v23.10

RELEASE_NOTES

@@ -73,6 +73,19 @@ If not what makes it happen or not happen?
 Change Log
 ----------
 
+v23.10
+
+x86_64: Fixed scrypt, scryptn2 algos SSE2. 
+Fixed sha512d256d algo AVX2, SSE2, NEON.
+Fixed a bug in Skein N-way that reduced performance.
+ARM: Skein algo optimized for NEON & SHA2.
+Skein2 algo 2-way optimized for NEON & SSE2.
+
+v23.9
+
+x86_64: fixed minotaurx crash, broken in 23.7.
+ARM: #407 fix compile error due to incorrect type casting for vrev instruction argument.
+
 v23.8
 
 Cpuminer-opt is no longer dependant on OpenSSL.

algo/blake/blake256-hash.c

@@ -429,7 +429,7 @@ void blake256_transform_le( uint32_t *H, const uint32_t *buf,
 #define BLAKE256_4X32_BLOCK_BSWAP32 \
 { \
    v128_t shuf_bswap32 = v128_set64( 0x0c0d0e0f08090a0b, \
-                                          0x0405060700010203 ); \
+                                     0x0405060700010203 ); \
    M0 = _mm_shuffle_epi8( buf[ 0], shuf_bswap32 ); \
    M1 = _mm_shuffle_epi8( buf[ 1], shuf_bswap32 ); \
    M2 = _mm_shuffle_epi8( buf[ 2], shuf_bswap32 ); \
@@ -931,14 +931,14 @@ void blake256_4x32_final_rounds_le( void *final_hash, const void *midstate,
    const v128_t shuf_bswap32 =
                       v128_set64( 0x0c0d0e0f08090a0b, 0x0405060700010203 );
 
-   H[0] = _mm_shuffle_epi8( mm128_xor3( V8, V0, h[0] ), shuf_bswap32 );
-   H[1] = _mm_shuffle_epi8( mm128_xor3( V9, V1, h[1] ), shuf_bswap32 );
-   H[2] = _mm_shuffle_epi8( mm128_xor3( VA, V2, h[2] ), shuf_bswap32 );
-   H[3] = _mm_shuffle_epi8( mm128_xor3( VB, V3, h[3] ), shuf_bswap32 );
-   H[4] = _mm_shuffle_epi8( mm128_xor3( VC, V4, h[4] ), shuf_bswap32 );
-   H[5] = _mm_shuffle_epi8( mm128_xor3( VD, V5, h[5] ), shuf_bswap32 );
-   H[6] = _mm_shuffle_epi8( mm128_xor3( VE, V6, h[6] ), shuf_bswap32 );
-   H[7] = _mm_shuffle_epi8( mm128_xor3( VF, V7, h[7] ), shuf_bswap32 );
+   H[0] = _mm_shuffle_epi8( v128_xor3( V8, V0, h[0] ), shuf_bswap32 );
+   H[1] = _mm_shuffle_epi8( v128_xor3( V9, V1, h[1] ), shuf_bswap32 );
+   H[2] = _mm_shuffle_epi8( v128_xor3( VA, V2, h[2] ), shuf_bswap32 );
+   H[3] = _mm_shuffle_epi8( v128_xor3( VB, V3, h[3] ), shuf_bswap32 );
+   H[4] = _mm_shuffle_epi8( v128_xor3( VC, V4, h[4] ), shuf_bswap32 );
+   H[5] = _mm_shuffle_epi8( v128_xor3( VD, V5, h[5] ), shuf_bswap32 );
+   H[6] = _mm_shuffle_epi8( v128_xor3( VE, V6, h[6] ), shuf_bswap32 );
+   H[7] = _mm_shuffle_epi8( v128_xor3( VF, V7, h[7] ), shuf_bswap32 );
 
 #else
 

algo/blake/blake512-hash.c

@@ -475,11 +475,12 @@ void blake512_update(blake512_context *sc, const void *data, size_t len)
 void blake512_close( blake512_context *sc, void *dst )
 {
    unsigned char buf[128] __attribute__((aligned(32)));
-   size_t ptr;
+   size_t ptr, k;
    unsigned bit_len;
    uint64_t th, tl;
 
    ptr = sc->ptr;
+   memcpy( buf, sc->buf, ptr );
    bit_len = ((unsigned)ptr << 3);
    buf[ptr] = 0x80;
    tl = sc->T0 + bit_len;
@@ -519,7 +520,8 @@ void blake512_close( blake512_context *sc, void *dst )
       blake512_update( sc, buf, 128 );
    }
    
-   v128_block_bswap64_512( dst, sc->H ); 
+   for ( k = 0; k < 8; k ++ )
+      ((uint64_t*)dst)[k] = bswap_64( sc->H[k] );
 }
 
 void blake512_full( blake512_context *sc, void *dst, const void *data,

algo/blake/sph_blake2b.c

@@ -131,47 +131,7 @@
    V[7] = v128_alignr64( V6, V7, 1 ); \
 }
 
-/*
-#elif defined(__SSE2__)
-// always true
-
-#define BLAKE2B_G( Va, Vb, Vc, Vd, Sa, Sb, Sc, Sd ) \
-{ \
-   Va = _mm_add_epi64( Va, _mm_add_epi64( Vb, \
-                 _mm_set_epi64x( m[ sigmaR[ Sc ] ], m[ sigmaR[ Sa ] ] ) ) ); \
-   Vd = mm128_swap64_32( _mm_xor_si128( Vd, Va ) ); \
-   Vc = _mm_add_epi64( Vc, Vd ); \
-   Vb = mm128_shuflr64_24( _mm_xor_si128( Vb, Vc ) ); \
-\
-   Va = _mm_add_epi64( Va, _mm_add_epi64( Vb, \
-                 _mm_set_epi64x( m[ sigmaR[ Sd ] ], m[ sigmaR[ Sb ] ] ) ) ); \
-   Vd = mm128_shuflr64_16( _mm_xor_si128( Vd, Va ) ); \
-   Vc = _mm_add_epi64( Vc, Vd ); \
-   Vb = mm128_ror_64( _mm_xor_si128( Vb, Vc ), 63 ); \
-}
-
-#define BLAKE2B_ROUND( R ) \
-{ \
-   v128_t *V = (v128_t*)v; \
-   v128_t V2, V3, V6, V7; \
-   const uint8_t *sigmaR = sigma[R]; \
-   BLAKE2B_G( V[0], V[2], V[4], V[6], 0, 1, 2, 3 ); \
-   BLAKE2B_G( V[1], V[3], V[5], V[7], 4, 5, 6, 7 ); \
-   V2 = mm128_alignr_64( V[3], V[2], 1 ); \
-   V3 = mm128_alignr_64( V[2], V[3], 1 ); \
-   V6 = mm128_alignr_64( V[6], V[7], 1 ); \
-   V7 = mm128_alignr_64( V[7], V[6], 1 ); \
-   BLAKE2B_G( V[0], V2, V[5], V6,  8,  9, 10, 11 ); \
-   BLAKE2B_G( V[1], V3, V[4], V7, 12, 13, 14, 15 ); \
-   V[2] = mm128_alignr_64( V2, V3, 1 ); \
-   V[3] = mm128_alignr_64( V3, V2, 1 ); \
-   V[6] = mm128_alignr_64( V7, V6, 1 ); \
-   V[7] = mm128_alignr_64( V6, V7, 1 ); \
-}
-*/
-
 #else
-// never used, SSE2 is always available
 
 #ifndef ROTR64
 #define ROTR64(x, y)  (((x) >> (y)) ^ ((x) << (64 - (y))))

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

@@ -62,78 +62,78 @@ static const uint32_t IV256[] = {
 */
 
 #define ss0(x) \
-   _mm_xor_si128( _mm_xor_si128( _mm_srli_epi32( (x), 1), \
-                                 _mm_slli_epi32( (x), 3) ), \
-                  _mm_xor_si128( mm128_rol_32( (x),  4), \
-                                 mm128_rol_32( (x), 19) ) )
+   v128_xor( v128_xor( v128_sr32( (x), 1), \
+                                 v128_sl32( (x), 3) ), \
+                  v128_xor( v128_rol32( (x),  4), \
+                                 v128_rol32( (x), 19) ) )
 
 #define ss1(x) \
-   _mm_xor_si128( _mm_xor_si128( _mm_srli_epi32( (x), 1), \
-                                 _mm_slli_epi32( (x), 2) ), \
-                  _mm_xor_si128( mm128_rol_32( (x),  8), \
-                                 mm128_rol_32( (x), 23) ) )
+   v128_xor( v128_xor( v128_sr32( (x), 1), \
+                                 v128_sl32( (x), 2) ), \
+                  v128_xor( v128_rol32( (x),  8), \
+                                 v128_rol32( (x), 23) ) )
 
 #define ss2(x) \
-   _mm_xor_si128( _mm_xor_si128( _mm_srli_epi32( (x), 2), \
-                                 _mm_slli_epi32( (x), 1) ), \
-                  _mm_xor_si128( mm128_rol_32( (x), 12), \
-                                 mm128_rol_32( (x), 25) ) )
+   v128_xor( v128_xor( v128_sr32( (x), 2), \
+                                 v128_sl32( (x), 1) ), \
+                  v128_xor( v128_rol32( (x), 12), \
+                                 v128_rol32( (x), 25) ) )
 
 #define ss3(x) \
-   _mm_xor_si128( _mm_xor_si128( _mm_srli_epi32( (x), 2), \
-                                 _mm_slli_epi32( (x), 2) ), \
-                  _mm_xor_si128( mm128_rol_32( (x), 15), \
-                                 mm128_rol_32( (x), 29) ) )
+   v128_xor( v128_xor( v128_sr32( (x), 2), \
+                                 v128_sl32( (x), 2) ), \
+                  v128_xor( v128_rol32( (x), 15), \
+                                 v128_rol32( (x), 29) ) )
 
 #define ss4(x) \
-  _mm_xor_si128( (x), _mm_srli_epi32( (x), 1 ) )
+  v128_xor( (x), v128_sr32( (x), 1 ) )
 
 #define ss5(x) \
-  _mm_xor_si128( (x), _mm_srli_epi32( (x), 2 ) )
+  v128_xor( (x), v128_sr32( (x), 2 ) )
 
-#define rs1(x)    mm128_rol_32( x,  3 ) 
-#define rs2(x)    mm128_rol_32( x,  7 ) 
-#define rs3(x)    mm128_rol_32( x, 13 ) 
-#define rs4(x)    mm128_rol_32( x, 16 ) 
-#define rs5(x)    mm128_rol_32( x, 19 ) 
-#define rs6(x)    mm128_rol_32( x, 23 ) 
-#define rs7(x)    mm128_rol_32( x, 27 ) 
+#define rs1(x)    v128_rol32( x,  3 ) 
+#define rs2(x)    v128_rol32( x,  7 ) 
+#define rs3(x)    v128_rol32( x, 13 ) 
+#define rs4(x)    v128_rol32( x, 16 ) 
+#define rs5(x)    v128_rol32( x, 19 ) 
+#define rs6(x)    v128_rol32( x, 23 ) 
+#define rs7(x)    v128_rol32( x, 27 ) 
 
 #define rol_off_32( M, j, off ) \
-   mm128_rol_32( M[ ( (j) + (off) ) & 0xF ] , \
+   v128_rol32( M[ ( (j) + (off) ) & 0xF ] , \
                 ( ( (j) + (off) ) & 0xF ) + 1 )
 
 #define add_elt_s( M, H, j ) \
-   _mm_xor_si128( \
-       _mm_add_epi32( \
-             _mm_sub_epi32( _mm_add_epi32( rol_off_32( M, j, 0 ), \
+   v128_xor( \
+       v128_add32( \
+             v128_sub32( v128_add32( rol_off_32( M, j, 0 ), \
                                            rol_off_32( M, j, 3 ) ), \
                             rol_off_32( M, j, 10 ) ), \
-       _mm_set1_epi32( ( (j)+16 ) * 0x05555555UL ) ), \
+       v128_32( ( (j)+16 ) * 0x05555555UL ) ), \
    H[ ( (j)+7 ) & 0xF ] )
 
 
 #define expand1s( qt, M, H, i ) \
-   _mm_add_epi32(  mm128_add4_32( \
-            mm128_add4_32( ss1( qt[ (i)-16 ] ), ss2( qt[ (i)-15 ] ), \
+   v128_add32(  v128_add4_32( \
+            v128_add4_32( ss1( qt[ (i)-16 ] ), ss2( qt[ (i)-15 ] ), \
                            ss3( qt[ (i)-14 ] ), ss0( qt[ (i)-13 ] ) ), \
-            mm128_add4_32( ss1( qt[ (i)-12 ] ), ss2( qt[ (i)-11 ] ), \
+            v128_add4_32( ss1( qt[ (i)-12 ] ), ss2( qt[ (i)-11 ] ), \
                            ss3( qt[ (i)-10 ] ), ss0( qt[ (i)- 9 ] ) ), \
-            mm128_add4_32( ss1( qt[ (i)- 8 ] ), ss2( qt[ (i)- 7 ] ), \
+            v128_add4_32( ss1( qt[ (i)- 8 ] ), ss2( qt[ (i)- 7 ] ), \
                            ss3( qt[ (i)- 6 ] ), ss0( qt[ (i)- 5 ] ) ),  \
-            mm128_add4_32( ss1( qt[ (i)- 4 ] ), ss2( qt[ (i)- 3 ] ), \
+            v128_add4_32( ss1( qt[ (i)- 4 ] ), ss2( qt[ (i)- 3 ] ), \
                            ss3( qt[ (i)- 2 ] ), ss0( qt[ (i)- 1 ] ) ) ), \
       add_elt_s( M, H, (i)-16 ) )
 
 #define expand2s( qt, M, H, i) \
-   _mm_add_epi32( mm128_add4_32( \
-            mm128_add4_32( qt[ (i)-16 ], rs1( qt[ (i)-15 ] ), \
+   v128_add32( v128_add4_32( \
+            v128_add4_32( qt[ (i)-16 ], rs1( qt[ (i)-15 ] ), \
                            qt[ (i)-14 ], rs2( qt[ (i)-13 ] ) ), \
-            mm128_add4_32( qt[ (i)-12 ], rs3( qt[ (i)-11 ] ), \
+            v128_add4_32( qt[ (i)-12 ], rs3( qt[ (i)-11 ] ), \
                            qt[ (i)-10 ], rs4( qt[ (i)- 9 ] ) ), \
-            mm128_add4_32( qt[ (i)- 8 ], rs5( qt[ (i)- 7 ] ), \
+            v128_add4_32( qt[ (i)- 8 ], rs5( qt[ (i)- 7 ] ), \
                            qt[ (i)- 6 ], rs6( qt[ (i)- 5 ] ) ), \
-            mm128_add4_32( qt[ (i)- 4 ], rs7( qt[ (i)- 3 ] ), \
+            v128_add4_32( qt[ (i)- 4 ], rs7( qt[ (i)- 3 ] ), \
                            ss4( qt[ (i)- 2 ] ), ss5( qt[ (i)- 1 ] ) ) ), \
       add_elt_s( M, H, (i)-16 ) )
 
@@ -141,169 +141,169 @@ static const uint32_t IV256[] = {
 // resulting in some sign changes compared to the reference code.
 
 #define Ws0 \
-   _mm_add_epi32( \
-      _mm_add_epi32( \
-         _mm_sub_epi32( _mm_xor_si128( M[ 5], H[ 5] ), \
-                        _mm_xor_si128( M[ 7], H[ 7] ) ), \
-         _mm_xor_si128( M[10], H[10] ) ), \
-      _mm_add_epi32( _mm_xor_si128( M[13], H[13] ), \
-                     _mm_xor_si128( M[14], H[14] ) ) )
+   v128_add32( \
+      v128_add32( \
+         v128_sub32( v128_xor( M[ 5], H[ 5] ), \
+                        v128_xor( M[ 7], H[ 7] ) ), \
+         v128_xor( M[10], H[10] ) ), \
+      v128_add32( v128_xor( M[13], H[13] ), \
+                     v128_xor( M[14], H[14] ) ) )
 
 #define Ws1 \
-   _mm_add_epi32( \
-       _mm_add_epi32( \
-          _mm_sub_epi32( _mm_xor_si128( M[ 6], H[ 6] ), \
-                         _mm_xor_si128( M[ 8], H[ 8] ) ), \
-          _mm_xor_si128( M[11], H[11] ) ), \
-       _mm_sub_epi32( _mm_xor_si128( M[14], H[14] ), \
-                      _mm_xor_si128( M[15], H[15] ) ) )
+   v128_add32( \
+       v128_add32( \
+          v128_sub32( v128_xor( M[ 6], H[ 6] ), \
+                         v128_xor( M[ 8], H[ 8] ) ), \
+          v128_xor( M[11], H[11] ) ), \
+       v128_sub32( v128_xor( M[14], H[14] ), \
+                      v128_xor( M[15], H[15] ) ) )
 
 #define Ws2 \
-   _mm_sub_epi32( \
-      _mm_add_epi32( \
-         _mm_add_epi32( _mm_xor_si128( M[ 0], H[ 0] ), \
-                        _mm_xor_si128( M[ 7], H[ 7] ) ), \
-         _mm_xor_si128( M[ 9], H[ 9] ) ), \
-      _mm_sub_epi32( _mm_xor_si128( M[12], H[12] ), \
-                     _mm_xor_si128( M[15], H[15] ) ) )
+   v128_sub32( \
+      v128_add32( \
+         v128_add32( v128_xor( M[ 0], H[ 0] ), \
+                        v128_xor( M[ 7], H[ 7] ) ), \
+         v128_xor( M[ 9], H[ 9] ) ), \
+      v128_sub32( v128_xor( M[12], H[12] ), \
+                     v128_xor( M[15], H[15] ) ) )
 
 #define Ws3 \
-   _mm_sub_epi32( \
-      _mm_add_epi32( \
-         _mm_sub_epi32( _mm_xor_si128( M[ 0], H[ 0] ), \
-                        _mm_xor_si128( M[ 1], H[ 1] ) ), \
-         _mm_xor_si128( M[ 8], H[ 8] ) ), \
-      _mm_sub_epi32( _mm_xor_si128( M[10], H[10] ), \
-                     _mm_xor_si128( M[13], H[13] ) ) )
+   v128_sub32( \
+      v128_add32( \
+         v128_sub32( v128_xor( M[ 0], H[ 0] ), \
+                        v128_xor( M[ 1], H[ 1] ) ), \
+         v128_xor( M[ 8], H[ 8] ) ), \
+      v128_sub32( v128_xor( M[10], H[10] ), \
+                     v128_xor( M[13], H[13] ) ) )
 
 #define Ws4 \
-   _mm_sub_epi32( \
-      _mm_add_epi32( \
-         _mm_add_epi32( _mm_xor_si128( M[ 1], H[ 1] ), \
-                        _mm_xor_si128( M[ 2], H[ 2] ) ), \
-         _mm_xor_si128( M[ 9], H[ 9] ) ), \
-      _mm_add_epi32( _mm_xor_si128( M[11], H[11] ), \
-                     _mm_xor_si128( M[14], H[14] ) ) )
+   v128_sub32( \
+      v128_add32( \
+         v128_add32( v128_xor( M[ 1], H[ 1] ), \
+                        v128_xor( M[ 2], H[ 2] ) ), \
+         v128_xor( M[ 9], H[ 9] ) ), \
+      v128_add32( v128_xor( M[11], H[11] ), \
+                     v128_xor( M[14], H[14] ) ) )
 
 #define Ws5 \
-   _mm_sub_epi32( \
-      _mm_add_epi32( \
-         _mm_sub_epi32( _mm_xor_si128( M[ 3], H[ 3] ), \
-                        _mm_xor_si128( M[ 2], H[ 2] ) ), \
-         _mm_xor_si128( M[10], H[10] ) ), \
-      _mm_sub_epi32( _mm_xor_si128( M[12], H[12] ), \
-                     _mm_xor_si128( M[15], H[15] ) ) )
+   v128_sub32( \
+      v128_add32( \
+         v128_sub32( v128_xor( M[ 3], H[ 3] ), \
+                        v128_xor( M[ 2], H[ 2] ) ), \
+         v128_xor( M[10], H[10] ) ), \
+      v128_sub32( v128_xor( M[12], H[12] ), \
+                     v128_xor( M[15], H[15] ) ) )
 
 #define Ws6 \
-   _mm_sub_epi32( \
-      _mm_sub_epi32( \
-         _mm_sub_epi32( _mm_xor_si128( M[ 4], H[ 4] ), \
-                        _mm_xor_si128( M[ 0], H[ 0] ) ), \
-         _mm_xor_si128( M[ 3], H[ 3] ) ), \
-      _mm_sub_epi32( _mm_xor_si128( M[11], H[11] ), \
-                     _mm_xor_si128( M[13], H[13] ) ) )
+   v128_sub32( \
+      v128_sub32( \
+         v128_sub32( v128_xor( M[ 4], H[ 4] ), \
+                        v128_xor( M[ 0], H[ 0] ) ), \
+         v128_xor( M[ 3], H[ 3] ) ), \
+      v128_sub32( v128_xor( M[11], H[11] ), \
+                     v128_xor( M[13], H[13] ) ) )
 
 #define Ws7 \
-   _mm_sub_epi32( \
-      _mm_sub_epi32( \
-         _mm_sub_epi32( _mm_xor_si128( M[ 1], H[ 1] ), \
-                        _mm_xor_si128( M[ 4], H[ 4] ) ), \
-         _mm_xor_si128( M[ 5], H[ 5] ) ), \
-      _mm_add_epi32( _mm_xor_si128( M[12], H[12] ), \
-                     _mm_xor_si128( M[14], H[14] ) ) )
+   v128_sub32( \
+      v128_sub32( \
+         v128_sub32( v128_xor( M[ 1], H[ 1] ), \
+                        v128_xor( M[ 4], H[ 4] ) ), \
+         v128_xor( M[ 5], H[ 5] ) ), \
+      v128_add32( v128_xor( M[12], H[12] ), \
+                     v128_xor( M[14], H[14] ) ) )
 
 #define Ws8 \
-   _mm_add_epi32( \
-      _mm_sub_epi32( \
-         _mm_sub_epi32( _mm_xor_si128( M[ 2], H[ 2] ), \
-                        _mm_xor_si128( M[ 5], H[ 5] ) ), \
-         _mm_xor_si128( M[ 6], H[ 6] ) ), \
-      _mm_sub_epi32( _mm_xor_si128( M[13], H[13] ), \
-                     _mm_xor_si128( M[15], H[15] ) ) )
+   v128_add32( \
+      v128_sub32( \
+         v128_sub32( v128_xor( M[ 2], H[ 2] ), \
+                        v128_xor( M[ 5], H[ 5] ) ), \
+         v128_xor( M[ 6], H[ 6] ) ), \
+      v128_sub32( v128_xor( M[13], H[13] ), \
+                     v128_xor( M[15], H[15] ) ) )
 #define Ws9 \
-   _mm_sub_epi32( \
-      _mm_add_epi32( \
-         _mm_sub_epi32( _mm_xor_si128( M[ 0], H[ 0] ), \
-                        _mm_xor_si128( M[ 3], H[ 3] ) ), \
-         _mm_xor_si128( M[ 6], H[ 6] ) ), \
-      _mm_sub_epi32( _mm_xor_si128( M[ 7], H[ 7] ), \
-                     _mm_xor_si128( M[14], H[14] ) ) )
+   v128_sub32( \
+      v128_add32( \
+         v128_sub32( v128_xor( M[ 0], H[ 0] ), \
+                        v128_xor( M[ 3], H[ 3] ) ), \
+         v128_xor( M[ 6], H[ 6] ) ), \
+      v128_sub32( v128_xor( M[ 7], H[ 7] ), \
+                     v128_xor( M[14], H[14] ) ) )
 
 #define Ws10 \
-   _mm_sub_epi32( \
-      _mm_sub_epi32( \
-         _mm_sub_epi32( _mm_xor_si128( M[ 8], H[ 8] ), \
-                        _mm_xor_si128( M[ 1], H[ 1] ) ), \
-         _mm_xor_si128( M[ 4], H[ 4] ) ), \
-      _mm_sub_epi32( _mm_xor_si128( M[ 7], H[ 7] ), \
-                     _mm_xor_si128( M[15], H[15] ) ) )
+   v128_sub32( \
+      v128_sub32( \
+         v128_sub32( v128_xor( M[ 8], H[ 8] ), \
+                        v128_xor( M[ 1], H[ 1] ) ), \
+         v128_xor( M[ 4], H[ 4] ) ), \
+      v128_sub32( v128_xor( M[ 7], H[ 7] ), \
+                     v128_xor( M[15], H[15] ) ) )
 
 #define Ws11 \
-   _mm_sub_epi32( \
-      _mm_sub_epi32( \
-         _mm_sub_epi32( _mm_xor_si128( M[ 8], H[ 8] ), \
-                        _mm_xor_si128( M[ 0], H[ 0] ) ), \
-         _mm_xor_si128( M[ 2], H[ 2] ) ), \
-      _mm_sub_epi32( _mm_xor_si128( M[ 5], H[ 5] ), \
-                     _mm_xor_si128( M[ 9], H[ 9] ) ) )
+   v128_sub32( \
+      v128_sub32( \
+         v128_sub32( v128_xor( M[ 8], H[ 8] ), \
+                        v128_xor( M[ 0], H[ 0] ) ), \
+         v128_xor( M[ 2], H[ 2] ) ), \
+      v128_sub32( v128_xor( M[ 5], H[ 5] ), \
+                     v128_xor( M[ 9], H[ 9] ) ) )
 
