diff --git a/RELEASE_NOTES b/RELEASE_NOTES
index 2ca3180..870df90 100644
--- a/RELEASE_NOTES
+++ b/RELEASE_NOTES
@@ -38,6 +38,12 @@ supported.
 Change Log
 ----------
 
+v3.9.5.4
+
+Fixed sha256q AVX2 poor performance.
+Fixed skein2 buffer overflow and restored bswap-interleave optimization.
+More restructuring.
+
 v3.9.5.3
 
 Fix crash mining hodl with aes-sse42.
@@ -45,7 +51,7 @@ More restructuring and share report tweaks.
 
 v3.9.5.2
 
-Revert bswap-interleave optiization for causing crashes on Windows.
+Revert bswap-interleave optimization for causing crashes on Windows.
 
 v3.9.5.1
 
diff --git a/algo/blake/blake-4way.c b/algo/blake/blake-4way.c
index a075fa3..8d1372f 100644
--- a/algo/blake/blake-4way.c
+++ b/algo/blake/blake-4way.c
@@ -27,25 +27,19 @@ int scanhash_blake_4way( struct work *work, uint32_t max_nonce,
    uint32_t *ptarget = work->target;
    const uint32_t first_nonce = pdata[19];
    uint32_t HTarget = ptarget[7];
-   uint32_t _ALIGN(32) edata[20];
+   __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
    uint32_t n = first_nonce;
    int thr_id = mythr->id;  // thr_id arg is deprecated
 
    if (opt_benchmark)
       HTarget = 0x7f;
 
-   // we need big endian data...
-   swab32_array( edata, pdata, 20 );
-   mm128_intrlv_4x32( vdata, edata, edata, edata, edata, 640 );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
    blake256r14_4way_init( &blake_4w_ctx );
    blake256r14_4way( &blake_4w_ctx, vdata, 64 );
 
-   uint32_t *noncep = vdata + 76;   // 19*4
    do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
+      *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
 
       blakehash_4way( hash, vdata );
 
@@ -76,9 +70,9 @@ void blakehash_8way( void *state, const void *input )
      memcpy( &ctx, &blake_8w_ctx, sizeof ctx );
      blake256r14_8way( &ctx, input + (64<<3), 16 );
      blake256r14_8way_close( &ctx, vhash );
-     mm256_dintrlv_8x32( state,     state+ 32, state+ 64, state+ 96,
-                              state+128, state+160, state+192, state+224,
-                              vhash, 256 );
+     _dintrlv_8x32( state,     state+ 32, state+ 64, state+ 96,
+                    state+128, state+160, state+192, state+224,
+                    vhash, 256 );
 }
 
 int scanhash_blake_8way( struct work *work, uint32_t max_nonce,
@@ -90,32 +84,21 @@ int scanhash_blake_8way( struct work *work, uint32_t max_nonce,
    uint32_t *ptarget = work->target;
    const uint32_t first_nonce = pdata[19];
    uint32_t HTarget = ptarget[7];
-   uint32_t _ALIGN(32) edata[20];
    uint32_t n = first_nonce;
+   __m256i  *noncev = (__m256i*)vdata + 19;   // aligned
    int thr_id = mythr->id;  // thr_id arg is deprecated
 
    if (opt_benchmark)
       HTarget = 0x7f;
 
-   // we need big endian data...
-   swab32_array( edata, pdata, 20 );
-
-   mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
-                                 edata, edata, edata, edata, 640 );
+   mm256_bswap32_intrlv80_8x32( vdata, pdata );
 
    blake256r14_8way_init( &blake_8w_ctx );
    blake256r14_8way( &blake_8w_ctx, vdata, 64 );
 
-   uint32_t *noncep = vdata + 152;   // 19*8
    do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
-      be32enc( noncep +4, n+4 );
-      be32enc( noncep +5, n+5 );
-      be32enc( noncep +6, n+6 );
-      be32enc( noncep +7, n+7 );
+      *noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
+                                                  n+3, n+2, n+1, n ) );
       pdata[19] = n;
 
       blakehash_8way( hash, vdata );
diff --git a/algo/blake/blake2s-4way.c b/algo/blake/blake2s-4way.c
index e26b587..9048566 100644
--- a/algo/blake/blake2s-4way.c
+++ b/algo/blake/blake2s-4way.c
@@ -16,9 +16,9 @@ void blake2s_8way_hash( void *output, const void *input )
    blake2s_8way_update( &ctx, input + (64<<3), 16 );
    blake2s_8way_final( &ctx, vhash, BLAKE2S_OUTBYTES );
 
-   mm256_dintrlv_8x32( output,     output+ 32, output+ 64, output+ 96,
-                            output+128, output+160, output+192, output+224,
-                            vhash, 256 );
+   dintrlv_8x32( output,     output+ 32, output+ 64, output+ 96,
+                 output+128, output+160, output+192, output+224,
+                 vhash, 256 );
 }
 
 int scanhash_blake2s_8way( struct work *work, uint32_t max_nonce,
@@ -28,28 +28,19 @@ int scanhash_blake2s_8way( struct work *work, uint32_t max_nonce,
    uint32_t hash[8*8] __attribute__ ((aligned (32)));
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
-   uint32_t _ALIGN(64) edata[20];
    const uint32_t Htarg = ptarget[7];
    const uint32_t first_nonce = pdata[19];
+   __m256i  *noncev = (__m256i*)vdata + 19;   // aligned
    uint32_t n = first_nonce;
-   uint32_t *noncep = vdata + 152;   // 19*8
    int thr_id = mythr->id;  // thr_id arg is deprecated
 
-   swab32_array( edata, pdata, 20 );
-   mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
-                                 edata, edata, edata, edata, 640 );
+   mm256_bswap32_intrlv80_8x32( vdata, pdata );
    blake2s_8way_init( &blake2s_8w_ctx, BLAKE2S_OUTBYTES );
    blake2s_8way_update( &blake2s_8w_ctx, vdata, 64 );
 
    do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
-      be32enc( noncep +4, n+4 );
-      be32enc( noncep +5, n+5 );
-      be32enc( noncep +6, n+6 );
-      be32enc( noncep +7, n+7 );
+      *noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
+                                                  n+3, n+2, n+1, n ) );
       pdata[19] = n;
 
       blake2s_8way_hash( hash, vdata );
@@ -94,23 +85,18 @@ int scanhash_blake2s_4way( struct work *work, uint32_t max_nonce,
    uint32_t hash[8*4] __attribute__ ((aligned (32)));
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
-   uint32_t _ALIGN(64) edata[20];
    const uint32_t Htarg = ptarget[7];
    const uint32_t first_nonce = pdata[19];
+   __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
    uint32_t n = first_nonce;
-   uint32_t *noncep = vdata + 76;   // 19*4
    int thr_id = mythr->id;  // thr_id arg is deprecated
 
-   swab32_array( edata, pdata, 20 );
-   mm128_intrlv_4x32( vdata, edata, edata, edata, edata, 640 );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
    blake2s_4way_init( &blake2s_4w_ctx, BLAKE2S_OUTBYTES );
    blake2s_4way_update( &blake2s_4w_ctx, vdata, 64 );
 
    do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
+      *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
       pdata[19] = n;
 
       blake2s_4way_hash( hash, vdata );
diff --git a/algo/blake/blakecoin-4way.c b/algo/blake/blakecoin-4way.c
index 4fbd91e..898cbe3 100644
--- a/algo/blake/blakecoin-4way.c
+++ b/algo/blake/blakecoin-4way.c
@@ -29,23 +29,18 @@ int scanhash_blakecoin_4way( struct work *work, uint32_t max_nonce,
    uint32_t *ptarget = work->target;
    const uint32_t first_nonce = pdata[19];
    uint32_t HTarget = ptarget[7];
-   uint32_t _ALIGN(32) edata[20];
    uint32_t n = first_nonce;
+   __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
    int thr_id = mythr->id;  // thr_id arg is deprecated
    if ( opt_benchmark )
       HTarget = 0x7f;
 
-   swab32_array( edata, pdata, 20 );
-   mm128_intrlv_4x32( vdata, edata, edata, edata, edata, 640 );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
    blake256r8_4way_init( &blakecoin_4w_ctx );
    blake256r8_4way( &blakecoin_4w_ctx, vdata, 64 );
 
-   uint32_t *noncep = vdata + 76;   // 19*4
    do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
+      *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
       pdata[19] = n;
       blakecoin_4way_hash( hash, vdata );
 
@@ -79,9 +74,8 @@ void blakecoin_8way_hash( void *state, const void *input )
      blake256r8_8way( &ctx, input + (64<<3), 16 );
      blake256r8_8way_close( &ctx, vhash );
 
-     mm256_dintrlv_8x32( state,     state+ 32, state+ 64, state+ 96,
-                              state+128, state+160, state+192, state+224,
-                              vhash, 256 );
+     dintrlv_8x32( state,     state+ 32, state+ 64, state+ 96, state+128,
+                   state+160, state+192, state+224, vhash, 256 );
 }
 
 int scanhash_blakecoin_8way( struct work *work, uint32_t max_nonce,
@@ -93,29 +87,19 @@ int scanhash_blakecoin_8way( struct work *work, uint32_t max_nonce,
    uint32_t *ptarget = work->target;
    const uint32_t first_nonce = pdata[19];
    uint32_t HTarget = ptarget[7];
-   uint32_t _ALIGN(32) edata[20];
    uint32_t n = first_nonce;
-   uint32_t *noncep = vdata + 152;   // 19*8
+   __m256i  *noncev = (__m256i*)vdata + 19;   // aligned
    int thr_id = mythr->id;  // thr_id arg is deprecated
    if ( opt_benchmark )
       HTarget = 0x7f;
 
-   // we need big endian data...
-   swab32_array( edata, pdata, 20 );
-   mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
-                                 edata, edata, edata, edata, 640 );
+   mm256_bswap32_intrlv80_8x32( vdata, pdata );
    blake256r8_8way_init( &blakecoin_8w_ctx );
    blake256r8_8way( &blakecoin_8w_ctx, vdata, 64 );
 
    do {
-      be32enc( noncep,    n   );
-      be32enc( noncep +1, n+1 );
-      be32enc( noncep +2, n+2 );
-      be32enc( noncep +3, n+3 );
-      be32enc( noncep +4, n+4 );
-      be32enc( noncep +5, n+5 );
-      be32enc( noncep +6, n+6 );
-      be32enc( noncep +7, n+7 );
+      *noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
+                                                  n+3, n+2, n+1, n ) );
       pdata[19] = n;
       blakecoin_8way_hash( hash, vdata );
 
diff --git a/algo/blake/pentablake-4way.c b/algo/blake/pentablake-4way.c
index 7033d4a..1b45afa 100644
--- a/algo/blake/pentablake-4way.c
+++ b/algo/blake/pentablake-4way.c
@@ -10,13 +10,8 @@
 #include "blake-hash-4way.h"
 #include "sph_blake.h"
 
-//#define DEBUG_ALGO
-
 extern void pentablakehash_4way( void *output, const void *input )
 {
-//	unsigned char _ALIGN(32) hash[128];
-//	// same as uint32_t hashA[16], hashB[16];
-//	#define hashB hash+64
 
      uint64_t hash0[8] __attribute__ ((aligned (64)));
      uint64_t hash1[8] __attribute__ ((aligned (64)));
@@ -29,22 +24,7 @@ extern void pentablakehash_4way( void *output, const void *input )
      blake512_4way_init( &ctx );
      blake512_4way( &ctx, input, 80 );
      blake512_4way_close( &ctx, vhash );
-/*
-uint64_t sin0[10], sin1[10], sin2[10], sin3[10];
-mm256_deinterleave_4x64( sin0, sin1, sin2, sin3, input, 640 );
-sph_blake512_context ctx2_blake;
-sph_blake512_init(&ctx2_blake);
-sph_blake512(&ctx2_blake, sin0, 80);
-sph_blake512_close(&ctx2_blake, (void*) hash);
 
-mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
-uint64_t* hash64 = (uint64_t*)hash;
-for( int i = 0; i < 8; i++ )
-{
-   if ( hash0[i] != hash64[i] )
-      printf("hash mismatch %u\n",i);
-}
-*/
      blake512_4way_init( &ctx );
      blake512_4way( &ctx, vhash, 64 );
      blake512_4way_close( &ctx, vhash );
@@ -61,42 +41,10 @@ for( int i = 0; i < 8; i++ )
      blake512_4way( &ctx, vhash, 64 );
      blake512_4way_close( &ctx, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
      memcpy( output,    hash0, 32 );
      memcpy( output+32, hash1, 32 );
      memcpy( output+64, hash2, 32 );
      memcpy( output+96, hash3, 32 );
-
-/*
-     uint64_t sin0[10] __attribute__ ((aligned (64)));
-     uint64_t sin1[10] __attribute__ ((aligned (64)));
-     uint64_t sin2[10] __attribute__ ((aligned (64)));
-     uint64_t sin3[10] __attribute__ ((aligned (64)));
-
-	sph_blake512_context     ctx_blake;
-
-	sph_blake512_init(&ctx_blake);
-	sph_blake512(&ctx_blake, input, 80);
-	sph_blake512_close(&ctx_blake, hash);
-
-        sph_blake512_init(&ctx_blake);
-	sph_blake512(&ctx_blake, hash, 64);
-	sph_blake512_close(&ctx_blake, hash);
-
-        sph_blake512_init(&ctx_blake);
-	sph_blake512(&ctx_blake, hash, 64);
-	sph_blake512_close(&ctx_blake, hash);
-
-        sph_blake512_init(&ctx_blake);
-	sph_blake512(&ctx_blake, hash, 64);
-	sph_blake512_close(&ctx_blake, hash);
-
-        sph_blake512_init(&ctx_blake);
-	sph_blake512(&ctx_blake, hash, 64);
-	sph_blake512_close(&ctx_blake, hash);
-
-	memcpy(output, hash, 32);
-*/
 }
 
 int scanhash_pentablake_4way( struct work *work,
@@ -137,7 +85,7 @@ int scanhash_pentablake_4way( struct work *work,
     swab32_array( endiandata, pdata, 20 );
 
     uint64_t *edata = (uint64_t*)endiandata;
-    mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+    intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
 
     for ( int m=0; m < 6; m++ )
     {
diff --git a/algo/cryptonight/cryptolight.c b/algo/cryptonight/cryptolight.c
index 18e2940..6295be5 100644
--- a/algo/cryptonight/cryptolight.c
+++ b/algo/cryptonight/cryptolight.c
@@ -242,6 +242,8 @@ void cryptolight_hash(void* output, const void* input, int len) {
 	free(ctx);
 }
 
+#if defined(__AES__)
+
 static void cryptolight_hash_ctx_aes_ni(void* output, const void* input,
                        int len, struct cryptonight_ctx* ctx)
 {
@@ -312,6 +314,8 @@ static void cryptolight_hash_ctx_aes_ni(void* output, const void* input,
 	oaes_free((OAES_CTX **) &ctx->aes_ctx);
 }
 
+#endif
+
 int scanhash_cryptolight( struct work *work,
 		uint32_t max_nonce, uint64_t *hashes_done, struct thr_info *mythr)
 {
diff --git a/algo/groestl/myrgr-4way.c b/algo/groestl/myrgr-4way.c
index ecf6a97..b7ba0eb 100644
--- a/algo/groestl/myrgr-4way.c
+++ b/algo/groestl/myrgr-4way.c
@@ -47,10 +47,6 @@ void myriad_4way_hash( void *output, const void *input )
 
      sha256_4way( &ctx.sha, vhash, 64 );
      sha256_4way_close( &ctx.sha, output );
-
-//     sha256_4way_close( &ctx.sha, vhash );
-//     mm128_dintrlv_4x32( output, output+32, output+64, output+96,
-//                           vhash, 256 );
 }
 
 int scanhash_myriad_4way( struct work *work, uint32_t max_nonce,
@@ -68,18 +64,10 @@ int scanhash_myriad_4way( struct work *work, uint32_t max_nonce,
    __m128i  *noncev = (__m128i*)vdata + 19;   // aligned
    int thr_id = mythr->id;  // thr_id arg is deprecated
 
-/*
-        uint32_t *pdata = work->data;
-        uint32_t *ptarget = work->target;
-
-	uint32_t _ALIGN(64) endiandata[20];
-	const uint32_t first_nonce = pdata[19];
-	uint32_t nonce = first_nonce;
-*/
    if ( opt_benchmark )
       ( (uint32_t*)ptarget )[7] = 0x0000ff;
 
-   mm128_bswap_intrlv80_4x32( vdata, pdata );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
    do {
       *noncev = mm128_bswap_32( _mm_set_epi32( n+3,n+2,n+1,n ) );
 
diff --git a/algo/jh/jha-4way.c b/algo/jh/jha-4way.c
index 8177cc9..2c76a33 100644
--- a/algo/jh/jha-4way.c
+++ b/algo/jh/jha-4way.c
@@ -3,7 +3,6 @@
 #include <stdint.h>
 #include <string.h>
 #include <stdio.h>
-//#include "avxdefs.h"
 
 #if defined(JHA_4WAY)
 
@@ -13,9 +12,6 @@
 #include "algo/keccak/keccak-hash-4way.h"
 #include "algo/groestl/aes_ni/hash-groestl.h"
 
-//static __thread keccak512_4way_context jha_kec_mid
-//                                   __attribute__ ((aligned (64)));
-
 void jha_hash_4way( void *out, const void *input )
 {
     uint64_t hash0[8] __attribute__ ((aligned (64)));
@@ -46,7 +42,7 @@ void jha_hash_4way( void *out, const void *input )
        vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256(
                vh[0], _mm256_set1_epi64x( 1 ) ), m256_zero );
 
-       mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+       dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
        init_groestl( &ctx_groestl, 64 );
        update_and_final_groestl( &ctx_groestl, (char*)hash0,
                                                (char*)hash0, 512 );
@@ -59,7 +55,7 @@ void jha_hash_4way( void *out, const void *input )
        init_groestl( &ctx_groestl, 64 );
        update_and_final_groestl( &ctx_groestl, (char*)hash3,
                                                (char*)hash3, 512 );
-       mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+       intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
 
        skein512_4way_init( &ctx_skein );
        skein512_4way( &ctx_skein, vhash, 64 );
@@ -79,8 +75,6 @@ void jha_hash_4way( void *out, const void *input )
        for ( int i = 0; i < 8; i++ )
           casti_m256i( out, i ) = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );
     }
-
-//    mm256_dintrlv_4x64( out, out+32, out+64, out+96, vhash, 256 );
 }
 
 int scanhash_jha_4way( struct work *work, uint32_t max_nonce,
@@ -88,7 +82,6 @@ int scanhash_jha_4way( struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[8*4] __attribute__ ((aligned (64)));
    uint32_t vdata[20*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t *hash7 = &(hash[25]);
    uint32_t lane_hash[8] __attribute__ ((aligned (32)));
    uint32_t *pdata = work->data;
@@ -116,11 +109,7 @@ int scanhash_jha_4way( struct work *work, uint32_t max_nonce,
 		0
 	};
 
-   for ( int i=0; i < 19; i++ )
-      be32enc( &edata[i], pdata[i] );
-
-   mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//   mm256_bswap_intrlv80_4x64( vdata, pdata );
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
    for ( int m = 0; m < 6; m++ )
    {
@@ -130,26 +119,17 @@ int scanhash_jha_4way( struct work *work, uint32_t max_nonce,
          do {
               *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
                 _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
-//              be32enc( noncep,   n   );
-//              be32enc( noncep+2, n+1 );
-//              be32enc( noncep+4, n+2 );
-//              be32enc( noncep+6, n+3 );
 
               jha_hash_4way( hash, vdata );
               pdata[19] = n;
 
-//              for ( int i = 0; i < 4; i++ )
-//              if ( ( !( (hash+(i<<3))[7] & mask ) == 0 )
-//                  && fulltest( hash+(i<<3), ptarget ) )
               for ( int i = 0; i < 4; i++ ) if ( !( (hash7[i] & mask ) == 0 ) )
               {
-                 mm256_extr_lane_4x64( lane_hash, hash, i, 256 );
+                 extr_lane_4x64( lane_hash, hash, i, 256 );
                  if ( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
                  {
                     pdata[19] = n+i;
                     submit_lane_solution( work, lane_hash, mythr, i );
-//                 nonces[ num_found++ ] = n+i;
-//                 work_set_target_ratio( work, hash+(i<<3) );
                  }
               }
               n += 4;
diff --git a/algo/keccak/keccak-4way.c b/algo/keccak/keccak-4way.c
index 5ab6dda..5d2c87d 100644
--- a/algo/keccak/keccak-4way.c
+++ b/algo/keccak/keccak-4way.c
@@ -20,8 +20,7 @@ int scanhash_keccak_4way( struct work *work, uint32_t max_nonce,
                           uint64_t *hashes_done, struct thr_info *mythr )
 {
    uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-   uint32_t hash[8*4] __attribute__ ((aligned (32)));
-   uint32_t edata[20] __attribute__ ((aligned (64)));
+   uint32_t hash[16*4] __attribute__ ((aligned (32)));
    uint32_t lane_hash[8] __attribute__ ((aligned (32)));
    uint32_t *hash7 = &(hash[25]);   // 3*8+1
    uint32_t *pdata = work->data;
@@ -32,9 +31,7 @@ int scanhash_keccak_4way( struct work *work, uint32_t max_nonce,
 //   const uint32_t Htarg = ptarget[7];
     int thr_id = mythr->id;  // thr_id arg is deprecated
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//   mm256_bswap_intrlv80_4x64( vdata, pdata );
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
    do {
        *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
                 _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
@@ -44,7 +41,7 @@ int scanhash_keccak_4way( struct work *work, uint32_t max_nonce,
       for ( int lane = 0; lane < 4; lane++ )
       if ( ( ( hash7[ lane<<1 ] & 0xFFFFFF00 ) == 0 ) )
       {
-          mm256_extr_lane_4x64( lane_hash, hash, lane, 256 );
+          extr_lane_4x64( lane_hash, hash, lane, 256 );
           if ( fulltest( lane_hash, ptarget ) )
           {
               pdata[19] = n + lane;
diff --git a/algo/lyra2/allium-4way.c b/algo/lyra2/allium-4way.c
index 56d6216..d1471a5 100644
--- a/algo/lyra2/allium-4way.c
+++ b/algo/lyra2/allium-4way.c
@@ -44,11 +44,11 @@ void allium_4way_hash( void *state, const void *input )
    blake256_4way( &ctx.blake, input + (64<<2), 16 );
    blake256_4way_close( &ctx.blake, vhash32 );
 
-   mm256_rintrlv_4x32_4x64( vhash64, vhash32, 256 );
+   rintrlv_4x32_4x64( vhash64, vhash32, 256 );
    keccak256_4way( &ctx.keccak, vhash64, 32 );
    keccak256_4way_close( &ctx.keccak, vhash64 );
 
-   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
+   dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
 
    LYRA2RE( hash0, 32, hash0, 32, hash0, 32, 1, 8, 8 );
    LYRA2RE( hash1, 32, hash1, 32, hash1, 32, 1, 8, 8 );
@@ -68,12 +68,12 @@ void allium_4way_hash( void *state, const void *input )
    LYRA2RE( hash2, 32, hash2, 32, hash2, 32, 1, 8, 8 );
    LYRA2RE( hash3, 32, hash3, 32, hash3, 32, 1, 8, 8 );
 
-   mm256_intrlv_4x64( vhash64, hash0, hash1, hash2, hash3, 256 );
+   intrlv_4x64( vhash64, hash0, hash1, hash2, hash3, 256 );
 
    skein256_4way( &ctx.skein, vhash64, 32 );
    skein256_4way_close( &ctx.skein, vhash64 );
 
-   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
+   dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
 
    update_and_final_groestl256( &ctx.groestl, state, hash0, 256 );
    memcpy( &ctx.groestl, &allium_4way_ctx.groestl,
@@ -103,7 +103,7 @@ int scanhash_allium_4way( struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ( (uint32_t*)ptarget )[7] = 0x0000ff;
 
-   mm128_bswap_intrlv80_4x32( vdata, pdata );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
    blake256_4way_init( &allium_4way_ctx.blake );
    blake256_4way( &allium_4way_ctx.blake, vdata, 64 );
 
diff --git a/algo/lyra2/lyra2h-4way.c b/algo/lyra2/lyra2h-4way.c
index 137b614..a76e68c 100644
--- a/algo/lyra2/lyra2h-4way.c
+++ b/algo/lyra2/lyra2h-4way.c
@@ -64,7 +64,7 @@ int scanhash_lyra2h_4way( struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ptarget[7] = 0x0000ff;
 
-   mm128_bswap_intrlv80_4x32( vdata, pdata );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
    lyra2h_4way_midstate( vdata );
 
    do {
diff --git a/algo/lyra2/lyra2rev2-4way.c b/algo/lyra2/lyra2rev2-4way.c
index e100a06..9832fb1 100644
--- a/algo/lyra2/lyra2rev2-4way.c
+++ b/algo/lyra2/lyra2rev2-4way.c
@@ -42,12 +42,12 @@ void lyra2rev2_4way_hash( void *state, const void *input )
    blake256_4way( &ctx.blake, input + (64<<2), 16 );
    blake256_4way_close( &ctx.blake, vhash );
 
-   mm256_rintrlv_4x32_4x64( vhash64, vhash, 256 );
+   rintrlv_4x32_4x64( vhash64, vhash, 256 );
 
    keccak256_4way( &ctx.keccak, vhash64, 32 );
    keccak256_4way_close( &ctx.keccak, vhash64 );
 
-   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
+   dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
 
    cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
    cubehashInit( &ctx.cube, 256, 16, 32 );
@@ -62,12 +62,12 @@ void lyra2rev2_4way_hash( void *state, const void *input )
    LYRA2REV2( l2v2_wholeMatrix, hash2, 32, hash2, 32, hash2, 32, 1, 4, 4 );
    LYRA2REV2( l2v2_wholeMatrix, hash3, 32, hash3, 32, hash3, 32, 1, 4, 4 );
 
-   mm256_intrlv_4x64( vhash64, hash0, hash1, hash2, hash3, 256 );
+   intrlv_4x64( vhash64, hash0, hash1, hash2, hash3, 256 );
 
    skein256_4way( &ctx.skein, vhash64, 32 );
    skein256_4way_close( &ctx.skein, vhash64 );
 
-   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
+   dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
 
    cubehashInit( &ctx.cube, 256, 16, 32 );
    cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
@@ -102,7 +102,7 @@ int scanhash_lyra2rev2_4way( struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ( (uint32_t*)ptarget )[7] = 0x0000ff;
 
-   mm128_bswap_intrlv80_4x32( vdata, pdata );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
 
    blake256_4way_init( &l2v2_4way_ctx.blake );
    blake256_4way( &l2v2_4way_ctx.blake, vdata, 64 );
diff --git a/algo/lyra2/lyra2rev3-4way.c b/algo/lyra2/lyra2rev3-4way.c
index 1246288..389aebf 100644
--- a/algo/lyra2/lyra2rev3-4way.c
+++ b/algo/lyra2/lyra2rev3-4way.c
@@ -41,7 +41,7 @@ void lyra2rev3_8way_hash( void *state, const void *input )
    blake256_8way( &ctx.blake, input, 80 );
    blake256_8way_close( &ctx.blake, vhash );
 
-   mm256_dintrlv_8x32( hash0, hash1, hash2, hash3,
+   dintrlv_8x32( hash0, hash1, hash2, hash3,
                        hash4, hash5, hash6, hash7, vhash, 256 );
 
