bool register_scrypt_algo( algo_gate_t* gate )
    opt_param_n = opt_param_n ? opt_param_n : 1024;
    applog( LOG_INFO,"Scrypt paramaters: N= %d, R= 1", opt_param_n );
 
-// scrypt_throughput defined at compile time and used to replace
+   switch ( SCRYPT_THROUGHPUT )
-// MAX_WAYS to reduce memory usage.
+   {
-   
+     case 16:  // AVX512
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
-//   scrypt_throughput = 16;
        if ( opt_param_n > 0x4000 )
          scratchbuf_size = opt_param_n * 3 * 128;  // 3 buf
        else      
          scratchbuf_size = opt_param_n * 4 * 128;  // 4 way
-#elif defined(__SHA__) || defined(__ARM_FEATURE_SHA2)
+     break;
-//   scrypt_throughput = 2;
+     case 2:  // SHA256
          scratchbuf_size = opt_param_n * 2 * 128;  // 2 buf
-#elif defined(__AVX2__)
+     break;
-//   scrypt_throughput = 8;   
+     case 8:  // AVX2
        if ( opt_param_n > 0x4000 )
          scratchbuf_size = opt_param_n * 3 * 128;  // 3 buf
      else
          scratchbuf_size = opt_param_n * 2 * 128;  // 2 way
-#else
+     break;
-//   scrypt_throughput = 4;
+     case 4:  // SSE2, NEON
        if ( opt_param_n > 0x4000 )
          scratchbuf_size = opt_param_n * 2 * 128;  // 2 buf
        else
          scratchbuf_size = opt_param_n * 4 * 128;  // 4 way
-#endif
+     default:
+         scratchbuf_size = opt_param_n;  // 1 way
+   }
 
    char t_units[4] = {0};
    char d_units[4] = {0};

algo/sha/hmac-sha256-hash-4way.c

@@ -31,7 +31,7 @@
 #include "hmac-sha256-hash-4way.h"
 #include "compat.h"
 
-#if defined(__SSE2__)
+#if defined(__SSE2__) || defined(__ARM_NEON)
 // HMAC 4-way SSE2
 
 /**
@@ -62,30 +62,30 @@ hmac_sha256_4way_init( hmac_sha256_4way_context *ctx, const void *_K,
 	/* If Klen > 64, the key is really SHA256(K). */
 	if ( Klen > 64 )
    {
-		sha256_4way_init( &ctx->ictx );
+		sha256_4x32_init( &ctx->ictx );
-		sha256_4way_update( &ctx->ictx, K, Klen );
+		sha256_4x32_update( &ctx->ictx, K, Klen );
-		sha256_4way_close( &ctx->ictx, khash );
+		sha256_4x32_close( &ctx->ictx, khash );
 		K = khash;
 		Klen = 32;
 	}
 
 	/* Inner SHA256 operation is SHA256(K xor [block of 0x36] || data). */
-   sha256_4way_init( &ctx->ictx );
+   sha256_4x32_init( &ctx->ictx );
 	memset( pad, 0x36, 64*4 );
 
    for ( i = 0; i < Klen; i++ )
-		casti_m128i( pad, i ) = _mm_xor_si128( casti_m128i( pad, i ),
+		casti_v128u32( pad, i ) = v128_xor( casti_v128u32( pad, i ),
-                                             casti_m128i( K, i ) );
+                                          casti_v128u32( K, i ) );
 
-   sha256_4way_update( &ctx->ictx, pad, 64 );
+   sha256_4x32_update( &ctx->ictx, pad, 64 );
 
 	/* Outer SHA256 operation is SHA256(K xor [block of 0x5c] || hash). */
-	sha256_4way_init( &ctx->octx );
+	sha256_4x32_init( &ctx->octx );
 	memset( pad, 0x5c, 64*4 );
 	for ( i = 0; i < Klen/4; i++ )
-		casti_m128i( pad, i ) = _mm_xor_si128( casti_m128i( pad, i ),
+		casti_v128u32( pad, i ) = v128_xor( casti_v128u32( pad, i ),
-                                             casti_m128i( K, i ) );
+                                          casti_v128u32( K, i ) );
-	sha256_4way_update( &ctx->octx, pad, 64 );
+	sha256_4x32_update( &ctx->octx, pad, 64 );
 }
 
 /* Add bytes to the HMAC-SHA256 operation. */
@@ -94,7 +94,7 @@ hmac_sha256_4way_update( hmac_sha256_4way_context *ctx, const void *in,
                          size_t len )
 {
 	/* Feed data to the inner SHA256 operation. */
-	sha256_4way_update( &ctx->ictx, in, len );
+	sha256_4x32_update( &ctx->ictx, in, len );
 }
 
 /* Finish an HMAC-SHA256 operation. */
@@ -104,13 +104,13 @@ hmac_sha256_4way_close( hmac_sha256_4way_context *ctx, void *digest )
 	unsigned char ihash[32*4] __attribute__ ((aligned (64)));
 
 	/* Finish the inner SHA256 operation. */
-	sha256_4way_close( &ctx->ictx, ihash );
+	sha256_4x32_close( &ctx->ictx, ihash );
 
 	/* Feed the inner hash to the outer SHA256 operation. */
-	sha256_4way_update( &ctx->octx, ihash, 32 );
+	sha256_4x32_update( &ctx->octx, ihash, 32 );
 
 	/* Finish the outer SHA256 operation. */
-	sha256_4way_close( &ctx->octx, digest );
+	sha256_4x32_close( &ctx->octx, digest );
 }
 
 /**
@@ -126,7 +126,7 @@ pbkdf2_sha256_4way( uint8_t *buf, size_t dkLen,
 	hmac_sha256_4way_context PShctx, hctx;
 	uint8_t _ALIGN(128) T[32*4];
 	uint8_t _ALIGN(128) U[32*4];
-   __m128i ivec;
+   v128u32_t ivec;
    size_t i, clen;
 	uint64_t j;
 	int k;
@@ -139,7 +139,7 @@ pbkdf2_sha256_4way( uint8_t *buf, size_t dkLen,
 	for ( i = 0; i * 32 < dkLen; i++ )
    {
 		/* Generate INT(i + 1). */
-      ivec = _mm_set1_epi32( bswap_32( i+1 ) ); 
+      ivec = v128_32( bswap_32( i+1 ) ); 
 
 		/* Compute U_1 = PRF(P, S || INT(i)). */
 		memcpy( &hctx, &PShctx, sizeof(hmac_sha256_4way_context) );
@@ -158,8 +158,8 @@ pbkdf2_sha256_4way( uint8_t *buf, size_t dkLen,
 
 			/* ... xor U_j ... */
 			for ( k = 0; k < 8; k++ )
-				casti_m128i( T, k ) = _mm_xor_si128( casti_m128i( T, k ),
+				casti_v128u32( T, k ) = v128_xor( casti_v128u32( T, k ),
-                                                 casti_m128i( U, k ) );
+                                              casti_v128u32( U, k ) );
 		}
 
 		/* Copy as many bytes as necessary into buf. */
@@ -199,30 +199,30 @@ hmac_sha256_8way_init( hmac_sha256_8way_context *ctx, const void *_K,
    /* If Klen > 64, the key is really SHA256(K). */
    if ( Klen > 64 )
    {
-      sha256_8way_init( &ctx->ictx );
+      sha256_8x32_init( &ctx->ictx );
-      sha256_8way_update( &ctx->ictx, K, Klen );
+      sha256_8x32_update( &ctx->ictx, K, Klen );
-      sha256_8way_close( &ctx->ictx, khash );
+      sha256_8x32_close( &ctx->ictx, khash );
       K = khash;
       Klen = 32;
    }
 
    /* Inner SHA256 operation is SHA256(K xor [block of 0x36] || data). */
-   sha256_8way_init( &ctx->ictx );
+   sha256_8x32_init( &ctx->ictx );
    memset( pad, 0x36, 64*8);
 
    for ( i = 0; i < Klen/4; i++ )
       casti_m256i( pad, i ) = _mm256_xor_si256( casti_m256i( pad, i ),
                                                 casti_m256i( K, i ) );
 
-   sha256_8way_update( &ctx->ictx, pad, 64 );
+   sha256_8x32_update( &ctx->ictx, pad, 64 );
 
    /* Outer SHA256 operation is SHA256(K xor [block of 0x5c] || hash). */
-   sha256_8way_init( &ctx->octx );
+   sha256_8x32_init( &ctx->octx );
    memset( pad, 0x5c, 64*8 );
    for ( i = 0; i < Klen/4; i++ )
       casti_m256i( pad, i ) = _mm256_xor_si256( casti_m256i( pad, i ),
                                                 casti_m256i( K, i ) );
-   sha256_8way_update( &ctx->octx, pad, 64 );
+   sha256_8x32_update( &ctx->octx, pad, 64 );
 }
 
 void
@@ -230,7 +230,7 @@ hmac_sha256_8way_update( hmac_sha256_8way_context *ctx, const void *in,
                          size_t len )
 {
    /* Feed data to the inner SHA256 operation. */
-   sha256_8way_update( &ctx->ictx, in, len );
+   sha256_8x32_update( &ctx->ictx, in, len );
 }
 
 /* Finish an HMAC-SHA256 operation. */
@@ -240,13 +240,13 @@ hmac_sha256_8way_close( hmac_sha256_8way_context *ctx, void *digest )
    unsigned char ihash[32*8] __attribute__ ((aligned (128)));
 
    /* Finish the inner SHA256 operation. */
-   sha256_8way_close( &ctx->ictx, ihash );
+   sha256_8x32_close( &ctx->ictx, ihash );
 
    /* Feed the inner hash to the outer SHA256 operation. */
-   sha256_8way_update( &ctx->octx, ihash, 32 );
+   sha256_8x32_update( &ctx->octx, ihash, 32 );
 
    /* Finish the outer SHA256 operation. */
-   sha256_8way_close( &ctx->octx, digest );
+   sha256_8x32_close( &ctx->octx, digest );
 }
 
 /**
@@ -306,7 +306,7 @@ pbkdf2_sha256_8way( uint8_t *buf, size_t dkLen, const uint8_t *passwd,
    }
 }
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 // HMAC 16-way AVX512
 
@@ -332,21 +332,21 @@ hmac_sha256_16way_init( hmac_sha256_16way_context *ctx, const void *_K,
    /* If Klen > 64, the key is really SHA256(K). */
    if ( Klen > 64 )
    {
-      sha256_16way_init( &ctx->ictx );
+      sha256_16x32_init( &ctx->ictx );
-      sha256_16way_update( &ctx->ictx, K, Klen );
+      sha256_16x32_update( &ctx->ictx, K, Klen );
-      sha256_16way_close( &ctx->ictx, khash );
+      sha256_16x32_close( &ctx->ictx, khash );
       K = khash;
       Klen = 32;
    }
 
    /* Inner SHA256 operation is SHA256(K xor [block of 0x36] || data). */
-   sha256_16way_init( &ctx->ictx );
+   sha256_16x32_init( &ctx->ictx );
    memset( pad, 0x36, 64*16 );
 
    for ( i = 0; i < Klen; i++ )
       casti_m512i( pad, i ) = _mm512_xor_si512( casti_m512i( pad, i ),
                                                 casti_m512i( K, i ) );
-   sha256_16way_update( &ctx->ictx, pad, 64 );
+   sha256_16x32_update( &ctx->ictx, pad, 64 );
 
    /* Outer SHA256 operation is SHA256(K xor [block of 0x5c] || hash). */
    sha256_16way_init( &ctx->octx );
@@ -354,7 +354,7 @@ hmac_sha256_16way_init( hmac_sha256_16way_context *ctx, const void *_K,
    for ( i = 0; i < Klen/4; i++ )
       casti_m512i( pad, i ) = _mm512_xor_si512( casti_m512i( pad, i ),
                                              casti_m512i( K, i ) );
-   sha256_16way_update( &ctx->octx, pad, 64 );
+   sha256_16x32_update( &ctx->octx, pad, 64 );
 }
    
 void
@@ -362,7 +362,7 @@ hmac_sha256_16way_update( hmac_sha256_16way_context *ctx, const void *in,
                          size_t len )
 {
    /* Feed data to the inner SHA256 operation. */
-   sha256_16way_update( &ctx->ictx, in, len );
+   sha256_16x32_update( &ctx->ictx, in, len );
 }
 
 /* Finish an HMAC-SHA256 operation. */
@@ -372,13 +372,13 @@ hmac_sha256_16way_close( hmac_sha256_16way_context *ctx, void *digest )
    unsigned char ihash[32*16] __attribute__ ((aligned (128)));
 