 #define Ws12 \
-   _mm_sub_epi32( \
-      _mm_sub_epi32( \
-         _mm_add_epi32( _mm_xor_si128( M[ 1], H[ 1] ), \
-                        _mm_xor_si128( M[ 3], H[ 3] ) ), \
-         _mm_xor_si128( M[ 6], H[ 6] ) ), \
-      _mm_sub_epi32( _mm_xor_si128( M[ 9], H[ 9] ), \
-                     _mm_xor_si128( M[10], H[10] ) ) )
+   v128_sub32( \
+      v128_sub32( \
+         v128_add32( v128_xor( M[ 1], H[ 1] ), \
+                        v128_xor( M[ 3], H[ 3] ) ), \
+         v128_xor( M[ 6], H[ 6] ) ), \
+      v128_sub32( v128_xor( M[ 9], H[ 9] ), \
+                     v128_xor( M[10], H[10] ) ) )
 
 #define Ws13 \
-   _mm_add_epi32( \
-      _mm_add_epi32( \
-         _mm_add_epi32( _mm_xor_si128( M[ 2], H[ 2] ), \
-                        _mm_xor_si128( M[ 4], H[ 4] ) ), \
-         _mm_xor_si128( M[ 7], H[ 7] ) ), \
-      _mm_add_epi32( _mm_xor_si128( M[10], H[10] ), \
-                     _mm_xor_si128( M[11], H[11] ) ) )
+   v128_add32( \
+      v128_add32( \
+         v128_add32( v128_xor( M[ 2], H[ 2] ), \
+                        v128_xor( M[ 4], H[ 4] ) ), \
+         v128_xor( M[ 7], H[ 7] ) ), \
+      v128_add32( v128_xor( M[10], H[10] ), \
+                     v128_xor( M[11], H[11] ) ) )
 
 #define Ws14 \
-   _mm_sub_epi32( \
-      _mm_add_epi32( \
-         _mm_sub_epi32( _mm_xor_si128( M[ 3], H[ 3] ), \
-                        _mm_xor_si128( M[ 5], H[ 5] ) ), \
-         _mm_xor_si128( M[ 8], H[ 8] ) ), \
-      _mm_add_epi32( _mm_xor_si128( M[11], H[11] ), \
-                     _mm_xor_si128( M[12], H[12] ) ) )
+   v128_sub32( \
+      v128_add32( \
+         v128_sub32( v128_xor( M[ 3], H[ 3] ), \
+                        v128_xor( M[ 5], H[ 5] ) ), \
+         v128_xor( M[ 8], H[ 8] ) ), \
+      v128_add32( v128_xor( M[11], H[11] ), \
+                     v128_xor( M[12], H[12] ) ) )
 
 #define Ws15 \
-   _mm_sub_epi32( \
-      _mm_sub_epi32( \
-         _mm_sub_epi32( _mm_xor_si128( M[12], H[12] ), \
-                        _mm_xor_si128( M[ 4], H[4] ) ), \
-         _mm_xor_si128( M[ 6], H[ 6] ) ), \
-      _mm_sub_epi32( _mm_xor_si128( M[ 9], H[ 9] ), \
-                     _mm_xor_si128( M[13], H[13] ) ) )
+   v128_sub32( \
+      v128_sub32( \
+         v128_sub32( v128_xor( M[12], H[12] ), \
+                        v128_xor( M[ 4], H[4] ) ), \
+         v128_xor( M[ 6], H[ 6] ) ), \
+      v128_sub32( v128_xor( M[ 9], H[ 9] ), \
+                     v128_xor( M[13], H[13] ) ) )
 
 
-void compress_small( const __m128i *M, const __m128i H[16], __m128i dH[16] )
+void compress_small( const v128u64_t *M, const v128u64_t H[16], v128u64_t dH[16] )
 {
-   __m128i qt[32], xl, xh; \
+   v128u64_t qt[32], xl, xh; \
 
-   qt[ 0] = _mm_add_epi32( ss0( Ws0 ), H[ 1] );
-   qt[ 1] = _mm_add_epi32( ss1( Ws1 ), H[ 2] );
-   qt[ 2] = _mm_add_epi32( ss2( Ws2 ), H[ 3] );
-   qt[ 3] = _mm_add_epi32( ss3( Ws3 ), H[ 4] );
-   qt[ 4] = _mm_add_epi32( ss4( Ws4 ), H[ 5] );
-   qt[ 5] = _mm_add_epi32( ss0( Ws5 ), H[ 6] );
-   qt[ 6] = _mm_add_epi32( ss1( Ws6 ), H[ 7] );
-   qt[ 7] = _mm_add_epi32( ss2( Ws7 ), H[ 8] );
-   qt[ 8] = _mm_add_epi32( ss3( Ws8 ), H[ 9] );
-   qt[ 9] = _mm_add_epi32( ss4( Ws9 ), H[10] );
-   qt[10] = _mm_add_epi32( ss0( Ws10), H[11] );
-   qt[11] = _mm_add_epi32( ss1( Ws11), H[12] );
-   qt[12] = _mm_add_epi32( ss2( Ws12), H[13] );
-   qt[13] = _mm_add_epi32( ss3( Ws13), H[14] );
-   qt[14] = _mm_add_epi32( ss4( Ws14), H[15] );
-   qt[15] = _mm_add_epi32( ss0( Ws15), H[ 0] );
+   qt[ 0] = v128_add32( ss0( Ws0 ), H[ 1] );
+   qt[ 1] = v128_add32( ss1( Ws1 ), H[ 2] );
+   qt[ 2] = v128_add32( ss2( Ws2 ), H[ 3] );
+   qt[ 3] = v128_add32( ss3( Ws3 ), H[ 4] );
+   qt[ 4] = v128_add32( ss4( Ws4 ), H[ 5] );
+   qt[ 5] = v128_add32( ss0( Ws5 ), H[ 6] );
+   qt[ 6] = v128_add32( ss1( Ws6 ), H[ 7] );
+   qt[ 7] = v128_add32( ss2( Ws7 ), H[ 8] );
+   qt[ 8] = v128_add32( ss3( Ws8 ), H[ 9] );
+   qt[ 9] = v128_add32( ss4( Ws9 ), H[10] );
+   qt[10] = v128_add32( ss0( Ws10), H[11] );
+   qt[11] = v128_add32( ss1( Ws11), H[12] );
+   qt[12] = v128_add32( ss2( Ws12), H[13] );
+   qt[13] = v128_add32( ss3( Ws13), H[14] );
+   qt[14] = v128_add32( ss4( Ws14), H[15] );
+   qt[15] = v128_add32( ss0( Ws15), H[ 0] );
    qt[16] = expand1s( qt, M, H, 16 );
    qt[17] = expand1s( qt, M, H, 17 );
    qt[18] = expand2s( qt, M, H, 18 );
@@ -321,92 +321,92 @@ void compress_small( const __m128i *M, const __m128i H[16], __m128i dH[16] )
    qt[30] = expand2s( qt, M, H, 30 );
    qt[31] = expand2s( qt, M, H, 31 );
 
-   xl = _mm_xor_si128( mm128_xor4( qt[16], qt[17], qt[18], qt[19] ),
-                       mm128_xor4( qt[20], qt[21], qt[22], qt[23] ) );
-   xh = _mm_xor_si128( xl, _mm_xor_si128(
-                             mm128_xor4( qt[24], qt[25], qt[26], qt[27] ),
-                             mm128_xor4( qt[28], qt[29], qt[30], qt[31] ) ) );
+   xl = v128_xor( v128_xor4( qt[16], qt[17], qt[18], qt[19] ),
+                       v128_xor4( qt[20], qt[21], qt[22], qt[23] ) );
+   xh = v128_xor( xl, v128_xor(
+                             v128_xor4( qt[24], qt[25], qt[26], qt[27] ),
+                             v128_xor4( qt[28], qt[29], qt[30], qt[31] ) ) );
 
-   dH[ 0] = _mm_add_epi32(
-                 _mm_xor_si128( M[0],
-                      _mm_xor_si128( _mm_slli_epi32( xh, 5 ),
-                                     _mm_srli_epi32( qt[16], 5 ) ) ),
-                 _mm_xor_si128( _mm_xor_si128( xl, qt[24] ), qt[ 0] ));
-   dH[ 1] = _mm_add_epi32(
-                 _mm_xor_si128( M[1],
-                      _mm_xor_si128( _mm_srli_epi32( xh, 7 ),
-                                     _mm_slli_epi32( qt[17], 8 ) ) ),
-                 _mm_xor_si128( _mm_xor_si128( xl, qt[25] ), qt[ 1] ));
-   dH[ 2] = _mm_add_epi32(
-                 _mm_xor_si128( M[2],
-                      _mm_xor_si128( _mm_srli_epi32( xh, 5 ),
-                                     _mm_slli_epi32( qt[18], 5 ) ) ),
-                 _mm_xor_si128( _mm_xor_si128( xl, qt[26] ), qt[ 2] ));
-   dH[ 3] = _mm_add_epi32(
-                 _mm_xor_si128( M[3],
-                      _mm_xor_si128( _mm_srli_epi32( xh, 1 ),
-                                     _mm_slli_epi32( qt[19], 5 ) ) ),
-                 _mm_xor_si128( _mm_xor_si128( xl, qt[27] ), qt[ 3] ));
-   dH[ 4] = _mm_add_epi32(
-                 _mm_xor_si128( M[4],
-                      _mm_xor_si128( _mm_srli_epi32( xh, 3 ),
-                                     _mm_slli_epi32( qt[20], 0 ) ) ),
-                 _mm_xor_si128( _mm_xor_si128( xl, qt[28] ), qt[ 4] ));
-   dH[ 5] = _mm_add_epi32(
-                 _mm_xor_si128( M[5],
-                      _mm_xor_si128( _mm_slli_epi32( xh, 6 ),
-                                     _mm_srli_epi32( qt[21], 6 ) ) ),
-                 _mm_xor_si128( _mm_xor_si128( xl, qt[29] ), qt[ 5] ));
-   dH[ 6] = _mm_add_epi32(
-                 _mm_xor_si128( M[6],
-                      _mm_xor_si128( _mm_srli_epi32( xh, 4 ),
-                                     _mm_slli_epi32( qt[22], 6 ) ) ),
-                 _mm_xor_si128( _mm_xor_si128( xl, qt[30] ), qt[ 6] ));
-   dH[ 7] = _mm_add_epi32(
-                 _mm_xor_si128( M[7],
-                      _mm_xor_si128( _mm_srli_epi32( xh, 11 ),
-                                     _mm_slli_epi32( qt[23], 2 ) ) ),
-                 _mm_xor_si128( _mm_xor_si128( xl, qt[31] ), qt[ 7] ));
-   dH[ 8] = _mm_add_epi32( _mm_add_epi32(
-                 mm128_rol_32( dH[4], 9 ),
-                 _mm_xor_si128( _mm_xor_si128( xh, qt[24] ), M[ 8] )),
-                 _mm_xor_si128( _mm_slli_epi32( xl, 8 ),
-                                _mm_xor_si128( qt[23], qt[ 8] ) ) );
-   dH[ 9] = _mm_add_epi32( _mm_add_epi32(
-                 mm128_rol_32( dH[5], 10 ),
-                 _mm_xor_si128( _mm_xor_si128( xh, qt[25] ), M[ 9] )),
-                 _mm_xor_si128( _mm_srli_epi32( xl, 6 ),
-                                _mm_xor_si128( qt[16], qt[ 9] ) ) );
-   dH[10] = _mm_add_epi32( _mm_add_epi32(
-                 mm128_rol_32( dH[6], 11 ),
-                 _mm_xor_si128( _mm_xor_si128( xh, qt[26] ), M[10] )),
-                 _mm_xor_si128( _mm_slli_epi32( xl, 6 ),
-                                _mm_xor_si128( qt[17], qt[10] ) ) );
-   dH[11] = _mm_add_epi32( _mm_add_epi32(
-                 mm128_rol_32( dH[7], 12 ),
-                 _mm_xor_si128( _mm_xor_si128( xh, qt[27] ), M[11] )),
-                 _mm_xor_si128( _mm_slli_epi32( xl, 4 ),
-                                _mm_xor_si128( qt[18], qt[11] ) ) );
-   dH[12] = _mm_add_epi32( _mm_add_epi32(
-                 mm128_rol_32( dH[0], 13 ),
-                 _mm_xor_si128( _mm_xor_si128( xh, qt[28] ), M[12] )),
-                 _mm_xor_si128( _mm_srli_epi32( xl, 3 ),
-                                _mm_xor_si128( qt[19], qt[12] ) ) );
-   dH[13] = _mm_add_epi32( _mm_add_epi32(
-                 mm128_rol_32( dH[1], 14 ),
-                 _mm_xor_si128( _mm_xor_si128( xh, qt[29] ), M[13] )),
-                 _mm_xor_si128( _mm_srli_epi32( xl, 4 ),
-                                _mm_xor_si128( qt[20], qt[13] ) ) );
-   dH[14] = _mm_add_epi32( _mm_add_epi32(
-                 mm128_rol_32( dH[2], 15 ),
-                 _mm_xor_si128( _mm_xor_si128( xh, qt[30] ), M[14] )),
-                 _mm_xor_si128( _mm_srli_epi32( xl, 7 ),
-                                _mm_xor_si128( qt[21], qt[14] ) ) );
-   dH[15] = _mm_add_epi32( _mm_add_epi32(
-                 mm128_rol_32( dH[3], 16 ),
-                 _mm_xor_si128( _mm_xor_si128( xh, qt[31] ), M[15] )),
-                 _mm_xor_si128( _mm_srli_epi32( xl, 2 ),
-                                _mm_xor_si128( qt[22], qt[15] ) ) );
+   dH[ 0] = v128_add32(
+                 v128_xor( M[0],
+                      v128_xor( v128_sl32( xh, 5 ),
+                                     v128_sr32( qt[16], 5 ) ) ),
+                 v128_xor( v128_xor( xl, qt[24] ), qt[ 0] ));
+   dH[ 1] = v128_add32(
+                 v128_xor( M[1],
+                      v128_xor( v128_sr32( xh, 7 ),
+                                     v128_sl32( qt[17], 8 ) ) ),
+                 v128_xor( v128_xor( xl, qt[25] ), qt[ 1] ));
+   dH[ 2] = v128_add32(
+                 v128_xor( M[2],
+                      v128_xor( v128_sr32( xh, 5 ),
+                                     v128_sl32( qt[18], 5 ) ) ),
+                 v128_xor( v128_xor( xl, qt[26] ), qt[ 2] ));
+   dH[ 3] = v128_add32(
+                 v128_xor( M[3],
+                      v128_xor( v128_sr32( xh, 1 ),
+                                     v128_sl32( qt[19], 5 ) ) ),
+                 v128_xor( v128_xor( xl, qt[27] ), qt[ 3] ));
+   dH[ 4] = v128_add32(
+                 v128_xor( M[4],
+                      v128_xor( v128_sr32( xh, 3 ),
+                                     v128_sl32( qt[20], 0 ) ) ),
+                 v128_xor( v128_xor( xl, qt[28] ), qt[ 4] ));
+   dH[ 5] = v128_add32(
+                 v128_xor( M[5],
+                      v128_xor( v128_sl32( xh, 6 ),
+                                     v128_sr32( qt[21], 6 ) ) ),
+                 v128_xor( v128_xor( xl, qt[29] ), qt[ 5] ));
+   dH[ 6] = v128_add32(
+                 v128_xor( M[6],
+                      v128_xor( v128_sr32( xh, 4 ),
+                                     v128_sl32( qt[22], 6 ) ) ),
+                 v128_xor( v128_xor( xl, qt[30] ), qt[ 6] ));
+   dH[ 7] = v128_add32(
+                 v128_xor( M[7],
+                      v128_xor( v128_sr32( xh, 11 ),
+                                     v128_sl32( qt[23], 2 ) ) ),
+                 v128_xor( v128_xor( xl, qt[31] ), qt[ 7] ));
+   dH[ 8] = v128_add32( v128_add32(
+                 v128_rol32( dH[4], 9 ),
+                 v128_xor( v128_xor( xh, qt[24] ), M[ 8] )),
+                 v128_xor( v128_sl32( xl, 8 ),
+                                v128_xor( qt[23], qt[ 8] ) ) );
+   dH[ 9] = v128_add32( v128_add32(
+                 v128_rol32( dH[5], 10 ),
+                 v128_xor( v128_xor( xh, qt[25] ), M[ 9] )),
+                 v128_xor( v128_sr32( xl, 6 ),
+                                v128_xor( qt[16], qt[ 9] ) ) );
+   dH[10] = v128_add32( v128_add32(
+                 v128_rol32( dH[6], 11 ),
+                 v128_xor( v128_xor( xh, qt[26] ), M[10] )),
+                 v128_xor( v128_sl32( xl, 6 ),
+                                v128_xor( qt[17], qt[10] ) ) );
+   dH[11] = v128_add32( v128_add32(
+                 v128_rol32( dH[7], 12 ),
+                 v128_xor( v128_xor( xh, qt[27] ), M[11] )),
+                 v128_xor( v128_sl32( xl, 4 ),
+                                v128_xor( qt[18], qt[11] ) ) );
+   dH[12] = v128_add32( v128_add32(
+                 v128_rol32( dH[0], 13 ),
+                 v128_xor( v128_xor( xh, qt[28] ), M[12] )),
+                 v128_xor( v128_sr32( xl, 3 ),
+                                v128_xor( qt[19], qt[12] ) ) );
+   dH[13] = v128_add32( v128_add32(
+                 v128_rol32( dH[1], 14 ),
+                 v128_xor( v128_xor( xh, qt[29] ), M[13] )),
+                 v128_xor( v128_sr32( xl, 4 ),
+                                v128_xor( qt[20], qt[13] ) ) );
+   dH[14] = v128_add32( v128_add32(
+                 v128_rol32( dH[2], 15 ),
+                 v128_xor( v128_xor( xh, qt[30] ), M[14] )),
+                 v128_xor( v128_sr32( xl, 7 ),
+                                v128_xor( qt[21], qt[14] ) ) );
+   dH[15] = v128_add32( v128_add32(
+                 v128_rol32( dH[3], 16 ),
+                 v128_xor( v128_xor( xh, qt[31] ), M[15] )),
+                 v128_xor( v128_sr32( xl, 2 ),
+                                v128_xor( qt[22], qt[15] ) ) );
 }
 
 static const uint32_t final_s[16][4] =
@@ -429,7 +429,7 @@ static const uint32_t final_s[16][4] =
    { 0xaaaaaaaf, 0xaaaaaaaf, 0xaaaaaaaf, 0xaaaaaaaf }
 };
 /*
-static const __m128i final_s[16] =
+static const v128u64_t final_s[16] =
 {
    { 0xaaaaaaa0aaaaaaa0, 0xaaaaaaa0aaaaaaa0 },
    { 0xaaaaaaa1aaaaaaa1, 0xaaaaaaa1aaaaaaa1 },
@@ -451,26 +451,26 @@ static const __m128i final_s[16] =
 */
 void bmw256_4way_init( bmw256_4way_context *ctx )
 {
-   ctx->H[ 0] = _mm_set1_epi64x( 0x4041424340414243 );
-   ctx->H[ 1] = _mm_set1_epi64x( 0x4445464744454647 );
-   ctx->H[ 2] = _mm_set1_epi64x( 0x48494A4B48494A4B );
-   ctx->H[ 3] = _mm_set1_epi64x( 0x4C4D4E4F4C4D4E4F );
-   ctx->H[ 4] = _mm_set1_epi64x( 0x5051525350515253 );
-   ctx->H[ 5] = _mm_set1_epi64x( 0x5455565754555657 );
-   ctx->H[ 6] = _mm_set1_epi64x( 0x58595A5B58595A5B );
-   ctx->H[ 7] = _mm_set1_epi64x( 0x5C5D5E5F5C5D5E5F );
-   ctx->H[ 8] = _mm_set1_epi64x( 0x6061626360616263 );
-   ctx->H[ 9] = _mm_set1_epi64x( 0x6465666764656667 );
-   ctx->H[10] = _mm_set1_epi64x( 0x68696A6B68696A6B );
-   ctx->H[11] = _mm_set1_epi64x( 0x6C6D6E6F6C6D6E6F );
-   ctx->H[12] = _mm_set1_epi64x( 0x7071727370717273 );
-   ctx->H[13] = _mm_set1_epi64x( 0x7475767774757677 );
-   ctx->H[14] = _mm_set1_epi64x( 0x78797A7B78797A7B );
-   ctx->H[15] = _mm_set1_epi64x( 0x7C7D7E7F7C7D7E7F );
+   ctx->H[ 0] = v128_64( 0x4041424340414243 );
+   ctx->H[ 1] = v128_64( 0x4445464744454647 );
+   ctx->H[ 2] = v128_64( 0x48494A4B48494A4B );
+   ctx->H[ 3] = v128_64( 0x4C4D4E4F4C4D4E4F );
+   ctx->H[ 4] = v128_64( 0x5051525350515253 );
+   ctx->H[ 5] = v128_64( 0x5455565754555657 );
+   ctx->H[ 6] = v128_64( 0x58595A5B58595A5B );
+   ctx->H[ 7] = v128_64( 0x5C5D5E5F5C5D5E5F );
+   ctx->H[ 8] = v128_64( 0x6061626360616263 );
+   ctx->H[ 9] = v128_64( 0x6465666764656667 );
+   ctx->H[10] = v128_64( 0x68696A6B68696A6B );
+   ctx->H[11] = v128_64( 0x6C6D6E6F6C6D6E6F );
+   ctx->H[12] = v128_64( 0x7071727370717273 );
+   ctx->H[13] = v128_64( 0x7475767774757677 );
+   ctx->H[14] = v128_64( 0x78797A7B78797A7B );
+   ctx->H[15] = v128_64( 0x7C7D7E7F7C7D7E7F );
 
 
 //   for ( int i = 0; i < 16; i++ )
-//      sc->H[i] = _mm_set1_epi32( iv[i] );
+//      sc->H[i] = v128_32( iv[i] );
    ctx->ptr = 0;
    ctx->bit_count = 0;
 }
@@ -478,10 +478,10 @@ void bmw256_4way_init( bmw256_4way_context *ctx )
 static void
 bmw32_4way(bmw_4way_small_context *sc, const void *data, size_t len)
 {
-   __m128i *vdata = (__m128i*)data;
-   __m128i *buf;
-   __m128i htmp[16];
-   __m128i *h1, *h2;
+   v128u64_t *vdata = (v128u64_t*)data;
+   v128u64_t *buf;
+   v128u64_t htmp[16];
+   v128u64_t *h1, *h2;
    size_t ptr;
    const int buf_size = 64;  // bytes of one lane, compatible with len
 
@@ -497,13 +497,13 @@ bmw32_4way(bmw_4way_small_context *sc, const void *data, size_t len)
       clen = buf_size - ptr;
       if ( clen > len )
          clen = len;
-      memcpy_128( buf + (ptr>>2), vdata, clen >> 2 );
+      v128_memcpy( buf + (ptr>>2), vdata, clen >> 2 );
       vdata += ( clen >> 2 );
       len -= clen;
       ptr += clen;
       if ( ptr == buf_size )
       {
-         __m128i *ht;
+         v128u64_t *ht;
          compress_small( buf, h1, h2 );
          ht = h1;
          h1 = h2;
@@ -513,46 +513,45 @@ bmw32_4way(bmw_4way_small_context *sc, const void *data, size_t len)
    }
    sc->ptr = ptr;
 
-
    if ( h1 != sc->H )
-        memcpy_128( sc->H, h1, 16 );
+        v128_memcpy( sc->H, h1, 16 );
 }
 
 static void
 bmw32_4way_close(bmw_4way_small_context *sc, unsigned ub, unsigned n,
 	void *dst, size_t out_size_w32)
 {
-   __m128i *buf;
-   __m128i h1[16], h2[16], *h;
+   v128u64_t *buf;
+   v128u64_t h1[16], h2[16], *h;
    size_t ptr, u, v;
    const int buf_size = 64;  // bytes of one lane, compatible with len
 
    buf = sc->buf;
    ptr = sc->ptr;
-   buf[ ptr>>2 ] = _mm_set1_epi64x( 0x0000008000000080 );
+   buf[ ptr>>2 ] = v128_64( 0x0000008000000080 );
    ptr += 4;
    h = sc->H;
 
    // assume bit_count fits in 32 bits 
    if ( ptr > buf_size - 4 )
    {
-      memset_zero_128( buf + (ptr>>2), (buf_size - ptr) >> 2 );
+      v128_memset_zero( buf + (ptr>>2), (buf_size - ptr) >> 2 );
       compress_small( buf, h, h1 );
       ptr = 0;
       h = h1;
    }
-   memset_zero_128( buf + (ptr>>2), (buf_size - 8 - ptr) >> 2 );
-   buf[ (buf_size - 8) >> 2 ] = _mm_set1_epi32( sc->bit_count + n );
-   buf[ (buf_size - 4) >> 2 ] = m128_zero;
+   v128_memset_zero( buf + (ptr>>2), (buf_size - 8 - ptr) >> 2 );
+   buf[ (buf_size - 8) >> 2 ] = v128_32( sc->bit_count + n );
+   buf[ (buf_size - 4) >> 2 ] = v128_zero;
    compress_small( buf, h, h2 );
 
    for ( u = 0; u < 16; u ++ )
       buf[u] = h2[u];
 
-   compress_small( buf, (__m128i*)final_s, h1 );
+   compress_small( buf, (v128u64_t*)final_s, h1 );
 
    for (u = 0, v = 16 - out_size_w32; u < out_size_w32; u ++, v ++)
-      casti_m128i( dst, u ) = h1[v];
+      casti_v128( dst, u ) = h1[v];
 }
 