    LYRA2REV3( l2v3_wholeMatrix, hash0, 32, hash0, 32, hash0, 32, 1, 4, 4 );
@@ -78,7 +78,7 @@ void lyra2rev3_8way_hash( void *state, const void *input )
    LYRA2REV3( l2v3_wholeMatrix, hash6, 32, hash6, 32, hash6, 32, 1, 4, 4 );
    LYRA2REV3( l2v3_wholeMatrix, hash7, 32, hash7, 32, hash7, 32, 1, 4, 4 );
 
-   mm256_intrlv_8x32( vhash, hash0, hash1, hash2, hash3,
+   intrlv_8x32( vhash, hash0, hash1, hash2, hash3,
                              hash4, hash5, hash6, hash7, 256 );
 
    bmw256_8way( &ctx.bmw, vhash, 32 );
@@ -91,7 +91,6 @@ int scanhash_lyra2rev3_8way( struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[8*8] __attribute__ ((aligned (64)));
    uint32_t vdata[20*8] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t *hash7 = &(hash[7<<3]);
    uint32_t lane_hash[8] __attribute__ ((aligned (32)));
    uint32_t *pdata = work->data;
@@ -105,10 +104,7 @@ int scanhash_lyra2rev3_8way( struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ( (uint32_t*)ptarget )[7] = 0x0000ff;
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
-                              edata, edata, edata, edata, 640 );
-//   mm256_bswap_intrlv80_8x32( vdata, pdata );
+   mm256_bswap32_intrlv80_8x32( vdata, pdata );
    do
    {
       *noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
@@ -119,7 +115,7 @@ int scanhash_lyra2rev3_8way( struct work *work, uint32_t max_nonce,
 
       for ( int lane = 0; lane < 8; lane++ ) if ( hash7[lane] <= Htarg )
       {
-         mm256_extr_lane_8x32( lane_hash, hash, lane, 256 );
+         extr_lane_8x32( lane_hash, hash, lane, 256 );
          if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
          {
               pdata[19] = n + lane;
@@ -208,7 +204,7 @@ int scanhash_lyra2rev3_4way( struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ( (uint32_t*)ptarget )[7] = 0x0000ff;
 
-   mm128_bswap_intrlv80_4x32( vdata, pdata );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
    do
    {
       *noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
diff --git a/algo/lyra2/lyra2z-4way.c b/algo/lyra2/lyra2z-4way.c
index 1123363..cf8ca0c 100644
--- a/algo/lyra2/lyra2z-4way.c
+++ b/algo/lyra2/lyra2z-4way.c
@@ -60,7 +60,7 @@ int scanhash_lyra2z_4way( struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ptarget[7] = 0x0000ff;
 
-   mm128_bswap_intrlv80_4x32( vdata, pdata );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
    lyra2z_4way_midstate( vdata );
 
    do {
@@ -119,8 +119,8 @@ void lyra2z_8way_hash( void *state, const void *input )
      blake256_8way( &ctx_blake, input + (64*8), 16 );
      blake256_8way_close( &ctx_blake, vhash );
 
-     mm256_dintrlv_8x32( hash0, hash1, hash2, hash3,
-                         hash4, hash5, hash6, hash7, vhash, 256 );
+     dintrlv_8x32( hash0, hash1, hash2, hash3,
+                   hash4, hash5, hash6, hash7, vhash, 256 );
 
      LYRA2Z( lyra2z_8way_matrix, hash0, 32, hash0, 32, hash0, 32, 8, 8, 8 );
      LYRA2Z( lyra2z_8way_matrix, hash1, 32, hash1, 32, hash1, 32, 8, 8, 8 );
@@ -146,7 +146,6 @@ int scanhash_lyra2z_8way( struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[8*8] __attribute__ ((aligned (64)));
    uint32_t vdata[20*8] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t Htarg = ptarget[7];
@@ -158,10 +157,7 @@ int scanhash_lyra2z_8way( struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ptarget[7] = 0x0000ff;
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
-                              edata, edata, edata, edata, 640 );
-//   mm256_bswap_intrlv80_8x32( vdata, pdata );
+   mm256_bswap32_intrlv80_8x32( vdata, pdata );
    lyra2z_8way_midstate( vdata );
 
    do {
diff --git a/algo/lyra2/phi2-4way.c b/algo/lyra2/phi2-4way.c
index ff9bff7..1c3f759 100644
--- a/algo/lyra2/phi2-4way.c
+++ b/algo/lyra2/phi2-4way.c
@@ -69,13 +69,13 @@ void phi2_hash_4way( void *state, const void *input )
    LYRA2RE( &hashA[3][0], 32, &hashB[3][0], 32, &hashB[3][0], 32, 1, 8, 8 );
    LYRA2RE( &hashA[3][8], 32, &hashB[3][8], 32, &hashB[3][8], 32, 1, 8, 8 );
 
-   mm256_intrlv_4x64( vhash, hashA[0], hashA[1], hashA[2], hashA[3], 512 );
+   intrlv_4x64( vhash, hashA[0], hashA[1], hashA[2], hashA[3], 512 );
 
    jh512_4way_init( &ctx.jh );
    jh512_4way( &ctx.jh, vhash, 64 );
    jh512_4way_close( &ctx.jh, vhash );
 
-   mm256_dintrlv_4x64( hash[0], hash[1], hash[2], hash[3], vhash, 512 );
+   dintrlv_4x64( hash[0], hash[1], hash[2], hash[3], vhash, 512 );
 
    if ( hash[0][0] & 1 )
   	{
@@ -141,7 +141,7 @@ void phi2_hash_4way( void *state, const void *input )
                           (const BitSequence *)hash[3], 512 );
    }
 
-   mm256_intrlv_4x64( vhash, hash[0], hash[1], hash[2], hash[3], 512 );
+   intrlv_4x64( vhash, hash[0], hash[1], hash[2], hash[3], 512 );
    
    skein512_4way_init( &ctx.skein );
 	skein512_4way( &ctx.skein, vhash, 64 );
@@ -217,7 +217,7 @@ int scanhash_phi2_4way( struct work *work, uint32_t max_nonce,
 
       for ( int lane = 0; lane < 4; lane++ ) if (  hash7[ lane<<1 ] < Htarg )
       {
-          mm256_extr_lane_4x64( lane_hash, hash, lane, 256 );
+          extr_lane_4x64( lane_hash, hash, lane, 256 );
           if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
           {
               pdata[19] = n + lane;
diff --git a/algo/nist5/nist5-4way.c b/algo/nist5/nist5-4way.c
index e3f8fce..1d09c72 100644
--- a/algo/nist5/nist5-4way.c
+++ b/algo/nist5/nist5-4way.c
@@ -12,9 +12,6 @@
 #include "algo/keccak/keccak-hash-4way.h"
 #include "algo/groestl/aes_ni/hash-groestl.h"
 
-// no improvement with midstate
-//static __thread blake512_4way_context ctx_mid;
-
 void nist5hash_4way( void *out, const void *input )
 {
      uint64_t hash0[8] __attribute__ ((aligned (64)));
@@ -28,14 +25,11 @@ void nist5hash_4way( void *out, const void *input )
      skein512_4way_context  ctx_skein;
      keccak512_4way_context ctx_keccak;
 
-//     memcpy( &ctx_blake, &ctx_mid, sizeof(ctx_mid) );
-//     blake512_4way( &ctx_blake, input + (64<<2), 16 );
-
      blake512_4way_init( &ctx_blake );
      blake512_4way( &ctx_blake, input, 80 );
      blake512_4way_close( &ctx_blake, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx_groestl, 64 );
      update_and_final_groestl( &ctx_groestl, (char*)hash0,
@@ -50,7 +44,7 @@ void nist5hash_4way( void *out, const void *input )
      update_and_final_groestl( &ctx_groestl, (char*)hash3,
                                (const char*)hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      jh512_4way_init( &ctx_jh );
      jh512_4way( &ctx_jh, vhash, 64 );
@@ -72,13 +66,12 @@ int scanhash_nist5_4way( struct work *work, uint32_t max_nonce,
      uint32_t *hash7 = &(hash[25]);
      uint32_t lane_hash[8] __attribute__ ((aligned (32)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-     uint32_t endiandata[20] __attribute__((aligned(64)));
      uint32_t *pdata = work->data;
      uint32_t *ptarget = work->target;
      uint32_t n = pdata[19];
      const uint32_t first_nonce = pdata[19];
      const uint32_t Htarg = ptarget[7];
-     uint32_t *noncep = vdata + 73;   // 9*8 + 1
+     __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
      int thr_id = mythr->id;  // thr_id arg is deprecated
 
      uint64_t htmax[] = {          0,
@@ -95,15 +88,7 @@ int scanhash_nist5_4way( struct work *work, uint32_t max_nonce,
                           0xFFFF0000,
                                    0 };
 
-     // we need bigendian data...
-     swab32_array( endiandata, pdata, 20 );
-
-     uint64_t *edata = (uint64_t*)endiandata;
-     mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
-
-     // precalc midstate
-//     blake512_4way_init( &ctx_mid );
-//     blake512_4way( &ctx_mid, vdata, 64 );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
      for ( int m=0; m < 6; m++ )
      {
@@ -112,17 +97,15 @@ int scanhash_nist5_4way( struct work *work, uint32_t max_nonce,
            uint32_t mask = masks[m];
 
            do {
-              be32enc( noncep,   n   );
-              be32enc( noncep+2, n+1 );
-              be32enc( noncep+4, n+2 );
-              be32enc( noncep+6, n+3 );
+              *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
+                 _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
 
               nist5hash_4way( hash, vdata );
 
               for ( int lane = 0; lane < 4; lane++ )
               if ( ( hash7[ lane ] & mask ) == 0 )
               {
-                 mm256_extr_lane_4x64( lane_hash, hash, lane, 256 );
+                 extr_lane_4x64( lane_hash, hash, lane, 256 );
                  if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
                  {
                     pdata[19] = n + lane;
diff --git a/algo/quark/anime-4way.c b/algo/quark/anime-4way.c
index b71398b..c5bdde3 100644
--- a/algo/quark/anime-4way.c
+++ b/algo/quark/anime-4way.c
@@ -62,7 +62,7 @@ void anime_4way_hash( void *state, const void *input )
 
     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
 
-    mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+    dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
     if ( hash0[0] & mask )
     {
@@ -88,7 +88,7 @@ void anime_4way_hash( void *state, const void *input )
                                                (char*)hash3, 512 );
     }
 
-    mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+    intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
 
     if ( mm256_anybits0( vh_mask ) )
     {
@@ -98,7 +98,7 @@ void anime_4way_hash( void *state, const void *input )
 
     mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
-    mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+    dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
     reinit_groestl( &ctx.groestl );
     update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -109,7 +109,7 @@ void anime_4way_hash( void *state, const void *input )
     reinit_groestl( &ctx.groestl );
     update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-    mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+    intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
     jh512_4way( &ctx.jh, vhash, 64 );
     jh512_4way_close( &ctx.jh, vhash );
@@ -155,7 +155,7 @@ void anime_4way_hash( void *state, const void *input )
 
     mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
-    mm256_dintrlv_4x64( state, state+32, state+64, state+96, vhash, 256 );
+    dintrlv_4x64( state, state+32, state+64, state+96, vhash, 256 );
 }
 
 int scanhash_anime_4way( struct work *work, uint32_t max_nonce,
@@ -163,7 +163,6 @@ int scanhash_anime_4way( struct work *work, uint32_t max_nonce,
 {
     uint32_t hash[4*8] __attribute__ ((aligned (64)));
     uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
     uint32_t *pdata = work->data;
     uint32_t *ptarget = work->target;
     uint32_t n = pdata[19];
@@ -188,9 +187,7 @@ int scanhash_anime_4way( struct work *work, uint32_t max_nonce,
                 0
         };
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//    mm256_bswap_intrlv80_4x64( vdata, pdata );
+    mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
     for (int m=0; m < 6; m++)
        if (Htarg <= htmax[m])
diff --git a/algo/quark/hmq1725-4way.c b/algo/quark/hmq1725-4way.c
index 7658bf1..3645f19 100644
--- a/algo/quark/hmq1725-4way.c
+++ b/algo/quark/hmq1725-4way.c
@@ -67,7 +67,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
      bmw512_4way( &ctx.bmw, input, 80 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_whirlpool_init( &ctx.whirlpool );
      sph_whirlpool( &ctx.whirlpool, hash0, 64 );
@@ -84,7 +84,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
 // first fork, A is groestl serial, B is skein parallel.
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
                                    m256_zero );
@@ -116,7 +116,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
                                                (char*)hash3, 512 );
 //     }
 
-     mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
 
 // B
 
@@ -158,7 +158,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
      mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
     
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_luffa( &ctx.luffa, 512 );
      update_and_final_luffa( &ctx.luffa, (BitSequence*)hash0,
@@ -186,7 +186,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
      cubehashUpdateDigest( &ctx.cube, (BitSequence *)hash3,
                                 (const BitSequence *)hash3, 64 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
 // A= keccak parallel, B= jh parallel
     
@@ -209,7 +209,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
      mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_shavite512_init( &ctx.shavite );
      sph_shavite512 ( &ctx.shavite, hash0, 64 );
@@ -240,7 +240,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 // A is whirlpool serial, B is haval parallel.
     
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
                                    m256_zero );
@@ -271,7 +271,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
         sph_whirlpool_close( &ctx.whirlpool, hash3 );
 //     }
 
-     mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
 
 // B
 
@@ -285,7 +285,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
      mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
     
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -300,13 +300,13 @@ extern void hmq1725_4way_hash(void *state, const void *input)
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                              (const BitSequence *)hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
      
      blake512_4way_init( &ctx.blake );
      blake512_4way( &ctx.blake, vhash, 64 );
      blake512_4way_close( &ctx.blake, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
 // shavite & luffa, both serial, select individually.
 
@@ -362,13 +362,13 @@ extern void hmq1725_4way_hash(void *state, const void *input)
                                     (const BitSequence *)hash3, 64 );
    }
 
-   mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+   intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
    hamsi512_4way_init( &ctx.hamsi );
    hamsi512_4way( &ctx.hamsi, vhash, 64 );
    hamsi512_4way_close( &ctx.hamsi, vhash );
 
-   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+   dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
    sph_fugue512_init( &ctx.fugue );
    sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -438,13 +438,13 @@ extern void hmq1725_4way_hash(void *state, const void *input)
                              (const BitSequence *)hash3, 512 );
    }
 
-   mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
+   intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
 
    shabal512_4way_init( &ctx.shabal );
    shabal512_4way( &ctx.shabal, vhash, 64 );
    shabal512_4way_close( &ctx.shabal, vhash );
 
-   mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
+   dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
 
    sph_whirlpool_init( &ctx.whirlpool );
    sph_whirlpool( &ctx.whirlpool, hash0, 64 );
@@ -461,7 +461,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
 // A = fugue serial, B = sha512 prarallel
    
-   mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+   intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
    vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
                                  m256_zero );
@@ -491,7 +491,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
       sph_fugue512_close( &ctx.fugue, hash3 );
 //   }
 
-   mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+   intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
 
 //   if ( mm256_any_clr_256( vh_mask ) )
 //   {
@@ -502,7 +502,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
 
    mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
-   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+   dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
    init_groestl( &ctx.groestl, 64 );
    update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -513,7 +513,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
    init_groestl( &ctx.groestl, 64 );
    update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-   mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+   intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
    sha512_4way_init( &ctx.sha512 ); 
    sha512_4way( &ctx.sha512, vhash, 64 );
@@ -524,7 +524,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
    vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
                                  m256_zero );
 
-   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+   dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
      
 //   if ( mm256_any_set_256( vh_mask ) ) //4
 //   {
@@ -559,7 +559,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
       sph_whirlpool_close( &ctx.whirlpool, hash3 );
 //   }
 
-   mm256_intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, 512 );
+   intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, 512 );
 
    mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
@@ -589,7 +589,7 @@ int scanhash_hmq1725_4way( struct work *work, uint32_t max_nonce,
    uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
                         0xFFFFF000, 0xFFFF0000,          0  };
 
-   mm256_bswap_intrlv80_4x64( vdata, pdata );
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
    for ( int m = 0; m < 6; m++ ) if ( Htarg <= htmax[m] )
    {
       uint32_t mask = masks[ m ];
diff --git a/algo/quark/quark-4way.c b/algo/quark/quark-4way.c
index 5e22c67..9c0fb5d 100644
--- a/algo/quark/quark-4way.c
+++ b/algo/quark/quark-4way.c
@@ -63,7 +63,7 @@ void quark_4way_hash( void *state, const void *input )
 
     vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
 
-    mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+    dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
     if ( hash0[0] & mask )
     {
@@ -89,7 +89,7 @@ void quark_4way_hash( void *state, const void *input )
                                                (char*)hash3, 512 );
     }
 
-    mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
+    intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
 
     if ( mm256_anybits0( vh_mask ) )   
     {
@@ -99,7 +99,7 @@ void quark_4way_hash( void *state, const void *input )
 
     mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
 
-    mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+    dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
     reinit_groestl( &ctx.groestl );
     update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -110,7 +110,7 @@ void quark_4way_hash( void *state, const void *input )
     reinit_groestl( &ctx.groestl );
     update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-    mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+    intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
     jh512_4way( &ctx.jh, vhash, 64 );
     jh512_4way_close( &ctx.jh, vhash );
@@ -168,7 +168,6 @@ int scanhash_quark_4way( struct work *work, uint32_t max_nonce,
 {
     uint32_t hash[4*8] __attribute__ ((aligned (64)));
     uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
     uint32_t lane_hash[8] __attribute__ ((aligned (64)));
     uint32_t *hash7 = &(hash[25]);
     uint32_t *pdata = work->data;
@@ -178,9 +177,7 @@ int scanhash_quark_4way( struct work *work, uint32_t max_nonce,
     __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
     int thr_id = mythr->id;  // thr_id arg is deprecated
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//    mm256_bswap_intrlv80_4x64( vdata, pdata );
+    mm256_bswap32_intrlv80_4x64( vdata, pdata );
     do
     {
        *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
@@ -192,7 +189,7 @@ int scanhash_quark_4way( struct work *work, uint32_t max_nonce,
        for ( int i = 0; i < 4; i++ )
        if ( ( hash7[ i<<1 ] & 0xFFFFFF00 ) == 0 )
        {
-          mm256_extr_lane_4x64( lane_hash, hash, i, 256 );
+          extr_lane_4x64( lane_hash, hash, i, 256 );
           if ( fulltest( lane_hash, ptarget ) && !opt_benchmark  )
           {
             pdata[19] = n+i;
diff --git a/algo/qubit/deep-2way.c b/algo/qubit/deep-2way.c
index 0060932..9ca6608 100644
--- a/algo/qubit/deep-2way.c
+++ b/algo/qubit/deep-2way.c
@@ -39,7 +39,7 @@ void deep_2way_hash( void *output, const void *input )
      memcpy( &ctx, &deep_2way_ctx, sizeof(deep_2way_ctx) );
      luffa_2way_update( &ctx.luffa, input + (64<<1), 16 );
      luffa_2way_close( &ctx.luffa, vhash );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
 
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0,
                            (const byte*) hash0, 64 );
@@ -86,7 +86,7 @@ int scanhash_deep_2way( struct work *work,uint32_t max_nonce,
      casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
 
      uint64_t *edata = (uint64_t*)endiandata;
-     mm256_intrlv_2x128( (uint64_t*)vdata, edata, edata, 640 );
+     intrlv_2x128( (uint64_t*)vdata, edata, edata, 640 );
 
      luffa_2way_init( &deep_2way_ctx.luffa, 512 );
      luffa_2way_update( &deep_2way_ctx.luffa, vdata, 64 );
diff --git a/algo/qubit/qubit-2way.c b/algo/qubit/qubit-2way.c
index 038cec0..8dc04b9 100644
--- a/algo/qubit/qubit-2way.c
+++ b/algo/qubit/qubit-2way.c
@@ -41,7 +41,7 @@ void qubit_2way_hash( void *output, const void *input )
      memcpy( &ctx, &qubit_2way_ctx, sizeof(qubit_2way_ctx) );
      luffa_2way_update( &ctx.luffa, input + (64<<1), 16 );
      luffa_2way_close( &ctx.luffa, vhash );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
 
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0,
                            (const byte*) hash0, 64 );
@@ -55,9 +55,9 @@ void qubit_2way_hash( void *output, const void *input )
      sph_shavite512( &ctx.shavite, hash1, 64 );
      sph_shavite512_close( &ctx.shavite, hash1 );
 
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
 
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
                        (const BitSequence *) hash0, 512 );
@@ -92,7 +92,7 @@ int scanhash_qubit_2way( struct work *work,uint32_t max_nonce,
      casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
 
      uint64_t *edata = (uint64_t*)endiandata;
-     mm256_intrlv_2x128( (uint64_t*)vdata, edata, edata, 640 );
+     intrlv_2x128( (uint64_t*)vdata, edata, edata, 640 );
 
      luffa_2way_init( &qubit_2way_ctx.luffa, 512 );
      luffa_2way_update( &qubit_2way_ctx.luffa, vdata, 64 );
diff --git a/algo/ripemd/lbry-4way.c b/algo/ripemd/lbry-4way.c
index 3e3fa7d..9ac5f53 100644
--- a/algo/ripemd/lbry-4way.c
+++ b/algo/ripemd/lbry-4way.c
@@ -40,9 +40,9 @@ void lbry_8way_hash( void* output, const void* input )
    sha256_8way_close( &ctx_sha256, vhashA );
 
    // reinterleave to do sha512 4-way 64 bit twice.
-   mm256_dintrlv_8x32( h0, h1, h2, h3, h4, h5, h6, h7, vhashA, 256 );
-   mm256_intrlv_4x64( vhashA, h0, h1, h2, h3, 256 );
-   mm256_intrlv_4x64( vhashB, h4, h5, h6, h7, 256 );
+   dintrlv_8x32( h0, h1, h2, h3, h4, h5, h6, h7, vhashA, 256 );
+   intrlv_4x64( vhashA, h0, h1, h2, h3, 256 );
+   intrlv_4x64( vhashB, h4, h5, h6, h7, 256 );
 
    sha512_4way_init( &ctx_sha512 );
    sha512_4way( &ctx_sha512, vhashA, 32 );
@@ -53,9 +53,9 @@ void lbry_8way_hash( void* output, const void* input )
    sha512_4way_close( &ctx_sha512, vhashB );
 
    // back to 8-way 32 bit
-   mm256_dintrlv_4x64( h0, h1, h2, h3, vhashA, 512 );
-   mm256_dintrlv_4x64( h4, h5, h6, h7, vhashB, 512 );
-   mm256_intrlv_8x32( vhashA, h0, h1, h2, h3, h4, h5, h6, h7, 512 );
+   dintrlv_4x64( h0, h1, h2, h3, vhashA, 512 );
+   dintrlv_4x64( h4, h5, h6, h7, vhashB, 512 );
+   intrlv_8x32( vhashA, h0, h1, h2, h3, h4, h5, h6, h7, 512 );
 
    ripemd160_8way_init( &ctx_ripemd );
    ripemd160_8way( &ctx_ripemd, vhashA, 32 );
@@ -97,11 +97,15 @@ int scanhash_lbry_8way( struct work *work, uint32_t max_nonce,
                         0xFFFFF000, 0xFFFF0000,          0 };
 
    // we need bigendian data...
-   casti_m256i( edata, 0 ) = mm256_bswap_32( casti_m256i( pdata, 0 ) );
-   casti_m256i( edata, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
-   casti_m256i( edata, 2 ) = mm256_bswap_32( casti_m256i( pdata, 2 ) );
-   casti_m256i( edata, 3 ) = mm256_bswap_32( casti_m256i( pdata, 3 ) );
-   mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
+   casti_m128i( edata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
+   casti_m128i( edata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
+   casti_m128i( edata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
+   casti_m128i( edata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
+   casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
+   casti_m128i( edata, 5 ) = mm128_bswap_32( casti_m128i( pdata, 5 ) );
+   casti_m128i( edata, 6 ) = mm128_bswap_32( casti_m128i( pdata, 6 ) );
+   casti_m128i( edata, 7 ) = mm128_bswap_32( casti_m128i( pdata, 7 ) );
+   intrlv_8x32( vdata, edata, edata, edata, edata,
                              edata, edata, edata, edata, 1024 );
    sha256_8way_init( &sha256_8w_mid );
    sha256_8way( &sha256_8w_mid, vdata, LBRY_MIDSTATE );
@@ -118,7 +122,7 @@ int scanhash_lbry_8way( struct work *work, uint32_t max_nonce,
          for ( int i = 0; i < 8; i++ )  if ( !( hash7[ i ] & mask ) )
          {
             // deinterleave hash for lane
-            mm256_extr_lane_8x32( lane_hash, hash, i, 256 );
+            extr_lane_8x32( lane_hash, hash, i, 256 );
             if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
             {
               pdata[27] = n + i;
diff --git a/algo/sha/md-helper-4way.c b/algo/sha/md-helper-4way.c
index eb5c05c..e67a7c5 100644
--- a/algo/sha/md-helper-4way.c
+++ b/algo/sha/md-helper-4way.c
@@ -196,9 +196,9 @@ SPH_XCAT( HASH, _addbits_and_close )(void *cc, 	unsigned ub, unsigned n,
     ptr = (unsigned)sc->count & (SPH_BLEN - 1U);
 
 #ifdef PW01
-    sc->buf[ptr>>3] = _mm256_set1_epi64x( 0x100 >> 8 );
+    sc->buf[ptr>>3] = m256_const1_64( 0x100 >> 8 );
 #else
-    sc->buf[ptr>>3] = _mm256_set1_epi64x( 0x80 );
+    sc->buf[ptr>>3] = m256_const1_64( 0x80 );
 #endif
     ptr += 8;
 
diff --git a/algo/sha/sha2-hash-4way.c b/algo/sha/sha2-hash-4way.c
index d3e5150..9516543 100644
--- a/algo/sha/sha2-hash-4way.c
+++ b/algo/sha/sha2-hash-4way.c
@@ -660,7 +660,7 @@ void sha512_4way_close( sha512_4way_context *sc, void *dst )
     const int pad = buf_size - 16;
 
     ptr = (unsigned)sc->count & (buf_size - 1U);
-    sc->buf[ ptr>>3 ] = _mm256_set1_epi64x( 0x80 );
+    sc->buf[ ptr>>3 ] = m256_const1_64( 0x80 );
     ptr += 8;
     if ( ptr > pad )
     {
diff --git a/algo/sha/sha256q-4way.c b/algo/sha/sha256q-4way.c
index 689577d..cc47b1c 100644
--- a/algo/sha/sha256q-4way.c
+++ b/algo/sha/sha256q-4way.c
@@ -36,7 +36,6 @@ int scanhash_sha256q_8way( struct work *work, uint32_t max_nonce,
 {
    uint32_t vdata[20*8] __attribute__ ((aligned (64)));
    uint32_t hash[8*8] __attribute__ ((aligned (32)));
-   uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t lane_hash[8] __attribute__ ((aligned (32)));
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
@@ -60,10 +59,7 @@ int scanhash_sha256q_8way( struct work *work, uint32_t max_nonce,
                                         0 };
 
    // Need big endian data
-   swab32_array( edata, pdata, 20 );
-   mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
-                             edata, edata, edata, edata, 640 );
-//   mm256_bswap_intrlv80_8x32( vdata, pdata );
+   mm256_bswap32_intrlv80_8x32( vdata, pdata );
    sha256_8way_init( &sha256_ctx8 );
    sha256_8way( &sha256_ctx8, vdata, 64 );
 
