v3.11.7

v3.11.6
v3.11.5
2025-09-17 23:44:27 +00:00 · 2020-01-26 04:33:39 -05:00 · 2020-01-23 00:11:08 -05:00 · 2020-01-18 15:14:27 -05:00 · 2020-01-08 14:44:47 -05:00
160 changed files with 4928 additions and 4761 deletions
--- a/3
+++ b/3
@@ -33,3 +33,6 @@ Jay D Dee
 xcouiz@gmail.com

 Cryply
+
+Colin Percival
+Alexander Peslyak
--- a/Makefile.am
+++ b/Makefile.am
@@ -80,7 +80,6 @@ cpuminer_SOURCES = \
  algo/cryptonight/cryptonight-common.c\
  algo/cryptonight/cryptonight-aesni.c\
  algo/cryptonight/cryptonight.c\
-  algo/cubehash/sph_cubehash.c \
  algo/cubehash/cubehash_sse2.c\
  algo/cubehash/cube-hash-2way.c \
  algo/echo/sph_echo.c \
@@ -89,6 +88,7 @@ cpuminer_SOURCES = \
  algo/gost/sph_gost.c \
  algo/groestl/groestl-gate.c \
  algo/groestl/groestl512-hash-4way.c \
+  algo/groestl/groestl256-hash-4way.c \
  algo/groestl/sph_groestl.c \
  algo/groestl/groestl.c \
  algo/groestl/groestl-4way.c \
@@ -120,6 +120,8 @@ cpuminer_SOURCES = \
  algo/keccak/keccak-hash-4way.c \
  algo/keccak/keccak-4way.c\
  algo/keccak/keccak-gate.c \
+  algo/keccak/sha3d-4way.c \
+  algo/keccak/sha3d.c \
  algo/lanehash/lane.c \
  algo/luffa/sph_luffa.c \
  algo/luffa/luffa.c \
@@ -179,6 +181,7 @@ cpuminer_SOURCES = \
  algo/sha/sph_sha2big.c \
  algo/sha/sha256-hash-4way.c \
  algo/sha/sha512-hash-4way.c \
+  algo/sha/hmac-sha256-hash.c \
  algo/sha/sha2.c \
  algo/sha/sha256t-gate.c \
  algo/sha/sha256t-4way.c \
@@ -291,12 +294,11 @@ cpuminer_SOURCES = \
  algo/x22/x25x.c \
  algo/x22/x25x-4way.c \
  algo/yescrypt/yescrypt.c \
-  algo/yescrypt/sha256_Y.c \
  algo/yescrypt/yescrypt-best.c \
  algo/yespower/yespower-gate.c \
  algo/yespower/yespower-blake2b.c \
  algo/yespower/crypto/blake2b-yp.c \
-  algo/yespower/sha256_p.c \
+  algo/yespower/yescrypt-r8g.c \
  algo/yespower/yespower-opt.c

 disable_flags =
--- a/README.md
+++ b/README.md
@@ -97,10 +97,10 @@ Supported Algorithms
                          qubit         Qubit
                          scrypt        scrypt(1024, 1, 1) (default)
                          scrypt:N      scrypt(N, 1, 1)
-                          scryptjane:nf
                          sha256d       Double SHA-256
                          sha256q       Quad SHA-256, Pyrite (PYE)
                          sha256t       Triple SHA-256, Onecoin (OC)
+                          sha3d         Double keccak256 (BSHA3)
                          shavite3      Shavite3
                          skein         Skein+Sha (Skeincoin)
                          skein2        Double Skein (Woodcoin)
@@ -134,6 +134,7 @@ Supported Algorithms
                          xevan         Bitsend (BSD)
                          yescrypt      Globalboost-Y (BSTY)
                          yescryptr8    BitZeny (ZNY)
+                          yescryptr8g   Koto (KOTO)
                          yescryptr16   Eli
                          yescryptr32   WAVI
                          yespower      Cryply
--- a/109
+++ b/109
@@ -8,9 +8,10 @@ Security warning

 Miner programs are often flagged as malware by antivirus programs. This is
 usually a false positive, they are flagged simply because they are
-cryptocurrency miners. However, some malware has been spread using the
-cover that miners are known to be subject to false positives. Always be on
-alert. The source code of cpuminer-opt is open for anyone to inspect.
+cryptocurrency miners. However, some malware masquerading as a miner has
+been spread using the cover that miners are known to be subject to false
+positives ans users will dismiss the AV alert. Always be on alert.
+The source code of cpuminer-opt is open for anyone to inspect.
 If you don't trust the software don't download it.

 The cryptographic hashing code has been taken from trusted sources but has been
@@ -29,12 +30,110 @@ Requirements
 Intel Core2 or newer, or AMD Steamroller or newer CPU. ARM CPUs are not
 supported.

-64 bit Linux or Windows operating system. Apple, Android and Rpi are
-not supported. FreeBSD YMMV.
+64 bit Linux or Windows operating system. Apple, Android and Raspberry Pi
+are not supported. FreeBSD YMMV.
+
+Reporting bugs
+--------------
+
+Bugs can be reported by sending am email to JayDDee246@gmail.com or opening
+an issue in git: https://github.com/JayDDee/cpuminer-opt/issues
+
+Please include the following information:
+
+1. CPU model, operating system, cpuminer-opt version (must be latest),
+   binary file for Windows, changes to default build procedure for Linux.
+
+2. Exact comand line (except user and pw) and intial output showing
+   the above requested info.
+
+3. Additional program output showing any error messages or other
+   pertinent data.
+
+4. A clear description of the problem including history, scope,
+   persistence or intermittance, and reproduceability. 
+
+In simpler terms:
+
+What is it doing?
+What should it be doing instead?
+Did it work in a previous release?
+Does it happen for all algos? All pools? All options? Solo?
+Does it happen all the time?
+If not what makes it happen or not happen? 

 Change Log
 ----------

+v3.11.7
+
+Added yescryptr8g algo fotr KOTO, including support for block version 5.
+
+Added sha3d algo for BSHA3.
+
+Removed memcmp and clean_job checks from get_new_work, now only check job_id.
+
+Small improvement to sha512 and sha256 parallel implementations that don't
+use SHA.
+
+v3.11.6
+
+Fixed CPU temperature regression from v3.11.5.
+
+More improvements to share log. More compact, highlight incremented counter,
+block height when solved, job id when stale.
+
+v3.11.5
+
+Fixed AVX512 detection that could cause compilation errors on CPUs
+without AVX512.
+
+Fixed "BLOCK SOLVED" log incorrectly displaying "Accepted" when a block
+is solved.
+Added share counter to share submitited & accepted logs
+Added job id to share submitted log.
+Share submitted log is no longer highlighted blue, there was too much blue.
+
+Another CPU temperature fix for Linux.
+
+Added bug reporting tips to RELEASE NOTES.
+
+v3.11.4
+
+Fixed scrypt segfault since v3.9.9.1.
+
+Stale shares counted and reported seperately from other rejected shares.
+
+Display of counters for solved blocks, rejects, stale shares suppressed in
+periodic summary when zero.
+
+v3.11.3
+
+Fixed x12 AVX2 again.
+
+More speed for allium: AVX2 +4%, AVX512 +6%, VAES +14%.
+
+Restored lost speed for x22i & x25x.
+
+v3.11.2
+
+Fixed x11gost (sib) AVX2 invalid shares.
+
+Fixed x16r, x16rv2, x16s, x16rt, x16rt-veil (veil), x21s.
+No shares were submitted when cube, shavite or echo were the first function
+in the hash order.
+
+Fixed all algos reporting stats problems when mining with SSE2.
+
+Faster Lyra2 AVX512: lyra2z +47%, lyra2rev3 +11%, allium +13%, x21s +6% 
+
+Other minor performance improvements.
+
+Known issue:
+
+Lyra2 AVX512 improvements paradoxically reduced performance on x22i and x25x.
+https://github.com/JayDDee/cpuminer-opt/issues/225
+
 v3.11.1

 Faster panama for x25x AVX2 & AVX512.
--- a/algo-gate-api.c
+++ b/algo-gate-api.c
@@ -209,6 +209,7 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
    case ALGO_SHA256D:       register_sha256d_algo       ( gate ); break;
    case ALGO_SHA256Q:       register_sha256q_algo       ( gate ); break;
    case ALGO_SHA256T:       register_sha256t_algo       ( gate ); break;
+    case ALGO_SHA3D:         register_sha3d_algo         ( gate ); break;
    case ALGO_SHAVITE3:      register_shavite_algo       ( gate ); break;
    case ALGO_SKEIN:         register_skein_algo         ( gate ); break;
    case ALGO_SKEIN2:        register_skein2_algo        ( gate ); break;
@@ -247,6 +248,7 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
 */
    case ALGO_YESCRYPT:      register_yescrypt_algo      ( gate ); break;
    case ALGO_YESCRYPTR8:    register_yescryptr8_algo    ( gate ); break;
+    case ALGO_YESCRYPTR8G:   register_yescryptr8g_algo   ( gate ); break;
    case ALGO_YESCRYPTR16:   register_yescryptr16_algo   ( gate ); break;
    case ALGO_YESCRYPTR32:   register_yescryptr32_algo   ( gate ); break;
    case ALGO_YESPOWER:      register_yespower_algo      ( gate ); break;
--- a/algo-gate-api.h
+++ b/algo-gate-api.h
@@ -121,54 +121,55 @@ void ( *hash_suw ) ( void*, const void* );

 // Allocate thread local buffers and other initialization specific to miner
 // threads.
-bool ( *miner_thread_init )      ( int );
+bool ( *miner_thread_init )     ( int );

 // Generate global blockheader from stratum data.
-void ( *stratum_gen_work )       ( struct stratum_ctx*, struct work* );
+void ( *stratum_gen_work )      ( struct stratum_ctx*, struct work* );

 // Get thread local copy of blockheader with unique nonce.
-void ( *get_new_work )           ( struct work*, struct work*, int, uint32_t*,
-                                   bool );
+void ( *get_new_work )          ( struct work*, struct work*, int, uint32_t* );

 // Return pointer to nonce in blockheader.
-uint32_t *( *get_nonceptr )      ( uint32_t* );
+uint32_t *( *get_nonceptr )     ( uint32_t* );

 // Decode getwork blockheader
-bool ( *work_decode )            ( const json_t*, struct work* );
+bool ( *work_decode )           ( const json_t*, struct work* );

 // Extra getwork data
-void ( *decode_extra_data )      ( struct work*, uint64_t* );
+void ( *decode_extra_data )     ( struct work*, uint64_t* );

-bool ( *submit_getwork_result )  ( CURL*, struct work* );
+bool ( *submit_getwork_result ) ( CURL*, struct work* );

-void ( *gen_merkle_root )        ( char*, struct stratum_ctx* );
+void ( *gen_merkle_root )       ( char*, struct stratum_ctx* );

 // Increment extranonce
-void ( *build_extraheader )      ( struct work*, struct stratum_ctx* );
+void ( *build_extraheader )     ( struct work*, struct stratum_ctx* );
+
+void ( *build_block_header )    ( struct work*, uint32_t, uint32_t*,
+	                                uint32_t*, uint32_t, uint32_t,
+                                   unsigned char* );

-void ( *build_block_header )     ( struct work*, uint32_t, uint32_t*,
-	                                uint32_t*, uint32_t, uint32_t );
 // Build mining.submit message
-void ( *build_stratum_request )  ( char*, struct work*, struct stratum_ctx* );
+void ( *build_stratum_request ) ( char*, struct work*, struct stratum_ctx* );

-char* ( *malloc_txs_request )    ( struct work* );
+char* ( *malloc_txs_request )   ( struct work* );

 // Big or little
-void ( *set_work_data_endian )   ( struct work* );
+void ( *set_work_data_endian )  ( struct work* );

-double ( *calc_network_diff )    ( struct work* );
+double ( *calc_network_diff )   ( struct work* );

 // Wait for first work
-bool ( *ready_to_mine )          ( struct work*, struct stratum_ctx*, int );
+bool ( *ready_to_mine )         ( struct work*, struct stratum_ctx*, int );

 // Diverge mining threads
-bool ( *do_this_thread )         ( int );
+bool ( *do_this_thread )        ( int );

 // After do_this_thread
-void ( *resync_threads )         ( struct work* );
+void ( *resync_threads )        ( struct work* );

-json_t* (*longpoll_rpc_call)     ( CURL*, int*, char* );
-bool ( *stratum_handle_response )( json_t* );
+json_t* (*longpoll_rpc_call)      ( CURL*, int*, char* );
+bool ( *stratum_handle_response ) ( json_t* );
 set_t optimizations;
 int  ( *get_work_data_size )     ();
 int  ntime_index;
@@ -225,7 +226,7 @@ uint32_t *std_get_nonceptr( uint32_t *work_data );
 uint32_t *jr2_get_nonceptr( uint32_t *work_data );

 void std_get_new_work( struct work *work, struct work *g_work, int thr_id,
-                       uint32_t* end_nonce_ptr, bool clean_job );
+                       uint32_t* end_nonce_ptr );
 void jr2_get_new_work( struct work *work, struct work *g_work, int thr_id,
                       uint32_t* end_nonce_ptr );

@@ -256,7 +257,8 @@ double std_calc_network_diff( struct work *work );

 void std_build_block_header( struct work* g_work, uint32_t version,
 	                          uint32_t *prevhash,  uint32_t *merkle_root,
-   	                       uint32_t ntime,      uint32_t nbits );
+   	                       uint32_t ntime,      uint32_t nbits,
+                             unsigned char *final_sapling_hash );

 void std_build_extraheader( struct work *work, struct stratum_ctx *sctx );

--- a/algo/argon2/argon2a/argon2a.c
+++ b/algo/argon2/argon2a/argon2a.c
@@ -62,9 +62,7 @@ int scanhash_argon2( struct work* work, uint32_t max_nonce,
 		argon2hash(hash, endiandata);
 		if (hash[7] <= Htarg && fulltest(hash, ptarget)) {
 			pdata[19] = nonce;
-			*hashes_done = pdata[19] - first_nonce;
-			work_set_target_ratio(work, hash);
-			return 1;
+         submit_solution( work, hash, mythr );
 		}
 		nonce++;
 	} while (nonce < max_nonce && !work_restart[thr_id].restart);
--- a/algo/blake/blake-4way.c
+++ b/algo/blake/blake-4way.c
@@ -13,7 +13,7 @@ void blakehash_4way(void *state, const void *input)
     uint32_t vhash[8*4] __attribute__ ((aligned (64)));
     blake256r14_4way_context ctx;
     memcpy( &ctx, &blake_4w_ctx, sizeof ctx );
-     blake256r14_4way( &ctx, input + (64<<2), 16 );
+     blake256r14_4way_update( &ctx, input + (64<<2), 16 );
     blake256r14_4way_close( &ctx, vhash );
     dintrlv_4x32( state, state+32, state+64, state+96, vhash, 256 );
 }
@@ -36,7 +36,7 @@ int scanhash_blake_4way( struct work *work, uint32_t max_nonce,

   mm128_bswap32_intrlv80_4x32( vdata, pdata );
   blake256r14_4way_init( &blake_4w_ctx );
-   blake256r14_4way( &blake_4w_ctx, vdata, 64 );
+   blake256r14_4way_update( &blake_4w_ctx, vdata, 64 );

   do {
      *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
--- a/algo/blake/blake-hash-4way.h
+++ b/algo/blake/blake-hash-4way.h
@@ -37,8 +37,6 @@
 #ifndef __BLAKE_HASH_4WAY__
 #define __BLAKE_HASH_4WAY__ 1

-//#ifdef __SSE4_2__
-
 #ifdef __cplusplus
 extern "C"{
 #endif
@@ -51,46 +49,41 @@ extern "C"{

 #define SPH_SIZE_blake512   512

-// With SSE4.2 only Blake-256 4 way is available.
-// With AVX2 Blake-256 8way & Blake-512 4 way are also available.
-
-// Blake-256 4 way
+//////////////////////////
+//
+//   Blake-256 4 way SSE2

 typedef struct {
   unsigned char buf[64<<2];
   uint32_t H[8<<2];
-//   __m128i buf[16] __attribute__ ((aligned (64)));
-//   __m128i H[8];
-//   __m128i S[4];    
   size_t ptr;
   uint32_t T0, T1;
   int rounds;   // 14 for blake, 8 for blakecoin & vanilla
 } blake_4way_small_context __attribute__ ((aligned (64)));

-// Default 14 rounds
+// Default, 14 rounds, blake, decred
 typedef blake_4way_small_context blake256_4way_context;
 void blake256_4way_init(void *ctx);
 void blake256_4way_update(void *ctx, const void *data, size_t len);
-#define blake256_4way blake256_4way_update
 void blake256_4way_close(void *ctx, void *dst);

 // 14 rounds, blake, decred
 typedef blake_4way_small_context blake256r14_4way_context;
 void blake256r14_4way_init(void *cc);
 void blake256r14_4way_update(void *cc, const void *data, size_t len);
-#define blake256r14_4way blake256r14_4way_update
 void blake256r14_4way_close(void *cc, void *dst);

 // 8 rounds, blakecoin, vanilla
 typedef blake_4way_small_context blake256r8_4way_context;
 void blake256r8_4way_init(void *cc);
 void blake256r8_4way_update(void *cc, const void *data, size_t len);
-#define blake256r8_4way blake256r8_4way_update
 void blake256r8_4way_close(void *cc, void *dst);

 #ifdef __AVX2__

-// Blake-256 8 way
+//////////////////////////
+//
+//   Blake-256 8 way AVX2

 typedef struct {
   __m256i buf[16] __attribute__ ((aligned (64)));
@@ -104,7 +97,6 @@ typedef struct {
 typedef blake_8way_small_context blake256_8way_context;
 void blake256_8way_init(void *cc);
 void blake256_8way_update(void *cc, const void *data, size_t len);
-//#define blake256_8way blake256_8way_update
 void blake256_8way_close(void *cc, void *dst);

 // 14 rounds, blake, decred
@@ -117,10 +109,9 @@ void blake256r14_8way_close(void *cc, void *dst);
 typedef blake_8way_small_context blake256r8_8way_context;
 void blake256r8_8way_init(void *cc);
 void blake256r8_8way_update(void *cc, const void *data, size_t len);
-#define blake256r8_8way blake256r8_8way_update
 void blake256r8_8way_close(void *cc, void *dst);

-// Blake-512 4 way
+// Blake-512 4 way AVX2

 typedef struct {
   __m256i buf[16];
@@ -134,14 +125,15 @@ typedef blake_4way_big_context blake512_4way_context;

 void blake512_4way_init( blake_4way_big_context *sc );
 void blake512_4way_update( void *cc, const void *data, size_t len );
-#define blake512_4way blake512_4way_update
 void blake512_4way_close( void *cc, void *dst );
-void blake512_4way_addbits_and_close( void *cc, unsigned ub, unsigned n,
-                                      void *dst );
+void blake512_4way_full( blake_4way_big_context *sc, void * dst,
+                         const void *data, size_t len );

 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)

-//Blake-256 16 way
+////////////////////////////
+//
+//   Blake-256 16 way AVX512

 typedef struct {
   __m512i buf[16];
@@ -169,8 +161,9 @@ void blake256r8_16way_init(void *cc);
 void blake256r8_16way_update(void *cc, const void *data, size_t len);
 void blake256r8_16way_close(void *cc, void *dst);

-
-// Blake-512 8 way
+////////////////////////////
+//
+//// Blake-512 8 way AVX512

 typedef struct {
   __m512i buf[16];
@@ -185,12 +178,10 @@ typedef blake_8way_big_context blake512_8way_context;
 void blake512_8way_init( blake_8way_big_context *sc );
 void blake512_8way_update( void *cc, const void *data, size_t len );
 void blake512_8way_close( void *cc, void *dst );
-void blake512_8way_addbits_and_close( void *cc, unsigned ub, unsigned n,
-                                      void *dst );
+void blake512_8way_full( blake_8way_big_context *sc, void * dst,
+                        const void *data, size_t len );

 #endif  // AVX512
-
-
 #endif  // AVX2

 #ifdef __cplusplus
--- a/algo/blake/blake2b-4way.c
+++ b/algo/blake/blake2b-4way.c
@@ -39,7 +39,7 @@ int scanhash_blake2b_8way( struct work *work, uint32_t max_nonce,
      blake2b_8way_final( &ctx, hash );

      for ( int lane = 0; lane < 8; lane++ )
-      if ( hash7[ lane<<1 ] < Htarg )
+      if ( hash7[ lane<<1 ] <= Htarg )
      {
          extr_lane_8x64( lane_hash, hash, lane, 256 );
          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
@@ -94,7 +94,7 @@ int scanhash_blake2b_4way( struct work *work, uint32_t max_nonce,
      blake2b_4way_final( &ctx, hash );

      for ( int lane = 0; lane < 4; lane++ )
-      if ( hash7[ lane<<1 ] < Htarg )
+      if ( hash7[ lane<<1 ] <= Htarg )
      {
          extr_lane_4x64( lane_hash, hash, lane, 256 );
          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
--- a/algo/blake/blake2b.c
+++ b/algo/blake/blake2b.c
@@ -43,17 +43,14 @@ int scanhash_blake2b( struct work *work, uint32_t max_nonce,

 	do {
 		be32enc(&endiandata[19], n);
-		//blake2b_hash_end(vhashcpu, endiandata);
 		blake2b_hash(vhashcpu, endiandata);

-		if (vhashcpu[7] < Htarg && fulltest(vhashcpu, ptarget)) {
-			work_set_target_ratio(work, vhashcpu);
-			*hashes_done = n - first_nonce + 1;
+		if (vhashcpu[7] <= Htarg && fulltest(vhashcpu, ptarget))
+      {
 			pdata[19] = n;
-			return 1;
-		}
-		n++;
-
+         submit_solution( work, vhashcpu, mythr );
+      }
+      n++;
 	} while (n < max_nonce && !work_restart[thr_id].restart);
 	*hashes_done = n - first_nonce + 1;
 	pdata[19] = n;
--- a/algo/blake/blake2s.c
+++ b/algo/blake/blake2s.c
@@ -56,7 +56,7 @@ int scanhash_blake2s( struct work *work,
 	do {
 		be32enc(&endiandata[19], n);
 		blake2s_hash( hash64, endiandata );
-		if (hash64[7] < Htarg && fulltest(hash64, ptarget)) {
+		if (hash64[7] <= Htarg && fulltest(hash64, ptarget)) {
 			*hashes_done = n - first_nonce + 1;
 			pdata[19] = n;
 			return true;
--- a/algo/blake/blake512-hash-4way.c
+++ b/algo/blake/blake512-hash-4way.c
@@ -267,22 +267,22 @@ static const sph_u64 CB[16] = {
 #define CBx_(n)     CBx__(n)
 #define CBx__(n)    CB ## n

-#define CB0   SPH_C64(0x243F6A8885A308D3)
-#define CB1   SPH_C64(0x13198A2E03707344)
-#define CB2   SPH_C64(0xA4093822299F31D0)
-#define CB3   SPH_C64(0x082EFA98EC4E6C89)
-#define CB4   SPH_C64(0x452821E638D01377)
-#define CB5   SPH_C64(0xBE5466CF34E90C6C)
-#define CB6   SPH_C64(0xC0AC29B7C97C50DD)
-#define CB7   SPH_C64(0x3F84D5B5B5470917)
-#define CB8   SPH_C64(0x9216D5D98979FB1B)
-#define CB9   SPH_C64(0xD1310BA698DFB5AC)
-#define CBA   SPH_C64(0x2FFD72DBD01ADFB7)
-#define CBB   SPH_C64(0xB8E1AFED6A267E96)
-#define CBC   SPH_C64(0xBA7C9045F12C7F99)
-#define CBD   SPH_C64(0x24A19947B3916CF7)
-#define CBE   SPH_C64(0x0801F2E2858EFC16)
-#define CBF   SPH_C64(0x636920D871574E69)
+#define CB0   0x243F6A8885A308D3
+#define CB1   0x13198A2E03707344
+#define CB2   0xA4093822299F31D0
+#define CB3   0x082EFA98EC4E6C89
+#define CB4   0x452821E638D01377
+#define CB5   0xBE5466CF34E90C6C
+#define CB6   0xC0AC29B7C97C50DD
+#define CB7   0x3F84D5B5B5470917
+#define CB8   0x9216D5D98979FB1B
+#define CB9   0xD1310BA698DFB5AC
+#define CBA   0x2FFD72DBD01ADFB7
+#define CBB   0xB8E1AFED6A267E96
+#define CBC   0xBA7C9045F12C7F99
+#define CBD   0x24A19947B3916CF7
+#define CBE   0x0801F2E2858EFC16
+#define CBF   0x636920D871574E69

 #define READ_STATE64(state)   do { \
      H0 = (state)->H[0]; \
@@ -349,9 +349,9 @@ static const sph_u64 CB[16] = {
 #define DECL_STATE64_8WAY \
   __m512i H0, H1, H2, H3, H4, H5, H6, H7; \
        __m512i S0, S1, S2, S3; \
-   sph_u64 T0, T1;
+   uint64_t T0, T1;

-#define COMPRESS64_8WAY   do \
+#define COMPRESS64_8WAY( buf )   do \
 { \
  __m512i M0, M1, M2, M3, M4, M5, M6, M7; \
  __m512i M8, M9, MA, MB, MC, MD, ME, MF; \
@@ -424,6 +424,84 @@ static const sph_u64 CB[16] = {
  H7 = mm512_xor4( VF, V7, S3, H7 ); \
 } while (0)

+void blake512_8way_compress( blake_8way_big_context *sc )
+{ 
+  __m512i M0, M1, M2, M3, M4, M5, M6, M7;
+  __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;
+  __m512i shuf_bswap64;
+
+  V0 = sc->H[0];
+  V1 = sc->H[1];
+  V2 = sc->H[2];
+  V3 = sc->H[3];
+  V4 = sc->H[4];
+  V5 = sc->H[5];
+  V6 = sc->H[6];
+  V7 = sc->H[7];
+  V8 = _mm512_xor_si512( sc->S[0], m512_const1_64( CB0 ) );
+  V9 = _mm512_xor_si512( sc->S[1], m512_const1_64( CB1 ) );
+  VA = _mm512_xor_si512( sc->S[2], m512_const1_64( CB2 ) );
+  VB = _mm512_xor_si512( sc->S[3], m512_const1_64( CB3 ) );
+  VC = _mm512_xor_si512( _mm512_set1_epi64( sc->T0 ),
+                            m512_const1_64( CB4 ) );
+  VD = _mm512_xor_si512( _mm512_set1_epi64( sc->T0 ),
+                            m512_const1_64( CB5 ) );
+  VE = _mm512_xor_si512( _mm512_set1_epi64( sc->T1 ),
+                            m512_const1_64( CB6 ) );
+  VF = _mm512_xor_si512( _mm512_set1_epi64( sc->T1 ),
+                            m512_const1_64( CB7 ) );
+
+  shuf_bswap64 = m512_const_64( 0x38393a3b3c3d3e3f, 0x3031323334353637,
+                                0x28292a2b2c2d2e2f, 0x2021222324252627,
+                                0x18191a1b1c1d1e1f, 0x1011121314151617,
+                                0x08090a0b0c0d0e0f, 0x0001020304050607 );
+
+  M0 = _mm512_shuffle_epi8( sc->buf[ 0], shuf_bswap64 );
+  M1 = _mm512_shuffle_epi8( sc->buf[ 1], shuf_bswap64 );
+  M2 = _mm512_shuffle_epi8( sc->buf[ 2], shuf_bswap64 );
+  M3 = _mm512_shuffle_epi8( sc->buf[ 3], shuf_bswap64 );
+  M4 = _mm512_shuffle_epi8( sc->buf[ 4], shuf_bswap64 );
+  M5 = _mm512_shuffle_epi8( sc->buf[ 5], shuf_bswap64 );
+  M6 = _mm512_shuffle_epi8( sc->buf[ 6], shuf_bswap64 );
+  M7 = _mm512_shuffle_epi8( sc->buf[ 7], shuf_bswap64 );
+  M8 = _mm512_shuffle_epi8( sc->buf[ 8], shuf_bswap64 );
+  M9 = _mm512_shuffle_epi8( sc->buf[ 9], shuf_bswap64 );
+  MA = _mm512_shuffle_epi8( sc->buf[10], shuf_bswap64 );
+  MB = _mm512_shuffle_epi8( sc->buf[11], shuf_bswap64 );
+  MC = _mm512_shuffle_epi8( sc->buf[12], shuf_bswap64 );
+  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);
+  ROUND_B_8WAY(2);
+  ROUND_B_8WAY(3);
+  ROUND_B_8WAY(4);
+  ROUND_B_8WAY(5);
+  ROUND_B_8WAY(6);
+  ROUND_B_8WAY(7);
+  ROUND_B_8WAY(8);
+  ROUND_B_8WAY(9);
+  ROUND_B_8WAY(0);
+  ROUND_B_8WAY(1);
+  ROUND_B_8WAY(2);
+  ROUND_B_8WAY(3);
+  ROUND_B_8WAY(4);
+  ROUND_B_8WAY(5);
+
+  sc->H[0] = mm512_xor4( V8, V0, sc->S[0], sc->H[0] );
+  sc->H[1] = mm512_xor4( V9, V1, sc->S[1], sc->H[1] );
+  sc->H[2] = mm512_xor4( VA, V2, sc->S[2], sc->H[2] );
+  sc->H[3] = mm512_xor4( VB, V3, sc->S[3], sc->H[3] );
+  sc->H[4] = mm512_xor4( VC, V4, sc->S[0], sc->H[4] );
+  sc->H[5] = mm512_xor4( VD, V5, sc->S[1], sc->H[5] );
+  sc->H[6] = mm512_xor4( VE, V6, sc->S[2], sc->H[6] );
+  sc->H[7] = mm512_xor4( VF, V7, sc->S[3], sc->H[7] );
+}
+
 void blake512_8way_init( blake_8way_big_context *sc )
 {
   __m512i zero = m512_zero;
@@ -455,39 +533,43 @@ blake64_8way( blake_8way_big_context *sc, const void *data, size_t len )

   const int buf_size = 128;  //  sizeof/8

+// 64, 80 bytes: 1st pass copy data. 2nd pass copy padding and compress.   
+// 128 bytes: 1st pass copy data, compress. 2nd pass copy padding, compress.
+   
   buf = sc->buf;
   ptr = sc->ptr;
   if ( len < (buf_size - ptr) )
   {
-   memcpy_512( buf + (ptr>>3), vdata, len>>3 );
-   ptr += len;
-   sc->ptr = ptr;
-   return;
+      memcpy_512( buf + (ptr>>3), vdata, len>>3 );
+      ptr += len;
+      sc->ptr = ptr;
+      return;
   }

   READ_STATE64(sc);
   while ( len > 0 )
   {
-   size_t clen;
+      size_t clen;

-   clen = buf_size - ptr;
-   if ( clen > len )
+      clen = buf_size - ptr;
+      if ( clen > len )
      clen = len;
-   memcpy_512( buf + (ptr>>3), vdata, clen>>3 );
-   ptr += clen;
-   vdata = vdata + (clen>>3);
-   len -= clen;
-   if ( ptr == buf_size )
-        {
-      if ( ( T0 = SPH_T64(T0 + 1024) ) < 1024 )
-         T1 = SPH_T64(T1 + 1);
-      COMPRESS64_8WAY;
-      ptr = 0;
-   }
+      memcpy_512( buf + (ptr>>3), vdata, clen>>3 );
+      ptr += clen;
+      vdata = vdata + (clen>>3);
+      len -= clen;
+      if ( ptr == buf_size )
+      {
+         if ( ( T0 = T0 + 1024 ) < 1024 )
+            T1 = T1 + 1;
+         COMPRESS64_8WAY( buf );
+         ptr = 0;
+      }
   }
   WRITE_STATE64(sc);
   sc->ptr = ptr;
-}
+
+   }

 static void
 blake64_8way_close( blake_8way_big_context *sc, void *dst )
@@ -495,26 +577,22 @@ blake64_8way_close( blake_8way_big_context *sc, void *dst )
   __m512i buf[16];
   size_t ptr;
   unsigned bit_len;
-//   uint64_t z, zz;
-   sph_u64 th, tl;
+   uint64_t th, tl;

   ptr = sc->ptr;
   bit_len = ((unsigned)ptr << 3);
-//   z = 0x80 >> n;
-//   zz = ((ub & -z) | z) & 0xFF;
-//   buf[ptr>>3] = _mm512_set1_epi64( zz );
   buf[ptr>>3] = m512_const1_64( 0x80 );
   tl = sc->T0 + bit_len;
   th = sc->T1;
   if (ptr == 0 )
   {
-   sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
-   sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
+   sc->T0 = 0xFFFFFFFFFFFFFC00ULL;
+   sc->T1 = 0xFFFFFFFFFFFFFFFFULL;
   }
   else if ( sc->T0 == 0 )
   {
-   sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL) + bit_len;
-   sc->T1 = SPH_T64(sc->T1 - 1);
+   sc->T0 = 0xFFFFFFFFFFFFFC00ULL + bit_len;
+   sc->T1 = sc->T1 - 1;
   }
   else
   {
@@ -535,8 +613,8 @@ blake64_8way_close( blake_8way_big_context *sc, void *dst )
       memset_zero_512( buf + (ptr>>3) + 1, (120 - ptr) >> 3 );

       blake64_8way( sc, buf + (ptr>>3), 128 - ptr );
-       sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
-       sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
+       sc->T0 = 0xFFFFFFFFFFFFFC00ULL;
+       sc->T1 = 0xFFFFFFFFFFFFFFFFULL;
       memset_zero_512( buf, 112>>3 );
       buf[104>>3] = m512_const1_64( 0x0100000000000000ULL );
       buf[112>>3] = m512_const1_64( bswap_64( th ) );
@@ -547,6 +625,79 @@ blake64_8way_close( blake_8way_big_context *sc, void *dst )
   mm512_block_bswap_64( (__m512i*)dst, sc->H );
 }

+// init, update & close
+void blake512_8way_full( blake_8way_big_context *sc, void * dst, 
+                        const void *data, size_t len )
+{
+   
+// init
+
+   casti_m512i( sc->H, 0 ) = m512_const1_64( 0x6A09E667F3BCC908 );
+   casti_m512i( sc->H, 1 ) = m512_const1_64( 0xBB67AE8584CAA73B );
+   casti_m512i( sc->H, 2 ) = m512_const1_64( 0x3C6EF372FE94F82B );
+   casti_m512i( sc->H, 3 ) = m512_const1_64( 0xA54FF53A5F1D36F1 );
+   casti_m512i( sc->H, 4 ) = m512_const1_64( 0x510E527FADE682D1 );
+   casti_m512i( sc->H, 5 ) = m512_const1_64( 0x9B05688C2B3E6C1F );
+   casti_m512i( sc->H, 6 ) = m512_const1_64( 0x1F83D9ABFB41BD6B );
+   casti_m512i( sc->H, 7 ) = m512_const1_64( 0x5BE0CD19137E2179 );
+
+   casti_m512i( sc->S, 0 ) = m512_zero;
+   casti_m512i( sc->S, 1 ) = m512_zero;
+   casti_m512i( sc->S, 2 ) = m512_zero;
+   casti_m512i( sc->S, 3 ) = m512_zero;
+
+   sc->T0 = sc->T1 = 0;
+   sc->ptr = 0;
+
+// update
+
+   memcpy_512( sc->buf, (__m512i*)data, len>>3 );
+   sc->ptr = len;
+   if ( len == 128 )
+   {
+      if ( ( sc->T0 = sc->T0 + 1024 ) < 1024 )
+            sc->T1 = sc->T1 + 1;
+      blake512_8way_compress( sc );
+      sc->ptr = 0;
+   }
+
+// close
+
+   size_t ptr64 = sc->ptr >> 3;
+   unsigned bit_len;
+   uint64_t th, tl;
+
+   bit_len = sc->ptr << 3;
+   sc->buf[ptr64] = m512_const1_64( 0x80 );
+   tl = sc->T0 + bit_len;
+   th = sc->T1;
+
+   if ( ptr64 == 0 )
+   {
+   sc->T0 = 0xFFFFFFFFFFFFFC00ULL;
+   sc->T1 = 0xFFFFFFFFFFFFFFFFULL;
+   }
+   else if ( sc->T0 == 0 )
+   {
+   sc->T0 = 0xFFFFFFFFFFFFFC00ULL + bit_len;
+   sc->T1 = sc->T1 - 1;
+   }
+   else
+      sc->T0 -= 1024 - bit_len;
+
+   memset_zero_512( sc->buf + ptr64 + 1, 13 - ptr64 );
+   sc->buf[13] = m512_const1_64( 0x0100000000000000ULL );
+   sc->buf[14] = m512_const1_64( bswap_64( th ) );
+   sc->buf[15] = m512_const1_64( bswap_64( tl ) );
+
+   if ( ( sc->T0 = sc->T0 + 1024 ) < 1024 )
+       sc->T1 = sc->T1 + 1;
+
+   blake512_8way_compress( sc );
+   
+   mm512_block_bswap_64( (__m512i*)dst, sc->H );
+}
+   
 void
 blake512_8way_update(void *cc, const void *data, size_t len)
 {
@@ -555,12 +706,6 @@ blake512_8way_update(void *cc, const void *data, size_t len)

 void
 blake512_8way_close(void *cc, void *dst)
-{
-   blake512_8way_addbits_and_close(cc, 0, 0, dst);
-}
-
-void
-blake512_8way_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
 {
   blake64_8way_close(cc, dst);
 }
@@ -596,7 +741,7 @@ blake512_8way_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
 #define DECL_STATE64_4WAY \
 	__m256i H0, H1, H2, H3, H4, H5, H6, H7; \
        __m256i S0, S1, S2, S3; \
-	sph_u64 T0, T1;
+	uint64_t T0, T1;

 #define COMPRESS64_4WAY   do \
 { \
@@ -670,6 +815,81 @@ blake512_8way_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
 } while (0)


+void blake512_4way_compress( blake_4way_big_context *sc )
+{
+  __m256i M0, M1, M2, M3, M4, M5, M6, M7;
+  __m256i M8, M9, MA, MB, MC, MD, ME, MF;
+  __m256i V0, V1, V2, V3, V4, V5, V6, V7;
+  __m256i V8, V9, VA, VB, VC, VD, VE, VF;
+  __m256i shuf_bswap64;
+
+  V0 = sc->H[0];
+  V1 = sc->H[1];
+  V2 = sc->H[2];
+  V3 = sc->H[3];
+  V4 = sc->H[4];
+  V5 = sc->H[5];
+  V6 = sc->H[6];
+  V7 = sc->H[7];
+  V8 = _mm256_xor_si256( sc->S[0], m256_const1_64( CB0 ) );
+  V9 = _mm256_xor_si256( sc->S[1], m256_const1_64( CB1 ) );
+  VA = _mm256_xor_si256( sc->S[2], m256_const1_64( CB2 ) );
+  VB = _mm256_xor_si256( sc->S[3], m256_const1_64( CB3 ) );
+  VC = _mm256_xor_si256( _mm256_set1_epi64x( sc->T0 ),
+                             m256_const1_64( CB4 ) );
+  VD = _mm256_xor_si256( _mm256_set1_epi64x( sc->T0 ),
+                             m256_const1_64( CB5 ) );
+  VE = _mm256_xor_si256( _mm256_set1_epi64x( sc->T1 ),
+                             m256_const1_64( CB6 ) );
+  VF = _mm256_xor_si256( _mm256_set1_epi64x( sc->T1 ),
+                             m256_const1_64( CB7 ) );
+  shuf_bswap64 = m256_const_64( 0x18191a1b1c1d1e1f, 0x1011121314151617,
+                                0x08090a0b0c0d0e0f, 0x0001020304050607 );
+
+  M0 = _mm256_shuffle_epi8( sc->buf[ 0], shuf_bswap64 );
+  M1 = _mm256_shuffle_epi8( sc->buf[ 1], shuf_bswap64 );
+  M2 = _mm256_shuffle_epi8( sc->buf[ 2], shuf_bswap64 );
+  M3 = _mm256_shuffle_epi8( sc->buf[ 3], shuf_bswap64 );
+  M4 = _mm256_shuffle_epi8( sc->buf[ 4], shuf_bswap64 );
+  M5 = _mm256_shuffle_epi8( sc->buf[ 5], shuf_bswap64 );
+  M6 = _mm256_shuffle_epi8( sc->buf[ 6], shuf_bswap64 );
+  M7 = _mm256_shuffle_epi8( sc->buf[ 7], shuf_bswap64 );
+  M8 = _mm256_shuffle_epi8( sc->buf[ 8], shuf_bswap64 );
+  M9 = _mm256_shuffle_epi8( sc->buf[ 9], shuf_bswap64 );
+  MA = _mm256_shuffle_epi8( sc->buf[10], shuf_bswap64 );
+  MB = _mm256_shuffle_epi8( sc->buf[11], shuf_bswap64 );
+  MC = _mm256_shuffle_epi8( sc->buf[12], shuf_bswap64 );
+  MD = _mm256_shuffle_epi8( sc->buf[13], shuf_bswap64 );
+  ME = _mm256_shuffle_epi8( sc->buf[14], shuf_bswap64 );
+  MF = _mm256_shuffle_epi8( sc->buf[15], shuf_bswap64 );
+
+  ROUND_B_4WAY(0);
+  ROUND_B_4WAY(1);
+  ROUND_B_4WAY(2);
+  ROUND_B_4WAY(3);
+  ROUND_B_4WAY(4);
+  ROUND_B_4WAY(5);
+  ROUND_B_4WAY(6);
+  ROUND_B_4WAY(7);
+  ROUND_B_4WAY(8);
+  ROUND_B_4WAY(9);
+  ROUND_B_4WAY(0);
+  ROUND_B_4WAY(1);
+  ROUND_B_4WAY(2);
+  ROUND_B_4WAY(3);
+  ROUND_B_4WAY(4);
+  ROUND_B_4WAY(5);
+
+  sc->H[0] = mm256_xor4( V8, V0, sc->S[0], sc->H[0] );
+  sc->H[1] = mm256_xor4( V9, V1, sc->S[1], sc->H[1] );
+  sc->H[2] = mm256_xor4( VA, V2, sc->S[2], sc->H[2] );
+  sc->H[3] = mm256_xor4( VB, V3, sc->S[3], sc->H[3] );
+  sc->H[4] = mm256_xor4( VC, V4, sc->S[0], sc->H[4] );
+  sc->H[5] = mm256_xor4( VD, V5, sc->S[1], sc->H[5] );
+  sc->H[6] = mm256_xor4( VE, V6, sc->S[2], sc->H[6] );
+  sc->H[7] = mm256_xor4( VF, V7, sc->S[3], sc->H[7] );
+}
+
 void blake512_4way_init( blake_4way_big_context *sc )
 {
   __m256i zero = m256_zero;
@@ -681,10 +901,12 @@ void blake512_4way_init( blake_4way_big_context *sc )
   casti_m256i( sc->H, 5 ) = m256_const1_64( 0x9B05688C2B3E6C1F );
   casti_m256i( sc->H, 6 ) = m256_const1_64( 0x1F83D9ABFB41BD6B );
   casti_m256i( sc->H, 7 ) = m256_const1_64( 0x5BE0CD19137E2179 );
+
   casti_m256i( sc->S, 0 ) = zero;
   casti_m256i( sc->S, 1 ) = zero;
   casti_m256i( sc->S, 2 ) = zero;
   casti_m256i( sc->S, 3 ) = zero;
+
   sc->T0 = sc->T1 = 0;
   sc->ptr = 0;
 }
@@ -703,31 +925,31 @@ blake64_4way( blake_4way_big_context *sc, const void *data, size_t len)
   ptr = sc->ptr;
   if ( len < (buf_size - ptr) )
   {
-	memcpy_256( buf + (ptr>>3), vdata, len>>3 );
-	ptr += len;
-	sc->ptr = ptr;
-	return;
+   	memcpy_256( buf + (ptr>>3), vdata, len>>3 );
+	   ptr += len;
+	   sc->ptr = ptr;
+	   return;
   }

   READ_STATE64(sc);
   while ( len > 0 )
   {
-	size_t clen;
+   	size_t clen;

-	clen = buf_size - ptr;
-	if ( clen > len )
-		clen = len;
-	memcpy_256( buf + (ptr>>3), vdata, clen>>3 );
-	ptr += clen;
-	vdata = vdata + (clen>>3);
-	len -= clen;
-	if (ptr == buf_size )
-        {
-		if ((T0 = SPH_T64(T0 + 1024)) < 1024)
-			T1 = SPH_T64(T1 + 1);
-		COMPRESS64_4WAY;
-		ptr = 0;
-	}
+	   clen = buf_size - ptr;
+	   if ( clen > len )
+		   clen = len;
+   	memcpy_256( buf + (ptr>>3), vdata, clen>>3 );
+	   ptr += clen;
+	   vdata = vdata + (clen>>3);
+	   len -= clen;
+	   if ( ptr == buf_size )
+      {
+		   if ( (T0 = T0 + 1024 ) < 1024 )
+			   T1 = SPH_T64(T1 + 1);
+	   	COMPRESS64_4WAY;
+		   ptr = 0;
+	   }
   }
   WRITE_STATE64(sc);
   sc->ptr = ptr;
@@ -739,7 +961,7 @@ blake64_4way_close( blake_4way_big_context *sc, void *dst )
   __m256i buf[16];
   size_t ptr;
   unsigned bit_len;
-   sph_u64 th, tl;
+   uint64_t th, tl;

   ptr = sc->ptr;
   bit_len = ((unsigned)ptr << 3);
@@ -748,13 +970,13 @@ blake64_4way_close( blake_4way_big_context *sc, void *dst )
   th = sc->T1;
   if (ptr == 0 )
   {
-	sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
-	sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
+	sc->T0 = 0xFFFFFFFFFFFFFC00ULL;
+	sc->T1 = 0xFFFFFFFFFFFFFFFFULL;
   }
   else if ( sc->T0 == 0 )
   {
-	sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL) + bit_len;
-	sc->T1 = SPH_T64(sc->T1 - 1);
+	sc->T0 = 0xFFFFFFFFFFFFFC00ULL + bit_len;
+	sc->T1 = sc->T1 - 1;
   } 
   else
   {
@@ -788,13 +1010,77 @@ blake64_4way_close( blake_4way_big_context *sc, void *dst )
   mm256_block_bswap_64( (__m256i*)dst, sc->H );
 }

-/*
-void
-blake512_4way_init(void *cc)
+// init, update & close
+void blake512_4way_full( blake_4way_big_context *sc, void * dst,
+                         const void *data, size_t len )
 {
-	blake64_4way_init(cc, IV512, salt_zero_big);
+
+// init
+
+   casti_m256i( sc->H, 0 ) = m256_const1_64( 0x6A09E667F3BCC908 );
+   casti_m256i( sc->H, 1 ) = m256_const1_64( 0xBB67AE8584CAA73B );
+   casti_m256i( sc->H, 2 ) = m256_const1_64( 0x3C6EF372FE94F82B );
+   casti_m256i( sc->H, 3 ) = m256_const1_64( 0xA54FF53A5F1D36F1 );
+   casti_m256i( sc->H, 4 ) = m256_const1_64( 0x510E527FADE682D1 );
+   casti_m256i( sc->H, 5 ) = m256_const1_64( 0x9B05688C2B3E6C1F );
+   casti_m256i( sc->H, 6 ) = m256_const1_64( 0x1F83D9ABFB41BD6B );
+   casti_m256i( sc->H, 7 ) = m256_const1_64( 0x5BE0CD19137E2179 );
+
+   casti_m256i( sc->S, 0 ) = m256_zero;
+   casti_m256i( sc->S, 1 ) = m256_zero;
+   casti_m256i( sc->S, 2 ) = m256_zero;
+   casti_m256i( sc->S, 3 ) = m256_zero;
+
+   sc->T0 = sc->T1 = 0;
+   sc->ptr = 0;
+
+// update
+
+   memcpy_256( sc->buf, (__m256i*)data, len>>3 );
+   sc->ptr += len;
+   if ( len == 128 )
+   {
+      if ( ( sc->T0 = sc->T0 + 1024 ) < 1024 )
+         sc->T1 =  sc->T1 + 1;
+      blake512_4way_compress( sc );
+      sc->ptr = 0;
+   }
+
+// close
+
+   size_t ptr64 = sc->ptr >> 3;
+   unsigned bit_len;
+   uint64_t th, tl;
+
+   bit_len = sc->ptr << 3;
+   sc->buf[ptr64] = m256_const1_64( 0x80 );
+   tl = sc->T0 + bit_len;
+   th = sc->T1;
+   if ( sc->ptr == 0 )
+   {
+      sc->T0 = 0xFFFFFFFFFFFFFC00ULL;
+      sc->T1 = 0xFFFFFFFFFFFFFFFFULL;
+   }
+   else if ( sc->T0 == 0 )
+   {
+      sc->T0 = 0xFFFFFFFFFFFFFC00ULL + bit_len;
+      sc->T1 = sc->T1 - 1;
+   }
+   else
+        sc->T0 -= 1024 - bit_len;
+
+   memset_zero_256( sc->buf + ptr64 + 1, 13 - ptr64 );
+   sc->buf[13] = m256_const1_64( 0x0100000000000000ULL );
+   sc->buf[14] = m256_const1_64( bswap_64( th ) );
+   sc->buf[15] = m256_const1_64( bswap_64( tl ) );
+
+   if ( ( sc->T0 = sc->T0 + 1024 ) < 1024 )
+       sc->T1 = sc->T1 + 1;
+
+   blake512_4way_compress( sc );
+
+   mm256_block_bswap_64( (__m256i*)dst, sc->H );
 }
-*/

 void
 blake512_4way_update(void *cc, const void *data, size_t len)
@@ -806,17 +1092,8 @@ void
 blake512_4way_close(void *cc, void *dst)
 {
   blake64_4way_close( cc, dst );
-
-//   blake512_4way_addbits_and_close(cc, dst);
 }

-/*
-void
-blake512_4way_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
-{
-	blake64_4way_close(cc, ub, n, dst, 8);
-}
-*/
 #ifdef __cplusplus
 }
 #endif
--- a/algo/blake/blakecoin-4way.c
+++ b/algo/blake/blakecoin-4way.c
@@ -14,7 +14,7 @@ void blakecoin_4way_hash(void *state, const void *input)
     blake256r8_4way_context ctx;

     memcpy( &ctx, &blakecoin_4w_ctx, sizeof ctx );
-     blake256r8_4way( &ctx, input + (64<<2), 16 );
+     blake256r8_4way_update( &ctx, input + (64<<2), 16 );
     blake256r8_4way_close( &ctx, vhash );

     dintrlv_4x32( state, state+32, state+64, state+96, vhash, 256 );
@@ -37,7 +37,7 @@ int scanhash_blakecoin_4way( struct work *work, uint32_t max_nonce,

   mm128_bswap32_intrlv80_4x32( vdata, pdata );
   blake256r8_4way_init( &blakecoin_4w_ctx );
-   blake256r8_4way( &blakecoin_4w_ctx, vdata, 64 );
+   blake256r8_4way_update( &blakecoin_4w_ctx, vdata, 64 );

   do {
      *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
@@ -71,7 +71,7 @@ void blakecoin_8way_hash( void *state, const void *input )
     blake256r8_8way_context ctx;

     memcpy( &ctx, &blakecoin_8w_ctx, sizeof ctx );
-     blake256r8_8way( &ctx, input + (64<<3), 16 );
+     blake256r8_8way_update( &ctx, input + (64<<3), 16 );
     blake256r8_8way_close( &ctx, vhash );

     dintrlv_8x32( state,     state+ 32, state+ 64, state+ 96, state+128,
@@ -95,7 +95,7 @@ int scanhash_blakecoin_8way( struct work *work, uint32_t max_nonce,

   mm256_bswap32_intrlv80_8x32( vdata, pdata );
   blake256r8_8way_init( &blakecoin_8w_ctx );
-   blake256r8_8way( &blakecoin_8w_ctx, vdata, 64 );
+   blake256r8_8way_update( &blakecoin_8w_ctx, vdata, 64 );

   do {
      *noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
--- a/algo/blake/decred-4way.c
+++ b/algo/blake/decred-4way.c
@@ -21,7 +21,7 @@ void decred_hash_4way( void *state, const void *input )
     blake256_4way_context ctx __attribute__ ((aligned (64)));

     memcpy( &ctx, &blake_mid, sizeof(blake_mid) );
-     blake256_4way( &ctx, tail, tail_len );
+     blake256_4way_update( &ctx, tail, tail_len );
     blake256_4way_close( &ctx, vhash );
     dintrlv_4x32( state, state+32, state+64, state+96, vhash, 256 );
 }
@@ -46,7 +46,7 @@ int scanhash_decred_4way( struct work *work, uint32_t max_nonce,
   mm128_intrlv_4x32x( vdata, edata, edata, edata, edata, 180*8 );

   blake256_4way_init( &blake_mid );
-   blake256_4way( &blake_mid, vdata, DECRED_MIDSTATE_LEN );
+   blake256_4way_update( &blake_mid, vdata, DECRED_MIDSTATE_LEN );

   uint32_t *noncep = vdata + DECRED_NONCE_INDEX * 4;
   do {
--- a/algo/blake/decred.c
+++ b/algo/blake/decred.c
@@ -77,25 +77,15 @@ int scanhash_decred( struct work *work, uint32_t max_nonce,
                be32enc(&endiandata[k], pdata[k]);
 #endif