    /* Finish the inner SHA256 operation. */
-   sha256_16way_close( &ctx->ictx, ihash );
+   sha256_16x32_close( &ctx->ictx, ihash );
 
    /* Feed the inner hash to the outer SHA256 operation. */
-   sha256_16way_update( &ctx->octx, ihash, 32 );
+   sha256_16x32_update( &ctx->octx, ihash, 32 );
 
    /* Finish the outer SHA256 operation. */
-   sha256_16way_close( &ctx->octx, digest );
+   sha256_16x32_close( &ctx->octx, digest );
 }
 
 /**

algo/sha/hmac-sha256-hash-4way.h

@@ -1,6 +1,6 @@
 /*-
  * Copyright 2005,2007,2009 Colin Percival
- * Copyright 2020 JayDDee@gmailcom
+ * Copyright 2020 JayDDee246@gmailcom
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -38,11 +38,12 @@
 #include "simd-utils.h"
 #include "sha256-hash.h"
 
-#if defined(__SSE2__)
+#if defined(__SSE2__) || defined(__ARM_NEON)
+
 typedef struct _hmac_sha256_4way_context
 {
-   sha256_4way_context ictx;
+   sha256_4x32_context ictx;
-   sha256_4way_context octx;
+   sha256_4x32_context octx;
 } hmac_sha256_4way_context;
 
 //void SHA256_Buf( const void *, size_t len, uint8_t digest[32] );
@@ -67,8 +68,8 @@ void pbkdf2_sha256_4way( uint8_t *, size_t, const uint8_t *, size_t,
 
 typedef struct _hmac_sha256_8way_context
 {
-   sha256_8way_context ictx;
+   sha256_8x32_context ictx;
-   sha256_8way_context octx;
+   sha256_8x32_context octx;
 } hmac_sha256_8way_context;
 
 //void SHA256_Buf( const void *, size_t len, uint8_t digest[32] );
@@ -84,12 +85,12 @@ void pbkdf2_sha256_8way( uint8_t *, size_t, const uint8_t *, size_t,
 
 #endif  // AVX2
        
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 typedef struct _hmac_sha256_16way_context
 {
-   sha256_16way_context ictx;
+   sha256_16x32_context ictx;
-   sha256_16way_context octx;
+   sha256_16x32_context octx;
 } hmac_sha256_16way_context;
 
 //void SHA256_Buf( const void *, size_t len, uint8_t digest[32] );

algo/sha/sha1-hash.c

@@ -205,7 +205,7 @@ void sha1_x86_sha_transform_be( uint32_t *state_out, const void *input,
 
 #endif
 
-#if defined(__aarch64__) && defined(__ARM_FEATURE_SHA2)
+#if defined(__ARM_NEON) && defined(__ARM_FEATURE_SHA2)
 
 #define sha1_neon_rounds( state_out, data, state_in ) \
 { \

algo/sha/sha256-hash-4way.c

@@ -30,6 +30,7 @@ static const uint32_t K256[64] =
    0x90BEFFFA, 0xA4506CEB, 0xBEF9A3F7, 0xC67178F2
 };
 
+#if defined(__SSE2__) || defined(__ARM_NEON)
 // SHA-256 4 way SSE2
 
 #define CHs(X, Y, Z) \
@@ -309,142 +310,6 @@ void sha256_4x32_final_rounds( v128_t *state_out, const v128_t *data,
    v128_store( state_out + 7,  H );
 }
 
-
-# if 0
-
-// Working correctly but still slower
-int sha256_4x32_transform_le_short( v128_t *state_out, const v128_t *data,
-                            const v128_t *state_in, const uint32_t *target )
-{
-   v128_t A, B, C, D, E, F, G, H, T0, T1, T2;
-   v128_t vmask, targ, hash;
-   int t6_mask, flip;
-   v128_t W[16];      v128_memcpy( W, data, 16 );
-
-   A = v128_load( state_in   );
-   B = v128_load( state_in+1 );
-   C = v128_load( state_in+2 );
-   D = v128_load( state_in+3 );
-   E = v128_load( state_in+4 );
-   F = v128_load( state_in+5 );
-   G = v128_load( state_in+6 );
-   H = v128_load( state_in+7 );
-
-   const v128_t IV7 = H;
-   const v128_t IV6 = G;
-
-   SHA256_4X32_16ROUNDS( A, B, C, D, E, F, G, H, 0 );
-   SHA256_4X32_MSG_EXPANSION( W );
-   SHA256_4X32_16ROUNDS( A, B, C, D, E, F, G, H, 16 );
-   SHA256_4X32_MSG_EXPANSION( W );
-   SHA256_4X32_16ROUNDS( A, B, C, D, E, F, G, H, 32 );
-
-   W[ 0] = SHA256_4X32_MEXP( W[14], W[ 9], W[ 1], W[ 0] );
-   W[ 1] = SHA256_4X32_MEXP( W[15], W[10], W[ 2], W[ 1] );
-   W[ 2] = SHA256_4X32_MEXP( W[ 0], W[11], W[ 3], W[ 2] );
-   W[ 3] = SHA256_4X32_MEXP( W[ 1], W[12], W[ 4], W[ 3] );
-   W[ 4] = SHA256_4X32_MEXP( W[ 2], W[13], W[ 5], W[ 4] );
-   W[ 5] = SHA256_4X32_MEXP( W[ 3], W[14], W[ 6], W[ 5] );
-   W[ 6] = SHA256_4X32_MEXP( W[ 4], W[15], W[ 7], W[ 6] );
-   W[ 7] = SHA256_4X32_MEXP( W[ 5], W[ 0], W[ 8], W[ 7] );
-   W[ 8] = SHA256_4X32_MEXP( W[ 6], W[ 1], W[ 9], W[ 8] );
-   W[ 9] = SHA256_4X32_MEXP( W[ 7], W[ 2], W[10], W[ 9] );
-   W[10] = SHA256_4X32_MEXP( W[ 8], W[ 3], W[11], W[10] );
-   W[11] = SHA256_4X32_MEXP( W[ 9], W[ 4], W[12], W[11] );
-   W[12] = SHA256_4X32_MEXP( W[10], W[ 5], W[13], W[12] );
-
-   v128_t X_xor_Y, Y_xor_Z = v128_xor( B, C );
-   
-   SHA256_4X32_ROUND( A, B, C, D, E, F, G, H,  0, 48 );
-   SHA256_4X32_ROUND( H, A, B, C, D, E, F, G,  1, 48 );
-   SHA256_4X32_ROUND( G, H, A, B, C, D, E, F,  2, 48 );
-   SHA256_4X32_ROUND( F, G, H, A, B, C, D, E,  3, 48 );
-   SHA256_4X32_ROUND( E, F, G, H, A, B, C, D,  4, 48 );
-   SHA256_4X32_ROUND( D, E, F, G, H, A, B, C,  5, 48 );
-   SHA256_4X32_ROUND( C, D, E, F, G, H, A, B,  6, 48 );
-   SHA256_4X32_ROUND( B, C, D, E, F, G, H, A,  7, 48 );
-   SHA256_4X32_ROUND( A, B, C, D, E, F, G, H,  8, 48 );
-   SHA256_4X32_ROUND( H, A, B, C, D, E, F, G,  9, 48 );
-
-   T0 = v128_add32( v128_32( K256[58] ),
-                   v128_add4_32( BSG2_1( C ), CHs( C, D, E ), W[10], F ) );
-   B = v128_add32( B, T0 );
-
-   T1 = v128_add32( v128_32( K256[59] ),
-                    v128_add4_32( BSG2_1( B ), CHs( B, C, D ), W[11], E ) );
-   A = v128_add32( A, T1 );
-
-   T2 = v128_add32( v128_32( K256[60] ),
-                    v128_add4_32( BSG2_1( A ), CHs( A, B, C ), W[12], D ) );
-   H = v128_add32( H, T2 );
-
-   targ = v128_32( target[7] );
-   hash = v128_bswap32( v128_add32( H, IV7 ) );
-
-   flip = ( (int)target[7] < 0 ? 0xf : 0 ) ^ v128_movmask32( hash );
-
-   if ( likely(
-             0xf == ( flip ^ v128_movmask32( v128_cmpgt32( hash, targ ) ) ) ))
-   return 0;
-
-   t6_mask = v128_movmask32( vmask = v128_cmpeq32( hash, targ ) );
-
-   // round 58 part 2
-   F = v128_add32( T0, v128_add32( BSG2_0( G ), MAJs( G, H, A ) ) );
-
-   // round 61  part 1
-   W[13] = SHA256_4X32_MEXP( W[11], W[ 6], W[14], W[13] );
-   T0 = v128_add32( v128_32( K256[61] ),
-                    v128_add4_32( BSG2_1( H ), CHs( H, A, B ), W[13], C ) );
-   G = v128_add32( G, T0 );
-
-   if ( t6_mask )
-   {
-      targ = v128_and( vmask, v128_32( target[6] ) );
-      hash = v128_bswap32( v128_add32( G, IV6 ) );
-
-      if ( ( 0 != ( t6_mask & v128_movmask32( v128_cmpeq32( hash, targ ) ) ) ))
-         return 0;
-      else
-      {
-         flip = ( (int)target[6] < 0 ? 0xf : 0 ) ^ v128_movmask32( hash );
-         if ( 0 != ( t6_mask & ( flip ^ v128_movmask32(
-                                             v128_cmpgt32( hash, targ ) ) ) ) )
-            return 0;
-          else if ( target[6] == 0x80000000 )
-          {
-             if ( 0 == ( t6_mask & v128_movmask32(
-                            v128_cmpgt32( hash, v128_xor( hash, hash ) ) ) ) )
-                return 0;
-          }
-       }
-   }
-   
-   // rounds 59 to 61 part 2
-   E = v128_add32( T1, v128_add32( BSG2_0( F ), MAJs( F, G, H ) ) );
-   D = v128_add32( T2, v128_add32( BSG2_0( E ), MAJs( E, F, G ) ) );
-   C = v128_add32( T0, v128_add32( BSG2_0( D ), MAJs( D, E, F ) ) );
-
-   // rounds 62 & 63
-   W[14] = SHA256_4X32_MEXP( W[12], W[ 7], W[15], W[14] );
-   W[15] = SHA256_4X32_MEXP( W[13], W[ 8], W[ 0], W[15] );
-
-   SHA256_4X32_ROUND( C, D, E, F, G, H, A, B, 14, 48 );
-   SHA256_4X32_ROUND( B, C, D, E, F, G, H, A, 15, 48 );
-
-   state_out[0] = v128_add32( state_in[0], A );
-   state_out[1] = v128_add32( state_in[1], B );
-   state_out[2] = v128_add32( state_in[2], C );
-   state_out[3] = v128_add32( state_in[3], D );
-   state_out[4] = v128_add32( state_in[4], E );
-   state_out[5] = v128_add32( state_in[5], F );
-   state_out[6] = v128_add32( state_in[6], G );
-   state_out[7] = v128_add32( state_in[7], H );
-return 1;
-}
-
-#endif
-
 void sha256_4x32_init( sha256_4x32_context *sc )
 {
    sc->count_high = sc->count_low = 0;
@@ -529,28 +394,30 @@ void sha256_4x32_full( void *dst, const void *data, size_t len )
    sha256_4x32_close( &ctx, dst );
 }
 
+#endif  // SSE2 || NEON
+
 #if defined(__AVX2__)
 
 // SHA-256 8 way
 
 #define BSG2_0x(x) \
-   _mm256_xor_si256( _mm256_xor_si256( mm256_ror_32( x,  2 ), \
+   mm256_xor3( mm256_ror_32( x,  2 ), \
-                                       mm256_ror_32( x, 13 ) ), \
+               mm256_ror_32( x, 13 ), \
                mm256_ror_32( x, 22 ) )
 
 #define BSG2_1x(x) \
-   _mm256_xor_si256( _mm256_xor_si256( mm256_ror_32( x,  6 ), \
+   mm256_xor3( mm256_ror_32( x,  6 ), \
-                                       mm256_ror_32( x, 11 ) ), \
+               mm256_ror_32( x, 11 ), \
                mm256_ror_32( x, 25 ) )
 
 #define SSG2_0x(x) \
-   _mm256_xor_si256( _mm256_xor_si256( mm256_ror_32( x,  7 ), \
+   mm256_xor3( mm256_ror_32( x,  7 ), \
-                                       mm256_ror_32( x, 18 ) ), \
+               mm256_ror_32( x, 18 ), \
                _mm256_srli_epi32( x, 3 ) ) 
 