 /*

algo/fugue/fugue-aesni.c

@@ -146,7 +146,7 @@ MYALIGN const unsigned int _IV512[] = {
 
 #define SUBSTITUTE(r0, _t2 )\
 	_t2 = _mm_shuffle_epi8(r0, M128(_inv_shift_rows));\
-	_t2 = _mm_aesenclast_si128( _t2, m128_zero )
+	_t2 = _mm_aesenclast_si128( _t2, v128_zero )
 
 #define SUPERMIX(t0, t1, t2, t3, t4)\
    t2 = t0;\
@@ -162,16 +162,16 @@ MYALIGN const unsigned int _IV512[] = {
    t1 = _mm_shuffle_epi8(t4, M128(_supermix1d));\
    t4 = _mm_xor_si128(t4, t1);\
    t1 = _mm_shuffle_epi8(t2, M128(_supermix1a));\
-   t2 = mm128_xor3(t2, t3, t0 );\
+   t2 = v128_xor3(t2, t3, t0 );\
    t2 = _mm_shuffle_epi8(t2, M128(_supermix7a));\
-   t4 = mm128_xor3( t4, t1, t2 ); \
+   t4 = v128_xor3( t4, t1, t2 ); \
    t2 = _mm_shuffle_epi8(t2, M128(_supermix7b));\
    t3 = _mm_shuffle_epi8(t3, M128(_supermix2a));\
    t1 = _mm_shuffle_epi8(t0, M128(_supermix4a));\
    t0 = _mm_shuffle_epi8(t0, M128(_supermix4b));\
-   t4 = mm128_xor3( t4, t2, t1 ); \
+   t4 = v128_xor3( t4, t2, t1 ); \
    t0 = _mm_xor_si128(t0, t3);\
-   t4 = mm128_xor3(t4, t0, _mm_shuffle_epi8(t0, M128(_supermix4c)));
+   t4 = v128_xor3(t4, t0, _mm_shuffle_epi8(t0, M128(_supermix4c)));
 
 /*
 #define SUPERMIX(t0, t1, t2, t3, t4)\
@@ -188,7 +188,7 @@ MYALIGN const unsigned int _IV512[] = {
 	t4 = _mm_xor_si128(t4, t1);\
 	t1 = _mm_shuffle_epi8(t2, M128(_supermix1a));\
 	t4 = _mm_xor_si128(t4, t1);\
-	t2 = mm128_xor3(t2, t3, t0 );\
+	t2 = v128_xor3(t2, t3, t0 );\
 	t2 = _mm_shuffle_epi8(t2, M128(_supermix7a));\
 	t4 = _mm_xor_si128(t4, t2);\
 	t2 = _mm_shuffle_epi8(t2, M128(_supermix7b));\
@@ -485,7 +485,7 @@ HashReturn fugue512_Init(hashState_fugue *ctx, int nHashSize)
 	ctx->uBlockLength = 4;
 
 	for(i = 0; i < 6; i++)
-		ctx->state[i] = m128_zero;
+		ctx->state[i] = v128_zero;
 
 	ctx->state[6]  = _mm_load_si128((__m128i*)_IV512 + 0);
 	ctx->state[7]  = _mm_load_si128((__m128i*)_IV512 + 1);

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

@@ -66,7 +66,40 @@ static const v128u32_t BLEND_MASK = { 0xffffffff, 0, 0, 0xffffffff };
 
 #define gr_shuffle32( v )      v128_blendv( v128_qrev32( v ), v, BLEND_MASK )
 
-//#define gr_shuffle32( v )       v128_shufflev32( v, vmask_d8 )
+/*
+#define TRANSP_MASK \
+     0xd,0x5,0x9,0x1,0xc,0x4,0x8,0x0,0xf,0x7,0xb,0x3,0xe,0x6,0xa,0x2
+#define SUBSH_MASK0 \
+     0xb,0xe,0x1,0x4,0x7,0xa,0xd,0x0,0x3,0x6,0x9,0xc,0xf,0x2,0x5,0x8
+#define SUBSH_MASK1 \
+     0xc,0xf,0x2,0x5,0x8,0xb,0xe,0x1,0x4,0x7,0xa,0xd,0x0,0x3,0x6,0x9
+#define SUBSH_MASK2 \
+     0xd,0x0,0x3,0x6,0x9,0xc,0xf,0x2,0x5,0x8,0xb,0xe,0x1,0x4,0x7,0xa
+#define SUBSH_MASK3 \
+     0xe,0x1,0x4,0x7,0xa,0xd,0x0,0x3,0x6,0x9,0xc,0xf,0x2,0x5,0x8,0xb
+#define SUBSH_MASK4  \
+     0xf,0x2,0x5,0x8,0xb,0xe,0x1,0x4,0x7,0xa,0xd,0x0,0x3,0x6,0x9,0xc
+#define SUBSH_MASK5 \
+     0x0,0x3,0x6,0x9,0xc,0xf,0x2,0x5,0x8,0xb,0xe,0x1,0x4,0x7,0xa,0xd
+#define SUBSH_MASK6 \
+     0x1,0x4,0x7,0xa,0xd,0x0,0x3,0x6,0x9,0xc,0xf,0x2,0x5,0x8,0xb,0xe
+#define SUBSH_MASK7 \
+     0x6,0x9,0xc,0xf,0x2,0x5,0x8,0xb,0xe,0x1,0x4,0x7,0xa,0xd,0x0,0x3
+
+//#define gr_shuffle8( v, c )    v128_shullfev8( v, c )
+
+
+#define gr_shuffle8( v, c15, c14, c13, c12, c11, c10, c09, c08, \
+                        c07, c06, c05, c04, c03, c02, c01, c00 ) \
+  v128_movlane8( v128_movlane8( v128_movlane8( v128_movlane8( \
+  v128_movlane8( v128_movlane8( v128_movlane8( v128_movlane8( \
+  v128_movlane8( v128_movlane8( v128_movlane8( v128_movlane8( \
+  v128_movlane8( v128_movlane8( v128_movlane8( v128_movlane8( \
+    v, 15, v, c15 ), 14, v, c14 ), 13, v, c13 ), 12, v, c12 ), \
+       11, v, c11 ), 10, v, c10 ),  9, v, c09 ),  8, v, c08 ), \
+        7, v, c07 ),  6, v, c06 ),  5, v, c05 ),  4, v, c04 ), \
+        3, v, c03 ),  2, v, c02 ),  1, v, c01 ),  0, v, c00 )
+*/
 
 #else
 

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

@@ -626,7 +626,7 @@ static const __m256i SUBSH_MASK7_2WAY =
 
 #define ROUND_2WAY(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
   /* AddRoundConstant */\
-  b1 = mm256_bcast_m128( mm128_mask_32( m128_neg1, 0x3 ) ); \
+  b1 = mm256_bcast_m128( mm128_mask_32( v128_neg1, 0x3 ) ); \
   a0 = _mm256_xor_si256( a0, mm256_bcast_m128( round_const_l0[i] ) );\
   a1 = _mm256_xor_si256( a1, b1 );\
   a2 = _mm256_xor_si256( a2, b1 );\

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

@@ -38,7 +38,7 @@
 #include <stddef.h>
 #include "simd-utils.h"
 
-// SSE2 or NEON Hamsi-512 2x64
+#if defined(__SSE4_2__) || defined(__ARM_NEON)
 
 typedef struct
 {
@@ -57,6 +57,8 @@ void hamsi512_2x64_ctx( hamsi512_2x64_context *sc, void *dst, const void *data,
                         size_t len );
 void hamsi512_2x64( void *dst, const void *data, size_t len );
 
+#endif
+
 #if defined (__AVX2__)
 
 // Hamsi-512 4x64

algo/luffa/luffa_for_sse2.c

@@ -75,16 +75,16 @@
 #define SUBCRUMB( a0, a1, a2, a3 ) \
 { \
     v128_t t = a0; \
-    a0 = mm128_xoror( a3, a0, a1 ); \
+    a0 = v128_xoror( a3, a0, a1 ); \
     a2 = v128_xor( a2, a3 ); \
     a1 = _mm_ternarylogic_epi64( a1, a3, t, 0x87 ); /* a1 xnor (a3 & t) */ \
-    a3 = mm128_xorand( a2, a3, t ); \
-    a2 = mm128_xorand( a1, a2, a0 ); \
+    a3 = v128_xorand( a2, a3, t ); \
+    a2 = v128_xorand( a1, a2, a0 ); \
     a1 = v128_or( a1, a3 ); \
     a3 = v128_xor( a3, a2 ); \
     t  = v128_xor( t, a1 ); \
     a2 = v128_and( a2, a1 ); \
-    a1 = mm128_xnor( a1, a0 ); \
+    a1 = v128_xnor( a1, a0 ); \
     a0 = t; \
 }
 

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

@@ -35,13 +35,13 @@ static const uint32_t IV[5] =
    _mm_xor_si128( _mm_and_si128( _mm_xor_si128( y, z ), x ), z )
 
 #define F3(x, y, z) \
-   _mm_xor_si128( _mm_or_si128( x, mm128_not( y ) ), z )
+   _mm_xor_si128( _mm_or_si128( x, v128_not( y ) ), z )
 
 #define F4(x, y, z) \
    _mm_xor_si128( _mm_and_si128( _mm_xor_si128( x, y ), z ), y )
 
 #define F5(x, y, z) \
-   _mm_xor_si128( x, _mm_or_si128( y, mm128_not( z ) ) )
+   _mm_xor_si128( x, _mm_or_si128( y, v128_not( z ) ) )
 
 #define RR(a, b, c, d, e, f, s, r, k) \
 do{ \

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

@@ -319,7 +319,7 @@ int sha256_4x32_transform_le_short( v128_t *state_out, const v128_t *data,
    v128_t A, B, C, D, E, F, G, H, T0, T1, T2;
    v128_t vmask, targ, hash;
    int t6_mask, flip;
-   v128_t W[16];      memcpy_128( W, data, 16 );
+   v128_t W[16];      v128_memcpy( W, data, 16 );
 
    A = v128_load( state_in   );
    B = v128_load( state_in+1 );

algo/sha/sha512256d-4way.c

@@ -5,11 +5,11 @@
 #include <stdint.h>
 
 #if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
- #define SHA512256D_8WAY 1
+#define SHA512256D_8WAY 1
 #elif defined(__AVX2__)
- #define SHA512256D_4WAY 1
+#define SHA512256D_4WAY 1
 #elif defined(__SSE2__) || defined(__ARM_NEON)
- #define SHA512256D_2WAY 1
+#define SHA512256D_2WAY 1
 #endif
 
 #if defined(SHA512256D_8WAY)
@@ -110,14 +110,13 @@ int scanhash_sha512256d_4way( struct work *work, uint32_t max_nonce,
     const uint32_t first_nonce = pdata[19];
     const uint32_t last_nonce = max_nonce - 4;
     uint32_t n = first_nonce;
-    __m256i  *noncev = (__m256i*)vdata + 9;
     const int thr_id = mythr->id;
     const bool bench = opt_benchmark;
     const __m256i four = v256_64( 0x0000000400000000 );
 
     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 );
+    casti_m256i( vdata,9 ) = mm256_intrlv_blend_32( _mm256_set_epi32(
+                     n+3, 0, n+2, 0, n+1, 0, n, 0 ), casti_m256i( vdata,9 ) );
     do
     {
        sha512256d_4way_init( &ctx );
@@ -138,7 +137,7 @@ int scanhash_sha512256d_4way( struct work *work, uint32_t max_nonce,
              submit_solution( work, lane_hash, mythr );
           }
        }
-       *noncev = _mm256_add_epi32( *noncev, four );
+       casti_m256i( vdata,9 ) = _mm256_add_epi32( casti_m256i( vdata,9 ), four );
        n += 4;
     } while ( (n < last_nonce) && !work_restart[thr_id].restart );
 
@@ -180,11 +179,10 @@ int scanhash_sha512256d_2x64( struct work *work, uint32_t max_nonce,
     v128u64_t *noncev = (v128u64_t*)vdata + 9;
     const int thr_id = mythr->id;
     const bool bench = opt_benchmark;
-    const v128u64_t two = v128_64( 0x0000000200000000 );
+    const v128_t two = v128_64( 0x0000000200000000 );
 
     v128_bswap32_intrlv80_2x64( vdata, pdata );
-    *noncev = v128_add32( v128_set32( 1, 0, 0, 0 ), *noncev );
-//    *noncev = v128_intrlv_blend_32( v128_set32( n+1, 0, n, 0 ), *noncev );
+    *noncev = v128_intrlv_blend_32( v128_set32( n+1, 0, n, 0 ), *noncev );
 
     do
     {
@@ -279,7 +277,7 @@ int scanhash_sha512256d( struct work *work,   uint32_t max_nonce,
 
 bool register_sha512256d_algo( algo_gate_t* gate )
 {
-   gate->optimizations = AVX2_OPT | AVX512_OPT;
+   gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT | NEON_OPT;
 #if defined(SHA512256D_8WAY)
    gate->scanhash = (void*)&scanhash_sha512256d_8way;
 #elif defined(SHA512256D_4WAY)

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

@@ -71,7 +71,7 @@ static const uint32_t IV512[] =
 static void
 c512_2way( shavite512_2way_context *ctx, const void *msg )
 {
-   const __m128i zero = _mm_setzero_si128();
+   const v128_t zero = v128_zero;
    __m256i p0, p1, p2, p3, x;
    __m256i k00, k01, k02, k03, k10, k11, k12, k13;
    __m256i *m = (__m256i*)msg;
@@ -278,7 +278,7 @@ c512_2way( shavite512_2way_context *ctx, const void *msg )
 void shavite512_2way_init( shavite512_2way_context *ctx )
 {
     __m256i *h = (__m256i*)ctx->h;
-    __m128i *iv = (__m128i*)IV512;
+    v128_t *iv = (v128_t*)IV512;
    
    h[0] = mm256_bcast_m128( iv[0] );
    h[1] = mm256_bcast_m128( iv[1] );
@@ -358,7 +358,7 @@ void shavite512_2way_close( shavite512_2way_context *ctx, void *dst )
     count.u32[3] = ctx->count3;
 
     casti_m256i( buf, 6 ) = mm256_bcast_m128(
-                  _mm_insert_epi16( m128_zero, count.u16[0], 7 ) ); 
+                  _mm_insert_epi16( v128_zero, count.u16[0], 7 ) ); 
     casti_m256i( buf, 7 ) = mm256_bcast_m128( _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] ) );
@@ -434,7 +434,7 @@ void shavite512_2way_update_close( shavite512_2way_context *ctx, void *dst,
    }
 
     casti_m256i( buf, 6 ) = mm256_bcast_m128(
-                  _mm_insert_epi16( m128_zero, count.u16[0], 7 ) ); 
+                  _mm_insert_epi16( v128_zero, count.u16[0], 7 ) ); 
     casti_m256i( buf, 7 ) = mm256_bcast_m128( _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] ) );
@@ -451,7 +451,7 @@ 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;
+    v128_t *iv = (v128_t*)IV512;
 
    h[0] = mm256_bcast_m128( iv[0] );
    h[1] = mm256_bcast_m128( iv[1] );
@@ -524,7 +524,7 @@ void shavite512_2way_full( shavite512_2way_context *ctx, void *dst,
    }
 
     casti_m256i( buf, 6 ) = mm256_bcast_m128(
-                  _mm_insert_epi16( m128_zero, count.u16[0], 7 ) );
+                  _mm_insert_epi16( v128_zero, count.u16[0], 7 ) );
     casti_m256i( buf, 7 ) = mm256_bcast_m128( _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] ) );

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

@@ -303,7 +303,7 @@ void shavite512_4way_close( shavite512_4way_context *ctx, void *dst )
     count.u32[3] = ctx->count3;
 
     casti_m512i( buf, 6 ) = mm512_bcast_m128(
-                  _mm_insert_epi16( m128_zero, count.u16[0], 7 ) ); 
+                  _mm_insert_epi16( v128_zero, count.u16[0], 7 ) ); 
     casti_m512i( buf, 7 ) = mm512_bcast_m128( _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] ) );
@@ -379,7 +379,7 @@ void shavite512_4way_update_close( shavite512_4way_context *ctx, void *dst,
    }
 
     casti_m512i( buf, 6 ) = mm512_bcast_m128(
-                  _mm_insert_epi16( m128_zero, count.u16[0], 7 ) ); 
+                  _mm_insert_epi16( v128_zero, count.u16[0], 7 ) ); 
     casti_m512i( buf, 7 ) = mm512_bcast_m128( _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] ) );
@@ -470,7 +470,7 @@ void shavite512_4way_full( shavite512_4way_context *ctx, void *dst,
    }
 
     casti_m512i( buf, 6 ) = mm512_bcast_m128(
-                  _mm_insert_epi16( m128_zero, count.u16[0], 7 ) );
+                  _mm_insert_epi16( v128_zero, count.u16[0], 7 ) );
     casti_m512i( buf, 7 ) = mm512_bcast_m128( _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] ) );

algo/skein/skein-4way.c

@@ -159,4 +159,69 @@ int scanhash_skein_4way( struct work *work, uint32_t max_nonce,
     return 0;
 }
 
+#elif defined(SKEIN_2WAY)
+
+static __thread skein512_2x64_context skein512_2x64_ctx
+                                            __attribute__ ((aligned (64)));
+
+void skeinhash_2x64( void *state, const void *input )
+{
+     uint64_t vhash64[8*2] __attribute__ ((aligned (32)));
+     uint32_t hash0[16] __attribute__ ((aligned (32)));
+     uint32_t hash1[16] __attribute__ ((aligned (32)));
+     skein512_2x64_context ctx_skein;
+     memcpy( &ctx_skein, &skein512_2x64_ctx, sizeof( ctx_skein ) );
+
+     skein512_2x64_final16( &ctx_skein, vhash64, input + (64*2) );
+
+     dintrlv_2x64( hash0, hash1, vhash64, 512 );
+
+     sha256_full( hash0, hash0, 64 );
+     sha256_full( hash1, hash1, 64 );
+
+     intrlv_2x32( state, hash0, hash1, 256 );
+}
+
+int scanhash_skein_2x64( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr )
+{
+    uint32_t vdata[20*2] __attribute__ ((aligned (32)));
+    uint32_t hash[8*2] __attribute__ ((aligned (32)));
+    uint32_t lane_hash[8] __attribute__ ((aligned (32)));
+    uint32_t *hash_d7 = &(hash[7<<1]);
+    uint32_t *pdata = work->data;
+    uint32_t *ptarget = work->target;
+    const uint32_t targ_d7 = ptarget[7];
+    const uint32_t first_nonce = pdata[19];
+    const uint32_t last_nonce = max_nonce - 2;
+    uint32_t n = first_nonce;
+    v128u32_t  *noncev = (v128u32_t*)vdata + 9;
+    const int thr_id = mythr->id;
+    const bool bench = opt_benchmark;
+
+   v128_bswap32_intrlv80_2x64( vdata, pdata );
+   skein512_2x64_prehash64( &skein512_2x64_ctx, vdata );
+   *noncev = v128_intrlv_blend_32( v128_set32( n+1, 0, n, 0 ), *noncev );
+   do
+   {
+       skeinhash_2x64( hash, vdata );
+       for ( int lane = 0; lane < 2; lane++ )
+       if ( unlikely( ( hash_d7[ lane ] <= targ_d7 ) && !bench ) )
+       {
+          extr_lane_2x32( lane_hash, hash, lane, 256 );
+          if ( valid_hash( lane_hash, ptarget ) )
+          {
+             pdata[19] = bswap_32( n + lane );
+             submit_solution( work, lane_hash, mythr );
+          }
+       }
+       *noncev = v128_add32( *noncev, v128_64( 0x0000000200000000 ) );
+       n += 2;
+    } while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
+
+    pdata[19] = n;
+    *hashes_done = n - first_nonce;
+    return 0;
+}
+
 #endif

algo/skein/skein-gate.c

@@ -3,16 +3,20 @@
 
 bool register_skein_algo( algo_gate_t* gate )
 {
-#if defined (SKEIN_8WAY)
-    gate->optimizations = AVX2_OPT | AVX512_OPT;
+#if defined(SKEIN_8WAY)
+    gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT;
     gate->scanhash  = (void*)&scanhash_skein_8way;
     gate->hash      = (void*)&skeinhash_8way;
-#elif defined (SKEIN_4WAY)
-    gate->optimizations = AVX2_OPT | AVX512_OPT | SHA_OPT;
+#elif defined(SKEIN_4WAY)
+    gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT | SHA_OPT | NEON_OPT;
     gate->scanhash  = (void*)&scanhash_skein_4way;
     gate->hash      = (void*)&skeinhash_4way;
+#elif defined(SKEIN_2WAY)
+    gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT | SHA_OPT | NEON_OPT;
+    gate->scanhash  = (void*)&scanhash_skein_2x64;
+    gate->hash      = (void*)&skeinhash_2x64;
 #else
-    gate->optimizations = AVX2_OPT | AVX512_OPT | SHA_OPT;
+    gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT | SHA_OPT | NEON_OPT;
     gate->scanhash  = (void*)&scanhash_skein;
     gate->hash      = (void*)&skeinhash;
 #endif
@@ -21,16 +25,15 @@ bool register_skein_algo( algo_gate_t* gate )
 
 bool register_skein2_algo( algo_gate_t* gate )
 {
-  gate->optimizations = AVX2_OPT | AVX512_OPT;
-#if defined (SKEIN_8WAY)
+  gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT | NEON_OPT;
+#if defined(SKEIN_8WAY)
   gate->scanhash  = (void*)&scanhash_skein2_8way;
-  gate->hash      = (void*)&skein2hash_8way;
-#elif defined (SKEIN_4WAY)
+#elif defined(SKEIN_4WAY)
   gate->scanhash  = (void*)&scanhash_skein2_4way;
-  gate->hash      = (void*)&skein2hash_4way;
+#elif defined(SKEIN_2WAY)
+  gate->scanhash  = (void*)&scanhash_skein2_2x64;
 #else
   gate->scanhash  = (void*)&scanhash_skein2;
-  gate->hash      = (void*)&skein2hash;
 #endif
   return true;
 };

algo/skein/skein-gate.h

@@ -7,6 +7,8 @@
   #define SKEIN_8WAY 1
 #elif defined(__AVX2__)
   #define SKEIN_4WAY 1
+#elif defined(__SSE2__) || defined(__ARM_NEON)
+  #define SKEIN_2WAY 1
 #endif
 
 #if defined(SKEIN_8WAY)
@@ -29,6 +31,16 @@ void skein2hash_4way( void *output, const void *input );
 int scanhash_skein2_4way( struct work *work, uint32_t max_nonce,
                           uint64_t* hashes_done, struct thr_info *mythr );
 
+#elif defined(SKEIN_2WAY)
+
+void skeinhash_2x64( void *output, const void *input );
+int scanhash_skein_2x64( struct work *work, uint32_t max_nonce,
+                         uint64_t *hashes_done, struct thr_info *mythr );
+
+void skein2hash_2x64( void *output, const void *input );
+int scanhash_skein2_2x64( struct work *work, uint32_t max_nonce,
+                          uint64_t* hashes_done, struct thr_info *mythr );
+
 #else
 
 void skeinhash( void *output, const void *input );

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

@@ -675,11 +675,13 @@ void skein512_8way_full( skein512_8way_context *sc, void *out, const void *data,
 
 // Close
 
-   unsigned et;
-
-   memset_zero_512( buf + (ptr>>3), (buf_size - ptr) >> 3 );
-   et = 352 + ((bcount == 0) << 7);
-   UBI_BIG_8WAY( et, ptr );
+   if ( ptr )
+   {
+      unsigned et;
+      memset_zero_512( buf + (ptr>>3), (buf_size - ptr) >> 3 );
+      et = 352 + ((bcount == 0) << 7);
+      UBI_BIG_8WAY( et, ptr );
+   }
 
    memset_zero_512( buf, buf_size >> 3 );
    bcount = 0;
@@ -970,11 +972,13 @@ skein512_4way_full( skein512_4way_context *sc, void *out, const void *data,
 
 // Close
 
-   unsigned et;
-
-   memset_zero_256( buf + (ptr>>3), (buf_size - ptr) >> 3 );
-   et = 352 + ((bcount == 0) << 7);
-   UBI_BIG_4WAY( et, ptr );
+   if ( ptr )
+   {
+      unsigned et;
+      memset_zero_256( buf + (ptr>>3), (buf_size - ptr) >> 3 );
+      et = 352 + ((bcount == 0) << 7);
+      UBI_BIG_4WAY( et, ptr );
+   }
 
    memset_zero_256( buf, buf_size >> 3 );
    bcount = 0;
@@ -1364,11 +1368,13 @@ skein512_2x64_full( skein512_2x64_context *sc, void *out, const void *data,
 
 // Close
 
-   unsigned et;
-
-   v128_memset_zero( buf + (ptr>>3), (buf_size - ptr) >> 3 );
-   et = 352 + ((bcount == 0) << 7);
-   UBI_BIG_2WAY( et, ptr );
+   if ( ptr )
+   {
+      unsigned et;
+      v128_memset_zero( buf + (ptr>>3), (buf_size - ptr) >> 3 );
+      et = 352 + ((bcount == 0) << 7);
+      UBI_BIG_2WAY( et, ptr );
+   }
 
    v128_memset_zero( buf, buf_size >> 3 );
    bcount = 0;

algo/skein/skein2-4way.c

@@ -5,19 +5,6 @@
 
 #if defined(SKEIN_8WAY)
 