-#ifdef DEBUG_ALGO
-        if (!thr_id) applog(LOG_DEBUG,"[%d] Target=%08x %08x", thr_id, ptarget[6], ptarget[7]);
-#endif
-
        do {
                //be32enc(&endiandata[DCR_NONCE_OFT32], n);
                endiandata[DECRED_NONCE_INDEX] = n;
                decred_hash(hash32, endiandata);

-                if (hash32[7] <= HTarget && fulltest(hash32, ptarget)) {
-                        work_set_target_ratio(work, hash32);
-                        *hashes_done = n - first_nonce + 1;
-#ifdef DEBUG_ALGO
-                        applog(LOG_BLUE, "Nonce : %08x %08x", n, swab32(n));
-                        applog_hash(ptarget);
-                        applog_compare_hash(hash32, ptarget);
-#endif
-                        pdata[DECRED_NONCE_INDEX] = n;
-                        return 1;
+                if (hash32[7] <= HTarget && fulltest(hash32, ptarget))
+                {
+                   pdata[DECRED_NONCE_INDEX] = n;
+                   submit_solution( work, hash32, mythr );
                }

                n++;
--- a/algo/blake/pentablake-4way.c
+++ b/algo/blake/pentablake-4way.c
@@ -22,23 +22,23 @@ extern void pentablakehash_4way( void *output, const void *input )


     blake512_4way_init( &ctx );
-     blake512_4way( &ctx, input, 80 );
+     blake512_4way_update( &ctx, input, 80 );
     blake512_4way_close( &ctx, vhash );

     blake512_4way_init( &ctx );
-     blake512_4way( &ctx, vhash, 64 );
+     blake512_4way_update( &ctx, vhash, 64 );
     blake512_4way_close( &ctx, vhash );

     blake512_4way_init( &ctx );
-     blake512_4way( &ctx, vhash, 64 );
+     blake512_4way_update( &ctx, vhash, 64 );
     blake512_4way_close( &ctx, vhash );

     blake512_4way_init( &ctx );
-     blake512_4way( &ctx, vhash, 64 );
+     blake512_4way_update( &ctx, vhash, 64 );
     blake512_4way_close( &ctx, vhash );

     blake512_4way_init( &ctx );
-     blake512_4way( &ctx, vhash, 64 );
+     blake512_4way_update( &ctx, vhash, 64 );
     blake512_4way_close( &ctx, vhash );

     memcpy( output,    hash0, 32 );
--- a/algo/bmw/bmw512-4way.c
+++ b/algo/bmw/bmw512-4way.c
@@ -40,7 +40,7 @@ int scanhash_bmw512_8way( struct work *work, uint32_t max_nonce,
      bmw512hash_8way( hash, vdata );

      for ( int lane = 0; lane < 8; lane++ )
-      if ( unlikely( hash7[ lane<<1 ] < Htarg ) )
+      if ( unlikely( hash7[ lane<<1 ] <= Htarg ) )
      {
          extr_lane_8x64( lane_hash, hash, lane, 256 );
          if ( fulltest( lane_hash, ptarget ) )
@@ -93,8 +93,7 @@ int scanhash_bmw512_4way( struct work *work, uint32_t max_nonce,
      bmw512hash_4way( hash, vdata );

      for ( int lane = 0; lane < 4; lane++ )
-      if ( unlikely( hash7[ lane<<1 ] < Htarg ) )
-//      if ( ( ( hash7[ lane<<1 ] & 0xFFFFFF00 ) == 0 ) )
+      if ( unlikely( hash7[ lane<<1 ] <= Htarg ) )
      {
          extr_lane_4x64( lane_hash, hash, lane, 256 );
          if ( fulltest( lane_hash, ptarget ) )
--- a/algo/cubehash/cube-hash-2way.c
+++ b/algo/cubehash/cube-hash-2way.c
@@ -168,6 +168,66 @@ int cube_4way_close( cube_4way_context *sp, void *output )
    return 0;
 }

+int cube_4way_full( cube_4way_context *sp, void *output,  int hashbitlen, 
+                    const void *data, size_t size )
+{
+    __m512i *h = (__m512i*)sp->h;
+    __m128i *iv = (__m128i*)( hashbitlen == 512 ? (__m128i*)IV512
+                                                : (__m128i*)IV256 );
+    sp->hashlen   = hashbitlen/128;
+    sp->blocksize = 32/16;
+    sp->rounds    = 16;
+    sp->pos       = 0;
+
+    h[ 0] = m512_const1_128( iv[0] );
+    h[ 1] = m512_const1_128( iv[1] );
+    h[ 2] = m512_const1_128( iv[2] );
+    h[ 3] = m512_const1_128( iv[3] );
+    h[ 4] = m512_const1_128( iv[4] );
+    h[ 5] = m512_const1_128( iv[5] );
+    h[ 6] = m512_const1_128( iv[6] );
+    h[ 7] = m512_const1_128( iv[7] );
+    h[ 0] = m512_const1_128( iv[0] );
+    h[ 1] = m512_const1_128( iv[1] );
+    h[ 2] = m512_const1_128( iv[2] );
+    h[ 3] = m512_const1_128( iv[3] );
+    h[ 4] = m512_const1_128( iv[4] );
+    h[ 5] = m512_const1_128( iv[5] );
+    h[ 6] = m512_const1_128( iv[6] );
+    h[ 7] = m512_const1_128( iv[7] );
+
+    const int len = size >> 4;
+    const __m512i *in = (__m512i*)data;
+    __m512i *hash = (__m512i*)output;
+    int i;
+
+    for ( i = 0; i < len; i++ )
+    {
+        sp->h[ sp->pos ] = _mm512_xor_si512( sp->h[ sp->pos ], in[i] );
+        sp->pos++;
+        if ( sp->pos == sp->blocksize )
+        {
+           transform_4way( sp );
+           sp->pos = 0;
+        }
+    }
+
+    // pos is zero for 64 byte data, 1 for 80 byte data.
+    sp->h[ sp->pos ] = _mm512_xor_si512( sp->h[ sp->pos ],
+                                    m512_const2_64( 0, 0x0000000000000080 ) );
+    transform_4way( sp );
+
+    sp->h[7] = _mm512_xor_si512( sp->h[7],
+                                    m512_const2_64( 0x0000000100000000, 0 ) );
+
+    for ( i = 0; i < 10; ++i )
+       transform_4way( sp );
+
+    memcpy( hash, sp->h, sp->hashlen<<6);
+    return 0;
+}
+
+
 int cube_4way_update_close( cube_4way_context *sp, void *output,
                               const void *data, size_t size )
 {
@@ -376,4 +436,62 @@ int cube_2way_update_close( cube_2way_context *sp, void *output,
    return 0;
 }

+int cube_2way_full( cube_2way_context *sp, void *output, int hashbitlen,
+                               const void *data, size_t size )
+{
+    __m256i *h = (__m256i*)sp->h;
+    __m128i *iv = (__m128i*)( hashbitlen == 512 ? (__m128i*)IV512
+                                                : (__m128i*)IV256 );
+    sp->hashlen   = hashbitlen/128;
+    sp->blocksize = 32/16;
+    sp->rounds    = 16;
+    sp->pos       = 0;
+
+    h[ 0] = m256_const1_128( iv[0] );
+    h[ 1] = m256_const1_128( iv[1] );
+    h[ 2] = m256_const1_128( iv[2] );
+    h[ 3] = m256_const1_128( iv[3] );
+    h[ 4] = m256_const1_128( iv[4] );
+    h[ 5] = m256_const1_128( iv[5] );
+    h[ 6] = m256_const1_128( iv[6] );
+    h[ 7] = m256_const1_128( iv[7] );
+    h[ 0] = m256_const1_128( iv[0] );
+    h[ 1] = m256_const1_128( iv[1] );
+    h[ 2] = m256_const1_128( iv[2] );
+    h[ 3] = m256_const1_128( iv[3] );
+    h[ 4] = m256_const1_128( iv[4] );
+    h[ 5] = m256_const1_128( iv[5] );
+    h[ 6] = m256_const1_128( iv[6] );
+    h[ 7] = m256_const1_128( iv[7] );
+
+    const int len = size >> 4;
+    const __m256i *in = (__m256i*)data;
+    __m256i *hash = (__m256i*)output;
+    int i;
+
+    for ( i = 0; i < len; i++ )
+    {
+        sp->h[ sp->pos ] = _mm256_xor_si256( sp->h[ sp->pos ], in[i] );
+        sp->pos++;
+        if ( sp->pos == sp->blocksize )
+        {
+           transform_2way( sp );
+           sp->pos = 0;
+        }
+    }
+
+    // pos is zero for 64 byte data, 1 for 80 byte data.
+    sp->h[ sp->pos ] = _mm256_xor_si256( sp->h[ sp->pos ],
+                                    m256_const2_64( 0, 0x0000000000000080 ) );
+    transform_2way( sp );
+
+    sp->h[7] = _mm256_xor_si256( sp->h[7],
+                                    m256_const2_64( 0x0000000100000000, 0 ) );
+
+    for ( i = 0; i < 10; ++i )    transform_2way( sp );
+
+    memcpy( hash, sp->h, sp->hashlen<<5 );
+    return 0;
+}
+
 #endif
--- a/algo/cubehash/cube-hash-2way.h
+++ b/algo/cubehash/cube-hash-2way.h
@@ -21,15 +21,12 @@ typedef struct _cube_4way_context cube_4way_context;

 int cube_4way_init( cube_4way_context* sp, int hashbitlen, int rounds,
                       int blockbytes );
-// reinitialize context with same parameters, much faster.
-int cube_4way_reinit( cube_4way_context *sp );
-
 int cube_4way_update( cube_4way_context *sp, const void *data, size_t size );
-
 int cube_4way_close( cube_4way_context *sp, void *output );
-
 int cube_4way_update_close( cube_4way_context *sp, void *output,
                            const void *data, size_t size );
+int cube_4way_full( cube_4way_context *sp, void *output, int hashbitlen,
+                    const void *data, size_t size );

 #endif

@@ -48,15 +45,12 @@ typedef struct _cube_2way_context cube_2way_context;

 int cube_2way_init( cube_2way_context* sp, int hashbitlen, int rounds,
                       int blockbytes );
-// reinitialize context with same parameters, much faster.
-int cube_2way_reinit( cube_2way_context *sp );
-
 int cube_2way_update( cube_2way_context *sp, const void *data, size_t size );
-
 int cube_2way_close( cube_2way_context *sp, void *output );
-
 int cube_2way_update_close( cube_2way_context *sp, void *output,
                            const void *data, size_t size );
+int cube_2way_full( cube_2way_context *sp, void *output, int hashbitlen,
+                    const void *data, size_t size );


 #endif
--- a/algo/echo/aes_ni/hash.c
+++ b/algo/echo/aes_ni/hash.c
@@ -7,7 +7,6 @@
 * - implements NIST hash api
 * - assumes that message lenght is multiple of 8-bits
 * - _ECHO_VPERM_ must be defined if compiling with ../main.c
- * -  define NO_AES_NI for aes_ni version
 *
 * Cagdas Calik
 * ccalik@metu.edu.tr
@@ -21,13 +20,7 @@
 #include "hash_api.h"
 //#include "vperm.h"
 #include <immintrin.h>
-/*
-#ifndef NO_AES_NI
-#include <wmmintrin.h>
-#else
-#include <tmmintrin.h>
-#endif
-*/
+#include "simd-utils.h"

 MYALIGN const unsigned int _k_s0F[] = {0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F};
 MYALIGN const unsigned int _k_ipt[] = {0x5A2A7000, 0xC2B2E898, 0x52227808, 0xCABAE090, 0x317C4D00, 0x4C01307D, 0xB0FDCC81, 0xCD80B1FC};
@@ -525,6 +518,165 @@ HashReturn update_final_echo( hashState_echo *state, BitSequence *hashval,
   return SUCCESS;
 }

+HashReturn echo_full( hashState_echo *state, BitSequence *hashval,
+            int nHashSize, const BitSequence *data, DataLength datalen )
+{
+   int i, j;
+
+   state->k = m128_zero;
+   state->processed_bits = 0;
+   state->uBufferBytes = 0;
+
+   switch( nHashSize )
+   {
+      case 256:
+         state->uHashSize = 256;
+         state->uBlockLength = 192;
+         state->uRounds = 8;
+         state->hashsize = m128_const_64( 0, 0x100 );
+         state->const1536 = m128_const_64( 0, 0x600 );
+         break;
+
+      case 512:
+         state->uHashSize = 512;
+         state->uBlockLength = 128;
+         state->uRounds = 10;
+         state->hashsize = m128_const_64( 0, 0x200 );
+         state->const1536 = m128_const_64( 0, 0x400 );
+         break;
+
+      default:
+         return BAD_HASHBITLEN;
+   }
+
+   for(i = 0; i < 4; i++)
+      for(j = 0; j < nHashSize / 256; j++)
+         state->state[i][j] = state->hashsize;
+
+   for(i = 0; i < 4; i++)
+      for(j = nHashSize / 256; j < 4; j++)
+         state->state[i][j] = m128_zero;
+
+
+   unsigned int uBlockCount, uRemainingBytes;
+
+   if( (state->uBufferBytes + datalen) >= state->uBlockLength )
+   {
+        if( state->uBufferBytes != 0 )
+        {
+           // Fill the buffer
+           memcpy( state->buffer + state->uBufferBytes,
+                   (void*)data, state->uBlockLength - state->uBufferBytes );
+
+           // Process buffer
+           Compress( state, state->buffer, 1 );
+           state->processed_bits += state->uBlockLength * 8;
+
+           data += state->uBlockLength - state->uBufferBytes;
+           datalen -= state->uBlockLength - state->uBufferBytes;
+        }
+
+        // buffer now does not contain any unprocessed bytes
+
+        uBlockCount = datalen / state->uBlockLength;
+        uRemainingBytes = datalen % state->uBlockLength;
+
+        if( uBlockCount > 0 )
+        {
+           Compress( state, data, uBlockCount );
+           state->processed_bits += uBlockCount * state->uBlockLength * 8;
+           data += uBlockCount * state->uBlockLength;
+        }
+
+        if( uRemainingBytes > 0 )
+        memcpy(state->buffer, (void*)data, uRemainingBytes);
+
+        state->uBufferBytes = uRemainingBytes;
+   }
+   else
+   {
+        memcpy( state->buffer + state->uBufferBytes, (void*)data, datalen );
+        state->uBufferBytes += datalen;
+   }
+
+   __m128i remainingbits;
+
+   // Add remaining bytes in the buffer
+   state->processed_bits += state->uBufferBytes * 8;
+
+   remainingbits = _mm_set_epi32( 0, 0, 0, state->uBufferBytes * 8 );
+
+   // Pad with 0x80
+   state->buffer[state->uBufferBytes++] = 0x80;
+   // Enough buffer space for padding in this block?
+   if( (state->uBlockLength - state->uBufferBytes) >= 18 )
+   {
+        // Pad with zeros
+        memset( state->buffer + state->uBufferBytes, 0, state->uBlockLength - (state->uBufferBytes + 18) );
+
+        // Hash size
+        *( (unsigned short*)(state->buffer + state->uBlockLength - 18) ) = state->uHashSize;
+
+        // Processed bits
+        *( (DataLength*)(state->buffer + state->uBlockLength - 16) ) =
+                   state->processed_bits;
+        *( (DataLength*)(state->buffer + state->uBlockLength - 8) ) = 0;
+
+        // Last block contains message bits?
+        if( state->uBufferBytes == 1 )
+        {
+           state->k = _mm_xor_si128( state->k, state->k );
+           state->k = _mm_sub_epi64( state->k, state->const1536 );
+        }
+        else
+        {
+           state->k = _mm_add_epi64( state->k, remainingbits );
+           state->k = _mm_sub_epi64( state->k, state->const1536 );
+        }
+
+        // Compress
+        Compress( state, state->buffer, 1 );
+   }
+   else
+   {
+        // Fill with zero and compress
+        memset( state->buffer + state->uBufferBytes, 0,
+                state->uBlockLength - state->uBufferBytes );
+        state->k = _mm_add_epi64( state->k, remainingbits );
+        state->k = _mm_sub_epi64( state->k, state->const1536 );
+        Compress( state, state->buffer, 1 );
+
+        // Last block
+        memset( state->buffer, 0, state->uBlockLength - 18 );
+
+        // Hash size
+        *( (unsigned short*)(state->buffer + state->uBlockLength - 18) ) =
+                 state->uHashSize;
+
+        // Processed bits
+        *( (DataLength*)(state->buffer + state->uBlockLength - 16) ) =
+                   state->processed_bits;
+        *( (DataLength*)(state->buffer + state->uBlockLength - 8) ) = 0;
+        // Compress the last block
+        state->k = _mm_xor_si128( state->k, state->k );
+        state->k = _mm_sub_epi64( state->k, state->const1536 );
+        Compress( state, state->buffer, 1) ;
+   }
+
+   // Store the hash value
+   _mm_store_si128( (__m128i*)hashval + 0, state->state[0][0] );
+   _mm_store_si128( (__m128i*)hashval + 1, state->state[1][0] );
+
+   if( state->uHashSize == 512 )
+   {
+        _mm_store_si128( (__m128i*)hashval + 2, state->state[2][0] );
+        _mm_store_si128( (__m128i*)hashval + 3, state->state[3][0] );
+
+   }
+   return SUCCESS;
+}
+
+

 HashReturn hash_echo(int hashbitlen, const BitSequence *data, DataLength databitlen, BitSequence *hashval)
 {
--- a/algo/echo/aes_ni/hash.c.test
+++ b/algo/echo/aes_ni/hash.c.test
@@ -1,620 +0,0 @@
-/*
- * file        : echo_vperm.c
- * version     : 1.0.208
- * date        : 14.12.2010
- * 
- * - vperm and aes_ni implementations of hash function ECHO
- * - implements NIST hash api
- * - assumes that message lenght is multiple of 8-bits
- * - _ECHO_VPERM_ must be defined if compiling with ../main.c
- * -  define NO_AES_NI for aes_ni version
- *
- * Cagdas Calik
- * ccalik@metu.edu.tr
- * Institute of Applied Mathematics, Middle East Technical University, Turkey.
- *
- */
-#if defined(__AES__)
-
-#include <memory.h>
-#include "miner.h"
-#include "hash_api.h"
-//#include "vperm.h"
-#include <immintrin.h>
-/*
-#ifndef NO_AES_NI
-#include <wmmintrin.h>
-#else
-#include <tmmintrin.h>
-#endif
-*/
-
-MYALIGN const unsigned int _k_s0F[] = {0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F, 0x0F0F0F0F};
-MYALIGN const unsigned int _k_ipt[] = {0x5A2A7000, 0xC2B2E898, 0x52227808, 0xCABAE090, 0x317C4D00, 0x4C01307D, 0xB0FDCC81, 0xCD80B1FC};
-MYALIGN const unsigned int _k_opt[] = {0xD6B66000, 0xFF9F4929, 0xDEBE6808, 0xF7974121, 0x50BCEC00, 0x01EDBD51, 0xB05C0CE0, 0xE10D5DB1};
-MYALIGN const unsigned int _k_inv[] = {0x0D080180, 0x0E05060F, 0x0A0B0C02, 0x04070309, 0x0F0B0780, 0x01040A06, 0x02050809, 0x030D0E0C};
-MYALIGN const unsigned int _k_sb1[] = {0xCB503E00, 0xB19BE18F, 0x142AF544, 0xA5DF7A6E, 0xFAE22300, 0x3618D415, 0x0D2ED9EF, 0x3BF7CCC1};
-MYALIGN const unsigned int _k_sb2[] = {0x0B712400, 0xE27A93C6, 0xBC982FCD, 0x5EB7E955, 0x0AE12900, 0x69EB8840, 0xAB82234A, 0xC2A163C8};
-MYALIGN const unsigned int _k_sb3[] = {0xC0211A00, 0x53E17249, 0xA8B2DA89, 0xFB68933B, 0xF0030A00, 0x5FF35C55, 0xA6ACFAA5, 0xF956AF09};
-MYALIGN const unsigned int _k_sb4[] = {0x3FD64100, 0xE1E937A0, 0x49087E9F, 0xA876DE97, 0xC393EA00, 0x3D50AED7, 0x876D2914, 0xBA44FE79};
-MYALIGN const unsigned int _k_sb5[] = {0xF4867F00, 0x5072D62F, 0x5D228BDB, 0x0DA9A4F9, 0x3971C900, 0x0B487AC2, 0x8A43F0FB, 0x81B332B8};
-MYALIGN const unsigned int _k_sb7[] = {0xFFF75B00, 0xB20845E9, 0xE1BAA416, 0x531E4DAC, 0x3390E000, 0x62A3F282, 0x21C1D3B1, 0x43125170};
-MYALIGN const unsigned int _k_sbo[] = {0x6FBDC700, 0xD0D26D17, 0xC502A878, 0x15AABF7A, 0x5FBB6A00, 0xCFE474A5, 0x412B35FA, 0x8E1E90D1};
-MYALIGN const unsigned int _k_h63[] = {0x63636363, 0x63636363, 0x63636363, 0x63636363};
-MYALIGN const unsigned int _k_hc6[] = {0xc6c6c6c6, 0xc6c6c6c6, 0xc6c6c6c6, 0xc6c6c6c6};
-MYALIGN const unsigned int _k_h5b[] = {0x5b5b5b5b, 0x5b5b5b5b, 0x5b5b5b5b, 0x5b5b5b5b};
-MYALIGN const unsigned int _k_h4e[] = {0x4e4e4e4e, 0x4e4e4e4e, 0x4e4e4e4e, 0x4e4e4e4e};
-MYALIGN const unsigned int _k_h0e[] = {0x0e0e0e0e, 0x0e0e0e0e, 0x0e0e0e0e, 0x0e0e0e0e};
-MYALIGN const unsigned int _k_h15[] = {0x15151515, 0x15151515, 0x15151515, 0x15151515};
-MYALIGN const unsigned int _k_aesmix1[] = {0x0f0a0500, 0x030e0904, 0x07020d08, 0x0b06010c};
-MYALIGN const unsigned int _k_aesmix2[] = {0x000f0a05, 0x04030e09, 0x0807020d, 0x0c0b0601};
-MYALIGN const unsigned int _k_aesmix3[] = {0x05000f0a, 0x0904030e, 0x0d080702, 0x010c0b06};
-MYALIGN const unsigned int _k_aesmix4[] = {0x0a05000f, 0x0e090403, 0x020d0807, 0x06010c0b};
-
-
-MYALIGN const unsigned int 	const1[]		= {0x00000001, 0x00000000, 0x00000000, 0x00000000};
-MYALIGN const unsigned int	mul2mask[]		= {0x00001b00, 0x00000000, 0x00000000, 0x00000000};
-MYALIGN const unsigned int	lsbmask[]		= {0x01010101, 0x01010101, 0x01010101, 0x01010101};
-MYALIGN const unsigned int	invshiftrows[]	= {0x070a0d00, 0x0b0e0104, 0x0f020508, 0x0306090c};
-MYALIGN const unsigned int	zero[]			= {0x00000000, 0x00000000, 0x00000000, 0x00000000};
-MYALIGN const unsigned int	mul2ipt[]		= {0x728efc00, 0x6894e61a, 0x3fc3b14d, 0x25d9ab57, 0xfd5ba600, 0x2a8c71d7, 0x1eb845e3, 0xc96f9234};
-
-
-#define ECHO_SUBBYTES(state, i, j) \
-	state[i][j] = _mm_aesenc_si128(state[i][j], k1);\
-	state[i][j] = _mm_aesenc_si128(state[i][j], M128(zero));\
-	k1 = _mm_add_epi32(k1, M128(const1))
-
-#define ECHO_MIXBYTES(state1, state2, j, t1, t2, s2) \
-	s2 = _mm_add_epi8(state1[0][j], state1[0][j]);\
-	t1 = _mm_srli_epi16(state1[0][j], 7);\
-	t1 = _mm_and_si128(t1, M128(lsbmask));\
-	t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
-	s2 = _mm_xor_si128(s2, t2);\
-	state2[0][j] = s2;\
-	state2[1][j] = state1[0][j];\
-	state2[2][j] = state1[0][j];\
-	state2[3][j] = _mm_xor_si128(s2, state1[0][j]);\
-	s2 = _mm_add_epi8(state1[1][(j + 1) & 3], state1[1][(j + 1) & 3]);\
-	t1 = _mm_srli_epi16(state1[1][(j + 1) & 3], 7);\
-	t1 = _mm_and_si128(t1, M128(lsbmask));\
-	t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
-	s2 = _mm_xor_si128(s2, t2);\
-	state2[0][j] = _mm_xor_si128(state2[0][j], _mm_xor_si128(s2, state1[1][(j + 1) & 3]));\
-	state2[1][j] = _mm_xor_si128(state2[1][j], s2);\
-	state2[2][j] = _mm_xor_si128(state2[2][j], state1[1][(j + 1) & 3]);\
-	state2[3][j] = _mm_xor_si128(state2[3][j], state1[1][(j + 1) & 3]);\
-	s2 = _mm_add_epi8(state1[2][(j + 2) & 3], state1[2][(j + 2) & 3]);\
-	t1 = _mm_srli_epi16(state1[2][(j + 2) & 3], 7);\
-	t1 = _mm_and_si128(t1, M128(lsbmask));\
-	t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
-	s2 = _mm_xor_si128(s2, t2);\
-	state2[0][j] = _mm_xor_si128(state2[0][j], state1[2][(j + 2) & 3]);\
-	state2[1][j] = _mm_xor_si128(state2[1][j], _mm_xor_si128(s2, state1[2][(j + 2) & 3]));\
-	state2[2][j] = _mm_xor_si128(state2[2][j], s2);\
-	state2[3][j] = _mm_xor_si128(state2[3][j], state1[2][(j + 2) & 3]);\
-	s2 = _mm_add_epi8(state1[3][(j + 3) & 3], state1[3][(j + 3) & 3]);\
-	t1 = _mm_srli_epi16(state1[3][(j + 3) & 3], 7);\
-	t1 = _mm_and_si128(t1, M128(lsbmask));\
-	t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
-	s2 = _mm_xor_si128(s2, t2);\
-	state2[0][j] = _mm_xor_si128(state2[0][j], state1[3][(j + 3) & 3]);\
-	state2[1][j] = _mm_xor_si128(state2[1][j], state1[3][(j + 3) & 3]);\
-	state2[2][j] = _mm_xor_si128(state2[2][j], _mm_xor_si128(s2, state1[3][(j + 3) & 3]));\
-	state2[3][j] = _mm_xor_si128(state2[3][j], s2)
-
-
-#define ECHO_ROUND_UNROLL2 \
-	ECHO_SUBBYTES(_state, 0, 0);\
-	ECHO_SUBBYTES(_state, 1, 0);\
-	ECHO_SUBBYTES(_state, 2, 0);\
-	ECHO_SUBBYTES(_state, 3, 0);\
-	ECHO_SUBBYTES(_state, 0, 1);\
-	ECHO_SUBBYTES(_state, 1, 1);\
-	ECHO_SUBBYTES(_state, 2, 1);\
-	ECHO_SUBBYTES(_state, 3, 1);\
-	ECHO_SUBBYTES(_state, 0, 2);\
-	ECHO_SUBBYTES(_state, 1, 2);\
-	ECHO_SUBBYTES(_state, 2, 2);\
-	ECHO_SUBBYTES(_state, 3, 2);\
-	ECHO_SUBBYTES(_state, 0, 3);\
-	ECHO_SUBBYTES(_state, 1, 3);\
-	ECHO_SUBBYTES(_state, 2, 3);\
-	ECHO_SUBBYTES(_state, 3, 3);\
-	ECHO_MIXBYTES(_state, _state2, 0, t1, t2, s2);\
-	ECHO_MIXBYTES(_state, _state2, 1, t1, t2, s2);\
-	ECHO_MIXBYTES(_state, _state2, 2, t1, t2, s2);\
-	ECHO_MIXBYTES(_state, _state2, 3, t1, t2, s2);\
-	ECHO_SUBBYTES(_state2, 0, 0);\
-	ECHO_SUBBYTES(_state2, 1, 0);\
-	ECHO_SUBBYTES(_state2, 2, 0);\
-	ECHO_SUBBYTES(_state2, 3, 0);\
-	ECHO_SUBBYTES(_state2, 0, 1);\
-	ECHO_SUBBYTES(_state2, 1, 1);\
-	ECHO_SUBBYTES(_state2, 2, 1);\
-	ECHO_SUBBYTES(_state2, 3, 1);\
-	ECHO_SUBBYTES(_state2, 0, 2);\
-	ECHO_SUBBYTES(_state2, 1, 2);\
-	ECHO_SUBBYTES(_state2, 2, 2);\
-	ECHO_SUBBYTES(_state2, 3, 2);\
-	ECHO_SUBBYTES(_state2, 0, 3);\
-	ECHO_SUBBYTES(_state2, 1, 3);\
-	ECHO_SUBBYTES(_state2, 2, 3);\
-	ECHO_SUBBYTES(_state2, 3, 3);\
-	ECHO_MIXBYTES(_state2, _state, 0, t1, t2, s2);\
-	ECHO_MIXBYTES(_state2, _state, 1, t1, t2, s2);\
-	ECHO_MIXBYTES(_state2, _state, 2, t1, t2, s2);\
-	ECHO_MIXBYTES(_state2, _state, 3, t1, t2, s2)
-
-
-
-#define SAVESTATE(dst, src)\
-	dst[0][0] = src[0][0];\
-	dst[0][1] = src[0][1];\
-	dst[0][2] = src[0][2];\
-	dst[0][3] = src[0][3];\
-	dst[1][0] = src[1][0];\
-	dst[1][1] = src[1][1];\
-	dst[1][2] = src[1][2];\
-	dst[1][3] = src[1][3];\
-	dst[2][0] = src[2][0];\
-	dst[2][1] = src[2][1];\
-	dst[2][2] = src[2][2];\
-	dst[2][3] = src[2][3];\
-	dst[3][0] = src[3][0];\
-	dst[3][1] = src[3][1];\
-	dst[3][2] = src[3][2];\
-	dst[3][3] = src[3][3]
-
-
-void Compress(hashState_echo *ctx, const unsigned char *pmsg, unsigned int uBlockCount)
-{
-   unsigned int r, b, i, j;
-   __m128i t1, t2, s2, k1;
-   __m128i _state[4][4], _state2[4][4], _statebackup[4][4]; 
-
-   for(i = 0; i < 4; i++)
-	for(j = 0; j < ctx->uHashSize / 256; j++)
-		_state[i][j] = ctx->state[i][j];
-
-   for(b = 0; b < uBlockCount; b++)
-   {
-	ctx->k = _mm_add_epi64(ctx->k, ctx->const1536);
-
-	// load message
-	for(j = ctx->uHashSize / 256; j < 4; j++)
-	{
-	   for(i = 0; i < 4; i++)
-	   {
-		_state[i][j] = _mm_loadu_si128((__m128i*)pmsg + 4 * (j - (ctx->uHashSize / 256)) + i);
-	   }
-	}
-
-uint64_t *b = (uint64_t*)_state;
-//printf("Ss3: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
-   
-	// save state
-	SAVESTATE(_statebackup, _state);
-
-	k1 = ctx->k;
-
-	for(r = 0; r < ctx->uRounds / 2; r++)
-	{
-		ECHO_ROUND_UNROLL2;
-	}
-
-//printf("Ss4: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
-   
-   
-	if(ctx->uHashSize == 256)
-	{
-	   for(i = 0; i < 4; i++)
-	   {
-		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][1]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][3]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][1]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][3]);
-	   }
-	}
-	else
-	{
-	   for(i = 0; i < 4; i++)
-	   {
-		_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
-		_state[i][1] = _mm_xor_si128(_state[i][1], _state[i][3]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
-		_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
-		_state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][1]);
-		_state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][3]);
-           }
-	}
-	pmsg += ctx->uBlockLength;
-   }
-	SAVESTATE(ctx->state, _state);
-
-}
-
-
-
-HashReturn init_echo(hashState_echo *ctx, int nHashSize)
-{
-	int i, j;
-
-        ctx->k = _mm_setzero_si128(); 
-	ctx->processed_bits = 0;
-	ctx->uBufferBytes = 0;
-
-	switch(nHashSize)
-	{
-		case 256:
-			ctx->uHashSize = 256;
-			ctx->uBlockLength = 192;
-			ctx->uRounds = 8;
-			ctx->hashsize = _mm_set_epi32(0, 0, 0, 0x00000100);
-			ctx->const1536 = _mm_set_epi32(0x00000000, 0x00000000, 0x00000000, 0x00000600);
-			break;
-
-		case 512:
-			ctx->uHashSize = 512;
-			ctx->uBlockLength = 128;
-			ctx->uRounds = 10;
-			ctx->hashsize = _mm_set_epi32(0, 0, 0, 0x00000200);
-			ctx->const1536 = _mm_set_epi32(0x00000000, 0x00000000, 0x00000000, 0x00000400);
-			break;
-
-		default:
-			return BAD_HASHBITLEN;
-	}
-
-
-	for(i = 0; i < 4; i++)
-		for(j = 0; j < nHashSize / 256; j++)
-			ctx->state[i][j] = ctx->hashsize;
-
-	for(i = 0; i < 4; i++)
-		for(j = nHashSize / 256; j < 4; j++)
-			ctx->state[i][j] = _mm_set_epi32(0, 0, 0, 0);
-
-	return SUCCESS;
-}
-
-HashReturn update_echo(hashState_echo *state, const BitSequence *data, DataLength databitlen)
-{
-	unsigned int uByteLength, uBlockCount, uRemainingBytes;
-
-	uByteLength = (unsigned int)(databitlen / 8);
-
-	if((state->uBufferBytes + uByteLength) >= state->uBlockLength)
-	{
-		if(state->uBufferBytes != 0)
-		{
-			// Fill the buffer
-			memcpy(state->buffer + state->uBufferBytes, (void*)data, state->uBlockLength - state->uBufferBytes);
-
-			// Process buffer
-			Compress(state, state->buffer, 1);
-			state->processed_bits += state->uBlockLength * 8;
-
-			data += state->uBlockLength - state->uBufferBytes;
-			uByteLength -= state->uBlockLength - state->uBufferBytes;
-		}
-
-		// buffer now does not contain any unprocessed bytes
-
-		uBlockCount = uByteLength / state->uBlockLength;
-		uRemainingBytes = uByteLength % state->uBlockLength;
-
-		if(uBlockCount > 0)
-		{
-			Compress(state, data, uBlockCount);
-
-			state->processed_bits += uBlockCount * state->uBlockLength * 8;
-			data += uBlockCount * state->uBlockLength;
-		}
-
-		if(uRemainingBytes > 0)
-		{
-			memcpy(state->buffer, (void*)data, uRemainingBytes);
-		}
-
-		state->uBufferBytes = uRemainingBytes;
-	}
-	else
-	{
-		memcpy(state->buffer + state->uBufferBytes, (void*)data, uByteLength);
-		state->uBufferBytes += uByteLength;
-	}
-
-	return SUCCESS;
-}
-
-HashReturn final_echo(hashState_echo *state, BitSequence *hashval)
-{
-	__m128i remainingbits;
-
-	// Add remaining bytes in the buffer
-	state->processed_bits += state->uBufferBytes * 8;
-
-	remainingbits = _mm_set_epi32(0, 0, 0, state->uBufferBytes * 8);
-
-	// Pad with 0x80
-	state->buffer[state->uBufferBytes++] = 0x80;
-	
-	// Enough buffer space for padding in this block?
-	if((state->uBlockLength - state->uBufferBytes) >= 18)
-	{
-		// Pad with zeros
-		memset(state->buffer + state->uBufferBytes, 0, state->uBlockLength - (state->uBufferBytes + 18));
-
-		// Hash size
-		*((unsigned short*)(state->buffer + state->uBlockLength - 18)) = state->uHashSize;
-
-		// Processed bits
-		*((DataLength*)(state->buffer + state->uBlockLength - 16)) = state->processed_bits;
-		*((DataLength*)(state->buffer + state->uBlockLength - 8)) = 0;
-
-		// Last block contains message bits?
-		if(state->uBufferBytes == 1)
-		{
-			state->k = _mm_xor_si128(state->k, state->k);
-			state->k = _mm_sub_epi64(state->k, state->const1536);
-		}
-		else
-		{
-			state->k = _mm_add_epi64(state->k, remainingbits);
-			state->k = _mm_sub_epi64(state->k, state->const1536);
-		}
-
-		// Compress
-		Compress(state, state->buffer, 1);
-	}
-	else
-	{
-		// Fill with zero and compress
-		memset(state->buffer + state->uBufferBytes, 0, state->uBlockLength - state->uBufferBytes);
-		state->k = _mm_add_epi64(state->k, remainingbits);
-		state->k = _mm_sub_epi64(state->k, state->const1536);
-		Compress(state, state->buffer, 1);
-
-		// Last block
-		memset(state->buffer, 0, state->uBlockLength - 18);
-
-		// Hash size
-		*((unsigned short*)(state->buffer + state->uBlockLength - 18)) = state->uHashSize;
-
-		// Processed bits
-		*((DataLength*)(state->buffer + state->uBlockLength - 16)) = state->processed_bits;
-		*((DataLength*)(state->buffer + state->uBlockLength - 8)) = 0;
-
-		// Compress the last block
-		state->k = _mm_xor_si128(state->k, state->k);
-		state->k = _mm_sub_epi64(state->k, state->const1536);
-		Compress(state, state->buffer, 1);
-	}
-
-	// Store the hash value
-	_mm_storeu_si128((__m128i*)hashval + 0, state->state[0][0]);
-	_mm_storeu_si128((__m128i*)hashval + 1, state->state[1][0]);
-
-	if(state->uHashSize == 512)
-	{
-		_mm_storeu_si128((__m128i*)hashval + 2, state->state[2][0]);
-		_mm_storeu_si128((__m128i*)hashval + 3, state->state[3][0]);
-	}
-
-	return SUCCESS;
-}
-
-HashReturn update_final_echo( hashState_echo *state, BitSequence *hashval,
-                              const BitSequence *data, DataLength databitlen )
-{
-   unsigned int uByteLength, uBlockCount, uRemainingBytes;
-
-   uByteLength = (unsigned int)(databitlen / 8);
-
-/*   
-   if( (state->uBufferBytes + uByteLength) >= state->uBlockLength )
-   {
-printf("full block\n");
-      if( state->uBufferBytes != 0 )
-        {
-           // Fill the buffer
-           memcpy( state->buffer + state->uBufferBytes,
-                   (void*)data, state->uBlockLength - state->uBufferBytes );
-
-           // Process buffer
-           Compress( state, state->buffer, 1 );
-           state->processed_bits += state->uBlockLength * 8;
-
-           data += state->uBlockLength - state->uBufferBytes;
-           uByteLength -= state->uBlockLength - state->uBufferBytes;
-        }
-
-        // buffer now does not contain any unprocessed bytes
-
-        uBlockCount = uByteLength / state->uBlockLength;
-        uRemainingBytes = uByteLength % state->uBlockLength;
-
-        if( uBlockCount > 0 )
-        {
-           Compress( state, data, uBlockCount );
-           state->processed_bits += uBlockCount * state->uBlockLength * 8;
-           data += uBlockCount * state->uBlockLength;
-        }
-
-        if( uRemainingBytes > 0 )
-        memcpy(state->buffer, (void*)data, uRemainingBytes);
-
-        state->uBufferBytes = uRemainingBytes;
-   }
-   else
-   {
-*/
-   memcpy( state->buffer + state->uBufferBytes, (void*)data, uByteLength );
-        state->uBufferBytes += uByteLength;
-//   }
-
-   __m128i remainingbits;
-
-   // Add remaining bytes in the buffer
-   state->processed_bits += state->uBufferBytes * 8;
-
-   remainingbits = _mm_set_epi32( 0, 0, 0, state->uBufferBytes * 8 );
-
-   // Pad with 0x80
-   state->buffer[state->uBufferBytes++] = 0x80;
-
-   // Enough buffer space for padding in this block?
-
-//   if( (state->uBlockLength - state->uBufferBytes) >= 18 )
-//   {
-        // Pad with zeros
-
-        memset( state->buffer + state->uBufferBytes, 0, state->uBlockLength - (state->uBufferBytes + 18) );
-
-        // Hash size
-        *( (unsigned short*)(state->buffer + state->uBlockLength - 18) ) = state->uHashSize;
-
-        // Processed bits
-        *( (DataLength*)(state->buffer + state->uBlockLength - 16) ) =
-                   state->processed_bits;
-        *( (DataLength*)(state->buffer + state->uBlockLength - 8) ) = 0;
-
-
-        // Last block contains message bits?
-        if( state->uBufferBytes == 1 )
-        {
-           state->k = _mm_xor_si128( state->k, state->k );
-           state->k = _mm_sub_epi64( state->k, state->const1536 );
-        }
-        else
-        {
-           state->k = _mm_add_epi64( state->k, remainingbits );
-           state->k = _mm_sub_epi64( state->k, state->const1536 );
-        }
-
-uint64_t *b = (uint64_t*)&state->k;
-/*
-printf("Sk: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
-b = (uint64_t*)state->buffer;
-printf("Sb: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
-printf("Sb: %016lx %016lx %016lx %016lx\n",b[4],b[5],b[6],b[7]);
-printf("Sb: %016lx %016lx %016lx %016lx\n",b[8],b[9],b[10],b[11]);
-printf("Sb: %016lx %016lx %016lx %016lx\n",b[12],b[13],b[14],b[15]);
-
-b = (uint64_t*)state->state;
-printf("Ss1: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
-printf("Ss1: %016lx %016lx %016lx %016lx\n",b[4],b[5],b[6],b[7]);
-printf("Ss1: %016lx %016lx %016lx %016lx\n",b[8],b[9],b[10],b[11]);
-printf("Ss1: %016lx %016lx %016lx %016lx\n",b[12],b[13],b[14],b[15]);
-*/        
-        // Compress
-        Compress( state, state->buffer, 1 );
-
-//printf("Ss2: %016lx %016lx %016lx %016lx\n",b[0],b[1],b[2],b[3]);
-
-        
-/*
-   }
-   else
-   {
-        // Fill with zero and compress
-        memset( state->buffer + state->uBufferBytes, 0,
-                state->uBlockLength - state->uBufferBytes );
-        state->k = _mm_add_epi64( state->k, remainingbits );
-        state->k = _mm_sub_epi64( state->k, state->const1536 );
-        Compress( state, state->buffer, 1 );
-
-        // Last block
-        memset( state->buffer, 0, state->uBlockLength - 18 );
-
-        // Hash size
-        *( (unsigned short*)(state->buffer + state->uBlockLength - 18) ) =
-                 state->uHashSize;
-
-        // Processed bits
-        *( (DataLength*)(state->buffer + state->uBlockLength - 16) ) =
-                   state->processed_bits;
-        *( (DataLength*)(state->buffer + state->uBlockLength - 8) ) = 0;
-        // Compress the last block
-        state->k = _mm_xor_si128( state->k, state->k );
-        state->k = _mm_sub_epi64( state->k, state->const1536 );
-        Compress( state, state->buffer, 1) ;
-   }
-*/
-
-   // Store the hash value
-   _mm_storeu_si128( (__m128i*)hashval + 0, state->state[0][0] );
-   _mm_storeu_si128( (__m128i*)hashval + 1, state->state[1][0] );
-
-   if( state->uHashSize == 512 )
-   {
-        _mm_storeu_si128( (__m128i*)hashval + 2, state->state[2][0] );
-        _mm_storeu_si128( (__m128i*)hashval + 3, state->state[3][0] );
-
-   }
-   return SUCCESS;
-}
-
-
-HashReturn hash_echo(int hashbitlen, const BitSequence *data, DataLength databitlen, BitSequence *hashval)
-{
-	HashReturn hRet;
-	hashState_echo hs;
-
-	/////
-	/*
-	__m128i a, b, c, d, t[4], u[4], v[4];
-
-	a = _mm_set_epi32(0x0f0e0d0c, 0x0b0a0908, 0x07060504, 0x03020100);
-	b = _mm_set_epi32(0x1f1e1d1c, 0x1b1a1918, 0x17161514, 0x13121110);
-	c = _mm_set_epi32(0x2f2e2d2c, 0x2b2a2928, 0x27262524, 0x23222120);
-	d = _mm_set_epi32(0x3f3e3d3c, 0x3b3a3938, 0x37363534, 0x33323130);
-
-	t[0] = _mm_unpacklo_epi8(a, b);
-	t[1] = _mm_unpackhi_epi8(a, b);
-	t[2] = _mm_unpacklo_epi8(c, d);
-	t[3] = _mm_unpackhi_epi8(c, d);
-
-	u[0] = _mm_unpacklo_epi16(t[0], t[2]);
-	u[1] = _mm_unpackhi_epi16(t[0], t[2]);
-	u[2] = _mm_unpacklo_epi16(t[1], t[3]);
-	u[3] = _mm_unpackhi_epi16(t[1], t[3]);
-
-
-	t[0] = _mm_unpacklo_epi16(u[0], u[1]);
-	t[1] = _mm_unpackhi_epi16(u[0], u[1]);
-	t[2] = _mm_unpacklo_epi16(u[2], u[3]);
-	t[3] = _mm_unpackhi_epi16(u[2], u[3]);
-
-	u[0] = _mm_unpacklo_epi8(t[0], t[1]);
-	u[1] = _mm_unpackhi_epi8(t[0], t[1]);
-	u[2] = _mm_unpacklo_epi8(t[2], t[3]);
-	u[3] = _mm_unpackhi_epi8(t[2], t[3]);
-
-	a = _mm_unpacklo_epi8(u[0], u[1]);
-	b = _mm_unpackhi_epi8(u[0], u[1]);
-	c = _mm_unpacklo_epi8(u[2], u[3]);
-	d = _mm_unpackhi_epi8(u[2], u[3]);
-	*/
-	/////
-
-	hRet = init_echo(&hs, hashbitlen);
-	if(hRet != SUCCESS)
-		return hRet;
-
-	hRet = update_echo(&hs, data, databitlen);
-	if(hRet != SUCCESS)
-		return hRet;
-
-	hRet = final_echo(&hs, hashval);
-	if(hRet != SUCCESS)
-		return hRet;
-
-	return SUCCESS;
-}
-
-#endif
--- a/algo/echo/aes_ni/hash_api.h
+++ b/algo/echo/aes_ni/hash_api.h
@@ -15,7 +15,7 @@
 #ifndef HASH_API_H
 #define HASH_API_H

-#ifndef NO_AES_NI
+#ifdef __AES__
 #define HASH_IMPL_STR	"ECHO-aesni"
 #else
 #define HASH_IMPL_STR	"ECHO-vperm"
@@ -55,6 +55,8 @@ HashReturn hash_echo(int hashbitlen, const BitSequence *data, DataLength databit

 HashReturn update_final_echo( hashState_echo *state, BitSequence *hashval,
                              const BitSequence *data, DataLength databitlen );
+HashReturn echo_full( hashState_echo *state, BitSequence *hashval,
+            int nHashSize, const BitSequence *data, DataLength databitlen );

 #endif // HASH_API_H

--- a/algo/echo/echo-hash-4way.c
+++ b/algo/echo/echo-hash-4way.c
@@ -313,4 +313,92 @@ int echo_4way_update_close( echo_4way_context *state, void *hashval,
   return 0;
 }

+int echo_4way_full( echo_4way_context *ctx, void *hashval, int nHashSize, 
+                    const void *data, int datalen )
+{
+   int i, j;
+   int databitlen = datalen * 8;
+   ctx->k = m512_zero;
+   ctx->processed_bits = 0;
+   ctx->uBufferBytes = 0;
+
+   switch( nHashSize )
+   {
+      case 256:
+         ctx->uHashSize = 256;
+         ctx->uBlockLength = 192;
+         ctx->uRounds = 8;
+         ctx->hashsize = _mm512_set4_epi32( 0, 0, 0, 0x100 );
+         ctx->const1536 = _mm512_set4_epi32( 0, 0, 0, 0x600 );
+         break;
+
+      case 512:
+         ctx->uHashSize = 512;
+         ctx->uBlockLength = 128;
+         ctx->uRounds = 10;
+         ctx->hashsize = _mm512_set4_epi32( 0, 0, 0, 0x200 );
+         ctx->const1536 = _mm512_set4_epi32( 0, 0, 0, 0x400);
+         break;
+
+      default:
+         return 1;
+   }
+
+   for( i = 0; i < 4; i++ )
+      for( j = 0; j < nHashSize / 256; j++ )
+         ctx->state[ i ][ j ] = ctx->hashsize;
+
+   for( i = 0; i < 4; i++ )
+      for( j = nHashSize / 256; j < 4; j++ )
+         ctx->state[ i ][ j ] = m512_zero;
+
+   
+// bytelen is either 32 (maybe), 64 or 80 or 128!
+// all are less than full block.
+
+   int vlen = datalen / 32;  
+   const int vblen = ctx->uBlockLength / 16; //  16 bytes per lane
+   __m512i remainingbits;
+
+   if ( databitlen == 1024 )
+   {
+      echo_4way_compress( ctx, data, 1 );
+      ctx->processed_bits = 1024;
+      remainingbits = m512_const2_64( 0, -1024 );
+      vlen = 0;
+   }
+   else
+   {
+      vlen = databitlen / 128;  // * 4 lanes / 128 bits per lane
+      memcpy_512( ctx->buffer, data, vlen );
+      ctx->processed_bits += (unsigned int)( databitlen );
+      remainingbits = _mm512_set4_epi32( 0, 0, 0, databitlen );
+
+   }
+
+   ctx->buffer[ vlen ] = _mm512_set4_epi32( 0, 0, 0, 0x80 );
+   memset_zero_512( ctx->buffer + vlen + 1, vblen - vlen - 2 );
+   ctx->buffer[ vblen-2 ] =
+                _mm512_set4_epi32( (uint32_t)ctx->uHashSize << 16, 0, 0, 0 );
+   ctx->buffer[ vblen-1 ] =
+                   _mm512_set4_epi64( 0, ctx->processed_bits,
+                                      0, ctx->processed_bits );
+
+   ctx->k = _mm512_add_epi64( ctx->k, remainingbits );
+   ctx->k = _mm512_sub_epi64( ctx->k, ctx->const1536 );
+
+   echo_4way_compress( ctx, ctx->buffer, 1 );
+
+   _mm512_store_si512( (__m512i*)hashval + 0, ctx->state[ 0 ][ 0] );
+   _mm512_store_si512( (__m512i*)hashval + 1, ctx->state[ 1 ][ 0] );
+
+   if ( ctx->uHashSize == 512 )
+   {
+      _mm512_store_si512( (__m512i*)hashval + 2, ctx->state[ 2 ][ 0 ] );
+      _mm512_store_si512( (__m512i*)hashval + 3, ctx->state[ 3 ][ 0 ] );
+   }
+   return 0;
+}
+
+
 #endif
--- a/algo/echo/echo-hash-4way.h
+++ b/algo/echo/echo-hash-4way.h
@@ -32,5 +32,8 @@ int echo_close( echo_4way_context *state, void *hashval );
 int echo_4way_update_close( echo_4way_context *state, void *hashval,
                              const void *data, int databitlen );