 #define SSG2_1x(x) \
-   _mm256_xor_si256( _mm256_xor_si256( mm256_ror_32( x, 17 ), \
+   mm256_xor3( mm256_ror_32( x, 17 ), \
-                                       mm256_ror_32( x, 19 ) ), \
+               mm256_ror_32( x, 19 ), \
                _mm256_srli_epi32( x, 10 ) )
 
 #define SHA256_8WAY_MEXP( a, b, c, d ) \
@@ -574,62 +441,6 @@ void sha256_4x32_full( void *dst, const void *data, size_t len )
       W[14] = SHA256_8WAY_MEXP( W[12], W[ 7], W[15], W[14] ); \
       W[15] = SHA256_8WAY_MEXP( W[13], W[ 8], W[ 0], W[15] ); 
 
-
-// With AVX512VL ternary logic optimizations are available.
-// If not optimize by forwarding the result of X^Y in MAJ to the next round
-// to avoid recalculating it as Y^Z. This optimization is not applicable
-// when MAJ is optimized with ternary logic.
-
-#if defined(__AVX512VL__)
-
-#define CHx(X, Y, Z)    _mm256_ternarylogic_epi32( X, Y, Z, 0xca )
-
-#define MAJx(X, Y, Z)   _mm256_ternarylogic_epi32( X, Y, Z, 0xe8 )
-
-#define SHA256_8WAY_ROUND( A, B, C, D, E, F, G, H, i, j ) \
-do { \
-  __m256i T0 = _mm256_add_epi32( v256_32( K256[ (j)+(i) ] ), \
-                                 W[ i ] ); \
-  __m256i T1 = BSG2_1x( E ); \
-  __m256i T2 = BSG2_0x( A ); \
-  T0 = _mm256_add_epi32( T0, CHx( E, F, G ) ); \
-  T1 = _mm256_add_epi32( T1, H ); \
-  T2 = _mm256_add_epi32( T2, MAJx( A, B, C ) ); \
-  T1 = _mm256_add_epi32( T1, T0 ); \
-  D  = _mm256_add_epi32( D,  T1 ); \
-  H  = _mm256_add_epi32( T1, T2 ); \
-} while (0)
-
-#define SHA256_8WAY_16ROUNDS( A, B, C, D, E, F, G, H, j ) \
-   SHA256_8WAY_ROUND( A, B, C, D, E, F, G, H,  0, j ); \
-   SHA256_8WAY_ROUND( H, A, B, C, D, E, F, G,  1, j ); \
-   SHA256_8WAY_ROUND( G, H, A, B, C, D, E, F,  2, j ); \
-   SHA256_8WAY_ROUND( F, G, H, A, B, C, D, E,  3, j ); \
-   SHA256_8WAY_ROUND( E, F, G, H, A, B, C, D,  4, j ); \
-   SHA256_8WAY_ROUND( D, E, F, G, H, A, B, C,  5, j ); \
-   SHA256_8WAY_ROUND( C, D, E, F, G, H, A, B,  6, j ); \
-   SHA256_8WAY_ROUND( B, C, D, E, F, G, H, A,  7, j ); \
-   SHA256_8WAY_ROUND( A, B, C, D, E, F, G, H,  8, j ); \
-   SHA256_8WAY_ROUND( H, A, B, C, D, E, F, G,  9, j ); \
-   SHA256_8WAY_ROUND( G, H, A, B, C, D, E, F, 10, j ); \
-   SHA256_8WAY_ROUND( F, G, H, A, B, C, D, E, 11, j ); \
-   SHA256_8WAY_ROUND( E, F, G, H, A, B, C, D, 12, j ); \
-   SHA256_8WAY_ROUND( D, E, F, G, H, A, B, C, 13, j ); \
-   SHA256_8WAY_ROUND( C, D, E, F, G, H, A, B, 14, j ); \
-   SHA256_8WAY_ROUND( B, C, D, E, F, G, H, A, 15, j );
-
-// Not used with AVX512, needed to satisfy the compiler
-#define SHA256_8WAY_ROUND_NOMSG( A, B, C, D, E, F, G, H, i, j ) \
-{ \
-   __m256i T1 = mm256_add4_32( H, BSG2_1x(E), CHx(E, F, G), \
-                              v256_32( K256[(i)+(j)] ) ); \
-   __m256i T2 = _mm256_add_epi32( BSG2_0x(A), MAJx(A, B, C) ); \
-   D  = _mm256_add_epi32( D,  T1 ); \
-   H  = _mm256_add_epi32( T1, T2 ); \
-}
-
-#else  // AVX2
-
 #define CHx(X, Y, Z) \
    _mm256_xor_si256( _mm256_and_si256( _mm256_xor_si256( Y, Z ), X ), Z ) 
 
@@ -641,61 +452,58 @@ do { \
 
 #define SHA256_8WAY_ROUND_NOMSG( A, B, C, D, E, F, G, H, i, j ) \
 { \
-   __m256i T1 = mm256_add4_32( H, BSG2_1x(E), CHx(E, F, G), \
+   H = mm256_add4_32( H, BSG2_1x(E), CHx(E, F, G), \
                               v256_32( K256[(i)+(j)] ) ); \
-   __m256i T2 = _mm256_add_epi32( BSG2_0x(A), MAJx(A, B, C) ); \
+   __m256i T = _mm256_add_epi32( BSG2_0x(A), MAJx(A, B, C) ); \
    Y_xor_Z = X_xor_Y; \
-   D  = _mm256_add_epi32( D,  T1 ); \
+   D  = _mm256_add_epi32( D, H ); \
-   H  = _mm256_add_epi32( T1, T2 ); \
+   H  = _mm256_add_epi32( H, T ); \
 }
 
 #define SHA256_8WAY_ROUND( A, B, C, D, E, F, G, H, i, j ) \
-do { \
+{ \
-  __m256i T0 = _mm256_add_epi32( v256_32( K256[(j)+(i)] ), W[i] ); \
+  __m256i T1 = _mm256_add_epi32( v256_32( K256[(j)+(i)] ), W[i] ); \
-  __m256i T1 = BSG2_1x( E ); \
+  H = _mm256_add_epi32( H, BSG2_1x( E ) ); \
   __m256i T2 = BSG2_0x( A ); \
-  T0 = _mm256_add_epi32( T0, CHx( E, F, G ) ); \
+  T1 = _mm256_add_epi32( T1, CHx( E, F, G ) ); \
-  T1 = _mm256_add_epi32( T1, H ); \
   T2 = _mm256_add_epi32( T2, MAJx( A, B, C ) ); \
-  T1 = _mm256_add_epi32( T1, T0 ); \
+  H = _mm256_add_epi32( H, T1 ); \
   Y_xor_Z = X_xor_Y; \
-  D  = _mm256_add_epi32( D,  T1 ); \
+  D  = _mm256_add_epi32( D,  H ); \
-  H  = _mm256_add_epi32( T1, T2 ); \
+  H  = _mm256_add_epi32( H, T2 ); \
-} while (0)
+}
 
 // read Y_xor_Z, update X_xor_Y
 #define MAJ_2step(X, Y, Z, X_xor_Y, Y_xor_Z ) \
   _mm256_xor_si256( Y, _mm256_and_si256( X_xor_Y = _mm256_xor_si256( X, Y ), \
                                          Y_xor_Z ) )
 
-// start with toc initialized to y^z:   toc = B ^ C
+// start with toc initialized to y^z, toc = B ^ C for first ound.
 // First round reads toc as Y_xor_Z and saves X_xor_Y as tic.
 // Second round reads tic as Y_xor_Z and saves X_xor_Y as toc.
 
 #define SHA256_8WAY_2ROUNDS( A, B, C, D, E, F, G, H, i0, i1, j ) \
-do { \
+{ \
-  __m256i T0 = _mm256_add_epi32( v256_32( K256[ (j)+(i0) ] ), \
+  __m256i T1 = _mm256_add_epi32( v256_32( K256[ (j)+(i0) ] ), \
                                  W[ i0 ] ); \
-  __m256i T1 = BSG2_1x( E ); \
+  H = _mm256_add_epi32( H, BSG2_1x( E ) ); \
   __m256i T2 = BSG2_0x( A ); \
-  T0 = _mm256_add_epi32( T0, CHx( E, F, G ) ); \
+  T1 = _mm256_add_epi32( T1, CHx( E, F, G ) ); \
-  T1 = _mm256_add_epi32( T1, H ); \
   T2 = _mm256_add_epi32( T2, MAJ_2step( A, B, C, tic, toc ) ); \
-  T1 = _mm256_add_epi32( T1, T0 ); \
+  H = _mm256_add_epi32( H, T1 ); \
-  D  = _mm256_add_epi32( D,  T1 ); \
+  D  = _mm256_add_epi32( D,  H ); \
-  H  = _mm256_add_epi32( T1, T2 ); \
+  H  = _mm256_add_epi32( H, T2 ); \
 \
-  T0 = _mm256_add_epi32( v256_32( K256[ (j)+(i1) ] ), \
+  T1 = _mm256_add_epi32( v256_32( K256[ (j)+(i1) ] ), \
                                  W[ (i1) ] ); \
-  T1 = BSG2_1x( D ); \
+  G = _mm256_add_epi32( G, BSG2_1x( D ) ); \
   T2 = BSG2_0x( H ); \
-  T0 = _mm256_add_epi32( T0, CHx( D, E, F ) ); \
+  T1 = _mm256_add_epi32( T1, CHx( D, E, F ) ); \
-  T1 = _mm256_add_epi32( T1, G ); \
   T2 = _mm256_add_epi32( T2, MAJ_2step( H, A, B, toc, tic ) ); \
-  T1 = _mm256_add_epi32( T1, T0 ); \
+  G = _mm256_add_epi32( G, T1 ); \
-  C  = _mm256_add_epi32( C,  T1 ); \
+  C  = _mm256_add_epi32( C,  G ); \
-  G  = _mm256_add_epi32( T1, T2 ); \
+  G  = _mm256_add_epi32( G, T2 ); \
-} while (0)
+}
 
 #define SHA256_8WAY_16ROUNDS( A, B, C, D, E, F, G, H, j ) \
 { \
@@ -710,8 +518,6 @@ do { \
    SHA256_8WAY_2ROUNDS( C, D, E, F, G, H, A, B, 14, 15, j ); \
 }
 
-#endif   // AVX512VL else AVX2
-
 static inline void SHA256_8WAY_TRANSFORM( __m256i *out, __m256i *W,
                                           const  __m256i *in ) \
 {
@@ -745,7 +551,7 @@ static inline void SHA256_8WAY_TRANSFORM( __m256i *out, __m256i *W,
 }
 
 // accepts LE input data
-void sha256_8way_transform_le( __m256i *state_out, const __m256i *data,
+void sha256_8x32_transform_le( __m256i *state_out, const __m256i *data,
                                const __m256i *state_in )
 {
    __m256i W[16];
@@ -754,7 +560,7 @@ void sha256_8way_transform_le( __m256i *state_out, const __m256i *data,
 }
 
 // Accepts BE input data, need to bswap
-void sha256_8way_transform_be( __m256i *state_out, const __m256i *data,
+void sha256_8x32_transform_be( __m256i *state_out, const __m256i *data,
                                const __m256i *state_in )
 {
    __m256i W[16];
@@ -764,7 +570,7 @@ void sha256_8way_transform_be( __m256i *state_out, const __m256i *data,
 }
 
 // Aggressive prehashing, LE byte order
-void sha256_8way_prehash_3rounds( __m256i *state_mid, __m256i *X,
+void sha256_8x32_prehash_3rounds( __m256i *state_mid, __m256i *X,
                                   const __m256i *W, const __m256i *state_in )
 {
    __m256i A, B, C, D, E, F, G, H, T1;
@@ -788,9 +594,7 @@ void sha256_8way_prehash_3rounds( __m256i *state_mid, __m256i *X,
    G = _mm256_load_si256( state_in + 6 );
    H = _mm256_load_si256( state_in + 7 );
 
-#if !defined(__AVX512VL__)
    __m256i X_xor_Y, Y_xor_Z = _mm256_xor_si256( B, C );
-#endif
 
    SHA256_8WAY_ROUND( A, B, C, D, E, F, G, H,  0, 0 );
    SHA256_8WAY_ROUND( H, A, B, C, D, E, F, G,  1, 0 );
@@ -813,7 +617,7 @@ void sha256_8way_prehash_3rounds( __m256i *state_mid, __m256i *X,
    _mm256_store_si256( state_mid + 7, H );
 }
 