- static __thread skein512_8way_context skein512_8way_ctx
-                                             __attribute__ ((aligned (64)));
-
-void skein2hash_8way( void *output, const void *input )
-{
-   uint64_t hash[16*8] __attribute__ ((aligned (128)));
-   skein512_8way_context ctx;
-   memcpy( &ctx, &skein512_8way_ctx, sizeof( ctx ) );
-
-   skein512_8way_final16( &ctx, hash, input + (64*8) );
-   skein512_8way_full( &ctx, output, hash, 64 );
-}
-
 int scanhash_skein2_8way( struct work *work, uint32_t max_nonce,
                           uint64_t *hashes_done, struct thr_info *mythr )
 {
@@ -68,19 +55,6 @@ int scanhash_skein2_8way( struct work *work, uint32_t max_nonce,
 
 #elif defined(SKEIN_4WAY)
 
-static __thread skein512_4way_context skein512_4way_ctx
-                                           __attribute__ ((aligned (64)));
-
-void skein2hash_4way( void *output, const void *input )
-{
-   skein512_4way_context ctx;
-   memcpy( &ctx, &skein512_4way_ctx, sizeof( ctx ) ); 
-   uint64_t hash[16*4] __attribute__ ((aligned (64)));
-
-   skein512_4way_final16( &ctx, hash, input + (64*4) );
-   skein512_4way_full( &ctx, output, hash, 64 );
-}
-
 int scanhash_skein2_4way( struct work *work, uint32_t max_nonce,
                           uint64_t *hashes_done, struct thr_info *mythr )
 {
@@ -128,4 +102,53 @@ int scanhash_skein2_4way( struct work *work, uint32_t max_nonce,
     return 0;
 }
 
+#elif defined(SKEIN_2WAY)
+
+int scanhash_skein2_2x64( struct work *work, uint32_t max_nonce,
+                          uint64_t *hashes_done, struct thr_info *mythr )
+{
+    uint64_t hash[8*2] __attribute__ ((aligned (64)));
+    uint32_t vdata[20*2] __attribute__ ((aligned (64)));
+    skein512_2x64_context ctx;
+    uint32_t lane_hash[8] __attribute__ ((aligned (32)));
+    uint64_t *hash_q3 = &(hash[3*2]);
+    uint32_t *pdata = work->data;
+    uint32_t *ptarget = work->target;
+    const uint64_t targ_q3 = ((uint64_t*)ptarget)[3];
+    const uint32_t first_nonce = pdata[19];
+    const uint32_t last_nonce = max_nonce - 4;
+    uint32_t n = first_nonce;
+    v128u64_t *noncev = (v128u64_t*)vdata + 9;
+    const int thr_id = mythr->id;
+    const bool bench = opt_benchmark;
+    const v128u64_t two = v128_64( 0x0000000200000000 );
+
+    v128_bswap32_intrlv80_2x64( vdata, pdata );
+    skein512_2x64_prehash64( &ctx, vdata );
+    *noncev = v128_intrlv_blend_32( v128_set32( n+1, 0, n, 0 ), *noncev );
+
+    do
+    {
+       skein512_2x64_final16( &ctx, hash, vdata + (16*2) );
+       skein512_2x64_full( &ctx, hash, hash, 64 );
+
+       for ( int lane = 0; lane < 2; lane++ )
+       if ( hash_q3[ lane ] <= targ_q3 )
+       {
+          extr_lane_2x64( lane_hash, hash, lane, 256 );
+          if ( valid_hash( lane_hash, ptarget ) && !bench )
+          {
+             pdata[19] = bswap_32( n + lane );
+             submit_solution( work, lane_hash, mythr );
+          }
+       }
+       *noncev = v128_add32( *noncev, two );
+       n += 2;
+    } while ( (n < last_nonce) && !work_restart[thr_id].restart );
+
+    pdata[19] = n;
+    *hashes_done = n - first_nonce;
+    return 0;
+}
+
 #endif

algo/verthash/Verthash.c

@@ -191,7 +191,7 @@ static void rotate_indexes( uint32_t *p )
    *(__m256i*)hash = _mm256_mullo_epi32( _mm256_xor_si256( \
                                  *(__m256i*)hash, *(__m256i*)blob_off ), k );
 
-#elif defined(__SSE4_1__) // || defined(__ARM_NEON)
+#elif defined(__SSE4_1__)  || defined(__ARM_NEON)
 
 #define MULXOR \
    casti_v128( hash, 0 ) = v128_mul32( v128_xor( \
@@ -251,7 +251,7 @@ void verthash_hash( const void *blob_bytes, const size_t blob_size,
                              / VH_BYTE_ALIGNMENT ) + 1;
 #if defined (__AVX2__)        
     const __m256i k = _mm256_set1_epi32( 0x1000193 );
-#elif defined(__SSE4_1__) // || defined(__ARM_NEON)
+#elif defined(__SSE4_1__)  || defined(__ARM_NEON)
     const v128u32_t k = v128_32( 0x1000193 );
 #endif
     

algo/verthash/verthash-gate.c

@@ -129,7 +129,7 @@ bool register_verthash_algo( algo_gate_t* gate )
 {
   opt_target_factor = 256.0;
   gate->scanhash  = (void*)&scanhash_verthash;
-  gate->optimizations = SSE42_OPT | AVX2_OPT;
+  gate->optimizations = SSE42_OPT | AVX2_OPT | NEON_OPT;
    
   const char *verthash_data_file = opt_data_file ? opt_data_file
                                                  : default_verthash_data_file;

algo/x16/minotaur.c

@@ -11,7 +11,9 @@
 #include "algo/keccak/sph_keccak.h"
 #include "algo/skein/sph_skein.h"
 #include "algo/shavite/sph_shavite.h"
+#include "algo/luffa/luffa_for_sse2.h"
 #include "algo/cubehash/cubehash_sse2.h"
+#include "algo/simd/simd-hash-2way.h"
 #if defined(__aarch64__)
   #include "algo/simd/sph_simd.h"
 #endif
@@ -20,20 +22,17 @@
 #include "algo/whirlpool/sph_whirlpool.h"
 #include "algo/sha/sph_sha2.h"
 #include "algo/yespower/yespower.h"
-//#if defined(__AES__) || defined(__ARM_FEATURE_AES)
+#if defined(__AES__) || defined(__ARM_FEATURE_AES)
   #include "algo/echo/aes_ni/hash_api.h"
   #include "algo/groestl/aes_ni/hash-groestl.h"
-//#else
-  #include "algo/echo/sph_echo.h"
-  #include "algo/groestl/sph_groestl.h"
-//#endif
+#endif
+#include "algo/echo/sph_echo.h"
+#include "algo/groestl/sph_groestl.h"
 #if defined(__AES__)
   #include "algo/fugue/fugue-aesni.h"
 #else
   #include "algo/fugue/sph_fugue.h"
 #endif
-#include "algo/luffa/luffa_for_sse2.h"
-#include "algo/simd/nist.h"
 
 // Config
 #define MINOTAUR_ALGO_COUNT	16
@@ -48,12 +47,15 @@ typedef struct TortureGarden TortureGarden;
 struct TortureGarden
 {
 #if defined(__AES__) // || defined(__ARM_FEATURE_AES)
-   hashState_echo          echo;
    hashState_groestl       groestl;
 #else
-   sph_echo512_context     echo;
    sph_groestl512_context  groestl;
 #endif
+#if defined(__AES__) || defined(__ARM_FEATURE_AES)
+   hashState_echo          echo;
+#else
+   sph_echo512_context     echo;
+#endif
 #if defined(__AES__)
    hashState_fugue         fugue;
 #else
@@ -67,11 +69,7 @@ struct TortureGarden
    cubehashParam           cube;
    shavite512_context      shavite;
    hashState_luffa         luffa;
-#if defined(__aarch64__)
-   sph_simd512_context     simd;
-#else
-   hashState_sd            simd;
-#endif
+   simd512_context         simd;
    sph_hamsi512_context    hamsi;
    sph_shabal512_context   shabal;
    sph_whirlpool_context   whirlpool;
@@ -93,9 +91,7 @@ static int get_hash( void *output, const void *input, TortureGarden *garden,
     switch ( algo )
     {
         case 0:
-            blake512_init( &garden->blake );
-            blake512_update( &garden->blake, input, 64 );
-            blake512_close( &garden->blake, hash );
+            blake512_full( &garden->blake, hash, input, 64 );
             break;
         case 1:
             sph_bmw512_init( &garden->bmw );
@@ -107,7 +103,7 @@ static int get_hash( void *output, const void *input, TortureGarden *garden,
             cubehashUpdateDigest( &garden->cube, hash, input, 64 );
             break;
         case 3:
-#if defined(__AES__) // || defined(__ARM_FEATURE_AES)
+#if defined(__AES__) || defined(__ARM_FEATURE_AES)
             echo_full( &garden->echo, hash, 512, input, 64 );
 #else
             sph_echo512_init( &garden->echo );
@@ -165,13 +161,7 @@ static int get_hash( void *output, const void *input, TortureGarden *garden,
             sph_shavite512_close( &garden->shavite, hash );          
             break;
         case 13:
-#if defined(__aarch64__)
-            sph_simd512_init( &garden->simd );
-            sph_simd512( &garden->simd, input, 64);
-            sph_simd512_close( &garden->simd, hash );
-#else
-            simd_full( &garden->simd, (BitSequence *)hash, input, 512 );
-#endif
+            simd512_ctx( &garden->simd, hash, input, 64 );
             break;
         case 14:
             sph_skein512_init( &garden->skein );

algo/x17/x17-4way.c

@@ -931,15 +931,19 @@ int scanhash_x17_4x64( struct work *work, uint32_t max_nonce,
 // Need sph in some cases
 #include "algo/luffa/luffa_for_sse2.h"
 #include "algo/cubehash/cubehash_sse2.h"
-#include "algo/simd/sph_simd.h"
-#include "algo/simd/nist.h"
-#include "algo/hamsi/sph_hamsi.h"
+//#include "algo/simd/sph_simd.h"
+//#include "algo/simd/nist.h"
+#if !( defined(__SSE4_2__) || defined(__ARM_NEON) )
+  #include "algo/hamsi/sph_hamsi.h"
+#endif
 #include "algo/shabal/sph_shabal.h"
 #include "algo/haval/sph-haval.h"
-//#if !( defined(__AES__) || defined(__ARM_FEATURE_AES) )
+#if !( defined(__AES__) ) //|| defined(__ARM_FEATURE_AES) )
   #include "algo/groestl/sph_groestl.h"
+#endif
+#if !( defined(__AES__) || defined(__ARM_FEATURE_AES) )
   #include "algo/echo/sph_echo.h"
-//#endif
+#endif
 #include "algo/fugue/sph_fugue.h"
 
 union _x17_context_overlay
@@ -967,12 +971,8 @@ union _x17_context_overlay
         hashState_luffa         luffa;
         cubehashParam           cube;
         sph_shavite512_context  shavite;
-#if defined(__x86_64__)
         simd512_context         simd;
-#else
-        sph_simd512_context     simd;
-#endif
-#if defined(__SSE4_2__) // || defined(__ARM_NEON)
+#if defined(__SSE4_2__) || defined(__ARM_NEON)
         hamsi_2x64_context      hamsi;
 #else
         sph_hamsi512_context    hamsi;
@@ -1033,17 +1033,8 @@ int x17_2x64_hash( void *output, const void *input, int thr_id )
     sph_shavite512( &ctx.shavite, hash1, 64 );
     sph_shavite512_close( &ctx.shavite, hash1 );
 
-#if defined(__x86_64__)
     simd512_ctx( &ctx.simd, hash0, hash0, 64 );
     simd512_ctx( &ctx.simd, hash1, hash1, 64 );
-#else
-    sph_simd512_init( &ctx.simd );
-    sph_simd512( &ctx.simd, hash0, 64 );
-    sph_simd512_close( &ctx.simd, hash0 );
-    sph_simd512_init( &ctx.simd );
-    sph_simd512( &ctx.simd, hash1, 64 );
-    sph_simd512_close( &ctx.simd, hash1 );
-#endif
 
 #if defined(__AES__) || defined(__ARM_FEATURE_AES)
     echo_full( &ctx.echo, hash0, 512, hash0, 64 );
@@ -1057,7 +1048,7 @@ int x17_2x64_hash( void *output, const void *input, int thr_id )
     sph_echo512_close( &ctx.echo, hash1 );
 #endif
 
-#if defined(__SSE4_2__) // || defined(__ARM_NEON)
+#if defined(__SSE4_2__) || defined(__ARM_NEON)
     intrlv_2x64( vhash, hash0, hash1, 512 );
     hamsi512_2x64_ctx( &ctx.hamsi, vhash, vhash, 64 );
     dintrlv_2x64( hash0, hash1, vhash, 512 );

configure

@@ -1,6 +1,6 @@
 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.71 for cpuminer-opt 23.8.
+# Generated by GNU Autoconf 2.71 for cpuminer-opt 23.10.
 #
 #
 # Copyright (C) 1992-1996, 1998-2017, 2020-2021 Free Software Foundation,
@@ -608,8 +608,8 @@ MAKEFLAGS=
 # Identity of this package.
 PACKAGE_NAME='cpuminer-opt'
 PACKAGE_TARNAME='cpuminer-opt'
-PACKAGE_VERSION='23.8'
-PACKAGE_STRING='cpuminer-opt 23.8'
+PACKAGE_VERSION='23.10'
+PACKAGE_STRING='cpuminer-opt 23.10'
 PACKAGE_BUGREPORT=''
 PACKAGE_URL=''
 
@@ -1360,7 +1360,7 @@ if test "$ac_init_help" = "long"; then
   # Omit some internal or obsolete options to make the list less imposing.
   # This message is too long to be a string in the A/UX 3.1 sh.
   cat <<_ACEOF
-\`configure' configures cpuminer-opt 23.8 to adapt to many kinds of systems.
+\`configure' configures cpuminer-opt 23.10 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1432,7 +1432,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of cpuminer-opt 23.8:";;
+     short | recursive ) echo "Configuration of cpuminer-opt 23.10:";;
    esac
   cat <<\_ACEOF
 
@@ -1538,7 +1538,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-cpuminer-opt configure 23.8
+cpuminer-opt configure 23.10
 generated by GNU Autoconf 2.71
 
 Copyright (C) 2021 Free Software Foundation, Inc.
@@ -1985,7 +1985,7 @@ cat >config.log <<_ACEOF
 This file contains any messages produced by compilers while
 running configure, to aid debugging if configure makes a mistake.
 
-It was created by cpuminer-opt $as_me 23.8, which was
+It was created by cpuminer-opt $as_me 23.10, which was
 generated by GNU Autoconf 2.71.  Invocation command line was
 
   $ $0$ac_configure_args_raw
@@ -3593,7 +3593,7 @@ fi
 
 # Define the identity of the package.
  PACKAGE='cpuminer-opt'
- VERSION='23.8'
+ VERSION='23.10'
 
 
 printf "%s\n" "#define PACKAGE \"$PACKAGE\"" >>confdefs.h
@@ -7508,7 +7508,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
 # report actual input values of CONFIG_FILES etc. instead of their
 # values after options handling.
 ac_log="
-This file was extended by cpuminer-opt $as_me 23.8, which was
+This file was extended by cpuminer-opt $as_me 23.10, which was
 generated by GNU Autoconf 2.71.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -7576,7 +7576,7 @@ ac_cs_config_escaped=`printf "%s\n" "$ac_cs_config" | sed "s/^ //; s/'/'\\\\\\\\
 cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
 ac_cs_config='$ac_cs_config_escaped'
 ac_cs_version="\\
-cpuminer-opt config.status 23.8
+cpuminer-opt config.status 23.10
 configured by $0, generated by GNU Autoconf 2.71,
   with options \\"\$ac_cs_config\\"
 

configure.ac

@@ -1,4 +1,4 @@
-AC_INIT([cpuminer-opt], [23.8])
+AC_INIT([cpuminer-opt], [23.10])
 
 AC_PREREQ([2.59c])
 AC_CANONICAL_SYSTEM

configure~

@@ -1,6 +1,6 @@
 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.71 for cpuminer-opt 23.8.
+# Generated by GNU Autoconf 2.71 for cpuminer-opt 23.9.
 #
 #
 # Copyright (C) 1992-1996, 1998-2017, 2020-2021 Free Software Foundation,
@@ -608,8 +608,8 @@ MAKEFLAGS=
 # Identity of this package.
 PACKAGE_NAME='cpuminer-opt'
 PACKAGE_TARNAME='cpuminer-opt'
-PACKAGE_VERSION='23.8'
-PACKAGE_STRING='cpuminer-opt 23.8'
+PACKAGE_VERSION='23.9'
+PACKAGE_STRING='cpuminer-opt 23.9'
 PACKAGE_BUGREPORT=''
 PACKAGE_URL=''
 
@@ -657,8 +657,6 @@ JANSSON_LIBS
 LIBCURL_CPPFLAGS
 LIBCURL_CFLAGS
 LIBCURL
-HAVE_MACOS_FALSE
-HAVE_MACOS_TRUE
 MINGW_FALSE
 MINGW_TRUE
 ARCH_ARM_FALSE
@@ -1362,7 +1360,7 @@ if test "$ac_init_help" = "long"; then
   # Omit some internal or obsolete options to make the list less imposing.
   # This message is too long to be a string in the A/UX 3.1 sh.
   cat <<_ACEOF
-\`configure' configures cpuminer-opt 23.8 to adapt to many kinds of systems.
+\`configure' configures cpuminer-opt 23.9 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1434,7 +1432,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of cpuminer-opt 23.8:";;
+     short | recursive ) echo "Configuration of cpuminer-opt 23.9:";;
    esac
   cat <<\_ACEOF
 
@@ -1540,7 +1538,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-cpuminer-opt configure 23.8
+cpuminer-opt configure 23.9
 generated by GNU Autoconf 2.71
 
 Copyright (C) 2021 Free Software Foundation, Inc.
@@ -1987,7 +1985,7 @@ cat >config.log <<_ACEOF
 This file contains any messages produced by compilers while
 running configure, to aid debugging if configure makes a mistake.
 
-It was created by cpuminer-opt $as_me 23.8, which was
+It was created by cpuminer-opt $as_me 23.9, which was
 generated by GNU Autoconf 2.71.  Invocation command line was
 
   $ $0$ac_configure_args_raw
@@ -3595,7 +3593,7 @@ fi
 
 # Define the identity of the package.
  PACKAGE='cpuminer-opt'
- VERSION='23.8'
+ VERSION='23.9'
 
 
 printf "%s\n" "#define PACKAGE \"$PACKAGE\"" >>confdefs.h
@@ -6881,14 +6879,6 @@ else
   MINGW_FALSE=
 fi
 
- if test "x$OS" = "xAPPLE"; then
-  HAVE_MACOS_TRUE=
-  HAVE_MACOS_FALSE='#'
-else
-  HAVE_MACOS_TRUE='#'
-  HAVE_MACOS_FALSE=
-fi
-
 
 if test x$request_jansson = xtrue ; then
 	JANSSON_LIBS="compat/jansson/libjansson.a"
@@ -7128,10 +7118,6 @@ if test -z "${MINGW_TRUE}" && test -z "${MINGW_FALSE}"; then
   as_fn_error $? "conditional \"MINGW\" was never defined.
 Usually this means the macro was only invoked conditionally." "$LINENO" 5
 fi
-if test -z "${HAVE_MACOS_TRUE}" && test -z "${HAVE_MACOS_FALSE}"; then
-  as_fn_error $? "conditional \"HAVE_MACOS\" was never defined.
-Usually this means the macro was only invoked conditionally." "$LINENO" 5
-fi
 
 : "${CONFIG_STATUS=./config.status}"
 ac_write_fail=0
@@ -7522,7 +7508,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
 # report actual input values of CONFIG_FILES etc. instead of their
 # values after options handling.
 ac_log="
-This file was extended by cpuminer-opt $as_me 23.8, which was
+This file was extended by cpuminer-opt $as_me 23.9, which was
 generated by GNU Autoconf 2.71.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -7590,7 +7576,7 @@ ac_cs_config_escaped=`printf "%s\n" "$ac_cs_config" | sed "s/^ //; s/'/'\\\\\\\\
 cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
 ac_cs_config='$ac_cs_config_escaped'
 ac_cs_version="\\
-cpuminer-opt config.status 23.8
+cpuminer-opt config.status 23.9
 configured by $0, generated by GNU Autoconf 2.71,
   with options \\"\$ac_cs_config\\"
 

cpu-miner.c

@@ -2968,8 +2968,12 @@ static bool cpu_capability( bool display_only )
         printf(" Linux\n");
      #elif defined(WIN32)
         printf(" Windows\n");
+     #elif defined(__APPLE__)
+        printf(" MacOS\n");
+#elif defined(__unix__) || defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__NetBSD__) 
+        printf(" Unix\n");
      #else
-       printf("\n");
+        printf("\n");
      #endif
 
      printf("CPU features: ");

miner.h

@@ -3,12 +3,16 @@
 
 #include <cpuminer-config.h>
 
+#if !( defined(__SSE2__) || ( defined(__aarch64__) && defined(__ARM_NEON) ) )
+#warning "Unknown or unsupported CPU, requires x86_64 with SSE2 or AArch64 with NEON." 
+#endif
+
 #if defined(__x86_64__)
    #define USER_AGENT_ARCH "x64"     // Intel, AMD x86_64
 #elif defined(__aarch64__)
    #define USER_AGENT_ARCH "arm"     // AArch64
 //#elif
-//  #define USER_AGENT_ARCH "R5"     // RISC-V             
+//  #define USER_AGENT_ARCH "r5"     // RISC-V             
 #else
    #define USER_AGENT_ARCH
 #endif

simd-utils/intrlv.h

@@ -411,11 +411,11 @@ static inline void v128_bswap32_80( void *d, void *s )
 {
   const v128_t bswap_shuf = _mm_set_epi64x( 0x0c0d0e0f08090a0b,
                                              0x0405060700010203 );
-  casti_m128i( d, 0 ) = _mm_shuffle_epi8( casti_m128i( s, 0 ), bswap_shuf );
-  casti_m128i( d, 1 ) = _mm_shuffle_epi8( casti_m128i( s, 1 ), bswap_shuf );
-  casti_m128i( d, 2 ) = _mm_shuffle_epi8( casti_m128i( s, 2 ), bswap_shuf );
-  casti_m128i( d, 3 ) = _mm_shuffle_epi8( casti_m128i( s, 3 ), bswap_shuf );
-  casti_m128i( d, 4 ) = _mm_shuffle_epi8( casti_m128i( s, 4 ), bswap_shuf );
+  casti_v128( d, 0 ) = _mm_shuffle_epi8( casti_v128( s, 0 ), bswap_shuf );
+  casti_v128( d, 1 ) = _mm_shuffle_epi8( casti_v128( s, 1 ), bswap_shuf );
+  casti_v128( d, 2 ) = _mm_shuffle_epi8( casti_v128( s, 2 ), bswap_shuf );
+  casti_v128( d, 3 ) = _mm_shuffle_epi8( casti_v128( s, 3 ), bswap_shuf );
+  casti_v128( d, 4 ) = _mm_shuffle_epi8( casti_v128( s, 4 ), bswap_shuf );
 }
 
 #elif defined(__aarch64__) && defined(__ARM_NEON)
@@ -461,11 +461,11 @@ static inline void v128_bswap32_80( void *d, void *s )
 
 static inline void v128_bswap32_intrlv80_4x32( void *d, const void *src )
 {
-  v128_t s0 = casti_m128i( src,0 );
-  v128_t s1 = casti_m128i( src,1 );
-  v128_t s2 = casti_m128i( src,2 );
-  v128_t s3 = casti_m128i( src,3 );
-  v128_t s4 = casti_m128i( src,4 );
+  v128_t s0 = casti_v128( src,0 );
+  v128_t s1 = casti_v128( src,1 );
+  v128_t s2 = casti_v128( src,2 );
+  v128_t s3 = casti_v128( src,3 );
+  v128_t s4 = casti_v128( src,4 );
 
 #if defined(__SSSE3__)
 
@@ -480,38 +480,38 @@ static inline void v128_bswap32_intrlv80_4x32( void *d, const void *src )
 
 #else
 
-  s0 = mm128_bswap_32( s0 );
-  s1 = mm128_bswap_32( s1 );
-  s2 = mm128_bswap_32( s2 );
-  s3 = mm128_bswap_32( s3 );
-  s4 = mm128_bswap_32( s4 );
+  s0 = v128_bswap32( s0 );
+  s1 = v128_bswap32( s1 );
+  s2 = v128_bswap32( s2 );
+  s3 = v128_bswap32( s3 );
+  s4 = v128_bswap32( s4 );
 
 #endif
 
-  casti_m128i( d, 0 ) = _mm_shuffle_epi32( s0, 0x00 );
-  casti_m128i( d, 1 ) = _mm_shuffle_epi32( s0, 0x55 );
-  casti_m128i( d, 2 ) = _mm_shuffle_epi32( s0, 0xaa );
-  casti_m128i( d, 3 ) = _mm_shuffle_epi32( s0, 0xff );
+  casti_v128( d, 0 ) = _mm_shuffle_epi32( s0, 0x00 );
+  casti_v128( d, 1 ) = _mm_shuffle_epi32( s0, 0x55 );
+  casti_v128( d, 2 ) = _mm_shuffle_epi32( s0, 0xaa );
+  casti_v128( d, 3 ) = _mm_shuffle_epi32( s0, 0xff );
 
-  casti_m128i( d, 4 ) = _mm_shuffle_epi32( s1, 0x00 );
-  casti_m128i( d, 5 ) = _mm_shuffle_epi32( s1, 0x55 );
-  casti_m128i( d, 6 ) = _mm_shuffle_epi32( s1, 0xaa );
-  casti_m128i( d, 7 ) = _mm_shuffle_epi32( s1, 0xff );
+  casti_v128( d, 4 ) = _mm_shuffle_epi32( s1, 0x00 );
+  casti_v128( d, 5 ) = _mm_shuffle_epi32( s1, 0x55 );
+  casti_v128( d, 6 ) = _mm_shuffle_epi32( s1, 0xaa );
+  casti_v128( d, 7 ) = _mm_shuffle_epi32( s1, 0xff );
 