@@ -84,7 +80,7 @@ int scanhash_sha256q_8way( struct work *work, uint32_t max_nonce,
          if ( !( hash7[ lane ] & mask ) )
          { 
             // deinterleave hash for lane
-	         mm256_extr_lane_8x32( lane_hash, hash, lane, 256 );
+	         extr_lane_8x32( lane_hash, hash, lane, 256 );
 
 	         if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
             {
@@ -133,7 +129,6 @@ int scanhash_sha256q_4way( struct work *work, uint32_t max_nonce,
 {
    uint32_t vdata[20*4] __attribute__ ((aligned (64)));
    uint32_t hash[8*4] __attribute__ ((aligned (32)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t *hash7 = &(hash[7<<2]);
    uint32_t lane_hash[8] __attribute__ ((aligned (32)));
    uint32_t *pdata = work->data;
@@ -157,9 +152,7 @@ int scanhash_sha256q_4way( struct work *work, uint32_t max_nonce,
                                0xFFFF0000,
                                         0 };
 
-   swab32_array( edata, pdata, 20 );
-   mm128_intrlv_4x32( vdata, edata, edata, edata, edata, 640 );
-//   mm128_bswap_intrlv80_4x32( vdata, pdata );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
    sha256_4way_init( &sha256_ctx4 );
    sha256_4way( &sha256_ctx4, vdata, 64 );
 
diff --git a/algo/sha/sha256t-4way.c b/algo/sha/sha256t-4way.c
index e6e404e..c17ea1d 100644
--- a/algo/sha/sha256t-4way.c
+++ b/algo/sha/sha256t-4way.c
@@ -72,7 +72,7 @@ int scanhash_sha256t_11way( struct work *work, uint32_t max_nonce,
    casti_m256i( dataz, 1 ) = mm256_bswap_32( casti_m256i( pdata, 1 ) );
    casti_m128i( dataz, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
 
-   mm256_intrlv_8x32( datax, dataz, dataz, dataz, dataz,
+   intrlv_8x32( datax, dataz, dataz, dataz, dataz,
                                  dataz, dataz, dataz, dataz, 640 );
    mm64_interleave_2x32( datay, dataz, dataz, 640 );
 
@@ -99,7 +99,7 @@ int scanhash_sha256t_11way( struct work *work, uint32_t max_nonce,
         for ( i = 0; i < 8; i++ ) if ( !( hash7[ i ] & mask ) )
         { 
             // deinterleave hash for lane
-            mm256_extr_lane_8x32( lane_hash, hashx, i, 256 );
+            extr_lane_8x32( lane_hash, hashx, i, 256 );
             if ( fulltest( lane_hash, ptarget ) )
             {
 	            pdata[19] = n + i;
@@ -163,7 +163,6 @@ int scanhash_sha256t_8way( struct work *work, uint32_t max_nonce,
 {
    uint32_t vdata[20*8]  __attribute__ ((aligned (64)));
    uint32_t hash[8*8]    __attribute__ ((aligned (32)));
-   uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t lane_hash[8] __attribute__ ((aligned (32)));
    uint32_t *hash7 = &(hash[7<<3]);
    uint32_t *pdata = work->data;
@@ -187,12 +186,9 @@ int scanhash_sha256t_8way( struct work *work, uint32_t max_nonce,
                                0xFFFF0000,
                                         0 };
 
-   swab32_array( edata, pdata, 20 );
-   mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
-                             edata, edata, edata, edata, 640 );
 
    // Need big endian data
-//   mm256_bswap_intrlv80_8x32( vdata, pdata );
+   mm256_bswap32_intrlv80_8x32( vdata, pdata );
    sha256_8way_init( &sha256_ctx8 );
    sha256_8way( &sha256_ctx8, vdata, 64 );
 
@@ -209,7 +205,7 @@ int scanhash_sha256t_8way( struct work *work, uint32_t max_nonce,
          if ( !( hash7[ lane ] & mask ) )
          {
             // deinterleave hash for lane
-            mm256_extr_lane_8x32( lane_hash, hash, lane, 256 );
+            extr_lane_8x32( lane_hash, hash, lane, 256 );
             if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
             {
               pdata[19] = n + lane;
@@ -253,7 +249,6 @@ int scanhash_sha256t_4way( struct work *work, uint32_t max_nonce,
 {
    uint32_t vdata[20*4] __attribute__ ((aligned (64)));
    uint32_t hash[8*4] __attribute__ ((aligned (32)));
-   uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t lane_hash[8] __attribute__ ((aligned (64)));
    uint32_t *hash7 = &(hash[7<<2]);
    uint32_t *pdata = work->data;
@@ -277,10 +272,7 @@ int scanhash_sha256t_4way( struct work *work, uint32_t max_nonce,
                               0xFFFF0000,
                                        0 };
 
-   swab32_array( edata, pdata, 20 );
-   mm128_intrlv_4x32( vdata, edata, edata, edata, edata, 640 );
-
-//   mm128_bswap_intrlv80_4x32( vdata, pdata );
+   mm128_bswap32_intrlv80_4x32( vdata, pdata );
    sha256_4way_init( &sha256_ctx4 );
    sha256_4way( &sha256_ctx4, vdata, 64 );
 
diff --git a/algo/sha/sha256t-gate.c b/algo/sha/sha256t-gate.c
index 91e8e83..0271234 100644
--- a/algo/sha/sha256t-gate.c
+++ b/algo/sha/sha256t-gate.c
@@ -11,7 +11,7 @@ bool register_sha256t_algo( algo_gate_t* gate )
     gate->scanhash   = (void*)&scanhash_sha256t_4way;
     gate->hash       = (void*)&sha256t_4way_hash;
 #else
-gate->optimizations = SHA_OPT;
+    gate->optimizations = SHA_OPT;
     gate->scanhash   = (void*)&scanhash_sha256t;
     gate->hash       = (void*)&sha256t_hash;
 #endif
@@ -21,7 +21,11 @@ gate->optimizations = SHA_OPT;
 
 bool register_sha256q_algo( algo_gate_t* gate )
 {
-#if defined(SHA256T_4WAY)
+#if defined(SHA256T_8WAY)
+    gate->optimizations = SSE2_OPT | AVX2_OPT | SHA_OPT;
+    gate->scanhash   = (void*)&scanhash_sha256q_8way;
+    gate->hash       = (void*)&sha256q_8way_hash;
+#elif defined(SHA256T_4WAY)
     gate->optimizations = SSE2_OPT | AVX2_OPT | SHA_OPT;
     gate->scanhash   = (void*)&scanhash_sha256q_4way;
     gate->hash       = (void*)&sha256q_4way_hash;
diff --git a/algo/sha/sha256t-gate.h b/algo/sha/sha256t-gate.h
index 30b2766..0d519aa 100644
--- a/algo/sha/sha256t-gate.h
+++ b/algo/sha/sha256t-gate.h
@@ -8,7 +8,7 @@
 #if !defined(__SHA__)
  #if defined(__AVX2__)
   #define SHA256T_8WAY
-#elif defined(__SSE2__)
+ #elif defined(__SSE2__)
   #define SHA256T_4WAY
  #endif
 #endif
diff --git a/algo/skein/skein-4way.c b/algo/skein/skein-4way.c
index 656adb5..d600e60 100644
--- a/algo/skein/skein-4way.c
+++ b/algo/skein/skein-4way.c
@@ -12,7 +12,7 @@
 
 void skeinhash_4way( void *state, const void *input )
 {
-     uint64_t vhash64[8*4] __attribute__ ((aligned (64)));
+     uint64_t vhash64[16*4] __attribute__ ((aligned (64)));
      skein512_4way_context ctx_skein;
 #if defined(__SHA__)
      uint32_t hash0[16] __attribute__ ((aligned (64)));
@@ -30,7 +30,7 @@ void skeinhash_4way( void *state, const void *input )
      skein512_4way_close( &ctx_skein, vhash64 );
 
 #if defined(__SHA__)      
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 512 );
 
      SHA256_Init( &ctx_sha256 );
      SHA256_Update( &ctx_sha256, (unsigned char*)hash0, 64 );
@@ -50,7 +50,7 @@ void skeinhash_4way( void *state, const void *input )
 
      intrlv_4x32( state, hash0, hash1, hash2, hash3, 256 );
 #else
-     mm256_rintrlv_4x64_4x32( vhash32, vhash64, 512 );
+     rintrlv_4x64_4x32( vhash32, vhash64, 512 );
 
      sha256_4way_init( &ctx_sha256 );
      sha256_4way( &ctx_sha256, vhash32, 64 );
@@ -62,8 +62,7 @@ int scanhash_skein_4way( struct work *work, uint32_t max_nonce,
                          uint64_t *hashes_done, struct thr_info *mythr )
 {
     uint32_t vdata[20*4] __attribute__ ((aligned (64)));
-    uint32_t hash[8*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
+    uint32_t hash[16*4] __attribute__ ((aligned (64)));
     uint32_t lane_hash[8] __attribute__ ((aligned (32)));
     uint32_t *hash7 = &(hash[7<<2]);
     uint32_t *pdata = work->data;
@@ -74,9 +73,7 @@ int scanhash_skein_4way( struct work *work, uint32_t max_nonce,
     __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
     int thr_id = mythr->id;  // thr_id arg is deprecated
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//   mm256_bswap_intrlv80_4x64( vdata, pdata );
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
    do
    {
        *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
diff --git a/algo/skein/skein-hash-4way.c b/algo/skein/skein-hash-4way.c
index ef8beb8..358ecd8 100644
--- a/algo/skein/skein-hash-4way.c
+++ b/algo/skein/skein-hash-4way.c
@@ -279,10 +279,7 @@ do { \
                                               _mm256_xor_si256( k2, k3 ) ), \
                             _mm256_xor_si256( _mm256_xor_si256( k4, k5 ), \
                                               _mm256_xor_si256( k6, k7 ) ) ), \
-                         _mm256_set_epi64x( SPH_C64(0x1BD11BDAA9FC1A22), \
-                                            SPH_C64(0x1BD11BDAA9FC1A22), \
-                                            SPH_C64(0x1BD11BDAA9FC1A22), \
-                                            SPH_C64(0x1BD11BDAA9FC1A22) ) ); \
+                         m256_const1_64( 0x1BD11BDAA9FC1A22) ); \
   t2 = t0 ^ t1; \
 } while (0)
 
@@ -294,13 +291,11 @@ do { \
   w3 = _mm256_add_epi64( w3, SKBI(k,s,3) ); \
   w4 = _mm256_add_epi64( w4, SKBI(k,s,4) ); \
   w5 = _mm256_add_epi64( w5, _mm256_add_epi64( SKBI(k,s,5), \
-                           _mm256_set_epi64x( SKBT(t,s,0), SKBT(t,s,0), \
-                                              SKBT(t,s,0), SKBT(t,s,0) ) ) ); \
+                                         m256_const1_64( SKBT(t,s,0) ) ) ); \
   w6 = _mm256_add_epi64( w6, _mm256_add_epi64( SKBI(k,s,6), \
-                           _mm256_set_epi64x( SKBT(t,s,1), SKBT(t,s,1), \
-                                              SKBT(t,s,1), SKBT(t,s,1) ) ) ); \
+                                         m256_const1_64( SKBT(t,s,1) ) ) ); \
   w7 = _mm256_add_epi64( w7, _mm256_add_epi64( SKBI(k,s,7), \
-                                      _mm256_set_epi64x( s, s, s, s ) ) ); \
+                                         m256_const1_64( s ) ) ); \
 } while (0)
 
 
diff --git a/algo/skein/skein2-4way.c b/algo/skein/skein2-4way.c
index 98680d4..b67fa78 100644
--- a/algo/skein/skein2-4way.c
+++ b/algo/skein/skein2-4way.c
@@ -8,7 +8,7 @@
 void skein2hash_4way( void *output, const void *input )
 {
    skein512_4way_context ctx;
-   uint64_t hash[8*4] __attribute__ ((aligned (64)));
+   uint64_t hash[16*4] __attribute__ ((aligned (64)));
 
    skein512_4way_init( &ctx );
    skein512_4way( &ctx, input, 80 );
@@ -22,8 +22,7 @@ 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 )
 {
-    uint32_t hash[8*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
+    uint32_t hash[16*4] __attribute__ ((aligned (64)));
     uint32_t vdata[20*4] __attribute__ ((aligned (64)));
     uint32_t lane_hash[8] __attribute__ ((aligned (64)));
     uint32_t *hash7 = &(hash[25]);
@@ -34,20 +33,10 @@ int scanhash_skein2_4way( struct work *work, uint32_t max_nonce,
     uint32_t n = first_nonce;
     __m256i  *noncev = (__m256i*)vdata + 9;   // aligned
     int thr_id = mythr->id;  // thr_id arg is deprecated
-//    uint32_t *noncep = vdata + 73;   // 9*8 + 1
 
-
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-
-//    mm256_bswap_intrlv80_4x64( vdata, pdata );
+    mm256_bswap32_intrlv80_4x64( vdata, pdata );
     do 
     {
-//       be32enc( noncep,   n   );
-//       be32enc( noncep+2, n+1 );
-//       be32enc( noncep+4, n+2 );
-//       be32enc( noncep+6, n+3 );
-
        *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
                 _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
 
@@ -56,7 +45,7 @@ int scanhash_skein2_4way( struct work *work, uint32_t max_nonce,
        for ( int lane = 0; lane < 4; lane++ )
        if ( hash7[ lane<<1 ] <= Htarg )
        {
-          mm256_extr_lane_4x64( lane_hash, hash, lane, 256 );
+          extr_lane_4x64( lane_hash, hash, lane, 256 );
           if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
           {
              pdata[19] = n + lane;
diff --git a/algo/x11/c11-4way.c b/algo/x11/c11-4way.c
index f82dec3..e2922bc 100644
--- a/algo/x11/c11-4way.c
+++ b/algo/x11/c11-4way.c
@@ -69,7 +69,7 @@ void c11_4way_hash( void *state, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 3 Groestl
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -81,7 +81,7 @@ void c11_4way_hash( void *state, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
      // 4way
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      // 4 JH
      jh512_4way( &ctx.jh, vhash, 64 );
@@ -96,16 +96,16 @@ void c11_4way_hash( void *state, const void *input )
      skein512_4way_close( &ctx.skein, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 7 Luffa
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
-     mm256_intrlv_2x128( vhashB, hash2, hash3, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhashB, hash2, hash3, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhashB, 512 );
 
      // 8 Cubehash
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
@@ -133,13 +133,13 @@ void c11_4way_hash( void *state, const void *input )
      sph_shavite512_close( &ctx.shavite, hash3 );
 
      // 10 Simd
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
-     mm256_intrlv_2x128( vhashB, hash2, hash3, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhashB, hash2, hash3, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhashB, 512 );
 
      // 11 Echo
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -165,7 +165,6 @@ int scanhash_c11_4way( struct work *work, uint32_t max_nonce,
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-     uint32_t edata[20] __attribute__((aligned(64)));
      uint32_t *pdata = work->data;
      uint32_t *ptarget = work->target;
      uint32_t n = pdata[19];
@@ -178,9 +177,7 @@ int scanhash_c11_4way( struct work *work, uint32_t max_nonce,
      uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//     mm256_bswap_intrlv80_4x64( vdata, pdata );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
      for (int m=0; m < 6; m++) 
        if (Htarg <= htmax[m])
diff --git a/algo/x11/timetravel-4way.c b/algo/x11/timetravel-4way.c
index 155ad11..ba3199c 100644
--- a/algo/x11/timetravel-4way.c
+++ b/algo/x11/timetravel-4way.c
@@ -87,19 +87,16 @@ void timetravel_4way_hash(void *output, const void *input)
            blake512_4way( &ctx.blake, vhashA, dataLen );
            blake512_4way_close( &ctx.blake, vhashB );
            if ( i == 7 )
-              mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                       vhashB, dataLen<<3 );
+              dintrlv_4x64( hash0, hash1, hash2, hash3, vhashB, dataLen<<3 );
         break;
         case 1:
            bmw512_4way( &ctx.bmw, vhashA, dataLen );
            bmw512_4way_close( &ctx.bmw, vhashB );
            if ( i == 7 )
-              mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                       vhashB, dataLen<<3 );
+              dintrlv_4x64( hash0, hash1, hash2, hash3, vhashB, dataLen<<3 );
         break;
         case 2:
-           mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                    vhashA, dataLen<<3 );
+           dintrlv_4x64( hash0, hash1, hash2, hash3, vhashA, dataLen<<3 );
            update_and_final_groestl( &ctx.groestl, (char*)hash0,
                                                    (char*)hash0, dataLen<<3 );
            reinit_groestl( &ctx.groestl );
@@ -112,47 +109,40 @@ void timetravel_4way_hash(void *output, const void *input)
            update_and_final_groestl( &ctx.groestl, (char*)hash3,
                                                    (char*)hash3, dataLen<<3 );
            if ( i != 7 )
-              mm256_intrlv_4x64( vhashB,
-                                     hash0, hash1, hash2, hash3, dataLen<<3 );
+              intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, dataLen<<3 );
         break;
         case 3:
            skein512_4way( &ctx.skein, vhashA, dataLen );
            skein512_4way_close( &ctx.skein, vhashB );
            if ( i == 7 )
-              mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                       vhashB, dataLen<<3 );
+              dintrlv_4x64( hash0, hash1, hash2, hash3, vhashB, dataLen<<3 );
         break;
         case 4:
            jh512_4way( &ctx.jh, vhashA, dataLen );
            jh512_4way_close( &ctx.jh, vhashB );
            if ( i == 7 )
-              mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                       vhashB, dataLen<<3 );
+              dintrlv_4x64( hash0, hash1, hash2, hash3, vhashB, dataLen<<3 );
         break;
         case 5:
            keccak512_4way( &ctx.keccak, vhashA, dataLen );
            keccak512_4way_close( &ctx.keccak, vhashB );
            if ( i == 7 )
-              mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                       vhashB, dataLen<<3 );
+              dintrlv_4x64( hash0, hash1, hash2, hash3, vhashB, dataLen<<3 );
         break;
         case 6:
-           mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                    vhashA, dataLen<<3 );
-           mm256_intrlv_2x128( vhashA, hash0, hash1, dataLen<<3 );
+           dintrlv_4x64( hash0, hash1, hash2, hash3, vhashA, dataLen<<3 );
+           intrlv_2x128( vhashA, hash0, hash1, dataLen<<3 );
            luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, dataLen );
-           mm256_dintrlv_2x128( hash0, hash1, vhashA, dataLen<<3 );
-           mm256_intrlv_2x128( vhashA, hash2, hash3, dataLen<<3 );
+           dintrlv_2x128( hash0, hash1, vhashA, dataLen<<3 );
+           intrlv_2x128( vhashA, hash2, hash3, dataLen<<3 );
            luffa_2way_init( &ctx.luffa, 512 );
            luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, dataLen );
-           mm256_dintrlv_2x128( hash2, hash3, vhashA, dataLen<<3 );
+           dintrlv_2x128( hash2, hash3, vhashA, dataLen<<3 );
            if ( i != 7 )           
-              mm256_intrlv_4x64( vhashB,
-                                     hash0, hash1, hash2, hash3, dataLen<<3 );
+              intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, dataLen<<3 );
         break;
         case 7:
-           mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                    vhashA, dataLen<<3 );
+           dintrlv_4x64( hash0, hash1, hash2, hash3, vhashA, dataLen<<3 );
            cubehashUpdateDigest( &ctx.cube, (byte*)hash0,
                                       (const byte*)hash0, dataLen );
            memcpy( &ctx.cube, &tt8_4way_ctx.cube, sizeof(cubehashParam) );
@@ -165,8 +155,7 @@ void timetravel_4way_hash(void *output, const void *input)
            cubehashUpdateDigest( &ctx.cube, (byte*)hash3,
                                       (const byte*)hash3, dataLen );
            if ( i != 7 )           
-              mm256_intrlv_4x64( vhashB,
-                                     hash0, hash1, hash2, hash3, dataLen<<3 );
+              intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, dataLen<<3 );
         break;
         default:
            applog(LOG_ERR,"SWERR: timetravel invalid permutation");
@@ -215,7 +204,7 @@ int scanhash_timetravel_4way( struct work *work, uint32_t max_nonce,
    }
 
    uint64_t *edata = (uint64_t*)endiandata;
-   mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+   intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
 
    do
    {
diff --git a/algo/x11/timetravel10-4way.c b/algo/x11/timetravel10-4way.c
index 463c3a9..5dab3c8 100644
--- a/algo/x11/timetravel10-4way.c
+++ b/algo/x11/timetravel10-4way.c
@@ -93,19 +93,16 @@ void timetravel10_4way_hash(void *output, const void *input)
            blake512_4way( &ctx.blake, vhashA, dataLen );
            blake512_4way_close( &ctx.blake, vhashB );
            if ( i == 9 )
-              mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                       vhashB, dataLen<<3 );
+              dintrlv_4x64( hash0, hash1, hash2, hash3, vhashB, dataLen<<3 );
         break;
         case 1:
            bmw512_4way( &ctx.bmw, vhashA, dataLen );
            bmw512_4way_close( &ctx.bmw, vhashB );
            if ( i == 9 )
-              mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                       vhashB, dataLen<<3 );
+              dintrlv_4x64( hash0, hash1, hash2, hash3, vhashB, dataLen<<3 );
         break;
         case 2:
-           mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                    vhashA, dataLen<<3 );
+           dintrlv_4x64( hash0, hash1, hash2, hash3, vhashA, dataLen<<3 );
            update_and_final_groestl( &ctx.groestl, (char*)hash0,
                                                    (char*)hash0, dataLen<<3 );
            reinit_groestl( &ctx.groestl );
@@ -118,46 +115,40 @@ void timetravel10_4way_hash(void *output, const void *input)
            update_and_final_groestl( &ctx.groestl, (char*)hash3,
                                                    (char*)hash3, dataLen<<3 );
            if ( i != 9 )
-              mm256_intrlv_4x64( vhashB,
-                                     hash0, hash1, hash2, hash3, dataLen<<3 );
+              intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, dataLen<<3 );
         break;
         case 3:
            skein512_4way( &ctx.skein, vhashA, dataLen );
            skein512_4way_close( &ctx.skein, vhashB );
            if ( i == 9 )
-              mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                       vhashB, dataLen<<3 );
+              dintrlv_4x64( hash0, hash1, hash2, hash3, vhashB, dataLen<<3 );
         break;
         case 4:
            jh512_4way( &ctx.jh, vhashA, dataLen );
            jh512_4way_close( &ctx.jh, vhashB );
            if ( i == 9 )
-              mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                       vhashB, dataLen<<3 );
+              dintrlv_4x64( hash0, hash1, hash2, hash3, vhashB, dataLen<<3 );
         break;
         case 5:
            keccak512_4way( &ctx.keccak, vhashA, dataLen );
            keccak512_4way_close( &ctx.keccak, vhashB );
            if ( i == 9 )
-              mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                       vhashB, dataLen<<3 );
+              dintrlv_4x64( hash0, hash1, hash2, hash3, vhashB, dataLen<<3 );
         break;
         case 6:
-           mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                    vhashA, dataLen<<3 );
-           mm256_intrlv_2x128( vhashA, hash0, hash1, dataLen<<3 );
+           dintrlv_4x64( hash0, hash1, hash2, hash3, vhashA, dataLen<<3 );
+           intrlv_2x128( vhashA, hash0, hash1, dataLen<<3 );
            luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, dataLen );
-           mm256_dintrlv_2x128( hash0, hash1, vhashA, dataLen<<3 );
-           mm256_intrlv_2x128( vhashA, hash2, hash3, dataLen<<3 );
+           dintrlv_2x128( hash0, hash1, vhashA, dataLen<<3 );
+           intrlv_2x128( vhashA, hash2, hash3, dataLen<<3 );
            luffa_2way_init( &ctx.luffa, 512 );
            luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, dataLen );
-           mm256_dintrlv_2x128( hash2, hash3, vhashA, dataLen<<3 );
+           dintrlv_2x128( hash2, hash3, vhashA, dataLen<<3 );
            if ( i != 9 )           
-              mm256_intrlv_4x64( vhashB,
-                                     hash0, hash1, hash2, hash3, dataLen<<3 );
+              intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, dataLen<<3 );
         break;
         case 7:
-           mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
+           dintrlv_4x64( hash0, hash1, hash2, hash3,
                                     vhashA, dataLen<<3 );
            cubehashUpdateDigest( &ctx.cube, (byte*)hash0,
                                       (const byte*)hash0, dataLen );
@@ -171,12 +162,10 @@ void timetravel10_4way_hash(void *output, const void *input)
            cubehashUpdateDigest( &ctx.cube, (byte*)hash3,
                                       (const byte*)hash3, dataLen );
            if ( i != 9 )           
-              mm256_intrlv_4x64( vhashB,
-                                     hash0, hash1, hash2, hash3, dataLen<<3 );
+              intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, dataLen<<3 );
         break;
         case 8:
-           mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                    vhashA, dataLen<<3 );
+           dintrlv_4x64( hash0, hash1, hash2, hash3, vhashA, dataLen<<3 );
            sph_shavite512( &ctx.shavite, hash0, dataLen );
            sph_shavite512_close( &ctx.shavite, hash0 );
            memcpy( &ctx.shavite, &tt10_4way_ctx.shavite, sizeof ctx.shavite );
@@ -189,22 +178,19 @@ void timetravel10_4way_hash(void *output, const void *input)
            sph_shavite512( &ctx.shavite, hash3, dataLen );
            sph_shavite512_close( &ctx.shavite, hash3 );
            if ( i != 9 )
-              mm256_intrlv_4x64( vhashB,
-                                     hash0, hash1, hash2, hash3, dataLen<<3 );
+              intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, dataLen<<3 );
         break;
         case 9:
-           mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                    vhashA, dataLen<<3 );
-           mm256_intrlv_2x128( vhashA, hash0, hash1, dataLen<<3 );
+           dintrlv_4x64( hash0, hash1, hash2, hash3, vhashA, dataLen<<3 );
+           intrlv_2x128( vhashA, hash0, hash1, dataLen<<3 );
            simd_2way_update_close( &ctx.simd, vhashA, vhashA, dataLen<<3 );
-           mm256_dintrlv_2x128( hash0, hash1, vhashA, dataLen<<3 );
-           mm256_intrlv_2x128( vhashA, hash2, hash3, dataLen<<3 );
+           dintrlv_2x128( hash0, hash1, vhashA, dataLen<<3 );
+           intrlv_2x128( vhashA, hash2, hash3, dataLen<<3 );
            simd_2way_init( &ctx.simd, 512 );
            simd_2way_update_close( &ctx.simd, vhashA, vhashA, dataLen<<3 );
-           mm256_dintrlv_2x128( hash2, hash3, vhashA, dataLen<<3 );
+           dintrlv_2x128( hash2, hash3, vhashA, dataLen<<3 );
            if ( i != 9 )
-              mm256_intrlv_4x64( vhashB,
-                                     hash0, hash1, hash2, hash3, dataLen<<3 );
+              intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, dataLen<<3 );
         break;
         default:
            applog(LOG_ERR,"SWERR: timetravel invalid permutation");
@@ -253,7 +239,7 @@ int scanhash_timetravel10_4way( struct work *work,
    }
 
    uint64_t *edata = (uint64_t*)endiandata;
-   mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+   intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
 
    do
    {
diff --git a/algo/x11/tribus-4way.c b/algo/x11/tribus-4way.c
index 2e17ad3..d7f6194 100644
--- a/algo/x11/tribus-4way.c
+++ b/algo/x11/tribus-4way.c
@@ -37,7 +37,7 @@ void tribus_hash_4way(void *state, const void *input)
      keccak512_4way( &ctx_keccak, vhash, 64 );
      keccak512_4way_close( &ctx_keccak, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // hash echo serially
      init_echo( &ctx_echo, 512 );
@@ -64,7 +64,6 @@ int scanhash_tribus_4way( struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[4*8] __attribute__ ((aligned (64)));
    uint32_t vdata[20*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t first_nonce = pdata[19];
@@ -87,9 +86,7 @@ int scanhash_tribus_4way( struct work *work, uint32_t max_nonce,
                         0xFFFF0000,
                                  0 };
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//   mm256_bswap_intrlv80_4x64( vdata, pdata );
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
    // precalc midstate
    // doing it one way then then interleaving would be faster but too
diff --git a/algo/x11/x11-4way.c b/algo/x11/x11-4way.c
index 26fb57b..52a3c1b 100644
--- a/algo/x11/x11-4way.c
+++ b/algo/x11/x11-4way.c
@@ -69,7 +69,7 @@ void x11_4way_hash( void *state, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 3 Groestl
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -81,7 +81,7 @@ void x11_4way_hash( void *state, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
      // 4way
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      // 4 Skein
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -95,16 +95,16 @@ void x11_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 7 Luffa parallel 2 way 128 bit
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
-     mm256_intrlv_2x128( vhashB, hash2, hash3, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhashB, hash2, hash3, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhashB, 512 );
 