-void sha256_8way_final_rounds( __m256i *state_out, const __m256i *data,
+void sha256_8x32_final_rounds( __m256i *state_out, const __m256i *data,
           const __m256i *state_in, const __m256i *state_mid, const __m256i *X )
 {
    __m256i A, B, C, D, E, F, G, H;
@@ -830,9 +634,7 @@ void sha256_8way_final_rounds( __m256i *state_out, const __m256i *data,
    G = _mm256_load_si256( state_mid + 6 );
    H = _mm256_load_si256( state_mid + 7 );
 
-#if !defined(__AVX512VL__)
    __m256i X_xor_Y, Y_xor_Z = _mm256_xor_si256( F, G );
-#endif
 
    // round 3 part 2, add nonces  
    A = _mm256_add_epi32( A, W[3] );
@@ -914,15 +716,13 @@ void sha256_8way_final_rounds( __m256i *state_out, const __m256i *data,
    _mm256_store_si256( state_out + 7,  H );
 }
 
-int sha256_8way_transform_le_short( __m256i *state_out, const __m256i *data,
+int sha256_8x32_transform_le_short( __m256i *state_out, const __m256i *data,
                            const __m256i *state_in, const uint32_t *target )
 {
-   __m256i A, B, C, D, E, F, G, H, T0, T1, T2;
+   __m256i A, B, C, D, E, F, G, H, G57, H56;
    __m256i vmask, targ, hash;
    __m256i W[16];  memcpy_256( W, data, 16 );
-   const __m256i bswap_shuf = mm256_bcast_m128( _mm_set_epi64x(
+   uint8_t flip, t6_mask, t7_mask;
-                              0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
-   uint8_t flip, t6_mask;
 
    A = _mm256_load_si256( state_in   );
    B = _mm256_load_si256( state_in+1 );
@@ -933,12 +733,10 @@ int sha256_8way_transform_le_short( __m256i *state_out, const __m256i *data,
    G = _mm256_load_si256( state_in+6 );
    H = _mm256_load_si256( state_in+7 );
 
-   const __m256i IV7 = H;
+   const __m256i istate6 = G;
-   const __m256i IV6 = G;
+   const __m256i istate7 = H;
 
-#if !defined(__AVX512VL__)
    __m256i X_xor_Y, Y_xor_Z = _mm256_xor_si256( B, C );
-#endif
 
    // rounds 0 to 16, ignore zero padding W[9..14]
    SHA256_8WAY_ROUND(       A, B, C, D, E, F, G, H,  0, 0 );
@@ -981,11 +779,9 @@ int sha256_8way_transform_le_short( __m256i *state_out, const __m256i *data,
    W[11] = SHA256_8WAY_MEXP( W[ 9], W[ 4], W[12], W[11] );
    W[12] = SHA256_8WAY_MEXP( W[10], W[ 5], W[13], W[12] );
 
-#if !defined(__AVX512VL__)
    Y_xor_Z = _mm256_xor_si256( B, C );
-#endif
 
-   // rounds 48 to 57
+   // Rounds 48 to 55
    SHA256_8WAY_ROUND( A, B, C, D, E, F, G, H,  0, 48 );
    SHA256_8WAY_ROUND( H, A, B, C, D, E, F, G,  1, 48 );
    SHA256_8WAY_ROUND( G, H, A, B, C, D, E, F,  2, 48 );
@@ -994,77 +790,83 @@ int sha256_8way_transform_le_short( __m256i *state_out, const __m256i *data,
    SHA256_8WAY_ROUND( D, E, F, G, H, A, B, C,  5, 48 );
    SHA256_8WAY_ROUND( C, D, E, F, G, H, A, B,  6, 48 );
    SHA256_8WAY_ROUND( B, C, D, E, F, G, H, A,  7, 48 );
-   SHA256_8WAY_ROUND( A, B, C, D, E, F, G, H,  8, 48 );
-   SHA256_8WAY_ROUND( H, A, B, C, D, E, F, G,  9, 48 );
 
-   // round 58 to 60 part 1
+   // Round 56
-   T0 = _mm256_add_epi32( v256_32( K256[58] ),
+   H = _mm256_add_epi32( v256_32( K256[56] ),
+                 mm256_add4_32( BSG2_1x( E ), CHx( E, F, G ), W[ 8], H ) );
+   D = _mm256_add_epi32( D, H );
+   H56 = _mm256_add_epi32( H, _mm256_add_epi32( BSG2_0x( A ),
+                                                   MAJx( A, B, C ) ) );
+   Y_xor_Z = X_xor_Y; 
+   
+   // Rounds 57 to 60 part 1
+   G = _mm256_add_epi32( v256_32( K256[57] ),
+                 mm256_add4_32( BSG2_1x( D ), CHx( D, E, F ), W[ 9], G ) );
+   C = _mm256_add_epi32( C, G );
+   G57 = _mm256_add_epi32( G, MAJx( H56, A, B ) );
+   
+   F = _mm256_add_epi32( v256_32( K256[58] ),
                  mm256_add4_32( BSG2_1x( C ), CHx( C, D, E ), W[10], F ) );
-   B = _mm256_add_epi32( B, T0 );
+   B = _mm256_add_epi32( B, F );
 
-   T1 = _mm256_add_epi32( v256_32( K256[59] ),
+   E = _mm256_add_epi32( v256_32( K256[59] ),
                  mm256_add4_32( BSG2_1x( B ), CHx( B, C, D ), W[11], E ) );
-   A = _mm256_add_epi32( A, T1 );
+   A = _mm256_add_epi32( A, E );
 
-   T2 = _mm256_add_epi32( v256_32( K256[60] ),
+   D = _mm256_add_epi32( v256_32( K256[60] ),
                  mm256_add4_32( BSG2_1x( A ), CHx( A, B, C ), W[12], D ) );
-   H = _mm256_add_epi32( H, T2 );
+   H = _mm256_add_epi32( H56, D );
 
    // Got H, test it.
+   hash = mm256_bswap_32( _mm256_add_epi32( H, istate7 ) );
    targ = v256_32( target[7] );
-   hash = _mm256_shuffle_epi8( _mm256_add_epi32( H, IV7 ), bswap_shuf );
+   // A simple unsigned LE test is complicated by the lack of a cmple
-   if ( target[7] )
+   // instruction, and lack of unsigned compares in AVX2.
-   {
    flip = ( (int)target[7] < 0 ? -1 : 0 ) ^ mm256_movmask_32( hash );
-      if ( likely( 0xff == ( flip ^
+   if ( likely( 0xff == ( t7_mask = ( flip ^
-                    mm256_movmask_32( _mm256_cmpgt_epi32( hash, targ ) ) ) ))
+                    mm256_movmask_32( _mm256_cmpgt_epi32( hash, targ ) ) ) )))
       return 0;
-   }
    t6_mask = mm256_movmask_32( vmask =_mm256_cmpeq_epi32( hash, targ ) );
 
-   // round 58 part 2
+   // Round 57 part 2
-   F = _mm256_add_epi32( T0, _mm256_add_epi32( BSG2_0x( G ),
+   G57 = _mm256_add_epi32( G57, BSG2_0x( H56 ) );
-                                               MAJx( G, H, A ) ) );
+   Y_xor_Z = X_xor_Y;
-   // round 61 part 1
-   W[13] = SHA256_8WAY_MEXP( W[11], W[ 6], W[14], W[13] );
-   T0 = _mm256_add_epi32( v256_32( K256[61] ),
-                 mm256_add4_32( BSG2_1x( H ), CHx( H, A, B ), W[13], C ) );
-   G = _mm256_add_epi32( G, T0 );
 
-   if ( t6_mask )
+   // Round 61 part 1
+   W[13] = SHA256_8WAY_MEXP( W[11], W[ 6], W[14], W[13] );
+   C = _mm256_add_epi32( v256_32( K256[61] ),
+                 mm256_add4_32( BSG2_1x( H ), CHx( H, A, B ), W[13], C ) );
+   G = _mm256_add_epi32( G57, C );
+
+   if ( t6_mask == (0xff & ~t7_mask ) )
    { 
       // Testing H was inconclusive: hash7 == target7, need to test G
       targ = _mm256_and_si256( vmask, v256_32( target[6] ) );
-      hash = _mm256_shuffle_epi8( _mm256_add_epi32( G, IV6 ), bswap_shuf );
+      hash = mm256_bswap_32( _mm256_add_epi32( G, istate6 ) );
-
-      if ( likely( 0 == ( t6_mask & mm256_movmask_32(
-                                      _mm256_cmpeq_epi32( hash, targ ) ) ) ))
-      {
       flip = ( (int)target[6] < 0 ? -1 : 0 ) ^ mm256_movmask_32( hash );
       if ( likely( 0 != ( t6_mask & ( flip ^
                    mm256_movmask_32( _mm256_cmpgt_epi32( hash, targ ) ) ) ) ))
          return 0;
-          if ( likely( ( target[6] == 0x80000000 )
-             && ( 0 == ( t6_mask & mm256_movmask_32( _mm256_cmpgt_epi32(
-                              hash, _mm256_xor_si256( hash, hash ) ) ) ) ) ))
-             return 0;
-       } 
-//     else inconclusive, testing targ5 isn't practical, fininsh hashing  
    }
 
-// At this point either the hash will be good or the test was inconclusive.
+   // Rounds 58 to 61 part 2
-// If the latter it's probably a high target difficulty with a nearly equal
+   F = _mm256_add_epi32( F, _mm256_add_epi32( BSG2_0x( G57 ),
-// high difficulty hash that has a good chance of being good.  
+                                                 MAJx( G57, H, A ) ) );
+   Y_xor_Z = X_xor_Y;
 
-   // rounds 59 to 61 part 2
+   E = _mm256_add_epi32( E, _mm256_add_epi32( BSG2_0x( F ),
-   E = _mm256_add_epi32( T1, _mm256_add_epi32( BSG2_0x( F ),
+                                                 MAJx( F, G57, H ) ) );
-                                               MAJx( F, G, H ) ) );
+   Y_xor_Z = X_xor_Y;
-   D = _mm256_add_epi32( T2, _mm256_add_epi32( BSG2_0x( E ),
+
-                                               MAJx( E, F, G ) ) );
+   D = _mm256_add_epi32( D, _mm256_add_epi32( BSG2_0x( E ),
-   C = _mm256_add_epi32( T0, _mm256_add_epi32( BSG2_0x( D ),
+                                                 MAJx( E, F, G57 ) ) );
+   Y_xor_Z = X_xor_Y;
+
+   C = _mm256_add_epi32( C, _mm256_add_epi32( BSG2_0x( D ),
                                                  MAJx( D, E, F ) ) );
+   Y_xor_Z = X_xor_Y;
 
-   // rounds 62 & 63
+   // Rounds 62 & 63
    W[14] = SHA256_8WAY_MEXP( W[12], W[ 7], W[15], W[14] );
    W[15] = SHA256_8WAY_MEXP( W[13], W[ 8], W[ 0], W[15] );
 
@@ -1083,8 +885,7 @@ int sha256_8way_transform_le_short( __m256i *state_out, const __m256i *data,
    return 1;
 }
 
-
+void sha256_8x32_init( sha256_8x32_context *sc )
-void sha256_8way_init( sha256_8way_context *sc )
 {
    sc->count_high = sc->count_low = 0;
    sc->val[0] = v256_32( sha256_iv[0] );
@@ -1100,7 +901,7 @@ void sha256_8way_init( sha256_8way_context *sc )
 // need to handle odd byte length for yespower.
 // Assume only last update is odd.
 