-  casti_m128i( d, 8 ) = _mm_shuffle_epi32( s2, 0x00 );
-  casti_m128i( d, 9 ) = _mm_shuffle_epi32( s2, 0x55 );
-  casti_m128i( d,10 ) = _mm_shuffle_epi32( s2, 0xaa );
-  casti_m128i( d,11 ) = _mm_shuffle_epi32( s2, 0xff );
+  casti_v128( d, 8 ) = _mm_shuffle_epi32( s2, 0x00 );
+  casti_v128( d, 9 ) = _mm_shuffle_epi32( s2, 0x55 );
+  casti_v128( d,10 ) = _mm_shuffle_epi32( s2, 0xaa );
+  casti_v128( d,11 ) = _mm_shuffle_epi32( s2, 0xff );
 
-  casti_m128i( d,12 ) = _mm_shuffle_epi32( s3, 0x00 );
-  casti_m128i( d,13 ) = _mm_shuffle_epi32( s3, 0x55 );
-  casti_m128i( d,14 ) = _mm_shuffle_epi32( s3, 0xaa );
-  casti_m128i( d,15 ) = _mm_shuffle_epi32( s3, 0xff );
+  casti_v128( d,12 ) = _mm_shuffle_epi32( s3, 0x00 );
+  casti_v128( d,13 ) = _mm_shuffle_epi32( s3, 0x55 );
+  casti_v128( d,14 ) = _mm_shuffle_epi32( s3, 0xaa );
+  casti_v128( d,15 ) = _mm_shuffle_epi32( s3, 0xff );
 
-  casti_m128i( d,16 ) = _mm_shuffle_epi32( s4, 0x00 );
-  casti_m128i( d,17 ) = _mm_shuffle_epi32( s4, 0x55 );
-  casti_m128i( d,18 ) = _mm_shuffle_epi32( s4, 0xaa );
-  casti_m128i( d,19 ) = _mm_shuffle_epi32( s4, 0xff );
+  casti_v128( d,16 ) = _mm_shuffle_epi32( s4, 0x00 );
+  casti_v128( d,17 ) = _mm_shuffle_epi32( s4, 0x55 );
+  casti_v128( d,18 ) = _mm_shuffle_epi32( s4, 0xaa );
+  casti_v128( d,19 ) = _mm_shuffle_epi32( s4, 0xff );
 }
 
 #elif defined(__aarch64__) && defined(__ARM_NEON)
@@ -797,11 +797,11 @@ static inline void mm256_bswap32_intrlv80_8x32( void *d, const void *src )
    const __m256i c1 = v256_32( 0x04050607 );
    const __m256i c2 = v256_32( 0x08090a0b );
    const __m256i c3 = v256_32( 0x0c0d0e0f );
-   const v128_t s0 = casti_m128i( src,0 );
-   const v128_t s1 = casti_m128i( src,1 );
-   const v128_t s2 = casti_m128i( src,2 );
-   const v128_t s3 = casti_m128i( src,3 );
-   const v128_t s4 = casti_m128i( src,4 );
+   const v128_t s0 = casti_v128( src,0 );
+   const v128_t s1 = casti_v128( src,1 );
+   const v128_t s2 = casti_v128( src,2 );
+   const v128_t s3 = casti_v128( src,3 );
+   const v128_t s4 = casti_v128( src,4 );
 
    casti_m256i( d, 0 ) = _mm256_permutexvar_epi8( c0,
                           _mm256_castsi128_si256( s0 ) );
@@ -855,11 +855,11 @@ static inline void mm256_bswap32_intrlv80_8x32( void *d, const void *src )
   const __m256i c2 = _mm256_add_epi32( c1, c1 );
   const __m256i c3 = _mm256_add_epi32( c2, c1 );
 
-  v128_t s0 = casti_m128i( src,0 );
-  v128_t s1 = casti_m128i( src,1 );
-  v128_t s2 = casti_m128i( src,2 );
-  v128_t s3 = casti_m128i( src,3 );
-  v128_t s4 = casti_m128i( src,4 );
+  v128_t s0 = casti_v128( src,0 );
+  v128_t s1 = casti_v128( src,1 );
+  v128_t s2 = casti_v128( src,2 );
+  v128_t s3 = casti_v128( src,3 );
+  v128_t s4 = casti_v128( src,4 );
 
   s0 = _mm_shuffle_epi8( s0, bswap_shuf );
   s1 = _mm_shuffle_epi8( s1, bswap_shuf );
@@ -1303,11 +1303,11 @@ static inline void mm512_bswap32_intrlv80_16x32( void *d, const void *src )
    const __m512i c1 = v512_32( 0x04050607 );
    const __m512i c2 = v512_32( 0x08090a0b );
    const __m512i c3 = v512_32( 0x0c0d0e0f );
-   const v128_t s0 = casti_m128i( src,0 );
-   const v128_t s1 = casti_m128i( src,1 );
-   const v128_t s2 = casti_m128i( src,2 );
-   const v128_t s3 = casti_m128i( src,3 );
-   const v128_t s4 = casti_m128i( src,4 );
+   const v128_t s0 = casti_v128( src,0 );
+   const v128_t s1 = casti_v128( src,1 );
+   const v128_t s2 = casti_v128( src,2 );
+   const v128_t s3 = casti_v128( src,3 );
+   const v128_t s4 = casti_v128( src,4 );
  
    casti_m512i( d, 0 ) = _mm512_permutexvar_epi8( c0,
                           _mm512_castsi128_si512( s0 ) );
@@ -1360,11 +1360,11 @@ static inline void mm512_bswap32_intrlv80_16x32( void *d, const void *src )
   const __m512i c1 = v512_32( 1 );
   const __m512i c2 = _mm512_add_epi32( c1, c1 );
   const __m512i c3 = _mm512_add_epi32( c2, c1 );
-  v128_t s0 = casti_m128i( src,0 );
-  v128_t s1 = casti_m128i( src,1 );
-  v128_t s2 = casti_m128i( src,2 );
-  v128_t s3 = casti_m128i( src,3 );
-  v128_t s4 = casti_m128i( src,4 );
+  v128_t s0 = casti_v128( src,0 );
+  v128_t s1 = casti_v128( src,1 );
+  v128_t s2 = casti_v128( src,2 );
+  v128_t s3 = casti_v128( src,3 );
+  v128_t s4 = casti_v128( src,4 );
 
   s0 = _mm_shuffle_epi8( s0, bswap_shuf );
   s1 = _mm_shuffle_epi8( s1, bswap_shuf );
@@ -1492,20 +1492,20 @@ static inline void v128_bswap32_intrlv80_2x64( void *d, const void *src )
 
 #if defined(__SSE2__)
 
-  casti_m128i( d,0 ) = _mm_shuffle_epi32( s0, 0x44 );
-  casti_m128i( d,1 ) = _mm_shuffle_epi32( s0, 0xee );
+  casti_v128( d,0 ) = _mm_shuffle_epi32( s0, 0x44 );
+  casti_v128( d,1 ) = _mm_shuffle_epi32( s0, 0xee );
 
-  casti_m128i( d,2 ) = _mm_shuffle_epi32( s1, 0x44 );
-  casti_m128i( d,3 ) = _mm_shuffle_epi32( s1, 0xee );
+  casti_v128( d,2 ) = _mm_shuffle_epi32( s1, 0x44 );
+  casti_v128( d,3 ) = _mm_shuffle_epi32( s1, 0xee );
 
-  casti_m128i( d,4 ) = _mm_shuffle_epi32( s2, 0x44 );
-  casti_m128i( d,5 ) = _mm_shuffle_epi32( s2, 0xee );
+  casti_v128( d,4 ) = _mm_shuffle_epi32( s2, 0x44 );
+  casti_v128( d,5 ) = _mm_shuffle_epi32( s2, 0xee );
 
-  casti_m128i( d,6 ) = _mm_shuffle_epi32( s3, 0x44 );
-  casti_m128i( d,7 ) = _mm_shuffle_epi32( s3, 0xee );
+  casti_v128( d,6 ) = _mm_shuffle_epi32( s3, 0x44 );
+  casti_v128( d,7 ) = _mm_shuffle_epi32( s3, 0xee );
 
-  casti_m128i( d,8 ) = _mm_shuffle_epi32( s4, 0x44 );
-  casti_m128i( d,9 ) = _mm_shuffle_epi32( s4, 0xee );
+  casti_v128( d,8 ) = _mm_shuffle_epi32( s4, 0x44 );
+  casti_v128( d,9 ) = _mm_shuffle_epi32( s4, 0xee );
 
 #elif defined(__ARM_NEON)
 
@@ -1719,7 +1719,7 @@ static inline void mm256_intrlv80_4x64( void *d, const void *src )
 {
   __m256i s0 = casti_m256i( src,0 );
   __m256i s1 = casti_m256i( src,1 );
-  v128_t s4 = casti_m128i( src,4 );
+  v128_t s4 = casti_v128( src,4 );
 
   casti_m256i( d, 0 ) = _mm256_permute4x64_epi64( s0, 0x00 );
   casti_m256i( d, 1 ) = _mm256_permute4x64_epi64( s0, 0x55 );
@@ -1747,11 +1747,11 @@ static inline void mm256_bswap32_intrlv80_4x64( void *d, const void *src )
 {
    const __m256i c0 = v256_64( 0x0405060700010203 );
    const __m256i c1 = v256_64( 0x0c0d0e0f08090a0b );
-   const v128_t s0 = casti_m128i( src,0 );
-   const v128_t s1 = casti_m128i( src,1 );
-   const v128_t s2 = casti_m128i( src,2 );
-   const v128_t s3 = casti_m128i( src,3 );
-   const v128_t s4 = casti_m128i( src,4 );
+   const v128_t s0 = casti_v128( src,0 );
+   const v128_t s1 = casti_v128( src,1 );
+   const v128_t s2 = casti_v128( src,2 );
+   const v128_t s3 = casti_v128( src,3 );
+   const v128_t s4 = casti_v128( src,4 );
 
    casti_m256i( d,0 ) = _mm256_permutexvar_epi8( c0,
                          _mm256_castsi128_si256( s0 ) );
@@ -1783,7 +1783,7 @@ static inline void mm256_bswap32_intrlv80_4x64( void *d, const void *src )
                     _mm_set_epi64x( 0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
   __m256i s0 = casti_m256i( src,0 );
   __m256i s1 = casti_m256i( src,1 );
-  v128_t s4 = casti_m128i( src,4 );
+  v128_t s4 = casti_v128( src,4 );
 
   s0 = _mm256_shuffle_epi8( s0, bswap_shuf );
   s1 = _mm256_shuffle_epi8( s1, bswap_shuf );
@@ -2162,11 +2162,11 @@ static inline void mm512_bswap32_intrlv80_8x64( void *d, const void *src )
 {
    const __m512i c0 = v512_64( 0x0405060700010203 );
    const __m512i c1 = v512_64( 0x0c0d0e0f08090a0b );
-   const v128_t s0 = casti_m128i( src,0 );
-   const v128_t s1 = casti_m128i( src,1 );
-   const v128_t s2 = casti_m128i( src,2 );
-   const v128_t s3 = casti_m128i( src,3 );
-   const v128_t s4 = casti_m128i( src,4 );
+   const v128_t s0 = casti_v128( src,0 );
+   const v128_t s1 = casti_v128( src,1 );
+   const v128_t s2 = casti_v128( src,2 );
+   const v128_t s3 = casti_v128( src,3 );
+   const v128_t s4 = casti_v128( src,4 );
 
    casti_m512i( d,0 ) = _mm512_permutexvar_epi8( c0,
                          _mm512_castsi128_si512( s0 ) );
@@ -2197,11 +2197,11 @@ static inline void mm512_bswap32_intrlv80_8x64( void *d, const void *src )
   const v128_t bswap_shuf = _mm_set_epi64x( 0x0c0d0e0f08090a0b,
                                              0x0405060700010203 );
   const __m512i c1 = v512_64( 1 );
-  v128_t s0 = casti_m128i( src,0 );
-  v128_t s1 = casti_m128i( src,1 );
-  v128_t s2 = casti_m128i( src,2 );
-  v128_t s3 = casti_m128i( src,3 );
-  v128_t s4 = casti_m128i( src,4 );
+  v128_t s0 = casti_v128( src,0 );
+  v128_t s1 = casti_v128( src,1 );
+  v128_t s2 = casti_v128( src,2 );
+  v128_t s3 = casti_v128( src,3 );
+  v128_t s4 = casti_v128( src,4 );
 
   s0 = _mm_shuffle_epi8( s0, bswap_shuf );
   s1 = _mm_shuffle_epi8( s1, bswap_shuf );
@@ -2391,11 +2391,11 @@ static inline void mm512_bswap32_intrlv80_4x128( void *d, const void *src )
 {
   const __m512i bswap_shuf = mm512_bcast_m128(
                     _mm_set_epi64x( 0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
-  const v128_t s0 = casti_m128i( src,0 );
-  const v128_t s1 = casti_m128i( src,1 );
-  const v128_t s2 = casti_m128i( src,2 );
-  const v128_t s3 = casti_m128i( src,3 );
-  const v128_t s4 = casti_m128i( src,4 );
+  const v128_t s0 = casti_v128( src,0 );
+  const v128_t s1 = casti_v128( src,1 );
+  const v128_t s2 = casti_v128( src,2 );
+  const v128_t s3 = casti_v128( src,3 );
+  const v128_t s4 = casti_v128( src,4 );
 
   casti_m512i( d,0 ) = _mm512_permutexvar_epi8( _mm512_castsi128_si512( s0 ),
                                                  bswap_shuf );
@@ -2415,11 +2415,11 @@ static inline void mm512_bswap32_intrlv80_4x128( void *d, const void *src )
 {
   const v128_t bswap_shuf = _mm_set_epi64x( 0x0c0d0e0f08090a0b,
                                              0x0405060700010203 );
-  v128_t s0 = casti_m128i( src,0 );
-  v128_t s1 = casti_m128i( src,1 );
-  v128_t s2 = casti_m128i( src,2 );
-  v128_t s3 = casti_m128i( src,3 );
-  v128_t s4 = casti_m128i( src,4 );
+  v128_t s0 = casti_v128( src,0 );
+  v128_t s1 = casti_v128( src,1 );
+  v128_t s2 = casti_v128( src,2 );
+  v128_t s3 = casti_v128( src,3 );
+  v128_t s4 = casti_v128( src,4 );
 
   s0 = _mm_shuffle_epi8( s0, bswap_shuf );
   s1 = _mm_shuffle_epi8( s1, bswap_shuf );
@@ -2489,44 +2489,44 @@ static inline void rintrlv_4x64_4x32( void *dst, const void *src,
    const v128_t *s = (const v128_t*)src;
    v128_t *d = (v128_t*)dst;
 
-   d[ 0] = mm128_shuffle2_32( s[ 0], s[ 1], 0x88 );
-   d[ 1] = mm128_shuffle2_32( s[ 0], s[ 1], 0xdd );
-   d[ 2] = mm128_shuffle2_32( s[ 2], s[ 3], 0x88 );
-   d[ 3] = mm128_shuffle2_32( s[ 2], s[ 3], 0xdd );
-   d[ 4] = mm128_shuffle2_32( s[ 4], s[ 5], 0x88 );
-   d[ 5] = mm128_shuffle2_32( s[ 4], s[ 5], 0xdd );
-   d[ 6] = mm128_shuffle2_32( s[ 6], s[ 7], 0x88 );
-   d[ 7] = mm128_shuffle2_32( s[ 6], s[ 7], 0xdd );
+   d[ 0] = v128_shuffle2_32( s[ 0], s[ 1], 0x88 );
+   d[ 1] = v128_shuffle2_32( s[ 0], s[ 1], 0xdd );
+   d[ 2] = v128_shuffle2_32( s[ 2], s[ 3], 0x88 );
+   d[ 3] = v128_shuffle2_32( s[ 2], s[ 3], 0xdd );
+   d[ 4] = v128_shuffle2_32( s[ 4], s[ 5], 0x88 );
+   d[ 5] = v128_shuffle2_32( s[ 4], s[ 5], 0xdd );
+   d[ 6] = v128_shuffle2_32( s[ 6], s[ 7], 0x88 );
+   d[ 7] = v128_shuffle2_32( s[ 6], s[ 7], 0xdd );
 
    if ( bit_len <= 256 ) return;
 
-   d[ 8] = mm128_shuffle2_32( s[ 8], s[ 9], 0x88 );
-   d[ 9] = mm128_shuffle2_32( s[ 8], s[ 9], 0xdd );
-   d[10] = mm128_shuffle2_32( s[10], s[11], 0x88 );
-   d[11] = mm128_shuffle2_32( s[10], s[11], 0xdd );
-   d[12] = mm128_shuffle2_32( s[12], s[13], 0x88 );
-   d[13] = mm128_shuffle2_32( s[12], s[13], 0xdd );
-   d[14] = mm128_shuffle2_32( s[14], s[15], 0x88 );
-   d[15] = mm128_shuffle2_32( s[14], s[15], 0xdd );
+   d[ 8] = v128_shuffle2_32( s[ 8], s[ 9], 0x88 );
+   d[ 9] = v128_shuffle2_32( s[ 8], s[ 9], 0xdd );
+   d[10] = v128_shuffle2_32( s[10], s[11], 0x88 );
+   d[11] = v128_shuffle2_32( s[10], s[11], 0xdd );
+   d[12] = v128_shuffle2_32( s[12], s[13], 0x88 );
+   d[13] = v128_shuffle2_32( s[12], s[13], 0xdd );
+   d[14] = v128_shuffle2_32( s[14], s[15], 0x88 );
+   d[15] = v128_shuffle2_32( s[14], s[15], 0xdd );
 
    if ( bit_len <= 512 ) return;
 
-   d[16] = mm128_shuffle2_32( s[16], s[17], 0x88 );
-   d[17] = mm128_shuffle2_32( s[16], s[17], 0xdd );
-   d[18] = mm128_shuffle2_32( s[18], s[19], 0x88 );
-   d[19] = mm128_shuffle2_32( s[18], s[19], 0xdd );
-   d[20] = mm128_shuffle2_32( s[20], s[21], 0x88 );
-   d[21] = mm128_shuffle2_32( s[20], s[21], 0xdd );
-   d[22] = mm128_shuffle2_32( s[22], s[23], 0x88 );
-   d[23] = mm128_shuffle2_32( s[22], s[23], 0xdd );
-   d[24] = mm128_shuffle2_32( s[24], s[25], 0x88 );
-   d[25] = mm128_shuffle2_32( s[24], s[25], 0xdd );
-   d[26] = mm128_shuffle2_32( s[26], s[27], 0x88 );
-   d[27] = mm128_shuffle2_32( s[26], s[27], 0xdd );
-   d[28] = mm128_shuffle2_32( s[28], s[29], 0x88 );
-   d[29] = mm128_shuffle2_32( s[28], s[29], 0xdd );
-   d[30] = mm128_shuffle2_32( s[30], s[31], 0x88 );
-   d[31] = mm128_shuffle2_32( s[30], s[31], 0xdd );
+   d[16] = v128_shuffle2_32( s[16], s[17], 0x88 );
+   d[17] = v128_shuffle2_32( s[16], s[17], 0xdd );
+   d[18] = v128_shuffle2_32( s[18], s[19], 0x88 );
+   d[19] = v128_shuffle2_32( s[18], s[19], 0xdd );
+   d[20] = v128_shuffle2_32( s[20], s[21], 0x88 );
+   d[21] = v128_shuffle2_32( s[20], s[21], 0xdd );
+   d[22] = v128_shuffle2_32( s[22], s[23], 0x88 );
+   d[23] = v128_shuffle2_32( s[22], s[23], 0xdd );
+   d[24] = v128_shuffle2_32( s[24], s[25], 0x88 );
+   d[25] = v128_shuffle2_32( s[24], s[25], 0xdd );
+   d[26] = v128_shuffle2_32( s[26], s[27], 0x88 );
+   d[27] = v128_shuffle2_32( s[26], s[27], 0xdd );
+   d[28] = v128_shuffle2_32( s[28], s[29], 0x88 );
+   d[29] = v128_shuffle2_32( s[28], s[29], 0xdd );
+   d[30] = v128_shuffle2_32( s[30], s[31], 0x88 );
+   d[31] = v128_shuffle2_32( s[30], s[31], 0xdd );
 
 // if ( bit_len <= 1024 ) return;
 }
@@ -2537,77 +2537,77 @@ static inline void rintrlv_8x64_8x32( void *dst, const void *src,
    const v128_t *s = (const v128_t*)src;
    v128_t *d = (v128_t*)dst;
 
-   d[ 0] = mm128_shuffle2_32( s[ 0], s[ 1], 0x88 );
-   d[ 1] = mm128_shuffle2_32( s[ 2], s[ 3], 0x88 );
-   d[ 2] = mm128_shuffle2_32( s[ 0], s[ 1], 0xdd );
-   d[ 3] = mm128_shuffle2_32( s[ 2], s[ 3], 0xdd );
-   d[ 4] = mm128_shuffle2_32( s[ 4], s[ 5], 0x88 );
-   d[ 5] = mm128_shuffle2_32( s[ 6], s[ 7], 0x88 );
-   d[ 6] = mm128_shuffle2_32( s[ 4], s[ 5], 0xdd );
-   d[ 7] = mm128_shuffle2_32( s[ 6], s[ 7], 0xdd );
-   d[ 8] = mm128_shuffle2_32( s[ 8], s[ 9], 0x88 );
-   d[ 9] = mm128_shuffle2_32( s[10], s[11], 0x88 );
-   d[10] = mm128_shuffle2_32( s[ 8], s[ 9], 0xdd );
-   d[11] = mm128_shuffle2_32( s[10], s[11], 0xdd );
-   d[12] = mm128_shuffle2_32( s[12], s[13], 0x88 );
-   d[13] = mm128_shuffle2_32( s[14], s[15], 0x88 );
-   d[14] = mm128_shuffle2_32( s[12], s[13], 0xdd );
-   d[15] = mm128_shuffle2_32( s[14], s[15], 0xdd );
+   d[ 0] = v128_shuffle2_32( s[ 0], s[ 1], 0x88 );
+   d[ 1] = v128_shuffle2_32( s[ 2], s[ 3], 0x88 );
+   d[ 2] = v128_shuffle2_32( s[ 0], s[ 1], 0xdd );
+   d[ 3] = v128_shuffle2_32( s[ 2], s[ 3], 0xdd );
+   d[ 4] = v128_shuffle2_32( s[ 4], s[ 5], 0x88 );
+   d[ 5] = v128_shuffle2_32( s[ 6], s[ 7], 0x88 );
+   d[ 6] = v128_shuffle2_32( s[ 4], s[ 5], 0xdd );
+   d[ 7] = v128_shuffle2_32( s[ 6], s[ 7], 0xdd );
+   d[ 8] = v128_shuffle2_32( s[ 8], s[ 9], 0x88 );
+   d[ 9] = v128_shuffle2_32( s[10], s[11], 0x88 );
+   d[10] = v128_shuffle2_32( s[ 8], s[ 9], 0xdd );
+   d[11] = v128_shuffle2_32( s[10], s[11], 0xdd );
+   d[12] = v128_shuffle2_32( s[12], s[13], 0x88 );
+   d[13] = v128_shuffle2_32( s[14], s[15], 0x88 );
+   d[14] = v128_shuffle2_32( s[12], s[13], 0xdd );
+   d[15] = v128_shuffle2_32( s[14], s[15], 0xdd );
 
    if ( bit_len <= 256 ) return;
 
-   d[16] = mm128_shuffle2_32( s[16], s[17], 0x88 );
-   d[17] = mm128_shuffle2_32( s[18], s[19], 0x88 );
-   d[18] = mm128_shuffle2_32( s[16], s[17], 0xdd );
-   d[19] = mm128_shuffle2_32( s[18], s[19], 0xdd );
-   d[20] = mm128_shuffle2_32( s[20], s[21], 0x88 );
-   d[21] = mm128_shuffle2_32( s[22], s[23], 0x88 );
-   d[22] = mm128_shuffle2_32( s[20], s[21], 0xdd );
-   d[23] = mm128_shuffle2_32( s[22], s[23], 0xdd );
-   d[24] = mm128_shuffle2_32( s[24], s[25], 0x88 );
-   d[25] = mm128_shuffle2_32( s[26], s[27], 0x88 );
-   d[26] = mm128_shuffle2_32( s[24], s[25], 0xdd );
-   d[27] = mm128_shuffle2_32( s[26], s[27], 0xdd );
-   d[28] = mm128_shuffle2_32( s[28], s[29], 0x88 );
-   d[29] = mm128_shuffle2_32( s[30], s[31], 0x88 );
-   d[30] = mm128_shuffle2_32( s[28], s[29], 0xdd );
-   d[31] = mm128_shuffle2_32( s[30], s[31], 0xdd );
+   d[16] = v128_shuffle2_32( s[16], s[17], 0x88 );
+   d[17] = v128_shuffle2_32( s[18], s[19], 0x88 );
+   d[18] = v128_shuffle2_32( s[16], s[17], 0xdd );
+   d[19] = v128_shuffle2_32( s[18], s[19], 0xdd );
+   d[20] = v128_shuffle2_32( s[20], s[21], 0x88 );
+   d[21] = v128_shuffle2_32( s[22], s[23], 0x88 );
+   d[22] = v128_shuffle2_32( s[20], s[21], 0xdd );
+   d[23] = v128_shuffle2_32( s[22], s[23], 0xdd );
+   d[24] = v128_shuffle2_32( s[24], s[25], 0x88 );
+   d[25] = v128_shuffle2_32( s[26], s[27], 0x88 );
+   d[26] = v128_shuffle2_32( s[24], s[25], 0xdd );
+   d[27] = v128_shuffle2_32( s[26], s[27], 0xdd );
+   d[28] = v128_shuffle2_32( s[28], s[29], 0x88 );
+   d[29] = v128_shuffle2_32( s[30], s[31], 0x88 );
+   d[30] = v128_shuffle2_32( s[28], s[29], 0xdd );
+   d[31] = v128_shuffle2_32( s[30], s[31], 0xdd );
 
    if ( bit_len <= 512 ) return;
 