+int echo_4way_full( echo_4way_context *ctx, void *hashval, int nHashSize,
+                    const void *data, int datalen );
+
 #endif 
 #endif
--- a/algo/gost/sph_gost.c
+++ b/algo/gost/sph_gost.c
@@ -4,7 +4,7 @@
 #include <stdlib.h>
 #include <memory.h>
 #include <math.h>
-
+#include "simd-utils.h"
 #include "sph_gost.h"

 #ifdef __cplusplus
@@ -696,9 +696,26 @@ static void AddModulo512(const void *a,const void *b,void *c)

 static void AddXor512(const void *a,const void *b,void *c)
 {
-	const unsigned long long *A=a, *B=b;
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+   casti_m512i( c, 0 ) = _mm512_xor_si512( casti_m512i( a, 0 ),
+                                           casti_m512i( b, 0 ) );
+#elif defined(__AVX2__)
+   casti_m256i( c, 0 ) = _mm256_xor_si256( casti_m256i( a, 0 ),
+                                           casti_m256i( b, 0 ) );
+   casti_m256i( c, 1 ) = _mm256_xor_si256( casti_m256i( a, 1 ),
+                                           casti_m256i( b, 1 ) );
+#elif defined(__SSE2__)
+   casti_m128i( c, 0 ) = _mm_xor_si128( casti_m128i( a, 0 ),
+                                        casti_m128i( b, 0 ) );
+   casti_m128i( c, 1 ) = _mm_xor_si128( casti_m128i( a, 1 ),
+                                        casti_m128i( b, 1 ) );
+   casti_m128i( c, 2 ) = _mm_xor_si128( casti_m128i( a, 2 ),
+                                        casti_m128i( b, 2 ) );
+   casti_m128i( c, 3 ) = _mm_xor_si128( casti_m128i( a, 3 ),
+                                        casti_m128i( b, 3 ) );
+#else
+   const unsigned long long *A=a, *B=b;
 	unsigned long long *C=c;
-#ifdef FULL_UNROLL
 	C[0] = A[0] ^ B[0];
 	C[1] = A[1] ^ B[1];
 	C[2] = A[2] ^ B[2];
@@ -707,12 +724,6 @@ static void AddXor512(const void *a,const void *b,void *c)
 	C[5] = A[5] ^ B[5];
 	C[6] = A[6] ^ B[6];
 	C[7] = A[7] ^ B[7];
-#else
-	int i = 0;
-
-	for(i=0; i<8; i++) {
-		C[i] = A[i] ^ B[i];
-	}
 #endif
 }

@@ -893,31 +904,32 @@ static void g_N(const unsigned char *N,unsigned char *h,const unsigned char *m)

 static void hash_X(unsigned char *IV,const unsigned char *message,unsigned long long length,unsigned char *out)
 {
-	unsigned char v512[64] = {
+	unsigned char v512[64] __attribute__((aligned(64))) = {
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x02,0x00
-	};
-	unsigned char v0[64] = {
+   };
+	unsigned char v0[64]  __attribute__((aligned(64))) = {
+		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
+		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
+		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
+		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
+   };
+	unsigned char Sigma[64] __attribute__((aligned(64))) = {
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
 	};
-	unsigned char Sigma[64] = {
+	unsigned char N[64] __attribute__((aligned(64))) = {
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
 	};
-	unsigned char N[64] = {
-		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
-		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
-		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
-		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
-	};
-	unsigned char m[64], *hash = IV;
+	unsigned char m[64] __attribute__((aligned(64)));
+   unsigned char *hash = IV;
 	unsigned long long len = length;

 	// Stage 2
@@ -952,7 +964,7 @@ static void hash_X(unsigned char *IV,const unsigned char *message,unsigned long

 static void hash_512(const unsigned char *message, unsigned long long length, unsigned char *out)
 {
-	unsigned char IV[64] = {
+	unsigned char IV[64] __attribute__((aligned(64))) = {
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
 		0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,
--- a/algo/gost/sph_gost.h
+++ b/algo/gost/sph_gost.h
@@ -81,9 +81,9 @@ typedef struct {
 */
 typedef struct {
 #ifndef DOXYGEN_IGNORE
-	unsigned char buf[64];    /* first field, for alignment */
+	unsigned char buf[64] __attribute__((aligned(64))); 
+   sph_u32 V[5][8] __attribute__((aligned(64)));
 	size_t ptr;
-	sph_u32 V[5][8];
 #endif
 } sph_gost512_context;

--- a/algo/groestl/aes_ni/groestl-version.h
+++ b/algo/groestl/aes_ni/groestl-version.h
@@ -2,13 +2,6 @@
 //#define TASM
 #define TINTR

-//#define AES_NI
-
-//#ifdef AES_NI
-// specify AES-NI, AVX (with AES-NI) or vector-permute implementation
-
-//#ifndef NO_AES_NI
-
 // Not to be confused with AVX512VAES
 #define VAES
 // #define VAVX
--- a/algo/groestl/aes_ni/hash-groestl.c
+++ b/algo/groestl/aes_ni/hash-groestl.c
@@ -14,7 +14,7 @@
 #include "miner.h"
 #include "simd-utils.h"

-#ifndef NO_AES_NI
+#ifdef __AES__

 #include "groestl-version.h"

@@ -67,8 +67,12 @@ HashReturn_gr init_groestl( hashState_groestl* ctx, int hashlen )
     ctx->chaining[i] = _mm_setzero_si128();
     ctx->buffer[i]   = _mm_setzero_si128();
  }
-  ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
-  INIT(ctx->chaining);
+
+  // The only non-zero in the IV is len. It can be hard coded.
+  ctx->chaining[ 6 ] = m128_const_64( 0x0200000000000000, 0 );
+//  ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
+//  INIT(ctx->chaining);
+
  ctx->buf_ptr = 0;
  ctx->rem_ptr = 0;

@@ -87,8 +91,9 @@ HashReturn_gr reinit_groestl( hashState_groestl* ctx )
     ctx->chaining[i] = _mm_setzero_si128();
     ctx->buffer[i]   = _mm_setzero_si128();
  }
-  ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
-  INIT(ctx->chaining);
+  ctx->chaining[ 6 ] = m128_const_64( 0x0200000000000000, 0 );
+//  ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
+//  INIT(ctx->chaining);
  ctx->buf_ptr = 0;
  ctx->rem_ptr = 0;

@@ -180,6 +185,82 @@ HashReturn_gr final_groestl( hashState_groestl* ctx, void* output )
   return SUCCESS_GR;
 }

+int groestl512_full( hashState_groestl* ctx, void* output,
+                                const void* input, uint64_t databitlen )
+{
+
+  int i;
+
+  ctx->hashlen = 64;
+  SET_CONSTANTS();
+
+  for ( i = 0; i < SIZE512; i++ )
+  {
+     ctx->chaining[i] = _mm_setzero_si128();
+     ctx->buffer[i]   = _mm_setzero_si128();
+  }
+  ctx->chaining[ 6 ] = m128_const_64( 0x0200000000000000, 0 );
+  ctx->buf_ptr = 0;
+  ctx->rem_ptr = 0;
+
+
+   const int len = (int)databitlen / 128;
+   const int hashlen_m128i = ctx->hashlen / 16;   // bytes to __m128i
+   const int hash_offset = SIZE512 - hashlen_m128i;
+   int rem = ctx->rem_ptr;
+   uint64_t blocks = len / SIZE512;
+   __m128i* in = (__m128i*)input;
+
+   // --- update ---
+
+   // digest any full blocks, process directly from input 
+   for ( i = 0; i < blocks; i++ )
+      TF1024( ctx->chaining, &in[ i * SIZE512 ] );
+   ctx->buf_ptr = blocks * SIZE512;
+
+   // copy any remaining data to buffer, it may already contain data
+   // from a previous update for a midstate precalc
+   for ( i = 0; i < len % SIZE512; i++ )
+       ctx->buffer[ rem + i ] = in[ ctx->buf_ptr + i ];
+   i += rem;    // use i as rem_ptr in final
+
+   //--- final ---
+
+   blocks++;      // adjust for final block
+
+   if ( i == len -1 )
+   {
+       // only 128 bits left in buffer, all padding at once
+       ctx->buffer[i] = _mm_set_epi8( blocks,0,0,0, 0,0,0,0,
+                                           0,0,0,0, 0,0,0,0x80 );
+   }
+   else
+   {
+       // add first padding
+       ctx->buffer[i] = _mm_set_epi8( 0,0,0,0, 0,0,0,0,
+                                      0,0,0,0, 0,0,0,0x80 );
+       // add zero padding
+       for ( i += 1; i < SIZE512 - 1; i++ )
+           ctx->buffer[i] = _mm_setzero_si128();
+
+       // add length padding, second last byte is zero unless blocks > 255
+       ctx->buffer[i] = _mm_set_epi8( blocks, blocks>>8, 0,0, 0,0,0,0,
+                                           0,         0 ,0,0, 0,0,0,0 );
+   }
+
+   // digest final padding block and do output transform
+   TF1024( ctx->chaining, ctx->buffer );
+
+   OF1024( ctx->chaining );
+
+   // store hash result in output 
+   for ( i = 0; i < hashlen_m128i; i++ )
+      casti_m128i( output, i ) = ctx->chaining[ hash_offset + i ];
+
+   return 0;
+}
+   
+
 HashReturn_gr update_and_final_groestl( hashState_groestl* ctx, void* output,
                                const void* input, DataLength_gr databitlen )
 {
--- a/algo/groestl/aes_ni/hash-groestl.h
+++ b/algo/groestl/aes_ni/hash-groestl.h
@@ -87,5 +87,6 @@ HashReturn_gr final_groestl( hashState_groestl*, void* );

 HashReturn_gr update_and_final_groestl( hashState_groestl*,  void*,
                                        const void*, DataLength_gr );
+int groestl512_full( hashState_groestl*,  void*, const void*, uint64_t );

 #endif /* __hash_h */
--- a/algo/groestl/aes_ni/hash-groestl256.c
+++ b/algo/groestl/aes_ni/hash-groestl256.c
@@ -11,7 +11,7 @@
 #include "miner.h"
 #include "simd-utils.h"

-#ifndef NO_AES_NI
+#ifdef __AES__

 #include "groestl-version.h"

@@ -86,8 +86,11 @@ HashReturn_gr reinit_groestl256(hashState_groestl256* ctx)
     ctx->chaining[i] = _mm_setzero_si128();
     ctx->buffer[i]   = _mm_setzero_si128();
  }
-  ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
-  INIT256(ctx->chaining);
+
+  ctx->chaining[ 3 ] = m128_const_64( 0, 0x0100000000000000 );
+
+//  ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
+//  INIT256(ctx->chaining);
  ctx->buf_ptr = 0;
  ctx->rem_ptr = 0;

--- a/algo/groestl/aes_ni/hash-groestl256.h
+++ b/algo/groestl/aes_ni/hash-groestl256.h
@@ -93,9 +93,6 @@ typedef enum
 typedef struct {
  __attribute__ ((aligned (32))) __m128i chaining[SIZE256];
  __attribute__ ((aligned (32))) __m128i buffer[SIZE256];
-//  __attribute__ ((aligned (32))) u64 chaining[SIZE/8];      /* actual state */
-//  __attribute__ ((aligned (32))) BitSequence_gr buffer[SIZE];  /* data buffer */
-//  u64 block_counter;        /* message block counter */
  int hashlen;              // bytes
  int blk_count;
  int buf_ptr;              /* data buffer pointer */
--- a/algo/groestl/groestl-4way.c
+++ b/algo/groestl/groestl-4way.c
@@ -49,7 +49,7 @@ int scanhash_groestl_4way( struct work *work, uint32_t max_nonce,
        pdata[19] = n;

        for ( int lane = 0; lane < 4; lane++ )
-        if ( ( hash+(lane<<3) )[7] < Htarg )
+        if ( ( hash+(lane<<3) )[7] <= Htarg )
        if ( fulltest( hash+(lane<<3), ptarget) && !opt_benchmark )
        {
           pdata[19] = n + lane;
--- a/algo/groestl/groestl.c
+++ b/algo/groestl/groestl.c
@@ -3,19 +3,18 @@
 #include <stdlib.h>
 #include <stdint.h>
 #include <string.h>
-
-#ifdef NO_AES_NI
-  #include "sph_groestl.h"
-#else
+#ifdef __AES__
  #include "algo/groestl/aes_ni/hash-groestl.h"
+#else
+  #include "sph_groestl.h"
 #endif

 typedef struct
 {
-#ifdef NO_AES_NI
-    sph_groestl512_context groestl1, groestl2;
-#else
+#ifdef __AES__
    hashState_groestl groestl1, groestl2;
+#else
+    sph_groestl512_context groestl1, groestl2;
 #endif

 } groestl_ctx_holder;
@@ -24,12 +23,12 @@ static groestl_ctx_holder groestl_ctx;

 void init_groestl_ctx()
 {
-#ifdef NO_AES_NI
-    sph_groestl512_init( &groestl_ctx.groestl1 );
-    sph_groestl512_init( &groestl_ctx.groestl2 );
-#else
+#ifdef __AES__
    init_groestl( &groestl_ctx.groestl1, 64 );
    init_groestl( &groestl_ctx.groestl2, 64 );
+#else
+    sph_groestl512_init( &groestl_ctx.groestl1 );
+    sph_groestl512_init( &groestl_ctx.groestl2 );
 #endif
 }

@@ -39,18 +38,18 @@ void groestlhash( void *output, const void *input )
     groestl_ctx_holder ctx __attribute__ ((aligned (64)));
     memcpy( &ctx, &groestl_ctx, sizeof(groestl_ctx) );

-#ifdef NO_AES_NI
-     sph_groestl512(&ctx.groestl1, input, 80);
-     sph_groestl512_close(&ctx.groestl1, hash);
-
-     sph_groestl512(&ctx.groestl2, hash, 64);
-     sph_groestl512_close(&ctx.groestl2, hash);
-#else
+#ifdef __AES__
     update_and_final_groestl( &ctx.groestl1, (char*)hash,
                               (const char*)input, 640 );

     update_and_final_groestl( &ctx.groestl2, (char*)hash,
                               (const char*)hash, 512 );
+#else
+     sph_groestl512(&ctx.groestl1, input, 80);
+     sph_groestl512_close(&ctx.groestl1, hash);
+
+     sph_groestl512(&ctx.groestl2, hash, 64);
+     sph_groestl512_close(&ctx.groestl2, hash);
 #endif
     memcpy(output, hash, 32);
 }
--- a/algo/groestl/groestl256-hash-4way.c
+++ b/algo/groestl/groestl256-hash-4way.c
@@ -1,4 +1,5 @@
 /* hash.c     Aug 2011
+ * groestl512-hash-4way https://github.com/JayDDee/cpuminer-opt  2019-12.
 *
 * Groestl implementation for different versions.
 * Author: Krystian Matusiewicz, Günther A. Roland, Martin Schläffer
@@ -6,51 +7,18 @@
 * This code is placed in the public domain
 */

+// Optimized for hash and data length that are integrals of __m128i 
+
+
 #include <memory.h>
-#include "hash-groestl256.h"
+#include "groestl256-intr-4way.h"
 #include "miner.h"
 #include "simd-utils.h"

-#ifndef NO_AES_NI
+#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)

-#include "groestl-version.h"

-#ifdef TASM
-  #ifdef VAES
-    #include "groestl256-asm-aes.h"
-  #else
-    #ifdef VAVX
-      #include "groestl256-asm-avx.h"
-    #else
-      #ifdef VVPERM
-        #include "groestl256-asm-vperm.h"
-      #else
-        #error NO VERSION SPECIFIED (-DV[AES/AVX/VVPERM])
-      #endif
-    #endif
-  #endif
-#else
-  #ifdef TINTR
-    #ifdef VAES
-      #include "groestl256-intr-aes.h"
-    #else
-      #ifdef VAVX
-        #include "groestl256-intr-avx.h"
-      #else
-        #ifdef VVPERM
-          #include "groestl256-intr-vperm.h"
-        #else
-          #error NO VERSION SPECIFIED (-DV[AES/AVX/VVPERM])
-        #endif
-      #endif
-    #endif
-  #else
-    #error NO TYPE SPECIFIED (-DT[ASM/INTR])
-  #endif
-#endif
-
-/* initialise context */
-HashReturn_gr init_groestl256( hashState_groestl256* ctx, int hashlen )
+int groestl256_4way_init( groestl256_4way_context* ctx, uint64_t hashlen )
 {
  int i;

@@ -58,223 +26,84 @@ HashReturn_gr init_groestl256( hashState_groestl256* ctx, int hashlen )
  SET_CONSTANTS();

  if (ctx->chaining == NULL || ctx->buffer == NULL)
-    return FAIL_GR;
+    return 1;

  for ( i = 0; i < SIZE256; i++ )
  {
-     ctx->chaining[i] = _mm_setzero_si128();
-     ctx->buffer[i]   = _mm_setzero_si128();
+     ctx->chaining[i] = m512_zero;
+     ctx->buffer[i]   = m512_zero;
  }
-  ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
-  INIT256( ctx->chaining );
+
+  // The only non-zero in the IV is len. It can be hard coded.
+  ctx->chaining[ 3 ] = m512_const2_64( 0, 0x0100000000000000 );
+//  uint64_t len = U64BIG((uint64_t)LENGTH);
+//  ctx->chaining[ COLS/2 -1 ] = _mm512_set4_epi64( len, 0, len, 0 );
+//  INIT256_4way(ctx->chaining);
+
  ctx->buf_ptr = 0;
  ctx->rem_ptr = 0;

-  return SUCCESS_GR;
+  return 0;
 }

-
-HashReturn_gr reinit_groestl256(hashState_groestl256* ctx)
- {
-  int i;
-
-  if (ctx->chaining == NULL || ctx->buffer == NULL)
-    return FAIL_GR;
-
-  for ( i = 0; i < SIZE256; i++ )
-  {
-     ctx->chaining[i] = _mm_setzero_si128();
-     ctx->buffer[i]   = _mm_setzero_si128();
-  }
-  ((u64*)ctx->chaining)[COLS-1] = U64BIG((u64)LENGTH);
-  INIT256(ctx->chaining);
-  ctx->buf_ptr = 0;
-  ctx->rem_ptr = 0;
-
-  return SUCCESS_GR;
-}
-
-// Use this only for midstate and never for cryptonight
-HashReturn_gr update_groestl256( hashState_groestl256* ctx, const void* input,
-                                 DataLength_gr databitlen )
-{
-   __m128i* in = (__m128i*)input;
-   const int len = (int)databitlen / 128;  // bits to __m128i
-   const int blocks = len / SIZE256;    // __M128i to blocks
-   int rem = ctx->rem_ptr;
-   int i;
-
-   ctx->blk_count = blocks;
-   ctx->databitlen = databitlen;
-
-   // digest any full blocks 
-   for ( i = 0; i < blocks; i++ )
-       TF512( ctx->chaining, &in[ i * SIZE256 ] );
-   // adjust buf_ptr to last block
-   ctx->buf_ptr = blocks * SIZE256;
-
-   // Copy any remainder to buffer
-   for ( i = 0; i < len % SIZE256; i++ )
-       ctx->buffer[ rem + i ] = in[ ctx->buf_ptr + i ];
-   // adjust rem_ptr for new data
-   ctx->rem_ptr += i;
-
-   return SUCCESS_GR;
-}
-
-// don't use this at all
-HashReturn_gr final_groestl256( hashState_groestl256* ctx, void* output )
-{
-   const int len = (int)ctx->databitlen / 128;  // bits to __m128i 
-   const int blocks = ctx->blk_count + 1;       // adjust for final block
-   const int rem_ptr = ctx->rem_ptr;      // end of data start of padding
-   const int hashlen_m128i = ctx->hashlen / 16;  // bytes to __m128i
-   const int hash_offset = SIZE256 - hashlen_m128i;  // where in buffer
-   int i;
-
-   // first pad byte = 0x80, last pad byte = block count
-   // everything in between is zero
-
-   if ( rem_ptr == len - 1 )
-   {
-       // all padding at once
-       ctx->buffer[rem_ptr] = _mm_set_epi8( blocks,0,0,0, 0,0,0,0,
-                                                  0,0,0,0, 0,0,0,0x80 );
-   }
-   else
-   {
-       // add first padding
-       ctx->buffer[rem_ptr] = _mm_set_epi8( 0,0,0,0, 0,0,0,0,
-                                            0,0,0,0, 0,0,0,0x80 );
-       // add zero padding
-       for ( i = rem_ptr + 1; i < SIZE256 - 1; i++ )
-           ctx->buffer[i] = _mm_setzero_si128();
-       // add length padding
-       // cheat since we know the block count is trivial, good if block < 256
-       ctx->buffer[i] = _mm_set_epi8( blocks,0,0,0, 0,0,0,0,
-                                           0,0,0,0, 0,0,0,0 );
-   }
-
-   // digest final padding block and do output transform
-   TF512( ctx->chaining, ctx->buffer );
-   OF512( ctx->chaining );
-
-   // store hash result in output 
-   for ( i = 0; i < hashlen_m128i; i++ )
-      casti_m128i( output, i ) = ctx->chaining[ hash_offset + i];
-
-   return SUCCESS_GR;
-}
-
-HashReturn_gr update_and_final_groestl256( hashState_groestl256* ctx,
-                   void* output, const void* input, DataLength_gr databitlen )
+int groestl256_4way_update_close( groestl256_4way_context* ctx, void* output,
+                                const void* input, uint64_t databitlen )
 {
   const int len = (int)databitlen / 128;
   const int hashlen_m128i = ctx->hashlen / 16;   // bytes to __m128i
   const int hash_offset = SIZE256 - hashlen_m128i;
   int rem = ctx->rem_ptr;
   int blocks = len / SIZE256;
-   __m128i* in = (__m128i*)input;
+   __m512i* in = (__m512i*)input;
   int i;

   // --- update ---

   // digest any full blocks, process directly from input 
   for ( i = 0; i < blocks; i++ )
-      TF512( ctx->chaining, &in[ i * SIZE256 ] );
+      TF512_4way( ctx->chaining, &in[ i * SIZE256 ] );
   ctx->buf_ptr = blocks * SIZE256;

-   // cryptonight has 200 byte input, an odd number of __m128i
-   // remainder is only 8 bytes, ie u64.
-   if ( databitlen % 128 !=0 )
-   {
-      // must be cryptonight, copy 64 bits of data
-      *(uint64_t*)(ctx->buffer) = *(uint64_t*)(&in[ ctx->buf_ptr ] );
-      i = -1; // signal for odd length
-   }
-   else   
-   { 
-      // Copy any remaining data to buffer for final transform
-      for ( i = 0; i < len % SIZE256; i++ )
-          ctx->buffer[ rem + i ] = in[ ctx->buf_ptr + i ];
-      i += rem;   // use i as rem_ptr in final
-   }
+   // copy any remaining data to buffer, it may already contain data
+   // from a previous update for a midstate precalc
+   for ( i = 0; i < len % SIZE256; i++ )
+       ctx->buffer[ rem + i ] = in[ ctx->buf_ptr + i ];
+   i += rem;    // use i as rem_ptr in final

   //--- final ---

-   // adjust for final block
-   blocks++;
+   blocks++;      // adjust for final block

-   if ( i == len - 1 )
-   {
-       // all padding at once
-       ctx->buffer[i] = _mm_set_epi8( blocks,blocks>>8,0,0, 0,0,0,0,
-                                           0,        0,0,0, 0,0,0,0x80 );
-   }
+   if ( i == SIZE256 - 1 )
+   {        
+       // only 1 vector left in buffer, all padding at once
+       ctx->buffer[i] = m512_const1_128( _mm_set_epi8(
+                      blocks, blocks>>8,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0x80 ) );
+   }   
   else
   {
-      if ( i == -1 )
-      {
-         // cryptonight odd length
-         ((uint64_t*)ctx->buffer)[ 1 ] = 0x80ull;
-         // finish the block with zero and length padding as normal
-         i = 0;
-       }
-       else
-       {
-          // add first padding
-          ctx->buffer[i] = _mm_set_epi8( 0,0,0,0, 0,0,0,0,
-                                         0,0,0,0, 0,0,0,0x80 );
-       }
+       // add first padding
+       ctx->buffer[i] = m512_const4_64( 0, 0x80, 0, 0x80 );
       // add zero padding
       for ( i += 1; i < SIZE256 - 1; i++ )
-           ctx->buffer[i] = _mm_setzero_si128();
-       // add length padding
-       // cheat since we know the block count is trivial, good if block < 256
-       ctx->buffer[i] = _mm_set_epi8( blocks,blocks>>8,0,0, 0,0,0,0,
-                                           0,        0,0,0, 0,0,0,0 );
+           ctx->buffer[i] = m512_zero;
+
+       // add length padding, second last byte is zero unless blocks > 255
+       ctx->buffer[i] = m512_const1_128( _mm_set_epi8(
+                   blocks, blocks>>8, 0,0, 0,0, 0,0, 0,0, 0,0, 0,0, 0,0 ) );
   }

-   // digest final padding block and do output transform
-   TF512( ctx->chaining, ctx->buffer );
-   OF512( ctx->chaining );
+// digest final padding block and do output transform
+   TF512_4way( ctx->chaining, ctx->buffer );
+
+   OF512_4way( ctx->chaining );

   // store hash result in output 
   for ( i = 0; i < hashlen_m128i; i++ )
-      casti_m128i( output, i ) = ctx->chaining[ hash_offset + i ];
+      casti_m512i( output, i ) = ctx->chaining[ hash_offset + i ];

-   return SUCCESS_GR;
+   return 0;
 }

-/* hash bit sequence */
-HashReturn_gr hash_groestl256(int hashbitlen,
-                const BitSequence_gr* data,
-                DataLength_gr databitlen,
-                BitSequence_gr* hashval) {
-  HashReturn_gr ret;
-  hashState_groestl256 context;
+#endif   // VAES

-  /* initialise */
-  if ((ret = init_groestl256(&context, hashbitlen/8)) != SUCCESS_GR)
-    return ret;
-
-  /* process message */
-  if ((ret = update_groestl256(&context, data, databitlen)) != SUCCESS_GR)
-    return ret;
-
-  /* finalise */
-  ret = final_groestl256(&context, hashval);
-
-  return ret;
-}
-
-/* eBash API */
-//#ifdef crypto_hash_BYTES
-//int crypto_hash(unsigned char *out, const unsigned char *in, unsigned long long inlen)
-//{
-//  if (hash_groestl(crypto_hash_BYTES * 8, in, inlen * 8,out) == SUCCESS_GR) return 0;
-//  return -1;
-//}
-//#endif
-
-#endif
--- a/algo/groestl/groestl256-hash-4way.h
+++ b/algo/groestl/groestl256-hash-4way.h
@@ -6,56 +6,39 @@
 * This code is placed in the public domain
 */

-#ifndef __hash_h
-#define __hash_h
+#if !defined(GROESTL256_HASH_4WAY_H__)
+#define GROESTL256_HASH_4WAY_H__ 1

+#include "simd-utils.h"
 #include <immintrin.h>
+#include <stdint.h>
 #include <stdio.h>
 #if defined(_WIN64) || defined(__WINDOWS__)
 #include <windows.h>
 #endif
 #include <stdlib.h>

-/* eBash API begin */
-/*
-#include "crypto_hash.h"
-#ifdef crypto_hash_BYTES
-
-#include <crypto_uint8.h>
-#include <crypto_uint32.h>
-#include <crypto_uint64.h>
-typedef crypto_uint8 u8;
-typedef crypto_uint32 u32;
-typedef crypto_uint64 u64;
-#endif
- */
-/* eBash API end */
-
-//#define LENGTH (512)
-
-#include "brg_endian.h"
-#define NEED_UINT_64T
-#include "algo/sha/brg_types.h"
-
-#ifdef IACA_TRACE
-  #include IACA_MARKS
-#endif
-
+#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+   
 #define LENGTH (256)

+//#include "brg_endian.h"
+//#define NEED_UINT_64T
+//#include "algo/sha/brg_types.h"
+
 /* some sizes (number of bytes) */
 #define ROWS (8)
 #define LENGTHFIELDLEN (ROWS)
 #define COLS512 (8)
 //#define COLS1024 (16)
 #define SIZE_512 ((ROWS)*(COLS512))
-//#define SIZE1024 ((ROWS)*(COLS1024))
+//#define SIZE_1024 ((ROWS)*(COLS1024))
 #define ROUNDS512 (10)
 //#define ROUNDS1024 (14)

 //#if LENGTH<=256
 #define COLS (COLS512)
-//#define SIZE (SIZE512)
+#define SIZE (SIZE512)
 #define ROUNDS (ROUNDS512)
 //#else
 //#define COLS (COLS1024)
@@ -63,59 +46,30 @@ typedef crypto_uint64 u64;
 //#define ROUNDS (ROUNDS1024)
 //#endif

-#define ROTL64(a,n) ((((a)<<(n))|((a)>>(64-(n))))&li_64(ffffffffffffffff))
-
-#if (PLATFORM_BYTE_ORDER == IS_BIG_ENDIAN)
-#define EXT_BYTE(var,n) ((u8)((u64)(var) >> (8*(7-(n)))))
-#define U64BIG(a) (a)
-#endif /* IS_BIG_ENDIAN */
-
-#if (PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN)
-#define EXT_BYTE(var,n) ((u8)((u64)(var) >> (8*n)))
-#define U64BIG(a) \
-  ((ROTL64(a, 8) & li_64(000000FF000000FF)) | \
-   (ROTL64(a,24) & li_64(0000FF000000FF00)) | \
-   (ROTL64(a,40) & li_64(00FF000000FF0000)) | \
-   (ROTL64(a,56) & li_64(FF000000FF000000)))
-#endif /* IS_LITTLE_ENDIAN */
-
-typedef unsigned char BitSequence_gr;
-typedef unsigned long long DataLength_gr;
-typedef enum
-{
-    SUCCESS_GR = 0,
-    FAIL_GR = 1,
-    BAD_HASHBITLEN_GR = 2
-} HashReturn_gr;
-
 #define SIZE256 (SIZE_512/16)

 typedef struct {
-  __attribute__ ((aligned (32))) __m128i chaining[SIZE256];
-  __attribute__ ((aligned (32))) __m128i buffer[SIZE256];
-//  __attribute__ ((aligned (32))) u64 chaining[SIZE/8];      /* actual state */
-//  __attribute__ ((aligned (32))) BitSequence_gr buffer[SIZE];  /* data buffer */
-//  u64 block_counter;        /* message block counter */
-  int hashlen;              // bytes
-  int blk_count;
-  int buf_ptr;              /* data buffer pointer */
+  __attribute__ ((aligned (128))) __m512i chaining[SIZE256];
+  __attribute__ ((aligned (64))) __m512i buffer[SIZE256];
+  int hashlen;       // byte
+  int blk_count;     // SIZE_m128i
+  int buf_ptr;       // __m128i offset
  int rem_ptr;
-  int databitlen;
-} hashState_groestl256;
+  int databitlen;    // bits
+} groestl256_4way_context;

-HashReturn_gr init_groestl256( hashState_groestl256*, int );

-HashReturn_gr reinit_groestl256( hashState_groestl256* );
+int groestl256_4way_init( groestl256_4way_context*, uint64_t );

-HashReturn_gr update_groestl256( hashState_groestl256*, const void*,
-                              DataLength_gr );
+//int reinit_groestl( hashState_groestl* );

-HashReturn_gr final_groestl256( hashState_groestl256*, void* );
+//int groestl512_4way_update( groestl256_4way_context*, const void*,
+//                              uint64_t );

-HashReturn_gr hash_groestli256( int, const BitSequence_gr*, DataLength_gr,
-                            BitSequence_gr* );
+//int groestl512_4way_close( groestl512_4way_context*, void* );

-HashReturn_gr update_and_final_groestl256( hashState_groestl256*, void*,
-                                           const void*, DataLength_gr );
+int groestl256_4way_update_close( groestl256_4way_context*,  void*,
+                                        const void*, uint64_t );

-#endif /* __hash_h */
+#endif
+#endif 
--- a/algo/groestl/groestl256-intr-4way.h
+++ b/algo/groestl/groestl256-intr-4way.h
@@ -7,35 +7,37 @@
 * This code is placed in the public domain
 */

-#include <smmintrin.h>
-#include <wmmintrin.h>
-#include "hash-groestl256.h"
+
+#if !defined(GROESTL256_INTR_4WAY_H__)
+#define GROESTL256_INTR_4WAY_H__ 1
+      
+#include "groestl256-hash-4way.h"
+
+#if defined(__VAES__)

 /* global constants  */
-__m128i ROUND_CONST_Lx;
-__m128i ROUND_CONST_L0[ROUNDS512];
-__m128i ROUND_CONST_L7[ROUNDS512];
-//__m128i ROUND_CONST_P[ROUNDS1024];
-//__m128i ROUND_CONST_Q[ROUNDS1024];
-__m128i TRANSP_MASK;
-__m128i SUBSH_MASK[8];
-__m128i ALL_1B;
-__m128i ALL_FF;
-
+__m512i ROUND_CONST_Lx;
+__m512i ROUND_CONST_L0[ROUNDS512];
+__m512i ROUND_CONST_L7[ROUNDS512];
+//__m512i ROUND_CONST_P[ROUNDS1024];
+//__m512i ROUND_CONST_Q[ROUNDS1024];
+__m512i TRANSP_MASK;
+__m512i SUBSH_MASK[8];
+__m512i ALL_1B;
+__m512i ALL_FF;

 #define tos(a)    #a
 #define tostr(a)  tos(a)

-
 /* xmm[i] will be multiplied by 2
 * xmm[j] will be lost
 * xmm[k] has to be all 0x1b */
 #define MUL2(i, j, k){\
-  j = _mm_xor_si128(j, j);\
-  j = _mm_cmpgt_epi8(j, i);\
-  i = _mm_add_epi8(i, i);\
-  j = _mm_and_si128(j, k);\
-  i = _mm_xor_si128(i, j);\
+  j = _mm512_xor_si512(j, j);\
+  j = _mm512_movm_epi8( _mm512_cmpgt_epi8_mask(j, i) );\
+  i = _mm512_add_epi8(i, i);\
+  j = _mm512_and_si512(j, k);\
+  i = _mm512_xor_si512(i, j);\
 } 

 /**/
@@ -61,152 +63,188 @@ __m128i ALL_FF;
  /* t_i = a_i + a_{i+1} */\
  b6 = a0;\
  b7 = a1;\
-  a0 = _mm_xor_si128(a0, a1);\
+  a0 = _mm512_xor_si512(a0, a1);\
  b0 = a2;\
-  a1 = _mm_xor_si128(a1, a2);\
+  a1 = _mm512_xor_si512(a1, a2);\
  b1 = a3;\
-  a2 = _mm_xor_si128(a2, a3);\
+  a2 = _mm512_xor_si512(a2, a3);\
  b2 = a4;\
-  a3 = _mm_xor_si128(a3, a4);\
+  a3 = _mm512_xor_si512(a3, a4);\
  b3 = a5;\
-  a4 = _mm_xor_si128(a4, a5);\
+  a4 = _mm512_xor_si512(a4, a5);\
  b4 = a6;\
-  a5 = _mm_xor_si128(a5, a6);\
+  a5 = _mm512_xor_si512(a5, a6);\
  b5 = a7;\
-  a6 = _mm_xor_si128(a6, a7);\
-  a7 = _mm_xor_si128(a7, b6);\
+  a6 = _mm512_xor_si512(a6, a7);\
+  a7 = _mm512_xor_si512(a7, b6);\
  \
  /* build y4 y5 y6 ... in regs xmm8, xmm9, xmm10 by adding t_i*/\
-  b0 = _mm_xor_si128(b0, a4);\
-  b6 = _mm_xor_si128(b6, a4);\
-  b1 = _mm_xor_si128(b1, a5);\
-  b7 = _mm_xor_si128(b7, a5);\
-  b2 = _mm_xor_si128(b2, a6);\
-  b0 = _mm_xor_si128(b0, a6);\
+  b0 = _mm512_xor_si512(b0, a4);\
+  b6 = _mm512_xor_si512(b6, a4);\
+  b1 = _mm512_xor_si512(b1, a5);\
+  b7 = _mm512_xor_si512(b7, a5);\
+  b2 = _mm512_xor_si512(b2, a6);\
+  b0 = _mm512_xor_si512(b0, a6);\
  /* spill values y_4, y_5 to memory */\
  TEMP0 = b0;\
-  b3 = _mm_xor_si128(b3, a7);\
-  b1 = _mm_xor_si128(b1, a7);\
+  b3 = _mm512_xor_si512(b3, a7);\
+  b1 = _mm512_xor_si512(b1, a7);\
  TEMP1 = b1;\
-  b4 = _mm_xor_si128(b4, a0);\
-  b2 = _mm_xor_si128(b2, a0);\
+  b4 = _mm512_xor_si512(b4, a0);\
+  b2 = _mm512_xor_si512(b2, a0);\
  /* save values t0, t1, t2 to xmm8, xmm9 and memory */\
  b0 = a0;\
-  b5 = _mm_xor_si128(b5, a1);\
-  b3 = _mm_xor_si128(b3, a1);\
+  b5 = _mm512_xor_si512(b5, a1);\
+  b3 = _mm512_xor_si512(b3, a1);\
  b1 = a1;\
-  b6 = _mm_xor_si128(b6, a2);\
-  b4 = _mm_xor_si128(b4, a2);\
+  b6 = _mm512_xor_si512(b6, a2);\
+  b4 = _mm512_xor_si512(b4, a2);\
  TEMP2 = a2;\
-  b7 = _mm_xor_si128(b7, a3);\
-  b5 = _mm_xor_si128(b5, a3);\
+  b7 = _mm512_xor_si512(b7, a3);\
+  b5 = _mm512_xor_si512(b5, a3);\
  \
  /* compute x_i = t_i + t_{i+3} */\
-  a0 = _mm_xor_si128(a0, a3);\
-  a1 = _mm_xor_si128(a1, a4);\
-  a2 = _mm_xor_si128(a2, a5);\
-  a3 = _mm_xor_si128(a3, a6);\
-  a4 = _mm_xor_si128(a4, a7);\
-  a5 = _mm_xor_si128(a5, b0);\
-  a6 = _mm_xor_si128(a6, b1);\
-  a7 = _mm_xor_si128(a7, TEMP2);\
+  a0 = _mm512_xor_si512(a0, a3);\
+  a1 = _mm512_xor_si512(a1, a4);\
+  a2 = _mm512_xor_si512(a2, a5);\
+  a3 = _mm512_xor_si512(a3, a6);\
+  a4 = _mm512_xor_si512(a4, a7);\
+  a5 = _mm512_xor_si512(a5, b0);\
+  a6 = _mm512_xor_si512(a6, b1);\
+  a7 = _mm512_xor_si512(a7, TEMP2);\
  \
  /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
  /* compute w_i : add y_{i+4} */\
-  b1 = ALL_1B;\
+  b1 = m512_const1_64( 0x1b1b1b1b1b1b1b1b );\
  MUL2(a0, b0, b1);\
-  a0 = _mm_xor_si128(a0, TEMP0);\
+  a0 = _mm512_xor_si512(a0, TEMP0);\
  MUL2(a1, b0, b1);\
-  a1 = _mm_xor_si128(a1, TEMP1);\
+  a1 = _mm512_xor_si512(a1, TEMP1);\
  MUL2(a2, b0, b1);\
-  a2 = _mm_xor_si128(a2, b2);\
+  a2 = _mm512_xor_si512(a2, b2);\
  MUL2(a3, b0, b1);\
-  a3 = _mm_xor_si128(a3, b3);\
+  a3 = _mm512_xor_si512(a3, b3);\
  MUL2(a4, b0, b1);\
-  a4 = _mm_xor_si128(a4, b4);\
+  a4 = _mm512_xor_si512(a4, b4);\
  MUL2(a5, b0, b1);\
-  a5 = _mm_xor_si128(a5, b5);\
+  a5 = _mm512_xor_si512(a5, b5);\
  MUL2(a6, b0, b1);\
-  a6 = _mm_xor_si128(a6, b6);\
+  a6 = _mm512_xor_si512(a6, b6);\
  MUL2(a7, b0, b1);\
-  a7 = _mm_xor_si128(a7, b7);\
+  a7 = _mm512_xor_si512(a7, b7);\
  \
  /* compute v_i : double w_i      */\
  /* add to y_4 y_5 .. v3, v4, ... */\
  MUL2(a0, b0, b1);\
-  b5 = _mm_xor_si128(b5, a0);\
+  b5 = _mm512_xor_si512(b5, a0);\
  MUL2(a1, b0, b1);\
-  b6 = _mm_xor_si128(b6, a1);\
+  b6 = _mm512_xor_si512(b6, a1);\
  MUL2(a2, b0, b1);\
-  b7 = _mm_xor_si128(b7, a2);\
+  b7 = _mm512_xor_si512(b7, a2);\
  MUL2(a5, b0, b1);\
-  b2 = _mm_xor_si128(b2, a5);\
+  b2 = _mm512_xor_si512(b2, a5);\
  MUL2(a6, b0, b1);\
-  b3 = _mm_xor_si128(b3, a6);\
+  b3 = _mm512_xor_si512(b3, a6);\
  MUL2(a7, b0, b1);\
-  b4 = _mm_xor_si128(b4, a7);\
+  b4 = _mm512_xor_si512(b4, a7);\
  MUL2(a3, b0, b1);\
  MUL2(a4, b0, b1);\
  b0 = TEMP0;\
  b1 = TEMP1;\
-  b0 = _mm_xor_si128(b0, a3);\
-  b1 = _mm_xor_si128(b1, a4);\
+  b0 = _mm512_xor_si512(b0, a3);\
+  b1 = _mm512_xor_si512(b1, a4);\
 }/*MixBytes*/

-#define SET_CONSTANTS(){\
-   ALL_1B = _mm_set_epi32(0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b, 0x1b1b1b1b);\
-  TRANSP_MASK = _mm_set_epi32(0x0f070b03, 0x0e060a02, 0x0d050901, 0x0c040800);\
-  SUBSH_MASK[0] = _mm_set_epi32(0x03060a0d, 0x08020509, 0x0c0f0104, 0x070b0e00);\
-  SUBSH_MASK[1] = _mm_set_epi32(0x04070c0f, 0x0a03060b, 0x0e090205, 0x000d0801);\
-  SUBSH_MASK[2] = _mm_set_epi32(0x05000e09, 0x0c04070d, 0x080b0306, 0x010f0a02);\
-  SUBSH_MASK[3] = _mm_set_epi32(0x0601080b, 0x0e05000f, 0x0a0d0407, 0x02090c03);\
-  SUBSH_MASK[4] = _mm_set_epi32(0x0702090c, 0x0f060108, 0x0b0e0500, 0x030a0d04);\
-  SUBSH_MASK[5] = _mm_set_epi32(0x00030b0e, 0x0907020a, 0x0d080601, 0x040c0f05);\
-  SUBSH_MASK[6] = _mm_set_epi32(0x01040d08, 0x0b00030c, 0x0f0a0702, 0x050e0906);\
-  SUBSH_MASK[7] = _mm_set_epi32(0x02050f0a, 0x0d01040e, 0x090c0003, 0x06080b07);\
-  for(i = 0; i < ROUNDS512; i++)\
-  {\
-    ROUND_CONST_L0[i] = _mm_set_epi32(0xffffffff, 0xffffffff, 0x70605040 ^ (i * 0x01010101), 0x30201000 ^ (i * 0x01010101));\
-    ROUND_CONST_L7[i] = _mm_set_epi32(0x8f9fafbf ^ (i * 0x01010101), 0xcfdfefff ^ (i * 0x01010101), 0x00000000, 0x00000000);\
-  }\
-  ROUND_CONST_Lx = _mm_set_epi32(0xffffffff, 0xffffffff, 0x00000000, 0x00000000);\
-}while(0); \
+// calculate the round constants seperately and load at startup
+
+#define SET_CONSTANTS(){\
+  ALL_1B = _mm512_set1_epi32( 0x1b1b1b1b );\
+  TRANSP_MASK   = _mm512_set_epi32( \
+                         0x3f373b33, 0x3e363a32, 0x3d353931, 0x3c343830, \
+                         0x2f272b23, 0x2e262a22, 0x2d252921, 0x2c242820, \
+                         0x1f171b13, 0x1e161a12, 0x1d151911, 0x1c141810, \
+                         0x0f070b03, 0x0e060a02, 0x0d050901, 0x0c040800 ); \
+  SUBSH_MASK[0] = _mm512_set_epi32( \
+                         0x33363a3d, 0x38323539, 0x3c3f3134, 0x373b3e30, \
+                         0x23262a2d, 0x28222529, 0x2c2f2124, 0x272b2e20, \
+                         0x13161a1d, 0x18121519, 0x1c1f1114, 0x171b1e10, \
+                         0x03060a0d, 0x08020509, 0x0c0f0104, 0x070b0e00 ); \
+  SUBSH_MASK[1] = _mm512_set_epi32( \
+                         0x34373c3f, 0x3a33363b, 0x3e393235, 0x303d3831, \
+                         0x24272c2f, 0x2a23262b, 0x2e292225, 0x202d2821, \
+                         0x14171c1f, 0x1a13161b, 0x1e191215, 0x101d1801, \
+                         0x04070c0f, 0x0a03060b, 0x0e090205, 0x000d0801 );\
+  SUBSH_MASK[2] = _mm512_set_epi32( \
+                         0x35303e39, 0x3c34373d, 0x383b3336, 0x313f3a32, \
+                         0x25202e29, 0x2c24272d, 0x282b2326, 0x212f2a22, \
+                         0x15101e19, 0x1c14171d, 0x181b1316, 0x111f1a12, \
+                         0x05000e09, 0x0c04070d, 0x080b0306, 0x010f0a02 );\
+  SUBSH_MASK[3] = _mm512_set_epi32( \
+                         0x3631383b, 0x3e35303f, 0x3a3d3437, 0x32393c33, \
+                         0x2621282b, 0x2e25202f, 0x2a2d2427, 0x22292c23, \
+                         0x1611181b, 0x1e15101f, 0x1a1d1417, 0x12191c13, \
+                         0x0601080b, 0x0e05000f, 0x0a0d0407, 0x02090c03 );\
+  SUBSH_MASK[4] = _mm512_set_epi32( \
+                         0x3732393c, 0x3f363138, 0x3b3e3530, 0x333a3d34, \
+                         0x2722292c, 0x2f262128, 0x2b2e2520, 0x232a2d24, \
+                         0x1712191c, 0x1f161118, 0x1b1e1510, 0x131a1d14, \
+                         0x0702090c, 0x0f060108, 0x0b0e0500, 0x030a0d04 );\
+  SUBSH_MASK[5] = _mm512_set_epi32( \
+                         0x30333b3e, 0x3937323a, 0x3d383631, 0x343c3f35, \
+                         0x20232b2e, 0x2927222a, 0x2d282621, 0x242c2f25, \
+                         0x10131b1e, 0x1917121a, 0x1d181611, 0x141c1f15, \
+                         0x00030b0e, 0x0907020a, 0x0d080601, 0x040c0f05 );\
+  SUBSH_MASK[6] = _mm512_set_epi32( \
+                         0x31343d38, 0x3b30333c, 0x3f3a3732, 0x353e3936, \
+                         0x21242d28, 0x2b20232c, 0x2f2a2722, 0x252e2926, \
+                         0x11141d18, 0x1b10131c, 0x1f1a1712, 0x151e1916, \
+                         0x01040d08, 0x0b00030c, 0x0f0a0702, 0x050e0906 );\
+  SUBSH_MASK[7] = _mm512_set_epi32( \
+                         0x32353f3a, 0x3d31343e, 0x393c3033, 0x36383b37, \
+                         0x22252f2a, 0x2d21242e, 0x292c2023, 0x26282b27, \
+                         0x12151f1a, 0x1d11141e, 0x191c1013, 0x16181b17, \
+                         0x02050f0a, 0x0d01040e, 0x090c0003, 0x06080b07 );\
+  for ( i = 0; i < ROUNDS512; i++ ) \
+  {\
+    ROUND_CONST_L0[i] = _mm512_set4_epi32( 0xffffffff, 0xffffffff, \
+          0x70605040 ^ ( i * 0x01010101 ), 0x30201000 ^ ( i * 0x01010101 ) ); \
+    ROUND_CONST_L7[i] = _mm512_set4_epi32( 0x8f9fafbf ^ ( i * 0x01010101 ), \
+          0xcfdfefff ^ ( i * 0x01010101 ), 0x00000000, 0x00000000 ); \
+  }\
+  ROUND_CONST_Lx = _mm512_set4_epi32( 0xffffffff, 0xffffffff, \
+                                      0x00000000, 0x00000000 ); \
+}while(0);\

-/* one round
- * i = round number
- * a0-a7 = input rows
- * b0-b7 = output rows
- */
 #define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
  /* AddRoundConstant */\
  b1 = ROUND_CONST_Lx;\
-  a0 = _mm_xor_si128(a0, (ROUND_CONST_L0[i]));\
-  a1 = _mm_xor_si128(a1, b1);\
-  a2 = _mm_xor_si128(a2, b1);\
-  a3 = _mm_xor_si128(a3, b1);\
-  a4 = _mm_xor_si128(a4, b1);\
-  a5 = _mm_xor_si128(a5, b1);\
-  a6 = _mm_xor_si128(a6, b1);\
-  a7 = _mm_xor_si128(a7, (ROUND_CONST_L7[i]));\
+  a0 = _mm512_xor_si512( a0, (ROUND_CONST_L0[i]) );\
+  a1 = _mm512_xor_si512( a1, b1 );\
+  a2 = _mm512_xor_si512( a2, b1 );\
+  a3 = _mm512_xor_si512( a3, b1 );\
+  a4 = _mm512_xor_si512( a4, b1 );\
+  a5 = _mm512_xor_si512( a5, b1 );\
+  a6 = _mm512_xor_si512( a6, b1 );\
+  a7 = _mm512_xor_si512( a7, (ROUND_CONST_L7[i]) );\
  \
  /* ShiftBytes + SubBytes (interleaved) */\
-  b0 = _mm_xor_si128(b0,  b0);\
-  a0 = _mm_shuffle_epi8(a0, (SUBSH_MASK[0]));\
-  a0 = _mm_aesenclast_si128(a0, b0);\
-  a1 = _mm_shuffle_epi8(a1, (SUBSH_MASK[1]));\
-  a1 = _mm_aesenclast_si128(a1, b0);\
-  a2 = _mm_shuffle_epi8(a2, (SUBSH_MASK[2]));\
-  a2 = _mm_aesenclast_si128(a2, b0);\
-  a3 = _mm_shuffle_epi8(a3, (SUBSH_MASK[3]));\
-  a3 = _mm_aesenclast_si128(a3, b0);\
-  a4 = _mm_shuffle_epi8(a4, (SUBSH_MASK[4]));\
-  a4 = _mm_aesenclast_si128(a4, b0);\
-  a5 = _mm_shuffle_epi8(a5, (SUBSH_MASK[5]));\
-  a5 = _mm_aesenclast_si128(a5, b0);\
-  a6 = _mm_shuffle_epi8(a6, (SUBSH_MASK[6]));\
-  a6 = _mm_aesenclast_si128(a6, b0);\
-  a7 = _mm_shuffle_epi8(a7, (SUBSH_MASK[7]));\
-  a7 = _mm_aesenclast_si128(a7, b0);\
+  b0 = _mm512_xor_si512( b0, b0 );\
+  a0 = _mm512_shuffle_epi8( a0, (SUBSH_MASK[0]) );\
+  a0 = _mm512_aesenclast_epi128(a0, b0 );\
+  a1 = _mm512_shuffle_epi8( a1, (SUBSH_MASK[1]) );\
+  a1 = _mm512_aesenclast_epi128(a1, b0 );\
+  a2 = _mm512_shuffle_epi8( a2, (SUBSH_MASK[2]) );\
+  a2 = _mm512_aesenclast_epi128(a2, b0 );\
+  a3 = _mm512_shuffle_epi8( a3, (SUBSH_MASK[3]) );\
+  a3 = _mm512_aesenclast_epi128(a3, b0 );\
+  a4 = _mm512_shuffle_epi8( a4, (SUBSH_MASK[4]) );\
+  a4 = _mm512_aesenclast_epi128(a4, b0 );\
+  a5 = _mm512_shuffle_epi8( a5, (SUBSH_MASK[5]) );\
+  a5 = _mm512_aesenclast_epi128(a5, b0 );\
+  a6 = _mm512_shuffle_epi8( a6, (SUBSH_MASK[6]) );\
+  a6 = _mm512_aesenclast_epi128(a6, b0 );\
+  a7 = _mm512_shuffle_epi8( a7, (SUBSH_MASK[7]) );\
+  a7 = _mm512_aesenclast_epi128( a7, b0 );\
  \
  /* MixBytes */\
  MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7);\
@@ -237,31 +275,31 @@ __m128i ALL_FF;
 #define Matrix_Transpose_A(i0, i1, i2, i3, o1, o2, o3, t0){\
  t0 = TRANSP_MASK;\
  \
-  i0 = _mm_shuffle_epi8(i0, t0);\
-  i1 = _mm_shuffle_epi8(i1, t0);\
-  i2 = _mm_shuffle_epi8(i2, t0);\
-  i3 = _mm_shuffle_epi8(i3, t0);\
+  i0 = _mm512_shuffle_epi8( i0, t0 );\
+  i1 = _mm512_shuffle_epi8( i1, t0 );\
+  i2 = _mm512_shuffle_epi8( i2, t0 );\
+  i3 = _mm512_shuffle_epi8( i3, t0 );\
  \
  o1 = i0;\
  t0 = i2;\
  \
-  i0 = _mm_unpacklo_epi16(i0, i1);\
-  o1 = _mm_unpackhi_epi16(o1, i1);\
-  i2 = _mm_unpacklo_epi16(i2, i3);\
-  t0 = _mm_unpackhi_epi16(t0, i3);\
+  i0 = _mm512_unpacklo_epi16( i0, i1 );\
+  o1 = _mm512_unpackhi_epi16( o1, i1 );\
+  i2 = _mm512_unpacklo_epi16( i2, i3 );\
+  t0 = _mm512_unpackhi_epi16( t0, i3 );\
  \
-  i0 = _mm_shuffle_epi32(i0, 216);\
-  o1 = _mm_shuffle_epi32(o1, 216);\
-  i2 = _mm_shuffle_epi32(i2, 216);\
-  t0 = _mm_shuffle_epi32(t0, 216);\
+  i0 = _mm512_shuffle_epi32( i0, 216 );\
+  o1 = _mm512_shuffle_epi32( o1, 216 );\
+  i2 = _mm512_shuffle_epi32( i2, 216 );\
+  t0 = _mm512_shuffle_epi32( t0, 216 );\
  \
  o2 = i0;\
  o3 = o1;\
  \
-  i0 = _mm_unpacklo_epi32(i0, i2);\
-  o1 = _mm_unpacklo_epi32(o1, t0);\
-  o2 = _mm_unpackhi_epi32(o2, i2);\
-  o3 = _mm_unpackhi_epi32(o3, t0);\
+  i0 = _mm512_unpacklo_epi32( i0, i2 );\
+  o1 = _mm512_unpacklo_epi32( o1, t0 );\
+  o2 = _mm512_unpackhi_epi32( o2, i2 );\
+  o3 = _mm512_unpackhi_epi32( o3, t0 );\
 }/**/