-void sha256_8way_update( sha256_8way_context *sc, const void *data, size_t len )
+void sha256_8x32_update( sha256_8x32_context *sc, const void *data, size_t len )
 {
    __m256i *vdata = (__m256i*)data;
    size_t ptr;
@@ -1121,7 +922,7 @@ void sha256_8way_update( sha256_8way_context *sc, const void *data, size_t len )
       len -= clen;
       if ( ptr == buf_size )
       {
-         sha256_8way_transform_be( sc->val, sc->buf, sc->val );
+         sha256_8x32_transform_be( sc->val, sc->buf, sc->val );
          ptr = 0;
       }
       clow = sc->count_low;
@@ -1132,7 +933,7 @@ void sha256_8way_update( sha256_8way_context *sc, const void *data, size_t len )
    }
 }
 
-void sha256_8way_close( sha256_8way_context *sc, void *dst )
+void sha256_8x32_close( sha256_8x32_context *sc, void *dst )
 {
     unsigned ptr;
     uint32_t low, high;
@@ -1146,7 +947,7 @@ void sha256_8way_close( sha256_8way_context *sc, void *dst )
     if ( ptr > pad )
     {
          memset_zero_256( sc->buf + (ptr>>2), (buf_size - ptr) >> 2 );
-         sha256_8way_transform_be( sc->val, sc->buf, sc->val );
+         sha256_8x32_transform_be( sc->val, sc->buf, sc->val );
          memset_zero_256( sc->buf, pad >> 2 );
     }
     else
@@ -1159,20 +960,20 @@ void sha256_8way_close( sha256_8way_context *sc, void *dst )
     sc->buf[   pad     >> 2 ] = v256_32( bswap_32( high ) );
     sc->buf[ ( pad+4 ) >> 2 ] = v256_32( bswap_32( low ) );
 
-    sha256_8way_transform_be( sc->val, sc->buf, sc->val );
+    sha256_8x32_transform_be( sc->val, sc->buf, sc->val );
 
     mm256_block_bswap_32( dst, sc->val );
 }
 
-void sha256_8way_full( void *dst, const void *data, size_t len )
+void sha256_8x32_full( void *dst, const void *data, size_t len )
 {
-   sha256_8way_context ctx;
+   sha256_8x32_context ctx;
-   sha256_8way_init( &ctx );
+   sha256_8x32_init( &ctx );
-   sha256_8way_update( &ctx, data, len );
+   sha256_8x32_update( &ctx, data, len );
-   sha256_8way_close( &ctx, dst );
+   sha256_8x32_close( &ctx, dst );
 }
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
 // SHA-256 16 way
 
@@ -1218,40 +1019,26 @@ void sha256_8way_full( void *dst, const void *data, size_t len )
    W[15] = SHA256_16WAY_MEXP( W[13], W[ 8], W[ 0], W[15] );
 
 #define SHA256_16WAY_ROUND( A, B, C, D, E, F, G, H, i, j ) \
-do { \
+{ \
-  __m512i T0 = _mm512_add_epi32( v512_32( K256[(j)+(i)] ), W[i] ); \
+  __m512i T1 = _mm512_add_epi32( v512_32( K256[(j)+(i)] ), W[i] ); \
-  __m512i T1 = BSG2_1x16( E ); \
+  H = _mm512_add_epi32( H, BSG2_1x16( E ) ); \
   __m512i T2 = BSG2_0x16( A ); \
-  T0 = _mm512_add_epi32( T0, CHx16( E, F, G ) ); \
+  T1 = _mm512_add_epi32( T1, CHx16( E, F, G ) ); \
-  T1 = _mm512_add_epi32( T1, H ); \
   T2 = _mm512_add_epi32( T2, MAJx16( A, B, C ) ); \
-  T1 = _mm512_add_epi32( T1, T0 ); \
+  H = _mm512_add_epi32( H, T1 ); \
-  D  = _mm512_add_epi32( D,  T1 ); \
+  D  = _mm512_add_epi32( D,  H ); \
-  H  = _mm512_add_epi32( T1, T2 ); \
+  H  = _mm512_add_epi32( H, T2 ); \
-} while (0)
+}
    
 #define SHA256_16WAY_ROUND_NOMSG( A, B, C, D, E, F, G, H, i, j ) \
 { \
-   __m512i T1 = mm512_add4_32( H, BSG2_1x16(E), CHx16(E, F, G), \
+   H = mm512_add4_32( H, BSG2_1x16(E), CHx16(E, F, G), \
                               v512_32( K256[(i)+(j)] ) ); \
-   __m512i T2 = _mm512_add_epi32( BSG2_0x16(A), MAJx16(A, B, C) ); \
+   __m512i T = _mm512_add_epi32( BSG2_0x16(A), MAJx16(A, B, C) ); \
-   D  = _mm512_add_epi32( D,  T1 ); \
+   D  = _mm512_add_epi32( D, H ); \
-   H  = _mm512_add_epi32( T1, T2 ); \
+   H  = _mm512_add_epi32( H, T ); \
 }
 
-/*
-#define SHA256_16WAY_ROUND(A, B, C, D, E, F, G, H, i, j) \
-do { \
-  __m512i T1, T2; \
-  __m512i K = v512_32( K256[( (j)+(i) )] ); \
-  T1 = _mm512_add_epi32( H, mm512_add4_32( BSG2_1x16(E), CHx16(E, F, G), \
-                                           K, W[i] ) ); \
-  T2 = _mm512_add_epi32( BSG2_0x16(A), MAJx16(A, B, C) ); \
-  D  = _mm512_add_epi32( D,  T1 ); \
-  H  = _mm512_add_epi32( T1, T2 ); \
-} while (0)
-*/
-
 #define SHA256_16WAY_16ROUNDS( A, B, C, D, E, F, G, H, j ) \
    SHA256_16WAY_ROUND( A, B, C, D, E, F, G, H,  0, j ); \
    SHA256_16WAY_ROUND( H, A, B, C, D, E, F, G,  1, j ); \
@@ -1302,7 +1089,7 @@ static inline void SHA256_16WAY_TRANSFORM( __m512i *out, __m512i *W,
 }
 
 // accepts LE input data
-void sha256_16way_transform_le( __m512i *state_out, const __m512i *data,
+void sha256_16x32_transform_le( __m512i *state_out, const __m512i *data,
                                 const __m512i *state_in )
 {
    __m512i W[16];
@@ -1311,7 +1098,7 @@ void sha256_16way_transform_le( __m512i *state_out, const __m512i *data,
 }
 
 // Accepts BE input data, need to bswap
-void sha256_16way_transform_be( __m512i *state_out, const __m512i *data,
+void sha256_16x32_transform_be( __m512i *state_out, const __m512i *data,
                                 const __m512i *state_in )
 {
    __m512i W[16];
@@ -1321,7 +1108,7 @@ void sha256_16way_transform_be( __m512i *state_out, const __m512i *data,
 }
  
 // Aggressive prehashing, LE byte order
-void sha256_16way_prehash_3rounds( __m512i *state_mid, __m512i *X, 
+void sha256_16x32_prehash_3rounds( __m512i *state_mid, __m512i *X, 
                                    const __m512i *W, const __m512i *state_in )
 {
    __m512i A, B, C, D, E, F, G, H, T1;
@@ -1369,7 +1156,7 @@ void sha256_16way_prehash_3rounds( __m512i *state_mid, __m512i *X,
    _mm512_store_si512( state_mid + 7, H );
 }   
 
-void sha256_16way_final_rounds( __m512i *state_out, const __m512i *data,
+void sha256_16x32_final_rounds( __m512i *state_out, const __m512i *data,
           const __m512i *state_in, const __m512i *state_mid, const __m512i *X )
 {
    __m512i A, B, C, D, E, F, G, H;
@@ -1470,15 +1257,12 @@ void sha256_16way_final_rounds( __m512i *state_out, const __m512i *data,
 
 // returns 0 if hash aborted early and invalid,
 // returns 1 for completed hash with at least one valid candidate.
-int sha256_16way_transform_le_short( __m512i *state_out, const __m512i *data,
+int sha256_16x32_transform_le_short( __m512i *state_out, const __m512i *data,
                               const __m512i *state_in, const uint32_t *target )
 {
-   __m512i A, B, C, D, E, F, G, H, hash, targ;
+   __m512i A, B, C, D, E, F, G, H, hash, targ, G57, H56;
-   __m512i T0, T1, T2;
    __m512i W[16];      memcpy_512( W, data, 16 );
-   __mmask16 t6_mask;
+   __mmask16 mask;
-   const __m512i bswap_shuf = mm512_bcast_m128( _mm_set_epi64x(
-                              0x0c0d0e0f08090a0b, 0x0405060700010203 ) ); 
    
    A = _mm512_load_si512( state_in   );
    B = _mm512_load_si512( state_in+1 );
@@ -1489,8 +1273,8 @@ int sha256_16way_transform_le_short( __m512i *state_out, const __m512i *data,
    G = _mm512_load_si512( state_in+6 );
    H = _mm512_load_si512( state_in+7 );
 
-   const __m512i IV6 = G;
+   const __m512i istate6 = G;
-   const __m512i IV7 = H; 
+   const __m512i istate7 = H;
    
    // rounds 0 to 8
    SHA256_16WAY_ROUND( A, B, C, D, E, F, G, H,  0, 0 );
@@ -1562,7 +1346,7 @@ int sha256_16way_transform_le_short( __m512i *state_out, const __m512i *data,
    W[11] = SHA256_16WAY_MEXP( W[ 9], W[ 4], W[12], W[11] );
    W[12] = SHA256_16WAY_MEXP( W[10], W[ 5], W[13], W[12] );
    
-   // Rounds 48 to 57
+   // Rounds 48 to 55
    SHA256_16WAY_ROUND( A, B, C, D, E, F, G, H,  0, 48 );
    SHA256_16WAY_ROUND( H, A, B, C, D, E, F, G,  1, 48 );
    SHA256_16WAY_ROUND( G, H, A, B, C, D, E, F,  2, 48 );
@@ -1571,62 +1355,67 @@ int sha256_16way_transform_le_short( __m512i *state_out, const __m512i *data,
    SHA256_16WAY_ROUND( D, E, F, G, H, A, B, C,  5, 48 );
    SHA256_16WAY_ROUND( C, D, E, F, G, H, A, B,  6, 48 );
    SHA256_16WAY_ROUND( B, C, D, E, F, G, H, A,  7, 48 );
-   SHA256_16WAY_ROUND( A, B, C, D, E, F, G, H,  8, 48 );
-   SHA256_16WAY_ROUND( H, A, B, C, D, E, F, G,  9, 48 );
 
-   // rounds 58 to 60 part 1
+   // Round 56
-   T0 = _mm512_add_epi32( v512_32( K256[58] ),
+   H = _mm512_add_epi32( v512_32( K256[56] ),
+                 mm512_add4_32( BSG2_1x16( E ), CHx16( E, F, G ), W[ 8], H ) );
+   D = _mm512_add_epi32( D, H );
+   H56 = _mm512_add_epi32( H, _mm512_add_epi32( BSG2_0x16( A ),
+                                                   MAJx16( A, B, C ) ) );
+   
+   // Rounds 57 to 60 part 1
+   G = _mm512_add_epi32( v512_32( K256[57] ),
+                 mm512_add4_32( BSG2_1x16( D ), CHx16( D, E, F ), W[ 9], G ) );
+   C = _mm512_add_epi32( C, G );
+   G57 = _mm512_add_epi32( G, MAJx16( H56, A, B ) );
+   
+   F = _mm512_add_epi32( v512_32( K256[58] ),
                  mm512_add4_32( BSG2_1x16( C ), CHx16( C, D, E ), W[10], F ) );
-   B = _mm512_add_epi32( B, T0 );
+   B = _mm512_add_epi32( B, F );
    
-   T1 = _mm512_add_epi32( v512_32( K256[59] ),
+   E = _mm512_add_epi32( v512_32( K256[59] ),
                  mm512_add4_32( BSG2_1x16( B ), CHx16( B, C, D ), W[11], E ) );
-   A = _mm512_add_epi32( A, T1 );
+   A = _mm512_add_epi32( A, E );
 
-   T2 = _mm512_add_epi32( v512_32( K256[60] ),
+   D = _mm512_add_epi32( v512_32( K256[60] ),
                  mm512_add4_32( BSG2_1x16( A ), CHx16( A, B, C ), W[12], D ) );
-   H = _mm512_add_epi32( H, T2 );
+   H = _mm512_add_epi32( H56, D );
 
-   // got H, test it against target[7]
+   // got final H, test it against target[7]
-   hash = _mm512_shuffle_epi8( _mm512_add_epi32( H , IV7 ), bswap_shuf );
+   hash = mm512_bswap_32( _mm512_add_epi32( H , istate7 ) );
    targ = v512_32( target[7] );
-   if ( target[7] )
+   if ( likely( 0 == ( mask = _mm512_cmple_epu32_mask( hash, targ ) ) ))
-   if ( likely( 0 == _mm512_cmple_epu32_mask( hash, targ ) ))
       return 0;
-   t6_mask = _mm512_cmpeq_epi32_mask( hash, targ );
 