      // 8 Cubehash
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
@@ -132,13 +132,13 @@ void x11_4way_hash( void *state, const void *input )
      sph_shavite512_close( &ctx.shavite, hash3 );
 
      // 10 Simd
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
-     mm256_intrlv_2x128( vhashB, hash2, hash3, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhashB, hash2, hash3, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhashB, 512 );
 
      // 11 Echo
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -164,7 +164,6 @@ int scanhash_x11_4way( struct work *work, uint32_t max_nonce,
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-     uint32_t edata[20] __attribute__((aligned(64)));
      uint32_t *pdata = work->data;
      uint32_t *ptarget = work->target;
      uint32_t n = pdata[19];
@@ -177,9 +176,7 @@ int scanhash_x11_4way( struct work *work, uint32_t max_nonce,
      uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//     mm256_bswap_intrlv80_4x64( vdata, pdata );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
      for (int m=0; m < 6; m++) 
        if (Htarg <= htmax[m])
diff --git a/algo/x11/x11evo-4way.c b/algo/x11/x11evo-4way.c
index eebc28d..8fe1512 100644
--- a/algo/x11/x11evo-4way.c
+++ b/algo/x11/x11evo-4way.c
@@ -87,19 +87,16 @@ void x11evo_4way_hash( void *state, const void *input )
          case 0:
             blake512_4way( &ctx.blake, input, 80 );
             blake512_4way_close( &ctx.blake, vhash );
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                        vhash, 64<<3 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 64<<3 );
          break;
          case 1:
             bmw512_4way( &ctx.bmw, vhash, 64 );
             bmw512_4way_close( &ctx.bmw, vhash );
             if ( i >= len-1 )
-               mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                        vhash, 64<<3 );
+               dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 64<<3 );
          break;
          case 2:
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                     vhash, 64<<3 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 64<<3 );
             update_and_final_groestl( &ctx.groestl, (char*)hash0,
                                                     (char*)hash0, 512 );
             reinit_groestl( &ctx.groestl );
@@ -112,47 +109,40 @@ void x11evo_4way_hash( void *state, const void *input )
             update_and_final_groestl( &ctx.groestl, (char*)hash3,
                                                       (char*)hash3, 512 );
             if ( i < len-1 )
-               mm256_intrlv_4x64( vhash,
-                                      hash0, hash1, hash2, hash3, 64<<3 );
+               intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 64<<3 );
          break;
          case 3:
             skein512_4way( &ctx.skein, vhash, 64 );
             skein512_4way_close( &ctx.skein, vhash );
             if ( i >= len-1 )
-               mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                        vhash, 64<<3 );
+               dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 64<<3 );
          break;
          case 4:
             jh512_4way( &ctx.jh, vhash, 64 );
             jh512_4way_close( &ctx.jh, vhash );
             if ( i >= len-1 )
-               mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                        vhash, 64<<3 );
+               dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 64<<3 );
          break;
          case 5:
             keccak512_4way( &ctx.keccak, vhash, 64 );
             keccak512_4way_close( &ctx.keccak, vhash );
             if ( i >= len-1 )
-               mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                        vhash, 64<<3 );
+               dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 64<<3 );
          break;
          case 6:
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                     vhash, 64<<3 );
-            mm256_intrlv_2x128( vhash, hash0, hash1, 64<<3 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 64<<3 );
+            intrlv_2x128( vhash, hash0, hash1, 64<<3 );
             luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-            mm256_dintrlv_2x128( hash0, hash1, vhash, 64<<3 );
-            mm256_intrlv_2x128( vhash, hash2, hash3, 64<<3 );
+            dintrlv_2x128( hash0, hash1, vhash, 64<<3 );
+            intrlv_2x128( vhash, hash2, hash3, 64<<3 );
             luffa_2way_init( &ctx.luffa, 512 );
             luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-            mm256_dintrlv_2x128( hash2, hash3, vhash, 64<<3 );
+            dintrlv_2x128( hash2, hash3, vhash, 64<<3 );
             if ( i < len-1 )
-               mm256_intrlv_4x64( vhash,
-                                      hash0, hash1, hash2, hash3, 64<<3 );
+               intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 64<<3 );
          break;
          case 7:
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                     vhash, 64<<3 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 64<<3 );
             cubehashUpdateDigest( &ctx.cube, (byte*)hash0,
                                       (const byte*) hash0, 64 );
             memcpy( &ctx.cube, &x11evo_4way_ctx.cube, sizeof(cubehashParam) );
@@ -165,12 +155,10 @@ void x11evo_4way_hash( void *state, const void *input )
             cubehashUpdateDigest( &ctx.cube, (byte*)hash3,
                                       (const byte*) hash3, 64 );
             if ( i < len-1 )
-               mm256_intrlv_4x64( vhash,
-                                      hash0, hash1, hash2, hash3, 64<<3 );
+               intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 64<<3 );
          break;
          case 8:
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                     vhash, 64<<3 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 64<<3 );
             sph_shavite512( &ctx.shavite, hash0, 64 );
             sph_shavite512_close( &ctx.shavite, hash0 );
             memcpy( &ctx.shavite, &x11evo_4way_ctx.shavite,
@@ -186,26 +174,22 @@ void x11evo_4way_hash( void *state, const void *input )
             sph_shavite512( &ctx.shavite, hash3, 64 );
             sph_shavite512_close( &ctx.shavite, hash3 );
             if ( i < len-1 )
-               mm256_intrlv_4x64( vhash,
-                                      hash0, hash1, hash2, hash3, 64<<3 );
+               intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 64<<3 );
          break;
          case 9:
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                     vhash, 64<<3 );
-            mm256_intrlv_2x128( vhash, hash0, hash1, 64<<3 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 64<<3 );
+            intrlv_2x128( vhash, hash0, hash1, 64<<3 );
             simd_2way_update_close( &ctx.simd, vhash, vhash, 64<<3 );
-            mm256_dintrlv_2x128( hash0, hash1, vhash, 64<<3 );
-            mm256_intrlv_2x128( vhash, hash2, hash3, 64<<3 );
+            dintrlv_2x128( hash0, hash1, vhash, 64<<3 );
+            intrlv_2x128( vhash, hash2, hash3, 64<<3 );
             simd_2way_init( &ctx.simd, 512 );
             simd_2way_update_close( &ctx.simd, vhash, vhash, 64<<3 );
-            mm256_dintrlv_2x128( hash2, hash3, vhash, 64<<3 );
+            dintrlv_2x128( hash2, hash3, vhash, 64<<3 );
             if ( i < len-1 )
-               mm256_intrlv_4x64( vhash,
-                                      hash0, hash1, hash2, hash3, 64<<3 );
+               intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 64<<3 );
          break;
          case 10:
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3,
-                                     vhash, 64<<3 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 64<<3 );
             update_final_echo( &ctx.echo, (BitSequence *)hash0,
                                    (const BitSequence *) hash0, 512 );
             memcpy( &ctx.echo, &x11evo_4way_ctx.echo, sizeof(hashState_echo) );
@@ -218,8 +202,7 @@ void x11evo_4way_hash( void *state, const void *input )
             update_final_echo( &ctx.echo, (BitSequence *)hash3,
                                    (const BitSequence *) hash3, 512 );
             if ( i < len-1 )
-               mm256_intrlv_4x64( vhash,
-                                      hash0, hash1, hash2, hash3, 64<<3 );
+               intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 64<<3 );
          break;
       }
    }
@@ -269,7 +252,7 @@ int scanhash_x11evo_4way( struct work* work, uint32_t max_nonce,
       }
 
      uint64_t *edata = (uint64_t*)endiandata;
-     mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+     intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
 
      do
      {
diff --git a/algo/x11/x11gost-4way.c b/algo/x11/x11gost-4way.c
index 21e03a8..90b1ebd 100644
--- a/algo/x11/x11gost-4way.c
+++ b/algo/x11/x11gost-4way.c
@@ -70,7 +70,7 @@ void x11gost_4way_hash( void *state, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
      memcpy( &ctx.groestl, &x11gost_4way_ctx.groestl,
@@ -84,7 +84,7 @@ void x11gost_4way_hash( void *state, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
      // 4way
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way( &ctx.skein, vhash, 64 );
      skein512_4way_close( &ctx.skein, vhash );
@@ -96,7 +96,7 @@ void x11gost_4way_hash( void *state, const void *input )
      keccak512_4way_close( &ctx.keccak, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_gost512( &ctx.gost, hash0, 64 );
      sph_gost512_close( &ctx.gost, hash0 );
@@ -110,13 +110,13 @@ void x11gost_4way_hash( void *state, const void *input )
      sph_gost512( &ctx.gost, hash3, 64 );
      sph_gost512_close( &ctx.gost, hash3 );
 
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
      memcpy( &ctx.cube, &x11gost_4way_ctx.cube, sizeof(cubehashParam) );
@@ -141,12 +141,12 @@ void x11gost_4way_hash( void *state, const void *input )
      sph_shavite512( &ctx.shavite, hash3, 64 );
      sph_shavite512_close( &ctx.shavite, hash3 );
 
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
                        (const BitSequence *) hash0, 512 );
@@ -171,7 +171,6 @@ int scanhash_x11gost_4way( struct work *work, uint32_t max_nonce,
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
      uint32_t *pdata = work->data;
      uint32_t *ptarget = work->target;
      uint32_t n = pdata[19];
@@ -184,9 +183,7 @@ int scanhash_x11gost_4way( struct work *work, uint32_t max_nonce,
      uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//     mm256_bswap_intrlv80_4x64( vdata, pdata );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
      for (int m=0; m < 6; m++) 
        if (Htarg <= htmax[m])
diff --git a/algo/x12/x12-4way.c b/algo/x12/x12-4way.c
index 3a0cde7..80fae6d 100644
--- a/algo/x12/x12-4way.c
+++ b/algo/x12/x12-4way.c
@@ -33,7 +33,6 @@ typedef struct {
     simd_2way_context       simd;
     hashState_echo          echo;
     hamsi512_4way_context   hamsi;
-//    sph_fugue512_context    fugue;
 } x12_4way_ctx_holder;
 
 x12_4way_ctx_holder x12_4way_ctx __attribute__ ((aligned (64)));
@@ -52,7 +51,6 @@ void init_x12_4way_ctx()
      simd_2way_init( &x12_4way_ctx.simd, 512 );
      init_echo( &x12_4way_ctx.echo, 512 );
      hamsi512_4way_init( &x12_4way_ctx.hamsi );
-//     sph_fugue512_init( &x12_4way_ctx.fugue );
 };
 
 void x12_4way_hash( void *state, const void *input )
@@ -74,7 +72,7 @@ void x12_4way_hash( void *state, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 3 Groestl
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -86,7 +84,7 @@ void x12_4way_hash( void *state, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
      // Parallel 4way 64 bit
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      // 4 Skein
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -101,16 +99,16 @@ void x12_4way_hash( void *state, const void *input )
      keccak512_4way_close( &ctx.keccak, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 7 Luffa
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     intrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 8 Cubehash
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
@@ -138,13 +136,13 @@ void x12_4way_hash( void *state, const void *input )
      sph_shavite512_close( &ctx.shavite, hash3 );
 
      // 10 Simd
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 11 Echo
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -160,11 +158,11 @@ void x12_4way_hash( void *state, const void *input )
                        (const BitSequence *) hash3, 512 );
 
      // 12 Hamsi parallel 4way 32 bit
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_dintrlv_4x64( state, state+32, state+64, state+96, vhash, 256 );
+     dintrlv_4x64( state, state+32, state+64, state+96, vhash, 256 );
 }
 
 int scanhash_x12_4way( struct work *work, uint32_t max_nonce,
@@ -189,7 +187,7 @@ int scanhash_x12_4way( struct work *work, uint32_t max_nonce,
      swab32_array( endiandata, pdata, 20 );
 
      uint64_t *edata = (uint64_t*)endiandata;
-     mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+     intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
 
      for ( int m=0; m < 6; m++ )
        if ( Htarg <= htmax[m] )
diff --git a/algo/x13/phi1612-4way.c b/algo/x13/phi1612-4way.c
index add72c8..a62b6f3 100644
--- a/algo/x13/phi1612-4way.c
+++ b/algo/x13/phi1612-4way.c
@@ -53,7 +53,7 @@ void phi1612_4way_hash( void *state, const void *input )
      jh512_4way_close( &ctx.jh, vhash );
 
      // Serial to the end
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // Cubehash
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
@@ -114,7 +114,6 @@ int scanhash_phi1612_4way( struct work *work, uint32_t max_nonce,
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-     uint32_t edata[20] __attribute__ ((aligned (64)));
      uint32_t *pdata = work->data;
      uint32_t *ptarget = work->target;
      const uint32_t first_nonce = pdata[19];
@@ -125,9 +124,7 @@ int scanhash_phi1612_4way( struct work *work, uint32_t max_nonce,
 
      if ( opt_benchmark )
           ( (uint32_t*)ptarget )[7] = 0x0cff;
-     swab32_array( edata, pdata, 20 );
-     mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//     mm256_bswap_intrlv80_4x64( vdata, pdata );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
      do {
            *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
diff --git a/algo/x13/skunk-4way.c b/algo/x13/skunk-4way.c
index 461d80c..165047e 100644
--- a/algo/x13/skunk-4way.c
+++ b/algo/x13/skunk-4way.c
@@ -33,7 +33,7 @@ void skunk_4way_hash( void *output, const void *input )
 
      skein512_4way( &ctx.skein, input, 80 );
      skein512_4way_close( &ctx.skein, vhash );
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*)hash0, 64 );
      memcpy( &ctx.cube, &skunk_4way_ctx.cube, sizeof(cubehashParam) );
@@ -78,7 +78,6 @@ int scanhash_skunk_4way( struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[4*8] __attribute__ ((aligned (64)));
    uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-   uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t first_nonce = pdata[19];
@@ -91,9 +90,7 @@ int scanhash_skunk_4way( struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ((uint32_t*)ptarget)[7] = 0x0cff;
 
-   swab32_array( edata, pdata, 20 );
-   mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//   mm256_bswap_intrlv80_4x64( vdata, pdata );
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
    do
    {
       *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
diff --git a/algo/x13/x13-4way.c b/algo/x13/x13-4way.c
index b8d8220..970f8e3 100644
--- a/algo/x13/x13-4way.c
+++ b/algo/x13/x13-4way.c
@@ -74,7 +74,7 @@ void x13_4way_hash( void *state, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 3 Groestl
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -86,7 +86,7 @@ void x13_4way_hash( void *state, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
      // Parallel 4way 64 bit
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      // 4 Skein
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -101,16 +101,16 @@ void x13_4way_hash( void *state, const void *input )
      keccak512_4way_close( &ctx.keccak, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 7 Luffa
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 8 Cubehash
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
@@ -138,13 +138,13 @@ void x13_4way_hash( void *state, const void *input )
      sph_shavite512_close( &ctx.shavite, hash3 );
 
      // 10 Simd
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 11 Echo
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -160,10 +160,10 @@ void x13_4way_hash( void *state, const void *input )
                        (const BitSequence *) hash3, 512 );
 
      // 12 Hamsi parallel 4way 32 bit
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 13 Fugue serial
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -189,7 +189,6 @@ int scanhash_x13_4way( struct work *work, uint32_t max_nonce,
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-     uint32_t edata[20] __attribute__ ((aligned (64)));
      uint32_t *pdata = work->data;
      uint32_t *ptarget = work->target;
      uint32_t n = pdata[19];
@@ -202,9 +201,7 @@ int scanhash_x13_4way( struct work *work, uint32_t max_nonce,
      uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
-     swab32_array( edata, pdata, 20 );
-     mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//     mm256_bswap_intrlv80_4x64( vdata, pdata );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
      for ( int m=0; m < 6; m++ )
        if ( Htarg <= htmax[m] )
diff --git a/algo/x13/x13sm3-4way.c b/algo/x13/x13sm3-4way.c
index 3289ed1..a107627 100644
--- a/algo/x13/x13sm3-4way.c
+++ b/algo/x13/x13sm3-4way.c
@@ -81,7 +81,7 @@ void x13sm3_4way_hash( void *state, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // Groestl
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -93,7 +93,7 @@ void x13sm3_4way_hash( void *state, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
      // Parallel 4way
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      // Skein
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -108,16 +108,16 @@ void x13sm3_4way_hash( void *state, const void *input )
      keccak512_4way_close( &ctx.keccak, vhash );
 
      // Serial to the end
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // Luffa
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // Cubehash
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
@@ -145,13 +145,13 @@ void x13sm3_4way_hash( void *state, const void *input )
      sph_shavite512_close( &ctx.shavite, hash3 );
 
      // Simd
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // Echo
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -185,10 +185,10 @@ void x13sm3_4way_hash( void *state, const void *input )
      dintrlv_4x32( hash0, hash1, hash2, hash3, sm3_vhash, 512 );
 
      // Hamsi parallel 4x32x2
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // Fugue serial
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -214,7 +214,6 @@ int scanhash_x13sm3_4way( struct work *work, uint32_t max_nonce,
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-     uint32_t edata[20] __attribute__ ((aligned (64)));
      uint32_t *pdata = work->data;
      uint32_t *ptarget = work->target;
      uint32_t n = pdata[19];
@@ -227,9 +226,7 @@ int scanhash_x13sm3_4way( struct work *work, uint32_t max_nonce,
      uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
-     swab32_array( edata, pdata, 20 );
-     mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//     mm256_bswap_intrlv80_4x64( vdata, pdata );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
      blake512_4way_init( &x13sm3_ctx_mid );
      blake512_4way( &x13sm3_ctx_mid, vdata, 64 );
diff --git a/algo/x14/polytimos-4way.c b/algo/x14/polytimos-4way.c
index 4740bfd..3e1cc69 100644
--- a/algo/x14/polytimos-4way.c
+++ b/algo/x14/polytimos-4way.c
@@ -39,7 +39,7 @@ void polytimos_4way_hash( void *output, const void *input )
 
      // Need to convert from 64 bit interleaved to 32 bit interleaved.
      uint32_t vhash32[16*4];
-     mm256_rintrlv_4x64_4x32( vhash32, vhash, 512 );
+     rintrlv_4x64_4x32( vhash32, vhash, 512 );
      shabal512_4way_init( &ctx.shabal );
      shabal512_4way( &ctx.shabal, vhash32, 64 );
      shabal512_4way_close( &ctx.shabal, vhash32 );
@@ -58,15 +58,15 @@ void polytimos_4way_hash( void *output, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -105,7 +105,6 @@ int scanhash_polytimos_4way( struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[4*8] __attribute__ ((aligned (64)));
    uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t *pdata = work->data;
    uint32_t *ptarget = work->target;
    const uint32_t first_nonce = pdata[19];
@@ -118,9 +117,7 @@ int scanhash_polytimos_4way( struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ptarget[7] = 0x0cff;
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//   mm256_bswap_intrlv80_4x64( vdata, pdata );
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
    do {
       *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
                  _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
diff --git a/algo/x14/veltor-4way.c b/algo/x14/veltor-4way.c
index 15842a7..4f35161 100644
--- a/algo/x14/veltor-4way.c
+++ b/algo/x14/veltor-4way.c
@@ -40,7 +40,7 @@ void veltor_4way_hash( void *output, const void *input )
 
      skein512_4way( &ctx.skein, input, 80 );
      skein512_4way_close( &ctx.skein, vhash );
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_shavite512( &ctx.shavite, hash0, 64 );
      sph_shavite512_close( &ctx.shavite, hash0 );
@@ -82,7 +82,6 @@ int scanhash_veltor_4way( struct work *work, uint32_t max_nonce,
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
      uint32_t *pdata = work->data;
      uint32_t *ptarget = work->target;
      const uint32_t Htarg = ptarget[7];
@@ -95,9 +94,7 @@ int scanhash_veltor_4way( struct work *work, uint32_t max_nonce,
      if ( opt_benchmark )
         ptarget[7] = 0x0cff;
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//     mm256_bswap_intrlv80_4x64( vdata, pdata );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
      do
      {
diff --git a/algo/x14/x14-4way.c b/algo/x14/x14-4way.c
index de66208..5267d78 100644
--- a/algo/x14/x14-4way.c
+++ b/algo/x14/x14-4way.c
@@ -78,7 +78,7 @@ void x14_4way_hash( void *state, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 3 Groestl
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -90,7 +90,7 @@ void x14_4way_hash( void *state, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
      // Parallel 4way
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      // 4 Skein
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -105,16 +105,16 @@ void x14_4way_hash( void *state, const void *input )
      keccak512_4way_close( &ctx.keccak, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 7 Luffa
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 8 Cubehash
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
@@ -142,13 +142,13 @@ void x14_4way_hash( void *state, const void *input )
      sph_shavite512_close( &ctx.shavite, hash3 );
 
      // 10 Simd
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 11 Echo
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -164,10 +164,10 @@ void x14_4way_hash( void *state, const void *input )
                        (const BitSequence *) hash3, 512 );
 
      // 12 Hamsi parallel 4way 32 bit
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 13 Fugue serial
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -193,7 +193,6 @@ int scanhash_x14_4way( struct work *work, uint32_t max_nonce,
 {
      uint32_t hash[4*16] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
      uint32_t *pdata = work->data;
      uint32_t *ptarget = work->target;
      uint32_t n = pdata[19];
@@ -206,9 +205,7 @@ int scanhash_x14_4way( struct work *work, uint32_t max_nonce,
      uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//     mm256_bswap_intrlv80_4x64( vdata, pdata );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
      for ( int m=0; m < 6; m++ )
        if ( Htarg <= htmax[m] )
diff --git a/algo/x15/x15-4way.c b/algo/x15/x15-4way.c
index e4b395d..87fe361 100644
--- a/algo/x15/x15-4way.c
+++ b/algo/x15/x15-4way.c
@@ -81,7 +81,7 @@ void x15_4way_hash( void *state, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 3 Groestl
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -93,7 +93,7 @@ void x15_4way_hash( void *state, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
      // Parallel 4way
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      // 4 Skein
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -108,16 +108,16 @@ void x15_4way_hash( void *state, const void *input )
      keccak512_4way_close( &ctx.keccak, vhash );
 
      // Serial to the end
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 7 Luffa
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 8 Cubehash
      cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
@@ -145,13 +145,13 @@ void x15_4way_hash( void *state, const void *input )
      sph_shavite512_close( &ctx.shavite, hash3 );
 
      // 10 Simd
-     mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
+     intrlv_2x128( vhash, hash0, hash1, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-     mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
+     dintrlv_2x128( hash0, hash1, vhash, 512 );
+     intrlv_2x128( vhash, hash2, hash3, 512 );
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+     dintrlv_2x128( hash2, hash3, vhash, 512 );
 
      // 11 Echo
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -167,10 +167,10 @@ void x15_4way_hash( void *state, const void *input )
                        (const BitSequence *) hash3, 512 );
 
      // 12 Hamsi parallel 4way 32 bit
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 13 Fugue
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -217,7 +217,6 @@ int scanhash_x15_4way( struct work *work, uint32_t max_nonce,
                        uint64_t *hashes_done, struct thr_info *mythr )
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
-     uint32_t edata[20] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
      uint32_t *pdata = work->data;
      uint32_t *ptarget = work->target;
@@ -231,10 +230,8 @@ int scanhash_x15_4way( struct work *work, uint32_t max_nonce,
      uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
-     swab32_array( edata, pdata, 20 );
-     mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
 
-//     mm256_bswap_intrlv80_4x64( vdata, pdata );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
 
      for ( int m=0; m < 6; m++ )
        if ( Htarg <= htmax[m] )
diff --git a/algo/x16/x16r-4way.c b/algo/x16/x16r-4way.c
index 85eed51..56cce06 100644
--- a/algo/x16/x16r-4way.c
+++ b/algo/x16/x16r-4way.c
@@ -67,7 +67,7 @@ void x16r_4way_hash( void* output, const void* input )
    void *in3 = (void*) hash3;
    int size = 80;
 
-   mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, input, 640 );
+   dintrlv_4x64( hash0, hash1, hash2, hash3, input, 640 );
  
    if ( s_ntime == UINT32_MAX )
    {
@@ -96,11 +96,11 @@ void x16r_4way_hash( void* output, const void* input )
                blake512_4way( &ctx.blake, input, size );
             else
             {
-               mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
+               intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
                blake512_4way( &ctx.blake, vhash, size );
             }
             blake512_4way_close( &ctx.blake, vhash );
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case BMW:
             bmw512_4way_init( &ctx.bmw );
@@ -108,11 +108,11 @@ void x16r_4way_hash( void* output, const void* input )
                bmw512_4way( &ctx.bmw, input, size );
             else
             {
-               mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
+               intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
                bmw512_4way( &ctx.bmw, vhash, size );
             }
             bmw512_4way_close( &ctx.bmw, vhash );
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case GROESTL:
                init_groestl( &ctx.groestl, 64 );
@@ -134,11 +134,11 @@ void x16r_4way_hash( void* output, const void* input )
                skein512_4way( &ctx.skein, input, size );
             else
             {
-               mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
+               intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
                skein512_4way( &ctx.skein, vhash, size );
             }
             skein512_4way_close( &ctx.skein, vhash );
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case JH:
             jh512_4way_init( &ctx.jh );
@@ -146,11 +146,11 @@ void x16r_4way_hash( void* output, const void* input )
                jh512_4way( &ctx.jh, input, size );
             else
             {
-               mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
+               intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
                jh512_4way( &ctx.jh, vhash, size );
             }
             jh512_4way_close( &ctx.jh, vhash );
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case KECCAK:
             keccak512_4way_init( &ctx.keccak );
@@ -158,21 +158,21 @@ void x16r_4way_hash( void* output, const void* input )
                keccak512_4way( &ctx.keccak, input, size );
             else
             {
-               mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
+               intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
                keccak512_4way( &ctx.keccak, vhash, size );
             }
             keccak512_4way_close( &ctx.keccak, vhash );
-            mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+            dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case LUFFA:
-            mm256_intrlv_2x128( vhash, in0, in1, size<<3 );
+            intrlv_2x128( vhash, in0, in1, size<<3 );
             luffa_2way_init( &ctx.luffa, 512 );
             luffa_2way_update_close( &ctx.luffa, vhash, vhash, size );
-            mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-            mm256_intrlv_2x128( vhash, in2, in3, size<<3 );
+            dintrlv_2x128( hash0, hash1, vhash, 512 );
+            intrlv_2x128( vhash, in2, in3, size<<3 );
             luffa_2way_init( &ctx.luffa, 512 );
             luffa_2way_update_close( &ctx.luffa, vhash, vhash, size);
-            mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+            dintrlv_2x128( hash2, hash3, vhash, 512 );
          break;
          case CUBEHASH:
             cubehashInit( &ctx.cube, 512, 16, 32 );
@@ -203,14 +203,14 @@ void x16r_4way_hash( void* output, const void* input )
             sph_shavite512_close( &ctx.shavite, hash3 );
          break;
          case SIMD:
-            mm256_intrlv_2x128( vhash, in0, in1, size<<3 );
+            intrlv_2x128( vhash, in0, in1, size<<3 );
             simd_2way_init( &ctx.simd, 512 );
             simd_2way_update_close( &ctx.simd, vhash, vhash, size<<3 );
-            mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
-            mm256_intrlv_2x128( vhash, in2, in3, size<<3 );
+            dintrlv_2x128( hash0, hash1, vhash, 512 );
+            intrlv_2x128( vhash, in2, in3, size<<3 );
             simd_2way_init( &ctx.simd, 512 );
             simd_2way_update_close( &ctx.simd, vhash, vhash, size<<3 );
-            mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
+            dintrlv_2x128( hash2, hash3, vhash, 512 );
          break;
          case ECHO:
              init_echo( &ctx.echo, 512 );
@@ -227,11 +227,11 @@ void x16r_4way_hash( void* output, const void* input )
                                 (const BitSequence*)in3, size<<3 );
          break;
          case HAMSI:
-             mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
+             intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
              hamsi512_4way_init( &ctx.hamsi );
              hamsi512_4way( &ctx.hamsi, vhash, size );
              hamsi512_4way_close( &ctx.hamsi, vhash );
-             mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+             dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
          case FUGUE:
              sph_fugue512_init( &ctx.fugue );
@@ -269,11 +269,11 @@ void x16r_4way_hash( void* output, const void* input )
              sph_whirlpool_close( &ctx.whirlpool, hash3 );
          break;
          case SHA_512:
-             mm256_intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
+             intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
              sha512_4way_init( &ctx.sha512 );
              sha512_4way( &ctx.sha512, vhash, size );
              sha512_4way_close( &ctx.sha512, vhash );
-             mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+             dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
          break;
       }
       size = 64;
@@ -316,7 +316,7 @@ int scanhash_x16r_4way( struct work *work, uint32_t max_nonce,
       ptarget[7] = 0x0cff;
 
    uint64_t *edata = (uint64_t*)endiandata;
-   mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
+   intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
 
    do
    {
diff --git a/algo/x17/sonoa-4way.c b/algo/x17/sonoa-4way.c
index 18c1065..bda67b4 100644
--- a/algo/x17/sonoa-4way.c
+++ b/algo/x17/sonoa-4way.c
@@ -69,7 +69,7 @@ void sonoa_4way_hash( void *state, const void *input )
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -80,7 +80,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -94,7 +94,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -116,8 +116,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
+     dintrlv_2x128_512( hash0, hash1, vhashA );
+     dintrlv_2x128_512( hash2, hash3, vhashB );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -134,13 +134,13 @@ void sonoa_4way_hash( void *state, const void *input )
 