 /* Matrix Transpose Step 2
@@ -279,19 +317,19 @@ __m128i ALL_FF;
 #define Matrix_Transpose_B(i0, i1, i2, i3, i4, i5, i6, i7, o1, o2, o3, o4, o5, o6, o7){\
  o1 = i0;\
  o2 = i1;\
-  i0 = _mm_unpacklo_epi64(i0, i4);\
-  o1 = _mm_unpackhi_epi64(o1, i4);\
+  i0 = _mm512_unpacklo_epi64( i0, i4 );\
+  o1 = _mm512_unpackhi_epi64( o1, i4 );\
  o3 = i1;\
  o4 = i2;\
-  o2 = _mm_unpacklo_epi64(o2, i5);\
-  o3 = _mm_unpackhi_epi64(o3, i5);\
+  o2 = _mm512_unpacklo_epi64( o2, i5 );\
+  o3 = _mm512_unpackhi_epi64( o3, i5 );\
  o5 = i2;\
  o6 = i3;\
-  o4 = _mm_unpacklo_epi64(o4, i6);\
-  o5 = _mm_unpackhi_epi64(o5, i6);\
+  o4 = _mm512_unpacklo_epi64( o4, i6 );\
+  o5 = _mm512_unpackhi_epi64( o5, i6 );\
  o7 = i3;\
-  o6 = _mm_unpacklo_epi64(o6, i7);\
-  o7 = _mm_unpackhi_epi64(o7, i7);\
+  o6 = _mm512_unpacklo_epi64( o6, i7 );\
+  o7 = _mm512_unpackhi_epi64( o7, i7 );\
 }/**/

 /* Matrix Transpose Inverse Step 2
@@ -302,19 +340,20 @@ __m128i ALL_FF;
 */
 #define Matrix_Transpose_B_INV(i0, i1, i2, i3, i4, i5, i6, i7, o0, o1, o2, o3){\
  o0 = i0;\
-  i0 = _mm_unpacklo_epi64(i0, i1);\
-  o0 = _mm_unpackhi_epi64(o0, i1);\
+  i0 = _mm512_unpacklo_epi64( i0, i1 );\
+  o0 = _mm512_unpackhi_epi64( o0, i1 );\
  o1 = i2;\
-  i2 = _mm_unpacklo_epi64(i2, i3);\
-  o1 = _mm_unpackhi_epi64(o1, i3);\
+  i2 = _mm512_unpacklo_epi64( i2, i3 );\
+  o1 = _mm512_unpackhi_epi64( o1, i3 );\
  o2 = i4;\
-  i4 = _mm_unpacklo_epi64(i4, i5);\
-  o2 = _mm_unpackhi_epi64(o2, i5);\
+  i4 = _mm512_unpacklo_epi64( i4, i5 );\
+  o2 = _mm512_unpackhi_epi64( o2, i5 );\
  o3 = i6;\
-  i6 = _mm_unpacklo_epi64(i6, i7);\
-  o3 = _mm_unpackhi_epi64(o3, i7);\
+  i6 = _mm512_unpacklo_epi64( i6, i7 );\
+  o3 = _mm512_unpackhi_epi64( o3, i7 );\
 }/**/

+
 /* Matrix Transpose Output Step 2
 * input is one 512-bit state with two rows in one xmm
 * output is one 512-bit state with one row in the low 64-bits of one xmm
@@ -322,19 +361,19 @@ __m128i ALL_FF;
 * outputs: (i0-7) = (0|S)
 */
 #define Matrix_Transpose_O_B(i0, i1, i2, i3, i4, i5, i6, i7, t0){\
-  t0 = _mm_xor_si128(t0, t0);\
+  t0 = _mm512_xor_si512( t0, t0 );\
  i1 = i0;\
  i3 = i2;\
  i5 = i4;\
  i7 = i6;\
-  i0 = _mm_unpacklo_epi64(i0, t0);\
-  i1 = _mm_unpackhi_epi64(i1, t0);\
-  i2 = _mm_unpacklo_epi64(i2, t0);\
-  i3 = _mm_unpackhi_epi64(i3, t0);\
-  i4 = _mm_unpacklo_epi64(i4, t0);\
-  i5 = _mm_unpackhi_epi64(i5, t0);\
-  i6 = _mm_unpacklo_epi64(i6, t0);\
-  i7 = _mm_unpackhi_epi64(i7, t0);\
+  i0 = _mm512_unpacklo_epi64( i0, t0 );\
+  i1 = _mm512_unpackhi_epi64( i1, t0 );\
+  i2 = _mm512_unpacklo_epi64( i2, t0 );\
+  i3 = _mm512_unpackhi_epi64( i3, t0 );\
+  i4 = _mm512_unpacklo_epi64( i4, t0 );\
+  i5 = _mm512_unpackhi_epi64( i5, t0 );\
+  i6 = _mm512_unpacklo_epi64( i6, t0 );\
+  i7 = _mm512_unpackhi_epi64( i7, t0 );\
 }/**/

 /* Matrix Transpose Output Inverse Step 2
@@ -344,17 +383,18 @@ __m128i ALL_FF;
 * outputs: (i0, i2, i4, i6) = S
 */
 #define Matrix_Transpose_O_B_INV(i0, i1, i2, i3, i4, i5, i6, i7){\
-  i0 = _mm_unpacklo_epi64(i0, i1);\
-  i2 = _mm_unpacklo_epi64(i2, i3);\
-  i4 = _mm_unpacklo_epi64(i4, i5);\
-  i6 = _mm_unpacklo_epi64(i6, i7);\
+  i0 = _mm512_unpacklo_epi64( i0, i1 );\
+  i2 = _mm512_unpacklo_epi64( i2, i3 );\
+  i4 = _mm512_unpacklo_epi64( i4, i5 );\
+  i6 = _mm512_unpacklo_epi64( i6, i7 );\
 }/**/


-void INIT256( __m128i* chaining )
+
+void INIT256_4way( __m512i* chaining )
 {
-  static __m128i xmm0, /*xmm1,*/ xmm2, /*xmm3, xmm4, xmm5,*/ xmm6, xmm7;
-  static __m128i /*xmm8, xmm9, xmm10, xmm11,*/ xmm12, xmm13, xmm14, xmm15;
+  static __m512i xmm0, xmm2, xmm6, xmm7;
+  static __m512i xmm12, xmm13, xmm14, xmm15;

  /* load IV into registers xmm12 - xmm15 */
  xmm12 = chaining[0];
@@ -373,17 +413,13 @@ void INIT256( __m128i* chaining )
  chaining[3] = xmm7;
 }

-void TF512( __m128i* chaining, __m128i* message )
+void TF512_4way( __m512i* chaining, __m512i* message )
 {
-  static __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m128i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
-  static __m128i TEMP0;
-  static __m128i TEMP1;
-  static __m128i TEMP2;
-
-#ifdef IACA_TRACE
-  IACA_START;
-#endif
+  static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  static __m512i TEMP0;
+  static __m512i TEMP1;
+  static __m512i TEMP2;

  /* load message into registers xmm12 - xmm15 */
  xmm12 = message[0];
@@ -404,10 +440,10 @@ void TF512( __m128i* chaining, __m128i* message )

  /* xor message to CV get input of P */
  /* result: CV+M in xmm8, xmm0, xmm4, xmm5 */
-  xmm8 = _mm_xor_si128(xmm8, xmm12);
-  xmm0 = _mm_xor_si128(xmm0, xmm2);
-  xmm4 = _mm_xor_si128(xmm4, xmm6);
-  xmm5 = _mm_xor_si128(xmm5, xmm7);
+  xmm8 = _mm512_xor_si512( xmm8, xmm12 );
+  xmm0 = _mm512_xor_si512( xmm0, xmm2 );
+  xmm4 = _mm512_xor_si512( xmm4, xmm6 );
+  xmm5 = _mm512_xor_si512( xmm5, xmm7 );

  /* there are now 2 rows of the Groestl state (P and Q) in each xmm register */
  /* unpack to get 1 row of P (64 bit) and Q (64 bit) into one xmm register */
@@ -422,17 +458,17 @@ void TF512( __m128i* chaining, __m128i* message )

  /* xor output of P and Q */
  /* result: P(CV+M)+Q(M) in xmm0...xmm3 */
-  xmm0 = _mm_xor_si128(xmm0, xmm8);
-  xmm1 = _mm_xor_si128(xmm1, xmm10);
-  xmm2 = _mm_xor_si128(xmm2, xmm12);
-  xmm3 = _mm_xor_si128(xmm3, xmm14);
+  xmm0 = _mm512_xor_si512( xmm0, xmm8 );
+  xmm1 = _mm512_xor_si512( xmm1, xmm10 );
+  xmm2 = _mm512_xor_si512( xmm2, xmm12 );
+  xmm3 = _mm512_xor_si512( xmm3, xmm14 );

  /* xor CV (feed-forward) */
  /* result: P(CV+M)+Q(M)+CV in xmm0...xmm3 */
-  xmm0 = _mm_xor_si128(xmm0, (chaining[0]));
-  xmm1 = _mm_xor_si128(xmm1, (chaining[1]));
-  xmm2 = _mm_xor_si128(xmm2, (chaining[2]));
-  xmm3 = _mm_xor_si128(xmm3, (chaining[3]));
+  xmm0 = _mm512_xor_si512( xmm0, (chaining[0]) );
+  xmm1 = _mm512_xor_si512( xmm1, (chaining[1]) );
+  xmm2 = _mm512_xor_si512( xmm2, (chaining[2]) );
+  xmm3 = _mm512_xor_si512( xmm3, (chaining[3]) );

  /* store CV */
  chaining[0] = xmm0;
@@ -440,19 +476,16 @@ void TF512( __m128i* chaining, __m128i* message )
  chaining[2] = xmm2;
  chaining[3] = xmm3;

-#ifdef IACA_TRACE
-  IACA_END;
-#endif
  return;
 }

-void OF512( __m128i* chaining )
+void OF512_4way( __m512i* chaining )
 {
-  static __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
-  static __m128i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
-  static __m128i TEMP0;
-  static __m128i TEMP1;
-  static __m128i TEMP2;
+  static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
+  static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
+  static __m512i TEMP0;
+  static __m512i TEMP1;
+  static __m512i TEMP2;

  /* load CV into registers xmm8, xmm10, xmm12, xmm14 */
  xmm8 = chaining[0];
@@ -475,10 +508,10 @@ void OF512( __m128i* chaining )

  /* xor CV to P output (feed-forward) */
  /* result: P(CV)+CV in xmm8, xmm10, xmm12, xmm14 */
-  xmm8 = _mm_xor_si128(xmm8,  (chaining[0]));
-  xmm10 = _mm_xor_si128(xmm10, (chaining[1]));
-  xmm12 = _mm_xor_si128(xmm12, (chaining[2]));
-  xmm14 = _mm_xor_si128(xmm14, (chaining[3]));
+  xmm8  = _mm512_xor_si512( xmm8,  (chaining[0]) );
+  xmm10 = _mm512_xor_si512( xmm10, (chaining[1]) );
+  xmm12 = _mm512_xor_si512( xmm12, (chaining[2]) );
+  xmm14 = _mm512_xor_si512( xmm14, (chaining[3]) );

  /* transform state back from row ordering into column ordering */
  /* result: final hash value in xmm9, xmm11 */
@@ -489,4 +522,5 @@ void OF512( __m128i* chaining )
  chaining[3] = xmm11;
 }

-
+#endif  // VAES
+#endif  // GROESTL512_INTR_4WAY_H__
--- a/algo/groestl/groestl512-hash-4way.c
+++ b/algo/groestl/groestl512-hash-4way.c
@@ -15,36 +15,23 @@
 #include "miner.h"
 #include "simd-utils.h"

-#if defined(__VAES__)
-
-#define ROTL64(a,n) \
-   ( ( ( (a)<<(n) ) | ( (a) >> (64-(n)) ) ) & 0xffffffffffffffff )
-     
-#define U64BIG(a) \
-  ( ( ROTL64(a, 8) & 0x000000FF000000FF ) | \
-    ( ROTL64(a,24) & 0x0000FF000000FF00 ) | \
-    ( ROTL64(a,40) & 0x00FF000000FF0000 ) | \
-    ( ROTL64(a,56) & 0xFF000000FF000000 ) )
+#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)

 int groestl512_4way_init( groestl512_4way_context* ctx, uint64_t hashlen )
 {
  int i;

-  ctx->hashlen = hashlen;
  SET_CONSTANTS();

  if (ctx->chaining == NULL || ctx->buffer == NULL)
    return 1;

-  for ( i = 0; i < SIZE512; i++ )
-  {
-     ctx->chaining[i] = m512_zero;
-     ctx->buffer[i]   = m512_zero;
-  }
+  memset_zero_512( ctx->chaining, SIZE512 );
+  memset_zero_512( ctx->buffer, SIZE512 );
+
+  // The only non-zero in the IV is len. It can be hard coded.
+  ctx->chaining[ 6 ] = m512_const2_64( 0x0200000000000000, 0 );

-  uint64_t len = U64BIG((uint64_t)LENGTH);
-  ctx->chaining[ COLS/2 -1 ] = _mm512_set4_epi64( len, 0, len, 0 );
-  INIT_4way(ctx->chaining);
  ctx->buf_ptr = 0;
  ctx->rem_ptr = 0;

@@ -55,7 +42,7 @@ int groestl512_4way_update_close( groestl512_4way_context* ctx, void* output,
                                const void* input, uint64_t databitlen )
 {
   const int len = (int)databitlen / 128;
-   const int hashlen_m128i = ctx->hashlen / 16;   // bytes to __m128i
+   const int hashlen_m128i = 64 / 16;   // bytes to __m128i
   const int hash_offset = SIZE512 - hashlen_m128i;
   int rem = ctx->rem_ptr;
   int blocks = len / SIZE512;
@@ -64,16 +51,13 @@ int groestl512_4way_update_close( groestl512_4way_context* ctx, void* output,

   // --- update ---

-   // digest any full blocks, process directly from input 
   for ( i = 0; i < blocks; i++ )
      TF1024_4way( ctx->chaining, &in[ i * SIZE512 ] );
   ctx->buf_ptr = blocks * SIZE512;

-   // copy any remaining data to buffer, it may already contain data
-   // from a previous update for a midstate precalc
   for ( i = 0; i < len % SIZE512; i++ )
       ctx->buffer[ rem + i ] = in[ ctx->buf_ptr + i ];
-   i += rem;    // use i as rem_ptr in final
+   i += rem; 

   //--- final ---

@@ -87,23 +71,71 @@ int groestl512_4way_update_close( groestl512_4way_context* ctx, void* output,
   }   
   else
   {
-       // add first padding
       ctx->buffer[i] = m512_const4_64( 0, 0x80, 0, 0x80 );
-       // add zero padding
       for ( i += 1; i < SIZE512 - 1; i++ )
           ctx->buffer[i] = m512_zero;
-
-       // add length padding, second last byte is zero unless blocks > 255
       ctx->buffer[i] = m512_const1_128( _mm_set_epi8(
                   blocks, blocks>>8, 0,0, 0,0, 0,0, 0,0, 0,0, 0,0, 0,0 ) );
   }

-// digest final padding block and do output transform
   TF1024_4way( ctx->chaining, ctx->buffer );
-
   OF1024_4way( ctx->chaining );

-   // store hash result in output 
+   for ( i = 0; i < hashlen_m128i; i++ )
+      casti_m512i( output, i ) = ctx->chaining[ hash_offset + i ];
+
+   return 0;
+}
+
+int groestl512_4way_full( groestl512_4way_context* ctx, void* output,
+                          const void* input, uint64_t datalen )
+{
+   const int len = (int)datalen >> 4;
+   const int hashlen_m128i = 64 >> 4;   // bytes to __m128i
+   const int hash_offset = SIZE512 - hashlen_m128i;
+   uint64_t blocks = len / SIZE512;
+   __m512i* in = (__m512i*)input;
+   int i;
+
+   // --- init ---
+
+   SET_CONSTANTS();
+   memset_zero_512( ctx->chaining, SIZE512 );
+   memset_zero_512( ctx->buffer, SIZE512 );
+   ctx->chaining[ 6 ] = m512_const2_64( 0x0200000000000000, 0 );
+   ctx->buf_ptr = 0;
+   ctx->rem_ptr = 0;
+
+   // --- update ---
+
+   for ( i = 0; i < blocks; i++ )
+      TF1024_4way( ctx->chaining, &in[ i * SIZE512 ] );
+   ctx->buf_ptr = blocks * SIZE512;
+
+   for ( i = 0; i < len % SIZE512; i++ )
+       ctx->buffer[ ctx->rem_ptr + i ] = in[ ctx->buf_ptr + i ];
+   i += ctx->rem_ptr;
+
+   // --- close ---
+
+   blocks++;   
+
+   if ( i == SIZE512 - 1 )
+   {
+       // only 1 vector left in buffer, all padding at once
+       ctx->buffer[i] = m512_const2_64( blocks << 56, 0x80 );
+   }
+   else
+   {
+       ctx->buffer[i] = m512_const4_64( 0, 0x80, 0, 0x80 );
+       for ( i += 1; i < SIZE512 - 1; i++ )
+           ctx->buffer[i] = m512_zero;
+       ctx->buffer[i] = m512_const2_64( blocks << 56, 0 );
+   }
+
+   TF1024_4way( ctx->chaining, ctx->buffer );
+   OF1024_4way( ctx->chaining );
+
   for ( i = 0; i < hashlen_m128i; i++ )
      casti_m512i( output, i ) = ctx->chaining[ hash_offset + i ];

--- a/algo/groestl/groestl512-hash-4way.h
+++ b/algo/groestl/groestl512-hash-4way.h
@@ -1,11 +1,3 @@
-/* hash.h     Aug 2011
- *
- * Groestl implementation for different versions.
- * Author: Krystian Matusiewicz, Günther A. Roland, Martin Schläffer
- *
- * This code is placed in the public domain
- */
-
 #if !defined(GROESTL512_HASH_4WAY_H__)
 #define GROESTL512_HASH_4WAY_H__ 1

@@ -18,11 +10,9 @@
 #endif
 #include <stdlib.h>

-#define LENGTH (512)
+#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)

-//#include "brg_endian.h"
-//#define NEED_UINT_64T
-//#include "algo/sha/brg_types.h"
+#define LENGTH (512)

 /* some sizes (number of bytes) */
 #define ROWS (8)
@@ -44,34 +34,11 @@
 #define ROUNDS (ROUNDS1024)
 //#endif

-/*
-#define ROTL64(a,n) ((((a)<<(n))|((a)>>(64-(n))))&li_64(ffffffffffffffff))
-
-#if (PLATFORM_BYTE_ORDER == IS_BIG_ENDIAN)
-#define EXT_BYTE(var,n) ((u8)((u64)(var) >> (8*(7-(n)))))
-#define U64BIG(a) (a)
-#endif // IS_BIG_ENDIAN 
-
-#if (PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN)
-#define EXT_BYTE(var,n) ((u8)((u64)(var) >> (8*n)))
-#define U64BIG(a) \
-  ((ROTL64(a, 8) & li_64(000000FF000000FF)) | \
-   (ROTL64(a,24) & li_64(0000FF000000FF00)) | \
-   (ROTL64(a,40) & li_64(00FF000000FF0000)) | \
-   (ROTL64(a,56) & li_64(FF000000FF000000)))
-#endif // IS_LITTLE_ENDIAN 
-
-typedef unsigned char BitSequence_gr;
-typedef unsigned long long DataLength_gr;
-typedef enum { SUCCESS_GR = 0, FAIL_GR = 1, BAD_HASHBITLEN_GR = 2} HashReturn_gr;
-*/
-
 #define SIZE512 (SIZE_1024/16)

 typedef struct {
  __attribute__ ((aligned (128))) __m512i chaining[SIZE512];
  __attribute__ ((aligned (64))) __m512i buffer[SIZE512];
-  int hashlen;       // byte
  int blk_count;     // SIZE_m128i
  int buf_ptr;       // __m128i offset
  int rem_ptr;
@@ -85,10 +52,11 @@ int groestl512_4way_init( groestl512_4way_context*, uint64_t );

 int groestl512_4way_update( groestl512_4way_context*, const void*,
                              uint64_t );
-
 int groestl512_4way_close( groestl512_4way_context*, void* );
-
 int groestl512_4way_update_close( groestl512_4way_context*,  void*,
                                        const void*, uint64_t );
+int groestl512_4way_full( groestl512_4way_context*,  void*,
+                          const void*, uint64_t );

-#endif /* __hash_h */
+#endif   // VAES
+#endif   // GROESTL512_HASH_4WAY_H__
--- a/algo/groestl/groestl512-intr-4way.h
+++ b/algo/groestl/groestl512-intr-4way.h
@@ -115,7 +115,7 @@ __m512i ALL_FF;
  \
  /* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
  /* compute w_i : add y_{i+4} */\
-  b1 = ALL_1B;\
+  b1 = m512_const1_64( 0x1b1b1b1b1b1b1b1b );\
  MUL2(a0, b0, b1);\
  a0 = _mm512_xor_si512(a0, TEMP0);\
  MUL2(a1, b0, b1);\
@@ -276,7 +276,7 @@ __m512i ALL_FF;
  for ( round_counter = 0; round_counter < 14; round_counter += 2) \
  { \
    /* AddRoundConstant Q1024 */\
-    xmm1 = ALL_FF;\
+    xmm1 = m512_neg1;\
    xmm8  = _mm512_xor_si512( xmm8,  xmm1 );\
    xmm9  = _mm512_xor_si512( xmm9,  xmm1 );\
    xmm10 = _mm512_xor_si512( xmm10, xmm1 );\
@@ -298,7 +298,7 @@ __m512i ALL_FF;
    SUBMIX(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
    \
    /* AddRoundConstant Q1024 */\
-    xmm9 = ALL_FF;\
+    xmm9 = m512_neg1;\
    xmm0 = _mm512_xor_si512( xmm0, xmm9 );\
    xmm1 = _mm512_xor_si512( xmm1, xmm9 );\
    xmm2 = _mm512_xor_si512( xmm2, xmm9 );\
--- a/algo/groestl/myr-groestl.c
+++ b/algo/groestl/myr-groestl.c
@@ -1,22 +1,20 @@
 #include "myrgr-gate.h"
-
 #include <stdio.h>
 #include <stdlib.h>
 #include <stdint.h>
 #include <string.h>
-
-#ifdef NO_AES_NI
-  #include "sph_groestl.h"
-#else
+#ifdef __AES__
  #include "aes_ni/hash-groestl.h"
+#else
+  #include "sph_groestl.h"
 #endif
 #include <openssl/sha.h>

 typedef struct {
-#ifdef NO_AES_NI
-    sph_groestl512_context  groestl;
-#else
+#ifdef __AES__
    hashState_groestl       groestl;
+#else
+    sph_groestl512_context  groestl;
 #endif
    SHA256_CTX              sha;
 } myrgr_ctx_holder;
@@ -25,10 +23,10 @@ myrgr_ctx_holder myrgr_ctx;

 void init_myrgr_ctx()
 {
-#ifdef NO_AES_NI
-     sph_groestl512_init( &myrgr_ctx.groestl );
-#else
+#ifdef __AES__
     init_groestl ( &myrgr_ctx.groestl, 64 );
+#else
+     sph_groestl512_init( &myrgr_ctx.groestl );
 #endif
     SHA256_Init( &myrgr_ctx.sha );
 }
@@ -40,12 +38,12 @@ void myriad_hash(void *output, const void *input)

 	uint32_t _ALIGN(32) hash[16];

-#ifdef NO_AES_NI
-	sph_groestl512(&ctx.groestl, input, 80);
-	sph_groestl512_close(&ctx.groestl, hash);
-#else
+#ifdef __AES__
   update_groestl( &ctx.groestl, (char*)input, 640 );
   final_groestl( &ctx.groestl, (char*)hash);
+#else
+	sph_groestl512(&ctx.groestl, input, 80);
+	sph_groestl512_close(&ctx.groestl, hash);
 #endif

   SHA256_Update( &ctx.sha, (unsigned char*)hash, 64 );
--- a/algo/heavy/bastion.c
+++ b/algo/heavy/bastion.c
@@ -1,13 +1,10 @@
 #include "algo-gate-api.h"
-
 #include <stdio.h>
 #include <string.h>
 #include <openssl/sha.h>
 #include <stdint.h>
 #include <stdlib.h>
-
 #include "sph_hefty1.h"
-
 #include "algo/luffa/sph_luffa.h"
 #include "algo/fugue/sph_fugue.h"
 #include "algo/skein/sph_skein.h"
@@ -16,8 +13,7 @@
 #include "algo/echo/sph_echo.h"
 #include "algo/hamsi/sph_hamsi.h"
 #include "algo/luffa/luffa_for_sse2.h"
-
-#ifndef NO_AES_NI
+#ifdef __AES__
  #include "algo/echo/aes_ni/hash_api.h"
 #endif

@@ -25,30 +21,23 @@ void bastionhash(void *output, const void *input)
 {
 	unsigned char hash[64] __attribute__ ((aligned (64)));

-#ifdef NO_AES_NI
-        sph_echo512_context     ctx_echo;
+#ifdef __AES__
+   hashState_echo          ctx_echo;
 #else
-        hashState_echo          ctx_echo;
+   sph_echo512_context     ctx_echo;
 #endif
-        hashState_luffa         ctx_luffa;
+   hashState_luffa         ctx_luffa;
 	sph_fugue512_context ctx_fugue;
 	sph_whirlpool_context ctx_whirlpool;
 	sph_shabal512_context ctx_shabal;
   sph_hamsi512_context ctx_hamsi;
 	sph_skein512_context ctx_skein;

-//        unsigned char hashbuf[128] __attribute__ ((aligned (16)));
-//        sph_u64 hashctA;
-//        sph_u64 hashctB;
-//        size_t hashptr;
-
 	HEFTY1(input, 80, hash);

        init_luffa( &ctx_luffa, 512 );
        update_and_final_luffa( &ctx_luffa, (BitSequence*)hash,
                                (const BitSequence*)hash, 64 );
-//        update_luffa( &ctx_luffa, hash, 64 );
-//        final_luffa( &ctx_luffa, hash );

 	if (hash[0] & 0x8)
 	{
@@ -71,23 +60,19 @@ void bastionhash(void *output, const void *input)

 	if (hash[0] & 0x8)
 	{
-#ifdef NO_AES_NI
+#ifdef __AES__
+      init_echo( &ctx_echo, 512 );
+      update_final_echo ( &ctx_echo,(BitSequence*)hash,
+                              (const BitSequence*)hash, 512 );
+#else
 		sph_echo512_init(&ctx_echo);
 		sph_echo512(&ctx_echo, hash, 64);
 		sph_echo512_close(&ctx_echo, hash);
-#else
-                init_echo( &ctx_echo, 512 );
-                update_final_echo ( &ctx_echo,(BitSequence*)hash,
-                                    (const BitSequence*)hash, 512 );
-//                update_echo ( &ctx_echo, hash, 512 );
-//                final_echo( &ctx_echo,  hash );
 #endif
 	} else {
-                init_luffa( &ctx_luffa, 512 );
-                update_and_final_luffa( &ctx_luffa, (BitSequence*)hash,
-                                        (const BitSequence*)hash, 64 );
-//                update_luffa( &ctx_luffa, hash, 64 );
-//                final_luffa( &ctx_luffa, hash );
+      init_luffa( &ctx_luffa, 512 );
+      update_and_final_luffa( &ctx_luffa, (BitSequence*)hash,
+                                    (const BitSequence*)hash, 64 );
 	}

 	sph_shabal512_init(&ctx_shabal);
@@ -119,11 +104,9 @@ void bastionhash(void *output, const void *input)
 		sph_hamsi512(&ctx_hamsi, hash, 64);
 		sph_hamsi512_close(&ctx_hamsi, hash);
 	} else {
-                init_luffa( &ctx_luffa, 512 );
-                update_and_final_luffa( &ctx_luffa, (BitSequence*)hash,
-                                        (const BitSequence*)hash, 64 );
-//                update_luffa( &ctx_luffa, hash, 64 );
-//                final_luffa( &ctx_luffa, hash );
+      init_luffa( &ctx_luffa, 512 );
+      update_and_final_luffa( &ctx_luffa, (BitSequence*)hash,
+                                    (const BitSequence*)hash, 64 );
 	}

 	memcpy(output, hash, 32);
@@ -150,10 +133,8 @@ int scanhash_bastion( struct work *work, uint32_t max_nonce,
 		be32enc(&endiandata[19], n);
 		bastionhash(hash32, endiandata);
 		if (hash32[7] < Htarg && fulltest(hash32, ptarget)) {
-			work_set_target_ratio(work, hash32);
-			*hashes_done = n - first_nonce + 1;
 			pdata[19] = n;
-			return true;
+         submit_solution( work, hash32, mythr );
 		}
 		n++;

--- a/algo/hodl/hodl-gate.c
+++ b/algo/hodl/hodl-gate.c
@@ -161,7 +161,7 @@ bool register_hodl_algo( algo_gate_t* gate )
 //     return false;
 //  }
  pthread_barrier_init( &hodl_barrier, NULL, opt_n_threads );
-  gate->optimizations         = AES_OPT | AVX_OPT | AVX2_OPT;
+  gate->optimizations         = SSE42_OPT | AES_OPT | AVX2_OPT;
  gate->scanhash              = (void*)&hodl_scanhash;
  gate->get_new_work          = (void*)&hodl_get_new_work;
  gate->longpoll_rpc_call     = (void*)&hodl_longpoll_rpc_call;
--- a/algo/jh/jh-hash-4way.c
+++ b/algo/jh/jh-hash-4way.c
@@ -41,57 +41,10 @@
 extern "C"{
 #endif

-
-#if SPH_SMALL_FOOTPRINT && !defined SPH_SMALL_FOOTPRINT_JH
-#define SPH_SMALL_FOOTPRINT_JH   1
-#endif
-
-#if !defined SPH_JH_64 && SPH_64_TRUE
-#define SPH_JH_64   1
-#endif
-
-#if !SPH_64
-#undef SPH_JH_64
-#endif
-
 #ifdef _MSC_VER
 #pragma warning (disable: 4146)
 #endif

-/*
- * The internal bitslice representation may use either big-endian or
- * little-endian (true bitslice operations do not care about the bit
- * ordering, and the bit-swapping linear operations in JH happen to
- * be invariant through endianness-swapping). The constants must be
- * defined according to the chosen endianness; we use some
- * byte-swapping macros for that.
- */
-
-#if SPH_LITTLE_ENDIAN
-
-#if SPH_64
-#define C64e(x)     ((SPH_C64(x) >> 56) \
-                    | ((SPH_C64(x) >> 40) & SPH_C64(0x000000000000FF00)) \
-                    | ((SPH_C64(x) >> 24) & SPH_C64(0x0000000000FF0000)) \
-                    | ((SPH_C64(x) >>  8) & SPH_C64(0x00000000FF000000)) \
-                    | ((SPH_C64(x) <<  8) & SPH_C64(0x000000FF00000000)) \
-                    | ((SPH_C64(x) << 24) & SPH_C64(0x0000FF0000000000)) \
-                    | ((SPH_C64(x) << 40) & SPH_C64(0x00FF000000000000)) \
-                    | ((SPH_C64(x) << 56) & SPH_C64(0xFF00000000000000)))
-#define dec64e_aligned   sph_dec64le_aligned
-#define enc64e           sph_enc64le
-#endif
-
-#else
-
-#if SPH_64
-#define C64e(x)     SPH_C64(x)
-#define dec64e_aligned   sph_dec64be_aligned
-#define enc64e           sph_enc64be
-#endif
-
-#endif
-
 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)

 #define Sb_8W(x0, x1, x2, x3, c) \
@@ -152,8 +105,97 @@ do { \
    x3 = _mm256_xor_si256( x3, x4 ); \
 } while (0)

-#if SPH_JH_64
+static const uint64_t C[] =
+{
+   0x67f815dfa2ded572, 0x571523b70a15847b,
+   0xf6875a4d90d6ab81, 0x402bd1c3c54f9f4e,
+   0x9cfa455ce03a98ea, 0x9a99b26699d2c503,
+   0x8a53bbf2b4960266, 0x31a2db881a1456b5,
+   0xdb0e199a5c5aa303, 0x1044c1870ab23f40,
+   0x1d959e848019051c, 0xdccde75eadeb336f,
+   0x416bbf029213ba10, 0xd027bbf7156578dc,
+   0x5078aa3739812c0a, 0xd3910041d2bf1a3f,
+   0x907eccf60d5a2d42, 0xce97c0929c9f62dd,
+   0xac442bc70ba75c18, 0x23fcc663d665dfd1,
+   0x1ab8e09e036c6e97, 0xa8ec6c447e450521,
+   0xfa618e5dbb03f1ee, 0x97818394b29796fd,
+   0x2f3003db37858e4a, 0x956a9ffb2d8d672a,
+   0x6c69b8f88173fe8a, 0x14427fc04672c78a,
+   0xc45ec7bd8f15f4c5, 0x80bb118fa76f4475,
+   0xbc88e4aeb775de52, 0xf4a3a6981e00b882,
+   0x1563a3a9338ff48e, 0x89f9b7d524565faa,
+   0xfde05a7c20edf1b6, 0x362c42065ae9ca36,
+   0x3d98fe4e433529ce, 0xa74b9a7374f93a53,
+   0x86814e6f591ff5d0, 0x9f5ad8af81ad9d0e,
+   0x6a6234ee670605a7, 0x2717b96ebe280b8b,
+   0x3f1080c626077447, 0x7b487ec66f7ea0e0,
+   0xc0a4f84aa50a550d, 0x9ef18e979fe7e391,
+   0xd48d605081727686, 0x62b0e5f3415a9e7e,
+   0x7a205440ec1f9ffc, 0x84c9f4ce001ae4e3,
+   0xd895fa9df594d74f, 0xa554c324117e2e55,
+   0x286efebd2872df5b, 0xb2c4a50fe27ff578,
+   0x2ed349eeef7c8905, 0x7f5928eb85937e44,
+   0x4a3124b337695f70, 0x65e4d61df128865e,
+   0xe720b95104771bc7, 0x8a87d423e843fe74,
+   0xf2947692a3e8297d, 0xc1d9309b097acbdd,
+   0xe01bdc5bfb301b1d, 0xbf829cf24f4924da,
+   0xffbf70b431bae7a4, 0x48bcf8de0544320d,
+   0x39d3bb5332fcae3b, 0xa08b29e0c1c39f45,
+   0x0f09aef7fd05c9e5, 0x34f1904212347094,
+   0x95ed44e301b771a2, 0x4a982f4f368e3be9,
+   0x15f66ca0631d4088, 0xffaf52874b44c147,
+   0x30c60ae2f14abb7e, 0xe68c6eccc5b67046,
+   0x00ca4fbd56a4d5a4, 0xae183ec84b849dda,
+   0xadd1643045ce5773, 0x67255c1468cea6e8,
+   0x16e10ecbf28cdaa3, 0x9a99949a5806e933,
+   0x7b846fc220b2601f, 0x1885d1a07facced1,
+   0xd319dd8da15b5932, 0x46b4a5aac01c9a50,
+   0xba6b04e467633d9f, 0x7eee560bab19caf6,
+   0x742128a9ea79b11f, 0xee51363b35f7bde9,
+   0x76d350755aac571d, 0x01707da3fec2463a,
+   0x42d8a498afc135f7, 0x79676b9e20eced78,
+   0xa8db3aea15638341, 0x832c83324d3bc3fa,
+   0xf347271c1f3b40a7, 0x9a762db734f04059,
+   0xfd4f21d26c4e3ee7, 0xef5957dc398dfdb8,
+   0xdaeb492b490c9b8d, 0x0d70f36849d7a25b,
+   0x84558d7ad0ae3b7d, 0x658ef8e4f0e9a5f5,
+   0x533b1036f4a2b8a0, 0x5aec3e759e07a80c,
+   0x4f88e85692946891, 0x4cbcbaf8555cb05b,
+   0x7b9487f3993bbbe3, 0x5d1c6b72d6f4da75,
+   0x6db334dc28acae64, 0x71db28b850a5346c,
+   0x2a518d10f2e261f8, 0xfc75dd593364dbe3,
+   0xa23fce43f1bcac1c, 0xb043e8023cd1bb67,
+   0x75a12988ca5b0a33, 0x5c5316b44d19347f,
+   0x1e4d790ec3943b92, 0x3fafeeb6d7757479,
+   0x21391abef7d4a8ea, 0x5127234c097ef45c,
+   0xd23c32ba5324a326, 0xadd5a66d4a17a344,
+   0x08c9f2afa63e1db5, 0x563c6b91983d5983,
+   0x4d608672a17cf84c, 0xf6c76e08cc3ee246,
+   0x5e76bcb1b333982f, 0x2ae6c4efa566d62b,
+   0x36d4c1bee8b6f406, 0x6321efbc1582ee74,
+   0x69c953f40d4ec1fd, 0x26585806c45a7da7,
+   0x16fae0061614c17e, 0x3f9d63283daf907e,
+   0x0cd29b00e3f2c9d2, 0x300cd4b730ceaa5f,
+   0x9832e0f216512a74, 0x9af8cee3d830eb0d,
+   0x9279f1b57b9ec54b, 0xd36886046ee651ff,
+   0x316796e6574d239b, 0x05750a17f3a6e6cc,
+   0xce6c3213d98176b1, 0x62a205f88452173c,
+   0x47154778b3cb2bf4, 0x486a9323825446ff,
+   0x65655e4e0758df38, 0x8e5086fc897cfcf2,
+   0x86ca0bd0442e7031, 0x4e477830a20940f0,
+   0x8338f7d139eea065, 0xbd3a2ce437e95ef7,
+   0x6ff8130126b29721, 0xe7de9fefd1ed44a3,
+   0xd992257615dfa08b, 0xbe42dc12f6f7853c,
+   0x7eb027ab7ceca7d8, 0xdea83eaada7d8d53,
+   0xd86902bd93ce25aa, 0xf908731afd43f65a,
+   0xa5194a17daef5fc0, 0x6a21fd4c33664d97,
+   0x701541db3198b435, 0x9b54cdedbb0f1eea,
+   0x72409751a163d09a, 0xe26f4791bf9d75f6
+};

+// Big endian version
+
+/*
 static const sph_u64 C[] = {
 	C64e(0x72d5dea2df15f867), C64e(0x7b84150ab7231557),
 	C64e(0x81abd6904d5a87f6), C64e(0x4e9f4fc5c3d12b40),
@@ -240,6 +282,7 @@ static const sph_u64 C[] = {
 	C64e(0x35b49831db411570), C64e(0xea1e0fbbedcd549b),
 	C64e(0x9ad063a151974072), C64e(0xf6759dbf91476fe2)
 };
+*/

 #define Ceven_hi(r)   (C[((r) << 2) + 0])
 #define Ceven_lo(r)   (C[((r) << 2) + 1])
@@ -427,7 +470,7 @@ do { \
   h7h = _mm256_xor_si256( h7h, m3h ); \
   h7l = _mm256_xor_si256( h7l, m3l ); \

-
+/*
 static const sph_u64 IV256[] = {
 	C64e(0xeb98a3412c20d3eb), C64e(0x92cdbe7b9cb245c1),
 	C64e(0x1c93519160d4c7fa), C64e(0x260082d67e508a03),
@@ -450,11 +493,8 @@ static const sph_u64 IV512[] = {
 	C64e(0xcf57f6ec9db1f856), C64e(0xa706887c5716b156),
 	C64e(0xe3c2fcdfe68517fb), C64e(0x545a4678cc8cdd4b)
 };
+*/

-#else
-
-
-#endif

 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)

@@ -484,57 +524,6 @@ static const sph_u64 IV512[] = {
 		W ## ro(h7); \
 	} while (0)

-#if SPH_SMALL_FOOTPRINT_JH
-
-#if SPH_JH_64
-
-/*
- * The "small footprint" 64-bit version just uses a partially unrolled
- * loop.
- */
-
-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
-
-#define E8_8W   do { \
-      unsigned r; \
-      for (r = 0; r < 42; r += 7) { \
-         SL_8W(0); \
-         SL_8W(1); \
-         SL_8W(2); \
-         SL_8W(3); \
-         SL_8W(4); \
-         SL_8W(5); \
-         SL_8W(6); \
-      } \
-   } while (0)
-
-#endif
-
-#define E8   do { \
-		unsigned r; \
-		for (r = 0; r < 42; r += 7) { \
-			SL(0); \
-			SL(1); \
-			SL(2); \
-			SL(3); \
-			SL(4); \
-			SL(5); \
-			SL(6); \
-		} \
-	} while (0)
-
-#else
-
-
-#endif
-
-#else
-
-#if SPH_JH_64
-
-/*
- * On a "true 64-bit" architecture, we can unroll at will.
- */

 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)

@@ -585,6 +574,7 @@ static const sph_u64 IV512[] = {

 #endif  // AVX512

+
 #define E8   do { \
      SLu( 0, 0); \
      SLu( 1, 1); \
@@ -630,13 +620,6 @@ static const sph_u64 IV512[] = {
      SLu(41, 6); \
   } while (0)

-#else
-
-
-#endif
-
-#endif
-
 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)

 void jh256_8way_init( jh_8way_context *sc )
@@ -732,12 +715,12 @@ jh_8way_core( jh_8way_context *sc, const void *data, size_t len )

 static void
 jh_8way_close( jh_8way_context *sc, unsigned ub, unsigned n, void *dst,
-               size_t out_size_w32, const void *iv )
+               size_t out_size_w32 )
 {
   __m512i buf[16*4];
   __m512i *dst512 = (__m512i*)dst;
   size_t numz, u;
-   sph_u64 l0, l1, l0e, l1e;
+   uint64_t l0, l1;

   buf[0] = m512_const1_64( 0x80ULL );

@@ -748,12 +731,10 @@ jh_8way_close( jh_8way_context *sc, unsigned ub, unsigned n, void *dst,

   memset_zero_512( buf+1, (numz>>3) - 1 );

-   l0 = SPH_T64(sc->block_count << 9) + (sc->ptr << 3);
-   l1 = SPH_T64(sc->block_count >> 55);
-   sph_enc64be( &l0e, l0 );
-   sph_enc64be( &l1e, l1 );
-   *(buf + (numz>>3)    ) = _mm512_set1_epi64( l1e );
-   *(buf + (numz>>3) + 1) = _mm512_set1_epi64( l0e );
+   l0 = ( sc->block_count << 9 ) + ( sc->ptr << 3 );
+   l1 = ( sc->block_count >> 55 );
+   *(buf + (numz>>3)    ) = _mm512_set1_epi64( bswap_64( l1 ) );
+   *(buf + (numz>>3) + 1) = _mm512_set1_epi64( bswap_64( l0 ) );

   jh_8way_core( sc, buf, numz + 16 );

@@ -772,7 +753,7 @@ jh256_8way_update(void *cc, const void *data, size_t len)
 void
 jh256_8way_close(void *cc, void *dst)
 {
-   jh_8way_close(cc, 0, 0, dst, 8, IV256);
+   jh_8way_close(cc, 0, 0, dst, 8);
 }

 void
@@ -784,7 +765,7 @@ jh512_8way_update(void *cc, const void *data, size_t len)
 void
 jh512_8way_close(void *cc, void *dst)
 {
-   jh_8way_close(cc, 0, 0, dst, 16, IV512);
+   jh_8way_close(cc, 0, 0, dst, 16);
 }

 #endif
@@ -882,12 +863,12 @@ jh_4way_core( jh_4way_context *sc, const void *data, size_t len )

 static void
 jh_4way_close( jh_4way_context *sc, unsigned ub, unsigned n, void *dst,
-               size_t out_size_w32, const void *iv )
+               size_t out_size_w32 )
 {
   __m256i buf[16*4];
   __m256i *dst256 = (__m256i*)dst;
   size_t numz, u;
-   sph_u64 l0, l1, l0e, l1e;
+   uint64_t l0, l1;

   buf[0] = m256_const1_64( 0x80ULL );

@@ -898,12 +879,10 @@ jh_4way_close( jh_4way_context *sc, unsigned ub, unsigned n, void *dst,

   memset_zero_256( buf+1, (numz>>3) - 1 );   

-   l0 = SPH_T64(sc->block_count << 9) + (sc->ptr << 3);
-   l1 = SPH_T64(sc->block_count >> 55);
-   sph_enc64be( &l0e, l0 );
-   sph_enc64be( &l1e, l1 );
-   *(buf + (numz>>3)    ) = _mm256_set1_epi64x( l1e );
-   *(buf + (numz>>3) + 1) = _mm256_set1_epi64x( l0e ); 
+   l0 = ( sc->block_count << 9 ) + ( sc->ptr << 3 );
+   l1 = ( sc->block_count >> 55 );
+   *(buf + (numz>>3)    ) = _mm256_set1_epi64x( bswap_64( l1 ) );
+   *(buf + (numz>>3) + 1) = _mm256_set1_epi64x( bswap_64( l0 ) );

   jh_4way_core( sc, buf, numz + 16 );

@@ -922,7 +901,7 @@ jh256_4way_update(void *cc, const void *data, size_t len)
 void
 jh256_4way_close(void *cc, void *dst)
 {
-	jh_4way_close(cc, 0, 0, dst, 8, IV256);
+	jh_4way_close(cc, 0, 0, dst, 8 );
 }

 void
@@ -934,7 +913,7 @@ jh512_4way_update(void *cc, const void *data, size_t len)
 void
 jh512_4way_close(void *cc, void *dst)
 {
-	jh_4way_close(cc, 0, 0, dst, 16, IV512);
+	jh_4way_close(cc, 0, 0, dst, 16 );
 }


--- a/algo/jh/jh-hash-4way.h
+++ b/algo/jh/jh-hash-4way.h
@@ -43,7 +43,6 @@ extern "C"{
 #endif

 #include <stddef.h>
-#include "algo/sha/sph_types.h"
 #include "simd-utils.h"

 #define SPH_SIZE_jh256   256
--- a/algo/jh/jha-4way.c
+++ b/algo/jh/jha-4way.c
@@ -65,7 +65,7 @@ void jha_hash_4way( void *out, const void *input )
          vh[i] = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );

       blake512_4way_init( &ctx_blake );
-       blake512_4way( &ctx_blake, vhash, 64 );
+       blake512_4way_update( &ctx_blake, vhash, 64 );
       blake512_4way_close( &ctx_blake, vhashA );

       jh512_4way_init( &ctx_jh );
--- a/algo/jh/jha.c
+++ b/algo/jh/jha.c
@@ -1,19 +1,16 @@
 #include "jha-gate.h"
-
 #include <stdlib.h>
 #include <stdint.h>
 #include <string.h>
 #include <stdio.h>
-
 #include "algo/blake/sph_blake.h"
 #include "algo/jh/sph_jh.h"
 #include "algo/keccak/sph_keccak.h"
 #include "algo/skein/sph_skein.h"
-
-#ifdef NO_AES_NI
-  #include "algo/groestl/sph_groestl.h"
-#else
+#ifdef __AES__
  #include "algo/groestl/aes_ni/hash-groestl.h"
+#else
+  #include "algo/groestl/sph_groestl.h"
 #endif

 static __thread sph_keccak512_context jha_kec_mid __attribute__ ((aligned (64)));
@@ -28,12 +25,12 @@ void jha_hash(void *output, const void *input)
 {
 	uint8_t _ALIGN(128) hash[64];

-#ifdef NO_AES_NI
-	sph_groestl512_context ctx_groestl;
+#ifdef __AES__
+   hashState_groestl      ctx_groestl;
 #else
-        hashState_groestl      ctx_groestl;
+	sph_groestl512_context ctx_groestl;
 #endif
-        sph_blake512_context ctx_blake;
+   sph_blake512_context ctx_blake;
 	sph_jh512_context ctx_jh;
 	sph_keccak512_context ctx_keccak;
 	sph_skein512_context ctx_skein;
@@ -46,36 +43,36 @@ void jha_hash(void *output, const void *input)
 	for (int round = 0; round < 3; round++)
 	{
 	   if (hash[0] & 0x01)
-           {
-#ifdef NO_AES_NI
-		sph_groestl512_init(&ctx_groestl);
-		sph_groestl512(&ctx_groestl, hash, 64 );
-		sph_groestl512_close(&ctx_groestl, hash );
+      {
+#ifdef __AES__
+         init_groestl( &ctx_groestl, 64 );
+         update_and_final_groestl( &ctx_groestl, (char*)hash,
+                                              (char*)hash, 512 );
 #else
-                init_groestl( &ctx_groestl, 64 );
-                update_and_final_groestl( &ctx_groestl, (char*)hash,
-                                          (char*)hash, 512 );
+   		sph_groestl512_init(&ctx_groestl);
+	   	sph_groestl512(&ctx_groestl, hash, 64 );
+		   sph_groestl512_close(&ctx_groestl, hash );
 #endif
-	    }
-            else
-            {
-		sph_skein512_init(&ctx_skein);
-		sph_skein512(&ctx_skein, hash, 64);
-		sph_skein512_close(&ctx_skein, hash );
-	    }
+      }
+      else
+      {
+		   sph_skein512_init(&ctx_skein);
+		   sph_skein512(&ctx_skein, hash, 64);
+		   sph_skein512_close(&ctx_skein, hash );
+	   }