-   d[32] = mm128_shuffle2_32( s[32], s[33], 0x88 );
-   d[33] = mm128_shuffle2_32( s[34], s[35], 0x88 );
-   d[34] = mm128_shuffle2_32( s[32], s[33], 0xdd );
-   d[35] = mm128_shuffle2_32( s[34], s[35], 0xdd );
-   d[36] = mm128_shuffle2_32( s[36], s[37], 0x88 );
-   d[37] = mm128_shuffle2_32( s[38], s[39], 0x88 );
-   d[38] = mm128_shuffle2_32( s[36], s[37], 0xdd );
-   d[39] = mm128_shuffle2_32( s[38], s[39], 0xdd );
-   d[40] = mm128_shuffle2_32( s[40], s[41], 0x88 );
-   d[41] = mm128_shuffle2_32( s[42], s[43], 0x88 );
-   d[42] = mm128_shuffle2_32( s[40], s[41], 0xdd );
-   d[43] = mm128_shuffle2_32( s[42], s[43], 0xdd );
-   d[44] = mm128_shuffle2_32( s[44], s[45], 0x88 );
-   d[45] = mm128_shuffle2_32( s[46], s[47], 0x88 );
-   d[46] = mm128_shuffle2_32( s[44], s[45], 0xdd );
-   d[47] = mm128_shuffle2_32( s[46], s[47], 0xdd );
+   d[32] = v128_shuffle2_32( s[32], s[33], 0x88 );
+   d[33] = v128_shuffle2_32( s[34], s[35], 0x88 );
+   d[34] = v128_shuffle2_32( s[32], s[33], 0xdd );
+   d[35] = v128_shuffle2_32( s[34], s[35], 0xdd );
+   d[36] = v128_shuffle2_32( s[36], s[37], 0x88 );
+   d[37] = v128_shuffle2_32( s[38], s[39], 0x88 );
+   d[38] = v128_shuffle2_32( s[36], s[37], 0xdd );
+   d[39] = v128_shuffle2_32( s[38], s[39], 0xdd );
+   d[40] = v128_shuffle2_32( s[40], s[41], 0x88 );
+   d[41] = v128_shuffle2_32( s[42], s[43], 0x88 );
+   d[42] = v128_shuffle2_32( s[40], s[41], 0xdd );
+   d[43] = v128_shuffle2_32( s[42], s[43], 0xdd );
+   d[44] = v128_shuffle2_32( s[44], s[45], 0x88 );
+   d[45] = v128_shuffle2_32( s[46], s[47], 0x88 );
+   d[46] = v128_shuffle2_32( s[44], s[45], 0xdd );
+   d[47] = v128_shuffle2_32( s[46], s[47], 0xdd );
 
-   d[48] = mm128_shuffle2_32( s[48], s[49], 0x88 );
-   d[49] = mm128_shuffle2_32( s[50], s[51], 0x88 );
-   d[50] = mm128_shuffle2_32( s[48], s[49], 0xdd );
-   d[51] = mm128_shuffle2_32( s[50], s[51], 0xdd );
-   d[52] = mm128_shuffle2_32( s[52], s[53], 0x88 );
-   d[53] = mm128_shuffle2_32( s[54], s[55], 0x88 );
-   d[54] = mm128_shuffle2_32( s[52], s[53], 0xdd );
-   d[55] = mm128_shuffle2_32( s[54], s[55], 0xdd );
-   d[56] = mm128_shuffle2_32( s[56], s[57], 0x88 );
-   d[57] = mm128_shuffle2_32( s[58], s[59], 0x88 );
-   d[58] = mm128_shuffle2_32( s[56], s[57], 0xdd );
-   d[59] = mm128_shuffle2_32( s[58], s[59], 0xdd );
-   d[60] = mm128_shuffle2_32( s[60], s[61], 0x88 );
-   d[61] = mm128_shuffle2_32( s[62], s[63], 0x88 );
-   d[62] = mm128_shuffle2_32( s[60], s[61], 0xdd );
-   d[63] = mm128_shuffle2_32( s[62], s[63], 0xdd );
+   d[48] = v128_shuffle2_32( s[48], s[49], 0x88 );
+   d[49] = v128_shuffle2_32( s[50], s[51], 0x88 );
+   d[50] = v128_shuffle2_32( s[48], s[49], 0xdd );
+   d[51] = v128_shuffle2_32( s[50], s[51], 0xdd );
+   d[52] = v128_shuffle2_32( s[52], s[53], 0x88 );
+   d[53] = v128_shuffle2_32( s[54], s[55], 0x88 );
+   d[54] = v128_shuffle2_32( s[52], s[53], 0xdd );
+   d[55] = v128_shuffle2_32( s[54], s[55], 0xdd );
+   d[56] = v128_shuffle2_32( s[56], s[57], 0x88 );
+   d[57] = v128_shuffle2_32( s[58], s[59], 0x88 );
+   d[58] = v128_shuffle2_32( s[56], s[57], 0xdd );
+   d[59] = v128_shuffle2_32( s[58], s[59], 0xdd );
+   d[60] = v128_shuffle2_32( s[60], s[61], 0x88 );
+   d[61] = v128_shuffle2_32( s[62], s[63], 0x88 );
+   d[62] = v128_shuffle2_32( s[60], s[61], 0xdd );
+   d[63] = v128_shuffle2_32( s[62], s[63], 0xdd );
 
 // if ( bit_len <= 1024 ) return;
 }
@@ -3248,12 +3248,21 @@ static inline void rintrlv_2x256_8x64( void *dst, const void *src0,
 
 // blend 2 vectors while interleaving: { hi[n], lo[n-1], ... hi[1], lo[0] }
 #if defined(__SSE4_1__)
-// No SSE2 implementation.
 
-//#define mm128_intrlv_blend_64( hi, lo )   _mm_blend_epi16( hi, lo, 0x0f )
-//#define mm128_intrlv_blend_32( hi, lo )   _mm_blend_epi16( hi, lo, 0x33 )
+#define v128_intrlv_blend_64( hi, lo )   _mm_blend_epi16( hi, lo, 0x0f )
+#define v128_intrlv_blend_32( hi, lo )   _mm_blend_epi16( hi, lo, 0x33 )
 
-#endif   // SSE4_1
+#elif defined(__SSE2__) || defined(__ARM_NEON)
+
+#define v128_intrlv_blend_64( hi, lo )  \
+   v128_blendv( hi, lo, v128_set64( 0ull, 0xffffffffffffffffull ) )
+
+#define v128_intrlv_blend_32( hi, lo ) \
+   v128_blendv( hi, lo, v128_set64( 0xffffffffull, 0xffffffffull ) )
+
+#else
+// unknown, unsupported architecture
+#endif
 
 #if defined(__AVX2__)
 

simd-utils/simd-128.h

@@ -35,17 +35,17 @@
 ///////////////////////////////////////////////////////////////////////////////
 
 // New architecturally agnostic syntax: 
-//    All users of 128 bit SIMD should use new syntax or protect SSE2 only
-//    code segments.
-//    Other vector sizes continue with old syntax for now.
-//    Definitionns here will gradually be converted to new synytax.
-//    For consistency the larger vector utilities should do the same.
-
+//
+//           __m128i -> v128_t
+//           _mm_    -> v128_
+//           mm128_  -> v128_
+// 
+//    There is also new syntax to accomodate ARM's stricter type checking of
+//    vector element size. They have no effect on x86_64.
 
 // direct translation of native intrinsics
 
 #define v128_t                         __m128i 
-// Needed for ARM
 #define v128u64_t                      v128_t
 #define v128u32_t                      v128_t
 #define v128u16_t                      v128_t
@@ -56,17 +56,15 @@
 
 // Needed for ARM, Doesn't do anything special on x86_64
 #define v128_load1_64(p)               _mm_set1_epi64x(*(uint64_t*)(p) )
-#define v128_load1_32(p)               _mm_set_epi32(  *(uint32_t*)(p) )
-#define v128_load1_16(p)               _mm_set_epi16(  *(uint16_t*)(p) )
-#define v128_load1_8( p)               _mm_set_epi8(   *(uint8_t*) (p) )
+#define v128_load1_32(p)               _mm_set1_epi32( *(uint32_t*)(p) )
+#define v128_load1_16(p)               _mm_set1_epi16( *(uint16_t*)(p) )
+#define v128_load1_8( p)               _mm_set1_epi8(  *(uint8_t*) (p) )
 
 // arithmetic
 #define v128_add64                     _mm_add_epi64
 #define v128_add32                     _mm_add_epi32
 #define v128_add16                     _mm_add_epi16
 #define v128_add8                      _mm_add_epi8
-#define v128_add4_64                   mm128_add4_64
-#define v128_add4_32                   mm128_add4_32
 
 #define v128_sub64                     _mm_sub_epi64
 #define v128_sub32                     _mm_sub_epi32
@@ -82,7 +80,7 @@
 #define v128_mulw32                    _mm_mul_epu32
 #define v128_mulw16                    _mm_mul_epu16
 
-// compare
+// signed compare
 #define v128_cmpeq64                   _mm_cmpeq_epi64
 #define v128_cmpeq32                   _mm_cmpeq_epi32
 #define v128_cmpeq16                   _mm_cmpeq_epi16
@@ -120,27 +118,6 @@
 #define v128_xor                       _mm_xor_si128
 #define v128_xorq                      _mm_xor_si128
 #define v128_andnot                    _mm_andnot_si128
-#define v128_xnor( a, b )              mm128_not( _mm_xor_si128( a, b ) )
-#define v128_ornot( a, b )             mm128_or( a, mm128_not( b ) ) 
-
-// ternary
-#define v128_xorandnot( v2, v1, v0 ) \
-                               _mm_xor_si128( v2, _mm_andnot_si128( v1, v0 ) )
-#define v128_xor3( v2, v1, v0 ) \
-                               _mm_xor_si128( v2, _mm_xor_si128( v1, v0 ) )
-#define v128_and3( a, b, c )           _mm_and_si128( a, _mm_and_si128( b, c ) )
-#define v128_or3( a, b, c )            _mm_or_si128( a, _mm_or_si128( b, c ) )
-#define v128_xorand( a, b, c )         _mm_xor_si128( a, _mm_and_si128( b, c ) )
-#define v128_andxor( a, b, c )         _mm_and_si128( a, _mm_xor_si128( b, c ))
-#define v128_xoror( a, b, c )          _mm_xor_si128( a, _mm_or_si128( b, c ) )
-#define v128_orand( a, b, c )          _mm_or_si128( a, _mm_and_si128( b, c ) )
-
-// shift 2 concatenated vectors right
-#define v128_alignr64                  mm128_alignr_64
-#define v128_alignr32                  mm128_alignr_32
-#if defined(__SSSE3__)
-  #define v128_alignr8                 _mm_alignr_epi8
-#endif
 
 // unpack
 #define v128_unpacklo64                _mm_unpacklo_epi64
@@ -243,24 +220,22 @@ static inline __m128i mm128_mov32_128( const uint32_t n )
 
 // Pseudo constants
 #define v128_zero                       _mm_setzero_si128()
-#define m128_zero                       _mm_setzero_si128()
-
 
 #if defined(__SSE4_1__)
 
 // Bitwise AND, return 1 if result is all bits clear.
-#define v128_and_eq0                    _mm_testz_si128
+#define v128_and_eq0(v1, v0)            _mm_testz_si128(v1, v0)
 
+// v128_is_zero?
 static inline int v128_cmpeq0( v128_t v )
 {  return v128_and_eq0( v, v ); }
 
 #endif
 
 // Bitwise compare return 1 if all bits set.
-#define v128_cmpeq1                      _mm_test_all ones
+#define v128_cmpeq1(v)                   _mm_test_all ones(v)
 
-#define v128_one                         mm128_mov64_128( 1 )
-#define m128_one_128                     v128_one
+#define v128_one                         mm128_mov64_128(1)
 
 // ASM avoids the need to initialize return variable to avoid compiler warning.
 // Macro abstracts function parentheses to look like an identifier.
@@ -274,17 +249,14 @@ static inline __m128i v128_neg1_fn()
 #endif
    return a;
 }
-#define m128_neg1_fn                     v128_neg1_fn
 #define v128_neg1                        v128_neg1_fn()
-#define m128_neg1                        v128_neg1
 
 //
 // Vector pointer cast
 
 // p = any aligned pointer
 // returns p as pointer to vector type
-#define castp_m128i(p)    ((__m128i*)(p))
-#define castp_v128        castp_m128i
+#define castp_v128(p)     ((__m128i*)(p))
 #define castp_v128u64     castp_v128
 #define castp_v128u32     castp_v128
 #define castp_v128u16     castp_v128
@@ -292,8 +264,7 @@ static inline __m128i v128_neg1_fn()
 
 // p = any aligned pointer
 // returns *p, watch your pointer arithmetic
-#define cast_m128i(p)     (*((__m128i*)(p)))
-#define cast_v128         cast_m128i
+#define cast_v128(p)      (*((__m128i*)(p)))
 #define cast_v128u64      cast_v128
 #define cast_v128u32      cast_v128
 #define cast_v128u16      cast_v128
@@ -301,8 +272,8 @@ static inline __m128i v128_neg1_fn()
 
 // p = any aligned pointer, i = scaled array index
 // returns value p[i]
-#define casti_m128i(p,i)   (((__m128i*)(p))[(i)])
-#define casti_v128         casti_m128i
+#define casti_v128(p,i)    (((__m128i*)(p))[(i)])
+#define casti_m128i        casti_v128     // deprecated
 #define casti_v128u64      casti_v128
 #define casti_v128u32      casti_v128
 #define casti_v128u16      casti_v128
@@ -310,7 +281,7 @@ static inline __m128i v128_neg1_fn()
 
 // p = any aligned pointer, o = scaled offset
 // returns pointer p+o
-#define casto_m128i(p,o) (((__m128i*)(p))+(o))
+#define casto_v128(p,o) (((__m128i*)(p))+(o))
 
 #if defined(__SSE4_1__)
 #define v128_get64( v, l )         _mm_extract_epi64( v, l )
@@ -325,7 +296,7 @@ static inline __m128i v128_neg1_fn()
 
 /////////////////////////////////////////////////////////////
 //
-//      _mm_insert_ps( _mm128i v1, __m128i v2, imm8 c )
+//      _mm_insert_ps( __m128i v1, __m128i v2, imm8 c )
 //
 // Fast and powerful but very limited in its application.
 // It requires SSE4.1 but only works with 128 bit vectors with 32 bit
@@ -380,115 +351,112 @@ static inline __m128i v128_neg1_fn()
 #if defined(__AVX512VL__)
 //TODO Enable for AVX10_256
 
-static inline __m128i mm128_not( const __m128i v )
+static inline __m128i v128_not( const __m128i v )
 {  return _mm_ternarylogic_epi64( v, v, v, 1 ); }
 
 #else
 
-#define mm128_not( v )          _mm_xor_si128( v, m128_neg1 ) 
+#define v128_not( v )          _mm_xor_si128( v, v128_neg1 ) 
 
 #endif
-#define v128_not                mm128_not
 
-static inline __m128i mm128_negate_64( __m128i v )
+static inline v128u64_t v128_negate_64( v128u64_t v )
 { return _mm_sub_epi64( _mm_xor_si128( v, v ), v ); }
-#define v128_negate64                 mm128_negate_64
 
-static inline __m128i mm128_negate_32( __m128i v )
+static inline v128u32_t v128_negate_32( v128u32_t v )
 { return _mm_sub_epi32( _mm_xor_si128( v, v ), v ); }
-#define v128_negate32                 mm128_negate_32
 
-static inline __m128i mm128_negate_16( __m128i v ) 
+static inline v128u16_t v128_negate_16( v128u16_t v ) 
 { return _mm_sub_epi16( _mm_xor_si128( v, v ), v ); }
-#define v128_negate16                 mm128_negate_16
 
 
 // Add 4 values, fewer dependencies than sequential addition.
-#define mm128_add4_64( a, b, c, d ) \
+#define v128_add4_64( a, b, c, d ) \
    _mm_add_epi64( _mm_add_epi64( a, b ), _mm_add_epi64( c, d ) )
 
-#define mm128_add4_32( a, b, c, d ) \
+#define v128_add4_32( a, b, c, d ) \
    _mm_add_epi32( _mm_add_epi32( a, b ), _mm_add_epi32( c, d ) )
-#define v128_add4_32                   mm128_add4_32
 
-#define mm128_add4_16( a, b, c, d ) \
+#define v128_add4_16( a, b, c, d ) \
    _mm_add_epi16( _mm_add_epi16( a, b ), _mm_add_epi16( c, d ) )
 
-#define mm128_add4_8( a, b, c, d ) \
+#define v128_add4_8( a, b, c, d ) \
    _mm_add_epi8( _mm_add_epi8( a, b ), _mm_add_epi8( c, d ) )
 
-#define mm128_xor4( a, b, c, d ) \
+#define v128_xor4( a, b, c, d ) \
    _mm_xor_si128( _mm_xor_si128( a, b ), _mm_xor_si128( c, d ) )
 
+
 // Memory functions
 // Mostly for convenience, avoids calculating bytes.
 // Assumes data is alinged and integral.
 // n = number of __m128i, bytes/16
 
-static inline void memset_zero_128( __m128i *dst,  const int n )
-{   for ( int i = 0; i < n; i++ ) dst[i] = m128_zero; }
-#define v128_memset_zero               memset_zero_128
+static inline void v128_memset_zero( v128_t *dst,  const int n )
+{   for ( int i = 0; i < n; i++ ) dst[i] = v128_zero; }
+#define memset_zero_128      v128_memset_zero
 
-static inline void memset_128( __m128i *dst, const __m128i a, const int n )
+static inline void v128_memset( v128_t *dst, const v128_t a, const int n )
 {   for ( int i = 0; i < n; i++ ) dst[i] = a; }
-#define v128_memset                    memset_128
 
-static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
+static inline void v128_memcpy( v128_t *dst, const v128_t *src, const int n )
 {   for ( int i = 0; i < n; i ++ ) dst[i] = src[i]; }
-#define v128_memcpy                    memcpy_128
+#define  memcpy_128           v128_memcpy  
 
 #if defined(__AVX512VL__)
 //TODO Enable for AVX10_256
 
 // a ^ b ^ c
-#define mm128_xor3( a, b, c )      _mm_ternarylogic_epi64( a, b, c, 0x96 )
+#define v128_xor3( a, b, c )      _mm_ternarylogic_epi64( a, b, c, 0x96 )
 
 // a & b & c
-#define mm128_and3( a, b, c )      _mm_ternarylogic_epi64( a, b, c, 0x80 )
+#define v128_and3( a, b, c )      _mm_ternarylogic_epi64( a, b, c, 0x80 )
 
 // a | b | c
-#define mm128_or3( a, b, c )       _mm_ternarylogic_epi64( a, b, c, 0xfe )
+#define v128_or3( a, b, c )       _mm_ternarylogic_epi64( a, b, c, 0xfe )
 
 // a ^ ( b & c )
-#define mm128_xorand( a, b, c )    _mm_ternarylogic_epi64( a, b, c, 0x78 )
+#define v128_xorand( a, b, c )    _mm_ternarylogic_epi64( a, b, c, 0x78 )
 
 // a & ( b ^ c )
-#define mm128_andxor( a, b, c )    _mm_ternarylogic_epi64( a, b, c, 0x60 )
+#define v128_andxor( a, b, c )    _mm_ternarylogic_epi64( a, b, c, 0x60 )
 
 // a ^ ( b | c )
-#define mm128_xoror( a, b, c )     _mm_ternarylogic_epi64( a, b, c, 0x1e )
+#define v128_xoror( a, b, c )     _mm_ternarylogic_epi64( a, b, c, 0x1e )
 
 // a ^ ( ~b & c )
-#define mm128_xorandnot( a, b, c ) _mm_ternarylogic_epi64( a, b, c, 0xd2 )
+#define v128_xorandnot( a, b, c ) _mm_ternarylogic_epi64( a, b, c, 0xd2 )
 
 // a | ( b & c )
-#define mm128_orand( a, b, c )     _mm_ternarylogic_epi64( a, b, c, 0xf8 )
+#define v128_orand( a, b, c )     _mm_ternarylogic_epi64( a, b, c, 0xf8 )
 
 // ~( a ^ b ), same as (~a) ^ b
-#define mm128_xnor( a, b )         _mm_ternarylogic_epi64( a, b, b, 0x81 )
+#define v128_xnor( a, b )         _mm_ternarylogic_epi64( a, b, b, 0x81 )
 
 #else
 
-#define mm128_xor3( a, b, c )      _mm_xor_si128( a, _mm_xor_si128( b, c ) )
+#define v128_xor3( a, b, c )      _mm_xor_si128( a, _mm_xor_si128( b, c ) )
 
-#define mm128_and3( a, b, c )      _mm_and_si128( a, _mm_and_si128( b, c ) )
+#define v128_and3( a, b, c )      _mm_and_si128( a, _mm_and_si128( b, c ) )
 
-#define mm128_or3( a, b, c )       _mm_or_si128( a, _mm_or_si128( b, c ) )
+#define v128_or3( a, b, c )       _mm_or_si128( a, _mm_or_si128( b, c ) )
 
-#define mm128_xorand( a, b, c )    _mm_xor_si128( a, _mm_and_si128( b, c ) )
+#define v128_xorand( a, b, c )    _mm_xor_si128( a, _mm_and_si128( b, c ) )
 
-#define mm128_andxor( a, b, c )    _mm_and_si128( a, _mm_xor_si128( b, c ))
+#define v128_andxor( a, b, c )    _mm_and_si128( a, _mm_xor_si128( b, c ))
 
-#define mm128_xoror( a, b, c )     _mm_xor_si128( a, _mm_or_si128( b, c ) )
+#define v128_xoror( a, b, c )     _mm_xor_si128( a, _mm_or_si128( b, c ) )
 
-#define mm128_xorandnot( a, b, c ) _mm_xor_si128( a, _mm_andnot_si128( b, c ) )
+#define v128_xorandnot( a, b, c ) _mm_xor_si128( a, _mm_andnot_si128( b, c ) )
 
-#define mm128_orand( a, b, c )     _mm_or_si128( a, _mm_and_si128( b, c ) )
+#define v128_orand( a, b, c )     _mm_or_si128( a, _mm_and_si128( b, c ) )
 
-#define mm128_xnor( a, b )         mm128_not( _mm_xor_si128( a, b ) )
+#define v128_xnor( a, b )         mm128_not( _mm_xor_si128( a, b ) )
 
 #endif
 
+#define v128_ornot( a, b )             _mm_or_si128( a, v128_not( b ) )
+
 // Mask making
 // Equivalent of AVX512 _mm_movepi64_mask & _mm_movepi32_mask.
 // Returns 2 or 4 bit integer mask from MSBit of 64 or 32 bit elements.
@@ -514,7 +482,7 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
 #define v128_qrev16(v)      v128_shuffle16( v, 0x1b )
 #define v128_lrev16(v)      v128_shuffle16( v, 0xb1 )
 
-// These sgould never be callled from application code, use rol/ror.
+// These should never be callled from application code, use rol/ror.
 #define v128_ror64_sse2( v, c ) \
    _mm_or_si128( _mm_srli_epi64( v, c ), _mm_slli_epi64( v, 64-(c) ) )
 
@@ -530,12 +498,12 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
 #if defined(__AVX512VL__)
 
 // AVX512 fastest all rotations.
-#define mm128_ror_64                _mm_ror_epi64
-#define mm128_rol_64                _mm_rol_epi64
-#define mm128_ror_32                _mm_ror_epi32
-#define mm128_rol_32                _mm_rol_epi32
+#define v128_ror64                _mm_ror_epi64
+#define v128_rol64                _mm_rol_epi64
+#define v128_ror32                _mm_ror_epi32
+#define v128_rol32                _mm_rol_epi32
 
-// ror/rol will alway find the fastest but these names may fit better with
+// ror/rol will always find the fastest but these names may fit better with
 // application code performing shuffles rather than bit rotations.
 #define v128_shuflr64_8( v)         _mm_ror_epi64( v,  8 )
 #define v128_shufll64_8( v)         _mm_rol_epi64( v,  8 )
@@ -549,7 +517,7 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
 #define v128_shufll32_16(v)         _mm_rol_epi32( v, 16 )
 
 #elif defined(__SSSE3__)
-// SSE2: fastest 32 bit, very fast 16, fast 8
+// SSSE3: fastest 32 bit, very fast 16, fast 8
 
 #define v128_shuflr64_8( v ) \
     _mm_shuffle_epi8( v, _mm_set_epi64x( \
@@ -575,7 +543,7 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
    _mm_shuffle_epi8( v, _mm_set_epi64x( \
                                   0x0e0d0c0f0a09080b, 0x0605040702010003 ) )
 
-#define mm128_ror_64( v, c ) \
+#define v128_ror64( v, c ) \
    ( (c) ==  8 ) ? v128_shuflr64_8( v ) \
  : ( (c) == 16 ) ? v128_shuffle16( v, 0x39 ) \
  : ( (c) == 24 ) ? v128_shuflr64_24( v ) \
@@ -585,7 +553,7 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
  : ( (c) == 56 ) ? v128_shufll64_8( v ) \
  : v128_ror64_sse2( v, c ) 
 