-   // round 58 part 2
+   // Round 57 part 2
-   F = _mm512_add_epi32( T0, _mm512_add_epi32( BSG2_0x16( G ),
+   G57 = _mm512_add_epi32( G57, BSG2_0x16( H56 ) );
-                                               MAJx16( G, H, A ) ) );
    
-   // round 61 part 1
+   // Round 61 part 1
    W[13] = SHA256_16WAY_MEXP( W[11], W[ 6], W[14], W[13] );
-   T0 = _mm512_add_epi32( v512_32( K256[61] ),
+   C = _mm512_add_epi32( v512_32( K256[61] ),
                  mm512_add4_32( BSG2_1x16( H ), CHx16( H, A, B ), W[13], C ) );
-   G = _mm512_add_epi32( G, T0 );
+   G = _mm512_add_epi32( G57, C );
 
-   // got G, test it against target[6] if indicated
+   // got final G, test it against target[6] if indicated.
-   if ( (uint16_t)t6_mask )
+   if ( mask == _mm512_cmpeq_epi32_mask( hash, targ ) )
    {
-      hash = _mm512_shuffle_epi8( _mm512_add_epi32( G, IV6 ), bswap_shuf );
+      hash = mm512_bswap_32( _mm512_add_epi32( G, istate6 ) );
       targ = v512_32( target[6] );
-      if ( likely( 0 == _mm512_mask_cmple_epu32_mask( t6_mask, hash, targ ) ))
+      if ( likely( 0 == _mm512_mask_cmple_epu32_mask( mask, hash, targ ) ))
           return 0;
    }
 
-   // round 59 part 2
+   // Round 58 to 61 part 2
-   E = _mm512_add_epi32( T1, _mm512_add_epi32( BSG2_0x16( F ),
+   F = _mm512_add_epi32( F, _mm512_add_epi32( BSG2_0x16( G57 ),
-                                                  MAJx16( F, G, H ) ) );
+                                                 MAJx16( G57, H, A ) ) );
-   
+   E = _mm512_add_epi32( E, _mm512_add_epi32( BSG2_0x16( F ),
-   // round 60 part 2
+                                                 MAJx16( F, G57, H ) ) );
-   D = _mm512_add_epi32( T2, _mm512_add_epi32( BSG2_0x16( E ),
+   D = _mm512_add_epi32( D, _mm512_add_epi32( BSG2_0x16( E ),
-                                                  MAJx16( E, F, G ) ) );
+                                                 MAJx16( E, F, G57 ) ) );
-   
+   C = _mm512_add_epi32( C, _mm512_add_epi32( BSG2_0x16( D ),
-   // round 61 part 2
-   C = _mm512_add_epi32( T0, _mm512_add_epi32( BSG2_0x16( D ),
                                                  MAJx16( D, E, F ) ) );
 
-   // rounds 62, 63
+   // Rounds 62, 63
    W[14] = SHA256_16WAY_MEXP( W[12], W[ 7], W[15], W[14] );
    W[15] = SHA256_16WAY_MEXP( W[13], W[ 8], W[ 0], W[15] );
    
@@ -1644,7 +1433,7 @@ int sha256_16way_transform_le_short( __m512i *state_out, const __m512i *data,
    return 1;
 }
 
-void sha256_16way_init( sha256_16way_context *sc )
+void sha256_16x32_init( sha256_16x32_context *sc )
 {
    sc->count_high = sc->count_low = 0;
    sc->val[0] = v512_32( sha256_iv[0] );
@@ -1657,7 +1446,7 @@ void sha256_16way_init( sha256_16way_context *sc )
    sc->val[7] = v512_32( sha256_iv[7] );
 }
 
-void sha256_16way_update( sha256_16way_context *sc, const void *data,
+void sha256_16x32_update( sha256_16x32_context *sc, const void *data,
                            size_t len )
 {
    __m512i *vdata = (__m512i*)data;
@@ -1679,7 +1468,7 @@ void sha256_16way_update( sha256_16way_context *sc, const void *data,
       len -= clen;
       if ( ptr == buf_size )
       {
-         sha256_16way_transform_be( sc->val, sc->buf, sc->val );
+         sha256_16x32_transform_be( sc->val, sc->buf, sc->val );
          ptr = 0;
       }
       clow = sc->count_low;
@@ -1690,7 +1479,7 @@ void sha256_16way_update( sha256_16way_context *sc, const void *data,
    }
 }
 
-void sha256_16way_close( sha256_16way_context *sc, void *dst )
+void sha256_16x32_close( sha256_16x32_context *sc, void *dst )
 {
     unsigned ptr;
     uint32_t low, high;
@@ -1704,7 +1493,7 @@ void sha256_16way_close( sha256_16way_context *sc, void *dst )
     if ( ptr > pad )
     {
          memset_zero_512( sc->buf + (ptr>>2), (buf_size - ptr) >> 2 );
-         sha256_16way_transform_be( sc->val, sc->buf, sc->val );
+         sha256_16x32_transform_be( sc->val, sc->buf, sc->val );
          memset_zero_512( sc->buf, pad >> 2 );
     }
     else
@@ -1717,17 +1506,17 @@ void sha256_16way_close( sha256_16way_context *sc, void *dst )
     sc->buf[   pad     >> 2 ] = v512_32( bswap_32( high ) );
     sc->buf[ ( pad+4 ) >> 2 ] = v512_32( bswap_32( low ) );
 
-    sha256_16way_transform_be( sc->val, sc->buf, sc->val );
+    sha256_16x32_transform_be( sc->val, sc->buf, sc->val );
 
     mm512_block_bswap_32( dst, sc->val );
 }
 
-void sha256_16way_full( void *dst, const void *data, size_t len )
+void sha256_16x32_full( void *dst, const void *data, size_t len )
 {
-   sha256_16way_context ctx;
+   sha256_16x32_context ctx;
-   sha256_16way_init( &ctx );
+   sha256_16x32_init( &ctx );
-   sha256_16way_update( &ctx, data, len );
+   sha256_16x32_update( &ctx, data, len );
-   sha256_16way_close( &ctx, dst );
+   sha256_16x32_close( &ctx, dst );
 }
 
 #undef CH

algo/sha/sha256-hash.c

@@ -1,6 +1,6 @@
 #include "sha256-hash.h"
 
-#if ( defined(__x86_64__) && defined(__SHA__) ) || defined(__ARM_NEON) && defined(__ARM_FEATURE_SHA2)
+#if ( defined(__x86_64__) && defined(__SHA__) ) || ( defined(__ARM_NEON) && defined(__ARM_FEATURE_SHA2) )
 
 static const uint32_t SHA256_IV[8] =
 {
@@ -10,6 +10,28 @@ static const uint32_t SHA256_IV[8] =
 
 #if defined(__x86_64__) && defined(__SHA__)
 
+
+/* common code used for rounds 12 through 51  */
+
+#define sha256_generic_qround( s0, s1, m, a, b, c ) \
+    TMP = _mm_alignr_epi8( a, c, 4 ); \
+    s1 = _mm_sha256rnds2_epu32( s1, s0, m ); \
+    b  = _mm_add_epi32( b, TMP ); \
+    b  = _mm_sha256msg2_epu32( b, a ); \
+    m  = _mm_shuffle_epi32( m, 0x0e ); \
+    s0 = _mm_sha256rnds2_epu32( s0, s1, m ); \
+    c  = _mm_sha256msg1_epu32( c, a );
+
+// r12-15      
+// sha256_generic_qround( s0, s1, m, t3, t0, t2 )
+// r16-19
+// sha256_generic_qround( s0, s1, m, t0, t1, t3 )
+// r20-23
+// sha256_generic_qround( s0, s1, m, t1, t2, t0 ) 
+// r24-27
+// sha256_generic_qround( s0, s1, m, t2, t3, t1 ) ... 
+
+
 #define sha256_opt_rounds( state_out, input, state_in ) \
 { \
     __m128i STATE0, STATE1; \
@@ -189,7 +211,7 @@ static const uint32_t SHA256_IV[8] =
     _mm_store_si128( (__m128i*) &state_out[4], STATE1 ); \
 }
 
-void sha256_opt_transform_le( uint32_t *state_out, const void *input,
+void sha256_x86_sha_transform_le( uint32_t *state_out, const void *input,
                               const uint32_t *state_in )
 {
 #define load_msg( m, i ) casti_v128( m, i )
@@ -197,7 +219,7 @@ void sha256_opt_transform_le( uint32_t *state_out, const void *input,
 #undef load_msg
 }
 
-void sha256_opt_transform_be( uint32_t *state_out, const void *input,
+void sha256_x86_sha_transform_be( uint32_t *state_out, const void *input,
                               const uint32_t *state_in )
 {
 #define load_msg( m, i ) v128_bswap32( casti_v128( m, i ) )
@@ -517,7 +539,7 @@ void sha256_opt_transform_be( uint32_t *state_out, const void *input,
     _mm_store_si128( (__m128i*) &out_Y[4], STATE1_Y ); \
 }
 
-void sha256_ni2x_transform_le( uint32_t *out_X, uint32_t*out_Y,
+void sha256_x86_x2sha_transform_le( uint32_t *out_X, uint32_t*out_Y,
                                  const void *msg_X, const void *msg_Y,
                                  const uint32_t *in_X, const uint32_t *in_Y )
 {
@@ -526,7 +548,7 @@ void sha256_ni2x_transform_le( uint32_t *out_X, uint32_t*out_Y,
 #undef load_msg
 }
 
-void sha256_ni2x_transform_be( uint32_t *out_X, uint32_t*out_Y,
+void sha256_x86_x2sha_transform_be( uint32_t *out_X, uint32_t*out_Y,
                               const void *msg_X, const void *msg_Y,
                               const uint32_t *in_X, const uint32_t *in_Y )
 {
@@ -541,14 +563,14 @@ void sha256_ni2x_transform_be( uint32_t *out_X, uint32_t*out_Y,
 // The goal is to avoid any redundant processing in final. Prehash is almost
 // 4 rounds total, only missing the final addition of the nonce.
 // Nonce must be set to zero for prehash.
-void sha256_ni_prehash_3rounds( uint32_t *ostate, const void *msg,
+void sha256_x86_sha_prehash_3rounds( uint32_t *ostate, const void *msg,
                                 uint32_t *sstate, const uint32_t *istate )
 {
    __m128i STATE0, STATE1, MSG, TMP;
 
    // Load initial values
-   TMP    = casti_m128i( istate, 0 );
+   TMP    = casti_v128u32( istate, 0 );
-   STATE1 = casti_m128i( istate, 1 );
+   STATE1 = casti_v128u32( istate, 1 );
 
    TMP    = _mm_shuffle_epi32( TMP, 0xB1 );       // CDAB
    STATE1 = _mm_shuffle_epi32( STATE1, 0x1B );    // EFGH
@@ -556,20 +578,20 @@ void sha256_ni_prehash_3rounds( uint32_t *ostate, const void *msg,
    STATE1 = _mm_blend_epi16( STATE1, TMP, 0xF0 ); // CDGH
 
    // Save current hash
-   casti_m128i( sstate, 0 ) = STATE0;
+   casti_v128u32( sstate, 0 ) = STATE0;
-   casti_m128i( sstate, 1 ) = STATE1;
+   casti_v128u32( sstate, 1 ) = STATE1;
 
    // Rounds 0 to 3
-   MSG = casti_m128i( msg, 0 );
+   MSG = casti_v128u32( msg, 0 );
    TMP = _mm_set_epi64x( 0xE9B5DBA5B5C0FBCFULL, 0x71374491428A2F98ULL );
    MSG = _mm_add_epi32( MSG, TMP );
    STATE1 = _mm_sha256rnds2_epu32( STATE1, STATE0, MSG );
    MSG = _mm_shuffle_epi32( MSG, 0x0E );
-   casti_m128i( ostate, 0 ) = _mm_sha256rnds2_epu32( STATE0, STATE1, MSG );
+   casti_v128u32( ostate, 0 ) = _mm_sha256rnds2_epu32( STATE0, STATE1, MSG );
-   casti_m128i( ostate, 1 ) = STATE1;
+   casti_v128u32( ostate, 1 ) = STATE1;
 }
 
-void sha256_ni2x_final_rounds( uint32_t *out_X, uint32_t *out_Y,
+void sha256_x86_x2sha_final_rounds( uint32_t *out_X, uint32_t *out_Y,
                  const void *msg_X, const void *msg_Y,
                  const uint32_t *state_mid_X, const uint32_t *state_mid_Y,
                  const uint32_t *state_save_X, const uint32_t *state_save_Y )
@@ -579,22 +601,22 @@ void sha256_ni2x_final_rounds( uint32_t *out_X, uint32_t *out_Y,
     __m128i TMSG0_X, TMSG1_X, TMSG2_X, TMSG3_X;
     __m128i TMSG0_Y, TMSG1_Y, TMSG2_Y, TMSG3_Y;
 