-	    if (hash[0] & 0x01)
-            {
-		sph_blake512_init(&ctx_blake);
-		sph_blake512(&ctx_blake, hash, 64);
-		sph_blake512_close(&ctx_blake, hash );
-	    }
-            else
-            {
-		sph_jh512_init(&ctx_jh);
-		sph_jh512(&ctx_jh, hash, 64 );
-		sph_jh512_close(&ctx_jh, hash );
-	    }
+	   if (hash[0] & 0x01)
+      {
+		   sph_blake512_init(&ctx_blake);
+		   sph_blake512(&ctx_blake, hash, 64);
+		   sph_blake512_close(&ctx_blake, hash );
+	   }
+      else
+      {
+		   sph_jh512_init(&ctx_jh);
+		   sph_jh512(&ctx_jh, hash, 64 );
+		   sph_jh512_close(&ctx_jh, hash );
+	   }
 	}

 	memcpy(output, hash, 32);
@@ -117,9 +114,6 @@ int scanhash_jha( struct work *work, uint32_t max_nonce,

        jha_kec_midstate( endiandata );

-#ifdef DEBUG_ALGO
-	printf("[%d] Htarg=%X\n", thr_id, Htarg);
-#endif
 	for (int m=0; m < 6; m++) {
 		if (Htarg <= htmax[m]) {
 			uint32_t mask = masks[m];
@@ -127,25 +121,9 @@ int scanhash_jha( struct work *work, uint32_t max_nonce,
 				pdata[19] = ++n;
 				be32enc(&endiandata[19], n);
 				jha_hash(hash32, endiandata);
-#ifndef DEBUG_ALGO
-				if ((!(hash32[7] & mask)) && fulltest(hash32, ptarget)) {
-					work_set_target_ratio(work, hash32);
-					*hashes_done = n - first_nonce + 1;
-					return 1;
-				}
-#else
-				if (!(n % 0x1000) && !thr_id) printf(".");
-				if (!(hash32[7] & mask)) {
-					printf("[%d]",thr_id);
-					if (fulltest(hash32, ptarget)) {
-						work_set_target_ratio(work, hash32);
-						*hashes_done = n - first_nonce + 1;
-						return 1;
-					}
-				}
-#endif
+				if ((!(hash32[7] & mask)) && fulltest(hash32, ptarget)) 
+               submit_solution( work, hash32, mythr );
 			} while (n < max_nonce && !work_restart[thr_id].restart);
-			// see blake.c if else to understand the loop on htmax => mask
 			break;
 		}
 	}
--- a/algo/keccak/keccak-4way.c
+++ b/algo/keccak/keccak-4way.c
@@ -28,26 +28,28 @@ int scanhash_keccak_8way( struct work *work, uint32_t max_nonce,
   const uint32_t first_nonce = pdata[19];
   __m512i  *noncev = (__m512i*)vdata + 9;   // aligned
   const uint32_t Htarg = ptarget[7];
-   int thr_id = mythr->id;  
+   const int thr_id = mythr->id;  
+   const bool bench = opt_benchmark;

   mm512_bswap32_intrlv80_8x64( vdata, pdata );
+   *noncev = mm512_intrlv_blend_32(
+              _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 );
   do {
-       *noncev = mm512_intrlv_blend_32( mm512_bswap_32(
-                _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 );
-
      keccakhash_8way( hash, vdata );

      for ( int lane = 0; lane < 8; lane++ )
-      if ( hash7[ lane<<1 ] < Htarg ) 
+      if unlikely( hash7[ lane<<1 ] <= Htarg && !bench ) 
      {
          extr_lane_8x64( lane_hash, hash, lane, 256 );
-          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+          if ( valid_hash( lane_hash, ptarget ) )
          {
-              pdata[19] = n + lane;
+              pdata[19] = bswap_32( n + lane );
              submit_lane_solution( work, lane_hash, mythr, lane );
          }
      }
+      *noncev = _mm512_add_epi32( *noncev,
+                                  m512_const1_64( 0x0000000800000000 ) );
      n += 8;

   } while ( (n < max_nonce-8) && !work_restart[thr_id].restart);
@@ -79,27 +81,28 @@ int scanhash_keccak_4way( struct work *work, uint32_t max_nonce,
   const uint32_t first_nonce = pdata[19];
   __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
   const uint32_t Htarg = ptarget[7];
-   int thr_id = mythr->id;
+   const int thr_id = mythr->id;
+   const bool bench = opt_benchmark;

   mm256_bswap32_intrlv80_4x64( vdata, pdata );
+   *noncev = mm256_intrlv_blend_32(
+                _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
   do {
-       *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
-                _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
-	
      keccakhash_4way( hash, vdata );

      for ( int lane = 0; lane < 4; lane++ )
-      if ( hash7[ lane<<1 ] < Htarg )
+      if unlikely( hash7[ lane<<1 ] <= Htarg && !bench )
      {
          extr_lane_4x64( lane_hash, hash, lane, 256 );
-          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+          if ( valid_hash( lane_hash, ptarget ))
          {
-              pdata[19] = n + lane;
+              pdata[19] = bswap_32( n + lane );
              submit_lane_solution( work, lane_hash, mythr, lane );
          }
      }
+      *noncev = _mm256_add_epi32( *noncev,
+                                  m256_const1_64( 0x0000000400000000 ) );
      n += 4;
-
   } while ( (n < max_nonce-4) && !work_restart[thr_id].restart);

   *hashes_done = n - first_nonce + 1;
--- a/algo/keccak/keccak-gate.c
+++ b/algo/keccak/keccak-gate.c
@@ -1,5 +1,9 @@
 #include "keccak-gate.h"
+#include "sph_keccak.h"

+int hard_coded_eb = 1;
+
+// KECCAK

 bool register_keccak_algo( algo_gate_t* gate )
 {
@@ -19,6 +23,8 @@ bool register_keccak_algo( algo_gate_t* gate )
  return true;
 };

+// KECCAKC
+
 bool register_keccakc_algo( algo_gate_t* gate )
 {
  gate->optimizations = AVX2_OPT | AVX512_OPT;
@@ -37,3 +43,50 @@ bool register_keccakc_algo( algo_gate_t* gate )
  return true;
 };

+// SHA3D
+
+void sha3d( void *state, const void *input, int len )
+{
+	uint32_t _ALIGN(64) buffer[16], hash[16];
+	sph_keccak_context ctx_keccak;
+
+	sph_keccak256_init( &ctx_keccak );
+	sph_keccak256 ( &ctx_keccak, input, len );
+	sph_keccak256_close( &ctx_keccak, (void*) buffer );
+
+   sph_keccak256_init( &ctx_keccak );
+	sph_keccak256 ( &ctx_keccak, buffer, 32 );
+	sph_keccak256_close( &ctx_keccak, (void*) hash );
+
+	memcpy(state, hash, 32);
+}
+
+void sha3d_gen_merkle_root( char* merkle_root, struct stratum_ctx* sctx )
+{
+  sha3d( merkle_root, sctx->job.coinbase, (int) sctx->job.coinbase_size );
+  for ( int i = 0; i < sctx->job.merkle_count; i++ )
+  {
+     memcpy( merkle_root + 32, sctx->job.merkle[i], 32 );
+     sha256d( merkle_root, merkle_root, 64 );
+  }
+}
+
+bool register_sha3d_algo( algo_gate_t* gate )
+{
+  hard_coded_eb = 6;
+  opt_extranonce = false;
+  gate->optimizations = AVX2_OPT | AVX512_OPT;
+  gate->gen_merkle_root = (void*)&sha3d_gen_merkle_root;
+#if defined (KECCAK_8WAY)
+  gate->scanhash  = (void*)&scanhash_sha3d_8way;
+  gate->hash      = (void*)&sha3d_hash_8way;
+#elif defined (KECCAK_4WAY)
+  gate->scanhash  = (void*)&scanhash_sha3d_4way;
+  gate->hash      = (void*)&sha3d_hash_4way;
+#else
+  gate->scanhash  = (void*)&scanhash_sha3d;
+  gate->hash      = (void*)&sha3d_hash;
+#endif
+  return true;
+};
+
--- a/algo/keccak/keccak-gate.h
+++ b/algo/keccak/keccak-gate.h
@@ -10,24 +10,37 @@
  #define KECCAK_4WAY 1
 #endif

+extern int hard_coded_eb;
+
 #if defined(KECCAK_8WAY)

 void keccakhash_8way( void *state, const void *input );
 int scanhash_keccak_8way( struct work *work, uint32_t max_nonce,
                         uint64_t *hashes_done, struct thr_info *mythr );

+void sha3d_hash_8way( void *state, const void *input );
+int scanhash_sha3d_8way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr );
+
 #elif defined(KECCAK_4WAY)

 void keccakhash_4way( void *state, const void *input );
 int scanhash_keccak_4way( struct work *work, uint32_t max_nonce,
                         uint64_t *hashes_done, struct thr_info *mythr );

+void sha3d_hash_4way( void *state, const void *input );
+int scanhash_sha3d_4way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr );
+
 #else

 void keccakhash( void *state, const void *input );
 int scanhash_keccak( struct work *work, uint32_t max_nonce,
                    uint64_t *hashes_done, struct thr_info *mythr );

-#endif
+void sha3d_hash( void *state, const void *input );
+int scanhash_sha3d( struct work *work, uint32_t max_nonce,
+                    uint64_t *hashes_done, struct thr_info *mythr );

 #endif
+#endif
--- a/algo/keccak/keccak-hash-4way.c
+++ b/algo/keccak/keccak-hash-4way.c
@@ -1,6 +1,7 @@
 #include <stddef.h>
 #include <stdint.h>
 #include "keccak-hash-4way.h"
+#include "keccak-gate.h"

 static const uint64_t RC[] = {
        0x0000000000000001, 0x0000000000008082,
@@ -163,12 +164,12 @@ static void keccak64_8way_close( keccak64_ctx_m512i *kc, void *dst,
    unsigned eb;
    union {
       __m512i tmp[lim + 1];
-       sph_u64 dummy;   /* for alignment */
+       uint64_t dummy;   /* for alignment */
    } u;
    size_t j;
    size_t m512_len = byte_len >> 3;

-    eb = 0x100  >> 8;
+    eb = hard_coded_eb;
    if ( kc->ptr == (lim - 8) )
    {
        const uint64_t t = eb | 0x8000000000000000;
@@ -344,12 +345,12 @@ static void keccak64_close( keccak64_ctx_m256i *kc, void *dst, size_t byte_len,
    unsigned eb;
    union {
       __m256i tmp[lim + 1];
-       sph_u64 dummy;   /* for alignment */
+       uint64_t dummy;   /* for alignment */
    } u;
    size_t j;
    size_t m256_len = byte_len >> 3;

-    eb = 0x100  >> 8;
+    eb = hard_coded_eb;
    if ( kc->ptr == (lim - 8) )
    {
        const uint64_t t = eb | 0x8000000000000000;
--- a/algo/keccak/keccak-hash-4way.h
+++ b/algo/keccak/keccak-hash-4way.h
@@ -43,16 +43,8 @@ extern "C"{
 #ifdef  __AVX2__

 #include <stddef.h>
-#include "algo/sha/sph_types.h"
 #include "simd-utils.h"

-#define SPH_SIZE_keccak256   256
-
-/**
- * Output size (in bits) for Keccak-512.
- */
-#define SPH_SIZE_keccak512   512
-
 /**
 * This structure is a context for Keccak computations: it contains the
 * intermediate values and some data from the last entered block. Once a
--- a/algo/keccak/keccak.c
+++ b/algo/keccak/keccak.c
@@ -18,36 +18,34 @@ void keccakhash(void *state, const void *input)
 	memcpy(state, hash, 32);
 }

-int scanhash_keccak( struct work *work,
-	uint32_t max_nonce, uint64_t *hashes_done, struct thr_info *mythr )
+int scanhash_keccak( struct work *work, uint32_t max_nonce,
+                     uint64_t *hashes_done, struct thr_info *mythr )
 {
-        uint32_t *pdata = work->data;
-        uint32_t *ptarget = work->target;
-	uint32_t n = pdata[19] - 1;
-	const uint32_t first_nonce = pdata[19];
-	//const uint32_t Htarg = ptarget[7];
-   int thr_id = mythr->id;  // thr_id arg is deprecated
+   uint32_t _ALIGN(64) hash64[8];
+   uint32_t _ALIGN(64) endiandata[32];
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   uint32_t n = pdata[19];
+   const uint32_t first_nonce = pdata[19];
+   const uint32_t last_nonce = max_nonce;
+   const int thr_id = mythr->id;

-	uint32_t _ALIGN(32) hash64[8];
-	uint32_t endiandata[32];
+   for ( int i=0; i < 19; i++ )
+      be32enc( &endiandata[i], pdata[i] );

-        for (int i=0; i < 19; i++) 
-                be32enc(&endiandata[i], pdata[i]);
+   do {
+      be32enc( &endiandata[19], n );
+      keccakhash( hash64, endiandata );
+      if ( valid_hash( hash64, ptarget ) && !opt_benchmark )
+      {
+         pdata[19] = n;
+         submit_solution( work, hash64, mythr );
+      }
+      n++;
+   } while ( n < last_nonce && !work_restart[thr_id].restart );

-	do {
-	
-		pdata[19] = ++n;
-		be32enc(&endiandata[19], n); 
-		keccakhash(hash64, endiandata);
-        if (((hash64[7]&0xFFFFFF00)==0) && 
-				fulltest(hash64, ptarget)) {
-            *hashes_done = n - first_nonce + 1;
-			return true;
-		}
-	} while (n < max_nonce && !work_restart[thr_id].restart);
-	
-	*hashes_done = n - first_nonce + 1;
-	pdata[19] = n;
-	return 0;
+   *hashes_done = n - first_nonce;
+   pdata[19] = n;
+   return 0;
 }

--- a/algo/keccak/sha3d-4way.c
+++ b/algo/keccak/sha3d-4way.c
@@ -0,0 +1,126 @@
+#include "keccak-gate.h"
+#include <stdlib.h>
+#include <string.h>
+#include <stdint.h>
+#include "sph_keccak.h"
+#include "keccak-hash-4way.h"
+
+#if defined(KECCAK_8WAY)
+
+void sha3d_hash_8way(void *state, const void *input)
+{
+    uint32_t buffer[16*8] __attribute__ ((aligned (128)));
+    keccak256_8way_context ctx;
+
+    keccak256_8way_init( &ctx );
+    keccak256_8way_update( &ctx, input, 80 );
+    keccak256_8way_close( &ctx, buffer );
+
+    keccak256_8way_init( &ctx );
+    keccak256_8way_update( &ctx, buffer, 32 );
+    keccak256_8way_close( &ctx, state );
+}
+
+int scanhash_sha3d_8way( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t vdata[24*8] __attribute__ ((aligned (128)));
+   uint32_t hash[16*8] __attribute__ ((aligned (64)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (64)));
+   uint32_t *hash7 = &(hash[49]);   // 3*16+1
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   uint32_t n = pdata[19];
+   const uint32_t first_nonce = pdata[19];
+   const uint32_t last_nonce = max_nonce - 8;
+   __m512i  *noncev = (__m512i*)vdata + 9;   // aligned
+   const uint32_t Htarg = ptarget[7];
+   const int thr_id = mythr->id;  
+   const bool bench = opt_benchmark;
+
+   mm512_bswap32_intrlv80_8x64( vdata, pdata );
+   *noncev = mm512_intrlv_blend_32(
+              _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 );
+   do {
+      sha3d_hash_8way( hash, vdata );
+
+      for ( int lane = 0; lane < 8; lane++ )
+      if unlikely( hash7[ lane<<1 ] <= Htarg && !bench ) 
+      {
+          extr_lane_8x64( lane_hash, hash, lane, 256 );
+          if ( valid_hash( lane_hash, ptarget ) )
+          {
+              pdata[19] = bswap_32( n + lane );
+              submit_lane_solution( work, lane_hash, mythr, lane );
+          }
+      }
+      *noncev = _mm512_add_epi32( *noncev,
+                                  m512_const1_64( 0x0000000800000000 ) );
+      n += 8;
+
+   } while ( (n < last_nonce) && !work_restart[thr_id].restart);
+
+   *hashes_done = n - first_nonce;
+   return 0;
+}
+
+#elif defined(KECCAK_4WAY)
+
+void sha3d_hash_4way(void *state, const void *input)
+{
+    uint32_t buffer[16*4] __attribute__ ((aligned (64)));
+    keccak256_4way_context ctx;
+
+    keccak256_4way_init( &ctx );
+    keccak256_4way_update( &ctx, input, 80 );
+    keccak256_4way_close( &ctx, buffer );
+
+    keccak256_4way_init( &ctx );
+    keccak256_4way_update( &ctx, buffer, 32 );
+    keccak256_4way_close( &ctx, state );
+}
+
+int scanhash_sha3d_4way( struct work *work, uint32_t max_nonce,
+                          uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t vdata[24*4] __attribute__ ((aligned (64)));
+   uint32_t hash[16*4] __attribute__ ((aligned (32)));
+   uint32_t lane_hash[8] __attribute__ ((aligned (32)));
+   uint32_t *hash7 = &(hash[25]);   // 3*8+1
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+   uint32_t n = pdata[19];
+   const uint32_t first_nonce = pdata[19];
+   const uint32_t last_nonce = max_nonce - 4;
+   __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
+   const uint32_t Htarg = ptarget[7];
+   const int thr_id = mythr->id;
+   const bool bench = opt_benchmark;
+
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
+   *noncev = mm256_intrlv_blend_32( 
+                _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
+   do {
+      sha3d_hash_4way( hash, vdata );
+
+      for ( int lane = 0; lane < 4; lane++ )
+      if unlikely( hash7[ lane<<1 ] <= Htarg && !bench )
+      {
+          extr_lane_4x64( lane_hash, hash, lane, 256 );
+          if ( valid_hash( lane_hash, ptarget ) )
+          {
+              pdata[19] = bswap_32( n + lane );
+              submit_lane_solution( work, lane_hash, mythr, lane );
+          }
+      }
+      *noncev = _mm256_add_epi32( *noncev,
+                                  m256_const1_64( 0x0000000400000000 ) );
+      n += 4;
+   } while ( (n < last_nonce) && !work_restart[thr_id].restart);
+
+   *hashes_done = n - first_nonce;
+   return 0;
+}
+
+#endif
--- a/algo/keccak/sha3d.c
+++ b/algo/keccak/sha3d.c
@@ -0,0 +1,50 @@
+#include "algo-gate-api.h"
+#include <stdlib.h>
+#include <string.h>
+#include <stdint.h>
+#include "sph_keccak.h"
+
+void sha3d_hash(void *state, const void *input)
+{
+    uint32_t buffer[16];
+    sph_keccak256_context ctx_keccak;
+   
+    sph_keccak256_init( &ctx_keccak );
+    sph_keccak256 ( &ctx_keccak, input, 80 );
+    sph_keccak256_close( &ctx_keccak, buffer );
+    sph_keccak256_init( &ctx_keccak );
+    sph_keccak256 ( &ctx_keccak, buffer, 32 );
+    sph_keccak256_close( &ctx_keccak, state );
+}
+
+int scanhash_sha3d( struct work *work, uint32_t max_nonce,
+                    uint64_t *hashes_done, struct thr_info *mythr )
+{
+   uint32_t _ALIGN(64) hash64[8];
+   uint32_t _ALIGN(64) endiandata[32];
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
+	uint32_t n = pdata[19];
+	const uint32_t first_nonce = pdata[19];
+   const uint32_t last_nonce = max_nonce;
+   const int thr_id = mythr->id;
+
+   for ( int i=0; i < 19; i++ ) 
+      be32enc( &endiandata[i], pdata[i] );
+
+	do {
+		be32enc( &endiandata[19], n ); 
+		sha3d_hash( hash64, endiandata );
+      if ( valid_hash( hash64, ptarget ) && !opt_benchmark )
+      {
+         pdata[19] = n;
+         submit_solution( work, hash64, mythr );
+		}
+      n++;
+   } while ( n < last_nonce && !work_restart[thr_id].restart );
+	
+	*hashes_done = n - first_nonce;
+	pdata[19] = n;
+	return 0;
+}
+
--- a/algo/keccak/sph_keccak.c
+++ b/algo/keccak/sph_keccak.c
@@ -32,8 +32,8 @@

 #include <stddef.h>
 #include <string.h>
-
 #include "sph_keccak.h"
+#include "keccak-gate.h"

 #ifdef __cplusplus
 extern "C"{
@@ -1616,7 +1616,7 @@ keccak_core(sph_keccak_context *kc, const void *data, size_t len, size_t lim)
 		} u; \
 		size_t j; \
 \
-		eb = (0x100 | (ub & 0xFF)) >> (8 - n); \
+		eb = hard_coded_eb; \
 		if (kc->ptr == (lim - 1)) { \
 			if (n == 7) { \
 				u.tmp[0] = eb; \
--- a/algo/luffa/luffa-hash-2way.c
+++ b/algo/luffa/luffa-hash-2way.c
@@ -459,6 +459,11 @@ int luffa_4way_init( luffa_4way_context *state, int hashbitlen )
    return 0;
 }

+int luffa512_4way_init( luffa_4way_context *state )
+{
+   return luffa_4way_init( state, 512 );
+}
+   
 // Do not call luffa_update_close after having called luffa_update.
 // Once luffa_update has been called only call luffa_update or luffa_close.
 int luffa_4way_update( luffa_4way_context *state, const void *data,
@@ -496,6 +501,14 @@ int luffa_4way_update( luffa_4way_context *state, const void *data,
    return 0;
 }

+/*
+int luffa512_4way_update( luffa_4way_context *state, const void *data,
+                       size_t len )
+{
+   return luffa_4way_update( state, data, len );
+}
+*/
+
 int luffa_4way_close( luffa_4way_context *state, void *hashval )
 {
    __m512i *buffer = (__m512i*)state->buffer;
@@ -518,6 +531,77 @@ int luffa_4way_close( luffa_4way_context *state, void *hashval )
    return 0;
 }

+/*
+int luffa512_4way_close( luffa_4way_context *state, void *hashval )
+{
+   return luffa_4way_close( state, hashval );
+}
+*/
+
+int luffa512_4way_full( luffa_4way_context *state, void *output,
+                        const void *data, size_t inlen )
+{
+    state->hashbitlen = 512;
+    __m128i *iv = (__m128i*)IV;
+
+    state->chainv[0] = m512_const1_128( iv[0] );
+    state->chainv[1] = m512_const1_128( iv[1] );
+    state->chainv[2] = m512_const1_128( iv[2] );
+    state->chainv[3] = m512_const1_128( iv[3] );
+    state->chainv[4] = m512_const1_128( iv[4] );
+    state->chainv[5] = m512_const1_128( iv[5] );
+    state->chainv[6] = m512_const1_128( iv[6] );
+    state->chainv[7] = m512_const1_128( iv[7] );
+    state->chainv[8] = m512_const1_128( iv[8] );
+    state->chainv[9] = m512_const1_128( iv[9] );
+
+    ((__m512i*)state->buffer)[0] = m512_zero;
+    ((__m512i*)state->buffer)[1] = m512_zero;
+
+    const __m512i *vdata  = (__m512i*)data;
+    __m512i msg[2];
+    int i;
+    const int blocks = (int)( inlen >> 5 );
+    const __m512i shuff_bswap32 = m512_const_64(
+                                   0x3c3d3e3f38393a3b, 0x3435363730313233,
+                                   0x2c2d2e2f28292a2b, 0x2425262720212223,
+                                   0x1c1d1e1f18191a1b, 0x1415161710111213,
+                                   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[1] = _mm512_shuffle_epi8( vdata[ 1 ], shuff_bswap32 );
+       rnd512_4way( state, msg );
+    }
+
+    // 16 byte partial block exists for 80 byte len
+    if ( state->rembytes  )
+    {
+       // padding of partial block
+       msg[0] = _mm512_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[1] = m512_const2_64( 0, 0x0000000080000000 );
+       rnd512_4way( state, msg );
+    }
+    else
+    {
+       // empty pad block
+       msg[0] = m512_const2_64( 0, 0x0000000080000000 );
+       msg[1] = m512_zero;
+       rnd512_4way( state, msg );
+    }
+
+    finalization512_4way( state, (uint32*)output );
+
+    if ( state->hashbitlen > 512 )
+        finalization512_4way( state, (uint32*)( output+64 ) );
+
+    return 0;
+}
+
 int luffa_4way_update_close( luffa_4way_context *state,
                 void *output, const void *data, size_t inlen )
 {
@@ -1031,6 +1115,69 @@ int luffa_2way_close( luffa_2way_context *state, void *hashval )
    return 0;
 }

+int luffa512_2way_full( luffa_2way_context *state, void *output,
+                        const void *data, size_t inlen )
+{
+    state->hashbitlen = 512;
+    __m128i *iv = (__m128i*)IV;
+
+    state->chainv[0] = m256_const1_128( iv[0] );
+    state->chainv[1] = m256_const1_128( iv[1] );
+    state->chainv[2] = m256_const1_128( iv[2] );
+    state->chainv[3] = m256_const1_128( iv[3] );
+    state->chainv[4] = m256_const1_128( iv[4] );
+    state->chainv[5] = m256_const1_128( iv[5] );
+    state->chainv[6] = m256_const1_128( iv[6] );
+    state->chainv[7] = m256_const1_128( iv[7] );
+    state->chainv[8] = m256_const1_128( iv[8] );
+    state->chainv[9] = m256_const1_128( iv[9] );
+
+    ((__m256i*)state->buffer)[0] = m256_zero;
+    ((__m256i*)state->buffer)[1] = m256_zero;
+
+    const __m256i *vdata  = (__m256i*)data;
+    __m256i msg[2];
+    int i;
+    const int blocks = (int)( inlen >> 5 );
+    const __m256i shuff_bswap32 = m256_const_64( 0x1c1d1e1f18191a1b,
+                                                 0x1415161710111213,
+                                                 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[1] = _mm256_shuffle_epi8( vdata[ 1 ], shuff_bswap32 );
+       rnd512_2way( state, msg );
+    }
+
+    // 16 byte partial block exists for 80 byte len
+    if ( state->rembytes  )
+    {
+       // padding of partial block
+       msg[0] = _mm256_shuffle_epi8( vdata[ 0 ], shuff_bswap32 );
+       msg[1] = m256_const2_64( 0, 0x0000000080000000 );
+       rnd512_2way( state, msg );
+    }
+    else
+    {
+       // empty pad block
+       msg[0] = m256_const2_64( 0, 0x0000000080000000 );
+       msg[1] = m256_zero;
+       rnd512_2way( state, msg );
+    }
+
+    finalization512_2way( state, (uint32*)output );
+
+    if ( state->hashbitlen > 512 )
+        finalization512_2way( state, (uint32*)( output+32 ) );
+
+    return 0;
+}
+
 int luffa_2way_update_close( luffa_2way_context *state,
                 void *output, const void *data, size_t inlen )
 {
--- a/algo/luffa/luffa-hash-2way.h
+++ b/algo/luffa/luffa-hash-2way.h
@@ -61,11 +61,23 @@ typedef struct {
 } luffa_4way_context __attribute((aligned(128)));

 int luffa_4way_init( luffa_4way_context *state, int hashbitlen );
-int luffa_4way_update( luffa_4way_context *state, const void *data,
-                       size_t len );
-int luffa_4way_close( luffa_4way_context *state, void *hashval );
+//int luffa_4way_update( luffa_4way_context *state, const void *data,
+//                       size_t len );
+//int luffa_4way_close( luffa_4way_context *state, void *hashval );
 int luffa_4way_update_close( luffa_4way_context *state, void *output,
                                   const void *data, size_t inlen );
+int luffa512_4way_full( luffa_4way_context *state, void *output,
+                         const void *data, size_t inlen );
+int luffa512_4way_init( luffa_4way_context *state );
+int luffa512_4way_update( luffa_4way_context *state, const void *data,
+                       size_t len );
+int luffa512_4way_close( luffa_4way_context *state, void *hashval );
+int luffa512_4way_update_close( luffa_4way_context *state, void *output,
+                                const void *data, size_t inlen );
+
+#define luffa_4way_update       luffa512_4way_update
+#define luffa_4way_close        luffa512_4way_close
+#define luffa_4way_update_close luffa512_4way_update_close

 #endif

@@ -82,6 +94,8 @@ int luffa_2way_update( luffa_2way_context *state, const void *data,
 int luffa_2way_close( luffa_2way_context *state, void *hashval );
 int luffa_2way_update_close( luffa_2way_context *state, void *output,
                                   const void *data, size_t inlen );
+int luffa512_2way_full( luffa_2way_context *state, void *output,
+                         const void *data, size_t inlen );

 #endif
 #endif
--- a/algo/lyra2/allium-4way.c
+++ b/algo/lyra2/allium-4way.c
@@ -7,33 +7,44 @@
 #include "algo/cubehash/cubehash_sse2.h"
 #include "algo/cubehash/cube-hash-2way.h"
 #include "algo/groestl/aes_ni/hash-groestl256.h"
+#if defined(__VAES__)
+  #include "algo/groestl/groestl256-hash-4way.h"
+#endif

-#if defined (ALLIUM_8WAY)  
+#if defined (ALLIUM_16WAY)  

 typedef struct {
-   blake256_8way_context     blake;
+   blake256_16way_context     blake;
   keccak256_8way_context    keccak;
   cube_4way_context          cube;
   skein256_8way_context     skein;
+#if defined(__VAES__)
+   groestl256_4way_context groestl;
+#else
   hashState_groestl256      groestl;
-} allium_8way_ctx_holder;
+#endif
+} allium_16way_ctx_holder;

-static __thread allium_8way_ctx_holder allium_8way_ctx;
+static __thread allium_16way_ctx_holder allium_16way_ctx;

-bool init_allium_8way_ctx()
+bool init_allium_16way_ctx()
 {
-   keccak256_8way_init( &allium_8way_ctx.keccak );
-   cube_4way_init( &allium_8way_ctx.cube, 256, 16, 32 );
-   skein256_8way_init( &allium_8way_ctx.skein );
-   init_groestl256( &allium_8way_ctx.groestl, 32 );
+   keccak256_8way_init( &allium_16way_ctx.keccak );
+   cube_4way_init( &allium_16way_ctx.cube, 256, 16, 32 );
+   skein256_8way_init( &allium_16way_ctx.skein );
+#if defined(__VAES__)
+   groestl256_4way_init( &allium_16way_ctx.groestl, 32 );
+#else
+   init_groestl256( &allium_16way_ctx.groestl, 32 );
+#endif
   return true;
 }

-void allium_8way_hash( void *state, const void *input )
+void allium_16way_hash( void *state, const void *input )
 {
-   uint32_t vhash[8*8] __attribute__ ((aligned (128)));
-   uint32_t vhashA[8*8] __attribute__ ((aligned (64)));
-   uint32_t vhashB[8*8] __attribute__ ((aligned (64)));
+   uint32_t vhash[16*8] __attribute__ ((aligned (128)));
+   uint32_t vhashA[16*8] __attribute__ ((aligned (64)));
+   uint32_t vhashB[16*8] __attribute__ ((aligned (64)));
   uint32_t hash0[8] __attribute__ ((aligned (64)));
   uint32_t hash1[8] __attribute__ ((aligned (64)));
   uint32_t hash2[8] __attribute__ ((aligned (64)));
@@ -42,18 +53,39 @@ void allium_8way_hash( void *state, const void *input )
   uint32_t hash5[8] __attribute__ ((aligned (64)));
   uint32_t hash6[8] __attribute__ ((aligned (64)));
   uint32_t hash7[8] __attribute__ ((aligned (64)));
-   allium_8way_ctx_holder ctx __attribute__ ((aligned (64)));
+   uint32_t hash8[8] __attribute__ ((aligned (64)));
+   uint32_t hash9[8] __attribute__ ((aligned (64)));
+   uint32_t hash10[8] __attribute__ ((aligned (64)));
+   uint32_t hash11[8] __attribute__ ((aligned (64)));
+   uint32_t hash12[8] __attribute__ ((aligned (64)));
+   uint32_t hash13[8] __attribute__ ((aligned (64)));
+   uint32_t hash14[8] __attribute__ ((aligned (64)));
+   uint32_t hash15[8] __attribute__ ((aligned (64)));
+   allium_16way_ctx_holder ctx __attribute__ ((aligned (64)));

-   memcpy( &ctx, &allium_8way_ctx, sizeof(allium_8way_ctx) );
-   blake256_8way_update( &ctx.blake, input + (64<<3), 16 );
-   blake256_8way_close( &ctx.blake, vhash );
+   memcpy( &ctx, &allium_16way_ctx, sizeof(allium_16way_ctx) );
+   blake256_16way_update( &ctx.blake, input + (64<<4), 16 );
+   blake256_16way_close( &ctx.blake, vhash );

-   rintrlv_8x32_8x64( vhashA, vhash, 256 );
+   dintrlv_16x32( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
+                  hash8, hash9, hash10, hash11, hash12, hash13, hash14, hash15,
+                  vhash, 256 );
+   intrlv_8x64( vhashA, hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
+                256 );
+   intrlv_8x64( vhashB, hash8, hash9, hash10, hash11, hash12, hash13, hash14,
+                hash15, 256 );
+   
+//   rintrlv_8x32_8x64( vhashA, vhash, 256 );
   keccak256_8way_update( &ctx.keccak, vhashA, 32 );
-   keccak256_8way_close( &ctx.keccak, vhash );
+   keccak256_8way_close( &ctx.keccak, vhashA);
+   keccak256_8way_init( &ctx.keccak );
+   keccak256_8way_update( &ctx.keccak, vhashB, 32 );
+   keccak256_8way_close( &ctx.keccak, vhashB);

   dintrlv_8x64( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
-                 vhash, 256 );
+                 vhashA, 256 );
+   dintrlv_8x64( hash8, hash9, hash10, hash11, hash12, hash13, hash14, hash15,
+                 vhashB, 256 );

   intrlv_2x256( vhash, hash0, hash1, 256 );
   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
@@ -67,6 +99,18 @@ void allium_8way_hash( void *state, const void *input )
   intrlv_2x256( vhash, hash6, hash7, 256 );
   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
   dintrlv_2x256( hash6, hash7, vhash, 256 );
+   intrlv_2x256( vhash, hash8, hash9, 256 );
+   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
+   dintrlv_2x256( hash8, hash9, vhash, 256 );
+   intrlv_2x256( vhash, hash10, hash11, 256 );
+   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
+   dintrlv_2x256( hash10, hash11, vhash, 256 );
+   intrlv_2x256( vhash, hash12, hash13, 256 );
+   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
+   dintrlv_2x256( hash12, hash13, vhash, 256 );
+   intrlv_2x256( vhash, hash14, hash15, 256 );
+   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
+   dintrlv_2x256( hash14, hash15, vhash, 256 );
  
   intrlv_4x128( vhashA, hash0, hash1, hash2, hash3, 256 );
   intrlv_4x128( vhashB, hash4, hash5, hash6, hash7, 256 );
@@ -78,6 +122,17 @@ void allium_8way_hash( void *state, const void *input )
   dintrlv_4x128( hash0, hash1, hash2, hash3, vhashA, 256 );
   dintrlv_4x128( hash4, hash5, hash6, hash7, vhashB, 256 );

+   intrlv_4x128( vhashA, hash8, hash9, hash10, hash11, 256 );
+   intrlv_4x128( vhashB, hash12, hash13, hash14, hash15, 256 );
+
+   cube_4way_init( &ctx.cube, 256, 16, 32 );
+   cube_4way_update_close( &ctx.cube, vhashA, vhashA, 32 );
+   cube_4way_init( &ctx.cube, 256, 16, 32 );
+   cube_4way_update_close( &ctx.cube, vhashB, vhashB, 32 );
+
+   dintrlv_4x128( hash8, hash9, hash10, hash11, vhashA, 256 );
+   dintrlv_4x128( hash12, hash13, hash14, hash15, vhashB, 256 );
+
   intrlv_2x256( vhash, hash0, hash1, 256 );
   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
   dintrlv_2x256( hash0, hash1, vhash, 256 );
@@ -90,133 +145,216 @@ void allium_8way_hash( void *state, const void *input )
   intrlv_2x256( vhash, hash6, hash7, 256 );
   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
   dintrlv_2x256( hash6, hash7, vhash, 256 );
+   intrlv_2x256( vhash, hash8, hash9, 256 );
+   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
+   dintrlv_2x256( hash8, hash9, vhash, 256 );
+   intrlv_2x256( vhash, hash10, hash11, 256 );
+   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
+   dintrlv_2x256( hash10, hash11, vhash, 256 );
+   intrlv_2x256( vhash, hash12, hash13, 256 );
+   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
+   dintrlv_2x256( hash12, hash13, vhash, 256 );
+   intrlv_2x256( vhash, hash14, hash15, 256 );
+   LYRA2RE_2WAY( vhash, 32, vhash, 32, 1, 8, 8 );
+   dintrlv_2x256( hash14, hash15, vhash, 256 );

-   intrlv_8x64( vhash, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
+   intrlv_8x64( vhashA, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
                hash7, 256 );
+   intrlv_8x64( vhashB, hash8, hash9, hash10, hash11, hash12, hash13, hash14,
+                hash15, 256 );
+
+   skein256_8way_update( &ctx.skein, vhashA, 32 );
+   skein256_8way_close( &ctx.skein, vhashA );
+   skein256_8way_init( &ctx.skein );
+   skein256_8way_update( &ctx.skein, vhashB, 32 );
+   skein256_8way_close( &ctx.skein, vhashB );

-   skein256_8way_update( &ctx.skein, vhash, 32 );
-   skein256_8way_close( &ctx.skein, vhash );

   dintrlv_8x64( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
-                 vhash, 256 );
+                 vhashA, 256 );
+   dintrlv_8x64( hash8, hash9, hash10, hash11, hash12, hash13, hash14, hash15,
+                 vhashB, 256 );
+
+#if defined(__VAES__)
+
+   intrlv_4x128( vhash, hash0, hash1, hash2, hash3, 256 );
+
+   groestl256_4way_update_close( &ctx.groestl, vhash, vhash, 256 );
+
+   dintrlv_4x128( state, state+32, state+64, state+96, vhash, 256 );
+   intrlv_4x128( vhash, hash4, hash5, hash6, hash7, 256 );
+
+   groestl256_4way_init( &ctx.groestl, 32 );
+   groestl256_4way_update_close( &ctx.groestl, vhash, vhash, 256 );
+   
+   dintrlv_4x128( state+128, state+160, state+192, state+224, vhash, 256 );
+   intrlv_4x128( vhash, hash8, hash9, hash10, hash11, 256 );
+
+   groestl256_4way_init( &ctx.groestl, 32 );
+   groestl256_4way_update_close( &ctx.groestl, vhash, vhash, 256 );
+
+   dintrlv_4x128( state+256, state+288, state+320, state+352, vhash, 256 );
+   intrlv_4x128( vhash, hash12, hash13, hash14, hash15, 256 );
+
+   groestl256_4way_init( &ctx.groestl, 32 );
+   groestl256_4way_update_close( &ctx.groestl, vhash, vhash, 256 );
+ 
+   dintrlv_4x128( state+384, state+416, state+448, state+480, vhash, 256 );
+   
+#else

   update_and_final_groestl256( &ctx.groestl, state, hash0, 256 );
-   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
           sizeof(hashState_groestl256) );
   update_and_final_groestl256( &ctx.groestl, state+32, hash1, 256 );
-   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
           sizeof(hashState_groestl256) );
   update_and_final_groestl256( &ctx.groestl, state+64, hash2, 256 );
-   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
           sizeof(hashState_groestl256) );
   update_and_final_groestl256( &ctx.groestl, state+96, hash3, 256 );
-   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
           sizeof(hashState_groestl256) );
   update_and_final_groestl256( &ctx.groestl, state+128, hash4, 256 );
-   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
           sizeof(hashState_groestl256) );
   update_and_final_groestl256( &ctx.groestl, state+160, hash5, 256 );
-   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
           sizeof(hashState_groestl256) );
   update_and_final_groestl256( &ctx.groestl, state+192, hash6, 256 );
-   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
           sizeof(hashState_groestl256) );
   update_and_final_groestl256( &ctx.groestl, state+224, hash7, 256 );
-   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, state+256, hash8, 256 );
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, state+288, hash9, 256 );
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, state+320, hash10, 256 );
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, state+352, hash11, 256 );
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, state+384, hash12, 256 );
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, state+416, hash13, 256 );
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, state+448, hash14, 256 );
+   memcpy( &ctx.groestl, &allium_16way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, state+480, hash15, 256 );
+
+#endif
 }

-int scanhash_allium_8way( struct work *work, uint32_t max_nonce,
+int scanhash_allium_16way( struct work *work, uint32_t max_nonce,
                             uint64_t *hashes_done, struct thr_info *mythr )
 {
-   uint32_t hash[8*8] __attribute__ ((aligned (128)));
-   uint32_t vdata[20*8] __attribute__ ((aligned (64)));
+   uint32_t hash[8*16] __attribute__ ((aligned (128)));
+   uint32_t vdata[20*16] __attribute__ ((aligned (64)));
   uint32_t *pdata = work->data;
   uint32_t *ptarget = work->target;
   const uint32_t first_nonce = pdata[19];
   uint32_t n = first_nonce;
-   const uint32_t last_nonce = max_nonce - 8;
-   const uint32_t Htarg = ptarget[7];
-   __m256i  *noncev = (__m256i*)vdata + 19;   // aligned
-   int thr_id = mythr->id;  // thr_id arg is deprecated
+   const uint32_t last_nonce = max_nonce - 16;
+   __m512i  *noncev = (__m512i*)vdata + 19;   // aligned
+   const int thr_id = mythr->id;
+   const bool bench = opt_benchmark;

-   if ( opt_benchmark )
-      ( (uint32_t*)ptarget )[7] = 0x0000ff;
+   if ( bench ) ( (uint32_t*)ptarget )[7] = 0x0000ff;

-   mm256_bswap32_intrlv80_8x32( vdata, pdata );
-   blake256_8way_init( &allium_8way_ctx.blake );
-   blake256_8way_update( &allium_8way_ctx.blake, vdata, 64 );
+   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( &allium_16way_ctx.blake );
+   blake256_16way_update( &allium_16way_ctx.blake, vdata, 64 );

   do {
-     *noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
-                                                 n+3, n+2, n+1, n ) );
+     allium_16way_hash( hash, vdata );

-     allium_8way_hash( hash, vdata );
-     pdata[19] = n;
-
-     for ( int lane = 0; lane < 8; lane++ ) if ( (hash+(lane<<3))[7] <= Htarg )
+     for ( int lane = 0; lane < 16; lane++ ) 
+     if unlikely( valid_hash( hash+(lane<<3), ptarget ) && !bench )
     {
-        if ( fulltest( hash+(lane<<3), ptarget ) && !opt_benchmark )
-        {
-           pdata[19] = n + lane;
-           submit_lane_solution( work, hash+(lane<<3), mythr, lane );
-         }
+         pdata[19] = bswap_32( n + lane );
+         submit_lane_solution( work, hash+(lane<<3), mythr, lane );
     }
-     n += 8;
+     *noncev = _mm512_add_epi32( *noncev, m512_const1_32( 16 ) );
+     n += 16;
   } while ( (n < last_nonce) && !work_restart[thr_id].restart);
-
   *hashes_done = n - first_nonce;
   return 0;
 }

-
-#elif defined (ALLIUM_4WAY)  
-
+#elif defined (ALLIUM_8WAY)  

 typedef struct {
-   blake256_4way_context     blake;
+   blake256_8way_context     blake;
   keccak256_4way_context    keccak;
   cubehashParam             cube;
   skein256_4way_context     skein;
   hashState_groestl256      groestl;

-} allium_4way_ctx_holder;
+} allium_8way_ctx_holder;

-static __thread allium_4way_ctx_holder allium_4way_ctx;
+static __thread allium_8way_ctx_holder allium_8way_ctx;

-bool init_allium_4way_ctx()
+bool init_allium_8way_ctx()
 {
-   keccak256_4way_init( &allium_4way_ctx.keccak );
-   cubehashInit( &allium_4way_ctx.cube, 256, 16, 32 );
-   skein256_4way_init( &allium_4way_ctx.skein );
-   init_groestl256( &allium_4way_ctx.groestl, 32 );
+   keccak256_4way_init( &allium_8way_ctx.keccak );
+   cubehashInit( &allium_8way_ctx.cube, 256, 16, 32 );
+   skein256_4way_init( &allium_8way_ctx.skein );
+   init_groestl256( &allium_8way_ctx.groestl, 32 );
   return true;
 }

-void allium_4way_hash( void *state, const void *input )
+void allium_8way_hash( void *hash, const void *input )
 {
-   uint32_t hash0[8] __attribute__ ((aligned (64)));
-   uint32_t hash1[8] __attribute__ ((aligned (32)));
-   uint32_t hash2[8] __attribute__ ((aligned (32)));
-   uint32_t hash3[8] __attribute__ ((aligned (32)));
-   uint32_t vhash32[8*4] __attribute__ ((aligned (64)));
-   uint32_t vhash64[8*4] __attribute__ ((aligned (64)));
-   allium_4way_ctx_holder ctx __attribute__ ((aligned (64))); 
+   uint64_t vhashA[4*8] __attribute__ ((aligned (64)));
+   uint64_t vhashB[4*8] __attribute__ ((aligned (64)));
+//   uint64_t hash[4*8] __attribute__ ((aligned (64)));
+   uint64_t *hash0 = (uint64_t*)hash;
+   uint64_t *hash1 = (uint64_t*)hash+ 4;
+   uint64_t *hash2 = (uint64_t*)hash+ 8;
+   uint64_t *hash3 = (uint64_t*)hash+12;
+   uint64_t *hash4 = (uint64_t*)hash+16;
+   uint64_t *hash5 = (uint64_t*)hash+20;
+   uint64_t *hash6 = (uint64_t*)hash+24;
+   uint64_t *hash7 = (uint64_t*)hash+28;
+   allium_8way_ctx_holder ctx __attribute__ ((aligned (64))); 

-   memcpy( &ctx, &allium_4way_ctx, sizeof(allium_4way_ctx) );
-   blake256_4way_update( &ctx.blake, input + (64<<2), 16 );
-   blake256_4way_close( &ctx.blake, vhash32 );
+   memcpy( &ctx, &allium_8way_ctx, sizeof(allium_8way_ctx) );
+   blake256_8way_update( &ctx.blake, input + (64<<3), 16 );
+   blake256_8way_close( &ctx.blake, vhashA );

-   rintrlv_4x32_4x64( vhash64, vhash32, 256 );
-   keccak256_4way_update( &ctx.keccak, vhash64, 32 );
-   keccak256_4way_close( &ctx.keccak, vhash64 );
+   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 );

-   dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
+   keccak256_4way_update( &ctx.keccak, vhashA, 32 );
+   keccak256_4way_close( &ctx.keccak, vhashA );
+   keccak256_4way_init( &ctx.keccak );
+   keccak256_4way_update( &ctx.keccak, vhashB, 32 );
+   keccak256_4way_close( &ctx.keccak, vhashB );
+
+   dintrlv_4x64( hash0, hash1, hash2, hash3, vhashA, 256 );
+   dintrlv_4x64( hash4, hash5, hash6, hash7, vhashB, 256 );

   LYRA2RE( hash0, 32, hash0, 32, hash0, 32, 1, 8, 8 );
   LYRA2RE( hash1, 32, hash1, 32, hash1, 32, 1, 8, 8 );
   LYRA2RE( hash2, 32, hash2, 32, hash2, 32, 1, 8, 8 );
   LYRA2RE( hash3, 32, hash3, 32, hash3, 32, 1, 8, 8 );
+   LYRA2RE( hash4, 32, hash4, 32, hash4, 32, 1, 8, 8 );
+   LYRA2RE( hash5, 32, hash5, 32, hash5, 32, 1, 8, 8 );
+   LYRA2RE( hash6, 32, hash6, 32, hash6, 32, 1, 8, 8 );
+   LYRA2RE( hash7, 32, hash7, 32, hash7, 32, 1, 8, 8 );

   cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*)hash0, 32 );
   cubehashInit( &ctx.cube, 256, 16, 32 );
@@ -225,69 +363,97 @@ void allium_4way_hash( void *state, const void *input )
   cubehashUpdateDigest( &ctx.cube, (byte*)hash2, (const byte*)hash2, 32 );
   cubehashInit( &ctx.cube, 256, 16, 32 );
   cubehashUpdateDigest( &ctx.cube, (byte*)hash3, (const byte*)hash3, 32 );
+   cubehashInit( &ctx.cube, 256, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*)hash4, (const byte*)hash4, 32 );
+   cubehashInit( &ctx.cube, 256, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*)hash5, (const byte*)hash5, 32 );
+   cubehashInit( &ctx.cube, 256, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*)hash6, (const byte*)hash6, 32 );
+   cubehashInit( &ctx.cube, 256, 16, 32 );
+   cubehashUpdateDigest( &ctx.cube, (byte*)hash7, (const byte*)hash7, 32 );

   LYRA2RE( hash0, 32, hash0, 32, hash0, 32, 1, 8, 8 );
   LYRA2RE( hash1, 32, hash1, 32, hash1, 32, 1, 8, 8 );
   LYRA2RE( hash2, 32, hash2, 32, hash2, 32, 1, 8, 8 );
   LYRA2RE( hash3, 32, hash3, 32, hash3, 32, 1, 8, 8 );
+   LYRA2RE( hash4, 32, hash4, 32, hash4, 32, 1, 8, 8 );
+   LYRA2RE( hash5, 32, hash5, 32, hash5, 32, 1, 8, 8 );
+   LYRA2RE( hash6, 32, hash6, 32, hash6, 32, 1, 8, 8 );
+   LYRA2RE( hash7, 32, hash7, 32, hash7, 32, 1, 8, 8 );

-   intrlv_4x64( vhash64, hash0, hash1, hash2, hash3, 256 );
+   intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 256 );
+   intrlv_4x64( vhashB, hash4, hash5, hash6, hash7, 256 );

-   skein256_4way_update( &ctx.skein, vhash64, 32 );
-   skein256_4way_close( &ctx.skein, vhash64 );
+   skein256_4way_update( &ctx.skein, vhashA, 32 );
+   skein256_4way_close( &ctx.skein, vhashA );
+   skein256_4way_init( &ctx.skein );
+   skein256_4way_update( &ctx.skein, vhashB, 32 );
+   skein256_4way_close( &ctx.skein, vhashB );

-   dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
+   dintrlv_4x64( hash0, hash1, hash2, hash3, vhashA, 256 );
+   dintrlv_4x64( hash4, hash5, hash6, hash7, vhashB, 256 );

-   update_and_final_groestl256( &ctx.groestl, state, hash0, 256 );
-   memcpy( &ctx.groestl, &allium_4way_ctx.groestl,
+   update_and_final_groestl256( &ctx.groestl, hash0, hash0, 256 );
+   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
           sizeof(hashState_groestl256) );
-   update_and_final_groestl256( &ctx.groestl, state+32, hash1, 256 );
-   memcpy( &ctx.groestl, &allium_4way_ctx.groestl,
+   update_and_final_groestl256( &ctx.groestl, hash1, hash1, 256 );
+   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
           sizeof(hashState_groestl256) );
-   update_and_final_groestl256( &ctx.groestl, state+64, hash2, 256 );
-   memcpy( &ctx.groestl, &allium_4way_ctx.groestl,
+   update_and_final_groestl256( &ctx.groestl, hash2, hash2, 256 );
+   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
           sizeof(hashState_groestl256) );
-   update_and_final_groestl256( &ctx.groestl, state+96, hash3, 256 );
+   update_and_final_groestl256( &ctx.groestl, hash3, hash3, 256 );
+   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, hash4, hash4, 256 );
+   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, hash5, hash5, 256 );
+   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, hash6, hash6, 256 );
+   memcpy( &ctx.groestl, &allium_8way_ctx.groestl,
+           sizeof(hashState_groestl256) );
+   update_and_final_groestl256( &ctx.groestl, hash7, hash7, 256 );
 }