 // 2
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      bmw512_4way_init( &ctx.bmw );
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -151,7 +151,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -165,7 +165,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -187,8 +187,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
+     dintrlv_2x128_512( hash0, hash1, vhashA );
+     dintrlv_2x128_512( hash2, hash3, vhashB );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -203,7 +203,7 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
@@ -215,7 +215,7 @@ void sonoa_4way_hash( void *state, const void *input )
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -226,7 +226,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -240,7 +240,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -262,8 +262,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
+     dintrlv_2x128_512( hash0, hash1, vhashA );
+     dintrlv_2x128_512( hash2, hash3, vhashB );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -278,13 +278,13 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -300,13 +300,13 @@ void sonoa_4way_hash( void *state, const void *input )
      sph_fugue512_close( &ctx.fugue, hash3 );
 
 // 4
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      bmw512_4way_init( &ctx.bmw );
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -317,7 +317,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -331,7 +331,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -353,8 +353,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
+     dintrlv_2x128_512( hash0, hash1, vhashA );
+     dintrlv_2x128_512( hash2, hash3, vhashB );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -369,13 +369,13 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -396,13 +396,13 @@ void sonoa_4way_hash( void *state, const void *input )
      shabal512_4way( &ctx.shabal, vhash, 64 );
      shabal512_4way_close( &ctx.shabal, vhash );
 
-     mm256_rintrlv_4x32_4x64( vhashB, vhash, 512 ); 
+     rintrlv_4x32_4x64( vhashB, vhash, 512 ); 
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhashB, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -417,8 +417,8 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_intrlv_2x128( vhashA, hash0, hash1, 512 );
-     mm256_intrlv_2x128( vhashB, hash2, hash3, 512 );
+     intrlv_2x128_512( vhashA, hash0, hash1 );
+     intrlv_2x128_512( vhashB, hash2, hash3 );
 
      shavite512_2way_init( &ctx.shavite );
      shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
@@ -426,13 +426,13 @@ void sonoa_4way_hash( void *state, const void *input )
      shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
 
 // 5
-     mm256_rintrlv_2x128_4x64( vhash, vhashA, vhashB, 512 );
+     rintrlv_2x128_4x64( vhash, vhashA, vhashB, 512 );
 
      bmw512_4way_init( &ctx.bmw );
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_rintrlv_4x64_4x32( vhashB, vhash,  512 );
+     rintrlv_4x64_4x32( vhashB, vhash,  512 );
 
      shabal512_4way_init( &ctx.shabal );
      shabal512_4way( &ctx.shabal, vhashB, 64 );
@@ -449,7 +449,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -463,7 +463,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -485,8 +485,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
+     dintrlv_2x128_512( hash0, hash1, vhashA );
+     dintrlv_2x128_512( hash2, hash3, vhashB );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -501,13 +501,13 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -545,13 +545,13 @@ void sonoa_4way_hash( void *state, const void *input )
 
 // 6
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
      
      bmw512_4way_init( &ctx.bmw );
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -562,7 +562,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -576,7 +576,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -598,8 +598,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
+     dintrlv_2x128_512( hash0, hash1, vhashA );
+     dintrlv_2x128_512( hash2, hash3, vhashB );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -614,13 +614,13 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -656,13 +656,13 @@ void sonoa_4way_hash( void *state, const void *input )
      sph_whirlpool( &ctx.whirlpool, hash3, 64 );
      sph_whirlpool_close( &ctx.whirlpool, hash3 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      sha512_4way_init( &ctx.sha512 );
      sha512_4way( &ctx.sha512, vhash, 64 );
      sha512_4way_close( &ctx.sha512, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_whirlpool_init( &ctx.whirlpool );
      sph_whirlpool( &ctx.whirlpool, hash0, 64 );
@@ -679,13 +679,13 @@ void sonoa_4way_hash( void *state, const void *input )
 
 // 7
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      bmw512_4way_init( &ctx.bmw );
      bmw512_4way( &ctx.bmw, vhash, 64 );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -696,7 +696,7 @@ void sonoa_4way_hash( void *state, const void *input )
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, 64 );
@@ -710,7 +710,7 @@ void sonoa_4way_hash( void *state, const void *input )
      keccak512_4way( &ctx.keccak, vhash, 64 );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -732,8 +732,8 @@ void sonoa_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_dintrlv_2x128( hash0, hash1, vhashA, 512 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, 512 );
+     dintrlv_2x128_512( hash0, hash1, vhashA );
+     dintrlv_2x128_512( hash2, hash3, vhashB );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -748,13 +748,13 @@ void sonoa_4way_hash( void *state, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, 512 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -790,13 +790,13 @@ void sonoa_4way_hash( void *state, const void *input )
      sph_whirlpool( &ctx.whirlpool, hash3, 64 );
      sph_whirlpool_close( &ctx.whirlpool, hash3 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      sha512_4way_init( &ctx.sha512 );
      sha512_4way( &ctx.sha512, vhash, 64 );
      sha512_4way_close( &ctx.sha512, vhash );
 
-     mm256_rintrlv_4x64_4x32( vhashB, vhash,  512 );
+     rintrlv_4x64_4x32( vhashB, vhash,  512 );
 
      haval256_5_4way_init( &ctx.haval );
      haval256_5_4way( &ctx.haval, vhashB, 64 );
@@ -808,7 +808,6 @@ int scanhash_sonoa_4way( struct work *work, uint32_t max_nonce,
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
      uint32_t lane_hash[8] __attribute__ ((aligned (32)));
      uint32_t *hash7 = &(hash[7<<2]);
      uint32_t *pdata = work->data;
@@ -824,9 +823,7 @@ int scanhash_sonoa_4way( struct work *work, uint32_t max_nonce,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
      // Need big endian data
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//     mm256_bswap_intrlv80_4x64( vdata, pdata );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
      for ( int m=0; m < 6; m++ ) if ( Htarg <= htmax[m] )
      {
         uint32_t mask = masks[m];
diff --git a/algo/x17/x17-4way.c b/algo/x17/x17-4way.c
index e0cec73..8c8ad6a 100644
--- a/algo/x17/x17-4way.c
+++ b/algo/x17/x17-4way.c
@@ -68,7 +68,7 @@ void x17_4way_hash( void *state, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serialize
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 3 Groestl
      init_groestl( &ctx.groestl, 64 );
@@ -81,7 +81,7 @@ void x17_4way_hash( void *state, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
 
      // Parallellize
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      // 4 Skein parallel 4 way 64 bit 
      skein512_4way_init( &ctx.skein );
@@ -99,7 +99,7 @@ void x17_4way_hash( void *state, const void *input )
      keccak512_4way_close( &ctx.keccak, vhash );
 
      // 7 Luffa  parallel 2 way 128 bit
-     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
+     rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
@@ -124,8 +124,8 @@ void x17_4way_hash( void *state, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
 
-     mm256_dintrlv_2x128_512( hash0, hash1, vhashA );
-     mm256_dintrlv_2x128_512( hash2, hash3, vhashB );
+     dintrlv_2x128_512( hash0, hash1, vhashA );
+     dintrlv_2x128_512( hash2, hash3, vhashB );
 
      // 11 Echo serial
      init_echo( &ctx.echo, 512 );
@@ -142,13 +142,13 @@ void x17_4way_hash( void *state, const void *input )
                        (const BitSequence *) hash3, 512 );
 
      // 12 Hamsi parallel 4 way 64 bit
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, 64 );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
 
      // 13 Fugue serial
      sph_fugue512_init( &ctx.fugue );
@@ -188,14 +188,14 @@ void x17_4way_hash( void *state, const void *input )
      sph_whirlpool_close( &ctx.whirlpool, hash3 );
 
      // 16 SHA512 parallel 64 bit 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 
      sha512_4way_init( &ctx.sha512 );
      sha512_4way( &ctx.sha512, vhash, 64 );
      sha512_4way_close( &ctx.sha512, vhash );     
 
      // 17 Haval parallel 32 bit
-     mm256_rintrlv_4x64_4x32( vhashB, vhash,  512 );
+     rintrlv_4x64_4x32( vhashB, vhash,  512 );
 
      haval256_5_4way_init( &ctx.haval );
      haval256_5_4way( &ctx.haval, vhashB, 64 );
@@ -207,7 +207,6 @@ int scanhash_x17_4way( struct work *work, uint32_t max_nonce,
 {
      uint32_t hash[4*8] __attribute__ ((aligned (64)));
      uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
      uint32_t lane_hash[8] __attribute__ ((aligned (32)));
      uint32_t *hash7 = &(hash[7<<2]);
      uint32_t *pdata = work->data;
@@ -223,9 +222,7 @@ int scanhash_x17_4way( struct work *work, uint32_t max_nonce,
                           0xFFFFF000, 0xFFFF0000,          0  };
 
      // Need big endian data
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//     mm256_bswap_intrlv80_4x64( vdata, pdata );
+     mm256_bswap32_intrlv80_4x64( vdata, pdata );
      for ( int m = 0; m < 6; m++ ) if ( Htarg <= htmax[m] )
      {
         uint32_t mask = masks[ m ];
@@ -235,7 +232,7 @@ int scanhash_x17_4way( struct work *work, uint32_t max_nonce,
 	              _mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
            x17_4way_hash( hash, vdata );
 
-	     for ( int lane = 0; lane < 4; lane++ )
+           for ( int lane = 0; lane < 4; lane++ )
            if ( ( hash7[ lane ] & mask ) == 0 )
            {
               extr_lane_4x32( lane_hash, hash, lane, 256 );
diff --git a/algo/x17/xevan-4way.c b/algo/x17/xevan-4way.c
index 8b5a11c..f547276 100644
--- a/algo/x17/xevan-4way.c
+++ b/algo/x17/xevan-4way.c
@@ -71,7 +71,7 @@ void xevan_4way_hash( void *output, const void *input )
      bmw512_4way_close( &ctx.bmw, vhash );
 
      // Serial
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0,
@@ -87,7 +87,7 @@ void xevan_4way_hash( void *output, const void *input )
                                dataLen<<3 );
 
      // Parallel 4way
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, dataLen );
@@ -101,7 +101,7 @@ void xevan_4way_hash( void *output, const void *input )
      keccak512_4way( &ctx.keccak, vhash, dataLen );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, dataLen<<3 );
+     rintrlv_4x64_2x128( vhashA, vhashB, vhash, dataLen<<3 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, dataLen );
@@ -123,8 +123,8 @@ void xevan_4way_hash( void *output, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, dataLen<<3 );
 
-     mm256_dintrlv_2x128( hash0, hash1, vhashA, dataLen<<3 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, dataLen<<3 );
+     dintrlv_2x128( hash0, hash1, vhashA, dataLen<<3 );
+     dintrlv_2x128( hash2, hash3, vhashB, dataLen<<3 );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -139,13 +139,13 @@ void xevan_4way_hash( void *output, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, dataLen<<3 );
      // Parallel
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, dataLen );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, dataLen );
@@ -183,19 +183,19 @@ void xevan_4way_hash( void *output, const void *input )
      sph_whirlpool( &ctx.whirlpool, hash3, dataLen );
      sph_whirlpool_close( &ctx.whirlpool, hash3 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
 
      sha512_4way_init( &ctx.sha512 );
      sha512_4way( &ctx.sha512, vhash, dataLen );
      sha512_4way_close( &ctx.sha512, vhash );
 
-     mm256_rintrlv_4x64_4x32( vhashA, vhash, dataLen<<3 );
+     rintrlv_4x64_4x32( vhashA, vhash, dataLen<<3 );
 
      haval256_5_4way_init( &ctx.haval );
      haval256_5_4way( &ctx.haval, vhashA, dataLen );
      haval256_5_4way_close( &ctx.haval, vhashA );
 
-     mm256_rintrlv_4x32_4x64( vhash, vhashA, dataLen<<3 );
+     rintrlv_4x32_4x64( vhash, vhashA, dataLen<<3 );
 
      memset( &vhash[ 4<<2 ], 0, (dataLen-32) << 2 );
 
@@ -207,7 +207,7 @@ void xevan_4way_hash( void *output, const void *input )
      bmw512_4way( &ctx.bmw, vhash, dataLen );
      bmw512_4way_close( &ctx.bmw, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
 
      init_groestl( &ctx.groestl, 64 );
      update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0,
@@ -222,7 +222,7 @@ void xevan_4way_hash( void *output, const void *input )
      update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3,
                                dataLen<<3 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
 
      skein512_4way_init( &ctx.skein );
      skein512_4way( &ctx.skein, vhash, dataLen );
@@ -236,7 +236,7 @@ void xevan_4way_hash( void *output, const void *input )
      keccak512_4way( &ctx.keccak, vhash, dataLen );
      keccak512_4way_close( &ctx.keccak, vhash );
 
-     mm256_rintrlv_4x64_2x128( vhashA, vhashB, vhash, dataLen<<3 );
+     rintrlv_4x64_2x128( vhashA, vhashB, vhash, dataLen<<3 );
 
      luffa_2way_init( &ctx.luffa, 512 );
      luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, dataLen );
@@ -258,8 +258,8 @@ void xevan_4way_hash( void *output, const void *input )
      simd_2way_init( &ctx.simd, 512 );
      simd_2way_update_close( &ctx.simd, vhashB, vhashB, dataLen<<3 );
 
-     mm256_dintrlv_2x128( hash0, hash1, vhashA, dataLen<<3 );
-     mm256_dintrlv_2x128( hash2, hash3, vhashB, dataLen<<3 );
+     dintrlv_2x128( hash0, hash1, vhashA, dataLen<<3 );
+     dintrlv_2x128( hash2, hash3, vhashB, dataLen<<3 );
 
      init_echo( &ctx.echo, 512 );
      update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -274,13 +274,13 @@ void xevan_4way_hash( void *output, const void *input )
      update_final_echo( &ctx.echo, (BitSequence *)hash3,
                        (const BitSequence *) hash3, dataLen<<3 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
 
      hamsi512_4way_init( &ctx.hamsi );
      hamsi512_4way( &ctx.hamsi, vhash, dataLen );
      hamsi512_4way_close( &ctx.hamsi, vhash );
 
-     mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
+     dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
 
      sph_fugue512_init( &ctx.fugue );
      sph_fugue512( &ctx.fugue, hash0, dataLen );
@@ -316,13 +316,13 @@ void xevan_4way_hash( void *output, const void *input )
      sph_whirlpool( &ctx.whirlpool, hash3, dataLen );
      sph_whirlpool_close( &ctx.whirlpool, hash3 );
 
-     mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
+     intrlv_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
 
      sha512_4way_init( &ctx.sha512 );
      sha512_4way( &ctx.sha512, vhash, dataLen );
      sha512_4way_close( &ctx.sha512, vhash );
 
-     mm256_rintrlv_4x64_4x32( vhashA, vhash, dataLen<<3 );
+     rintrlv_4x64_4x32( vhashA, vhash, dataLen<<3 );
 
      haval256_5_4way_init( &ctx.haval );
      haval256_5_4way( &ctx.haval, vhashA, dataLen );
@@ -334,7 +334,6 @@ int scanhash_xevan_4way( struct work *work, uint32_t max_nonce,
 {
    uint32_t hash[4*8] __attribute__ ((aligned (64)));
    uint32_t vdata[24*4] __attribute__ ((aligned (64)));
-    uint32_t edata[20] __attribute__ ((aligned (64)));
    uint32_t lane_hash[8] __attribute__ ((aligned (32)));
    uint32_t *hash7 = &(hash[7<<2]);
    uint32_t *pdata = work->data;
@@ -349,9 +348,7 @@ int scanhash_xevan_4way( struct work *work, uint32_t max_nonce,
    if ( opt_benchmark )
       ptarget[7] = 0x0cff;
 
-    swab32_array( edata, pdata, 20 );
-    mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
-//   mm256_bswap_intrlv80_4x64( vdata, pdata );
+   mm256_bswap32_intrlv80_4x64( vdata, pdata );
    do {
       *noncev = mm256_intrlv_blend_32( mm256_bswap_32(
                _mm256_set_epi32( n+3, 0,n+2, 0,n+1, 0, n, 0 ) ), *noncev );
diff --git a/configure b/configure
index 6dae213..0c55d03 100755
--- a/configure
+++ b/configure
@@ -1,6 +1,6 @@
 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.69 for cpuminer-opt 3.9.5.3.
+# Generated by GNU Autoconf 2.69 for cpuminer-opt 3.9.5.4.
 #
 #
 # Copyright (C) 1992-1996, 1998-2012 Free Software Foundation, Inc.
@@ -577,8 +577,8 @@ MAKEFLAGS=
 # Identity of this package.
 PACKAGE_NAME='cpuminer-opt'
 PACKAGE_TARNAME='cpuminer-opt'
-PACKAGE_VERSION='3.9.5.3'
-PACKAGE_STRING='cpuminer-opt 3.9.5.3'
+PACKAGE_VERSION='3.9.5.4'
+PACKAGE_STRING='cpuminer-opt 3.9.5.4'
 PACKAGE_BUGREPORT=''
 PACKAGE_URL=''
 
@@ -1332,7 +1332,7 @@ 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 3.9.5.3 to adapt to many kinds of systems.
+\`configure' configures cpuminer-opt 3.9.5.4 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1404,7 +1404,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of cpuminer-opt 3.9.5.3:";;
+     short | recursive ) echo "Configuration of cpuminer-opt 3.9.5.4:";;
    esac
   cat <<\_ACEOF
 
@@ -1509,7 +1509,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-cpuminer-opt configure 3.9.5.3
+cpuminer-opt configure 3.9.5.4
 generated by GNU Autoconf 2.69
 
 Copyright (C) 2012 Free Software Foundation, Inc.
@@ -2012,7 +2012,7 @@ 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 3.9.5.3, which was
+It was created by cpuminer-opt $as_me 3.9.5.4, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   $ $0 $@
@@ -2993,7 +2993,7 @@ fi
 
 # Define the identity of the package.
  PACKAGE='cpuminer-opt'
- VERSION='3.9.5.3'
+ VERSION='3.9.5.4'
 
 
 cat >>confdefs.h <<_ACEOF
@@ -6690,7 +6690,7 @@ 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 3.9.5.3, which was
+This file was extended by cpuminer-opt $as_me 3.9.5.4, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -6756,7 +6756,7 @@ _ACEOF
 cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
 ac_cs_config="`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`"
 ac_cs_version="\\
-cpuminer-opt config.status 3.9.5.3
+cpuminer-opt config.status 3.9.5.4
 configured by $0, generated by GNU Autoconf 2.69,
   with options \\"\$ac_cs_config\\"
 
diff --git a/configure.ac b/configure.ac
index c1b4825..ea6d5e0 100644
--- a/configure.ac
+++ b/configure.ac
@@ -1,4 +1,4 @@
-AC_INIT([cpuminer-opt], [3.9.5.3])
+AC_INIT([cpuminer-opt], [3.9.5.4])
 
 AC_PREREQ([2.59c])
 AC_CANONICAL_SYSTEM
diff --git a/simd-utils.h b/simd-utils.h
index 3bc0743..fb61eb9 100644
--- a/simd-utils.h
+++ b/simd-utils.h
@@ -175,32 +175,25 @@
 
 // 64 bit vectors
 #include "simd-utils/simd-64.h"
-#include "simd-utils/intrlv-mmx.h"
+//#include "simd-utils/intrlv-mmx.h"
 
 #if defined(__SSE2__)
 
 // 128 bit vectors
 #include "simd-utils/simd-128.h"
-#include "simd-utils/intrlv-sse2.h"
 
 #if defined(__AVX__)
 
 // 256 bit vector basics
 #include "simd-utils/simd-256.h"
-#include "simd-utils/intrlv-avx.h"
 
 #if defined(__AVX2__)
 
-// 256 bit everything else
-//#include "simd-utils/simd-avx2.h"
-#include "simd-utils/intrlv-avx2.h"
-
 // Skylake-X has all these
 #if defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
 
 // 512 bit vectors
 #include "simd-utils/simd-512.h"
-#include "simd-utils/intrlv-avx512.h"
 
 #endif  // MMX
 #endif  // SSE2
@@ -208,7 +201,6 @@
 #endif  // AVX2
 #endif  // AVX512
 
-// Picks implementation based on available CPU features.
-#include "simd-utils/intrlv-selector.h"
+#include "simd-utils/intrlv.h"
 