-    STATE0_X = casti_m128i( state_mid_X, 0 );
+    STATE0_X = casti_v128u32( state_mid_X, 0 );
-    STATE1_X = casti_m128i( state_mid_X, 1 );
+    STATE1_X = casti_v128u32( state_mid_X, 1 );
-    STATE0_Y = casti_m128i( state_mid_Y, 0 );
+    STATE0_Y = casti_v128u32( state_mid_Y, 0 );
-    STATE1_Y = casti_m128i( state_mid_Y, 1 );
+    STATE1_Y = casti_v128u32( state_mid_Y, 1 );
 
     // Add the nonces (msg[0] lane 3) to A & E (STATE0 lanes 1 & 3)
-    TMSG0_X = casti_m128i( msg_X, 0 );
+    TMSG0_X = casti_v128u32( msg_X, 0 );
-    TMSG0_Y = casti_m128i( msg_Y, 0 );
+    TMSG0_Y = casti_v128u32( msg_Y, 0 );
     TMP_X = v128_xim32( TMSG0_X, TMSG0_X, 0xd5 );
     TMP_Y = v128_xim32( TMSG0_Y, TMSG0_Y, 0xd5 );
     STATE0_X = _mm_add_epi32( STATE0_X, TMP_X );
     STATE0_Y = _mm_add_epi32( STATE0_Y, TMP_Y );
 
     // Rounds 4 to 7
-    TMSG1_X = casti_m128i( msg_X, 1 );
+    TMSG1_X = casti_v128u32( msg_X, 1 );
-    TMSG1_Y = casti_m128i( msg_Y, 1 );
+    TMSG1_Y = casti_v128u32( msg_Y, 1 );
     TMP_X = _mm_set_epi64x( 0xAB1C5ED5923F82A4ULL, 0x59F111F13956C25BULL );
     MSG_X = _mm_add_epi32( TMSG1_X, TMP_X );
     MSG_Y = _mm_add_epi32( TMSG1_Y, TMP_X );
@@ -616,8 +638,8 @@ void sha256_ni2x_final_rounds( uint32_t *out_X, uint32_t *out_Y,
     STATE0_Y = _mm_sha256rnds2_epu32( STATE0_Y, STATE1_Y, MSG_X );
 
     // Rounds 12 to 15
-    TMSG3_X = casti_m128i( msg_X, 3 );
+    TMSG3_X = casti_v128u32( msg_X, 3 );
-    TMSG3_Y = casti_m128i( msg_Y, 3 );
+    TMSG3_Y = casti_v128u32( msg_Y, 3 );
     TMP_X = _mm_set_epi64x( 0xC19BF1749BDC06A7ULL, 0x80DEB1FE72BE5D74ULL );
     MSG_X = _mm_add_epi32( TMSG3_X, TMP_X );
     MSG_Y = _mm_add_epi32( TMSG3_Y, TMP_X );
@@ -845,20 +867,20 @@ void sha256_ni2x_final_rounds( uint32_t *out_X, uint32_t *out_Y,
     STATE0_Y = _mm_sha256rnds2_epu32( STATE0_Y, STATE1_Y, MSG_Y );
 
     // Add saved state to new state
-    STATE0_X = _mm_add_epi32( STATE0_X, casti_m128i( state_save_X, 0 ) );
+    STATE0_X = _mm_add_epi32( STATE0_X, casti_v128u32( state_save_X, 0 ) );
-    STATE1_X = _mm_add_epi32( STATE1_X, casti_m128i( state_save_X, 1 ) );
+    STATE1_X = _mm_add_epi32( STATE1_X, casti_v128u32( state_save_X, 1 ) );
-    STATE0_Y = _mm_add_epi32( STATE0_Y, casti_m128i( state_save_Y, 0 ) );
+    STATE0_Y = _mm_add_epi32( STATE0_Y, casti_v128u32( state_save_Y, 0 ) );
-    STATE1_Y = _mm_add_epi32( STATE1_Y, casti_m128i( state_save_Y, 1 ) );
+    STATE1_Y = _mm_add_epi32( STATE1_Y, casti_v128u32( state_save_Y, 1 ) );
 
     // Unshuffle & save state    
     TMP_X = _mm_shuffle_epi32( STATE0_X, 0x1B );                        // FEBA
     TMP_Y = _mm_shuffle_epi32( STATE0_Y, 0x1B );
     STATE1_X = _mm_shuffle_epi32( STATE1_X, 0xB1 );                     // DCHG
     STATE1_Y = _mm_shuffle_epi32( STATE1_Y, 0xB1 );
-    casti_m128i( out_X, 0 ) = _mm_blend_epi16( TMP_X, STATE1_X, 0xF0 ); // DCBA
+    casti_v128u32( out_X, 0 ) = _mm_blend_epi16( TMP_X, STATE1_X, 0xF0 ); // DCBA
-    casti_m128i( out_Y, 0 ) = _mm_blend_epi16( TMP_Y, STATE1_Y, 0xF0 );
+    casti_v128u32( out_Y, 0 ) = _mm_blend_epi16( TMP_Y, STATE1_Y, 0xF0 );
-    casti_m128i( out_X, 1 ) = _mm_alignr_epi8( STATE1_X, TMP_X, 8 );    // ABEF
+    casti_v128u32( out_X, 1 ) = _mm_alignr_epi8( STATE1_X, TMP_X, 8 );    // ABEF
-    casti_m128i( out_Y, 1 ) = _mm_alignr_epi8( STATE1_Y, TMP_Y, 8 );
+    casti_v128u32( out_Y, 1 ) = _mm_alignr_epi8( STATE1_Y, TMP_Y, 8 );
 }
 
 #endif     // SHA
@@ -887,14 +909,14 @@ static const uint32_t K256[64] =
 
 #define sha256_neon_rounds( state_out, input, state_in ) \
 { \
-    uint32x4_t STATE0, STATE1, ABEF_SAVE, CDGH_SAVE; \
+    uint32x4_t STATE0, STATE1, ABCD_SAVE, EFGH_SAVE; \
     uint32x4_t MSG0, MSG1, MSG2, MSG3; \
     uint32x4_t TMP0, TMP1, TMP2; \
 \
     STATE0 = vld1q_u32( state_in   ); \
     STATE1 = vld1q_u32( state_in+4 ); \
-    ABEF_SAVE = STATE0; \
+    ABCD_SAVE = STATE0; \
-    CDGH_SAVE = STATE1; \
+    EFGH_SAVE = STATE1; \
 \
     MSG0 = load_msg( input, 0 ); \
     MSG1 = load_msg( input, 1 ); \
@@ -1004,8 +1026,8 @@ static const uint32_t K256[64] =
     TMP2 = STATE0; \
     STATE0 = vsha256hq_u32( STATE0, STATE1, TMP1 ); \
     STATE1 = vsha256h2q_u32( STATE1, TMP2, TMP1 ); \
-    STATE0 = vaddq_u32( STATE0, ABEF_SAVE ); \
+    STATE0 = vaddq_u32( STATE0, ABCD_SAVE ); \
-    STATE1 = vaddq_u32( STATE1, CDGH_SAVE ); \
+    STATE1 = vaddq_u32( STATE1, EFGH_SAVE ); \
     vst1q_u32( state_out  , STATE0 ); \
     vst1q_u32( state_out+4, STATE1 ); \
 }
@@ -1029,8 +1051,8 @@ void sha256_neon_sha_transform_le( uint32_t *state_out, const void *input,
 #define sha256_neon_x2sha_rounds( state_out_X, state_out_Y, input_X, \
                               input_Y, state_in_X, state_in_Y ) \
 { \
-    uint32x4_t STATE0_X, STATE1_X, ABEF_SAVE_X, CDGH_SAVE_X; \
+    uint32x4_t STATE0_X, STATE1_X, ABCD_SAVE_X, EFGH_SAVE_X; \
-    uint32x4_t STATE0_Y, STATE1_Y, ABEF_SAVE_Y, CDGH_SAVE_Y; \
+    uint32x4_t STATE0_Y, STATE1_Y, ABCD_SAVE_Y, EFGH_SAVE_Y; \
     uint32x4_t MSG0_X, MSG1_X, MSG2_X, MSG3_X; \
     uint32x4_t MSG0_Y, MSG1_Y, MSG2_Y, MSG3_Y; \
     uint32x4_t TMP0_X, TMP1_X, TMP2_X; \
@@ -1040,10 +1062,10 @@ void sha256_neon_sha_transform_le( uint32_t *state_out, const void *input,
     STATE0_Y = vld1q_u32( state_in_Y   ); \
     STATE1_X = vld1q_u32( state_in_X+4 ); \
     STATE1_Y = vld1q_u32( state_in_Y+4 ); \
-    ABEF_SAVE_X = STATE0_X; \
+    ABCD_SAVE_X = STATE0_X; \
-    ABEF_SAVE_Y = STATE0_Y; \
+    ABCD_SAVE_Y = STATE0_Y; \
-    CDGH_SAVE_X = STATE1_X; \
+    EFGH_SAVE_X = STATE1_X; \
-    CDGH_SAVE_Y = STATE1_Y; \
+    EFGH_SAVE_Y = STATE1_Y; \
 \
     MSG0_X = load_msg( input_X, 0 ); \
     MSG0_Y = load_msg( input_Y, 0 ); \
@@ -1245,10 +1267,10 @@ void sha256_neon_sha_transform_le( uint32_t *state_out, const void *input,
     STATE0_Y = vsha256hq_u32( STATE0_Y, STATE1_Y, TMP1_Y ); \
     STATE1_X = vsha256h2q_u32( STATE1_X, TMP2_X, TMP1_X ); \
     STATE1_Y = vsha256h2q_u32( STATE1_Y, TMP2_Y, TMP1_Y ); \
-    STATE0_X = vaddq_u32( STATE0_X, ABEF_SAVE_X ); \
+    STATE0_X = vaddq_u32( STATE0_X, ABCD_SAVE_X ); \
-    STATE0_Y = vaddq_u32( STATE0_Y, ABEF_SAVE_Y ); \
+    STATE0_Y = vaddq_u32( STATE0_Y, ABCD_SAVE_Y ); \
-    STATE1_X = vaddq_u32( STATE1_X, CDGH_SAVE_X ); \
+    STATE1_X = vaddq_u32( STATE1_X, EFGH_SAVE_X ); \
-    STATE1_Y = vaddq_u32( STATE1_Y, CDGH_SAVE_Y ); \
+    STATE1_Y = vaddq_u32( STATE1_Y, EFGH_SAVE_Y ); \
     vst1q_u32( state_out_X  , STATE0_X ); \
     vst1q_u32( state_out_Y  , STATE0_Y ); \
     vst1q_u32( state_out_X+4, STATE1_X ); \

algo/sha/sha256-hash.h

@@ -5,27 +5,21 @@
 #include "simd-utils.h"
 #include "cpuminer-config.h"
 
-// generic interface 
+static const uint32_t SHA256_IV[8];
+
+#if defined(__x86_64__) && defined(__SHA__)
 
 typedef struct
 {
-   unsigned char buf[64];    /* first field, for alignment */
+   unsigned char buf[64];
    uint32_t state[8];
    uint64_t count;
 } sha256_context __attribute__((aligned(64)));
 
-static const uint32_t SHA256_IV[8];
-
 void sha256_full( void *hash, const void *data, size_t len );
 void sha256_update( sha256_context *ctx, const void *data, size_t len );
 void sha256_final( sha256_context *ctx, void *hash );
 void sha256_ctx_init( sha256_context *ctx );
-void sha256_transform_le( uint32_t *state_out, const uint32_t *data,
-                          const uint32_t *state_in );
-void sha256_transform_be( uint32_t *state_out, const uint32_t *data,
-                          const uint32_t *state_in );
-
-#if defined(__x86_64__) && defined(__SHA__)
 
 void sha256_x86_sha_transform_le( uint32_t *state_out, const void *input,
                                   const uint32_t *state_in );
@@ -50,14 +44,6 @@ void sha256_x86_x2sha_final_rounds( uint32_t *state_out_X, uint32_t *state_out_Y
                  const uint32_t *state_mid_X, const uint32_t *state_mid_Y,
                  const uint32_t *state_save_X, const uint32_t *state_save_Y );
 