-int scanhash_allium_4way( struct work *work, uint32_t max_nonce,
+int scanhash_allium_8way( struct work *work, uint32_t max_nonce,
                             uint64_t *hashes_done, struct thr_info *mythr )
 {
-   uint32_t hash[8*4] __attribute__ ((aligned (64)));
-   uint32_t vdata[20*4] __attribute__ ((aligned (64)));
+   uint64_t hash[4*8] __attribute__ ((aligned (64)));
+   uint32_t vdata[20*8] __attribute__ ((aligned (64)));
   uint32_t *pdata = work->data;
-   uint32_t *ptarget = work->target;
+   uint64_t *ptarget = (uint64_t*)work->target;
   const uint32_t first_nonce = pdata[19];
+   const uint32_t last_nonce = max_nonce - 8;
   uint32_t n = first_nonce;
-   const uint32_t Htarg = ptarget[7];
-   __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
-   int thr_id = mythr->id;  // thr_id arg is deprecated
+   __m256i  *noncev = (__m256i*)vdata + 19;   // aligned
+   const int thr_id = mythr->id;  
+   const bool bench = opt_benchmark;

-   if ( opt_benchmark )
-      ( (uint32_t*)ptarget )[7] = 0x0000ff;
+   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 );

-   mm128_bswap32_intrlv80_4x32( vdata, pdata );
-   blake256_4way_init( &allium_4way_ctx.blake );
-   blake256_4way( &allium_4way_ctx.blake, vdata, 64 );
+   blake256_8way_init( &allium_8way_ctx.blake );
+   blake256_8way_update( &allium_8way_ctx.blake, vdata, 64 );

   do {
-     *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
+     allium_8way_hash( hash, vdata );

-     allium_4way_hash( hash, vdata );
-     pdata[19] = n;
-
-     for ( int lane = 0; lane < 4; lane++ ) if ( (hash+(lane<<3))[7] <= Htarg )
+     for ( int lane = 0; lane < 8; lane++ )
     {
-        if ( fulltest( hash+(lane<<3), ptarget ) && !opt_benchmark )
+        const uint64_t *lane_hash = hash + (lane<<2);
+        if unlikely( valid_hash( lane_hash, ptarget ) && !bench )
        {
-           pdata[19] = n + lane;
-           submit_lane_solution( work, hash+(lane<<3), mythr, lane );
-         }
+           pdata[19] = bswap_32( n + lane );
+           submit_lane_solution( work, lane_hash, mythr, lane );
+        }
     }
-     n += 4;
-   } while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
-
-   *hashes_done = n - first_nonce + 1;
+     n += 8;
+     *noncev = _mm256_add_epi32( *noncev, m256_const1_32( 8 ) );
+   } while likely( (n <= last_nonce) && !work_restart[thr_id].restart );
+   pdata[19] = n;
+   *hashes_done = n - first_nonce;
   return 0;
 }

--- a/algo/lyra2/lyra2-gate.c
+++ b/algo/lyra2/lyra2-gate.c
@@ -78,7 +78,7 @@ bool register_lyra2rev3_algo( algo_gate_t* gate )
  gate->scanhash  = (void*)&scanhash_lyra2rev3;
  gate->hash      = (void*)&lyra2rev3_hash;
 #endif
-  gate->optimizations = SSE2_OPT | SSE42_OPT | AVX2_OPT | AVX512_OPT;
+  gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT;
  gate->miner_thread_init = (void*)&lyra2rev3_thread_init;
  opt_target_factor = 256.0;
  return true;
@@ -119,7 +119,7 @@ bool register_lyra2rev2_algo( algo_gate_t* gate )
  gate->scanhash  = (void*)&scanhash_lyra2rev2;
  gate->hash      = (void*)&lyra2rev2_hash;
 #endif
-  gate->optimizations = SSE2_OPT | AES_OPT | SSE42_OPT | AVX2_OPT | AVX512_OPT;
+  gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT;
  gate->miner_thread_init = (void*)&lyra2rev2_thread_init;
  opt_target_factor = 256.0;
  return true;
@@ -146,7 +146,7 @@ bool register_lyra2z_algo( algo_gate_t* gate )
  gate->scanhash   = (void*)&scanhash_lyra2z;
  gate->hash       = (void*)&lyra2z_hash;
 #endif
-  gate->optimizations = SSE42_OPT | AVX2_OPT | AVX512_OPT;
+  gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT;
  opt_target_factor = 256.0;
  return true;
 };
@@ -165,7 +165,7 @@ bool register_lyra2h_algo( algo_gate_t* gate )
  gate->scanhash   = (void*)&scanhash_lyra2h;
  gate->hash       = (void*)&lyra2h_hash;
 #endif
-  gate->optimizations = SSE42_OPT | AVX2_OPT;
+  gate->optimizations = SSE2_OPT | AVX2_OPT;
  opt_target_factor = 256.0;
  return true;
 };
@@ -174,20 +174,20 @@ bool register_lyra2h_algo( algo_gate_t* gate )

 bool register_allium_algo( algo_gate_t* gate )
 {
-#if defined (ALLIUM_8WAY)
+#if defined (ALLIUM_16WAY)
+  gate->miner_thread_init = (void*)&init_allium_16way_ctx;
+  gate->scanhash  = (void*)&scanhash_allium_16way;
+  gate->hash      = (void*)&allium_16way_hash;
+#elif defined (ALLIUM_8WAY)
  gate->miner_thread_init = (void*)&init_allium_8way_ctx;
  gate->scanhash  = (void*)&scanhash_allium_8way;
  gate->hash      = (void*)&allium_8way_hash;
-#elif defined (ALLIUM_4WAY)
-  gate->miner_thread_init = (void*)&init_allium_4way_ctx;
-  gate->scanhash  = (void*)&scanhash_allium_4way;
-  gate->hash      = (void*)&allium_4way_hash;
 #else
  gate->miner_thread_init = (void*)&init_allium_ctx;
  gate->scanhash  = (void*)&scanhash_allium;
  gate->hash      = (void*)&allium_hash;
 #endif
-  gate->optimizations = SSE2_OPT | AES_OPT | SSE42_OPT | AVX2_OPT | AVX512_OPT;
+  gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
  opt_target_factor = 256.0;
  return true;
 };
@@ -220,7 +220,7 @@ void phi2_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
   // Assemble block header
   algo_gate.build_block_header( g_work, le32dec( sctx->job.version ),
                  (uint32_t*) sctx->job.prevhash, (uint32_t*) merkle_tree,
-                  le32dec( sctx->job.ntime ), le32dec(sctx->job.nbits) );
+                  le32dec( sctx->job.ntime ), le32dec(sctx->job.nbits), NULL );
   for ( t = 0; t < 16; t++ )
      g_work->data[ 20+t ] = ((uint32_t*)sctx->job.extra)[t];
 }
@@ -229,7 +229,7 @@ void phi2_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
 bool register_phi2_algo( algo_gate_t* gate )
 {
 //   init_phi2_ctx();
-   gate->optimizations = SSE2_OPT | AES_OPT | SSE42_OPT | AVX2_OPT | AVX512_OPT;
+   gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT;
   gate->get_work_data_size = (void*)&phi2_get_work_data_size;
   gate->decode_extra_data  = (void*)&phi2_decode_extra_data;
   gate->build_extraheader  = (void*)&phi2_build_extraheader;
--- a/algo/lyra2/lyra2-gate.h
+++ b/algo/lyra2/lyra2-gate.h
@@ -153,27 +153,27 @@ bool lyra2h_thread_init();
 //////////////////////////////////

 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
-  #define ALLIUM_8WAY 1
+  #define ALLIUM_16WAY 1
 #elif defined(__AVX2__) && defined(__AES__)
-  #define ALLIUM_4WAY 1
+  #define ALLIUM_8WAY 1
 #endif

 bool register_allium_algo( algo_gate_t* gate );

-#if defined(ALLIUM_8WAY)
+#if defined(ALLIUM_16WAY)
+
+void allium_16way_hash( void *state, const void *input );
+int scanhash_allium_16way( struct work *work, uint32_t max_nonce,
+                          uint64_t *hashes_done, struct thr_info *mythr );
+bool init_allium_16way_ctx();
+
+#elif defined(ALLIUM_8WAY)

 void allium_8way_hash( void *state, const void *input );
 int scanhash_allium_8way( struct work *work, uint32_t max_nonce,
                          uint64_t *hashes_done, struct thr_info *mythr );
 bool init_allium_8way_ctx();

-#elif defined(ALLIUM_4WAY)
-
-void allium_4way_hash( void *state, const void *input );
-int scanhash_allium_4way( struct work *work, uint32_t max_nonce,
-                          uint64_t *hashes_done, struct thr_info *mythr );
-bool init_allium_4way_ctx();
-
 #else

 void allium_hash( void *state, const void *input );
--- a/algo/lyra2/lyra2-hash-2way.c
+++ b/algo/lyra2/lyra2-hash-2way.c
@@ -575,4 +575,138 @@ int LYRA2RE_2WAY( void *K, uint64_t kLen, const void *pwd,
   return 0;
 }

+int LYRA2X_2WAY( void *K, uint64_t kLen, const void *pwd,
+                  const uint64_t pwdlen, const uint64_t timeCost,
+                  const uint64_t nRows, const uint64_t nCols )
+{
+   //====================== Basic variables ============================//
+   uint64_t _ALIGN(256) state[32];
+   int64_t row = 2; //index of row to be processed
+   int64_t prev = 1; //index of prev (last row ever computed/modified)
+   int64_t rowa0 = 0;
+   int64_t rowa1 = 0;
+   int64_t tau; //Time Loop iterator
+   int64_t step = 1; //Visitation step (used during Setup and Wandering phases)
+   int64_t window = 2; //Visitation window (used to define which rows can be revisited during Setup)
+   int64_t gap = 1; //Modifier to the step, assuming the values 1 or -1
+   int64_t i; //auxiliary iteration counter
+   //====================================================================/
+
+   //=== Initializing the Memory Matrix and pointers to it =============//
+   //Tries to allocate enough space for the whole memory matrix
+
+   const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * nCols;
+   const int64_t ROW_LEN_BYTES = ROW_LEN_INT64 * 8;
+   // for Lyra2REv2, nCols = 4, v1 was using 8
+   const int64_t BLOCK_LEN = (nCols == 4) ? BLOCK_LEN_BLAKE2_SAFE_INT64
+                                          : BLOCK_LEN_BLAKE2_SAFE_BYTES;
+
+   i = (int64_t)ROW_LEN_BYTES * nRows;
+   uint64_t *wholeMatrix = _mm_malloc( 2*i, 64 );
+   if (wholeMatrix == NULL)
+      return -1;
+
+   memset_zero_512( (__m512i*)wholeMatrix, i>>5 );
+
+   uint64_t *ptrWord = wholeMatrix;
+   uint64_t *pw = (uint64_t*)pwd;
+
+   //First, we clean enough blocks for the password, salt, basil and padding
+   int64_t nBlocksInput = ( ( pwdlen + pwdlen + 6 * sizeof(uint64_t) )
+                              / BLOCK_LEN_BLAKE2_SAFE_BYTES ) + 1;
+
+   uint64_t *ptr = wholeMatrix;
+
+   memcpy( ptr, pw, 2*pwdlen ); // password 
+   ptr += pwdlen>>2;
+   memcpy( ptr, pw, 2*pwdlen ); // password lane 1
+   ptr += pwdlen>>2;
+
+   // now build the rest interleaving on the fly.
+
+   ptr[0] = ptr[ 4] = kLen;
+   ptr[1] = ptr[ 5] = pwdlen;
+   ptr[2] = ptr[ 6] = pwdlen;   // saltlen
+   ptr[3] = ptr[ 7] = timeCost;
+   ptr[8] = ptr[12] = nRows;
+   ptr[9] = ptr[13] = nCols;
+   ptr[10] = ptr[14] = 0x80;
+   ptr[11] = ptr[15] = 0x0100000000000000;
+
+   absorbBlockBlake2Safe_2way( state, ptrWord, nBlocksInput, BLOCK_LEN );
+
+   //Initializes M[0] and M[1]
+   reducedSqueezeRow0_2way( state, &wholeMatrix[0], nCols ); //The locally copied password is most likely overwritten here
+
+   reducedDuplexRow1_2way( state, &wholeMatrix[0],
+                                  &wholeMatrix[ 2 * ROW_LEN_INT64], nCols );
+
+   do
+   {
+      //M[row] = rand; //M[row*] = M[row*] XOR rotW(rand)
+
+      reducedDuplexRowSetup_2way( state, &wholeMatrix[ 2* prev*ROW_LEN_INT64 ],
+                                         &wholeMatrix[ 2* rowa0*ROW_LEN_INT64 ],
+                                         &wholeMatrix[ 2* row*ROW_LEN_INT64 ],
+                                         nCols );
+
+      //updates the value of row* (deterministically picked during Setup))
+      rowa0 = (rowa0 + step) & (window - 1);
+      //update prev: it now points to the last row ever computed
+
+      prev = row;
+      //updates row: goes to the next row to be computed
+      row++;
+
+      //Checks if all rows in the window where visited.
+      if (rowa0 == 0)
+      {
+         step = window + gap; //changes the step: approximately doubles its value
+         window *= 2; //doubles the size of the re-visitation window
+         gap = -gap; //inverts the modifier to the step
+      }
+
+   } while (row < nRows);
+
+   //===================== Wandering Phase =============================//
+   row = 0; //Resets the visitation to the first row of the memory matrix
+   for (tau = 1; tau <= timeCost; tau++)
+   {
+      step = ((tau & 1) == 0) ? -1 : (nRows >> 1) - 1;
+      do
+      {
+        rowa0 = state[ 0 ] & (unsigned int)(nRows-1);
+        rowa1 = state[ 4 ] & (unsigned int)(nRows-1);
+
+        reducedDuplexRow_2way_X( state, &wholeMatrix[ 2* prev * ROW_LEN_INT64 ],
+                                      &wholeMatrix[ 2* rowa0 * ROW_LEN_INT64 ],
+                                      &wholeMatrix[ 2* rowa1 * ROW_LEN_INT64 ],
+                                      &wholeMatrix[ 2* row *ROW_LEN_INT64 ],
+                                      nCols );
+
+           //update prev: it now points to the last row ever computed
+           prev = row;
+
+           //updates row: goes to the next row to be computed
+           //----------------------------------------------------
+           row = (row + step) & (unsigned int)(nRows-1); //(USE THIS IF nRows IS A POWER OF 2)
+           //row = (row + step) % nRows; //(USE THIS FOR THE "GENERIC" CASE)
+           //----------------------------------------------------
+
+       } while (row != 0);
+   }
+
+   //===================== Wrap-up Phase ===============================//
+   //Absorbs the last block of the memory matrix
+   absorbBlock_2way( state, &wholeMatrix[ 2 * rowa0 *ROW_LEN_INT64],
+                            &wholeMatrix[ 2 * rowa1 *ROW_LEN_INT64] );
+   //Squeezes the key
+   squeeze_2way( state, K, (unsigned int) kLen );
+
+   //================== Freeing the memory =============================//
+   _mm_free(wholeMatrix);
+
+   return 0;
+}
+      
 #endif
--- a/algo/lyra2/lyra2.h
+++ b/algo/lyra2/lyra2.h
@@ -74,6 +74,9 @@ int LYRA2REV3_2WAY( uint64_t*, void *K, uint64_t kLen, const void *pwd,
 int LYRA2Z_2WAY( uint64_t*, void *K, uint64_t kLen, const void *pwd,
          uint64_t pwdlen, uint64_t timeCost, uint64_t nRows, uint64_t nCols );

+int LYRA2X_2WAY( void *K, uint64_t kLen, const void *pwd, uint64_t pwdlen,
+                  uint64_t timeCost, uint64_t nRows, uint64_t nCols );
+
 #endif

 #endif /* LYRA2_H_ */
--- a/algo/lyra2/lyra2h-4way.c
+++ b/algo/lyra2/lyra2h-4way.c
@@ -33,7 +33,7 @@ void lyra2h_4way_hash( void *state, const void *input )
     blake256_4way_context ctx_blake __attribute__ ((aligned (64)));

     memcpy( &ctx_blake, &l2h_4way_blake_mid, sizeof l2h_4way_blake_mid );
-     blake256_4way( &ctx_blake, input + (64*4), 16 );
+     blake256_4way_update( &ctx_blake, input + (64*4), 16 );
     blake256_4way_close( &ctx_blake, vhash );

     dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
--- a/algo/lyra2/sponge-2way.c
+++ b/algo/lyra2/sponge-2way.c
@@ -196,7 +196,6 @@ inline void reducedDuplexRowSetup_2way( uint64_t *State, uint64_t *rowIn,
    __m512i* in    = (__m512i*)rowIn;
    __m512i* inout = (__m512i*)rowInOut;
    __m512i* out   = (__m512i*)rowOut + ( (nCols-1) * BLOCK_LEN_M256I );
-    __m512i  t0, t1, t2;

    state0 = _mm512_load_si512( (__m512i*)State     );
    state1 = _mm512_load_si512( (__m512i*)State + 1 );
@@ -218,24 +217,27 @@ inline void reducedDuplexRowSetup_2way( uint64_t *State, uint64_t *rowIn,
       out[1] = _mm512_xor_si512( state1, in[1] );
       out[2] = _mm512_xor_si512( state2, in[2] );

-       //M[row*][col] = M[row*][col] XOR rotW(rand)
-       t0 = _mm512_permutex_epi64( state0, 0x93 );
-       t1 = _mm512_permutex_epi64( state1, 0x93 );
-       t2 = _mm512_permutex_epi64( state2, 0x93 );
+      {
+        register __m512i t0, t1, t2;
+       
+        //M[row*][col] = M[row*][col] XOR rotW(rand)
+        t0 = _mm512_permutex_epi64( state0, 0x93 );
+        t1 = _mm512_permutex_epi64( state1, 0x93 );
+        t2 = _mm512_permutex_epi64( state2, 0x93 );

-       inout[0] = _mm512_xor_si512( inout[0],
-                                 _mm512_mask_blend_epi32( 0x0303, t0, t2 ) );
-       inout[1] = _mm512_xor_si512( inout[1],
-                                 _mm512_mask_blend_epi32( 0x0303, t1, t0 ) );
-       inout[2] = _mm512_xor_si512( inout[2],
-                                 _mm512_mask_blend_epi32( 0x0303, t2, t1 ) );
+        inout[0] = _mm512_xor_si512( inout[0],
+                                 _mm512_mask_blend_epi64( 0x11, t0, t2 ) );
+        inout[1] = _mm512_xor_si512( inout[1],
+                                 _mm512_mask_blend_epi64( 0x11, t1, t0 ) );
+        inout[2] = _mm512_xor_si512( inout[2],
+                                 _mm512_mask_blend_epi64( 0x11, t2, t1 ) );
+      }

-
-       //Inputs: next column (i.e., next block in sequence)
-       in    += BLOCK_LEN_M256I;
-       inout += BLOCK_LEN_M256I;
-       //Output: goes to previous column
-       out   -= BLOCK_LEN_M256I;
+      //Inputs: next column (i.e., next block in sequence)
+      in    += BLOCK_LEN_M256I;
+      inout += BLOCK_LEN_M256I;
+      //Output: goes to previous column
+      out   -= BLOCK_LEN_M256I;
    }

    _mm512_store_si512( (__m512i*)State,     state0 );
@@ -244,12 +246,235 @@ inline void reducedDuplexRowSetup_2way( uint64_t *State, uint64_t *rowIn,
    _mm512_store_si512( (__m512i*)State + 3, state3 );
 }

-// big ugly workaound for pointer aliasing, use a union of pointers.
-// Access matrix using m512i for in and out, m256i for inout
+// reduced duplex row has three version depending on rows inout.
+// If they are the same the fastest version can be used, equivalent to 
+// linear version.
+// If either rowinout overlaps with rowout the slowest version is used,
+// to refresh local data after overwriting rowout.
+// Otherwise the normal version is used, slower than unified, faster than
+// overlap.
+//
+// The likelyhood of each case depends on the number of rows. More rows
+// means unified and overlap are both less likely.
+// Unified has a 1 in Nrows chances,
+// Overlap has 2 in Nrows chance reduced to 1 in Nrows because if both
+// overlap it's unified.
+// As a result normal is Nrows-2 / Nrows.
+// for 4 rows: 1 unified, 2 overlap, 1 normal.
+// for 8 rows: 1 unified, 2 overlap, 56 normal.

-inline void reducedDuplexRow_2way( uint64_t *State, uint64_t *rowIn,
-                            uint64_t *rowInOut0, uint64_t *rowInOut1,
+static inline void reducedDuplexRow_2way_normal( uint64_t *State,
+                   uint64_t *rowIn, uint64_t *rowInOut0, uint64_t *rowInOut1,
                            uint64_t *rowOut, uint64_t nCols)
+{
+   int i;
+   register __m512i state0, state1, state2, state3;
+   __m512i *in = (__m512i*)rowIn;
+   __m512i *inout0 = (__m512i*)rowInOut0;
+   __m512i *inout1 = (__m512i*)rowInOut1;
+   __m512i *out = (__m512i*)rowOut;
+   register __m512i io0, io1, io2;
+
+   state0 = _mm512_load_si512( (__m512i*)State     );
+   state1 = _mm512_load_si512( (__m512i*)State + 1 );
+   state2 = _mm512_load_si512( (__m512i*)State + 2 );
+   state3 = _mm512_load_si512( (__m512i*)State + 3 );
+
+   for ( i = 0; i < nCols; i++ )
+   {
+     //Absorbing "M[prev] [+] M[row*]"
+     io0 = _mm512_mask_blend_epi64( 0xf0,
+                                    _mm512_load_si512( (__m512i*)inout0 ),
+                                    _mm512_load_si512( (__m512i*)inout1 ) );
+     io1 = _mm512_mask_blend_epi64( 0xf0,
+                                    _mm512_load_si512( (__m512i*)inout0 +1 ),
+                                    _mm512_load_si512( (__m512i*)inout1 +1 ) );
+     io2 = _mm512_mask_blend_epi64( 0xf0,
+                                    _mm512_load_si512( (__m512i*)inout0 +2 ),
+                                    _mm512_load_si512( (__m512i*)inout1 +2 ) );
+
+     state0 = _mm512_xor_si512( state0, _mm512_add_epi64( in[0], io0 ) );
+     state1 = _mm512_xor_si512( state1, _mm512_add_epi64( in[1], io1 ) );
+     state2 = _mm512_xor_si512( state2, _mm512_add_epi64( in[2], io2 ) );
+
+     //Applies the reduced-round transformation f to the sponge's state
+     LYRA_ROUND_2WAY_AVX512( state0, state1, state2, state3 );
+
+     {
+       register __m512i t0, t1, t2;
+
+       //M[rowInOut][col] = M[rowInOut][col] XOR rotW(rand)
+       t0 = _mm512_permutex_epi64( state0, 0x93 );
+       t1 = _mm512_permutex_epi64( state1, 0x93 );
+       t2 = _mm512_permutex_epi64( state2, 0x93 );
+
+       io0 = _mm512_xor_si512( io0, _mm512_mask_blend_epi64( 0x11, t0, t2 ) );
+       io1 = _mm512_xor_si512( io1, _mm512_mask_blend_epi64( 0x11, t1, t0 ) );
+       io2 = _mm512_xor_si512( io2, _mm512_mask_blend_epi64( 0x11, t2, t1 ) );
+
+       //M[rowOut][col] = M[rowOut][col] XOR rand
+       out[0] = _mm512_xor_si512( out[0], state0 );
+       out[1] = _mm512_xor_si512( out[1], state1 );
+       out[2] = _mm512_xor_si512( out[2], state2 );
+     }
+
+     _mm512_mask_store_epi64( inout0,    0x0f, io0 );
+     _mm512_mask_store_epi64( inout1,    0xf0, io0 );
+     _mm512_mask_store_epi64( inout0 +1, 0x0f, io1 );
+     _mm512_mask_store_epi64( inout1 +1, 0xf0, io1 );
+     _mm512_mask_store_epi64( inout0 +2, 0x0f, io2 );
+     _mm512_mask_store_epi64( inout1 +2, 0xf0, io2 );
+
+      //Goes to next block
+      in     += BLOCK_LEN_M256I;
+      inout0 += BLOCK_LEN_M256I;
+      inout1 += BLOCK_LEN_M256I;
+      out    += BLOCK_LEN_M256I;
+   }
+
+   _mm512_store_si512( (__m512i*)State,     state0 );
+   _mm512_store_si512( (__m512i*)State + 1, state1 );
+   _mm512_store_si512( (__m512i*)State + 2, state2 );
+   _mm512_store_si512( (__m512i*)State + 3, state3 );
+}
+
+static inline void reducedDuplexRow_2way_overlap( uint64_t *State,
+                   uint64_t *rowIn, uint64_t *rowInOut0, uint64_t *rowInOut1,
+                            uint64_t *rowOut, uint64_t nCols)
+{
+   int i;
+   register __m512i state0, state1, state2, state3;
+   __m512i *in = (__m512i*)rowIn;
+   __m512i *inout0 = (__m512i*)rowInOut0;
+   __m512i *inout1 = (__m512i*)rowInOut1;
+   __m512i *out = (__m512i*)rowOut;
+//   inout_ovly io;
+   ovly_512 io0, io1, io2;
+
+   state0 = _mm512_load_si512( (__m512i*)State     );
+   state1 = _mm512_load_si512( (__m512i*)State + 1 );
+   state2 = _mm512_load_si512( (__m512i*)State + 2 );
+   state3 = _mm512_load_si512( (__m512i*)State + 3 );
+    
+   for ( i = 0; i < nCols; i++ )
+   {
+     //Absorbing "M[prev] [+] M[row*]"
+     io0.v512 = _mm512_mask_blend_epi64( 0xf0,
+                                  _mm512_load_si512( (__m512i*)inout0 ),
+                                  _mm512_load_si512( (__m512i*)inout1 ) );
+     io1.v512 = _mm512_mask_blend_epi64( 0xf0,
+                                  _mm512_load_si512( (__m512i*)inout0 +1 ),
+                                  _mm512_load_si512( (__m512i*)inout1 +1 ) );
+     io2.v512 = _mm512_mask_blend_epi64( 0xf0,
+                                  _mm512_load_si512( (__m512i*)inout0 +2 ),
+                                  _mm512_load_si512( (__m512i*)inout1 +2 ) );
+
+     state0 = _mm512_xor_si512( state0, _mm512_add_epi64( in[0], io0.v512 ) );
+     state1 = _mm512_xor_si512( state1, _mm512_add_epi64( in[1], io1.v512 ) );
+     state2 = _mm512_xor_si512( state2, _mm512_add_epi64( in[2], io2.v512 ) );
+     
+/* 
+     io.v512[0] = _mm512_mask_blend_epi64( 0xf0,
+                                  _mm512_load_si512( (__m512i*)inout0 ),
+                                  _mm512_load_si512( (__m512i*)inout1 ) );
+     io.v512[1] = _mm512_mask_blend_epi64( 0xf0,
+                                  _mm512_load_si512( (__m512i*)inout0 +1 ),
+                                  _mm512_load_si512( (__m512i*)inout1 +1 ) );
+     io.v512[2] = _mm512_mask_blend_epi64( 0xf0,
+                                  _mm512_load_si512( (__m512i*)inout0 +2 ),
+                                  _mm512_load_si512( (__m512i*)inout1 +2 ) );
+
+     state0 = _mm512_xor_si512( state0, _mm512_add_epi64( in[0], io.v512[0] ) );
+     state1 = _mm512_xor_si512( state1, _mm512_add_epi64( in[1], io.v512[1] ) );
+     state2 = _mm512_xor_si512( state2, _mm512_add_epi64( in[2], io.v512[2] ) );
+*/
+
+     //Applies the reduced-round transformation f to the sponge's state
+     LYRA_ROUND_2WAY_AVX512( state0, state1, state2, state3 );
+
+     {
+       __m512i t0, t1, t2;
+
+       //M[rowOut][col] = M[rowOut][col] XOR rand
+       out[0] = _mm512_xor_si512( out[0], state0 );
+       out[1] = _mm512_xor_si512( out[1], state1 );
+       out[2] = _mm512_xor_si512( out[2], state2 );
+
+       // if out is the same row as inout, update with new data.
+       if ( rowOut == rowInOut0 )
+       {
+          io0.v512 = _mm512_mask_blend_epi64( 0x0f, io0.v512, out[0] );
+          io1.v512 = _mm512_mask_blend_epi64( 0x0f, io1.v512, out[1] );
+          io2.v512 = _mm512_mask_blend_epi64( 0x0f, io2.v512, out[2] );
+
+       }
+       if ( rowOut == rowInOut1 )
+       {
+          io0.v512 = _mm512_mask_blend_epi64( 0xf0, io0.v512, out[0] );
+          io1.v512 = _mm512_mask_blend_epi64( 0xf0, io1.v512, out[1] );
+          io2.v512 = _mm512_mask_blend_epi64( 0xf0, io2.v512, out[2] );
+       }
+
+/*
+       if ( rowOut == rowInOut0 )
+       {
+          io.v512[0] = _mm512_mask_blend_epi64( 0x0f, io.v512[0], out[0] );
+          io.v512[1] = _mm512_mask_blend_epi64( 0x0f, io.v512[1], out[1] );
+          io.v512[2] = _mm512_mask_blend_epi64( 0x0f, io.v512[2], out[2] );
+
+       }
+       if ( rowOut == rowInOut1 )
+       {
+          io.v512[0] = _mm512_mask_blend_epi64( 0xf0, io.v512[0], out[0] );
+          io.v512[1] = _mm512_mask_blend_epi64( 0xf0, io.v512[1], out[1] );
+          io.v512[2] = _mm512_mask_blend_epi64( 0xf0, io.v512[2], out[2] );
+       }
+*/
+
+       //M[rowInOut][col] = M[rowInOut][col] XOR rotW(rand)
+       t0 = _mm512_permutex_epi64( state0, 0x93 );
+       t1 = _mm512_permutex_epi64( state1, 0x93 );
+       t2 = _mm512_permutex_epi64( state2, 0x93 );
+
+       io0.v512 = _mm512_xor_si512( io0.v512,
+                                 _mm512_mask_blend_epi64( 0x11, t0, t2 ) );
+       io1.v512 = _mm512_xor_si512( io1.v512,
+                                 _mm512_mask_blend_epi64( 0x11, t1, t0 ) );
+       io2.v512 = _mm512_xor_si512( io2.v512,
+                                 _mm512_mask_blend_epi64( 0x11, t2, t1 ) );
+     }
+
+      casti_m256i( inout0, 0 ) = io0.v256lo;
+      casti_m256i( inout1, 1 ) = io0.v256hi;
+      casti_m256i( inout0, 2 ) = io1.v256lo;
+      casti_m256i( inout1, 3 ) = io1.v256hi;
+      casti_m256i( inout0, 4 ) = io2.v256lo;
+      casti_m256i( inout1, 5 ) = io2.v256hi;
+/*     
+     _mm512_mask_store_epi64( inout0,    0x0f, io.v512[0] );
+     _mm512_mask_store_epi64( inout1,    0xf0, io.v512[0] );
+     _mm512_mask_store_epi64( inout0 +1, 0x0f, io.v512[1] );
+     _mm512_mask_store_epi64( inout1 +1, 0xf0, io.v512[1] );
+     _mm512_mask_store_epi64( inout0 +2, 0x0f, io.v512[2] );
+     _mm512_mask_store_epi64( inout1 +2, 0xf0, io.v512[2] );
+*/
+      //Goes to next block
+      in     += BLOCK_LEN_M256I;
+      inout0 += BLOCK_LEN_M256I;
+      inout1 += BLOCK_LEN_M256I;
+      out    += BLOCK_LEN_M256I;
+   }
+
+   _mm512_store_si512( (__m512i*)State,     state0 );
+   _mm512_store_si512( (__m512i*)State + 1, state1 );
+   _mm512_store_si512( (__m512i*)State + 2, state2 );
+   _mm512_store_si512( (__m512i*)State + 3, state3 );
+
+}
+
+static inline void reducedDuplexRow_2way_overlap_X( uint64_t *State,
+                    uint64_t *rowIn, uint64_t *rowInOut0, uint64_t *rowInOut1,
+                    uint64_t *rowOut, uint64_t nCols)
 {
   int i;
   register __m512i state0, state1, state2, state3;
@@ -257,30 +482,14 @@ inline void reducedDuplexRow_2way( uint64_t *State, uint64_t *rowIn,
   __m256i *inout0 = (__m256i*)rowInOut0;
   __m256i *inout1 = (__m256i*)rowInOut1;
   __m512i *out = (__m512i*)rowOut;
-   __m512i io[3];
-   povly inout;
-   inout.v512 = &io[0];
-    __m512i t0, t1, t2;
+   inout_ovly inout;
+   __m512i t0, t1, t2;

   state0 = _mm512_load_si512( (__m512i*)State     );
   state1 = _mm512_load_si512( (__m512i*)State + 1 );
   state2 = _mm512_load_si512( (__m512i*)State + 2 );
   state3 = _mm512_load_si512( (__m512i*)State + 3 );
-    
-    _mm_prefetch( in,     _MM_HINT_T0 );
-    _mm_prefetch( inout0,     _MM_HINT_T0 );
-    _mm_prefetch( inout1,     _MM_HINT_T0 );
-    _mm_prefetch( in     + 2, _MM_HINT_T0 );
-    _mm_prefetch( inout0 + 2, _MM_HINT_T0 );
-    _mm_prefetch( inout1 + 2, _MM_HINT_T0 );
-    _mm_prefetch( in     + 4, _MM_HINT_T0 );
-    _mm_prefetch( inout0 + 4, _MM_HINT_T0 );
-    _mm_prefetch( inout1 + 4, _MM_HINT_T0 );
-    _mm_prefetch( in     + 6, _MM_HINT_T0 );
-    _mm_prefetch( inout0 + 6, _MM_HINT_T0 );
-    _mm_prefetch( inout1 + 6, _MM_HINT_T0 );

-    
    for ( i = 0; i < nCols; i++ )
    {

@@ -311,15 +520,15 @@ inline void reducedDuplexRow_2way( uint64_t *State, uint64_t *rowIn,
      // if inout is the same row as out it was just overwritten, reload.
      if ( rowOut == rowInOut0 )
      {
-         inout.v256[0] = inout0[0];
-         inout.v256[2] = inout0[2];
-         inout.v256[4] = inout0[4];
+         inout.v256[0] = ( (__m256i*)out )[0];
+         inout.v256[2] = ( (__m256i*)out )[2];
+         inout.v256[4] = ( (__m256i*)out )[4];
      }
      if ( rowOut == rowInOut1 )
      {
-         inout.v256[1] = inout1[1];
-         inout.v256[3] = inout1[3];
-         inout.v256[5] = inout1[5];
+         inout.v256[1] = ( (__m256i*)out )[1];
+         inout.v256[3] = ( (__m256i*)out )[3];
+         inout.v256[5] = ( (__m256i*)out )[5];
      }

      //M[rowInOut][col] = M[rowInOut][col] XOR rotW(rand)
@@ -328,12 +537,12 @@ inline void reducedDuplexRow_2way( uint64_t *State, uint64_t *rowIn,
      t2 = _mm512_permutex_epi64( state2, 0x93 );

      inout.v512[0] = _mm512_xor_si512( inout.v512[0],
-                                   _mm512_mask_blend_epi32( 0x0303, t0, t2 ) );
+                                   _mm512_mask_blend_epi64( 0x11, t0, t2 ) );
      inout.v512[1] = _mm512_xor_si512( inout.v512[1],
-                                   _mm512_mask_blend_epi32( 0x0303, t1, t0 ) );
+                                   _mm512_mask_blend_epi64( 0x11, t1, t0 ) );
      inout.v512[2] = _mm512_xor_si512( inout.v512[2],
-                                   _mm512_mask_blend_epi32( 0x0303, t2, t1 ) );
-      
+                                   _mm512_mask_blend_epi64( 0x11, t2, t1 ) );
+
      inout0[0] = inout.v256[0];
      inout1[1] = inout.v256[1];
      inout0[2] = inout.v256[2];
@@ -354,4 +563,108 @@ inline void reducedDuplexRow_2way( uint64_t *State, uint64_t *rowIn,
   _mm512_store_si512( (__m512i*)State + 3, state3 );
 }

+// rowInOut0 == rowInOut1, fastest, least likely: 1 / nrows
+static inline void reducedDuplexRow_2way_unified( uint64_t *State,
+                   uint64_t *rowIn, uint64_t *rowInOut0,
+                            uint64_t *rowOut, uint64_t nCols)
+{
+   int i;
+   register __m512i state0, state1, state2, state3;
+   __m512i *in = (__m512i*)rowIn;
+   __m512i *inout = (__m512i*)rowInOut0;
+   __m512i *out = (__m512i*)rowOut;
+
+   state0 = _mm512_load_si512( (__m512i*)State     );
+   state1 = _mm512_load_si512( (__m512i*)State + 1 );
+   state2 = _mm512_load_si512( (__m512i*)State + 2 );
+   state3 = _mm512_load_si512( (__m512i*)State + 3 );
+
+   for ( i = 0; i < nCols; i++ )
+   {
+     //Absorbing "M[prev] [+] M[row*]"
+     state0 = _mm512_xor_si512( state0, _mm512_add_epi64( in[0], inout[0] ) );
+     state1 = _mm512_xor_si512( state1, _mm512_add_epi64( in[1], inout[1] ) );
+     state2 = _mm512_xor_si512( state2, _mm512_add_epi64( in[2], inout[2] ) );
+
+     //Applies the reduced-round transformation f to the sponge's state
+     LYRA_ROUND_2WAY_AVX512( state0, state1, state2, state3 );
+
+     {
+       register __m512i t0, t1, t2;
+
+       //M[rowInOut][col] = M[rowInOut][col] XOR rotW(rand)
+       t0 = _mm512_permutex_epi64( state0, 0x93 );
+       t1 = _mm512_permutex_epi64( state1, 0x93 );
+       t2 = _mm512_permutex_epi64( state2, 0x93 );
+
+       inout[0] = _mm512_xor_si512( inout[0],
+                                    _mm512_mask_blend_epi64( 0x11, t0, t2 ) );
+       inout[1] = _mm512_xor_si512( inout[1],
+                                    _mm512_mask_blend_epi64( 0x11, t1, t0 ) );
+       inout[2] = _mm512_xor_si512( inout[2],
+                                    _mm512_mask_blend_epi64( 0x11, t2, t1 ) );
+
+       out[0] = _mm512_xor_si512( out[0], state0 );
+       out[1] = _mm512_xor_si512( out[1], state1 );
+       out[2] = _mm512_xor_si512( out[2], state2 );
+
+     }
+
+     //Goes to next block
+     in    += BLOCK_LEN_M256I;
+     inout += BLOCK_LEN_M256I;
+     out   += BLOCK_LEN_M256I;
+   }
+
+   _mm512_store_si512( (__m512i*)State,     state0 );
+   _mm512_store_si512( (__m512i*)State + 1, state1 );
+   _mm512_store_si512( (__m512i*)State + 2, state2 );
+   _mm512_store_si512( (__m512i*)State + 3, state3 );
+}
+
+// Multi level specialization.
+// There are three cases that need to be handled:
+// unified: inout data is contiguous, fastest, unlikely.
+// normal: inout data is not contiguous with no overlap with out, likely. 
+// overlap: inout data is not contiguous and one lane overlaps with out
+//          slowest, unlikely.
+//
+// In adition different algos prefer different coding. x25x and x22i prefer
+// 256 bit memory acceses to handle the diverged data while all other
+// algos prefer 512 bit memory accesses with masking and blending.
+
+ 
+//  Wrapper
+inline void reducedDuplexRow_2way( uint64_t *State, uint64_t *rowIn,
+                                   uint64_t *rowInOut0, uint64_t *rowInOut1,
+                                   uint64_t *rowOut, uint64_t nCols )
+{
+  if ( rowInOut0 == rowInOut1 )
+     reducedDuplexRow_2way_unified( State, rowIn, rowInOut0, rowOut, nCols );
+  else if ( ( rowInOut0 == rowOut ) || ( rowInOut1 == rowOut ) )
+     reducedDuplexRow_2way_overlap( State, rowIn, rowInOut0, rowInOut1,
+                                    rowOut, nCols );
+  else
+     reducedDuplexRow_2way_normal( State, rowIn, rowInOut0, rowInOut1,
+                                   rowOut, nCols );
+}
+
+inline void reducedDuplexRow_2way_X( uint64_t *State, uint64_t *rowIn,
+                                     uint64_t *rowInOut0, uint64_t *rowInOut1,
+                                     uint64_t *rowOut, uint64_t nCols )
+{
+   if ( rowInOut0 == rowInOut1 )
+      reducedDuplexRow_2way_unified( State, rowIn, rowInOut0, rowOut, nCols );
+   else if ( ( rowInOut0 == rowOut ) || ( rowInOut1 == rowOut ) )
+   {
+      asm volatile ( "nop" );  // Prevent GCC from optimizing
+      reducedDuplexRow_2way_overlap_X( State, rowIn, rowInOut0, rowInOut1,
+                                       rowOut, nCols );
+   }
+   else
+      reducedDuplexRow_2way_normal( State, rowIn, rowInOut0, rowInOut1,
+                                    rowOut, nCols );
+}
+
+
 #endif // AVX512
--- a/algo/lyra2/sponge.h
+++ b/algo/lyra2/sponge.h
@@ -203,13 +203,24 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){

 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)

-union _povly
+union _ovly_512
 {
-   __m512i *v512;
-   __m256i *v256;
-   uint64_t *u64;
+  __m512i v512;
+  struct
+  {
+     __m256i v256lo;
+     __m256i v256hi;
+  };
 };
-typedef union _povly povly;
+typedef union _ovly_512 ovly_512;
+
+
+union _inout_ovly
+{
+   __m512i v512[3];
+   __m256i v256[6];
+};
+typedef union _inout_ovly inout_ovly;

 //---- Housekeeping
 void initState_2way( uint64_t State[/*16*/] );
@@ -234,6 +245,10 @@ void reducedDuplexRow_2way( uint64_t *State, uint64_t *rowIn,
                            uint64_t *rowInOut0, uint64_t *rowInOut1,
                            uint64_t *rowOut, uint64_t nCols);

+void reducedDuplexRow_2way_X( uint64_t *State, uint64_t *rowIn,
+                              uint64_t *rowInOut0, uint64_t *rowInOut1,
+                              uint64_t *rowOut, uint64_t nCols);
+
 #endif


--- a/algo/m7m.c
+++ b/algo/m7m.c
@@ -149,7 +149,7 @@ int scanhash_m7m_hash( struct work* work, uint64_t max_nonce,
    char data_str[161], hash_str[65], target_str[65];
    //uint8_t *bdata = 0;
    uint8_t bdata[8192] __attribute__ ((aligned (64)));
-    int rc = 0, i, digits;
+    int i, digits;
    int bytes;
    size_t p = sizeof(unsigned long), a = 64/p, b = 32/p;

@@ -267,47 +267,41 @@ int scanhash_m7m_hash( struct work* work, uint64_t max_nonce,
            SHA256_Final( (unsigned char*) hash, &ctxf_sha256 );
        }

-        const unsigned char *hash_ = (const unsigned char *)hash;
-        const unsigned char *target_ = (const unsigned char *)ptarget;
-        for ( i = 31; i >= 0; i-- )
+
+        if ( unlikely( valid_hash( (uint64_t*)hash, (uint64_t*)ptarget ) 
+             && !opt_benchmark ) )
+
+
+//        if ( unlikely( hash[7] <= ptarget[7] ) )
+//        if ( likely( fulltest( hash, ptarget ) && !opt_benchmark ) )        
        {
-	        if ( hash_[i] != target_[i] )
+           if ( opt_debug )
           {
-		        rc = hash_[i] < target_[i];
-		        break;
-	        }
-        }
-        if ( unlikely(rc) )
-        {
-            if ( opt_debug )
-            {
-                bin2hex(hash_str, (unsigned char *)hash, 32);
-                bin2hex(target_str, (unsigned char *)ptarget, 32);
-                bin2hex(data_str, (unsigned char *)data, 80);
-                applog(LOG_DEBUG, "DEBUG: [%d thread] Found share!\ndata   %s\nhash   %s\ntarget %s", thr_id, 
-                    data_str,
-                    hash_str,
-                    target_str);
+                bin2hex( hash_str, (unsigned char *)hash, 32 );
+                bin2hex( target_str, (unsigned char *)ptarget, 32 );
+                bin2hex( data_str, (unsigned char *)data, 80 );
+                applog( LOG_DEBUG, "DEBUG: [%d thread] Found share!\ndata   %s\nhash   %s\ntarget %s",
+                      thr_id, data_str, hash_str, target_str );
            }
            pdata[19] = data[19];
            submit_solution( work, hash, mythr );
        }
-    } while (n < max_nonce && !work_restart[thr_id].restart);
+    } while ( n < max_nonce && !work_restart[thr_id].restart );

     pdata[19] = n;

-     mpf_set_prec_raw(magifpi, prec0);
-     mpf_set_prec_raw(magifpi0, prec0);
-     mpf_set_prec_raw(mptmp, prec0);
-     mpf_set_prec_raw(mpt1, prec0);
-     mpf_set_prec_raw(mpt2, prec0);
-     mpf_clear(magifpi);
-     mpf_clear(magifpi0);
-     mpf_clear(mpten);
-     mpf_clear(mptmp);
-     mpf_clear(mpt1);
-     mpf_clear(mpt2);
-     mpz_clears(magipi, magisw, product, bns0, bns1, NULL);
+     mpf_set_prec_raw( magifpi, prec0 );
+     mpf_set_prec_raw( magifpi0, prec0 );
+     mpf_set_prec_raw( mptmp, prec0 );
+     mpf_set_prec_raw( mpt1, prec0 );
+     mpf_set_prec_raw( mpt2, prec0 );
+     mpf_clear( magifpi );
+     mpf_clear( magifpi0 );
+     mpf_clear( mpten );
+     mpf_clear( mptmp );
+     mpf_clear( mpt1 );
+     mpf_clear( mpt2 );
+     mpz_clears( magipi, magisw, product, bns0, bns1, NULL );

    *hashes_done = n - first_nonce + 1;
    return 0;
--- a/algo/nist5/nist5-4way.c
+++ b/algo/nist5/nist5-4way.c
@@ -102,7 +102,7 @@ int scanhash_nist5_8way( struct work *work, uint32_t max_nonce,
        nist5hash_8way( hash, vdata );

        for ( int lane = 0; lane < 8; lane++ )
-        if ( hash7[ lane<<1 ] < Htarg )
+        if ( hash7[ lane<<1 ] <= Htarg )
        {
           extr_lane_8x64( lane_hash, hash, lane, 256 );
           if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
@@ -190,7 +190,7 @@ int scanhash_nist5_4way( struct work *work, uint32_t max_nonce,
        nist5hash_4way( hash, vdata );

        for ( int lane = 0; lane < 4; lane++ )
-        if ( hash7[ lane<<1 ] < Htarg )
+        if ( hash7[ lane<<1 ] <= Htarg )
        {
           extr_lane_4x64( lane_hash, hash, lane, 256 );
           if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
--- a/algo/nist5/nist5.c
+++ b/algo/nist5/nist5.c
@@ -25,7 +25,6 @@ void nist5hash(void *output, const void *input)
   sph_skein512_context    ctx_skein;
   sph_jh512_context       ctx_jh;
   sph_keccak512_context   ctx_keccak;
-   uint32_t mask = 8;

   sph_blake512_init( &ctx_blake );
   sph_blake512( &ctx_blake, input, 80 );
@@ -59,10 +58,10 @@ void nist5hash(void *output, const void *input)
 int scanhash_nist5( struct work *work, uint32_t max_nonce,
                    uint64_t *hashes_done, struct thr_info *mythr)
 {
-        uint32_t endiandata[20] __attribute__((aligned(64)));
-        uint32_t hash64[8] __attribute__((aligned(32)));
-        uint32_t *pdata = work->data;
-        uint32_t *ptarget = work->target;
+   uint32_t endiandata[20] __attribute__((aligned(64)));
+   uint32_t hash64[8] __attribute__((aligned(32)));
+   uint32_t *pdata = work->data;
+   uint32_t *ptarget = work->target;
 	uint32_t n = pdata[19] - 1;
 	const uint32_t first_nonce = pdata[19];
   int thr_id = mythr->id;  // thr_id arg is deprecated
@@ -88,9 +87,6 @@ int scanhash_nist5( struct work *work, uint32_t max_nonce,
 	// we need bigendian data...
        swab32_array( endiandata, pdata, 20 );

-#ifdef DEBUG_ALGO
-	printf("[%d] Htarg=%X\n", thr_id, Htarg);
-#endif
 	for (int m=0; m < 6; m++) {
 		if (Htarg <= htmax[m]) {
 			uint32_t mask = masks[m];
@@ -98,24 +94,9 @@ int scanhash_nist5( struct work *work, uint32_t max_nonce,
 				pdata[19] = ++n;
 				be32enc(&endiandata[19], n);
 				nist5hash(hash64, endiandata);
-#ifndef DEBUG_ALGO
-				if ((!(hash64[7] & mask)) && fulltest(hash64, ptarget)) {
-					*hashes_done = n - first_nonce + 1;
-					return true;
-				}
-#else
-				if (!(n % 0x1000) && !thr_id) printf(".");
-				if (!(hash64[7] & mask)) {
-					printf("[%d]",thr_id);
-					if (fulltest(hash64, ptarget)) {
-                                                work_set_target_ratio( work, hash64 );
-						*hashes_done = n - first_nonce + 1;
-						return true;
-					}
-				}
-#endif
+				if ((!(hash64[7] & mask)) && fulltest(hash64, ptarget))
+                submit_solution( work, hash64, mythr );
 			} while (n < max_nonce && !work_restart[thr_id].restart);
-			// see blake.c if else to understand the loop on htmax => mask
 			break;
 		}
 	}
--- a/algo/nist5/zr5.c
+++ b/algo/nist5/zr5.c
@@ -142,11 +142,7 @@ int scanhash_zr5( struct work *work, uint32_t max_nonce,
       {
         pdata[0] = tmpdata[0];
         pdata[19] = nonce;
-         *hashes_done = pdata[19] - first_nonce + 1;
-         work_set_target_ratio( work, hash );
-         if (opt_debug)
-           applog(LOG_INFO, "found nonce %x", nonce);
-         return 1;
+         submit_solution( work, hash, mythr );
       }
    }
    nonce++;
@@ -158,14 +154,13 @@ int scanhash_zr5( struct work *work, uint32_t max_nonce,
 }

 void zr5_get_new_work( struct work* work, struct work* g_work, int thr_id,
-                       uint32_t* end_nonce_ptr, bool clean_job )
+                       uint32_t* end_nonce_ptr )
 {
   // ignore POK in first word
-// const int nonce_i = 19;
   const int wkcmp_sz = 72;  // (19-1) * sizeof(uint32_t)
   uint32_t *nonceptr = algo_gate.get_nonceptr( work->data );
   if ( memcmp( &work->data[1], &g_work->data[1], wkcmp_sz )
-      && ( clean_job || ( *nonceptr >= *end_nonce_ptr ) ) )
+      || ( *nonceptr >= *end_nonce_ptr ) )
   {
      work_free( work );
      work_copy( work, g_work );
--- a/algo/panama/panama-hash-4way.c
+++ b/algo/panama/panama-hash-4way.c
@@ -6,26 +6,25 @@

 #define M17( macro ) \
 do { \
-      macro( 0,  1,  2,  4); \
-      macro( 1,  2,  3,  5); \
-      macro( 2,  3,  4,  6); \
-      macro( 3,  4,  5,  7); \
-      macro( 4,  5,  6,  8); \
-      macro( 5,  6,  7,  9); \
-      macro( 6,  7,  8, 10); \
-      macro( 7,  8,  9, 11); \
-      macro( 8,  9, 10, 12); \
-      macro( 9, 10, 11, 13); \
-      macro(10, 11, 12, 14); \
-      macro(11, 12, 13, 15); \
-      macro(12, 13, 14, 16); \
-      macro(13, 14, 15,  0); \
-      macro(14, 15, 16,  1); \
-      macro(15, 16,  0,  2); \
-      macro(16,  0,  1,  3); \
+      macro(  0,  1,  2,  4 ); \
+      macro(  1,  2,  3,  5 ); \
+      macro(  2,  3,  4,  6 ); \
+      macro(  3,  4,  5,  7 ); \
+      macro(  4,  5,  6,  8 ); \
+      macro(  5,  6,  7,  9 ); \
+      macro(  6,  7,  8, 10 ); \
+      macro(  7,  8,  9, 11 ); \
+      macro(  8,  9, 10, 12 ); \
+      macro(  9, 10, 11, 13 ); \
+      macro( 10, 11, 12, 14 ); \
+      macro( 11, 12, 13, 15 ); \
+      macro( 12, 13, 14, 16 ); \
+      macro( 13, 14, 15,  0 ); \
+      macro( 14, 15, 16,  1 ); \
+      macro( 15, 16,  0,  2 ); \
+      macro( 16,  0,  1,  3 ); \
 } while (0)