-#define mm128_rol_64( v, c ) \
+#define v128_rol64( v, c ) \
    ( (c) ==  8 ) ? v128_shufll64_8( v ) \
  : ( (c) == 16 ) ? v128_shuffle16( v, 0x93 ) \
  : ( (c) == 24 ) ? v128_shufll64_24( v ) \
@@ -595,13 +563,13 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
  : ( (c) == 56 ) ? v128_shuflr64_8( v ) \
  : v128_rol64_sse2( v, c ) 
 
-#define mm128_ror_32( v, c ) \
+#define v128_ror32( v, c ) \
    ( (c) ==  8 ) ? v128_shuflr32_8( v ) \
  : ( (c) == 16 ) ? v128_lrev16( v ) \
  : ( (c) == 24 ) ? v128_shufll32_8( v ) \
  : v128_ror32_sse2( v, c ) 
 
-#define mm128_rol_32( v, c ) \
+#define v128_rol32( v, c ) \
    ( (c) ==  8 ) ? v128_shufll32_8( v ) \
  : ( (c) == 16 ) ? v128_lrev16( v ) \
  : ( (c) == 24 ) ? v128_shuflr32_8( v ) \
@@ -610,42 +578,41 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
 #elif defined(__SSE2__)
 // SSE2: fastest 32 bit, very fast 16
 
-#define mm128_ror_64( v, c ) \
+#define v128_ror64( v, c ) \
    ( (c) == 16 ) ? v128_shuffle16( v, 0x39 ) \
  : ( (c) == 32 ) ? _mm_shuffle_epi32( v, 0xb1 ) \
  : ( (c) == 48 ) ? v128_shuffle16( v, 0x93 ) \
  : v128_ror64_sse2( v, c )
 
-#define mm128_rol_64( v, c ) \
+#define v128_rol64( v, c ) \
    ( (c) == 16 ) ? v128_shuffle16( v, 0x93 ) \
  : ( (c) == 32 ) ? _mm_shuffle_epi32( v, 0xb1 ) \
  : ( (c) == 48 ) ? v128_shuffle16( v, 0x39 ) \
  : v128_rol64_sse2( v, c )
 
-#define mm128_ror_32( v, c ) \
+#define v128_ror32( v, c ) \
   ( (c) == 16 ) ? v128_lrev16( v ) \
  : v128_ror32_sse2( v, c )
 
-#define mm128_rol_32( v, c ) \
+#define v128_rol32( v, c ) \
   ( (c) == 16 ) ? v128_lrev16( v ) \
  : v128_rol32_sse2( v, c )
 
 #else 
 
-#define mm128_ror_64         v128_ror64_sse2
-#define mm128_rol_64         v128_rol64_sse2
-#define mm128_ror_32         v128_ror32_sse2
-#define mm128_rol_32         v128_rol32_sse2
+#define v128_ror64         v128_ror64_sse2
+#define v128_rol64         v128_rol64_sse2
+#define v128_ror32         v128_ror32_sse2
+#define v128_rol32         v128_rol32_sse2
 
 #endif
 
-// Generic names for portable code
-#define v128_ror64            mm128_ror_64
-#define v128_rol64            mm128_rol_64
-#define v128_ror32            mm128_ror_32
-#define v128_rol32            mm128_rol_32
-
+//#define v128_ror64            mm128_ror_64
+//#define v128_rol64            mm128_rol_64
+//#define v128_ror32            mm128_ror_32
+#define mm128_rol_32        v128_rol32
 
+/* not used
 // x2 rotates elements in 2 individual vectors in a double buffered
 // optimization for SSE2, does nothing for AVX512 but is there for
 // transparency.
@@ -653,25 +620,25 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
 #if defined(__AVX512VL__)
 //TODO Enable for AVX10_256
 
-#define mm128_rorx2_64( v1, v0, c ) \
+#define v128_2ror64( v1, v0, c ) \
    _mm_ror_epi64( v0, c ); \
    _mm_ror_epi64( v1, c )
 
-#define mm128_rolx2_64( v1, v0, c ) \
+#define v128_2rol64( v1, v0, c ) \
    _mm_rol_epi64( v0, c ); \
    _mm_rol_epi64( v1, c )
 
-#define mm128_rorx2_32( v1, v0, c ) \
+#define v128_2ror32( v1, v0, c ) \
    _mm_ror_epi32( v0, c ); \
    _mm_ror_epi32( v1, c )
 
-#define mm128_rolx2_32( v1, v0, c ) \
+#define mm128_2rol32( v1, v0, c ) \
    _mm_rol_epi32( v0, c ); \
    _mm_rol_epi32( v1, c )
 
 #else  // SSE2
 
-#define mm128_rorx2_64( v1, v0, c ) \
+#define v128_2ror64( v1, v0, c ) \
 { \
  __m128i t0 = _mm_srli_epi64( v0, c ); \
  __m128i t1 = _mm_srli_epi64( v1, c ); \
@@ -681,7 +648,7 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
  v1 = _mm_or_si256( v1, t1 ); \
 }
 
-#define mm128_rolx2_64( v1, v0, c ) \
+#define v128_2rol64( v1, v0, c ) \
 { \
  __m128i t0 = _mm_slli_epi64( v0, c ); \
  __m128i t1 = _mm_slli_epi64( v1, c ); \
@@ -691,7 +658,7 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
  v1 = _mm_or_si256( v1, t1 ); \
 }
 
-#define mm128_rorx2_32( v1, v0, c ) \
+#define v128_2ror32( v1, v0, c ) \
 { \
  __m128i t0 = _mm_srli_epi32( v0, c ); \
  __m128i t1 = _mm_srli_epi32( v1, c ); \
@@ -701,7 +668,7 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
  v1 = _mm_or_si256( v1, t1 ); \
 }
 
-#define mm128_rolx2_32( v1, v0, c ) \
+#define v128_2rol32( v1, v0, c ) \
 { \
  __m128i t0 = _mm_slli_epi32( v0, c ); \
  __m128i t1 = _mm_slli_epi32( v1, c ); \
@@ -712,12 +679,7 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
 }
 
 #endif   // AVX512 else SSE2
-
-#define v128_2ror64            mm128_rorx2_64
-#define v128_2rol64            mm128_rolx2_64
-#define v128_2ror32            mm128_rorx2_32
-#define v128_2rol32            mm128_rolx2_32
-
+*/
 
 // Cross lane shuffles
 
@@ -756,95 +718,76 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
 #define v128_shuflr16(v)    v128_shuffle16( v, 0x39 )
 #define v128_shufll16(v)    v128_shuffle16( v, 0x93 )
 
-
-//TODO fix this
-// alias bswap
-//#define v128_qrev8(v)       _mm_shuffle_epi8( v, v128_8( 0,1,2,3,4,5,6,7 ) )
-//#define v128_lrev8(v)       _mm_shuffle_epi8( v, v128_8( 4,5,6,7, 0,1,2,3 ) )
-//#define v128_wrev8(v)       _mm_shuffle_epi8( v, v128_8( 6,7, 4,5, 2,3, 1,0 ) )
-   
-// reverse bits, can it be done?
-//#define v128_bitrev8( v )              vrbitq_u8
-
-/* Not used
-#if defined(__SSSE3__)
-
-// Rotate right by c bytes, no SSE2 equivalent.
-static inline __m128i mm128_shuflr_x8( const __m128i v, const int c )
-{ return _mm_alignr_epi8( v, v, c ); }
-
-#endif
-*/
-
 // Endian byte swap.
 
 #if defined(__SSSE3__)
 
-#define mm128_bswap_128( v ) \
+#define v128_bswap128( v ) \
    _mm_shuffle_epi8( v, _mm_set_epi64x( 0x0001020304050607, \
                                         0x08090a0b0c0d0e0f ) )
 
-#define mm128_bswap_64( v ) \
+#define v128_bswap64( v ) \
    _mm_shuffle_epi8( v, _mm_set_epi64x( 0x08090a0b0c0d0e0f, \
                                         0x0001020304050607 ) )
 
-#define mm128_bswap_32( v ) \
+#define v128_bswap32( v ) \
    _mm_shuffle_epi8( v, _mm_set_epi64x( 0x0c0d0e0f08090a0b, \
                                         0x0405060700010203 ) )
+#define mm128_bswap_32      v128_bswap32
 
-#define mm128_bswap_16( v ) \
+#define v128_bswap16( v ) \
    _mm_shuffle_epi8( v, _mm_set_epi64x( 0x0e0f0c0d0a0b0809, \
                                         0x0607040502030001 )
 
 // 8 byte qword * 8 qwords * 2 lanes = 128 bytes
 #define mm128_block_bswap_64( d, s ) \
 { \
-   __m128i ctl = _mm_set_epi64x(  0x08090a0b0c0d0e0f, 0x0001020304050607 ); \
-  casti_m128i( d,0 ) = _mm_shuffle_epi8( casti_m128i( s,0 ), ctl ); \
-  casti_m128i( d,1 ) = _mm_shuffle_epi8( casti_m128i( s,1 ), ctl ); \
-  casti_m128i( d,2 ) = _mm_shuffle_epi8( casti_m128i( s,2 ), ctl ); \
-  casti_m128i( d,3 ) = _mm_shuffle_epi8( casti_m128i( s,3 ), ctl ); \
-  casti_m128i( d,4 ) = _mm_shuffle_epi8( casti_m128i( s,4 ), ctl ); \
-  casti_m128i( d,5 ) = _mm_shuffle_epi8( casti_m128i( s,5 ), ctl ); \
-  casti_m128i( d,6 ) = _mm_shuffle_epi8( casti_m128i( s,6 ), ctl ); \
-  casti_m128i( d,7 ) = _mm_shuffle_epi8( casti_m128i( s,7 ), ctl ); \
+  v128_t ctl = _mm_set_epi64x(  0x08090a0b0c0d0e0f, 0x0001020304050607 ); \
+  casti_v128( d,0 ) = _mm_shuffle_epi8( casti_v128( s,0 ), ctl ); \
+  casti_v128( d,1 ) = _mm_shuffle_epi8( casti_v128( s,1 ), ctl ); \
+  casti_v128( d,2 ) = _mm_shuffle_epi8( casti_v128( s,2 ), ctl ); \
+  casti_v128( d,3 ) = _mm_shuffle_epi8( casti_v128( s,3 ), ctl ); \
+  casti_v128( d,4 ) = _mm_shuffle_epi8( casti_v128( s,4 ), ctl ); \
+  casti_v128( d,5 ) = _mm_shuffle_epi8( casti_v128( s,5 ), ctl ); \
+  casti_v128( d,6 ) = _mm_shuffle_epi8( casti_v128( s,6 ), ctl ); \
+  casti_v128( d,7 ) = _mm_shuffle_epi8( casti_v128( s,7 ), ctl ); \
 }
 #define mm128_block_bswap64_512    mm128_block_bswap_64
 #define v128_block_bswap64_512     mm128_block_bswap_64
 
 #define v128_block_bswap64_1024( d, s ) \
 { \
-   __m128i ctl = _mm_set_epi64x(  0x08090a0b0c0d0e0f, 0x0001020304050607 ); \
-  casti_m128i( d, 0 ) = _mm_shuffle_epi8( casti_m128i( s, 0 ), ctl ); \
-  casti_m128i( d, 1 ) = _mm_shuffle_epi8( casti_m128i( s, 1 ), ctl ); \
-  casti_m128i( d, 2 ) = _mm_shuffle_epi8( casti_m128i( s, 2 ), ctl ); \
-  casti_m128i( d, 3 ) = _mm_shuffle_epi8( casti_m128i( s, 3 ), ctl ); \
-  casti_m128i( d, 4 ) = _mm_shuffle_epi8( casti_m128i( s, 4 ), ctl ); \
-  casti_m128i( d, 5 ) = _mm_shuffle_epi8( casti_m128i( s, 5 ), ctl ); \
-  casti_m128i( d, 6 ) = _mm_shuffle_epi8( casti_m128i( s, 6 ), ctl ); \
-  casti_m128i( d, 7 ) = _mm_shuffle_epi8( casti_m128i( s, 7 ), ctl ); \
-  casti_m128i( d, 8 ) = _mm_shuffle_epi8( casti_m128i( s, 8 ), ctl ); \
-  casti_m128i( d, 9 ) = _mm_shuffle_epi8( casti_m128i( s, 9 ), ctl ); \
-  casti_m128i( d,10 ) = _mm_shuffle_epi8( casti_m128i( s,10 ), ctl ); \
-  casti_m128i( d,11 ) = _mm_shuffle_epi8( casti_m128i( s,11 ), ctl ); \
-  casti_m128i( d,12 ) = _mm_shuffle_epi8( casti_m128i( s,12 ), ctl ); \
-  casti_m128i( d,13 ) = _mm_shuffle_epi8( casti_m128i( s,13 ), ctl ); \
-  casti_m128i( d,14 ) = _mm_shuffle_epi8( casti_m128i( s,14 ), ctl ); \
-  casti_m128i( d,15 ) = _mm_shuffle_epi8( casti_m128i( s,15 ), ctl ); \
+  v128_t ctl = _mm_set_epi64x(  0x08090a0b0c0d0e0f, 0x0001020304050607 ); \
+  casti_v128( d, 0 ) = _mm_shuffle_epi8( casti_v128( s, 0 ), ctl ); \
+  casti_v128( d, 1 ) = _mm_shuffle_epi8( casti_v128( s, 1 ), ctl ); \
+  casti_v128( d, 2 ) = _mm_shuffle_epi8( casti_v128( s, 2 ), ctl ); \
+  casti_v128( d, 3 ) = _mm_shuffle_epi8( casti_v128( s, 3 ), ctl ); \
+  casti_v128( d, 4 ) = _mm_shuffle_epi8( casti_v128( s, 4 ), ctl ); \
+  casti_v128( d, 5 ) = _mm_shuffle_epi8( casti_v128( s, 5 ), ctl ); \
+  casti_v128( d, 6 ) = _mm_shuffle_epi8( casti_v128( s, 6 ), ctl ); \
+  casti_v128( d, 7 ) = _mm_shuffle_epi8( casti_v128( s, 7 ), ctl ); \
+  casti_v128( d, 8 ) = _mm_shuffle_epi8( casti_v128( s, 8 ), ctl ); \
+  casti_v128( d, 9 ) = _mm_shuffle_epi8( casti_v128( s, 9 ), ctl ); \
+  casti_v128( d,10 ) = _mm_shuffle_epi8( casti_v128( s,10 ), ctl ); \
+  casti_v128( d,11 ) = _mm_shuffle_epi8( casti_v128( s,11 ), ctl ); \
+  casti_v128( d,12 ) = _mm_shuffle_epi8( casti_v128( s,12 ), ctl ); \
+  casti_v128( d,13 ) = _mm_shuffle_epi8( casti_v128( s,13 ), ctl ); \
+  casti_v128( d,14 ) = _mm_shuffle_epi8( casti_v128( s,14 ), ctl ); \
+  casti_v128( d,15 ) = _mm_shuffle_epi8( casti_v128( s,15 ), ctl ); \
 }
 
 // 4 byte dword * 8 dwords * 4 lanes = 128 bytes
 #define mm128_block_bswap_32( d, s ) \
 { \
-   __m128i ctl = _mm_set_epi64x( 0x0c0d0e0f08090a0b, 0x0405060700010203 ); \
-  casti_m128i( d,0 ) = _mm_shuffle_epi8( casti_m128i( s,0 ), ctl ); \
-  casti_m128i( d,1 ) = _mm_shuffle_epi8( casti_m128i( s,1 ), ctl ); \
-  casti_m128i( d,2 ) = _mm_shuffle_epi8( casti_m128i( s,2 ), ctl ); \
-  casti_m128i( d,3 ) = _mm_shuffle_epi8( casti_m128i( s,3 ), ctl ); \
-  casti_m128i( d,4 ) = _mm_shuffle_epi8( casti_m128i( s,4 ), ctl ); \
-  casti_m128i( d,5 ) = _mm_shuffle_epi8( casti_m128i( s,5 ), ctl ); \
-  casti_m128i( d,6 ) = _mm_shuffle_epi8( casti_m128i( s,6 ), ctl ); \
-  casti_m128i( d,7 ) = _mm_shuffle_epi8( casti_m128i( s,7 ), ctl ); \
+  v128_t ctl = _mm_set_epi64x( 0x0c0d0e0f08090a0b, 0x0405060700010203 ); \
+  casti_v128( d,0 ) = _mm_shuffle_epi8( casti_v128( s,0 ), ctl ); \
+  casti_v128( d,1 ) = _mm_shuffle_epi8( casti_v128( s,1 ), ctl ); \
+  casti_v128( d,2 ) = _mm_shuffle_epi8( casti_v128( s,2 ), ctl ); \
+  casti_v128( d,3 ) = _mm_shuffle_epi8( casti_v128( s,3 ), ctl ); \
+  casti_v128( d,4 ) = _mm_shuffle_epi8( casti_v128( s,4 ), ctl ); \
+  casti_v128( d,5 ) = _mm_shuffle_epi8( casti_v128( s,5 ), ctl ); \
+  casti_v128( d,6 ) = _mm_shuffle_epi8( casti_v128( s,6 ), ctl ); \
+  casti_v128( d,7 ) = _mm_shuffle_epi8( casti_v128( s,7 ), ctl ); \
 }
 #define mm128_block_bswap32_256      mm128_block_bswap_32
 #define v128_block_bswap32_256       mm128_block_bswap_32
@@ -852,129 +795,127 @@ static inline __m128i mm128_shuflr_x8( const __m128i v, const int c )
 
 #define mm128_block_bswap32_128( d, s ) \
 { \
-   __m128i ctl = _mm_set_epi64x( 0x0c0d0e0f08090a0b, 0x0405060700010203 ); \
-  casti_m128i( d,0 ) = _mm_shuffle_epi8( casti_m128i( s,0 ), ctl ); \
-  casti_m128i( d,1 ) = _mm_shuffle_epi8( casti_m128i( s,1 ), ctl ); \
-  casti_m128i( d,2 ) = _mm_shuffle_epi8( casti_m128i( s,2 ), ctl ); \
-  casti_m128i( d,3 ) = _mm_shuffle_epi8( casti_m128i( s,3 ), ctl ); \
+  v128_t ctl = _mm_set_epi64x( 0x0c0d0e0f08090a0b, 0x0405060700010203 ); \
+  casti_v128( d,0 ) = _mm_shuffle_epi8( casti_v128( s,0 ), ctl ); \
+  casti_v128( d,1 ) = _mm_shuffle_epi8( casti_v128( s,1 ), ctl ); \
+  casti_v128( d,2 ) = _mm_shuffle_epi8( casti_v128( s,2 ), ctl ); \
+  casti_v128( d,3 ) = _mm_shuffle_epi8( casti_v128( s,3 ), ctl ); \
 }   
 
 #define v128_block_bswap32_512( d, s ) \
 { \
-   __m128i ctl = _mm_set_epi64x( 0x0c0d0e0f08090a0b, 0x0405060700010203 ); \
-  casti_m128i( d, 0 ) = _mm_shuffle_epi8( casti_m128i( s, 0 ), ctl ); \
-  casti_m128i( d, 1 ) = _mm_shuffle_epi8( casti_m128i( s, 1 ), ctl ); \
-  casti_m128i( d, 2 ) = _mm_shuffle_epi8( casti_m128i( s, 2 ), ctl ); \
-  casti_m128i( d, 3 ) = _mm_shuffle_epi8( casti_m128i( s, 3 ), ctl ); \
-  casti_m128i( d, 4 ) = _mm_shuffle_epi8( casti_m128i( s, 4 ), ctl ); \
-  casti_m128i( d, 5 ) = _mm_shuffle_epi8( casti_m128i( s, 5 ), ctl ); \
-  casti_m128i( d, 6 ) = _mm_shuffle_epi8( casti_m128i( s, 6 ), ctl ); \
-  casti_m128i( d, 7 ) = _mm_shuffle_epi8( casti_m128i( s, 7 ), ctl ); \
-  casti_m128i( d, 8 ) = _mm_shuffle_epi8( casti_m128i( s, 8 ), ctl ); \
-  casti_m128i( d, 9 ) = _mm_shuffle_epi8( casti_m128i( s, 9 ), ctl ); \
-  casti_m128i( d,10 ) = _mm_shuffle_epi8( casti_m128i( s,10 ), ctl ); \
-  casti_m128i( d,11 ) = _mm_shuffle_epi8( casti_m128i( s,11 ), ctl ); \
-  casti_m128i( d,12 ) = _mm_shuffle_epi8( casti_m128i( s,12 ), ctl ); \
-  casti_m128i( d,13 ) = _mm_shuffle_epi8( casti_m128i( s,13 ), ctl ); \
-  casti_m128i( d,14 ) = _mm_shuffle_epi8( casti_m128i( s,14 ), ctl ); \
-  casti_m128i( d,15 ) = _mm_shuffle_epi8( casti_m128i( s,15 ), ctl ); \
+  v128_t ctl = _mm_set_epi64x( 0x0c0d0e0f08090a0b, 0x0405060700010203 ); \
+  casti_v128( d, 0 ) = _mm_shuffle_epi8( casti_v128( s, 0 ), ctl ); \
+  casti_v128( d, 1 ) = _mm_shuffle_epi8( casti_v128( s, 1 ), ctl ); \
+  casti_v128( d, 2 ) = _mm_shuffle_epi8( casti_v128( s, 2 ), ctl ); \
+  casti_v128( d, 3 ) = _mm_shuffle_epi8( casti_v128( s, 3 ), ctl ); \
+  casti_v128( d, 4 ) = _mm_shuffle_epi8( casti_v128( s, 4 ), ctl ); \
+  casti_v128( d, 5 ) = _mm_shuffle_epi8( casti_v128( s, 5 ), ctl ); \
+  casti_v128( d, 6 ) = _mm_shuffle_epi8( casti_v128( s, 6 ), ctl ); \
+  casti_v128( d, 7 ) = _mm_shuffle_epi8( casti_v128( s, 7 ), ctl ); \
+  casti_v128( d, 8 ) = _mm_shuffle_epi8( casti_v128( s, 8 ), ctl ); \
+  casti_v128( d, 9 ) = _mm_shuffle_epi8( casti_v128( s, 9 ), ctl ); \
+  casti_v128( d,10 ) = _mm_shuffle_epi8( casti_v128( s,10 ), ctl ); \
+  casti_v128( d,11 ) = _mm_shuffle_epi8( casti_v128( s,11 ), ctl ); \
+  casti_v128( d,12 ) = _mm_shuffle_epi8( casti_v128( s,12 ), ctl ); \
+  casti_v128( d,13 ) = _mm_shuffle_epi8( casti_v128( s,13 ), ctl ); \
+  casti_v128( d,14 ) = _mm_shuffle_epi8( casti_v128( s,14 ), ctl ); \
+  casti_v128( d,15 ) = _mm_shuffle_epi8( casti_v128( s,15 ), ctl ); \
 }
 
 #else  // SSE2
 
-static inline __m128i mm128_bswap_64( __m128i v )
+static inline v128_t v128_bswap64( __m128i v )
 {
       v = _mm_or_si128( _mm_slli_epi16( v, 8 ), _mm_srli_epi16( v, 8 ) );
       v = _mm_shufflelo_epi16( v, _MM_SHUFFLE( 0, 1, 2, 3 ) );
   return  _mm_shufflehi_epi16( v, _MM_SHUFFLE( 0, 1, 2, 3 ) );
 }
 
-static inline __m128i mm128_bswap_32( __m128i v )
+static inline v128_t v128_bswap32( __m128i v )
 {
       v = _mm_or_si128( _mm_slli_epi16( v, 8 ), _mm_srli_epi16( v, 8 ) );
       v = _mm_shufflelo_epi16( v, _MM_SHUFFLE( 2, 3, 0, 1 ) );
   return  _mm_shufflehi_epi16( v, _MM_SHUFFLE( 2, 3, 0, 1 ) );
 }
+#define mm128_bswap_32      v128_bswap32 
 
-static inline __m128i mm128_bswap_16( __m128i v )
+static inline v128_t v128_bswap16( __m128i v )
 {
   return _mm_or_si128( _mm_slli_epi16( v, 8 ), _mm_srli_epi16( v, 8 ) );
 }
 