-// Temporary during name transition
-#define sha256_opt_transform_le   sha256_x86_sha_transform_le
-#define sha256_opt_transform_be   sha256_x86_sha_transform_be
-#define sha256_ni2x_transform_le  sha256_x86_x2sha_transform_le
-#define sha256_ni2x_transform_be  sha256_x86_x2sha_transform_be
-#define sha256_ni_prehash_3rounds sha256_x86_sha_prehash_3rounds
-#define sha256_ni2x_final_rounds  sha256_x86_x2sha_final_rounds
-
 // generic API
 #define sha256_transform_le        sha256_x86_sha_transform_le
 #define sha256_transform_be        sha256_x86_sha_transform_be
@@ -68,6 +54,20 @@ void sha256_x86_x2sha_final_rounds( uint32_t *state_out_X, uint32_t *state_out_Y
 
 #elif defined(__ARM_NEON) && defined(__ARM_FEATURE_SHA2)
 
+// SHA-256 AArch64 with NEON & SHA2
+
+typedef struct
+{
+   unsigned char buf[64];
+   uint32_t state[8];
+   uint64_t count;
+} sha256_context __attribute__((aligned(64)));
+
+void sha256_full( void *hash, const void *data, size_t len );
+void sha256_update( sha256_context *ctx, const void *data, size_t len );
+void sha256_final( sha256_context *ctx, void *hash );
+void sha256_ctx_init( sha256_context *ctx );
+
 void sha256_neon_sha_transform_be( uint32_t *state_out, const void *input,
                                    const uint32_t *state_in );
 void sha256_neon_sha_transform_le( uint32_t *state_out, const void *input,
@@ -89,14 +89,6 @@ void sha256_neon_x2sha_final_rounds( uint32_t *state_out_X,
                  const uint32_t *state_mid_X, const uint32_t *state_mid_Y,
                  const uint32_t *state_save_X, const uint32_t *state_save_Y );
 
-// Temporary during name transition
-#define sha256_transform_le         sha256_neon_sha_transform_le
-#define sha256_transform_be         sha256_neon_sha_transform_be
-#define sha256_2x_transform_le      sha256_neon_x2sha_transform_le
-#define sha256_2x_transform_be      sha256_neon_x2sha_transform_be
-#define sha256_prehash_3rounds      sha256_neon_sha_prehash_3rounds
-#define sha256_2x_final_rounds      sha256_neon_x2sha_final_rounds
-
 // generic API
 #define sha256_transform_le        sha256_neon_sha_transform_le
 #define sha256_transform_be        sha256_neon_sha_transform_be
@@ -106,9 +98,11 @@ void sha256_neon_x2sha_final_rounds( uint32_t *state_out_X,
 #define sha256_2x_final_rounds     sha256_neon_x2sha_final_rounds
 
 #else
+
 // without HW acceleration...
 #include "sph_sha2.h"
 
+#define sha256_context              sph_sha256_context
 #define sha256_full                 sph_sha256_full
 #define sha256_ctx_init             sph_sha256_init
 #define sha256_update               sph_sha256
@@ -117,12 +111,11 @@ void sha256_neon_x2sha_final_rounds( uint32_t *state_out_X,
 #define sha256_transform_be         sph_sha256_transform_be
 #define sha256_prehash_3rounds      sph_sha256_prehash_3rounds
 
-
 #endif
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
 
-// SHA-256 16 way
+// SHA-256 16 way x86_64
 
 typedef struct
 {
@@ -162,7 +155,7 @@ int sha256_16x32_transform_le_short( __m512i *state_out, const __m512i *data,
 
 #if defined (__AVX2__)
 
-// SHA-256 8 way
+// SHA-256 8 way x86_64
 
 typedef struct
 {
@@ -187,21 +180,10 @@ void sha256_8x32_final_rounds( __m256i *state_out, const __m256i *data,
 int sha256_8x32_transform_le_short( __m256i *state_out, const __m256i *data,
                              const __m256i *state_in, const uint32_t *target );
 
-// Temporary API during naming transition
-#define sha256_8way_context               sha256_8x32_context
-#define sha256_8way_init                  sha256_8x32_init
-#define sha256_8way_update                sha256_8x32_update
-#define sha256_8way_close                 sha256_8x32_close
-#define sha256_8way_full                  sha256_8x32_full
-#define sha256_8way_transform_le          sha256_8x32_transform_le
-#define sha256_8way_transform_be          sha256_8x32_transform_be
-#define sha256_8way_prehash_3rounds       sha256_8x32_prehash_3rounds
-#define sha256_8way_final_rounds          sha256_8x32_final_rounds
-#define sha256_8way_transform_le_short    sha256_8x32_transform_le_short
-
 #endif  // AVX2
 
-// SHA-256 4 way
+#if defined(__SSE2__) || defined(__ARM_NEON)
+// SHA-256 4 way x86_64 with SSE2 or AArch64 with NEON
 
 typedef struct
 {
@@ -226,16 +208,5 @@ void sha256_4x32_final_rounds( v128_t *state_out, const v128_t *data,
 int sha256_4x32_transform_le_short( v128_t *state_out, const v128_t *data,
                             const v128_t *state_in, const uint32_t *target );
 
-// Temporary API during naming transition
+#endif // SSE2 || NEON
-#define sha256_4way_context              sha256_4x32_context
+#endif // SHA256_HASH_H__
-#define sha256_4way_init                 sha256_4x32_init
-#define sha256_4way_update               sha256_4x32_update
-#define sha256_4way_close                sha256_4x32_close
-#define sha256_4way_full                 sha256_4x32_full
-#define sha256_4way_transform_le         sha256_4x32_transform_le
-#define sha256_4way_transform_be         sha256_4x32_transform_be
-#define sha256_4way_prehash_3rounds      sha256_4x32_prehash_3rounds
-#define sha256_4way_final_rounds         sha256_4x32_final_rounds
-#define sha256_4way_transform_le_short   sha256_4x32_transform_le_short
-
-#endif

algo/sha/sha256d-4way.c

@@ -32,8 +32,6 @@ int scanhash_sha256d_sha( struct work *work, uint32_t max_nonce,
    uint32_t n = first_nonce;
    const int thr_id = mythr->id;
    const bool bench = opt_benchmark;
-   const v128_t shuf_bswap32 =
-           v128_set64( 0x0c0d0e0f08090a0bULL, 0x0405060700010203ULL );
 
    // hash first 64 byte block of data
    sha256_transform_le( mstatea, pdata, sha256_iv );
@@ -69,10 +67,8 @@ int scanhash_sha256d_sha( struct work *work, uint32_t max_nonce,
 
       if ( unlikely( bswap_32( hasha[7] ) <= ptarget[7] ) )
       {
-          casti_v128( hasha, 0 ) =
+          casti_v128( hasha, 0 ) = v128_bswap32( casti_v128( hasha, 0 ) );
-               _mm_shuffle_epi8( casti_v128( hasha, 0 ), shuf_bswap32 );
+          casti_v128( hasha, 1 ) = v128_bswap32( casti_v128( hasha, 1 ) );
-          casti_v128( hasha, 1 ) =
-               _mm_shuffle_epi8( casti_v128( hasha, 1 ), shuf_bswap32 );
           if ( likely( valid_hash( hasha, ptarget ) && !bench ) )
           {
              pdata[19] = n;
@@ -81,10 +77,8 @@ int scanhash_sha256d_sha( struct work *work, uint32_t max_nonce,
       }
       if ( unlikely( bswap_32( hashb[7] ) <= ptarget[7] ) )
       {
-         casti_v128( hashb, 0 ) =
+         casti_v128( hashb, 0 ) = v128_bswap32( casti_v128( hashb, 0 ) );
-               _mm_shuffle_epi8( casti_v128( hashb, 0 ), shuf_bswap32 );
+         casti_v128( hashb, 1 ) = v128_bswap32( casti_v128( hashb, 1 ) );
-         casti_v128( hashb, 1 ) =
-               _mm_shuffle_epi8( casti_v128( hashb, 1 ), shuf_bswap32 );
          if ( likely( valid_hash( hashb, ptarget ) && !bench ) )
          {
             pdata[19] = n+1;
@@ -204,8 +198,6 @@ int scanhash_sha256d_16way( struct work *work, const uint32_t max_nonce,
    const int thr_id = mythr->id;
    const __m512i sixteen = v512_32( 16 );
    const bool bench = opt_benchmark;
-   const __m256i bswap_shuf = mm256_bcast_m128( _mm_set_epi64x(
-                                0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
 
    // prehash first block directly from pdata
    sha256_transform_le( phash, pdata, sha256_iv );
@@ -231,7 +223,7 @@ int scanhash_sha256d_16way( struct work *work, const uint32_t max_nonce,
    buf[15] = v512_32( 80*8 );  // bit count
 
    // partially pre-expand & prehash second message block, avoiding the nonces
-   sha256_16way_prehash_3rounds( mstate2, mexp_pre, buf, mstate1 );
+   sha256_16x32_prehash_3rounds( mstate2, mexp_pre, buf, mstate1 );
 
    // vectorize IV for second hash
    istate[0] = v512_32( sha256_iv[0] );
@@ -250,15 +242,14 @@ int scanhash_sha256d_16way( struct work *work, const uint32_t max_nonce,
 
    do
    {
-      sha256_16way_final_rounds( block, buf, mstate1, mstate2, mexp_pre );
+      sha256_16x32_final_rounds( block, buf, mstate1, mstate2, mexp_pre );
-      if ( unlikely( sha256_16way_transform_le_short(
+      if ( unlikely( sha256_16x32_transform_le_short(
                                   hash32, block, istate, ptarget ) ) )
       {
          for ( int lane = 0; lane < 16; lane++ )
          {
             extr_lane_16x32( phash, hash32, lane, 256 );
-            casti_m256i( phash, 0 ) =
+            casti_m256i( phash, 0 ) = mm256_bswap_32( casti_m256i( phash, 0 ) ); 
-                   _mm256_shuffle_epi8( casti_m256i( phash, 0 ), bswap_shuf ); 
             if ( likely( valid_hash( phash, ptarget ) && !bench ) )
             {
               pdata[19] = n + lane;
@@ -299,8 +290,6 @@ int scanhash_sha256d_8way( struct work *work, const uint32_t max_nonce,
    const bool bench = opt_benchmark;
    const __m256i last_byte = v256_32( 0x80000000 );
    const __m256i eight = v256_32( 8 );
-   const __m256i bswap_shuf = mm256_bcast_m128( _mm_set_epi64x(
-                                0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
 
    for ( int i = 0; i < 19; i++ )
       vdata[i] = v256_32( pdata[i] );
@@ -325,22 +314,22 @@ int scanhash_sha256d_8way( struct work *work, const uint32_t max_nonce,
    istate[6] = v256_32( sha256_iv[6] );
    istate[7] = v256_32( sha256_iv[7] );
 
-   sha256_8way_transform_le( mstate1, vdata, istate );
+   sha256_8x32_transform_le( mstate1, vdata, istate );
 
    // Do 3 rounds on the first 12 bytes of the next block
-   sha256_8way_prehash_3rounds( mstate2, mexp_pre, vdata + 16, mstate1 );
+   sha256_8x32_prehash_3rounds( mstate2, mexp_pre, vdata + 16, mstate1 );
    
    do
    {
-      sha256_8way_final_rounds( block, vdata+16, mstate1, mstate2, mexp_pre );
+      sha256_8x32_final_rounds( block, vdata+16, mstate1, mstate2, mexp_pre );
-      if ( unlikely( sha256_8way_transform_le_short( hash32, block,
+      if ( unlikely( sha256_8x32_transform_le_short( hash32, block,
                                                      istate, ptarget ) ) )
       {
          for ( int lane = 0; lane < 8; lane++ )
          {
             extr_lane_8x32( lane_hash, hash32, lane, 256 );
             casti_m256i( lane_hash, 0 ) =
-               _mm256_shuffle_epi8( casti_m256i( lane_hash, 0 ), bswap_shuf );
+                                mm256_bswap_32( casti_m256i( lane_hash, 0 ) );
             if ( likely( valid_hash( lane_hash, ptarget ) && !bench ) )
             {
                pdata[19] = n + lane;

algo/sha/sha256d-4way.h

@@ -4,7 +4,7 @@
 #include <stdint.h>
 #include "algo-gate-api.h"
 
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(SIMD512)
   #define SHA256D_16WAY 1
 #elif defined(__SHA__)
   #define SHA256D_SHA 1