-
 #define RSTATE(n0, n1, n2, n4)    (a ## n0 = sc->state[n0])

 #define WSTATE(n0, n1, n2, n4)    (sc->state[n0] = a ## n0)
@@ -50,9 +49,7 @@ do { \

 #define LVARS_4W   \
 	LVAR17_4W(a) \
-	LVAR17_4W(g) \
-	LVAR17_4W(p) \
-	LVAR17_4W(t)
+	LVAR17_4W(g)

 #define BUPDATE1_4W( n0, n2 ) \
 do { \
@@ -63,14 +60,14 @@ do { \

 #define BUPDATE_4W \
 do { \
-		BUPDATE1_4W(0, 2); \
-		BUPDATE1_4W(1, 3); \
-		BUPDATE1_4W(2, 4); \
-		BUPDATE1_4W(3, 5); \
-		BUPDATE1_4W(4, 6); \
-		BUPDATE1_4W(5, 7); \
-		BUPDATE1_4W(6, 0); \
-		BUPDATE1_4W(7, 1); \
+		BUPDATE1_4W( 0, 2 ); \
+		BUPDATE1_4W( 1, 3 ); \
+		BUPDATE1_4W( 2, 4 ); \
+		BUPDATE1_4W( 3, 5 ); \
+		BUPDATE1_4W( 4, 6 ); \
+		BUPDATE1_4W( 5, 7 ); \
+		BUPDATE1_4W( 6, 0 ); \
+		BUPDATE1_4W( 7, 1 ); \
 } while (0)

 #define GAMMA_4W(n0, n1, n2, n4)   \
@@ -78,46 +75,46 @@ do { \
                             _mm_or_si128( a ## n1, mm128_not( a ## n2 ) ) ) )

 #define PI_ALL_4W   do { \
-      p0  = g0; \
-      p1  = mm128_rol_32(  g7,  1 ); \
-      p2  = mm128_rol_32( g14,  3 ); \
-      p3  = mm128_rol_32(  g4,  6 ); \
-      p4  = mm128_rol_32( g11, 10 ); \
-      p5  = mm128_rol_32(  g1, 15 ); \
-      p6  = mm128_rol_32(  g8, 21 ); \
-      p7  = mm128_rol_32( g15, 28 ); \
-      p8  = mm128_rol_32(  g5,  4 ); \
-      p9  = mm128_rol_32( g12, 13 ); \
-      p10 = mm128_rol_32(  g2, 23 ); \
-      p11 = mm128_rol_32(  g9,  2 ); \
-      p12 = mm128_rol_32( g16, 14 ); \
-      p13 = mm128_rol_32(  g6, 27 ); \
-      p14 = mm128_rol_32( g13,  9 ); \
-      p15 = mm128_rol_32(  g3, 24 ); \
-      p16 = mm128_rol_32( g10,  8 ); \
+      a0  = g0; \
+      a1  = mm128_rol_32( g7,   1 ); \
+      a2  = mm128_rol_32( g14,  3 ); \
+      a3  = mm128_rol_32( g4,   6 ); \
+      a4  = mm128_rol_32( g11, 10 ); \
+      a5  = mm128_rol_32( g1,  15 ); \
+      a6  = mm128_rol_32( g8,  21 ); \
+      a7  = mm128_rol_32( g15, 28 ); \
+      a8  = mm128_rol_32( g5,   4 ); \
+      a9  = mm128_rol_32( g12, 13 ); \
+      a10 = mm128_rol_32( g2,  23 ); \
+      a11 = mm128_rol_32( g9,   2 ); \
+      a12 = mm128_rol_32( g16, 14 ); \
+      a13 = mm128_rol_32( g6,  27 ); \
+      a14 = mm128_rol_32( g13,  9 ); \
+      a15 = mm128_rol_32( g3,  24 ); \
+      a16 = mm128_rol_32( g10,  8 ); \
   } while (0)

 #define THETA_4W(n0, n1, n2, n4)   \
-   ( t ## n0 = _mm_xor_si128( p ## n0, _mm_xor_si128( p ## n1, p ## n4 ) ) )
+   ( g ## n0 = _mm_xor_si128( a ## n0, _mm_xor_si128( a ## n1, a ## n4 ) ) )

 #define SIGMA_ALL_4W   do { \
-		a0 = _mm_xor_si128( t0, _mm_set1_epi32( 1 ) ); \
-		a1 = _mm_xor_si128( t1, INW2( 0 ) ); \
-		a2 = _mm_xor_si128( t2, INW2( 1 ) ); \
-		a3 = _mm_xor_si128( t3, INW2( 2 ) ); \
-		a4 = _mm_xor_si128( t4, INW2( 3 ) ); \
-		a5 = _mm_xor_si128( t5, INW2( 4 ) ); \
-		a6 = _mm_xor_si128( t6, INW2( 5 ) ); \
-		a7 = _mm_xor_si128( t7, INW2( 6 ) ); \
-		a8 = _mm_xor_si128( t8, INW2( 7 ) ); \
-		a9  = _mm_xor_si128(  t9, sc->buffer[ ptr16 ] [0 ] ); \
-		a10 = _mm_xor_si128( t10, sc->buffer[ ptr16 ] [1 ] ); \
-		a11 = _mm_xor_si128( t11, sc->buffer[ ptr16 ] [2 ] ); \
-		a12 = _mm_xor_si128( t12, sc->buffer[ ptr16 ] [3 ] ); \
-		a13 = _mm_xor_si128( t13, sc->buffer[ ptr16 ] [4 ] ); \
-		a14 = _mm_xor_si128( t14, sc->buffer[ ptr16 ] [5 ] ); \
-		a15 = _mm_xor_si128( t15, sc->buffer[ ptr16 ] [6 ] ); \
-		a16 = _mm_xor_si128( t16, sc->buffer[ ptr16 ] [7 ] ); \
+		a0 = _mm_xor_si128( g0, m128_one_32 ); \
+		a1 = _mm_xor_si128( g1, INW2( 0 ) ); \
+		a2 = _mm_xor_si128( g2, INW2( 1 ) ); \
+		a3 = _mm_xor_si128( g3, INW2( 2 ) ); \
+		a4 = _mm_xor_si128( g4, INW2( 3 ) ); \
+		a5 = _mm_xor_si128( g5, INW2( 4 ) ); \
+		a6 = _mm_xor_si128( g6, INW2( 5 ) ); \
+		a7 = _mm_xor_si128( g7, INW2( 6 ) ); \
+		a8 = _mm_xor_si128( g8, INW2( 7 ) ); \
+		a9  = _mm_xor_si128( g9,  sc->buffer[ ptr16 ][0] ); \
+		a10 = _mm_xor_si128( g10, sc->buffer[ ptr16 ][1] ); \
+		a11 = _mm_xor_si128( g11, sc->buffer[ ptr16 ][2] ); \
+		a12 = _mm_xor_si128( g12, sc->buffer[ ptr16 ][3] ); \
+		a13 = _mm_xor_si128( g13, sc->buffer[ ptr16 ][4] ); \
+		a14 = _mm_xor_si128( g14, sc->buffer[ ptr16 ][5] ); \
+		a15 = _mm_xor_si128( g15, sc->buffer[ ptr16 ][6] ); \
+		a16 = _mm_xor_si128( g16, sc->buffer[ ptr16 ][7] ); \
 	} while (0)

 #define PANAMA_STEP_4W   do { \
@@ -145,9 +142,9 @@ panama_4way_push( panama_4way_context *sc, const unsigned char *pbuf,
 #define INW2(i)   INW1(i)

 	M17( RSTATE );
-
   ptr0 = sc->buffer_ptr;
-	while (num -- > 0) {
+	while ( num-- > 0 )
+   {
 		PANAMA_STEP_4W;
 		pbuf = (const unsigned char *)pbuf + 32*4;
 	}
@@ -173,14 +170,11 @@ panama_4way_pull( panama_4way_context *sc, unsigned num )
 #define INW2(i)     casti_m128i( sc->buffer[ptr4], i )

 	M17( RSTATE );
-
   ptr0 = sc->buffer_ptr;
-
-   while (num -- > 0) {
+   while ( num-- > 0 )
+   {
 		unsigned ptr4;
-
 		ptr4 = ( (ptr0 + 4) & 31 );
-
      PANAMA_STEP_4W;
 	}
 	M17( WSTATE );
@@ -192,18 +186,11 @@ panama_4way_pull( panama_4way_context *sc, unsigned num )
 }

 void
-panama_4way_init(void *cc)
+panama_4way_init( void *cc )
 {
 	panama_4way_context *sc;

 	sc = cc;
-	/*
-	 * This is not completely conformant, but "it will work
-	 * everywhere". Initial state consists of zeroes everywhere.
-	 * Conceptually, the sph_u32 type may have padding bits which
-	 * must not be set to 0; but such an architecture remains to
-	 * be seen.
-	 */
 	sc->data_ptr = 0;
 	memset( sc->buffer, 0, sizeof sc->buffer );
 	sc->buffer_ptr = 0;
@@ -217,7 +204,8 @@ panama_4way_short( void *cc, const void *data, size_t len )
 	unsigned current;
 	sc = cc;
 	current = sc->data_ptr;
-	while (len > 0) {
+	while ( len > 0 )
+   {
 		unsigned clen;

 		clen = ( (sizeof sc->data ) >> 2 ) - current;
@@ -283,11 +271,8 @@ panama_4way_close( void *cc, void *dst )
 	*(__m128i*)( sc->data + current ) = m128_one_32;
   current++;
   memset_zero_128( (__m128i*)sc->data + current, 32 - current );
-
   panama_4way_push( sc, sc->data, 1 );
-
   panama_4way_pull( sc, 32 );
-
   for ( i = 0; i < 8; i ++ )
      casti_m128i( dst, i ) = sc->state[i + 9];
 }
@@ -306,9 +291,7 @@ panama_4way_close( void *cc, void *dst )

 #define LVARS_8W   \
   LVAR17_8W(a) \
-   LVAR17_8W(g) \
-   LVAR17_8W(p) \
-   LVAR17_8W(t)
+   LVAR17_8W(g)

 #define BUPDATE1_8W( n0, n2 ) \
 do { \
@@ -319,14 +302,14 @@ do { \

 #define BUPDATE_8W \
 do { \
-      BUPDATE1_8W(0, 2); \
-      BUPDATE1_8W(1, 3); \
-      BUPDATE1_8W(2, 4); \
-      BUPDATE1_8W(3, 5); \
-      BUPDATE1_8W(4, 6); \
-      BUPDATE1_8W(5, 7); \
-      BUPDATE1_8W(6, 0); \
-      BUPDATE1_8W(7, 1); \
+      BUPDATE1_8W( 0, 2 ); \
+      BUPDATE1_8W( 1, 3 ); \
+      BUPDATE1_8W( 2, 4 ); \
+      BUPDATE1_8W( 3, 5 ); \
+      BUPDATE1_8W( 4, 6 ); \
+      BUPDATE1_8W( 5, 7 ); \
+      BUPDATE1_8W( 6, 0 ); \
+      BUPDATE1_8W( 7, 1 ); \
 } while (0)

 #define GAMMA_8W(n0, n1, n2, n4)   \
@@ -334,46 +317,47 @@ do { \
                         _mm256_or_si256( a ## n1, mm256_not( a ## n2 ) ) ) )

 #define PI_ALL_8W   do { \
-      p0  = g0; \
-      p1  = mm256_rol_32(  g7,  1 ); \
-      p2  = mm256_rol_32( g14,  3 ); \
-      p3  = mm256_rol_32(  g4,  6 ); \
-      p4  = mm256_rol_32( g11, 10 ); \
-      p5  = mm256_rol_32(  g1, 15 ); \
-      p6  = mm256_rol_32(  g8, 21 ); \
-      p7  = mm256_rol_32( g15, 28 ); \
-      p8  = mm256_rol_32(  g5,  4 ); \
-      p9  = mm256_rol_32( g12, 13 ); \
-      p10 = mm256_rol_32(  g2, 23 ); \
-      p11 = mm256_rol_32(  g9,  2 ); \
-      p12 = mm256_rol_32( g16, 14 ); \
-      p13 = mm256_rol_32(  g6, 27 ); \
-      p14 = mm256_rol_32( g13,  9 ); \
-      p15 = mm256_rol_32(  g3, 24 ); \
-      p16 = mm256_rol_32( g10,  8 ); \
+      a0  = g0; \
+      a1  = mm256_rol_32( g7,   1 ); \
+      a2  = mm256_rol_32( g14,  3 ); \
+      a3  = mm256_rol_32( g4,   6 ); \
+      a4  = mm256_rol_32( g11, 10 ); \
+      a5  = mm256_rol_32( g1,  15 ); \
+      a6  = mm256_rol_32( g8,  21 ); \
+      a7  = mm256_rol_32( g15, 28 ); \
+      a8  = mm256_rol_32( g5,   4 ); \
+      a9  = mm256_rol_32( g12, 13 ); \
+      a10 = mm256_rol_32( g2,  23 ); \
+      a11 = mm256_rol_32( g9,   2 ); \
+      a12 = mm256_rol_32( g16, 14 ); \
+      a13 = mm256_rol_32( g6,  27 ); \
+      a14 = mm256_rol_32( g13,  9 ); \
+      a15 = mm256_rol_32( g3,  24 ); \
+      a16 = mm256_rol_32( g10,  8 ); \
   } while (0)

 #define THETA_8W(n0, n1, n2, n4)   \
-   ( t ## n0 = _mm256_xor_si256( p ## n0, _mm256_xor_si256( p ## n1, p ## n4 ) ) )
+   ( g ## n0 = _mm256_xor_si256( a ## n0, _mm256_xor_si256( a ## n1, \
+                                                            a ## n4 ) ) )

 #define SIGMA_ALL_8W   do { \
-      a0  = _mm256_xor_si256( t0, m256_one_32 ); \
-      a1  = _mm256_xor_si256( t1, INW2( 0 ) ); \
-      a2  = _mm256_xor_si256( t2, INW2( 1 ) ); \
-      a3  = _mm256_xor_si256( t3, INW2( 2 ) ); \
-      a4  = _mm256_xor_si256( t4, INW2( 3 ) ); \
-      a5  = _mm256_xor_si256( t5, INW2( 4 ) ); \
-      a6  = _mm256_xor_si256( t6, INW2( 5 ) ); \
-      a7  = _mm256_xor_si256( t7, INW2( 6 ) ); \
-      a8  = _mm256_xor_si256( t8, INW2( 7 ) ); \
-      a9  = _mm256_xor_si256(  t9, sc->buffer[ ptr16 ] [0 ] ); \
-      a10 = _mm256_xor_si256( t10, sc->buffer[ ptr16 ] [1 ] ); \
-      a11 = _mm256_xor_si256( t11, sc->buffer[ ptr16 ] [2 ] ); \
-      a12 = _mm256_xor_si256( t12, sc->buffer[ ptr16 ] [3 ] ); \
-      a13 = _mm256_xor_si256( t13, sc->buffer[ ptr16 ] [4 ] ); \
-      a14 = _mm256_xor_si256( t14, sc->buffer[ ptr16 ] [5 ] ); \
-      a15 = _mm256_xor_si256( t15, sc->buffer[ ptr16 ] [6 ] ); \
-      a16 = _mm256_xor_si256( t16, sc->buffer[ ptr16 ] [7 ] ); \
+      a0  = _mm256_xor_si256( g0, m256_one_32 ); \
+      a1  = _mm256_xor_si256( g1, INW2( 0 ) ); \
+      a2  = _mm256_xor_si256( g2, INW2( 1 ) ); \
+      a3  = _mm256_xor_si256( g3, INW2( 2 ) ); \
+      a4  = _mm256_xor_si256( g4, INW2( 3 ) ); \
+      a5  = _mm256_xor_si256( g5, INW2( 4 ) ); \
+      a6  = _mm256_xor_si256( g6, INW2( 5 ) ); \
+      a7  = _mm256_xor_si256( g7, INW2( 6 ) ); \
+      a8  = _mm256_xor_si256( g8, INW2( 7 ) ); \
+      a9  = _mm256_xor_si256( g9,  sc->buffer[ ptr16 ][0] ); \
+      a10 = _mm256_xor_si256( g10, sc->buffer[ ptr16 ][1] ); \
+      a11 = _mm256_xor_si256( g11, sc->buffer[ ptr16 ][2] ); \
+      a12 = _mm256_xor_si256( g12, sc->buffer[ ptr16 ][3] ); \
+      a13 = _mm256_xor_si256( g13, sc->buffer[ ptr16 ][4] ); \
+      a14 = _mm256_xor_si256( g14, sc->buffer[ ptr16 ][5] ); \
+      a15 = _mm256_xor_si256( g15, sc->buffer[ ptr16 ][6] ); \
+      a16 = _mm256_xor_si256( g16, sc->buffer[ ptr16 ][7] ); \
   } while (0)

 #define PANAMA_STEP_8W   do { \
@@ -401,7 +385,6 @@ panama_8way_push( panama_8way_context *sc, const unsigned char *pbuf,
 #define INW2(i)   INW1(i)

   M17( RSTATE );
-
   ptr0 = sc->buffer_ptr;
   while ( num-- > 0 )
   {
@@ -429,9 +412,9 @@ panama_8way_pull( panama_8way_context *sc, unsigned num )

   ptr0 = sc->buffer_ptr;

-   while (num -- > 0) {
+   while ( num-- > 0 )
+   {
      unsigned ptr4;
-
      ptr4 = ( (ptr0 + 4) & 31 );
      PANAMA_STEP_8W;
   }
@@ -449,13 +432,6 @@ panama_8way_init( void *cc )
   panama_8way_context *sc;

   sc = cc;
-   /*
-    * This is not completely conformant, but "it will work
-    * everywhere". Initial state consists of zeroes everywhere.
-    * Conceptually, the sph_u32 type may have padding bits which
-    * must not be set to 0; but such an architecture remains to
-    * be seen.
-    */
   sc->data_ptr = 0;
   memset( sc->buffer, 0, sizeof sc->buffer );
   sc->buffer_ptr = 0;
@@ -469,7 +445,8 @@ panama_8way_short( void *cc, const void *data, size_t len )
   unsigned current;
   sc = cc;
   current = sc->data_ptr;
-   while (len > 0) {
+   while ( len > 0 )
+   {
      unsigned clen;

      clen = ( (sizeof sc->data ) >> 3 ) - current;
@@ -486,7 +463,6 @@ panama_8way_short( void *cc, const void *data, size_t len )
         panama_8way_push( sc, sc->data, 1 );
      }
   }
-
   sc->data_ptr = current;
 }

@@ -535,9 +511,7 @@ panama_8way_close( void *cc, void *dst )
   *(__m256i*)( sc->data + current ) = m256_one_32;
   current++;
   memset_zero_256( (__m256i*)sc->data + current, 32 - current );
-
   panama_8way_push( sc, sc->data, 1 );
-
   panama_8way_pull( sc, 32 );

   for ( i = 0; i < 8; i ++ )
--- a/algo/quark/anime.c
+++ b/algo/quark/anime.c
@@ -160,16 +160,12 @@ int scanhash_anime( struct work *work, uint32_t max_nonce,
              pdata[19] = n;

             if ( ( hash[7] & mask ) == 0 && fulltest( hash, ptarget ) ) 
-             {
-                work_set_target_ratio( work, hash );
-                *hashes_done = n - first_nonce + 1;
-                return true;
-             }
+                submit_solution( work, hash, mythr );
             n++;
          } while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
          break;
       }
-
+    *hashes_done = n - first_nonce + 1;
    pdata[19] = n;
    return 0;
 }
--- a/algo/quark/hmq1725-4way.c
+++ b/algo/quark/hmq1725-4way.c
@@ -1513,10 +1513,10 @@ int scanhash_hmq1725_4way( struct work *work, uint32_t max_nonce,
       hmq1725_4way_hash( hash, vdata );

       for ( int lane = 0; lane < 4; lane++ )
-       if ( hash7[ lane<<1 ] <= Htarg )
+       if ( unlikely( hash7[ lane<<1 ] <= Htarg ) )
       {
          extr_lane_4x64( lane_hash, hash, lane, 256 );
-          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
+          if ( likely( fulltest( lane_hash, ptarget ) && !opt_benchmark ) )
          {
             pdata[19] = n + lane;
             submit_lane_solution( work, lane_hash, mythr, lane );
--- a/algo/quark/hmq1725.c
+++ b/algo/quark/hmq1725.c
@@ -330,11 +330,8 @@ int scanhash_hmq1725( struct work *work, uint32_t max_nonce,
 			be32enc(&endiandata[19], n); 
 			hmq1725hash(hash64, endiandata);
 			if (((hash64[7]&0xFFFFFFFF)==0) && 
-					fulltest(hash64, ptarget)) {
-				*hashes_done = n - first_nonce + 1;
-            work_set_target_ratio( work, hash64 );
-				return true;
-			}
+					fulltest(hash64, ptarget)) 
+            submit_solution( work, hash64, mythr );
 		} while (n < max_nonce && !work_restart[thr_id].restart);	
 	} 
 	else if (ptarget[7]<=0xF) 
@@ -344,11 +341,8 @@ int scanhash_hmq1725( struct work *work, uint32_t max_nonce,
 			be32enc(&endiandata[19], n); 
 			hmq1725hash(hash64, endiandata);
 			if (((hash64[7]&0xFFFFFFF0)==0) && 
-					fulltest(hash64, ptarget)) {
-				*hashes_done = n - first_nonce + 1;
-            work_set_target_ratio( work, hash64 );
-				return true;
-			}
+					fulltest(hash64, ptarget)) 
+            submit_solution( work, hash64, mythr );
 		} while (n < max_nonce && !work_restart[thr_id].restart);	
 	} 
 	else if (ptarget[7]<=0xFF) 
@@ -358,11 +352,8 @@ int scanhash_hmq1725( struct work *work, uint32_t max_nonce,
 			be32enc(&endiandata[19], n); 
 			hmq1725hash(hash64, endiandata);
 			if (((hash64[7]&0xFFFFFF00)==0) && 
-					fulltest(hash64, ptarget)) {
-				*hashes_done = n - first_nonce + 1;
-            work_set_target_ratio( work, hash64 );
-				return true;
-			}
+					fulltest(hash64, ptarget)) 
+            submit_solution( work, hash64, mythr );
 		} while (n < max_nonce && !work_restart[thr_id].restart);	
 	} 
 	else if (ptarget[7]<=0xFFF) 
@@ -372,13 +363,9 @@ int scanhash_hmq1725( struct work *work, uint32_t max_nonce,
 			be32enc(&endiandata[19], n); 
 			hmq1725hash(hash64, endiandata);
 			if (((hash64[7]&0xFFFFF000)==0) && 
-					fulltest(hash64, ptarget)) {
-				*hashes_done = n - first_nonce + 1;
-            work_set_target_ratio( work, hash64 );
-				return true;
-			}
+					fulltest(hash64, ptarget)) 
+            submit_solution( work, hash64, mythr );
 		} while (n < max_nonce && !work_restart[thr_id].restart);	
-
 	} 
 	else if (ptarget[7]<=0xFFFF) 
 	{
@@ -387,13 +374,9 @@ int scanhash_hmq1725( struct work *work, uint32_t max_nonce,
 			be32enc(&endiandata[19], n); 
 			hmq1725hash(hash64, endiandata);
 			if (((hash64[7]&0xFFFF0000)==0) && 
-					fulltest(hash64, ptarget)) {
-				*hashes_done = n - first_nonce + 1;
-            work_set_target_ratio( work, hash64 );
-				return true;
-			}
+					fulltest(hash64, ptarget)) 
+                submit_solution( work, hash64, mythr );
 		} while (n < max_nonce && !work_restart[thr_id].restart);	
-
 	} 
 	else 
 	{
@@ -401,15 +384,10 @@ int scanhash_hmq1725( struct work *work, uint32_t max_nonce,
 			pdata[19] = ++n;
 			be32enc(&endiandata[19], n); 
 			hmq1725hash(hash64, endiandata);
-			if (fulltest(hash64, ptarget)) {
-				*hashes_done = n - first_nonce + 1;
-            work_set_target_ratio( work, hash64 );
-				return true;
-			}
+			if (fulltest(hash64, ptarget)) 
+                submit_solution( work, hash64, mythr );
 		} while (n < max_nonce && !work_restart[thr_id].restart);	
 	}
-	
-	
 	*hashes_done = n - first_nonce + 1;
 	pdata[19] = n;
 	return 0;
--- a/algo/quark/quark-4way.c
+++ b/algo/quark/quark-4way.c
@@ -248,11 +248,11 @@ void quark_8way_hash( void *state, const void *input )
       jh512_8way_close( &ctx.jh, vhashB );
    }

-    // Final blend, directly to state, only need 32 bytes.
-    casti_m512i( state,0 ) = _mm512_mask_blend_epi64( vh_mask, vhA[0], vhB[0] );
-    casti_m512i( state,1 ) = _mm512_mask_blend_epi64( vh_mask, vhA[1], vhB[1] );
-    casti_m512i( state,2 ) = _mm512_mask_blend_epi64( vh_mask, vhA[2], vhB[2] );
-    casti_m512i( state,3 ) = _mm512_mask_blend_epi64( vh_mask, vhA[3], vhB[3] );
+   // Final blend, directly to state, only need 32 bytes.
+   casti_m512i( state,0 ) = _mm512_mask_blend_epi64( vh_mask, vhA[0], vhB[0] );
+   casti_m512i( state,1 ) = _mm512_mask_blend_epi64( vh_mask, vhA[1], vhB[1] );
+   casti_m512i( state,2 ) = _mm512_mask_blend_epi64( vh_mask, vhA[2], vhB[2] );
+   casti_m512i( state,3 ) = _mm512_mask_blend_epi64( vh_mask, vhA[3], vhB[3] );
 }

 int scanhash_quark_8way( struct work *work, uint32_t max_nonce,
@@ -267,23 +267,24 @@ int scanhash_quark_8way( struct work *work, uint32_t max_nonce,
    uint32_t n = pdata[19];
    const uint32_t first_nonce = pdata[19];
    __m512i  *noncev = (__m512i*)vdata + 9;   // aligned
-    int thr_id = mythr->id;  // thr_id arg is deprecated
+    int thr_id = mythr->id; 
+    const uint32_t Htarg = ptarget[7];

    mm512_bswap32_intrlv80_8x64( vdata, pdata );
    do
    {
       *noncev = mm512_intrlv_blend_32( mm512_bswap_32(
-                _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 );
+              _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 );

       quark_8way_hash( hash, vdata );
       pdata[19] = n;

       for ( int i = 0; i < 8; i++ )
-       if ( ( hash7[ i<<1 ] & 0xFFFFFF00 ) == 0 )
+       if ( unlikely( hash7[ i<<1 ] <= Htarg ) )
       {
          extr_lane_8x64( lane_hash, hash, i, 256 );
-          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark  )
+          if ( likely( fulltest( lane_hash, ptarget ) && !opt_benchmark ) )
          {
            pdata[19] = n+i;
            submit_lane_solution( work, lane_hash, mythr, i );
@@ -296,7 +297,6 @@ int scanhash_quark_8way( struct work *work, uint32_t max_nonce,
    return 0;
 }

-
 #elif defined (QUARK_4WAY)

 typedef struct {
@@ -460,8 +460,9 @@ int scanhash_quark_4way( struct work *work, uint32_t max_nonce,
    uint32_t n = pdata[19];
    const uint32_t first_nonce = pdata[19];
    __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
-    int thr_id = mythr->id;  // thr_id arg is deprecated
-
+    int thr_id = mythr->id;
+    const uint32_t Htarg = ptarget[7];
+ 
    mm256_bswap32_intrlv80_4x64( vdata, pdata );
    do
    {
@@ -472,10 +473,10 @@ int scanhash_quark_4way( struct work *work, uint32_t max_nonce,
       pdata[19] = n;

       for ( int i = 0; i < 4; i++ )
-       if ( ( hash7[ i<<1 ] & 0xFFFFFF00 ) == 0 )
+       if ( unlikely( hash7[ i<<1 ] <= Htarg ) )
       {
          extr_lane_4x64( lane_hash, hash, i, 256 );
-          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark  )
+          if ( likely( fulltest( lane_hash, ptarget ) && !opt_benchmark ) )
          {
            pdata[19] = n+i;
            submit_lane_solution( work, lane_hash, mythr, i );
--- a/algo/quark/quark.c
+++ b/algo/quark/quark.c
@@ -126,15 +126,11 @@ int scanhash_quark( struct work *work, uint32_t max_nonce,
 		pdata[19] = ++n;
 		be32enc(&endiandata[19], n); 
 		quark_hash(hash64, &endiandata);
-                if ((hash64[7]&0xFFFFFF00)==0)
-                {
-                  if (fulltest(hash64, ptarget)) 
-                  {
-                    work_set_target_ratio( work, hash64 );
-                    *hashes_done = n - first_nonce + 1;
-		    return true;
-                  }
-               }
+      if ((hash64[7]&0xFFFFFF00)==0)
+      {
+         if (fulltest(hash64, ptarget)) 
+                submit_solution( work, hash64, mythr );
+      }
 	} while (n < max_nonce && !work_restart[thr_id].restart);
 	
 	*hashes_done = n - first_nonce + 1;
--- a/algo/qubit/deep.c
+++ b/algo/qubit/deep.c
@@ -5,7 +5,7 @@
 #include <stdio.h>
 #include "algo/luffa/luffa_for_sse2.h" 
 #include "algo/cubehash/cubehash_sse2.h" 
-#ifndef NO_AES_NI
+#ifdef __AES__
 #include "algo/echo/aes_ni/hash_api.h"
 #else
 #include "algo/echo/sph_echo.h"
@@ -15,10 +15,10 @@ typedef struct
 {
        hashState_luffa         luffa;
        cubehashParam           cubehash;
-#ifdef NO_AES_NI
-        sph_echo512_context echo;
-#else
+#ifdef __AES__
        hashState_echo          echo;
+#else
+        sph_echo512_context echo;
 #endif
 } deep_ctx_holder;

@@ -29,10 +29,10 @@ void init_deep_ctx()
 {
        init_luffa( &deep_ctx.luffa, 512 );
        cubehashInit( &deep_ctx.cubehash, 512, 16, 32 );
-#ifdef NO_AES_NI
-        sph_echo512_init( &deep_ctx.echo );
-#else
+#ifdef __AES__
        init_echo( &deep_ctx.echo, 512 );
+#else
+        sph_echo512_init( &deep_ctx.echo );
 #endif
 };

@@ -59,12 +59,12 @@ void deep_hash(void *output, const void *input)
        cubehashUpdateDigest( &ctx.cubehash, (byte*)hash, 
                              (const byte*) hash,64);

-#ifdef NO_AES_NI
-        sph_echo512 (&ctx.echo, (const void*) hash, 64);
-        sph_echo512_close(&ctx.echo, (void*) hash);
-#else
+#ifdef __AES__
        update_final_echo ( &ctx.echo, (BitSequence *) hash,
                          (const BitSequence *) hash, 512);
+#else
+        sph_echo512 (&ctx.echo, (const void*) hash, 64);
+        sph_echo512_close(&ctx.echo, (void*) hash);
 #endif

        asm volatile ("emms");
@@ -92,46 +92,21 @@ int scanhash_deep( struct work *work, uint32_t max_nonce,

        deep_luffa_midstate( endiandata );

-#ifdef DEBUG_ALGO
-	printf("[%d] Htarg=%X\n", thr_id, Htarg);
-#endif
 	for ( int m=0; m < 6; m++ )
-        {
+   {
 	    if ( Htarg <= htmax[m] )
-            {
+       {
 	        uint32_t mask = masks[m];
 	        do
-                {
+           {
 	            pdata[19] = ++n;
-		    be32enc( &endiandata[19], n );
-		    deep_hash( hash64, endiandata );
-#ifndef DEBUG_ALGO
-		    if (!(hash64[7] & mask))
-                    {
-                       if ( fulltest(hash64, ptarget) )
-                       {
-		          *hashes_done = n - first_nonce + 1;
-		          return true;
-                       }
-//                       else
-//                       {
-//                          applog(LOG_INFO, "Result does not validate on CPU!");
-//                       }
-                     }
-#else
-                    if (!(n % 0x1000) && !thr_id) printf(".");
-	        	if (!(hash64[7] & mask)) {
-		            printf("[%d]",thr_id);
-			    if (fulltest(hash64, ptarget)) {
-                             work_set_target_ratio( work, hash64 );
-                             *hashes_done = n - first_nonce + 1;
-				return true;
-	                    }
- 	                }
-#endif
-                } while ( n < max_nonce && !work_restart[thr_id].restart );
-                // see blake.c if else to understand the loop on htmax => mask
-            break;
+               be32enc( &endiandata[19], n );
+		         deep_hash( hash64, endiandata );
+		         if (!(hash64[7] & mask))
+               if ( fulltest(hash64, ptarget) )
+                   submit_solution( work, hash64, mythr );
+            } while ( n < max_nonce && !work_restart[thr_id].restart );
+          break;
          } 
        }

--- a/algo/qubit/qubit-2way.c
+++ b/algo/qubit/qubit-2way.c
@@ -149,8 +149,8 @@ int scanhash_qubit_4way( struct work *work,uint32_t max_nonce,
        pdata[19] = n;

        for ( int lane = 0; lane < 4; lane++ )
-        if ( ( hash+(lane<<3) )[7] < Htarg )
-        if ( fulltest( hash+(lane<<3), ptarget) && !opt_benchmark )
+        if ( unlikely( ( hash+(lane<<3) )[7] <= Htarg ) )
+        if ( likely( fulltest( hash+(lane<<3), ptarget) && !opt_benchmark ) )
        {
           pdata[19] = n + lane;
           submit_lane_solution( work, hash+(lane<<3), mythr, lane );
@@ -233,10 +233,6 @@ int scanhash_qubit_2way( struct work *work,uint32_t max_nonce,
     uint32_t *noncep = vdata + 32+3;   // 4*8 + 3
     int thr_id = mythr->id;  // thr_id arg is deprecated
     const uint32_t Htarg = ptarget[7];
-     uint64_t htmax[] = {          0,        0xF,       0xFF,
-                               0xFFF,     0xFFFF, 0x10000000  };
-     uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
-                          0xFFFFF000, 0xFFFF0000,          0  };

     casti_m256i( endiandata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
     casti_m256i( endiandata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
@@ -248,32 +244,27 @@ int scanhash_qubit_2way( struct work *work,uint32_t max_nonce,
     luffa_2way_init( &qubit_2way_ctx.luffa, 512 );
     luffa_2way_update( &qubit_2way_ctx.luffa, vdata, 64 );

-     for ( int m=0; m < 6; m++ ) if ( Htarg <= htmax[m] )
+     do
     {
-        uint32_t mask = masks[m];
-        do
-        {
-            be32enc( noncep,   n   );
-            be32enc( noncep+4, n+1 );
-            qubit_2way_hash( hash, vdata );
-            pdata[19] = n;
+         be32enc( noncep,   n   );
+         be32enc( noncep+4, n+1 );
+         qubit_2way_hash( hash, vdata );
+         pdata[19] = n;

-            if ( !( hash[7] & mask ) )
-            if ( fulltest( hash, ptarget) && !opt_benchmark )
-            {
-                pdata[19] = n;
-                submit_lane_solution( work, hash, mythr, 0 );
-            }
-            if ( !( (hash+8)[7] & mask ) )
-            if ( fulltest( hash+8, ptarget) && !opt_benchmark )
-            {
-               pdata[19] = n+1;
-               submit_lane_solution( work, hash+8, mythr, 1 );
-            }
-            n += 2;
-         } while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
-         break;
-     }
+         if ( unlikely( hash[7] <= Htarg ) )
+         if ( likely( fulltest( hash, ptarget) && !opt_benchmark ) )
+         {
+            pdata[19] = n;
+            submit_lane_solution( work, hash, mythr, 0 );
+         }
+         if ( unlikely( ( (hash+8))[7] <= Htarg ) )
+         if ( likely( fulltest( hash+8, ptarget) && !opt_benchmark ) )
+         {
+            pdata[19] = n+1;
+            submit_lane_solution( work, hash+8, mythr, 1 );
+         }
+         n += 2;
+     } while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
     *hashes_done = n - first_nonce + 1;
     return 0;
 }
--- a/algo/qubit/qubit.c
+++ b/algo/qubit/qubit.c
@@ -7,7 +7,7 @@
 #include "algo/cubehash/cubehash_sse2.h" 
 #include "algo/simd/nist.h"
 #include "algo/shavite/sph_shavite.h"
-#ifndef NO_AES_NI
+#ifdef __AES__
 #include "algo/echo/aes_ni/hash_api.h"
 #else
 #include "algo/echo/sph_echo.h"
@@ -19,10 +19,10 @@ typedef struct
        cubehashParam           cubehash;
        sph_shavite512_context  shavite;
        hashState_sd            simd;
-#ifdef NO_AES_NI
-        sph_echo512_context echo;
-#else
+#ifdef __AES__
        hashState_echo          echo;
+#else
+        sph_echo512_context echo;
 #endif
 } qubit_ctx_holder;

@@ -35,10 +35,10 @@ void init_qubit_ctx()
        cubehashInit(&qubit_ctx.cubehash,512,16,32);
        sph_shavite512_init(&qubit_ctx.shavite);
        init_sd(&qubit_ctx.simd,512);
-#ifdef NO_AES_NI
-        sph_echo512_init(&qubit_ctx.echo);
-#else
+#ifdef __AES__
        init_echo(&qubit_ctx.echo, 512);
+#else
+        sph_echo512_init(&qubit_ctx.echo);
 #endif
 };

@@ -71,12 +71,12 @@ void qubit_hash(void *output, const void *input)
        update_final_sd( &ctx.simd, (BitSequence *)hash,
                         (const BitSequence*)hash,  512 );

-#ifdef NO_AES_NI
-        sph_echo512 (&ctx.echo, (const void*) hash, 64);
-        sph_echo512_close(&ctx.echo, (void*) hash);
-#else
+#ifdef __AES__
        update_final_echo( &ctx.echo, (BitSequence *) hash,
                     (const BitSequence *) hash, 512 );
+#else
+        sph_echo512 (&ctx.echo, (const void*) hash, 64);
+        sph_echo512_close(&ctx.echo, (void*) hash);
 #endif

        asm volatile ("emms");
@@ -104,48 +104,23 @@ int scanhash_qubit( struct work *work,	uint32_t max_nonce,

        qubit_luffa_midstate( endiandata );

-#ifdef DEBUG_ALGO
-	printf("[%d] Htarg=%X\n", thr_id, Htarg);
-#endif
 	for ( int m=0; m < 6; m++ )
-        {
+   {
 	    if ( Htarg <= htmax[m] )
-            {
+       {
 	        uint32_t mask = masks[m];
 	        do
-                {
+           {
 	            pdata[19] = ++n;
-		    be32enc(&endiandata[19], n);
-		    qubit_hash(hash64, endiandata);
-#ifndef DEBUG_ALGO
-		    if (!(hash64[7] & mask))
-                    {
-                       if ( fulltest(hash64, ptarget) )
-                       {
-		          *hashes_done = n - first_nonce + 1;
-		          return true;
-                       }
-//                       else
-//                       {
-//                          applog(LOG_INFO, "Result does not validate on CPU!");
-//                       }
-                     }
-#else
-                    if (!(n % 0x1000) && !thr_id) printf(".");
-	        	if (!(hash64[7] & mask)) {
-		            printf("[%d]",thr_id);
-			    if (fulltest(hash64, ptarget)) {
-                             work_set_target_ratio( work, hash64 );
-                             *hashes_done = n - first_nonce + 1;
-				return true;
-	                    }
- 	                }
-#endif
-                } while ( n < max_nonce && !work_restart[thr_id].restart );
-                // see blake.c if else to understand the loop on htmax => mask
-            break;
-          } 
-        }
+		         be32enc(&endiandata[19], n);
+		         qubit_hash(hash64, endiandata);
+		         if (!(hash64[7] & mask))
+               if ( fulltest(hash64, ptarget) )
+                  submit_solution( work, hash64, mythr );
+           } while ( n < max_nonce && !work_restart[thr_id].restart );
+           break;
+       } 
+   }

 	*hashes_done = n - first_nonce + 1;
 	pdata[19] = n;
--- a/algo/scrypt/scrypt.c
+++ b/algo/scrypt/scrypt.c
@@ -753,10 +753,8 @@ extern int scanhash_scrypt( struct work *work, uint32_t max_nonce,
 		
 		for (i = 0; i < throughput; i++) {
 			if (unlikely(hash[i * 8 + 7] <= Htarg && fulltest(hash + i * 8, ptarget))) {
-				*hashes_done = n - pdata[19] + 1;
 				pdata[19] = data[i * 20 + 19];
-                                work_set_target_ratio( work, hash );
-				return 1;
+            submit_solution( work, hash, mythr );
 			}
 		}
 	} while (likely(n < max_nonce && !work_restart[thr_id].restart));
--- a/algo/sha/hmac-sha256-hash.c
+++ b/algo/sha/hmac-sha256-hash.c
@@ -28,46 +28,10 @@

 #include <stdint.h>
 #include <string.h>
-
-#include "sysendian.h"
-
-#include "sha256_p.h"
+#include "simd-utils.h"
+#include "hmac-sha256-hash.h"
 #include "compat.h"

-
-/* Elementary functions used by SHA256 */
-#define Ch(x, y, z)	((x & (y ^ z)) ^ z)
-#define Maj(x, y, z)	((x & (y | z)) | (y & z))
-#define SHR(x, n)	(x >> n)
-#define ROTR(x, n)	((x >> n) | (x << (32 - n)))
-#define S0(x)		(ROTR(x, 2) ^ ROTR(x, 13) ^ ROTR(x, 22))
-#define S1(x)		(ROTR(x, 6) ^ ROTR(x, 11) ^ ROTR(x, 25))
-#define s0(x)		(ROTR(x, 7) ^ ROTR(x, 18) ^ SHR(x, 3))
-#define s1(x)		(ROTR(x, 17) ^ ROTR(x, 19) ^ SHR(x, 10))
-
-/* SHA256 round function */
-#define RND(a, b, c, d, e, f, g, h, k)			\
-	t0 = h + S1(e) + Ch(e, f, g) + k;		\
-	t1 = S0(a) + Maj(a, b, c);			\
-	d += t0;					\
-	h  = t0 + t1;
-
-/* Adjusted round function for rotating state */
-#define RNDr(S, W, i, k)			\
-	RND(S[(64 - i) % 8], S[(65 - i) % 8],	\
-	    S[(66 - i) % 8], S[(67 - i) % 8],	\
-	    S[(68 - i) % 8], S[(69 - i) % 8],	\
-	    S[(70 - i) % 8], S[(71 - i) % 8],	\
-	    W[i] + k)
-
-/*
-static unsigned char PAD[64] = {
-	0x80, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
-};
-*/
 /**
 * SHA256_Buf(in, len, digest):
 * Compute the SHA256 hash of ${len} bytes from ${in} and write it to ${digest}.
@@ -76,9 +40,9 @@ void
 SHA256_Buf( const void * in, size_t len, uint8_t digest[32] )
 {
 	SHA256_CTX ctx;
-        SHA256_Init( &ctx );
-        SHA256_Update( &ctx, in, len );
-        SHA256_Final( digest, &ctx );
+   SHA256_Init( &ctx );
+   SHA256_Update( &ctx, in, len );
+   SHA256_Final( digest, &ctx );
 }

 /**
@@ -87,19 +51,18 @@ SHA256_Buf( const void * in, size_t len, uint8_t digest[32] )
 * length ${Klen}, and write the result to ${digest}.
 */
 void
-HMAC_SHA256_Buf(const void * K, size_t Klen, const void * in, size_t len,
-    uint8_t digest[32])
+HMAC_SHA256_Buf( const void *K, size_t Klen, const void *in, size_t len,
+                 uint8_t digest[32])
 {
-        HMAC_SHA256_CTX ctx;
-
-        HMAC_SHA256_Init( &ctx, K, Klen );
-        HMAC_SHA256_Update( &ctx, in, len );
-        HMAC_SHA256_Final( digest, &ctx );
+   HMAC_SHA256_CTX ctx;
+   HMAC_SHA256_Init( &ctx, K, Klen );
+   HMAC_SHA256_Update( &ctx, in, len );
+   HMAC_SHA256_Final( digest, &ctx );
 }

 /* Initialize an HMAC-SHA256 operation with the given key. */
 void
-HMAC_SHA256_Init( HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen )
+HMAC_SHA256_Init( HMAC_SHA256_CTX *ctx, const void *_K, size_t Klen )
 {
 	unsigned char pad[64];
 	unsigned char khash[32];
@@ -107,7 +70,8 @@ HMAC_SHA256_Init( HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen )
 	size_t i;

 	/* If Klen > 64, the key is really SHA256(K). */
-	if (Klen > 64) {
+	if ( Klen > 64 )
+   {
 		SHA256_Init( &ctx->ictx );
 		SHA256_Update( &ctx->ictx, K, Klen );
 		SHA256_Final( khash, &ctx->ictx );
@@ -116,7 +80,7 @@ HMAC_SHA256_Init( HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen )
 	}

 	/* Inner SHA256 operation is SHA256(K xor [block of 0x36] || data). */
-        SHA256_Init( &ctx->ictx );
+   SHA256_Init( &ctx->ictx );
 	memset( pad, 0x36, 64 );
 	for ( i = 0; i < Klen; i++ )
 		pad[i] ^= K[i];
@@ -128,23 +92,19 @@ HMAC_SHA256_Init( HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen )
 	for ( i = 0; i < Klen; i++ )
 		pad[i] ^= K[i];
 	SHA256_Update( &ctx->octx, pad, 64 );
-
-	/* Clean the stack. */
-	//memset(khash, 0, 32);
 }

 /* Add bytes to the HMAC-SHA256 operation. */
 void
-HMAC_SHA256_Update(HMAC_SHA256_CTX * ctx, const void *in, size_t len)
+HMAC_SHA256_Update( HMAC_SHA256_CTX *ctx, const void *in, size_t len )
 {
-
 	/* Feed data to the inner SHA256 operation. */
 	SHA256_Update( &ctx->ictx, in, len );
 }

 /* Finish an HMAC-SHA256 operation. */
 void
-HMAC_SHA256_Final(unsigned char digest[32], HMAC_SHA256_CTX * ctx )
+HMAC_SHA256_Final( unsigned char digest[32], HMAC_SHA256_CTX *ctx )
 {
 	unsigned char ihash[32];

@@ -156,9 +116,6 @@ HMAC_SHA256_Final(unsigned char digest[32], HMAC_SHA256_CTX * ctx )

 	/* Finish the outer SHA256 operation. */
 	SHA256_Final( digest, &ctx->octx );
-
-	/* Clean the stack. */
-	//memset(ihash, 0, 32);
 }

 /**
@@ -167,52 +124,51 @@ HMAC_SHA256_Final(unsigned char digest[32], HMAC_SHA256_CTX * ctx )
 * write the output to buf.  The value dkLen must be at most 32 * (2^32 - 1).
 */
 void
-PBKDF2_SHA256(const uint8_t * passwd, size_t passwdlen, const uint8_t * salt,
-    size_t saltlen, uint64_t c, uint8_t * buf, size_t dkLen)
+PBKDF2_SHA256( const uint8_t *passwd, size_t passwdlen, const uint8_t *salt,
+               size_t saltlen, uint64_t c, uint8_t *buf, size_t dkLen )
 {
 	HMAC_SHA256_CTX PShctx, hctx;
 	uint8_t _ALIGN(128) T[32];
 	uint8_t _ALIGN(128) U[32];
-	uint8_t ivec[4];
+   uint32_t ivec;
 	size_t i, clen;
 	uint64_t j;
 	int k;

 	/* Compute HMAC state after processing P and S. */
-	HMAC_SHA256_Init(&PShctx, passwd, passwdlen);
-	HMAC_SHA256_Update(&PShctx, salt, saltlen);
+	HMAC_SHA256_Init( &PShctx, passwd, passwdlen );
+	HMAC_SHA256_Update( &PShctx, salt, saltlen );

 	/* Iterate through the blocks. */
-	for (i = 0; i * 32 < dkLen; i++) {
+	for ( i = 0; i * 32 < dkLen; i++ )
+   {
 		/* Generate INT(i + 1). */
-		be32enc(ivec, (uint32_t)(i + 1));
+      ivec = bswap_32( i+1 );

 		/* Compute U_1 = PRF(P, S || INT(i)). */
-		memcpy(&hctx, &PShctx, sizeof(HMAC_SHA256_CTX));
-		HMAC_SHA256_Update(&hctx, ivec, 4);
-		HMAC_SHA256_Final(U, &hctx);
+		memcpy( &hctx, &PShctx, sizeof(HMAC_SHA256_CTX) );
+		HMAC_SHA256_Update( &hctx, &ivec, 4 );
+		HMAC_SHA256_Final( U, &hctx );

 		/* T_i = U_1 ... */
-		memcpy(T, U, 32);
+		memcpy( T, U, 32 );

-		for (j = 2; j <= c; j++) {
+		for ( j = 2; j <= c; j++ )
+      {
 			/* Compute U_j. */
-			HMAC_SHA256_Init(&hctx, passwd, passwdlen);
-			HMAC_SHA256_Update(&hctx, U, 32);
-			HMAC_SHA256_Final(U, &hctx);
+			HMAC_SHA256_Init( &hctx, passwd, passwdlen );
+			HMAC_SHA256_Update( &hctx, U, 32 );
+			HMAC_SHA256_Final( U, &hctx );

 			/* ... xor U_j ... */
-			for (k = 0; k < 32; k++)
+			for ( k = 0; k < 32; k++ )
 				T[k] ^= U[k];
 		}

 		/* Copy as many bytes as necessary into buf. */
 		clen = dkLen - i * 32;
-		if (clen > 32)
+		if ( clen > 32 )
 			clen = 32;
-		memcpy(&buf[i * 32], T, clen);
+		memcpy( &buf[i * 32], T, clen );
 	}
-
-	/* Clean PShctx, since we never called _Final on it. */
-	//memset(&PShctx, 0, sizeof(HMAC_SHA256_CTX_Y));
 }
--- a/algo/sha/hmac-sha256-hash.h
+++ b/algo/sha/hmac-sha256-hash.h
@@ -26,23 +26,24 @@
 * $FreeBSD: src/lib/libmd/sha256_Y.h,v 1.2 2006/01/17 15:35:56 phk Exp $
 */

-#ifndef _SHA256_H_
-#define _SHA256_H_
+#ifndef HMAC_SHA256_H__
+#define HMAC_SHA256_H__

 #include <sys/types.h>
 #include <stdint.h>
 #include <openssl/sha.h>

-typedef struct HMAC_SHA256Context {
-        SHA256_CTX ictx;
-        SHA256_CTX octx;
+typedef struct HMAC_SHA256Context
+{
+     SHA256_CTX ictx;
+     SHA256_CTX octx;
 } HMAC_SHA256_CTX;

-void SHA256_Buf( const void * in, size_t len, uint8_t digest[32] );
+void SHA256_Buf( const void *, size_t len, uint8_t digest[32] );
 void HMAC_SHA256_Init( HMAC_SHA256_CTX *, const void *, size_t );
 void HMAC_SHA256_Update( HMAC_SHA256_CTX *, const void *, size_t );
 void HMAC_SHA256_Final( unsigned char [32], HMAC_SHA256_CTX * );
-void HMAC_SHA256_Buf( const void * K, size_t Klen, const void * in,
+void HMAC_SHA256_Buf( const void *, size_t Klen, const void *,
                      size_t len, uint8_t digest[32] );

 /**
@@ -53,4 +54,4 @@ void HMAC_SHA256_Buf( const void * K, size_t Klen, const void * in,
 void PBKDF2_SHA256( const uint8_t *, size_t, const uint8_t *, size_t,
                    uint64_t, uint8_t *, size_t);