 #endif  // SIMD_UTILS_H__
diff --git a/simd-utils/intrlv-avx.h b/simd-utils/intrlv-avx.h
deleted file mode 100644
index 400f060..0000000
--- a/simd-utils/intrlv-avx.h
+++ /dev/null
@@ -1,911 +0,0 @@
-#if !defined(INTRLV_AVX_H__)
-#define INTRLV_AVX_H__ 1
-
-// philosophical discussion
-// 
-// transitions:
-//
-//  int32 <-> int64
-//  uint64_t = (uint64_t)int32_lo | ( (uint64_t)int32_hi << 32 )
-//  Efficient transition and post processing, 32 bit granularity is lost.
-//  
-//  int32 <-> m64
-//  More complex, 32 bit granularity maintained, limited number of mmx regs.
-//  int32 <-> int64 <-> m64 might be more efficient.
-//
-//  int32 <-> m128
-//  Expensive, current implementation.
-//
-//  int32 <-> m256
-//  Very expensive multi stage, current implementation.
-//
-//  int64/m64 <-> m128
-//  Efficient, agnostic to native element size. Common.
-//
-//  m128 <-> m256
-//  Expensive for a single instruction, unavoidable. Common.
-// 
-//  Multi stage options
-//
-//  int32 <-> int64 -> m128
-//  More efficient than insert32, granularity maintained. Common.
-//
-//  int64 <-> m128 -> m256
-//  Unavoidable, reasonably efficient. Common
-//
-//  int32 <-> int64 -> m128 -> m256
-//  Seems inevitable, most efficient despite number of stages. Common.
-//
-//  Implementation plan. 
-//
-//  1. Complete m128 <-> m256
-//  2. Implement int64 <-> m128
-//  3. Combine int64 <-> m128 <-> m256
-//  4. Implement int32 <-> int64 <-> m128
-//  5. Combine int32 <-> int64 <-> m128 <-> m256
-//
-
-#if  defined(__AVX__)
-
-// Convenient short cuts for local use only
-
-// Extract 64 bits from the low 128 bits of 256 bit vector.
-#define extr64_cast128_256( a, n ) \
-   _mm_extract_epi64( _mm256_castsi256_si128( a ), n )
-
-// Extract 32 bits from the low 128 bits of 256 bit vector.
-#define extr32_cast128_256( a, n ) \
-   _mm_extract_epi32( _mm256_castsi256_si128( a ), n )
-
-///////////////////////////////////////////////////////////
-//
-//          AVX 256 Bit Vectors
-//
-//  256 bit interleaving can be done with AVX.
-
-#define mm256_put_64( s0, s1, s2, s3) \
-  _mm256_set_epi64x( *((const uint64_t*)(s3)), *((const uint64_t*)(s2)), \
-                     *((const uint64_t*)(s1)), *((const uint64_t*)(s0)) )
-
-#define mm256_put_32( s00, s01, s02, s03, s04, s05, s06, s07 ) \
-  _mm256_set_epi32( *((const uint32_t*)(s07)), *((const uint32_t*)(s06)), \
-                    *((const uint32_t*)(s05)), *((const uint32_t*)(s04)), \
-                    *((const uint32_t*)(s03)), *((const uint32_t*)(s02)), \
-                    *((const uint32_t*)(s01)), *((const uint32_t*)(s00)) )
-
-#define mm256_get_64( s, i0, i1, i2, i3 ) \
-  _mm256_set_epi64x( ((const uint64_t*)(s))[i3], ((const uint64_t*)(s))[i2], \
-                     ((const uint64_t*)(s))[i1], ((const uint64_t*)(s))[i0] )
-
-#define mm256_get_32( s, i0, i1, i2, i3, i4, i5, i6, i7 ) \
-  _mm256_set_epi32( ((const uint32_t*)(s))[i7], ((const uint32_t*)(s))[i6], \
-                    ((const uint32_t*)(s))[i5], ((const uint32_t*)(s))[i4], \
-                    ((const uint32_t*)(s))[i3], ((const uint32_t*)(s))[i2], \
-                    ((const uint32_t*)(s))[i1], ((const uint32_t*)(s))[i0] )
-
-/*
-// Blend 2 vectors alternating hi & lo: { hi[n], lo[n-1], ... hi[1], lo[0] }
-#define mm256_intrlv_blend_128( hi, lo ) \
-                _mm256_blend_epi32( hi, lo, 0x0f )
-
-#define mm256_intrlv_blend_64( hi, lo ) \
-                _mm256_blend_epi32( hi, lo, 0x33 )
-
-#define mm256_intrlv_blend_32( hi, lo ) \
-           _mm256_blend_epi32( hi, lo, 0x55 )
-*/
-
-// Interleave 8x32_256
-#define mm256_intrlv_8x32_256( d, s0, s1, s2, s3, s4, s5, s6, s7 ) \
-{ \
-   __m128i s0hi = mm128_extr_hi128_256( s0 ); \
-   __m128i s1hi = mm128_extr_hi128_256( s1 ); \
-   __m128i s2hi = mm128_extr_hi128_256( s2 ); \
-   __m128i s3hi = mm128_extr_hi128_256( s3 ); \
-   __m128i s4hi = mm128_extr_hi128_256( s4 ); \
-   __m128i s5hi = mm128_extr_hi128_256( s5 ); \
-   __m128i s6hi = mm128_extr_hi128_256( s6 ); \
-   __m128i s7hi = mm128_extr_hi128_256( s7 ); \
-   casti_m256i( d,0 ) = _mm256_set_epi32( \
-                        extr32_cast128_256(s7,0), extr32_cast128_256(s6,0), \
-                        extr32_cast128_256(s5,0), extr32_cast128_256(s4,0), \
-                        extr32_cast128_256(s3,0), extr32_cast128_256(s2,0), \
-                        extr32_cast128_256(s1,0), extr32_cast128_256(s0,0) ); \
-   casti_m256i( d,1 ) = _mm256_set_epi32( \
-                        extr32_cast128_256(s7,1), extr32_cast128_256(s6,1), \
-                        extr32_cast128_256(s5,1), extr32_cast128_256(s4,1), \
-                        extr32_cast128_256(s3,1), extr32_cast128_256(s2,1), \
-                        extr32_cast128_256(s1,1), extr32_cast128_256(s0,1) ); \
-   casti_m256i( d,2 ) = _mm256_set_epi32( \
-                        extr32_cast128_256(s7,2), extr32_cast128_256(s6,2), \
-                        extr32_cast128_256(s5,2), extr32_cast128_256(s4,2), \
-                        extr32_cast128_256(s3,2), extr32_cast128_256(s2,2), \
-                        extr32_cast128_256(s1,2), extr32_cast128_256(s0,2) ); \
-   casti_m256i( d,3 ) = _mm256_set_epi32( \
-                        extr32_cast128_256(s7,3), extr32_cast128_256(s6,3), \
-                        extr32_cast128_256(s5,3), extr32_cast128_256(s4,3), \
-                        extr32_cast128_256(s3,3), extr32_cast128_256(s2,3), \
-                        extr32_cast128_256(s1,3), extr32_cast128_256(s0,3) ); \
-   casti_m256i( d,4 ) = _mm256_set_epi32( \
-                           mm128_extr_32(s7hi,0),    mm128_extr_32(s6hi,0), \
-                           mm128_extr_32(s5hi,0),    mm128_extr_32(s4hi,0), \
-                           mm128_extr_32(s3hi,0),    mm128_extr_32(s2hi,0), \
-                           mm128_extr_32(s1hi,0),    mm128_extr_32(s0hi,0) ); \
-   casti_m256i( d,5 ) = _mm256_set_epi32( \
-                           mm128_extr_32(s7hi,1),    mm128_extr_32(s6hi,1), \
-                           mm128_extr_32(s5hi,1),    mm128_extr_32(s4hi,1), \
-                           mm128_extr_32(s3hi,1),    mm128_extr_32(s2hi,1), \
-                           mm128_extr_32(s1hi,1),    mm128_extr_32(s0hi,1) ); \
-   casti_m256i( d,6 ) = _mm256_set_epi32( \
-                           mm128_extr_32(s7hi,2),    mm128_extr_32(s6hi,2), \
-                           mm128_extr_32(s5hi,2),    mm128_extr_32(s4hi,2), \
-                           mm128_extr_32(s3hi,2),    mm128_extr_32(s2hi,2), \
-                           mm128_extr_32(s1hi,2),    mm128_extr_32(s0hi,2) ); \
-   casti_m256i( d,7 ) = _mm256_set_epi32( \
-                           mm128_extr_32(s7hi,3),    mm128_extr_32(s6hi,3), \
-                           mm128_extr_32(s5hi,3),    mm128_extr_32(s4hi,3), \
-                           mm128_extr_32(s3hi,3),    mm128_extr_32(s2hi,3), \
-                           mm128_extr_32(s1hi,3),    mm128_extr_32(s0hi,3) ); \
-} while(0)
-
-#define mm256_intrlv_8x32_128( d, s0, s1, s2, s3, s4, s5, s6, s7 ) \
-{ \
-   casti_m256i( d,0 ) = _mm256_set_epi32( \
-                           mm128_extr_32(s7,0), mm128_extr_32(s6,0), \
-                           mm128_extr_32(s5,0), mm128_extr_32(s4,0), \
-                           mm128_extr_32(s3,0), mm128_extr_32(s2,0), \
-                           mm128_extr_32(s1,0), mm128_extr_32(s0,0) ); \
-   casti_m256i( d,1 ) = _mm256_set_epi32( \
-                           mm128_extr_32(s7,1), mm128_extr_32(s6,1), \
-                           mm128_extr_32(s5,1), mm128_extr_32(s4,1), \
-                           mm128_extr_32(s3,1), mm128_extr_32(s2,1), \
-                           mm128_extr_32(s1,1), mm128_extr_32(s0,1) ); \
-   casti_m256i( d,2 ) = _mm256_set_epi32( \
-                           mm128_extr_32(s7,2), mm128_extr_32(s6,2), \
-                           mm128_extr_32(s5,2), mm128_extr_32(s4,2), \
-                           mm128_extr_32(s3,2), mm128_extr_32(s2,2), \
-                           mm128_extr_32(s1,2), mm128_extr_32(s0,2) ); \
-   casti_m256i( d,3 ) = _mm256_set_epi32( \
-                           mm128_extr_32(s7,3), mm128_extr_32(s6,3), \
-                           mm128_extr_32(s5,3), mm128_extr_32(s4,3), \
-                           mm128_extr_32(s3,3), mm128_extr_32(s2,3), \
-                           mm128_extr_32(s1,3), mm128_extr_32(s0,3) ); \
-} while(0)
-
-/*
-#define mm256_bswap_intrlv_8x32_256( d, src ) \
-do { \
-  __m256i s0 = mm256_bswap_32( src ); \
-  __m128i s1 = _mm256_extracti128_si256( s0, 1 ); \
-  casti_m256i( d, 0 ) = _mm256_set1_epi32( _mm_extract_epi32( \
-                                     _mm256_castsi256_si128( s0 ), 0 ) ); \
-  casti_m256i( d, 1 ) = _mm256_set1_epi32( _mm_extract_epi32( \
-                                     _mm256_castsi256_si128( s0 ), 1 ) ); \
-  casti_m256i( d, 2 ) = _mm256_set1_epi32( _mm_extract_epi32( \
-                                     _mm256_castsi256_si128( s0 ), 2 ) ); \
-  casti_m256i( d, 3 ) = _mm256_set1_epi32( _mm_extract_epi32( \
-                                     _mm256_castsi256_si128( s0 ), 3 ) ); \
-  casti_m256i( d, 4 ) = _mm256_set1_epi32( _mm_extract_epi32( s1,   0 ) ); \
-  casti_m256i( d, 5 ) = _mm256_set1_epi32( _mm_extract_epi32( s1,   1 ) ); \
-  casti_m256i( d, 6 ) = _mm256_set1_epi32( _mm_extract_epi32( s1,   2 ) ); \
-  casti_m256i( d, 7 ) = _mm256_set1_epi32( _mm_extract_epi32( s1,   3 ) ); \
-} while(0)
-
-
-#define mm256_bswap_intrlv_8x32_128( d, src ) \
-do { \
-  __m128i ss = mm128_bswap_32( src ); \
-  casti_m256i( d, 0 ) = _mm256_set1_epi32( _mm_extract_epi32( ss, 0 ) ); \
-  casti_m256i( d, 1 ) = _mm256_set1_epi32( _mm_extract_epi32( ss, 1 ) ); \
-  casti_m256i( d, 2 ) = _mm256_set1_epi32( _mm_extract_epi32( ss, 2 ) ); \
-  casti_m256i( d, 3 ) = _mm256_set1_epi32( _mm_extract_epi32( ss, 3 ) ); \
-} while(0)
-*/
-
-#define mm256_dintrlv_8x32_256( d0, d1, d2, d3, d4, d5, d6, d7, s ) \
-do { \
-  __m256i s0 = casti_m256i(s,0); \
-  __m256i s1 = casti_m256i(s,1); \
-  __m256i s2 = casti_m256i(s,2); \
-  __m256i s3 = casti_m256i(s,3); \
-  __m256i s4 = casti_m256i(s,4); \
-  __m256i s5 = casti_m256i(s,5); \
-  __m256i s6 = casti_m256i(s,6); \
-  __m256i s7 = casti_m256i(s,7); \
-  __m128i s0hi = _mm256_extracti128_si256( s0, 1 ); \
-  __m128i s1hi = _mm256_extracti128_si256( s1, 1 ); \
-  __m128i s2hi = _mm256_extracti128_si256( s2, 1 ); \
-  __m128i s3hi = _mm256_extracti128_si256( s3, 1 ); \
-  __m128i s4hi = _mm256_extracti128_si256( s4, 1 ); \
-  __m128i s5hi = _mm256_extracti128_si256( s5, 1 ); \
-  __m128i s6hi = _mm256_extracti128_si256( s6, 1 ); \
-  __m128i s7hi = _mm256_extracti128_si256( s7, 1 ); \
-   d0 = _mm256_set_epi32( \
-              extr32_cast128_256( s7, 0 ), extr32_cast128_256( s6, 0 ), \
-              extr32_cast128_256( s5, 0 ), extr32_cast128_256( s4, 0 ), \
-              extr32_cast128_256( s3, 0 ), extr32_cast128_256( s2, 0 ), \
-              extr32_cast128_256( s1, 0 ), extr32_cast128_256( s0, 0 ) );\
-   d1 = _mm256_set_epi32( \
-              extr32_cast128_256( s7, 1 ), extr32_cast128_256( s6, 1 ), \
-              extr32_cast128_256( s5, 1 ), extr32_cast128_256( s4, 1 ), \
-              extr32_cast128_256( s3, 1 ), extr32_cast128_256( s2, 1 ), \
-              extr32_cast128_256( s1, 1 ), extr32_cast128_256( s0, 1 ) );\
-   d2 = _mm256_set_epi32( \
-              extr32_cast128_256( s7, 2 ), extr32_cast128_256( s6, 2 ), \
-              extr32_cast128_256( s5, 2 ), extr32_cast128_256( s4, 2 ), \
-              extr32_cast128_256( s3, 2 ), extr32_cast128_256( s2, 2 ), \
-              extr32_cast128_256( s1, 2 ), extr32_cast128_256( s0, 2 ) );\
-   d3 = _mm256_set_epi32( \
-              extr32_cast128_256( s7, 3 ), extr32_cast128_256( s6, 3 ), \
-              extr32_cast128_256( s5, 3 ), extr32_cast128_256( s4, 3 ), \
-              extr32_cast128_256( s3, 3 ), extr32_cast128_256( s2, 3 ), \
-              extr32_cast128_256( s1, 3 ), extr32_cast128_256( s0, 3 ) );\
-   d4 = _mm256_set_epi32( \
-              _mm_extract_epi32( s7hi, 0 ), _mm_extract_epi32( s6hi, 0 ), \
-              _mm_extract_epi32( s5hi, 0 ), _mm_extract_epi32( s4hi, 0 ), \
-              _mm_extract_epi32( s3hi, 0 ), _mm_extract_epi32( s2hi, 0 ), \
-              _mm_extract_epi32( s1hi, 0 ), _mm_extract_epi32( s0hi, 0 ) ); \
-   d5 = _mm256_set_epi32( \
-              _mm_extract_epi32( s7hi, 1 ), _mm_extract_epi32( s6hi, 1 ), \
-              _mm_extract_epi32( s5hi, 1 ), _mm_extract_epi32( s4hi, 1 ), \
-              _mm_extract_epi32( s3hi, 1 ), _mm_extract_epi32( s2hi, 1 ), \
-              _mm_extract_epi32( s1hi, 1 ), _mm_extract_epi32( s0hi, 1 ) ); \
-   d6 = _mm256_set_epi32( \
-              _mm_extract_epi32( s7hi, 2 ), _mm_extract_epi32( s6hi, 2 ), \
-              _mm_extract_epi32( s5hi, 2 ), _mm_extract_epi32( s4hi, 2 ), \
-              _mm_extract_epi32( s3hi, 2 ), _mm_extract_epi32( s2hi, 2 ), \
-              _mm_extract_epi32( s1hi, 2 ), _mm_extract_epi32( s0hi, 2 ) ); \
-   d7 = _mm256_set_epi32( \
-              _mm_extract_epi32( s7hi, 3 ), _mm_extract_epi32( s6hi, 3 ), \
-              _mm_extract_epi32( s5hi, 3 ), _mm_extract_epi32( s4hi, 3 ), \
-              _mm_extract_epi32( s3hi, 3 ), _mm_extract_epi32( s2hi, 3 ), \
-              _mm_extract_epi32( s1hi, 3 ), _mm_extract_epi32( s0hi, 3 ) ); \
-} while(0)
-
-#define mm128_dintrlv_8x32_128( d0, d1, d2, d3, d4, d5, d6, d7, s ) \
-do { \
-   __m128i s0 = casti_m128i(s,0); \
-   __m128i s1 = casti_m128i(s,1); \
-   __m128i s2 = casti_m128i(s,2); \
-   __m128i s3 = casti_m128i(s,3); \
-   d0 = _mm_set_epi32( \
-              _mm_extract_epi32( s3, 0 ), _mm_extract_epi32( s2, 0 ), \
-              _mm_extract_epi32( s1, 0 ), _mm_extract_epi32( s0, 0 ) ); \
-   d1 = _mm_set_epi32( \
-              _mm_extract_epi32( s3, 1 ), _mm_extract_epi32( s2, 0 ), \
-              _mm_extract_epi32( s1, 1 ), _mm_extract_epi32( s0, 0 ) ); \
-   d2 = _mm_set_epi32( \
-              _mm_extract_epi32( s3, 0 ), _mm_extract_epi32( s2, 0 ), \
-              _mm_extract_epi32( s1, 0 ), _mm_extract_epi32( s0, 0 ) ); \
-   d3 = _mm_set_epi32( \
-              _mm_extract_epi32( s3, 0 ), _mm_extract_epi32( s2, 0 ), \
-              _mm_extract_epi32( s1, 0 ), _mm_extract_epi32( s0, 0 ) ); \
-   d4 = _mm_set_epi32( \
-              _mm_extract_epi32( s3, 0 ), _mm_extract_epi32( s2, 0 ), \
-              _mm_extract_epi32( s1, 0 ), _mm_extract_epi32( s0, 0 ) ); \
-   d5 = _mm_set_epi32( \
-              _mm_extract_epi32( s3, 0 ), _mm_extract_epi32( s2, 0 ), \
-              _mm_extract_epi32( s1, 0 ), _mm_extract_epi32( s0, 0 ) ); \
-   d6 = _mm_set_epi32( \
-              _mm_extract_epi32( s3, 0 ), _mm_extract_epi32( s2, 0 ), \
-              _mm_extract_epi32( s1, 0 ), _mm_extract_epi32( s0, 0 ) ); \
-   d7 = _mm_set_epi32( \
-              _mm_extract_epi32( s3, 0 ), _mm_extract_epi32( s2, 0 ), \
-              _mm_extract_epi32( s1, 0 ), _mm_extract_epi32( s0, 0 ) ); \
-} while(0)
-
-#define mm256_intrlv_4x64_256( d, s0, s1, s2, s3 ) \
-do { \
-  __m128i s0hi = _mm256_extracti128_si256( s0, 1 ); \
-  __m128i s1hi = _mm256_extracti128_si256( s1, 1 ); \
-  __m128i s2hi = _mm256_extracti128_si256( s2, 1 ); \
-  __m128i s3hi = _mm256_extracti128_si256( s3, 1 ); \
-  casti_m256i( d,0 ) = _mm256_set_epi64x( \
-                extr64_cast128_256( s3, 0 ), extr64_cast128_256( s2, 0 ), \
-                extr64_cast128_256( s1, 0 ), extr64_cast128_256( s0, 0 ) ); \
-  casti_m256i( d,1 ) = _mm256_set_epi64x( \
-                extr64_cast128_256( s3, 1 ), extr64_cast128_256( s2, 1 ), \
-                extr64_cast128_256( s1, 1 ), extr64_cast128_256( s0, 1 ) ); \
-  casti_m256i( d,2 ) = _mm256_set_epi64x( \
-                  _mm_extract_epi64( s3hi,0 ), _mm_extract_epi64( s2hi,0 ), \
-                  _mm_extract_epi64( s1hi,0 ), _mm_extract_epi64( s0hi,0 ) ); \
-  casti_m256i( d,3 ) = _mm256_set_epi64x( \
-                  _mm_extract_epi64( s3hi,1 ), _mm_extract_epi64( s2hi,1 ), \
-                  _mm_extract_epi64( s1hi,1 ), _mm_extract_epi64( s0hi,1 ) ); \
-} while(0)
-
-#define mm256_intrlv_4x64_128( d, s0, s1, s2, s3 ) \
-do { \
-  casti_m256i( d,0 ) = _mm256_set_epi64x( \
-                  _mm_extract_epi64( s3, 0 ), _mm_extract_epi64( s2, 0 ), \
-                  _mm_extract_epi64( s1, 0 ), _mm_extract_epi64( s0, 0 ) ); \
-  casti_m256i( d,1 ) = _mm256_set_epi64x( \
-                  _mm_extract_epi64( s3, 1 ), _mm_extract_epi64( s2, 1 ), \
-                  _mm_extract_epi64( s1, 1 ), _mm_extract_epi64( s0, 1 ) ); \
-} while(0)
-
-/*
-#define mm256_bswap_intrlv_4x64_256( d, src ) \
-do { \
-  __m256i s0 = mm256_bswap_32( src ); \
-  __m128i s1 = _mm256_extracti128_si256( s0, 1 ); \
-  casti_m256i( d,0 ) = _mm256_set1_epi64x( _mm_extract_epi64( \
-                                        _mm256_castsi256_si128( s0 ), 0 ) ); \
-  casti_m256i( d,1 ) = _mm256_set1_epi64x( _mm_extract_epi64( \
-                                        _mm256_castsi256_si128( s0 ), 1 ) ); \
-  casti_m256i( d,2 ) = _mm256_set1_epi64x(   _mm_extract_epi64( s1,   0 ) ); \
-  casti_m256i( d,3 ) = _mm256_set1_epi64x(   _mm_extract_epi64( s1,   1 ) ); \
-} while(0)
-
-#define mm256_bswap_intrlv_4x64_128( d, src ) \
-do { \
-  __m128i ss = mm128_bswap_32( src ); \
-  casti_m256i( d,0 ) = _mm256_set1_epi64x( _mm_extract_epi64( ss, 0 ) ); \
-  casti_m256i( d,1 ) = _mm256_set1_epi64x( _mm_extract_epi64( ss, 1 ) ); \
-} while(0)
-*/
-
-// 4 lanes of 256 bits using 64 bit interleaving (standard final hash size)
-static inline void mm256_dintrlv_4x64_256( void *d0, void *d1, void *d2,
-                            void *d3, const int n, const void *src )
-{
-   __m256i s0   = *( (__m256i*) src     );            // s[0][1:0]
-   __m256i s1   = *( (__m256i*)(src+32) );            // s[1][1:0]
-   __m256i s2   = *( (__m256i*)(src+64) );            // s[2][1:0]
-   __m256i s3   = *( (__m256i*)(src+96) );            // s[3][2:0]
-   __m128i s0hi = _mm256_extracti128_si256( s0, 1 );  // s[0][3:2]
-   __m128i s1hi = _mm256_extracti128_si256( s1, 1 );  // s[1][3:2]
-   __m128i s2hi = _mm256_extracti128_si256( s2, 1 );  // s[2][3:2]
-   __m128i s3hi = _mm256_extracti128_si256( s3, 1 );  // s[3][3:2]
-
-   casti_m256i( d0,n ) = _mm256_set_epi64x(
-              extr64_cast128_256( s3, 0 ), extr64_cast128_256( s2, 0 ),
-              extr64_cast128_256( s1, 0 ), extr64_cast128_256( s0, 0 ) );
-   casti_m256i( d1,n ) = _mm256_set_epi64x(
-              extr64_cast128_256( s3, 1 ), extr64_cast128_256( s2, 1 ),
-              extr64_cast128_256( s1, 1 ), extr64_cast128_256( s0, 1 ) );
-   casti_m256i( d2,n ) = _mm256_set_epi64x(
-              _mm_extract_epi64( s3hi, 0 ), _mm_extract_epi64( s2hi, 0 ),
-              _mm_extract_epi64( s1hi, 0 ), _mm_extract_epi64( s0hi, 0 ) );
-   casti_m256i( d3,n ) = _mm256_set_epi64x(
-              _mm_extract_epi64( s3hi, 1 ), _mm_extract_epi64( s2hi, 1 ),
-              _mm_extract_epi64( s1hi, 1 ), _mm_extract_epi64( s0hi, 1 ) );
-}
-
-
-// quarter avx2 block, 16 bytes * 4 lanes
-// 4 lanes of 128 bits using 64 bit interleaving
-// Used for last 16 bytes of 80 byte input, only used for testing.
-static inline void mm128_dintrlv_4x64_128( void *d0, void *d1, void *d2,
-                                  void *d3, const int n, const void *src )
-{
-  __m256i s0 = *( (__m256i*) src     );
-  __m256i s1 = *( (__m256i*)(src+32) );
-  __m128i s0hi = _mm256_extracti128_si256( s0, 1 );
-  __m128i s1hi = _mm256_extracti128_si256( s1, 1 );
-
-  casti_m128i( d0,n ) = _mm_set_epi64x( extr64_cast128_256( s1  , 0 ),
-                                        extr64_cast128_256( s0  , 0 ) );
-  casti_m128i( d1,n ) = _mm_set_epi64x( extr64_cast128_256( s1  , 1 ),
-                                        extr64_cast128_256( s0  , 1 ) );
-  casti_m128i( d2,n ) = _mm_set_epi64x( _mm_extract_epi64(    s1hi, 0 ),
-                                        _mm_extract_epi64(    s0hi, 0 ) );
-  casti_m128i( d3,n ) = _mm_set_epi64x( _mm_extract_epi64(    s1hi, 1 ),
-                                        _mm_extract_epi64(    s0hi, 1 ) );
-}
-
-/*
-static inline void mm256_dintrlv_2x128x256( void *d0, void *d1,
-                                                 const int n, const void *s )
-{
-   casti_m256i( d0,n ) = mm256_get_64( s, 0, 1, 4, 5 );
-   casti_m256i( d1,n ) = mm256_get_64( s, 2, 3, 6, 7 );
-}
-*/
-//
-
-#define mm256_intrlv_4x32_256( d, s0, s1, s2, s3 ) \
-do { \
-   casti_m256i( d,0 ) = _mm256_set_epi32( \
-                            mm128_extr_32( s3, 1 ), mm128_extr_32( s2, 1 ), \
-                            mm128_extr_32( s1, 1 ), mm128_extr_32( s0, 1 ), \
-                            mm128_extr_32( s3, 0 ), mm128_extr_32( s2, 0 ), \
-                            mm128_extr_32( s1, 0 ), mm128_extr_32( s0, 0 ) ); \
-   casti_m256i( d,1 ) = _mm256_set_epi32( \
-                            mm128_extr_32( s3, 3 ), mm128_extr_32( s2, 3 ), \
-                            mm128_extr_32( s1, 3 ), mm128_extr_32( s0, 3 ), \
-                            mm128_extr_32( s3, 2 ), mm128_extr_32( s2, 2 ), \
-                            mm128_extr_32( s1, 2 ), mm128_extr_32( s0, 2 ) ); \
-   casti_m256i( d,2 ) = _mm256_set_epi32( \
-                            mm128_extr_32( s3, 5 ), mm128_extr_32( s2, 5 ), \
-                            mm128_extr_32( s1, 5 ), mm128_extr_32( s0, 5 ), \
-                            mm128_extr_32( s3, 4 ), mm128_extr_32( s2, 4 ), \
-                            mm128_extr_32( s1, 4 ), mm128_extr_32( s0, 4 ) ); \
-   casti_m256i( d,3 ) = _mm256_set_epi32( \