-#define mm128_bswap_128( v )   v128_qrev32( v128_bswap64( v ) )
+#define v128_bswap128( v )   v128_qrev32( v128_bswap64( v ) )
 
 static inline void mm128_block_bswap_64( __m128i *d, const __m128i *s )
 {
-   d[0] = mm128_bswap_64( s[0] );
-   d[1] = mm128_bswap_64( s[1] );
-   d[2] = mm128_bswap_64( s[2] );
-   d[3] = mm128_bswap_64( s[3] );
-   d[4] = mm128_bswap_64( s[4] );
-   d[5] = mm128_bswap_64( s[5] );
-   d[6] = mm128_bswap_64( s[6] );
-   d[7] = mm128_bswap_64( s[7] );
+   d[0] = v128_bswap64( s[0] );
+   d[1] = v128_bswap64( s[1] );
+   d[2] = v128_bswap64( s[2] );
+   d[3] = v128_bswap64( s[3] );
+   d[4] = v128_bswap64( s[4] );
+   d[5] = v128_bswap64( s[5] );
+   d[6] = v128_bswap64( s[6] );
+   d[7] = v128_bswap64( s[7] );
 }
 #define v128_block_bswap64_512 mm128_block_bswap_64
 
 static inline void mm128_block_bswap64_1024( __m128i *d, const __m128i *s )
 {
-   d[ 0] = mm128_bswap_64( s[ 0] );
-   d[ 1] = mm128_bswap_64( s[ 1] );
-   d[ 2] = mm128_bswap_64( s[ 2] );
-   d[ 3] = mm128_bswap_64( s[ 3] );
-   d[ 4] = mm128_bswap_64( s[ 4] );
-   d[ 5] = mm128_bswap_64( s[ 5] );
-   d[ 6] = mm128_bswap_64( s[ 6] );
-   d[ 7] = mm128_bswap_64( s[ 7] );
-   d[ 8] = mm128_bswap_64( s[ 8] );
-   d[ 9] = mm128_bswap_64( s[ 9] );
-   d[10] = mm128_bswap_64( s[10] );
-   d[11] = mm128_bswap_64( s[11] );
-   d[14] = mm128_bswap_64( s[12] );
-   d[13] = mm128_bswap_64( s[13] );
-   d[14] = mm128_bswap_64( s[14] );
-   d[15] = mm128_bswap_64( s[15] );
+   d[ 0] = v128_bswap64( s[ 0] );
+   d[ 1] = v128_bswap64( s[ 1] );
+   d[ 2] = v128_bswap64( s[ 2] );
+   d[ 3] = v128_bswap64( s[ 3] );
+   d[ 4] = v128_bswap64( s[ 4] );
+   d[ 5] = v128_bswap64( s[ 5] );
+   d[ 6] = v128_bswap64( s[ 6] );
+   d[ 7] = v128_bswap64( s[ 7] );
+   d[ 8] = v128_bswap64( s[ 8] );
+   d[ 9] = v128_bswap64( s[ 9] );
+   d[10] = v128_bswap64( s[10] );
+   d[11] = v128_bswap64( s[11] );
+   d[14] = v128_bswap64( s[12] );
+   d[13] = v128_bswap64( s[13] );
+   d[14] = v128_bswap64( s[14] );
+   d[15] = v128_bswap64( s[15] );
 }
 
 static inline void mm128_block_bswap_32( __m128i *d, const __m128i *s )
 {
-   d[0] = mm128_bswap_32( s[0] );
-   d[1] = mm128_bswap_32( s[1] );
-   d[2] = mm128_bswap_32( s[2] );
-   d[3] = mm128_bswap_32( s[3] );
-   d[4] = mm128_bswap_32( s[4] );
-   d[5] = mm128_bswap_32( s[5] );
-   d[6] = mm128_bswap_32( s[6] );
-   d[7] = mm128_bswap_32( s[7] );
+   d[0] = v128_bswap32( s[0] );
+   d[1] = v128_bswap32( s[1] );
+   d[2] = v128_bswap32( s[2] );
+   d[3] = v128_bswap32( s[3] );
+   d[4] = v128_bswap32( s[4] );
+   d[5] = v128_bswap32( s[5] );
+   d[6] = v128_bswap32( s[6] );
+   d[7] = v128_bswap32( s[7] );
 }
 #define mm128_block_bswap32_256 mm128_block_bswap_32
 #define v128_block_bswap32_256  mm128_block_bswap_32
 
 static inline void mm128_block_bswap32_512( __m128i *d, const __m128i *s )
 {
-   d[ 0] = mm128_bswap_32( s[ 0] );
-   d[ 1] = mm128_bswap_32( s[ 1] );
-   d[ 2] = mm128_bswap_32( s[ 2] );
-   d[ 3] = mm128_bswap_32( s[ 3] );
-   d[ 4] = mm128_bswap_32( s[ 4] );
-   d[ 5] = mm128_bswap_32( s[ 5] );
-   d[ 6] = mm128_bswap_32( s[ 6] );
-   d[ 7] = mm128_bswap_32( s[ 7] );
-   d[ 8] = mm128_bswap_32( s[ 8] );
-   d[ 9] = mm128_bswap_32( s[ 9] );
-   d[10] = mm128_bswap_32( s[10] );
-   d[11] = mm128_bswap_32( s[11] );
-   d[12] = mm128_bswap_32( s[12] );
-   d[13] = mm128_bswap_32( s[13] );
-   d[14] = mm128_bswap_32( s[14] );
-   d[15] = mm128_bswap_32( s[15] );
+   d[ 0] = v128_bswap32( s[ 0] );
+   d[ 1] = v128_bswap32( s[ 1] );
+   d[ 2] = v128_bswap32( s[ 2] );
+   d[ 3] = v128_bswap32( s[ 3] );
+   d[ 4] = v128_bswap32( s[ 4] );
+   d[ 5] = v128_bswap32( s[ 5] );
+   d[ 6] = v128_bswap32( s[ 6] );
+   d[ 7] = v128_bswap32( s[ 7] );
+   d[ 8] = v128_bswap32( s[ 8] );
+   d[ 9] = v128_bswap32( s[ 9] );
+   d[10] = v128_bswap32( s[10] );
+   d[11] = v128_bswap32( s[11] );
+   d[12] = v128_bswap32( s[12] );
+   d[13] = v128_bswap32( s[13] );
+   d[14] = v128_bswap32( s[14] );
+   d[15] = v128_bswap32( s[15] );
 }
 
 #endif // SSSE3 else SSE2
 
-#define v128_bswap32                   mm128_bswap_32
-#define v128_bswap64                   mm128_bswap_64
-#define v128_bswap128                  mm128_bswap_128
 #define v128_block_bswap32             mm128_block_bswap_32
 #define v128_block_bswap64             mm128_block_bswap_64
 
@@ -984,24 +925,20 @@ static inline void mm128_block_bswap32_512( __m128i *d, const __m128i *s )
 
 #if defined(__SSSE3__)
 
-#define mm128_alignr_64( hi, lo, c )    _mm_alignr_epi8( hi, lo, (c)*8 )
-#define mm128_alignr_32( hi, lo, c )    _mm_alignr_epi8( hi, lo, (c)*4 )
+#define v128_alignr8                   _mm_alignr_epi8
+#define v128_alignr64( hi, lo, c )     _mm_alignr_epi8( hi, lo, (c)*8 )
+#define v128_alignr32( hi, lo, c )     _mm_alignr_epi8( hi, lo, (c)*4 )
 
 #else
 
-#define mm128_alignr_64( hi, lo, c ) \
+#define v128_alignr64( hi, lo, c ) \
    _mm_or_si128( _mm_slli_si128( hi, (c)*8 ), _mm_srli_si128( lo, (c)*8 ) )
 
-#define mm128_alignr_32( hi, lo, c ) \
+#define v128_alignr32( hi, lo, c ) \
    _mm_or_si128( _mm_slli_si128( lo, (c)*4 ), _mm_srli_si128( hi, (c)*4 ) )
 
 #endif
 
-// NEON only uses vector mask. x86 blend selects second arg when control bit
-// is set. Blendv selects second arg when sign bit is set. And masking is the
-// opposite, elements are selected from the first arg if the mask bits are set.
-// Arm blend is a bit by bit blend while x76 is an elenet blend.
-// Reverse the logic so the use mask is consistent with both formats.
 #if defined(__SSE4_1__)
 
 #define v128_blendv                    _mm_blendv_epi8
@@ -1009,7 +946,7 @@ static inline void mm128_block_bswap32_512( __m128i *d, const __m128i *s )
 #else
 
 #define v128_blendv( v1, v0, mask ) \
-   v128_or( v128_andnot( mask, v0 ), v128_and( mask, v1 ) )
+   v128_or( v128_andnot( mask, v1 ), v128_and( mask, v0 ) )
 
 #endif
 

simd-utils/simd-256.h

@@ -90,7 +90,7 @@ typedef union
 // code and therefore can't be used as compile time initializers.
 
 #define m256_zero            _mm256_setzero_si256()
-#define m256_one_128         mm256_bcast_m128( m128_one_128 )
+#define m256_one_128         mm256_bcast_m128( v128_one )
 
 static inline __m256i mm256_neg1_fn()
 {

simd-utils/simd-neon.h

@@ -21,36 +21,36 @@
 //
 //  vornq( v1, v0 )        or( v1, not( v0 ) )
 
-#define v128_t                         uint32x4_t   // default, 
-#define v128u64_t                      uint64x2_t
-#define v128u32_t                      uint32x4_t
-#define v128u16_t                      uint16x8_t
-#define v128u8_t                       uint8x16_t
+#define v128_t                        uint32x4_t   // default, 
+#define v128u64_t                     uint64x2_t
+#define v128u32_t                     uint32x4_t
+#define v128u16_t                     uint16x8_t
+#define v128u8_t                      uint8x16_t
 
 // load & store
-#define v128_load( p )                 vld1q_u32( (uint32_t*)(p) )
-#define v128_store( p, v )             vst1q_u32( (uint32_t*)(p), v )
+#define v128_load( p )                vld1q_u32( (uint32_t*)(p) )
+#define v128_store( p, v )            vst1q_u32( (uint32_t*)(p), v )
 
-#define v128u64_load( p )              vld1q_u64( (uint64_t*)(p) )
-#define v128u64_store( p, v )          vst1q_u64( (uint64_t*)(p), v )
-#define v128u32_load( p )              vld1q_u32( (uint32_t*)(p) )
-#define v128u32_store( p, v )          vst1q_u32( (uint32_t*)(p), v )
-#define v128u16_load( p )              vld1q_u16( (uint16_t*)(p) )
-#define v128u16_store( p, v )          vst1q_u16( (uint16_t*)(p), v )
-#define v128u8_load( p )               vld1q_u16( (uint8_t*)(p) )
-#define v128u8_store( p, v )           vst1q_u16( (uint8_t*)(p), v )
+#define v128u64_load( p )             vld1q_u64( (uint64_t*)(p) )
+#define v128u64_store( p, v )         vst1q_u64( (uint64_t*)(p), v )
+#define v128u32_load( p )             vld1q_u32( (uint32_t*)(p) )
+#define v128u32_store( p, v )         vst1q_u32( (uint32_t*)(p), v )
+#define v128u16_load( p )             vld1q_u16( (uint16_t*)(p) )
+#define v128u16_store( p, v )         vst1q_u16( (uint16_t*)(p), v )
+#define v128u8_load( p )              vld1q_u16( (uint8_t*)(p) )
+#define v128u8_store( p, v )          vst1q_u16( (uint8_t*)(p), v )
 
 // load & set1 combined
-#define v128_load1_64(p)               vld1q_dup_u64( (uint64_t*)(p) )
-#define v128_load1_32(p)               vld1q_dup_u32( (uint32_t*)(p) )
-#define v128_load1_16(p)               vld1q_dup_u16( (uint16_t*)(p) )
-#define v128_load1_8( p)               vld1q_dup_u8(  (uint8_t*) (p) )
+#define v128_load1_64(p)              vld1q_dup_u64( (uint64_t*)(p) )
+#define v128_load1_32(p)              vld1q_dup_u32( (uint32_t*)(p) )
+#define v128_load1_16(p)              vld1q_dup_u16( (uint16_t*)(p) )
+#define v128_load1_8( p)              vld1q_dup_u8(  (uint8_t*) (p) )
 
 // arithmetic
-#define v128_add64                     vaddq_u64
-#define v128_add32                     vaddq_u32
-#define v128_add16                     vaddq_u16
-#define v128_add8                      vaddq_u8
+#define v128_add64                    vaddq_u64
+#define v128_add32                    vaddq_u32
+#define v128_add16                    vaddq_u16
+#define v128_add8                     vaddq_u8
 
 #define v128_add4_64( v3, v2, v1, v0 ) \
    vaddq_u64( vaddq_u64( v3, v2 ), vaddq_u64( v1, v0 ) )
@@ -58,15 +58,15 @@
 #define v128_add4_32( v3, v2, v1, v0 ) \
    vaddq_u32( vaddq_u32( v3, v2 ), vaddq_u32( v1, v0 ) )
 
-#define v128_sub64                     vsubq_u64
-#define v128_sub32                     vsubq_u32
-#define v128_sub16                     vsubq_u16
-#define v128_sub8                      vsubq_u8
+#define v128_sub64                    vsubq_u64
+#define v128_sub32                    vsubq_u32
+#define v128_sub16                    vsubq_u16
+#define v128_sub8                     vsubq_u8
 
 // returns low half, u64 undocumented, may not exist.
-#define v128_mul64                     vmulq_u64
-#define v128_mul32                     vmulq_u32
-#define v128_mul16                     vmulq_u16
+#define v128_mul64                    vmulq_u64
+#define v128_mul32                    vmulq_u32
+#define v128_mul16                    vmulq_u16
 
 // slow, tested with argon2d
 static inline uint64x2_t v128_mulw32( uint32x4_t v1, uint32x4_t v0 )
@@ -76,101 +76,102 @@ static inline uint64x2_t v128_mulw32( uint32x4_t v1, uint32x4_t v0 )
 }
 
 // compare
-#define v128_cmpeq64                   vceqq_u64
-#define v128_cmpeq32                   vceqq_u32
-#define v128_cmpeq16                   vceqq_u16
-#define v128_cmpeq8                    vceqq_u8
+#define v128_cmpeq64                  vceqq_u64
+#define v128_cmpeq32                  vceqq_u32
+#define v128_cmpeq16                  vceqq_u16
+#define v128_cmpeq8                   vceqq_u8
 
-#define v128_iszero                    vceqzq_u64
+// v128_cmp0, v128_cmpz, v128 testz
+#define v128_iszero                   vceqzq_u64
 
 // Not yet needed
 //#define v128_cmpeq1
 
-#define v128_cmpgt64                   vcgtq_u64
-#define v128_cmpgt32                   vcgtq_u32
-#define v128_cmpgt16                   vcgtq_u16
-#define v128_cmpgt8                    vcgtq_u8
+#define v128_cmpgt64( v1, v0 )        vcgtq_s64( (int64x2_t)v1, (int64x2_t)v0 )
+#define v128_cmpgt32( v1, v0 )        vcgtq_s32( (int32x4_t)v1, (int32x4_t)v0 )
+#define v128_cmpgt16( v1, v0 )        vcgtq_s16( (int16x8_t)v1, (int16x8_t)v0 )
+#define v128_cmpgt8( v1, v0 )         vcgtq_s8( (int8x16_t)v1, (int8x16_t)v0 )
 
-#define v128_cmplt64                   vcltq_u64
-#define v128_cmplt32                   vcltq_u32
-#define v128_cmplt16                   vcltq_u16
-#define v128_cmplt8                    vcltq_u8
+#define v128_cmplt64( v1, v0 )        vcltq_s64( (int64x2_t)v1, (int64x2_t)v0 )
+#define v128_cmplt32( v1, v0 )        vcltq_s32( (int32x4_t)v1, (int32x4_t)v0 )
+#define v128_cmplt16( v1, v0 )        vcltq_s16( (int16x8_t)v1, (int16x8_t)v0 )
+#define v128_cmplt8( v1, v0 )         vcltq_s8( (int8x16_t)v1, (int8x16_t)v0 )
 
 // bit shift
-#define v128_sl64                      vshlq_n_u64
-#define v128_sl32                      vshlq_n_u32
-#define v128_sl16                      vshlq_n_u16
-#define v128_sl8                       vshlq_n_u8
+#define v128_sl64                     vshlq_n_u64
+#define v128_sl32                     vshlq_n_u32
+#define v128_sl16                     vshlq_n_u16
+#define v128_sl8                      vshlq_n_u8
 
-#define v128_sr64                      vshrq_n_u64
-#define v128_sr32                      vshrq_n_u32
-#define v128_sr16                      vshrq_n_u16
-#define v128_sr8                       vshrq_n_u8
+#define v128_sr64                     vshrq_n_u64
+#define v128_sr32                     vshrq_n_u32
+#define v128_sr16                     vshrq_n_u16
+#define v128_sr8                      vshrq_n_u8
 
 // Unit tested, working.
-#define v128_sra64                     vshrq_n_s64
-#define v128_sra32                     vshrq_n_s32
-#define v128_sra16                     vshrq_n_s16
+#define v128_sra64( v, c )            vshrq_n_s64( (int64x2_t)v, c )
+#define v128_sra32( v, c )            vshrq_n_s32( (int32x4_t)v, c )
+#define v128_sra16( v, c )            vshrq_n_s16( (int16x8_t)v, c )
 
 // unary logic
-#define v128_not                       vmvnq_u32
+#define v128_not                      vmvnq_u32
 
 // binary logic
-#define v128_or                        vorrq_u32
-#define v128_and                       vandq_u32
-#define v128_xor                       veorq_u32
+#define v128_or                       vorrq_u32
+#define v128_and                      vandq_u32
+#define v128_xor                      veorq_u32
 
 // ~v1 & v0
-#define v128_andnot( v1, v0 )          vandq_u32( vmvnq_u32( v1 ), v0 )
+#define v128_andnot( v1, v0 )         vandq_u32( vmvnq_u32( v1 ), v0 )
 
 // ~( a ^ b ), same as (~a) ^ b
-#define v128_xnor( v1, v0 )            v128_not( v128_xor( v1, v0 ) )
+#define v128_xnor( v1, v0 )           v128_not( v128_xor( v1, v0 ) )
 
 // ~v1 | v0, x86_64 convention, first arg is  not'ed
-#define v128_ornot( v1, v0 )           vornq_u32( v0, v1 ) 
+#define v128_ornot( v1, v0 )          vornq_u32( v0, v1 ) 
 
 // ternary logic
 
 // v2 ^ v1 ^ v0
 // veorq_u32 not defined
 //#define v128_xor3                      veor3q_u32
-#define v128_xor3( v2, v1, v0 )        veorq_u32( v2, veorq_u32( v1, v0 ) )
+#define v128_xor3( v2, v1, v0 )       veorq_u32( v2, veorq_u32( v1, v0 ) )
 
 // v2 & v1 & v0
-#define v128_and3( v2, v1, v0 )        v128_and( v2, v128_and( v1, v0 ) )
+#define v128_and3( v2, v1, v0 )       v128_and( v2, v128_and( v1, v0 ) )
 
 // v2 | v1 | v0
-#define v128_or3( v2, v1, v0 )         v128_or( v2, v128_or( v1, v0 ) )
+#define v128_or3( v2, v1, v0 )        v128_or( v2, v128_or( v1, v0 ) )
 
 // a ^ ( ~b & c )
-#define v128_xorandnot( v2, v1, v0 )   v128_xor( v2, v128_andnot( v1, v0 ) )
+#define v128_xorandnot( v2, v1, v0 )  v128_xor( v2, v128_andnot( v1, v0 ) )
 
 // a ^ ( b & c )
-#define v128_xorand( v2, v1, v0 )      v128_xor( v2, v128_and( v1, v0 ) )
+#define v128_xorand( v2, v1, v0 )     v128_xor( v2, v128_and( v1, v0 ) )
 
 // a & ( b ^ c )
-#define v128_andxor( v2, v1, v0 )      v128_and( v2, v128_xor( v1, v0 ) )
+#define v128_andxor( v2, v1, v0 )     v128_and( v2, v128_xor( v1, v0 ) )
 
 // a ^ ( b | c )
-#define v128_xoror( v2, v1, v0 )       v128_xor( v2, v128_or( v1, v0 ) )
+#define v128_xoror( v2, v1, v0 )      v128_xor( v2, v128_or( v1, v0 ) )
 
 // v2 | ( v1 & v0 )
-#define v128_orand( v2, v1, v0 )       v128_or( v2, v128_and( v1, v0 ) )
+#define v128_orand( v2, v1, v0 )      v128_or( v2, v128_and( v1, v0 ) )
 
 // shift 2 concatenated vectors right.
-#define v128_alignr64( v1, v0, c )     vextq_u64( v0, v1, c )
-#define v128_alignr32( v1, v0, c )     vextq_u32( v0, v1, c )
-#define v128_alignr8(  v1, v0, c )     vextq_u8(  v0, v1, c ) 
+#define v128_alignr64( v1, v0, c )    vextq_u64( v0, v1, c )
+#define v128_alignr32( v1, v0, c )    vextq_u32( v0, v1, c )
+#define v128_alignr8(  v1, v0, c )    vextq_u8(  v0, v1, c ) 
 
 // Intetleave high or low half of 2 vectors.
-#define v128_unpacklo64( v1, v0 )      vzip1q_u64( v1, v0 )
-#define v128_unpackhi64( v1, v0 )      vzip2q_u64( v1, v0 )
-#define v128_unpacklo32( v1, v0 )      vzip1q_u32( v1, v0 )
-#define v128_unpackhi32( v1, v0 )      vzip2q_u32( v1, v0 )
-#define v128_unpacklo16( v1, v0 )      vzip1q_u16( v1, v0 )
-#define v128_unpackhi16( v1, v0 )      vzip2q_u16( v1, v0 )
-#define v128_unpacklo8(  v1, v0 )      vzip1q_u8(  v1, v0 )
-#define v128_unpackhi8(  v1, v0 )      vzip2q_u8(  v1, v0 )
+#define v128_unpacklo64( v1, v0 )     vzip1q_u64( v1, v0 )
+#define v128_unpackhi64( v1, v0 )     vzip2q_u64( v1, v0 )
+#define v128_unpacklo32( v1, v0 )     vzip1q_u32( v1, v0 )
+#define v128_unpackhi32( v1, v0 )     vzip2q_u32( v1, v0 )
+#define v128_unpacklo16( v1, v0 )     vzip1q_u16( v1, v0 )
+#define v128_unpackhi16( v1, v0 )     vzip2q_u16( v1, v0 )
+#define v128_unpacklo8(  v1, v0 )     vzip1q_u8(  v1, v0 )
+#define v128_unpackhi8(  v1, v0 )     vzip2q_u8(  v1, v0 )
 
 
 // AES
@@ -184,19 +185,19 @@ static inline uint64x2_t v128_mulw32( uint32x4_t v1, uint32x4_t v0 )
 #define v128_aesenclast( v, k ) \
    v128_xor( k, vaeseq_u8( v, v128_zero ) )
 
-#define v128_aesenclast_nokey( v, k ) \
+#define v128_aesenclast_nokey( v ) \
    vaeseq_u8( v, v128_zero )
 
 #define v128_aesdec( v, k ) \
     v128_xor( k, vaesimcq_u8( vaesdq_u8( v, v128_zero ) ) )
 
-#define v128_aesdec_nokey( v, k ) \
+#define v128_aesdec_nokey( v ) \
     vaesimcq_u8( vaesdq_u8( v, v128_zero ) )
 
 #define v128_aesdeclast( v, k ) \
     v128_xor( k, vaesdq_u8( v, v128_zero ) )
 
-#define v128_aesdeclast_nokey( v, k ) \
+#define v128_aesdeclast_nokey( v ) \
     vaesdq_u8( v, v128_zero )
 
 
@@ -336,27 +337,27 @@ static inline void v128_memcpy( void *dst, const void *src, const int n )
 
 // Bit rotation
 #define v128_ror64( v, c ) \
-  ( (c) == 32 ) ? (uint64x2_t)vrev64q_u32( ((uint64x2_t)v) ) \
+  ( (c) == 32 ) ? (uint64x2_t)vrev64q_u32( ((uint32x4_t)v) ) \
    : vsriq_n_u64( vshlq_n_u64( ((uint64x2_t)v), 64-c ), ((uint64x2_t)v), c )
 
 #define v128_rol64( v, c ) \
-  ( (c) == 32 ) ? (uint64x2_t)vrev64q_u32( ((uint64x2_t)v) ) \
+  ( (c) == 32 ) ? (uint64x2_t)vrev64q_u32( ((uint32x4_t)v) ) \
    : vsliq_n_u64( vshrq_n_u64( ((uint64x2_t)v), 64-c ), ((uint64x2_t)v), c )
 
 #define v128_ror32( v, c ) \
-  ( (c) == 16 ) ? (uint32x4_t)vrev32q_u16( ((uint32x4_t)v) ) \
+  ( (c) == 16 ) ? (uint32x4_t)vrev32q_u16( ((uint16x8_t)v) ) \
    : vsriq_n_u32( vshlq_n_u32( ((uint32x4_t)v), 32-c ), ((uint32x4_t)v), c )
 
 #define v128_rol32( v, c ) \
-  ( (c) == 16 ) ? (uint32x4_t)vrev32q_u16( ((uint32x4_t)v) ) \
+  ( (c) == 16 ) ? (uint32x4_t)vrev32q_u16( ((uint16x8_t)v) ) \
    : vsliq_n_u32( vshrq_n_u32( ((uint32x4_t)v), 32-c ), ((uint32x4_t)v), c )
 
 #define v128_ror16( v, c ) \
-  ( (c) == 8 ) ? (uint16x8_t)vrev16q_u8( ((uint16x8_t)v) ) \
+  ( (c) == 8 ) ? (uint16x8_t)vrev16q_u8( ((uint8x16_t)v) ) \
    : vsriq_n_u16( vshlq_n_u16( ((uint16x8_t)v), 16-c ), ((uint16x8_t)v), c )
 
 #define v128_rol16( v, c ) \
-  ( (c) == 8 ) ? (uint16x8_t)vrev16q_u8( ((uint16x8_t)v) ) \
+  ( (c) == 8 ) ? (uint16x8_t)vrev16q_u8( ((uint8x16_t)v) ) \
    : vsliq_n_u16( vshrq_n_u16( ((uint16x8_t)v), 16-c ), ((uint16x8_t)v), c )
 
 #define v128_ror8( v, c ) \
@@ -433,6 +434,7 @@ static inline void v128_memcpy( void *dst, const void *src, const int n )
               ((uint8_t*)&v)[ ((uint8_t*)(&vmask))[ 1] ], \
               ((uint8_t*)&v)[ ((uint8_t*)(&vmask))[ 0] ] )
 
+
 // sub-vector shuffles sometimes mirror bit rotation. Shuffle is faster.
 // Bit rotation already promotes faster widths. Usage is context sensitive.
 // preferred.