-#endif /* !_SHA256_H_ */
+#endif // HMAC_SHA256_H__
--- a/algo/sha/sha256-hash-4way.c
+++ b/algo/sha/sha256-hash-4way.c
@@ -94,6 +94,37 @@ static const uint32_t K256[64] =
   _mm_xor_si128( _mm_xor_si128( \
        mm128_ror_32(x, 17), mm128_ror_32(x, 19) ), _mm_srli_epi32(x, 10) )

+#define SHA2s_4WAY_STEP(A, B, C, D, E, F, G, H, i, j) \
+do { \
+  __m128i K = _mm_set1_epi32( K256[( (j)+(i) )] ); \
+  __m128i T1 = mm128_ror_32( E, 14 ); \
+  __m128i T2 = mm128_ror_32( A,  9 ); \
+  __m128i T3 = _mm_xor_si128( F, G ); \
+  __m128i T4 = _mm_or_si128( A, B ); \
+  __m128i T5 = _mm_and_si128( A, B ); \
+  K  = _mm_add_epi32( K, W[i] ); \
+  T1 = _mm_xor_si128( T1, E ); \
+  T2 = _mm_xor_si128( T2, A ); \
+  T3 = _mm_and_si128( T3, E ); \
+  T4 = _mm_and_si128( T4, C ); \
+  K  = _mm_add_epi32( H, K ); \
+  T1 = mm128_ror_32( T1,  5 ); \
+  T2 = mm128_ror_32( T2, 11 ); \
+  T3 = _mm_xor_si128( T3, G ); \
+  T4 = _mm_or_si128( T4, T5 ); \
+  T1 = _mm_xor_si128( T1, E ); \
+  T2 = _mm_xor_si128( T2, A ); \
+  T1 = mm128_ror_32( T1,  6 ); \
+  T2 = mm128_ror_32( T2,  2 ); \
+  T1 = _mm_add_epi32( T1, T3 ); \
+  T2 = _mm_add_epi32( T2, T4 ); \
+  T1 = _mm_add_epi32( T1, K ); \
+  H  = _mm_add_epi32( T1, T2 ); \
+  D  = _mm_add_epi32( D, T1 ); \
+} while (0)
+
+
+/*
 #define SHA2s_4WAY_STEP(A, B, C, D, E, F, G, H, i, j) \
 do { \
  __m128i T1, T2; \
@@ -104,6 +135,8 @@ do { \
  D  = _mm_add_epi32( D,  T1 ); \
  H  = _mm_add_epi32( T1, T2 ); \
 } while (0)
+*/
+

 static void
 sha256_4way_round( sha256_4way_context *ctx, __m128i *in, __m128i r[8] )
--- a/algo/sha/sha512-hash-4way.c
+++ b/algo/sha/sha512-hash-4way.c
@@ -319,7 +319,7 @@ void sha512_8way_close( sha512_8way_context *sc, void *dst )

 // SHA-512 4 way 64 bit

-
+/*
 #define CH(X, Y, Z) \
   _mm256_xor_si256( _mm256_and_si256( _mm256_xor_si256( Y, Z ), X ), Z ) 

@@ -327,6 +327,15 @@ void sha512_8way_close( sha512_8way_context *sc, void *dst )
   _mm256_or_si256( _mm256_and_si256( X, Y ), \
                    _mm256_and_si256( _mm256_or_si256( X, Y ), Z ) )

+#define BSG5_0(x) \
+  mm256_ror_64( _mm256_xor_si256( mm256_ror_64( \
+                   _mm256_xor_si256( mm256_ror_64( x,  5 ), x ), 6 ), x ), 28 )
+
+#define BSG5_1(x) \
+  mm256_ror_64( _mm256_xor_si256( mm256_ror_64( \
+                   _mm256_xor_si256( mm256_ror_64( x, 23 ), x ), 4 ), x ), 14 )
+*/
+/*
 #define BSG5_0(x) \
   _mm256_xor_si256( _mm256_xor_si256( \
        mm256_ror_64(x, 28), mm256_ror_64(x, 34) ), mm256_ror_64(x, 39) )
@@ -334,7 +343,8 @@ void sha512_8way_close( sha512_8way_context *sc, void *dst )
 #define BSG5_1(x) \
   _mm256_xor_si256( _mm256_xor_si256( \
        mm256_ror_64(x, 14), mm256_ror_64(x, 18) ), mm256_ror_64(x, 41) )
-
+*/
+/*
 #define SSG5_0(x) \
   _mm256_xor_si256( _mm256_xor_si256( \
        mm256_ror_64(x,  1), mm256_ror_64(x,  8) ), _mm256_srli_epi64(x, 7) ) 
@@ -342,7 +352,7 @@ void sha512_8way_close( sha512_8way_context *sc, void *dst )
 #define SSG5_1(x) \
   _mm256_xor_si256( _mm256_xor_si256( \
        mm256_ror_64(x, 19), mm256_ror_64(x, 61) ), _mm256_srli_epi64(x, 6) )
-
+*/
 // Interleave SSG0 & SSG1 for better throughput.
 // return ssg0(w0) + ssg1(w1)
 static inline __m256i ssg512_add( __m256i w0, __m256i w1 )
@@ -361,7 +371,7 @@ static inline __m256i ssg512_add( __m256i w0, __m256i w1 )
   return _mm256_add_epi64( w0a, w1a );
 }

-
+/*
 #define SSG512x2_0( w0, w1, i ) do \
 { \
   __m256i X0a, X1a, X0b, X1b; \
@@ -391,7 +401,51 @@ static inline __m256i ssg512_add( __m256i w0, __m256i w1 )
  w0  = _mm256_xor_si256( X0a, X0b ); \
  w1  = _mm256_xor_si256( X1a, X1b ); \
 } while(0)
+*/

+#define SHA3_4WAY_STEP(A, B, C, D, E, F, G, H, i) \
+do { \
+  __m256i K = _mm256_set1_epi64x( K512[ i ] ); \
+  __m256i T1 = mm256_ror_64( E, 23 ); \
+  __m256i T2 = mm256_ror_64( A,  5 ); \
+  __m256i T3 = _mm256_xor_si256( F, G ); \
+  __m256i T4 = _mm256_or_si256( A, B ); \
+  __m256i T5 = _mm256_and_si256( A, B ); \
+  K = _mm256_add_epi64( K, W[i] ); \
+  T1 = _mm256_xor_si256( T1, E ); \
+  T2 = _mm256_xor_si256( T2, A ); \
+  T3 = _mm256_and_si256( T3, E ); \
+  T4 = _mm256_and_si256( T4, C ); \
+  K = _mm256_add_epi64( H, K ); \
+  T1 = mm256_ror_64( T1, 4 ); \
+  T2 = mm256_ror_64( T2, 6 ); \
+  T3 = _mm256_xor_si256( T3, G ); \
+  T4 = _mm256_or_si256( T4, T5 ); \
+  T1 = _mm256_xor_si256( T1, E ); \
+  T2 = _mm256_xor_si256( T2, A ); \
+  T1 = mm256_ror_64( T1, 14 ); \
+  T2 = mm256_ror_64( T2, 28 ); \
+  T1 = _mm256_add_epi64( T1, T3 ); \
+  T2 = _mm256_add_epi64( T2, T4 ); \
+  T1 = _mm256_add_epi64( T1, K ); \
+  H  = _mm256_add_epi64( T1, T2 ); \
+  D  = _mm256_add_epi64( D, T1 ); \
+} while (0)
+
+/*
+#define SHA3_4WAY_STEP(A, B, C, D, E, F, G, H, i) \
+do { \
+  __m256i K = _mm256_add_epi64( W[i], _mm256_set1_epi64x( K512[ i ] ) ); \
+  __m256i T1 = BSG5_1(E); \
+  __m256i T2 = BSG5_0(A); \
+  T1 = mm256_add4_64( T1, H, CH(E, F, G), K ); \
+  T2 = _mm256_add_epi64( T2, MAJ(A, B, C) ); \
+  D  = _mm256_add_epi64( D, T1 ); \
+  H  = _mm256_add_epi64( T1, T2 ); \
+} while (0)
+*/
+
+/*
 #define SHA3_4WAY_STEP(A, B, C, D, E, F, G, H, i) \
 do { \
  __m256i T1, T2; \
@@ -402,7 +456,7 @@ do { \
  D  = _mm256_add_epi64( D, T1 ); \
  H  = _mm256_add_epi64( T1, T2 ); \
 } while (0)
-
+*/

 static void
 sha512_4way_round( sha512_4way_context *ctx,  __m256i *in, __m256i r[8] )
--- a/algo/shavite/shavite-hash-2way.c
+++ b/algo/shavite/shavite-hash-2way.c
@@ -3,11 +3,9 @@

 #include <stdio.h>

-// This implementation is deprecated, superseded by VAES in Icelake
-// which provides HW based 4 way aes.
-// It was created for AVX2 to eliminate interleaving between the 
-// preceding and following function.
-// This code can be removed when current users have reverted to one way.
+// This is a fake, it actually does not do parallel AES, that requires VAES.
+// This is only intended when the preceding and folllowing functions use the
+// same 2x128 interleave.

 #if defined(__AVX2__)

@@ -410,4 +408,94 @@ void shavite512_2way_update_close( shavite512_2way_context *ctx, void *dst,
   casti_m256i( dst, 3 ) = casti_m256i( ctx->h, 3 );
 }

+void shavite512_2way_full( shavite512_2way_context *ctx, void *dst,
+                           const void *data, size_t len )
+{
+    __m256i *h = (__m256i*)ctx->h;
+    __m128i *iv = (__m128i*)IV512;
+
+   h[0] = m256_const1_128( iv[0] );
+   h[1] = m256_const1_128( iv[1] );
+   h[2] = m256_const1_128( iv[2] );
+   h[3] = m256_const1_128( iv[3] );
+
+   ctx->ptr    =
+   ctx->count0 =
+   ctx->count1 =
+   ctx->count2 =
+   ctx->count3 = 0;
+
+   unsigned char *buf = ctx->buf;
+   size_t         ptr = ctx->ptr;
+
+   // process full blocks and load buf with remainder.
+   while ( len > 0 )
+   {
+      size_t clen;
+
+      clen = (sizeof ctx->buf) - ptr;
+      if ( clen > len << 1 )
+         clen = len << 1;
+      memcpy( buf + ptr, data, clen );
+      data = (const unsigned char *)data + clen;
+      ptr += clen;
+      len -= (clen >> 1);
+      if ( ptr == sizeof ctx->buf )
+      {
+         if ( ( ctx->count0 = ctx->count0 + 1024 )  == 0 )
+         {
+             ctx->count1 = ctx->count1 + 1;
+             if ( ctx->count1 == 0 )
+             {
+                ctx->count2 = ctx->count2 + 1;
+                if ( ctx->count2 == 0 )
+                   ctx->count3 = ctx->count3 + 1;
+             }
+         }
+         c512_2way( ctx, buf );
+         ptr = 0;
+      }
+   }
+
+   uint32_t vp = ptr>>5;
+   // Count = { 0, 16, 64, 80 }. Outsize = 16 u32 = 512 bits = 0x0200
+   // Count is misaligned to 16 bits and straddles 2 vectors.
+   // Use u32 overlay to stage then u16 to load buf.
+   union
+   {
+      uint32_t u32[4];
+      uint16_t u16[8];
+   } count;
+
+   count.u32[0] = ctx->count0 += (ptr << 2);  // ptr/2 * 8
+   count.u32[1] = ctx->count1;
+   count.u32[2] = ctx->count2;
+   count.u32[3] = ctx->count3;
+
+   if ( vp == 0 )    // empty buf, xevan.
+   {
+      casti_m256i( buf, 0 ) = m256_const2_64( 0, 0x0000000000000080 );
+      memset_zero_256( (__m256i*)buf + 1, 5 );
+      ctx->count0 = ctx->count1 = ctx->count2 = ctx->count3 = 0;
+   }
+   else     // half full buf, everyone else.
+   {
+    casti_m256i( buf, vp++ ) = m256_const2_64( 0, 0x0000000000000080 );
+      memset_zero_256( (__m256i*)buf + vp, 6 - vp );
+   }
+
+    casti_m256i( buf, 6 ) = m256_const1_128(
+                  _mm_insert_epi16( m128_zero, count.u16[0], 7 ) );
+    casti_m256i( buf, 7 ) = m256_const1_128( _mm_set_epi16(
+                  0x0200,       count.u16[7], count.u16[6], count.u16[5],
+                  count.u16[4], count.u16[3], count.u16[2], count.u16[1] ) );
+
+   c512_2way( ctx, buf);
+
+   casti_m256i( dst, 0 ) = casti_m256i( ctx->h, 0 );
+   casti_m256i( dst, 1 ) = casti_m256i( ctx->h, 1 );
+   casti_m256i( dst, 2 ) = casti_m256i( ctx->h, 2 );
+   casti_m256i( dst, 3 ) = casti_m256i( ctx->h, 3 );
+}
+   
 #endif // AVX2
--- a/algo/shavite/shavite-hash-2way.h
+++ b/algo/shavite/shavite-hash-2way.h
@@ -18,6 +18,8 @@ void shavite512_2way_update( shavite512_2way_context *ctx, const void *data,
 void shavite512_2way_close( shavite512_2way_context *ctx, void *dst );
 void shavite512_2way_update_close( shavite512_2way_context *ctx, void *dst,
 		                   const void *data, size_t len );
+void shavite512_2way_full( shavite512_2way_context *ctx, void *dst,
+                           const void *data, size_t len );

 #endif // AVX2

--- a/algo/shavite/shavite-hash-4way.c
+++ b/algo/shavite/shavite-hash-4way.c
@@ -1,6 +1,8 @@
 #include "shavite-hash-4way.h"
 #include <stdint.h>

+#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
 static const uint32_t IV512[] =
 {
        0x72FCCDD8, 0x79CA4727, 0x128A077B, 0x40D55AEC,
@@ -9,8 +11,6 @@ static const uint32_t IV512[] =
        0xE275EADE, 0x502D9FCD, 0xB9357178, 0x022A4B9A
 };

-#if defined(__VAES__) && defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
-
 #define mm512_ror2x512hi_1x32( a, b ) \
   _mm512_mask_blend_epi32( 0x8888, mm512_ror128_32( a ), \
                                    mm512_ror128_32( b ) )
@@ -396,4 +396,96 @@ void shavite512_4way_update_close( shavite512_4way_context *ctx, void *dst,
   casti_m512i( dst, 3 ) = casti_m512i( ctx->h, 3 );
 }

+
+void shavite512_4way_full( shavite512_4way_context *ctx, void *dst,
+                           const void *data, size_t len )
+{
+    __m512i *h = (__m512i*)ctx->h;
+    __m128i *iv = (__m128i*)IV512;
+
+   h[0] = m512_const1_128( iv[0] );
+   h[1] = m512_const1_128( iv[1] );
+   h[2] = m512_const1_128( iv[2] );
+   h[3] = m512_const1_128( iv[3] );
+
+   ctx->ptr    = 
+   ctx->count0 = 
+   ctx->count1 =
+   ctx->count2 =
+   ctx->count3 = 0;
+
+   unsigned char *buf = ctx->buf;
+   size_t         ptr = ctx->ptr;
+
+   // process full blocks and load buf with remainder.
+   while ( len > 0 )
+   {
+      size_t clen;
+
+      clen = (sizeof ctx->buf) - ptr;
+      if ( clen > len << 2 )
+         clen = len << 2;
+      memcpy( buf + ptr, data, clen );
+      data = (const unsigned char *)data + clen;
+      ptr += clen;
+      len -= (clen >> 2);
+      if ( ptr == sizeof ctx->buf )
+      {
+         if ( ( ctx->count0 = ctx->count0 + 1024 )  == 0 )
+         {
+             ctx->count1 = ctx->count1 + 1;
+             if ( ctx->count1 == 0 )
+             {
+                ctx->count2 = ctx->count2 + 1;
+                if ( ctx->count2 == 0 )
+                   ctx->count3 = ctx->count3 + 1;
+             }
+         }
+         c512_4way( ctx, buf );
+         ptr = 0;
+      }
+   }
+
+   uint32_t vp = ptr>>6;
+   // Count = { 0, 16, 64, 80 }. Outsize = 16 u32 = 512 bits = 0x0200
+   // Count is misaligned to 16 bits and straddles 2 vectors.
+   // Use u32 overlay to stage then u16 to load buf.
+   union
+   {
+      uint32_t u32[4];
+      uint16_t u16[8];
+   } count;
+
+   count.u32[0] = ctx->count0 += (ptr << 1);  // ptr/4 * 8
+   count.u32[1] = ctx->count1;
+   count.u32[2] = ctx->count2;
+   count.u32[3] = ctx->count3;
+
+   if ( vp == 0 )    // empty buf, xevan.
+   {
+      casti_m512i( buf, 0 ) = m512_const2_64( 0, 0x0000000000000080 );
+      memset_zero_512( (__m512i*)buf + 1, 5 );
+      ctx->count0 = ctx->count1 = ctx->count2 = ctx->count3 = 0;
+   }
+   else     // half full buf, everyone else.
+   {
+    casti_m512i( buf, vp++ ) = m512_const2_64( 0, 0x0000000000000080 );
+      memset_zero_512( (__m512i*)buf + vp, 6 - vp );
+   }
+
+    casti_m512i( buf, 6 ) = m512_const1_128(
+                  _mm_insert_epi16( m128_zero, count.u16[0], 7 ) );
+    casti_m512i( buf, 7 ) = m512_const1_128( _mm_set_epi16(
+                  0x0200,       count.u16[7], count.u16[6], count.u16[5],
+                  count.u16[4], count.u16[3], count.u16[2], count.u16[1] ) );
+
+   c512_4way( ctx, buf);
+
+   casti_m512i( dst, 0 ) = casti_m512i( ctx->h, 0 );
+   casti_m512i( dst, 1 ) = casti_m512i( ctx->h, 1 );
+   casti_m512i( dst, 2 ) = casti_m512i( ctx->h, 2 );
+   casti_m512i( dst, 3 ) = casti_m512i( ctx->h, 3 );
+}
+
+
 #endif // VAES
--- a/algo/shavite/shavite-hash-4way.h
+++ b/algo/shavite/shavite-hash-4way.h
@@ -18,6 +18,8 @@ void shavite512_4way_update( shavite512_4way_context *ctx, const void *data,
 void shavite512_4way_close( shavite512_4way_context *ctx, void *dst );
 void shavite512_4way_update_close( shavite512_4way_context *ctx, void *dst,
 		                   const void *data, size_t len );
+void shavite512_4way_full( shavite512_4way_context *ctx, void *dst,
+                           const void *data, size_t len );

 #endif // VAES

--- a/algo/simd/simd-hash-2way.c
+++ b/algo/simd/simd-hash-2way.c
@@ -1035,7 +1035,7 @@ int simd_4way_update( simd_4way_context *state, const void *data,

  while ( databitlen > 0 )
  {
-    if ( current == 0 && databitlen >= bs )
+    if ( ( current == 0 ) && ( databitlen >= bs ) )
    {
       // We can hash the data directly from the input buffer.
      SIMD_4way_Compress( state, data, 0 );
@@ -1049,13 +1049,13 @@ int simd_4way_update( simd_4way_context *state, const void *data,
      int len = bs - current;
      if ( databitlen < len )
      {
-        memcpy( state->buffer + 4*(current/8), data, 4*((databitlen+7)/8) );
+        memcpy( state->buffer + 4 * (current/8), data, 4 * (databitlen/8) );
        state->count += databitlen;
        return 0;
      }
      else
      {
-        memcpy( state->buffer + 4*(current/8), data, 4*(len/8) );
+        memcpy( state->buffer + 4 * (current / 8), data, 4 * (len / 8) );
        state->count += len;
        databitlen -= len;
        data += 4*(len/8);
@@ -1128,7 +1128,7 @@ int simd_4way_update_close( simd_4way_context *state, void *hashval,
      int len = bs - current;
      if ( databitlen < len )
      {
-        memcpy( state->buffer + 4*( current/8 ), data, 4*( (databitlen+7)/8 ) );
+        memcpy( state->buffer + 4*( current/8 ), data, 4*( (databitlen)/8 ) );
        state->count += databitlen;
        break;
      }
@@ -1149,7 +1149,7 @@ int simd_4way_update_close( simd_4way_context *state, void *hashval,
  // If there is still some data in the buffer, hash it
  if ( current )
  {
-    current = ( current+7 ) / 8;
+    current = current / 8;
    memset( state->buffer + 4*current, 0, 4*( state->blocksize/8 - current) );
    SIMD_4way_Compress( state, state->buffer, 0 );
  }
@@ -1173,6 +1173,91 @@ int simd_4way_update_close( simd_4way_context *state, void *hashval,
  return 0;
 }

+int simd512_4way_full( simd_4way_context *state, void *hashval,
+                    const void *data, int datalen )
+{
+  __m512i *A = (__m512i*)state->A;
+
+  state->hashbitlen = 512;
+  state->n_feistels = 8;
+  state->blocksize = 128*8;
+  state->count = 0;
+
+  for ( int i = 0; i < 8; i++ )
+       A[i] = _mm512_set4_epi32( SIMD_IV_512[4*i+3], SIMD_IV_512[4*i+2],
+                                 SIMD_IV_512[4*i+1], SIMD_IV_512[4*i+0] );
+
+  int current, i;
+  int bs = state->blocksize;  // bits in one lane
+  int isshort = 1;
+  uint64_t l;
+  int databitlen = datalen * 8;
+
+  current = state->count & (bs - 1);
+
+  while ( databitlen > 0 )
+  {
+    if ( current == 0 && databitlen >= bs )
+    {
+      // We can hash the data directly from the input buffer.
+      SIMD_4way_Compress( state, data, 0 );
+      databitlen -= bs;
+      data += 4*( bs/8 );
+      state->count += bs;
+    }
+    else
+    {
+      // Copy a chunk of data to the buffer
+      int len = bs - current;
+      if ( databitlen < len )
+      {
+        memcpy( state->buffer + 4*( current/8 ), data, 4*( (databitlen)/8 ) );
+        state->count += databitlen;
+        break;
+      }
+      else
+      {
+        memcpy( state->buffer + 4*(current/8), data, 4*(len/8) );
+        state->count += len;
+        databitlen -= len;
+        data += 4*( len/8 );
+        current = 0;
+        SIMD_4way_Compress( state, state->buffer, 0 );
+      }
+    }
+  }
+
+  current = state->count & (state->blocksize - 1);
+
+  // If there is still some data in the buffer, hash it
+  if ( current )
+  {
+    current = current / 8;
+    memset( state->buffer + 4*current, 0, 4*( state->blocksize/8 - current) );
+    SIMD_4way_Compress( state, state->buffer, 0 );
+  }
+
+  //* Input the message length as the last block
+  memset( state->buffer, 0, 4*( state->blocksize/8 ) );
+  l = state->count;
+  for ( i = 0; i < 8; i++ )
+  {
+    state->buffer[ i    ] = l & 0xff;
+    state->buffer[ i+16 ] = l & 0xff;
+    state->buffer[ i+32 ] = l & 0xff;
+    state->buffer[ i+48 ] = l & 0xff;
+    l >>= 8;
+  }
+  if ( state->count < 16384 )
+    isshort = 2;
+
+  SIMD_4way_Compress( state, state->buffer, isshort );
+  memcpy( hashval, state->A, 4*( state->hashbitlen / 8 ) );
+  return 0;
+}
+
+
+
 #endif // AVX512

 ////////////////////////////////////
@@ -1929,4 +2014,90 @@ int simd_2way_update_close( simd_2way_context *state, void *hashval,
  return 0;
 }

+int simd512_2way_full( simd_2way_context *state, void *hashval,
+                    const void *data, int datalen )
+{
+  __m256i *A = (__m256i*)state->A;
+
+  state->hashbitlen = 512;
+  state->n_feistels = 8;
+  state->blocksize = 128*8;
+  state->count = 0;
+
+  for ( int i = 0; i < 8; i++ )
+       A[i] = _mm256_set_epi32( SIMD_IV_512[4*i+3], SIMD_IV_512[4*i+2],
+                                SIMD_IV_512[4*i+1], SIMD_IV_512[4*i+0],
+                                SIMD_IV_512[4*i+3], SIMD_IV_512[4*i+2],
+                                SIMD_IV_512[4*i+1], SIMD_IV_512[4*i+0] );
+
+  int current, i;
+  int bs = state->blocksize;  // bits in one lane
+  int isshort = 1;
+  uint64_t l;
+  int databitlen = datalen * 8;
+
+  current = state->count & (bs - 1);
+
+  while ( databitlen > 0 )
+  {
+    if ( current == 0 && databitlen >= bs )
+    {
+      // We can hash the data directly from the input buffer.
+      SIMD_2way_Compress( state, data, 0 );
+
+      databitlen -= bs;
+      data += 2*( bs/8 );
+      state->count += bs;
+    }
+    else
+    {
+      // Copy a chunk of data to the buffer
+      int len = bs - current;
+      if ( databitlen < len )
+      {
+
+         memcpy( state->buffer + 2*( current/8 ), data, 2*( (databitlen+7)/8 ) );
+        state->count += databitlen;
+        break;
+      }
+      else
+      {
+        memcpy( state->buffer + 2*(current/8), data, 2*(len/8) );
+        state->count += len;
+        databitlen -= len;
+        data += 2*( len/8 );
+        current = 0;
+        SIMD_2way_Compress( state, state->buffer, 0 );
+      }
+    }
+  }
+
+  current = state->count & (state->blocksize - 1);
+
+  // If there is still some data in the buffer, hash it
+  if ( current )
+  {
+    current = ( current+7 ) / 8;
+    memset( state->buffer + 2*current, 0, 2*( state->blocksize/8 - current) );
+    SIMD_2way_Compress( state, state->buffer, 0 );
+  }
+
+  //* Input the message length as the last block
+  memset( state->buffer, 0, 2*( state->blocksize/8 ) );
+  l = state->count;
+  for ( i = 0; i < 8; i++ )
+  {
+    state->buffer[ i    ] = l & 0xff;
+    state->buffer[ i+16 ] = l & 0xff;
+    l >>= 8;
+  }
+  if ( state->count < 16384 )
+    isshort = 2;
+
+  SIMD_2way_Compress( state, state->buffer, isshort );
+  memcpy( hashval, state->A, 2*( state->hashbitlen / 8 ) );
+  return 0;
+}
+
+
 #endif
--- a/algo/simd/simd-hash-2way.h
+++ b/algo/simd/simd-hash-2way.h
@@ -26,6 +26,8 @@ int simd_4way_update( simd_4way_context *state, const void *data,
 int simd_4way_close( simd_4way_context *state, void *hashval );
 int simd_4way_update_close( simd_4way_context *state, void *hashval,
                            const void *data, int databitlen );
+int simd512_4way_full( simd_4way_context *state, void *hashval,
+                    const void *data, int datalen );

 #endif

@@ -45,5 +47,8 @@ int simd_2way_update( simd_2way_context *state, const void *data,
 int simd_2way_close( simd_2way_context *state, void *hashval );
 int simd_2way_update_close( simd_2way_context *state, void *hashval,
                            const void *data, int databitlen );
+int simd512_2way_full( simd_2way_context *state, void *hashval,
+                    const void *data, int datalen );
+
 #endif
 #endif
--- a/algo/skein/skein-hash-4way.c
+++ b/algo/skein/skein-hash-4way.c
@@ -45,18 +45,18 @@ extern "C"{
 #endif

 /*
-static const sph_u64 IV256[] = {
-   SPH_C64(0xCCD044A12FDB3E13), SPH_C64(0xE83590301A79A9EB),
-   SPH_C64(0x55AEA0614F816E6F), SPH_C64(0x2A2767A4AE9B94DB),
-   SPH_C64(0xEC06025E74DD7683), SPH_C64(0xE7A436CDC4746251),
-   SPH_C64(0xC36FBAF9393AD185), SPH_C64(0x3EEDBA1833EDFC13)
+static const uint64_t IV256[] = {
+   0xCCD044A12FDB3E13, 0xE83590301A79A9EB,
+   0x55AEA0614F816E6F, 0x2A2767A4AE9B94DB,
+   0xEC06025E74DD7683, 0xE7A436CDC4746251,
+   0xC36FBAF9393AD185, 0x3EEDBA1833EDFC13
 };

-static const sph_u64 IV512[] = {
-   SPH_C64(0x4903ADFF749C51CE), SPH_C64(0x0D95DE399746DF03),
-   SPH_C64(0x8FD1934127C79BCE), SPH_C64(0x9A255629FF352CB1),
-   SPH_C64(0x5DB62599DF6CA7B0), SPH_C64(0xEABE394CA9D5C3F4),
-   SPH_C64(0x991112C71A75B523), SPH_C64(0xAE18A40B660FCC33)
+static const uint64_t IV512[] = {
+   0x4903ADFF749C51CE, 0x0D95DE399746DF03,
+   0x8FD1934127C79BCE, 0x9A255629FF352CB1,
+   0x5DB62599DF6CA7B0, 0xEABE394CA9D5C3F4,
+   0x991112C71A75B523, 0xAE18A40B660FCC33
 };
 */
   
@@ -372,7 +372,7 @@ do { \

 #define UBI_BIG_8WAY(etype, extra) \
 do { \
-  sph_u64 t0, t1, t2; \
+  uint64_t t0, t1, t2; \
  __m512i h8; \
  __m512i m0 =  buf[0]; \
  __m512i m1 =  buf[1]; \
@@ -391,8 +391,8 @@ do { \
  __m512i p5 = m5; \
  __m512i p6 = m6; \
  __m512i p7 = m7; \
-  t0 = SPH_T64(bcount << 6) + (sph_u64)(extra); \
-  t1 = (bcount >> 58) + ((sph_u64)(etype) << 55); \
+  t0 = (uint64_t)(bcount << 6) + (uint64_t)(extra); \
+  t1 = (bcount >> 58) + ((uint64_t)(etype) << 55); \
  TFBIG_KINIT_8WAY(h0, h1, h2, h3, h4, h5, h6, h7, h8, t0, t1, t2); \
  TFBIG_8WAY_4e(0); \
  TFBIG_8WAY_4o(1); \
@@ -425,7 +425,7 @@ do { \

 #define DECL_STATE_BIG_8WAY \
  __m512i h0, h1, h2, h3, h4, h5, h6, h7; \
-  sph_u64 bcount;
+  uint64_t bcount;


 #endif // AVX512
@@ -488,7 +488,7 @@ do { \
 // scale buf offset by 4
 #define UBI_BIG_4WAY(etype, extra) \
 do { \
-  sph_u64 t0, t1, t2; \
+  uint64_t t0, t1, t2; \
  __m256i h8; \
  __m256i m0 =  buf[0]; \
  __m256i m1 =  buf[1]; \
@@ -507,8 +507,8 @@ do { \
  __m256i p5 = m5; \
  __m256i p6 = m6; \
  __m256i p7 = m7; \
-  t0 = SPH_T64(bcount << 6) + (sph_u64)(extra); \
-  t1 = (bcount >> 58) + ((sph_u64)(etype) << 55); \
+  t0 = (uint64_t)(bcount << 6) + (uint64_t)(extra); \
+  t1 = (bcount >> 58) + ((uint64_t)(etype) << 55); \
  TFBIG_KINIT_4WAY(h0, h1, h2, h3, h4, h5, h6, h7, h8, t0, t1, t2); \
  TFBIG_4WAY_4e(0); \
  TFBIG_4WAY_4o(1); \
@@ -542,7 +542,7 @@ do { \

 #define DECL_STATE_BIG_4WAY \
  __m256i h0, h1, h2, h3, h4, h5, h6, h7; \
-  sph_u64 bcount;
+  uint64_t bcount;

 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)

--- a/algo/skein/skein-hash-4way.h
+++ b/algo/skein/skein-hash-4way.h
@@ -48,14 +48,8 @@ extern "C"{
 #endif

 #include <stddef.h>
-#include "algo/sha/sph_types.h"
 #include "simd-utils.h"

-// Output size in bits
-#define SPH_SIZE_skein256   256
-#define SPH_SIZE_skein512   512
-
-
 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)

 typedef struct
@@ -63,11 +57,11 @@ typedef struct
   __m512i buf[8];
   __m512i h0, h1, h2, h3, h4, h5, h6, h7;
   size_t ptr;
-   sph_u64 bcount;
-} sph_skein_8way_big_context __attribute__ ((aligned (128)));
+   uint64_t bcount;
+} skein_8way_big_context __attribute__ ((aligned (128)));

-typedef sph_skein_8way_big_context skein512_8way_context;
-typedef sph_skein_8way_big_context skein256_8way_context;
+typedef skein_8way_big_context skein512_8way_context;
+typedef skein_8way_big_context skein256_8way_context;

 void skein512_8way_init( skein512_8way_context *sc );
 void skein512_8way_update( void *cc, const void *data, size_t len );
@@ -84,21 +78,19 @@ typedef struct
   __m256i buf[8];
   __m256i h0, h1, h2, h3, h4, h5, h6, h7;
   size_t ptr;
-	sph_u64 bcount;
-} sph_skein_4way_big_context __attribute__ ((aligned (128)));
+	uint64_t bcount;
+} skein_4way_big_context __attribute__ ((aligned (128)));

-typedef sph_skein_4way_big_context skein512_4way_context;
-typedef sph_skein_4way_big_context skein256_4way_context;
+typedef skein_4way_big_context skein512_4way_context;
+typedef skein_4way_big_context skein256_4way_context;

 void skein512_4way_init( skein512_4way_context *sc );
 void skein512_4way_update( void *cc, const void *data, size_t len );
 void skein512_4way_close( void *cc, void *dst );
-//#define skein512_4way skein512_4way_update

 void skein256_4way_init( skein256_4way_context *sc );
 void skein256_4way_update( void *cc, const void *data, size_t len );
 void skein256_4way_close( void *cc, void *dst );
-//#define skein256_4way skein256_4way_update

 #ifdef __cplusplus
 }
--- a/algo/whirlpool/whirlpool.c
+++ b/algo/whirlpool/whirlpool.c
@@ -78,17 +78,12 @@ int scanhash_whirlpool( struct work* work, uint32_t max_nonce,
 	do {
 		const uint32_t Htarg = ptarget[7];
 		uint32_t vhash[8];
-                pdata[19] = ++n;
+      pdata[19] = ++n;
 		be32enc(&endiandata[19], n );
 		whirlpool_hash(vhash, endiandata);

 		if (vhash[7] <= Htarg && fulltest(vhash, ptarget))
-                {
-			work_set_target_ratio(work, vhash);
-                       *hashes_done = n - first_nonce + 1;
-			return true;
-		}
-
+             submit_solution( work, vhash, mythr );
 	} while ( n < max_nonce && !work_restart[thr_id].restart);

 	*hashes_done = n - first_nonce + 1;
--- a/algo/whirlpool/whirlpoolx.c
+++ b/algo/whirlpool/whirlpoolx.c
@@ -48,11 +48,7 @@ int scanhash_whirlpoolx( struct work* work, uint32_t max_nonce,
 		whirlpoolx_hash(vhash, endiandata);

 		if (vhash[7] <= Htarg && fulltest(vhash, ptarget))
-                {
-			work_set_target_ratio(work, vhash);
-                       *hashes_done = n - first_nonce + 1;
-			return true;
-		}
+          submit_solution( work, vhash, mythr );

 	} while ( n < max_nonce && !work_restart[thr_id].restart);

--- a/algo/x11/c11-4way.c
+++ b/algo/x11/c11-4way.c
@@ -275,7 +275,7 @@ int scanhash_c11_8way( struct work *work, uint32_t max_nonce,
        pdata[19] = n;

        for ( int i = 0; i < 8; i++ )
-        if ( ( ( hash+(i<<3) )[7] < Htarg )
+        if ( ( ( hash+(i<<3) )[7] <= Htarg )
             && fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
        {
           pdata[19] = n+i;
@@ -443,37 +443,26 @@ int scanhash_c11_4way( struct work *work, uint32_t max_nonce,
     int thr_id = mythr->id;  // thr_id arg is deprecated
     __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
     const uint32_t Htarg = ptarget[7];
-     uint64_t htmax[] = {          0,        0xF,       0xFF,
-                               0xFFF,     0xFFFF, 0x10000000  };
-     uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
-                          0xFFFFF000, 0xFFFF0000,          0  };

     mm256_bswap32_intrlv80_4x64( vdata, pdata );

-     for (int m=0; m < 6; m++) 
-       if (Htarg <= htmax[m])
-       {
-         uint32_t mask = masks[m];
-         do
-         {
-           *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
-                 _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
+     do
+     {
+        *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
+             _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );

-            c11_4way_hash( hash, vdata );
-            pdata[19] = n;
-
-            for ( int i = 0; i < 4; i++ )
-            if ( ( ( (hash+(i<<3))[7] & mask ) == 0 )
-                 && fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
-            {
-               pdata[19] = n+i;
-               submit_lane_solution( work, hash+(i<<3), mythr, i );
-            }
-            n += 4;
-         } while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
-         break;
-       }
+        c11_4way_hash( hash, vdata );
+        pdata[19] = n;

+        for ( int i = 0; i < 4; i++ )
+        if ( ( ( hash+(i<<3) )[7] <= Htarg )
+            && fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
+        {
+           pdata[19] = n+i;
+           submit_lane_solution( work, hash+(i<<3), mythr, i );
+        }
+        n += 4;
+     } while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
     *hashes_done = n - first_nonce + 1;
     return 0;
 }
--- a/algo/x11/c11.c
+++ b/algo/x11/c11.c
@@ -78,11 +78,9 @@ void c11_hash( void *output, const void *input )
    sph_bmw512_close( &ctx.bmw, hash );

 #if defined(__AES__)
-    init_groestl( &ctx.groestl, 64 );
    update_and_final_groestl( &ctx.groestl, (char*)hash,
                                      (const char*)hash, 512 );
 #else
-    sph_groestl512_init( &ctx.groestl );
    sph_groestl512( &ctx.groestl, hash, 64 );
    sph_groestl512_close( &ctx.groestl, hash );
 #endif
@@ -108,12 +106,12 @@ void c11_hash( void *output, const void *input )
     update_final_sd( &ctx.simd, (BitSequence *)hash,
                      (const BitSequence *)hash, 512 );

-#ifdef NO_AES_NI
-     sph_echo512( &ctx.echo, hash, 64 );
-     sph_echo512_close( &ctx.echo, hash );
-#else
+#if defined(__AES__)
     update_final_echo ( &ctx.echo, (BitSequence *)hash,
                         (const BitSequence *)hash, 512 );
+#else
+     sph_echo512( &ctx.echo, hash, 64 );
+     sph_echo512_close( &ctx.echo, hash );
 #endif

        memcpy(output, hash, 32);
@@ -138,16 +136,12 @@ int scanhash_c11( struct work *work, uint32_t max_nonce,
        swab32_array( endiandata, pdata, 20 );

 	do
-        {
+   {
+      pdata[19] = nonce;
 		be32enc( &endiandata[19], nonce );
 		c11_hash( hash, endiandata );
 		if ( hash[7] <= Htarg && fulltest(hash, ptarget) )
-                {
-			pdata[19] = nonce;
-			*hashes_done = pdata[19] - first_nonce;
-                        work_set_target_ratio( work, hash );
- 			return 1;
-		}
+            submit_solution( work, hash, mythr );
 		nonce++;
 	} while ( nonce < max_nonce && !(*restart) );
 	pdata[19] = nonce;
--- a/algo/x11/timetravel.c
+++ b/algo/x11/timetravel.c
@@ -11,10 +11,10 @@
 #include "algo/skein/sph_skein.h"
 #include "algo/luffa/luffa_for_sse2.h"
 #include "algo/cubehash/cubehash_sse2.h"
-#ifdef NO_AES_NI
-  #include "algo/groestl/sph_groestl.h"
-#else
+#ifdef __AES__
  #include "algo/groestl/aes_ni/hash-groestl.h"
+#else
+  #include "algo/groestl/sph_groestl.h"
 #endif

 static __thread uint32_t s_ntime = UINT32_MAX;
@@ -28,10 +28,10 @@ typedef struct {
        sph_keccak512_context   keccak;
        hashState_luffa         luffa;
        cubehashParam           cube;
-#ifdef NO_AES_NI
-        sph_groestl512_context  groestl;
-#else
+#ifdef __AES__
        hashState_groestl       groestl;
+#else
+        sph_groestl512_context  groestl;
 #endif
 } tt_ctx_holder;

@@ -47,10 +47,10 @@ void init_tt8_ctx()
        sph_keccak512_init( &tt_ctx.keccak );
        init_luffa( &tt_ctx.luffa, 512 );
        cubehashInit( &tt_ctx.cube, 512, 16, 32 );
-#ifdef NO_AES_NI
-        sph_groestl512_init( &tt_ctx.groestl );
-#else
+#ifdef __AES__
        init_groestl( &tt_ctx.groestl, 64 );
+#else
+        sph_groestl512_init( &tt_ctx.groestl );
 #endif
 };

@@ -110,7 +110,10 @@ void timetravel_hash(void *output, const void *input)
        }
        break;
     case 2:
-#ifdef NO_AES_NI
+#ifdef __AES__
+           update_and_final_groestl( &ctx.groestl, (char*)hashB,
+                                    (char*)hashA, dataLen*8 );
+#else
        if ( i == 0 )
        {
           memcpy( &ctx.groestl, &tt_mid.groestl, sizeof tt_mid.groestl );
@@ -122,19 +125,6 @@ void timetravel_hash(void *output, const void *input)
           sph_groestl512( &ctx.groestl, hashA, dataLen );
           sph_groestl512_close( &ctx.groestl, hashB );
        }
-#else
-// groestl midstate is slower
-//        if ( i == 0 )
-//        {
-//           memcpy( &ctx.groestl, &tt_mid.groestl, sizeof tt_mid.groestl );
-//           update_and_final_groestl( &ctx.groestl, (char*)hashB,
-//                                    (char*)input + midlen, tail*8 );
-//        }
-//        else
-//        {
-           update_and_final_groestl( &ctx.groestl, (char*)hashB,
-                                    (char*)hashA, dataLen*8 );
-//        }
 #endif
        break;
     case 3:
@@ -253,13 +243,9 @@ int scanhash_timetravel( struct work *work, uint32_t max_nonce,
           sph_bmw512( &tt_mid.bmw, endiandata, 64 );
           break;
        case 2:
-#ifdef NO_AES_NI
+#ifndef __AES__
           memcpy( &tt_mid.groestl, &tt_ctx.groestl, sizeof(tt_mid.groestl ) );
           sph_groestl512( &tt_mid.groestl, endiandata, 64 );
-#else
-// groestl midstate is slower
-//         memcpy( &tt_mid.groestl, &tt_ctx.groestl, sizeof(tt_mid.groestl ) );
-//         update_groestl( &tt_mid.groestl, (char*)endiandata, 64*8 );
 #endif
           break;
        case 3:
@@ -293,14 +279,10 @@ int scanhash_timetravel( struct work *work, uint32_t max_nonce,

        if ( hash[7] <= Htarg && fulltest( hash, ptarget) )
        {
-              work_set_target_ratio( work, hash );
              pdata[19] = nonce;
-              *hashes_done = pdata[19] - first_nonce;
-              work_set_target_ratio( work, hash );
-              return 1;
-         }
-         nonce++;
-
+              submit_solution( work, hash, mythr );
+        }
+        nonce++;
        } while (nonce < max_nonce && !(*restart));

        pdata[19] = nonce;
--- a/algo/x11/timetravel10.c
+++ b/algo/x11/timetravel10.c
@@ -12,11 +12,10 @@
 #include "algo/cubehash/cubehash_sse2.h"
 #include "algo/shavite/sph_shavite.h"
 #include "algo/simd/nist.h"
-
-#ifdef NO_AES_NI
-  #include "algo/groestl/sph_groestl.h"
-#else
+#ifdef __AES__
  #include "algo/groestl/aes_ni/hash-groestl.h"
+#else
+  #include "algo/groestl/sph_groestl.h"
 #endif

 static __thread uint32_t s_ntime = UINT32_MAX;
@@ -32,10 +31,10 @@ typedef struct {
        cubehashParam           cube;
        sph_shavite512_context  shavite;
        hashState_sd            simd;
-#ifdef NO_AES_NI
-        sph_groestl512_context  groestl;
-#else
+#ifdef __AES__
        hashState_groestl       groestl;
+#else
+        sph_groestl512_context  groestl;
 #endif
 } tt10_ctx_holder;

@@ -53,10 +52,10 @@ void init_tt10_ctx()
        cubehashInit( &tt10_ctx.cube, 512, 16, 32 );
        sph_shavite512_init( &tt10_ctx.shavite );
        init_sd( &tt10_ctx.simd, 512 );
-#ifdef NO_AES_NI
-        sph_groestl512_init( &tt10_ctx.groestl );
-#else
+#ifdef __AES__
        init_groestl( &tt10_ctx.groestl, 64 );
+#else
+        sph_groestl512_init( &tt10_ctx.groestl );
 #endif
 };

@@ -116,7 +115,10 @@ void timetravel10_hash(void *output, const void *input)
        }
        break;
     case 2:
-#ifdef NO_AES_NI
+#ifdef __AES__
+           update_and_final_groestl( &ctx.groestl, (char*)hashB,
+                                    (char*)hashA, dataLen*8 );
+#else
        if ( i == 0 )
        {
           memcpy( &ctx.groestl, &tt10_mid.groestl, sizeof tt10_mid.groestl );
@@ -128,19 +130,6 @@ void timetravel10_hash(void *output, const void *input)
           sph_groestl512( &ctx.groestl, hashA, dataLen );
           sph_groestl512_close( &ctx.groestl, hashB );
        }
-#else
-// groestl midstate is slower
-//        if ( i == 0 )
-//        {
-//           memcpy( &ctx.groestl, &tt10_mid.groestl, sizeof tt10_mid.groestl );
-//           update_and_final_groestl( &ctx.groestl, (char*)hashB,
-//                                    (char*)input + midlen, tail*8 );
-//        }
-//        else
-//        {
-           update_and_final_groestl( &ctx.groestl, (char*)hashB,
-                                    (char*)hashA, dataLen*8 );
-//        }
 #endif
        break;
     case 3:
@@ -286,13 +275,9 @@ int scanhash_timetravel10( struct work *work, uint32_t max_nonce,
           sph_bmw512( &tt10_mid.bmw, endiandata, 64 );
           break;
        case 2:
-#ifdef NO_AES_NI
+#ifndef __AES__
           memcpy( &tt10_mid.groestl, &tt10_ctx.groestl, sizeof(tt10_mid.groestl ) );
           sph_groestl512( &tt10_mid.groestl, endiandata, 64 );
-#else
-// groestl midstate is slower
-//         memcpy( &tt10_mid.groestl, &tt10_ctx.groestl, sizeof(tt10_mid.groestl ) );
-//         update_groestl( &tt10_mid.groestl, (char*)endiandata, 64*8 );
 #endif
           break;
        case 3:
@@ -334,14 +319,10 @@ int scanhash_timetravel10( struct work *work, uint32_t max_nonce,

        if ( hash[7] <= Htarg && fulltest( hash, ptarget) )
        {
-              work_set_target_ratio( work, hash );
              pdata[19] = nonce;
-              work_set_target_ratio( work, hash );
-              *hashes_done = pdata[19] - first_nonce;
-              return 1;
-         }
-         nonce++;
-
+              submit_solution( work, hash, mythr );
+        }
+        nonce++;
        } while (nonce < max_nonce && !(*restart));

        pdata[19] = nonce;
--- a/algo/x11/tribus-4way.c
+++ b/algo/x11/tribus-4way.c
@@ -124,7 +124,7 @@ int scanhash_tribus_8way( struct work *work, uint32_t max_nonce,
     pdata[19] = n;

     for ( int i = 0; i < 8; i++ )
-     if ( (hash+(i<<3))[7] < Htarg )
+     if ( (hash+(i<<3))[7] <= Htarg )
     if ( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
     {
          pdata[19] = n+i;
@@ -209,7 +209,7 @@ int scanhash_tribus_4way( struct work *work, uint32_t max_nonce,
     pdata[19] = n;

     for ( int i = 0; i < 4; i++ )
-     if ( (hash+(i<<3))[7] < Htarg )
+     if ( (hash+(i<<3))[7] <= Htarg )
     if ( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
     {
          pdata[19] = n+i;
--- a/algo/x11/tribus.c
+++ b/algo/x11/tribus.c
@@ -7,19 +7,19 @@
 #include "algo/jh//sph_jh.h"
 #include "algo/keccak/sph_keccak.h"

-#ifdef NO_AES_NI
-  #include "algo/echo/sph_echo.h"
-#else
+#ifdef __AES__
  #include "algo/echo/aes_ni/hash_api.h"
+#else
+  #include "algo/echo/sph_echo.h"
 #endif

 typedef struct {
    sph_jh512_context     jh;
    sph_keccak512_context keccak;
-#ifdef NO_AES_NI
-    sph_echo512_context   echo;
-#else
+#ifdef __AES__
    hashState_echo        echo;
+#else
+    sph_echo512_context   echo;
 #endif
 } tribus_ctx_holder;

@@ -29,10 +29,10 @@ bool tribus_thread_init()
 {
   sph_jh512_init( &tribus_ctx.jh );
   sph_keccak512_init( &tribus_ctx.keccak );
-#ifdef NO_AES_NI
-   sph_echo512_init( &tribus_ctx.echo );
-#else
+#ifdef __AES__
   init_echo( &tribus_ctx.echo, 512 );
+#else
+   sph_echo512_init( &tribus_ctx.echo );
 #endif
  return true;
 }
@@ -49,12 +49,12 @@ void tribus_hash(void *state, const void *input)
     sph_keccak512( &ctx.keccak, (const void*) hash, 64 );
     sph_keccak512_close( &ctx.keccak, (void*) hash );

-#ifdef NO_AES_NI
-     sph_echo512( &ctx.echo, hash, 64 );
-     sph_echo512_close (&ctx.echo, hash );
-#else
+#ifdef __AES__
     update_final_echo( &ctx.echo, (BitSequence *) hash,
                        (const BitSequence *) hash, 512 );
+#else
+     sph_echo512( &ctx.echo, hash, 64 );
+     sph_echo512_close (&ctx.echo, hash );
 #endif

     memcpy(state, hash, 32);
@@ -98,9 +98,6 @@ int scanhash_tribus( struct work *work, uint32_t max_nonce,
        sph_jh512_init( &tribus_ctx.jh );
        sph_jh512( &tribus_ctx.jh, endiandata, 64 );

-#ifdef DEBUG_ALGO
-	printf("[%d] Htarg=%X\n", thr_id, Htarg);
-#endif
 	for (int m=0; m < 6; m++) {
 		if (Htarg <= htmax[m]) {
 			uint32_t mask = masks[m];
@@ -108,25 +105,9 @@ int scanhash_tribus( struct work *work, uint32_t max_nonce,
 				pdata[19] = ++n;
 				be32enc(&endiandata[19], n);
 				tribus_hash(hash32, endiandata);
-#ifndef DEBUG_ALGO
-				if ((!(hash32[7] & mask)) && fulltest(hash32, ptarget)) {
-					work_set_target_ratio(work, hash32);
-					*hashes_done = n - first_nonce + 1;
-					return 1;
-				}
-#else
-				if (!(n % 0x1000) && !thr_id) printf(".");
-				if (!(hash32[7] & mask)) {
-					printf("[%d]",thr_id);
-					if (fulltest(hash32, ptarget)) {
-						work_set_target_ratio(work, hash32);
-						*hashes_done = n - first_nonce + 1;
-						return 1;
-					}
-				}
-#endif
+				if ((!(hash32[7] & mask)) && fulltest(hash32, ptarget)) 
+                submit_solution( work, hash32, mythr );
 			} while (n < max_nonce && !work_restart[thr_id].restart);
-			// see blake.c if else to understand the loop on htmax => mask
 			break;
 		}
 	}
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Jay D Dee	88f81fda0b	v3.11.7	2020-01-26 04:33:39 -05:00
Jay D Dee	103e6ad36c	v3.11.6	2020-01-23 00:11:08 -05:00
Jay D Dee	1a7a573675	v3.11.5	2020-01-18 15:14:27 -05:00
Jay D Dee	70089d1224	v3.11.2	2020-01-08 14:44:47 -05:00