-                            mm128_extr_32( s3, 7 ), mm128_extr_32( s2, 7 ), \
-                            mm128_extr_32( s1, 7 ), mm128_extr_32( s0, 7 ), \
-                            mm128_extr_32( s3, 6 ), mm128_extr_32( s2, 6 ), \
-                            mm128_extr_32( s1, 6 ), mm128_extr_32( s0, 6 ) ); \
-} while(0)
-
-// 256 bit versions of commmon 128 bit functions.
-static inline void mm256_intrlv_4x32( void *d, const void *s0,
-               const void *s1, const void *s2, const void *s3, int bit_len )
-{
-   mm256_intrlv_4x32_256( d     ,casti_m256i(s0,0), casti_m256i(s1,0),
-                                 casti_m256i(s2,0), casti_m256i(s3,0) );
-   if ( bit_len <= 256 ) return;
-   mm256_intrlv_4x32_256( d+128 ,casti_m256i(s0,1), casti_m256i(s1,1),
-                                 casti_m256i(s2,1), casti_m256i(s3,1) );
-   if ( bit_len <= 512 ) return;
-   if ( bit_len <= 640 )
-   {
-      mm128_intrlv_4x32_128( d+256, casti_m128i(s0,4), casti_m128i(s1,4),
-                                    casti_m128i(s2,4), casti_m128i(s3,4) );
-      return;
-   }
-   mm256_intrlv_4x32_256( d+256 ,casti_m256i(s0,2), casti_m256i(s1,2),
-                                 casti_m256i(s2,2), casti_m256i(s3,2) );
-   mm256_intrlv_4x32_256( d+384 ,casti_m256i(s0,3), casti_m256i(s1,3),
-                                 casti_m256i(s2,3), casti_m256i(s3,3) );
-}
-
-static inline void mm256_dintrlv_4x32_256( void *d0, void *d1, void *d2,
-                                           void *d3, const void *src )
-{
-   __m256i s0 = *(__m256i*) src;
-   __m256i s1 = *(__m256i*)(src+32);
-   __m256i s2 = *(__m256i*)(src+64);
-   __m256i s3 = *(__m256i*)(src+96);
-   *(__m256i*)d0 = _mm256_set_epi32(
-                  _mm256_extract_epi32( s3,4 ), _mm256_extract_epi32( s3,0 ),
-                  _mm256_extract_epi32( s2,4 ), _mm256_extract_epi32( s2,0 ),
-                  _mm256_extract_epi32( s1,4 ), _mm256_extract_epi32( s1,0 ),
-                  _mm256_extract_epi32( s0,4 ), _mm256_extract_epi32( s0,0 ) );
-   *(__m256i*)d1 = _mm256_set_epi32(
-                  _mm256_extract_epi32( s3,5 ), _mm256_extract_epi32( s3,1 ),
-                  _mm256_extract_epi32( s2,5 ), _mm256_extract_epi32( s2,1 ),
-                  _mm256_extract_epi32( s1,5 ), _mm256_extract_epi32( s1,1 ),
-                  _mm256_extract_epi32( s0,5 ), _mm256_extract_epi32( s0,1 ) );
-   *(__m256i*)d2 = _mm256_set_epi32(
-                  _mm256_extract_epi32( s3,6 ), _mm256_extract_epi32( s3,2 ),
-                  _mm256_extract_epi32( s2,6 ), _mm256_extract_epi32( s2,2 ),
-                  _mm256_extract_epi32( s1,6 ), _mm256_extract_epi32( s1,2 ),
-                  _mm256_extract_epi32( s0,6 ), _mm256_extract_epi32( s0,2 ) );
-   *(__m256i*)d3 = _mm256_set_epi32(
-                  _mm256_extract_epi32( s3,7 ), _mm256_extract_epi32( s3,3 ),
-                  _mm256_extract_epi32( s2,7 ), _mm256_extract_epi32( s2,3 ),
-                  _mm256_extract_epi32( s1,7 ), _mm256_extract_epi32( s1,3 ),
-                  _mm256_extract_epi32( s0,7 ), _mm256_extract_epi32( s0,3 ) );
-}
-
-static inline void mm256_dintrlv_4x32( void *d0, void *d1, void *d2,
-                                       void *d3, const void *s, int bit_len )
-{
-   mm256_dintrlv_4x32_256( d0    , d1    , d2    , d3    , s     );
-   if ( bit_len <= 256 ) return;
-   mm256_dintrlv_4x32_256( d0+ 32, d1+ 32, d2+ 32, d3+ 32, s+128 );
-   if ( bit_len <= 512 ) return;
-   if ( bit_len <= 640 )
-   {
-      mm128_dintrlv_4x32_128( d0+ 64, d1+ 64, d2+ 64, d3+ 64, s+256 );
-      return;
-   }
-   mm256_dintrlv_4x32_256( d0+ 64, d1+ 64, d2+ 64, d3+ 64, s+256 );
-   mm256_dintrlv_4x32_256( d0+ 96, d1+ 96, d2+ 96, d3+ 96, s+384 );
-}
-
-static inline void mm256_extr_lane_4x32( void *d, const void *s,
-                                         const int lane, const int bit_len )
-{
-  casti_m256i( d, 0 ) = mm256_get_32( s, lane   , lane+ 4, lane+ 8, lane+12,
-                                         lane+16, lane+20, lane+24, lane+28 );
-  if ( bit_len <= 256 ) return;
-  casti_m256i( d, 1 ) = mm256_get_32( s, lane+32, lane+36, lane+40, lane+44,
-                                         lane+48, lane+52, lane+56, lane+60 );
-}
-
-// Interleave 8 source buffers containing 32 bit data into the destination
-// vector
-static inline void mm256_intrlv_8x32( void *d, const void *s0,
-        const void *s1, const void *s2, const void *s3, const void *s4,
-        const void *s5, const void *s6, const void *s7, int bit_len )
-{
-   mm256_intrlv_8x32_256( d    , casti_m256i( s0,0 ), casti_m256i( s1,0 ),
-            casti_m256i( s2,0 ), casti_m256i( s3,0 ), casti_m256i( s4,0 ),
-            casti_m256i( s5,0 ), casti_m256i( s6,0 ), casti_m256i( s7,0 ) );
-   if ( bit_len <= 256 ) return;
-   mm256_intrlv_8x32_256( d+256, casti_m256i( s0,1 ), casti_m256i( s1,1 ), 
-            casti_m256i( s2,1 ), casti_m256i( s3,1 ), casti_m256i( s4,1 ),
-            casti_m256i( s5,1 ), casti_m256i( s6,1 ), casti_m256i( s7,1 ) );
-   if ( bit_len <= 512 ) return;
-   if ( bit_len <= 640 )
-   {
-      mm256_intrlv_8x32_128( d+512, casti_m128i( s0,4 ), casti_m128i( s1,4 ),
-               casti_m128i( s2,4 ), casti_m128i( s3,4 ), casti_m128i( s4,4 ),
-               casti_m128i( s5,4 ), casti_m128i( s6,4 ), casti_m128i( s7,4 ) );
-      return;
-   }
-   mm256_intrlv_8x32_256( d+512, casti_m256i( s0,2 ), casti_m256i( s1,2 ), 
-            casti_m256i( s2,2 ), casti_m256i( s3,2 ), casti_m256i( s4,2 ),
-            casti_m256i( s5,2 ), casti_m256i( s6,2 ), casti_m256i( s7,2 ) );
-   mm256_intrlv_8x32_256( d+768, casti_m256i( s0,3 ), casti_m256i( s1,3 ), 
-            casti_m256i( s2,3 ), casti_m256i( s3,3 ), casti_m256i( s4,3 ),
-            casti_m256i( s5,3 ), casti_m256i( s6,3 ), casti_m256i( s7,3 ) );
-   // bit_len == 1024
-}
-
-// A couple of mining specifi functions.
-/*
-// Interleave 80 bytes of 32 bit data for 8 lanes.
-static inline void mm256_bswap_intrlv80_8x32( void *d, const void *s )
-{
-   mm256_bswap_intrlv_8x32_256( d    , casti_m256i( s, 0 ) );
-   mm256_bswap_intrlv_8x32_256( d+256, casti_m256i( s, 1 ) );
-   mm256_bswap_intrlv_8x32_128( d+512, casti_m128i( s, 4 ) );
-}
-*/
-
-// Deinterleave 8 buffers of 32 bit data from the source buffer.
-// Sub-function can be called directly for 32 byte final hash.
-static inline void mm256_dintrlv_8x32( void *d0, void *d1, void *d2,
-                        void *d3, void *d4, void *d5, void *d6, void *d7,
-                        const void *s, int bit_len )
-{
-   mm256_dintrlv_8x32_256( casti_m256i(d0,0), casti_m256i(d1,0),
-        casti_m256i(d2,0), casti_m256i(d3,0), casti_m256i(d4,0),
-        casti_m256i(d5,0), casti_m256i(d6,0), casti_m256i(d7,0), s );
-   if ( bit_len <= 256 ) return;
-   mm256_dintrlv_8x32_256( casti_m256i(d0,1), casti_m256i(d1,1), 
-        casti_m256i(d2,1), casti_m256i(d3,1), casti_m256i(d4,1),
-        casti_m256i(d5,1), casti_m256i(d6,1), casti_m256i(d7,1), s+256 );
-   if ( bit_len <= 512 ) return;
-   // short block, final 16 bytes of input data
-   if ( bit_len <= 640 )
-   {
-      mm128_dintrlv_8x32_128( casti_m128i(d0,2), casti_m128i(d1,2), 
-           casti_m128i(d2,2), casti_m128i(d3,2), casti_m128i(d4,2),
-           casti_m128i(d5,2), casti_m128i(d6,2), casti_m128i(d7,2), s+512 );
-      return;
-   }
-   // bitlen == 1024
-   mm256_dintrlv_8x32_256( casti_m256i(d0,2), casti_m256i(d1,2), 
-        casti_m256i(d2,2), casti_m256i(d3,2), casti_m256i(d4,2),
-        casti_m256i(d5,2), casti_m256i(d6,2), casti_m256i(d7,2), s+512 );
-   mm256_dintrlv_8x32_256( casti_m256i(d0,3), casti_m256i(d1,3), 
-        casti_m256i(d2,3), casti_m256i(d3,3), casti_m256i(d4,3),
-        casti_m256i(d5,3), casti_m256i(d6,3), casti_m256i(d7,3), s+768 );
-}
-
-static inline void mm256_extr_lane_8x32( void *d, const void *s,
-                                            const int lane, const int bit_len )
-{
-  casti_m256i( d,0 ) = mm256_get_32(s, lane   , lane+  8, lane+ 16, lane+ 24,
-                                       lane+32, lane+ 40, lane+ 48, lane+ 56 );
-  if ( bit_len <= 256 ) return;
-  casti_m256i( d,1 ) = mm256_get_32(s, lane+64, lane+ 72, lane+ 80, lane+ 88,
-                                       lane+96, lane+104, lane+112, lane+120 );
-  // bit_len == 512
-}
-
-// Interleave 4 source buffers containing 64 bit data into the destination
-// buffer. Only bit_len 256, 512, 640 & 1024 are supported.
-static inline void mm256_intrlv_4x64( void *d, const void *s0,
-            const void *s1, const void *s2, const void *s3, int bit_len )
-{
-  mm256_intrlv_4x64_256( d    , casti_m256i(s0,0), casti_m256i(s1,0),
-                                casti_m256i(s2,0), casti_m256i(s3,0) );
-  if ( bit_len <= 256 ) return;
-  mm256_intrlv_4x64_256( d+128, casti_m256i(s0,1), casti_m256i(s1,1),
-                                casti_m256i(s2,1), casti_m256i(s3,1) );
-  if ( bit_len <= 512 ) return;
-  if ( bit_len <= 640 )
-  {
-    mm256_intrlv_4x64_128( d+256, casti_m128i(s0,4), casti_m128i(s1,4),
-                                  casti_m128i(s2,4), casti_m128i(s3,4) );
-    return;
-  }
-  // bit_len == 1024
-  mm256_intrlv_4x64_256( d+256, casti_m256i(s0,2), casti_m256i(s1,2),
-                                casti_m256i(s2,2), casti_m256i(s3,2) );
-  mm256_intrlv_4x64_256( d+384, casti_m256i(s0,3), casti_m256i(s1,3),
-                                casti_m256i(s2,3), casti_m256i(s3,3) );
-}
-/*
-// Interleave 80 bytes of 32 bit data for 8 lanes.
-static inline void mm256_bswap_intrlv80_4x64( void *d, const void *s )
-{
-   mm256_bswap_intrlv_4x64_256( d    , casti_m256i( s, 0 ) );
-   mm256_bswap_intrlv_4x64_256( d+128, casti_m256i( s, 1 ) );
-   mm256_bswap_intrlv_4x64_128( d+256, casti_m128i( s, 4 ) );
-}
-
-// Blend 32 byte lanes of hash from 2 sources according to control mask.
-// macro due to 256 bit value arg.
-#define mm256_blend_hash_4x64( dst, a, b, mask ) \
-do { \
-    dst[0] = _mm256_blendv_epi8( a[0], b[0], mask ); \
-    dst[1] = _mm256_blendv_epi8( a[1], b[1], mask ); \
-    dst[2] = _mm256_blendv_epi8( a[2], b[2], mask ); \
-    dst[3] = _mm256_blendv_epi8( a[3], b[3], mask ); \
-    dst[4] = _mm256_blendv_epi8( a[4], b[4], mask ); \
-    dst[5] = _mm256_blendv_epi8( a[5], b[5], mask ); \
-    dst[6] = _mm256_blendv_epi8( a[6], b[6], mask ); \
-    dst[7] = _mm256_blendv_epi8( a[7], b[7], mask ); \
-} while(0)
-*/
-
-// Deinterleave 4 buffers of 64 bit data from the source buffer.
-// bit_len must be 256, 512, 640 or 1024 bits.
-// Requires overrun padding for 640 bit len.
-static inline void mm256_dintrlv_4x64( void *d0, void *d1, void *d2,
-                                    void *d3, const void *s, int bit_len )
-{
-   mm256_dintrlv_4x64_256( d0, d1, d2, d3, 0, s );
-   if ( bit_len <= 256 ) return;
-   mm256_dintrlv_4x64_256( d0, d1, d2, d3, 1, s+128 );
-   if ( bit_len <= 512 ) return;
-   // short block, final 16 bytes of input data
-   if ( bit_len <= 640 )
-   {
-      mm128_dintrlv_4x64_128( d0, d1, d2, d3, 4, s+256 );
-      return;
-   }
-   // bit_len == 1024
-   mm256_dintrlv_4x64_256( d0, d1, d2, d3, 2, s+256 );
-   mm256_dintrlv_4x64_256( d0, d1, d2, d3, 3, s+384 );
-}
-
-// extract and deinterleave specified lane.
-#define mm256_extr_lane_4x64_256 \
-      casti_m256i( d, 0 ) = mm256_get_64( s, lane, lane+4, lane+8, lane+12 )
-static inline void mm256_extr_lane_4x64( void *d, const void *s,
-                                            const int lane, const int bit_len )
-{
-  casti_m256i( d, 0 ) = mm256_get_64( s, lane, lane+4, lane+8, lane+12 );
-  if ( bit_len <= 256 ) return;
-  casti_m256i( d, 1 ) = mm256_get_64( s, lane+16, lane+20, lane+24, lane+28 );
-  return;
-}
-
-
-// Convert from 4x32 SSE2 interleaving to 4x64 AVX2.
-// Can't do it in place
-static inline void mm256_rintrlv_4x32_4x64( void *dst, void *src,
-                                            int  bit_len )
-{
-   __m256i* d = (__m256i*)dst;
-   uint32_t *s = (uint32_t*)src;
-
-   d[0] = _mm256_set_epi32( s[ 7],s[ 3],s[ 6],s[ 2],s[ 5],s[ 1],s[ 4],s[ 0] );
-   d[1] = _mm256_set_epi32( s[15],s[11],s[14],s[10],s[13],s[ 9],s[12],s[ 8] );
-   d[2] = _mm256_set_epi32( s[23],s[19],s[22],s[18],s[21],s[17],s[20],s[16] );
-   d[3] = _mm256_set_epi32( s[31],s[27],s[30],s[26],s[29],s[25],s[28],s[24] );
-
-   if ( bit_len <= 256 ) return;
-
-   d[4] = _mm256_set_epi32( s[39],s[35],s[38],s[34],s[37],s[33],s[36],s[32] );
-   d[5] = _mm256_set_epi32( s[47],s[43],s[46],s[42],s[45],s[41],s[44],s[40] );
-   d[6] = _mm256_set_epi32( s[55],s[51],s[54],s[50],s[53],s[49],s[52],s[48] );
-   d[7] = _mm256_set_epi32( s[63],s[59],s[62],s[58],s[61],s[57],s[60],s[56] );
-
-   if ( bit_len <= 512 ) return;
-
-   d[8] = _mm256_set_epi32( s[71],s[67],s[70],s[66],s[69],s[65],s[68],s[64] );
-   d[9] = _mm256_set_epi32( s[79],s[75],s[78],s[74],s[77],s[73],s[76],s[72] );
-
-   if ( bit_len <= 640 ) return;
-
-  d[10] = _mm256_set_epi32(s[87],s[83],s[86],s[82],s[85],s[81],s[84],s[80]);
-  d[11] = _mm256_set_epi32(s[95],s[91],s[94],s[90],s[93],s[89],s[92],s[88]);
-
-  d[12] = _mm256_set_epi32(s[103],s[99],s[102],s[98],s[101],s[97],s[100],s[96]);
-  d[13] = _mm256_set_epi32(s[111],s[107],s[110],s[106],s[109],s[105],s[108],s[104]);
-  d[14] = _mm256_set_epi32(s[119],s[115],s[118],s[114],s[117],s[113],s[116],s[112]);
-  d[15] = _mm256_set_epi32(s[127],s[123],s[126],s[122],s[125],s[121],s[124],s[120]);
-   // bit_len == 1024
-}
-
-// Convert 4x64 byte (256 bit) vectors to 4x32 (128 bit) vectors for AVX
-// bit_len must be multiple of 64
-static inline void mm256_rintrlv_4x64_4x32( void *dst, void *src,
-                                            int  bit_len )
-{
-   __m256i  *d = (__m256i*)dst;
-   uint32_t *s = (uint32_t*)src;
-
-   d[0] = _mm256_set_epi32( s[ 7],s[ 5],s[ 3],s[ 1],s[ 6],s[ 4],s[ 2],s[ 0] );
-   d[1] = _mm256_set_epi32( s[15],s[13],s[11],s[ 9],s[14],s[12],s[10],s[ 8] );
-   d[2] = _mm256_set_epi32( s[23],s[21],s[19],s[17],s[22],s[20],s[18],s[16] );
-   d[3] = _mm256_set_epi32( s[31],s[29],s[27],s[25],s[30],s[28],s[26],s[24] );
-
-   if ( bit_len <= 256 ) return;
-
-   d[4] = _mm256_set_epi32( s[39],s[37],s[35],s[33],s[38],s[36],s[34],s[32] );
-   d[5] = _mm256_set_epi32( s[47],s[45],s[43],s[41],s[46],s[44],s[42],s[40] );
-   d[6] = _mm256_set_epi32( s[55],s[53],s[51],s[49],s[54],s[52],s[50],s[48] );
-   d[7] = _mm256_set_epi32( s[63],s[61],s[59],s[57],s[62],s[60],s[58],s[56] );
-
-   if ( bit_len <= 512 ) return;
-
-   d[8] = _mm256_set_epi32( s[71],s[69],s[67],s[65],s[70],s[68],s[66],s[64] );
-   d[9] = _mm256_set_epi32( s[79],s[77],s[75],s[73],s[78],s[76],s[74],s[72] );
-
-   if ( bit_len <= 640 ) return;
-
-   d[10] = _mm256_set_epi32( s[87],s[85],s[83],s[81],s[86],s[84],s[82],s[80] );
-   d[11] = _mm256_set_epi32( s[95],s[93],s[91],s[89],s[94],s[92],s[90],s[88] );
-
-   d[12] = _mm256_set_epi32( s[103],s[101],s[99],s[97],s[102],s[100],s[98],s[96] );
-   d[13] = _mm256_set_epi32( s[111],s[109],s[107],s[105],s[110],s[108],s[106],s[104] );
-   d[14] = _mm256_set_epi32( s[119],s[117],s[115],s[113],s[118],s[116],s[114],s[112] );
-   d[15] = _mm256_set_epi32( s[127],s[125],s[123],s[121],s[126],s[124],s[122],s[120] );
-   // bit_len == 1024
-}
-
-static inline void mm256_rintrlv_4x64_2x128( void *dst0, void *dst1,
-                                              const void *src, int  bit_len )
-{
-   __m256i* d0 = (__m256i*)dst0;
-   __m256i* d1 = (__m256i*)dst1;
-   uint64_t *s = (uint64_t*)src;
-
-   d0[0] = _mm256_set_epi64x( s[ 5], s[ 1], s[ 4], s[ 0] );
-   d1[0] = _mm256_set_epi64x( s[ 7], s[ 3], s[ 6], s[ 2] );
-
-   d0[1] = _mm256_set_epi64x( s[13], s[ 9], s[12], s[ 8] );
-   d1[1] = _mm256_set_epi64x( s[15], s[11], s[14], s[10] );
-
-   if ( bit_len <= 256 ) return;
-
-   d0[2] = _mm256_set_epi64x( s[21], s[17], s[20], s[16] );
-   d1[2] = _mm256_set_epi64x( s[23], s[19], s[22], s[18] );
-
-   d0[3] = _mm256_set_epi64x( s[29], s[25], s[28], s[24] );
-   d1[3] = _mm256_set_epi64x( s[31], s[27], s[30], s[26] );
-
-   if ( bit_len <= 512 ) return;
-
-   d0[4] = _mm256_set_epi64x( s[37], s[33], s[36], s[32] );
-   d1[4] = _mm256_set_epi64x( s[39], s[35], s[38], s[34] );
-
-   d0[5] = _mm256_set_epi64x( s[45], s[41], s[44], s[40] );
-   d1[5] = _mm256_set_epi64x( s[47], s[43], s[46], s[42] );
-
-   d0[6] = _mm256_set_epi64x( s[53], s[49], s[52], s[48] );
-   d1[6] = _mm256_set_epi64x( s[55], s[51], s[54], s[50] );
-
-   d0[7] = _mm256_set_epi64x( s[61], s[57], s[60], s[56] );
-   d1[7] = _mm256_set_epi64x( s[63], s[59], s[62], s[58] );
-}
-
-static inline void mm256_rintrlv_2x128_4x64( void *dst, const void *src0,
-                                         const void *src1, int  bit_len )
-{
-   __m256i* d = (__m256i*)dst;
-   uint64_t *s0 = (uint64_t*)src0;
-   uint64_t *s1 = (uint64_t*)src1;
-
-   d[ 0] = _mm256_set_epi64x( s1[2], s1[0], s0[2], s0[0] );
-   d[ 1] = _mm256_set_epi64x( s1[3], s1[1], s0[3], s0[1] );
-   d[ 2] = _mm256_set_epi64x( s1[6], s1[4], s0[6], s0[4] );
-   d[ 3] = _mm256_set_epi64x( s1[7], s1[5], s0[7], s0[5] );
-
-   if ( bit_len <= 256 ) return;
-
-   d[ 4] = _mm256_set_epi64x( s1[10], s1[ 8], s0[10], s0[ 8] );
-   d[ 5] = _mm256_set_epi64x( s1[11], s1[ 9], s0[11], s0[ 9] );
-   d[ 6] = _mm256_set_epi64x( s1[14], s1[12], s0[14], s0[12] );
-   d[ 7] = _mm256_set_epi64x( s1[15], s1[13], s0[15], s0[13] );
-
-   if ( bit_len <= 512 ) return;
-
-   d[ 8] = _mm256_set_epi64x( s1[18], s1[16], s0[18], s0[16] );
-   d[ 9] = _mm256_set_epi64x( s1[19], s1[17], s0[19], s0[17] );
-   d[10] = _mm256_set_epi64x( s1[22], s1[20], s0[22], s0[20] );
-   d[11] = _mm256_set_epi64x( s1[23], s1[21], s0[23], s0[21] );
-
-   d[12] = _mm256_set_epi64x( s1[26], s1[24], s0[26], s0[24] );
-   d[13] = _mm256_set_epi64x( s1[27], s1[25], s0[27], s0[25] );
-   d[14] = _mm256_set_epi64x( s1[30], s1[28], s0[30], s0[28] );
-   d[15] = _mm256_set_epi64x( s1[31], s1[29], s0[31], s0[29] );
-}
-
-
-static inline void mm256_intrlv_2x128( const void *d, const void *s0,
-                                      void *s1, const int bit_len )
-{
-  __m128i s1hi = _mm256_extracti128_si256( casti_m256i( s1,0 ), 1 );
-  __m128i s0hi = _mm256_extracti128_si256( casti_m256i( s0,0 ), 1 );
-  casti_m256i( d,0 ) = mm256_concat_128(
-                           _mm256_castsi256_si128( casti_m256i( s1,0 ) ),
-                           _mm256_castsi256_si128( casti_m256i( s0,0 ) ) );
-  casti_m256i( d,1 ) = mm256_concat_128( s1hi, s0hi );                  
-
-  if ( bit_len <= 256 ) return;
-  s0hi = _mm256_extracti128_si256( casti_m256i( s0,1 ), 1 );
-  s1hi = _mm256_extracti128_si256( casti_m256i( s1,1 ), 1 );
-  casti_m256i( d,2 ) = mm256_concat_128(
-                           _mm256_castsi256_si128( casti_m256i( s1,1 ) ),
-                           _mm256_castsi256_si128( casti_m256i( s0,1 ) ) );
-  casti_m256i( d,3 ) = mm256_concat_128( s1hi, s0hi );        
-
-  if ( bit_len <= 512 ) return;
-  if ( bit_len <= 640 )
-  {
-     casti_m256i( d,4 ) = mm256_concat_128(
-                           _mm256_castsi256_si128( casti_m256i( s1,2 ) ),
-                           _mm256_castsi256_si128( casti_m256i( s0,2 ) ) );
-     return;
-  }
-
-  s0hi = _mm256_extracti128_si256( casti_m256i( s0,2 ), 1 );
-  s1hi = _mm256_extracti128_si256( casti_m256i( s1,2 ), 1 );
-  casti_m256i( d,4 ) = mm256_concat_128(
-                           _mm256_castsi256_si128( casti_m256i( s1,2 ) ),
-                           _mm256_castsi256_si128( casti_m256i( s0,2 ) ) );
-  casti_m256i( d,5 ) = mm256_concat_128( s1hi, s0hi );        
-
-  s0hi = _mm256_extracti128_si256( casti_m256i( s0,3 ), 1 );
-  s1hi = _mm256_extracti128_si256( casti_m256i( s1,3 ), 1 );
-  casti_m256i( d,6 ) = mm256_concat_128(
-                           _mm256_castsi256_si128( casti_m256i( s1,3 ) ),
-                           _mm256_castsi256_si128( casti_m256i( s0,3 ) ) );
-  casti_m256i( d,7 ) = mm256_concat_128( s1hi, s0hi );        
-}
-
-// 512 is the bit len used by most, eliminate the conditionals
-static inline void mm256_dintrlv_2x128_512( void *dst0, void *dst1,
-                                            const void *s )
-{
-   __m256i *d0 = (__m256i*)dst0;
-   __m256i *d1 = (__m256i*)dst1;
-
-   __m256i s0 = casti_m256i( s, 0 );
-   __m256i s1 = casti_m256i( s, 1 );
-   d0[0] = _mm256_permute2x128_si256( s0, s1, 0x20 );
-   d1[0] = _mm256_permute2x128_si256( s0, s1, 0x31 );
-
-   s0 = casti_m256i( s, 2 );
-   s1 = casti_m256i( s, 3 );
-   d0[1] = _mm256_permute2x128_si256( s0, s1, 0x20 );
-   d1[1] = _mm256_permute2x128_si256( s0, s1, 0x31 );
-}   
-
-// Phase out usage for all 512 bit data lengths
-static inline void mm256_dintrlv_2x128( void *dst0, void *dst1, const void *s,
-                                             int bit_len )
-{
-   __m256i *d0 = (__m256i*)dst0;
-   __m256i *d1 = (__m256i*)dst1;
-
-   __m256i s0 = casti_m256i( s, 0 );
-   __m256i s1 = casti_m256i( s, 1 );
-   d0[0] = _mm256_permute2x128_si256( s0, s1, 0x20 );
-   d1[0] = _mm256_permute2x128_si256( s0, s1, 0x31 );
-
-   if ( bit_len <= 256 ) return;
-
-   s0 = casti_m256i( s, 2 );
-   s1 = casti_m256i( s, 3 );
-   d0[1] = _mm256_permute2x128_si256( s0, s1, 0x20 );
-   d1[1] = _mm256_permute2x128_si256( s0, s1, 0x31 );
-
-   if ( bit_len <= 512 ) return;
-
-   s0 = casti_m256i( s, 4 );
-   s1 = casti_m256i( s, 5 );
-   d0[2] = _mm256_permute2x128_si256( s0, s1, 0x20 );
-   d1[2] = _mm256_permute2x128_si256( s0, s1, 0x31 );
-
-   s0 = casti_m256i( s, 6 );
-   s1 = casti_m256i( s, 7 );
-   d0[3] = _mm256_permute2x128_si256( s0, s1, 0x20 );
-   d1[3] = _mm256_permute2x128_si256( s0, s1, 0x31 );
-}
-
-#undef extr64_cast128_256
-#undef extr32_cast128_256
-
-#endif // AVX
-#endif // INTRLV_AVX_H__
diff --git a/simd-utils/intrlv-avx2.h b/simd-utils/intrlv-avx2.h
deleted file mode 100644
index d4e3f77..0000000
--- a/simd-utils/intrlv-avx2.h
+++ /dev/null
@@ -1,104 +0,0 @@
-#if !defined(INTRLV_AVX2_H__)
-#define INTRLV_AVX2_H__ 1
-
-#if  defined(__AVX2__)
-
-///////////////////////////////////////////////////////////
-//
-//          AVX2 256 Bit Vectors
-//
-//  A few functions that need AVX2 for 256 bit.
-
-
-// Blend 2 vectors alternating hi & lo: { hi[n], lo[n-1], ... hi[1], lo[0] }
-#define mm256_intrlv_blend_128( hi, lo ) \
-                _mm256_blend_epi32( hi, lo, 0x0f )
-
-#define mm256_intrlv_blend_64( hi, lo ) \
-                _mm256_blend_epi32( hi, lo, 0x33 )
-
-#define mm256_intrlv_blend_32( hi, lo ) \
-           _mm256_blend_epi32( hi, lo, 0x55 )
-
-
-#define mm256_bswap_intrlv_8x32_256( d, src ) \
-do { \
-  __m256i s0 = mm256_bswap_32( src ); \
-  __m128i s1 = _mm256_extracti128_si256( s0, 1 ); \
-  casti_m256i( d, 0 ) = _mm256_set1_epi32( _mm_extract_epi32( \
-                                     _mm256_castsi256_si128( s0 ), 0 ) ); \
-  casti_m256i( d, 1 ) = _mm256_set1_epi32( _mm_extract_epi32( \
-                                     _mm256_castsi256_si128( s0 ), 1 ) ); \
-  casti_m256i( d, 2 ) = _mm256_set1_epi32( _mm_extract_epi32( \
-                                     _mm256_castsi256_si128( s0 ), 2 ) ); \
-  casti_m256i( d, 3 ) = _mm256_set1_epi32( _mm_extract_epi32( \
-                                     _mm256_castsi256_si128( s0 ), 3 ) ); \
-  casti_m256i( d, 4 ) = _mm256_set1_epi32( _mm_extract_epi32( s1,   0 ) ); \
-  casti_m256i( d, 5 ) = _mm256_set1_epi32( _mm_extract_epi32( s1,   1 ) ); \
-  casti_m256i( d, 6 ) = _mm256_set1_epi32( _mm_extract_epi32( s1,   2 ) ); \
-  casti_m256i( d, 7 ) = _mm256_set1_epi32( _mm_extract_epi32( s1,   3 ) ); \
-} while(0)
-
-#define mm256_bswap_intrlv_8x32_128( d, src ) \
-do { \
-  __m128i ss = mm128_bswap_32( src ); \
-  casti_m256i( d, 0 ) = _mm256_set1_epi32( _mm_extract_epi32( ss, 0 ) ); \
-  casti_m256i( d, 1 ) = _mm256_set1_epi32( _mm_extract_epi32( ss, 1 ) ); \
-  casti_m256i( d, 2 ) = _mm256_set1_epi32( _mm_extract_epi32( ss, 2 ) ); \
-  casti_m256i( d, 3 ) = _mm256_set1_epi32( _mm_extract_epi32( ss, 3 ) ); \
-} while(0)
-
-#define mm256_bswap_intrlv_4x64_256( d, src ) \
-do { \
-  __m256i s0 = mm256_bswap_32( src ); \
-  __m128i s1 = _mm256_extracti128_si256( s0, 1 ); \
-  casti_m256i( d,0 ) = _mm256_set1_epi64x( _mm_extract_epi64( \
-                                        _mm256_castsi256_si128( s0 ), 0 ) ); \
-  casti_m256i( d,1 ) = _mm256_set1_epi64x( _mm_extract_epi64( \
-                                        _mm256_castsi256_si128( s0 ), 1 ) ); \
-  casti_m256i( d,2 ) = _mm256_set1_epi64x(   _mm_extract_epi64( s1,   0 ) ); \
-  casti_m256i( d,3 ) = _mm256_set1_epi64x(   _mm_extract_epi64( s1,   1 ) ); \
-} while(0)
-
-#define mm256_bswap_intrlv_4x64_128( d, src ) \
-do { \
-  __m128i ss = mm128_bswap_32( src ); \
-  casti_m256i( d,0 ) = _mm256_set1_epi64x( _mm_extract_epi64( ss, 0 ) ); \
-  casti_m256i( d,1 ) = _mm256_set1_epi64x( _mm_extract_epi64( ss, 1 ) ); \
-} while(0)
-
-
-// A couple of mining specifi functions.
-
-// Interleave 80 bytes of 32 bit data for 8 lanes.
-static inline void mm256_bswap_intrlv80_8x32( void *d, const void *s )
-{
-   mm256_bswap_intrlv_8x32_256( d    , casti_m256i( s, 0 ) );
-   mm256_bswap_intrlv_8x32_256( d+256, casti_m256i( s, 1 ) );
-   mm256_bswap_intrlv_8x32_128( d+512, casti_m128i( s, 4 ) );
-}
-
-// Interleave 80 bytes of 32 bit data for 8 lanes.
-static inline void mm256_bswap_intrlv80_4x64( void *d, const void *s )
-{
-   mm256_bswap_intrlv_4x64_256( d    , casti_m256i( s, 0 ) );
-   mm256_bswap_intrlv_4x64_256( d+128, casti_m256i( s, 1 ) );
-   mm256_bswap_intrlv_4x64_128( d+256, casti_m128i( s, 4 ) );
-}
-
-// Blend 32 byte lanes of hash from 2 sources according to control mask.
-// macro due to 256 bit value arg.
-#define mm256_blend_hash_4x64( dst, a, b, mask ) \
-do { \
-    dst[0] = _mm256_blendv_epi8( a[0], b[0], mask ); \
-    dst[1] = _mm256_blendv_epi8( a[1], b[1], mask ); \
-    dst[2] = _mm256_blendv_epi8( a[2], b[2], mask ); \
-    dst[3] = _mm256_blendv_epi8( a[3], b[3], mask ); \
-    dst[4] = _mm256_blendv_epi8( a[4], b[4], mask ); \
-    dst[5] = _mm256_blendv_epi8( a[5], b[5], mask ); \
-    dst[6] = _mm256_blendv_epi8( a[6], b[6], mask ); \
-    dst[7] = _mm256_blendv_epi8( a[7], b[7], mask ); \
-} while(0)
-
-#endif // AVX2
-#endif // INTRLV_AVX2_H__
diff --git a/simd-utils/intrlv-avx512.h b/simd-utils/intrlv-avx512.h
deleted file mode 100644
index 011c6d2..0000000
--- a/simd-utils/intrlv-avx512.h
+++ /dev/null
@@ -1,679 +0,0 @@
-#if !defined(INTRLV_AVX512_H__)
-#define INTRLV_AVX512_H__ 1
-
-#if defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
-
-// SSE2 functions used in AVX512 interleaving
-
-// AVX512 block is 64 * 64 bytes
-
-// quarter avx512 block, 16 bytes * 16 lanes 
-static inline void mm128_dintrlv_16x32x128( void *d00, void *d01,
-        void *d02, void *d03, void *d04, void *d05, void *d06, void *d07,
-        void *d08, void *d09, void *d10, void *d11, void *d12, void *d13,
-        void *d14, void *d15, const int n, const void *s )
-{
-   cast_m128i( d00 ) = mm128_get_32( s, 0,  16, 32, 48 );
-   cast_m128i( d01 ) = mm128_get_32( s, 1,  17, 33, 49 );
-   cast_m128i( d02 ) = mm128_get_32( s, 2,  18, 34, 50 );
-   cast_m128i( d03 ) = mm128_get_32( s, 3,  19, 35, 51 );
-   cast_m128i( d04 ) = mm128_get_32( s, 4,  20, 36, 52 );
-   cast_m128i( d05 ) = mm128_get_32( s, 5,  21, 37, 53 );
-   cast_m128i( d06 ) = mm128_get_32( s, 6,  22, 38, 54 );
-   cast_m128i( d07 ) = mm128_get_32( s, 7,  23, 39, 55 );
-   cast_m128i( d08 ) = mm128_get_32( s, 8,  24, 40, 56 );
-   cast_m128i( d09 ) = mm128_get_32( s, 9,  25, 41, 57 );
-   cast_m128i( d10 ) = mm128_get_32( s, 10, 26, 42, 58 );
-   cast_m128i( d11 ) = mm128_get_32( s, 11, 27, 43, 59 );
-   cast_m128i( d12 ) = mm128_get_32( s, 12, 28, 44, 60 );
-   cast_m128i( d13 ) = mm128_get_32( s, 13, 29, 45, 61 );
-   cast_m128i( d14 ) = mm128_get_32( s, 14, 30, 46, 62 );
-   cast_m128i( d15 ) = mm128_get_32( s, 15, 31, 47, 63 );
-}
-
-// quarter avx512 block, 32 bytes * 8 lanes
-// 8 lanes of 128 bits using 64 bit interleaving
-// Used for last 16 bytes of 80 byte input, only used for testing.
-static inline void mm128_dintrlv_8x64x128( void *d0, void *d1, void *d2,
-                         void *d3, void *d4, void *d5, void *d6, void *d7,
-                         const int n, const void *s )
-{
-   casti_m128i( d0,n ) = mm128_get_64( s, 0,  8 );
-   casti_m128i( d1,n ) = mm128_get_64( s, 1,  9 );
-   casti_m128i( d2,n ) = mm128_get_64( s, 2, 10 );
-   casti_m128i( d3,n ) = mm128_get_64( s, 3, 11 );
-   casti_m128i( d4,n ) = mm128_get_64( s, 4, 12 );
-   casti_m128i( d5,n ) = mm128_get_64( s, 5, 13 );
-   casti_m128i( d6,n ) = mm128_get_64( s, 6, 14 );
-   casti_m128i( d7,n ) = mm128_get_64( s, 7, 15 );
-}
-
-static inline void mm128_dintrlv_4x128x128( void *d0, void *d1, void *d2,
-                                void *d3, const int n, const void *s )
-{
-   casti_m128i( d0,n ) = mm128_get_64( s, 0, 1 );
-   casti_m128i( d1,n ) = mm128_get_64( s, 2, 3 );
-   casti_m128i( d2,n ) = mm128_get_64( s, 4, 5 );
-   casti_m128i( d3,n ) = mm128_get_64( s, 5, 7 );
-}
-
-// AVX2 functions Used in AVX512 interleaving
-
-static inline void mm256_dintrlv_16x32x256( void *d00, void *d01,
-                 void *d02, void *d03, void *d04, void *d05,
-            void *d06, void *d07, void *d08, void *d09,
-            void *d10, void *d11, void *d12, void *d13,
-            void *d14, void *d15, const int n, const void *s )
-{
-   casti_m256i( d00,n ) = mm256_get_32( s,  0, 16, 32, 48, 64, 80, 96,112 );
-   casti_m256i( d01,n ) = mm256_get_32( s,  1, 17, 33, 49, 65, 81, 97,113 );
-   casti_m256i( d02,n ) = mm256_get_32( s,  2, 18, 34, 50, 66, 82, 98,114 );
-   casti_m256i( d03,n ) = mm256_get_32( s,  3, 19, 35, 51, 67, 83, 99,115 );
-   casti_m256i( d04,n ) = mm256_get_32( s,  4, 20, 36, 52, 68, 84,100,116 );
-   casti_m256i( d05,n ) = mm256_get_32( s,  5, 21, 37, 53, 69, 85,101,117 );
-   casti_m256i( d06,n ) = mm256_get_32( s,  6, 22, 38, 54, 70, 86,102,118 );
-   casti_m256i( d07,n ) = mm256_get_32( s,  7, 23, 39, 55, 71, 87,103,119 );
-   casti_m256i( d08,n ) = mm256_get_32( s,  8, 24, 40, 56, 72, 88,104,120 );
-   casti_m256i( d09,n ) = mm256_get_32( s,  9, 25, 41, 57, 73, 89,105,121 );
-   casti_m256i( d10,n ) = mm256_get_32( s, 10, 26, 42, 58, 74, 90,106,122 );
-   casti_m256i( d11,n ) = mm256_get_32( s, 11, 27, 43, 59, 75, 91,107,123 );
-   casti_m256i( d12,n ) = mm256_get_32( s, 12, 28, 44, 60, 76, 92,108,124 );
-   casti_m256i( d13,n ) = mm256_get_32( s, 13, 29, 45, 61, 77, 93,109,125 );
-   casti_m256i( d14,n ) = mm256_get_32( s, 14, 30, 46, 62, 78, 94,110,126 );
-   casti_m256i( d15,n ) = mm256_get_32( s, 15, 31, 47, 63, 79, 95,111,127 );
-}
-
-// 8 lanes of 256 bits using 64 bit interleaving (standard final hash size)
-static inline void mm256_dintrlv_8x64x256( void *d0, void *d1, void *d2,
-                            void *d3, void *d4, void *d5, void *d6, void *d7,
-                            const int n, const void *s )
-{
-   casti_m256i( d0,n ) = mm256_get_64( s,  0,  8, 16, 24 );
-   casti_m256i( d1,n ) = mm256_get_64( s,  1,  9, 17, 25 );
-   casti_m256i( d2,n ) = mm256_get_64( s,  2, 10, 18, 26 );
-   casti_m256i( d3,n ) = mm256_get_64( s,  3, 11, 19, 27 );
-   casti_m256i( d4,n ) = mm256_get_64( s,  4, 12, 20, 28 );
-   casti_m256i( d5,n ) = mm256_get_64( s,  5, 13, 21, 29 );
-   casti_m256i( d6,n ) = mm256_get_64( s,  6, 14, 22, 30 );
-   casti_m256i( d7,n ) = mm256_get_64( s,  7, 15, 23, 31 );
-}
-
-static inline void mm256_dintrlv_4x128x256( void *d0, void *d1, void *d2,
-                                void *d3, const int n, const void *s )
-{
-   casti_m256i( d0,n ) = mm256_get_64( s,  0,  1,  8,  9 );
-   casti_m256i( d1,n ) = mm256_get_64( s,  2,  3, 10, 11 );
-   casti_m256i( d2,n ) = mm256_get_64( s,  4,  5, 12, 13 );
-   casti_m256i( d3,n ) = mm256_get_64( s,  6,  7, 14, 15 );
-}
-
-// AVX 512 helper functions.
-//
-// Macro functions returning vector.
-// Abstracted typecasting, avoid temp pointers.
-// Source arguments may be any 64 or 32 byte aligned pointer as appropriate.
-
-#define mm512_put_64( s0, s1, s2, s3, s4, s5, s6, s7 ) \
-  _mm512_set_epi64( *((const uint64_t*)(s7)), *((const uint64_t*)(s6)), \
-                    *((const uint64_t*)(s5)), *((const uint64_t*)(s4)), \
-                    *((const uint64_t*)(s3)), *((const uint64_t*)(s2)), \
-                    *((const uint64_t*)(s1)), *((const uint64_t*)(s0)) ) 
-
-#define mm512_put_32( s00, s01, s02, s03, s04, s05, s06, s07, \
-                      s08, s09, s10, s11, s12, s13, s14, s15 ) \
-  _mm512_set_epi32( *((const uint32_t*)(s15)), *((const uint32_t*)(s14)), \
-                    *((const uint32_t*)(s13)), *((const uint32_t*)(s12)), \
-                    *((const uint32_t*)(s11)), *((const uint32_t*)(s10)), \
-                    *((const uint32_t*)(s09)), *((const uint32_t*)(s08)), \
-                    *((const uint32_t*)(s07)), *((const uint32_t*)(s06)), \
-                    *((const uint32_t*)(s05)), *((const uint32_t*)(s04)), \
-                    *((const uint32_t*)(s03)), *((const uint32_t*)(s02)), \
-                    *((const uint32_t*)(s01)), *((const uint32_t*)(s00)) ) 
-
-#define mm512_get_64( s, i0, i1, i2, i3, i4, i5, i6, i7 ) \
-  _mm512_set_epi64( ((const uint64_t*)(s))[i7], ((const uint64_t*)(s))[i6], \
-                    ((const uint64_t*)(s))[i5], ((const uint64_t*)(s))[i4], \
-                    ((const uint64_t*)(s))[i3], ((const uint64_t*)(s))[i2], \
-                    ((const uint64_t*)(s))[i1], ((const uint64_t*)(s))[i0] )
-
-#define mm512_get_32( s, i00, i01, i02, i03, i04, i05, i06, i07, \
-                         i08, i09, i10, i11, i12, i13, i14, i15 ) \
-  _mm512_set_epi32( ((const uint32_t*)(s))[i15], ((const uint32_t*)(s))[i14], \
-                    ((const uint32_t*)(s))[i13], ((const uint32_t*)(s))[i12], \
-                    ((const uint32_t*)(s))[i11], ((const uint32_t*)(s))[i10], \
-                    ((const uint32_t*)(s))[i09], ((const uint32_t*)(s))[i08], \
-                    ((const uint32_t*)(s))[i07], ((const uint32_t*)(s))[i06], \
-                    ((const uint32_t*)(s))[i05], ((const uint32_t*)(s))[i04], \
-                    ((const uint32_t*)(s))[i03], ((const uint32_t*)(s))[i02], \
-                    ((const uint32_t*)(s))[i01], ((const uint32_t*)(s))[i00] )
-
-// AVX512 has no blend, can be done with permute2xvar but at what cost?
-// Can also be done with shifting and mask-or'ing for 3 instructins with
-// 1 dependency. Finally it can be done with 1 _mm512_set but with 8 64 bit
-// array index calculations and 8 pointer reads.
-
-// Blend 2 vectors alternating hi & lo: { hi[n], lo[n-1], ... hi[1]. lo[0] }
-#define mm512_interleave_blend_128( hi, lo ) \
-  _mm256_permute2xvar_epi64( hi, lo, _mm512_set_epi64( \
-                            0x7, 0x6, 0x5, 0x4, 0xb, 0xa, 0x9, 0x8 )
-
-#define mm512_interleave_blend_64( hi, lo ) \
-  _mm256_permute2xvar_epi64( hi, lo, _mm512_set_epi64( \
-                            0x7, 0x6, 0xd, 0xc, 0x3, 0x2, 0x9, 0x8 )
-
-#define mm512_interleave_blend_32( hi, lo ) \
-  _mm256_permute2xvar_epi32( hi, lo, _mm512_set_epi32( \
-                  0x0f, 0x1e, 0x0d, 0x1c, 0x0b, 0x1a, 0x09, 0x18, \
-                            0x07, 0x16, 0x05, 0x14, 0x03, 0x12, 0x01, 0x10 )
-//
-
-static inline void mm512_intrlv_16x32x512( void *d, const void *s00,
-     const void *s01, const void *s02, const void *s03, const void *s04,
-     const void *s05, const void *s06, const void *s07, const void *s08,
-     const void *s09, const void *s10, const void *s11, const void *s12,
-     const void *s13, const void *s14, const void *s15 )
-{
-   casti_m512i( d, 0 ) = mm512_put_32(
-         s00,    s01,    s02,    s03,    s04,    s05,    s06,    s07,
-              s08,    s09,    s10,    s11,    s12,    s13,    s14,    s15 );
-   casti_m512i( d, 1 ) = mm512_put_32(
-         s00+ 4, s01+ 4, s02+ 4, s03+ 4, s04+ 4, s05+ 4, s06+ 4, s07+ 4,
-              s08+ 4, s09+ 4, s10+ 4, s11+ 4, s12+ 4, s13+ 4, s14+ 4, s15+ 4 );
-   casti_m512i( d, 2 ) = mm512_put_32(
-              s00+ 8, s01+ 8, s02+ 8, s03+ 8, s04+ 8, s05+ 8, s06+ 8, s07+ 8,
-              s08+ 8, s09+ 8, s10+ 8, s11+ 8, s12+ 8, s13+ 8, s14+ 8, s15+ 8 );
-   casti_m512i( d, 3 ) = mm512_put_32(
-         s00+12, s01+12, s02+12, s03+12, s04+12, s05+12, s06+12, s07+12,
-              s08+12, s09+12, s10+12, s11+12, s12+12, s13+12, s14+12, s15+12 );
-   casti_m512i( d, 4 ) = mm512_put_32(
-         s00+16, s01+16, s02+16, s03+16, s04+16, s05+16, s06+16, s07+16,
-              s08+16, s09+16, s10+16, s11+16, s12+16, s13+16, s14+16, s15+16 );
-   casti_m512i( d, 5 ) = mm512_put_32(
-         s00+20, s01+20, s02+20, s03+20, s04+20, s05+20, s06+20, s07+20,
-              s08+20, s09+20, s10+20, s11+20, s12+20, s13+20, s14+20, s15+20 );
-   casti_m512i( d, 6 ) = mm512_put_32(
-         s00+24, s01+24, s02+24, s03+24, s04+24, s05+24, s06+24, s07+24,
-              s08+24, s09+24, s10+24, s11+24, s12+24, s13+24, s14+24, s15+24 );
-   casti_m512i( d, 7 ) = mm512_put_32(
-         s00+28, s01+28, s02+28, s03+28, s04+28, s05+28, s06+28, s07+28,
-              s08+28, s09+28, s10+28, s11+28, s12+28, s13+28, s14+28, s15+28 );
-   casti_m512i( d, 8 ) = mm512_put_32(
-         s00+32, s01+28, s02+28, s03+28, s04+32, s05+28, s06+28, s07+28,
-              s08+32, s09+28, s10+28, s11+28, s12+32, s13+28, s14+28, s15+28 );
-   casti_m512i( d, 9 ) = mm512_put_32(
-         s00+36, s01+28, s02+28, s03+28, s04+36, s05+28, s06+28, s07+28,
-              s08+36, s09+28, s10+28, s11+28, s12+36, s13+28, s14+28, s15+28 );
-   casti_m512i( d,10 ) = mm512_put_32(
-         s00+40, s01+28, s02+28, s03+28, s04+40, s05+28, s06+28, s07+28,
-              s08+40, s09+28, s10+28, s11+28, s12+40, s13+28, s14+28, s15+28 );
-   casti_m512i( d,11 ) = mm512_put_32(
-         s00+44, s01+28, s02+28, s03+28, s04+44, s05+28, s06+28, s07+28,
-              s08+44, s09+28, s10+28, s11+28, s12+44, s13+28, s14+28, s15+28 );
-   casti_m512i( d,12 ) = mm512_put_32(
-         s00+48, s01+28, s02+28, s03+28, s04+48, s05+28, s06+28, s07+28,
-              s08+48, s09+28, s10+28, s11+28, s12+48, s13+28, s14+28, s15+28 );
-   casti_m512i( d,13 ) = mm512_put_32(
-         s00+52, s01+28, s02+28, s03+28, s04+52, s05+28, s06+28, s07+28,
-              s08+52, s09+28, s10+28, s11+28, s12+52, s13+28, s14+28, s15+28 );
-   casti_m512i( d,14 ) = mm512_put_32(
-         s00+56, s01+28, s02+28, s03+28, s04+56, s05+28, s06+28, s07+28,
-              s08+56, s09+28, s10+28, s11+28, s12+56, s13+28, s14+28, s15+28 );
-   casti_m512i( d,15 ) = mm512_put_32(
-         s00+60, s01+28, s02+28, s03+28, s04+60, s05+28, s06+28, s07+28,
-              s08+60, s09+28, s10+28, s11+28, s12+60, s13+28, s14+28, s15+28 );
-}
-
-static inline void mm512_intrlv_16x32x256( void *d, const void *s00,
-     const void *s01, const void *s02, const void *s03, const void *s04,
-     const void *s05, const void *s06, const void *s07, const void *s08,
-     const void *s09, const void *s10, const void *s11, const void *s12,
-     const void *s13, const void *s14, const void *s15 )
-{
-   casti_m512i( d, 0 ) = mm512_put_32(
-             s00,    s01,    s02,    s03,    s04,    s05,    s06,    s07,
-             s08,    s09,    s10,    s11,    s12,    s13,    s14,    s15 );
-   casti_m512i( d, 1 ) = mm512_put_32(
-        s00+ 4, s01+ 4, s02+ 4, s03+ 4, s04+ 4, s05+ 4, s06+ 4, s07+ 4,
-             s08+ 4, s09+ 4, s10+ 4, s11+ 4, s12+ 4, s13+ 4, s14+ 4, s15+ 4 );
-   casti_m512i( d, 2 ) = mm512_put_32(
-        s00+ 8, s01+ 8, s02+ 8, s03+ 8, s04+ 8, s05+ 8, s06+ 8, s07+ 8,
-             s08+ 8, s09+ 8, s10+ 8, s11+ 8, s12+ 8, s13+ 8, s14+ 8, s15+ 8 );
-   casti_m512i( d, 3 ) = mm512_put_32(
-        s00+12, s01+12, s02+12, s03+12, s04+12, s05+12, s06+12, s07+12,
-             s08+12, s09+12, s10+12, s11+12, s12+12, s13+12, s14+12, s15+12 );
-   casti_m512i( d, 4 ) = mm512_put_32(
-             s00+16, s01+16, s02+16, s03+16, s04+16, s05+16, s06+16, s07+16,
-             s08+16, s09+16, s10+16, s11+16, s12+16, s13+16, s14+16, s15+16 );
-   casti_m512i( d, 5 ) = mm512_put_32(
-        s00+20, s01+20, s02+20, s03+20, s04+20, s05+20, s06+20, s07+20,
-             s08+20, s09+20, s10+20, s11+20, s12+20, s13+20, s14+20, s15+20 );
-   casti_m512i( d, 6 ) = mm512_put_32(
-        s00+24, s01+24, s02+24, s03+24, s04+24, s05+24, s06+24, s07+24,
-             s08+24, s09+24, s10+24, s11+24, s12+24, s13+24, s14+24, s15+24 );
-   casti_m512i( d, 7 ) = mm512_put_32(
-        s00+28, s01+28, s02+28, s03+28, s04+28, s05+28, s06+28, s07+28,
-             s08+28, s09+28, s10+28, s11+28, s12+28, s13+28, s14+28, s15+28 );
-}
-
-// Last 16 bytes of input
-static inline void mm512_intrlv_16x32x128( void *d, const void *s00,
-     const void *s01, const void *s02, const void *s03, const void *s04,
-     const void *s05, const void *s06, const void *s07, const void *s08,
-     const void *s09, const void *s10, const void *s11, const void *s12,
-     const void *s13, const void *s14, const void *s15 )
-{
-   casti_m512i( d, 0 ) = mm512_put_32(
-        s00,    s01,    s02,    s03,    s04,    s05,    s06,    s07,
-        s08,    s09,    s10,    s11,    s12,    s13,    s14,    s15 );
-   casti_m512i( d, 1 ) = mm512_put_32(
-        s00+ 4, s01+ 4, s02+ 4, s03+ 4, s04+ 4, s05+ 4, s06+ 4, s07+ 4,
-             s08+ 4, s09+ 4, s10+ 4, s11+ 4, s12+ 4, s13+ 4, s14+ 4, s15+ 4 );
-   casti_m512i( d, 2 ) = mm512_put_32(
-        s00+ 8, s01+ 8, s02+ 8, s03+ 8, s04+ 8, s05+ 8, s06+ 8, s07+ 8,
-             s08+ 8, s09+ 8, s10+ 8, s11+ 8, s12+ 8, s13+ 8, s14+ 8, s15+ 8 );
-   casti_m512i( d, 3 ) = mm512_put_32(
-        s00+12, s01+12, s02+12, s03+12, s04+12, s05+12, s06+12, s07+12,
-             s08+12, s09+12, s10+12, s11+12, s12+12, s13+12, s14+12, s15+12 );
-}
-
-// can be called directly for 64 byte hash.
-static inline void mm512_dintrlv_16x32x512( void *d00, void *d01,
-                void *d02, void *d03, void *d04, void *d05, void *d06,
-                void *d07, void *d08, void *d09, void *d10, void *d11,
-                void *d12, void *d13, void *d14, void *d15, const int n,
-      const void *s )
-{
-   casti_m512i(d00,n) = mm512_get_32( s,  0, 16, 32, 48, 64, 80, 96,112,
-                              128,144,160,176,192,208,224,240 );
-   casti_m512i(d01,n) = mm512_get_32( s,  1, 17, 33, 49, 65, 81, 97,113,
-                              129,145,161,177,193,209,225,241 );
-   casti_m512i(d02,n) = mm512_get_32( s,  2, 18, 34, 50, 66, 82, 98,114,
-                    130,146,162,178,194,210,226,242 );
-   casti_m512i(d03,n) = mm512_get_32( s,  3, 19, 35, 51, 67, 83, 99,115,
-                                        131,147,163,179,195,211,227,243 );
-   casti_m512i(d04,n) = mm512_get_32( s,  4, 20, 36, 52, 68, 84,100,116,
-                              132,148,164,180,196,212,228,244 );
-   casti_m512i(d05,n) = mm512_get_32( s,  5, 21, 37, 53, 69, 85,101,117,
-                                        133,149,165,181,197,213,229,245 );
-   casti_m512i(d06,n) = mm512_get_32( s,  6, 22, 38, 54, 70, 86,102,118,
-                                        134,150,166,182,198,214,230,246 );
-   casti_m512i(d07,n) = mm512_get_32( s,  7, 23, 39, 55, 71, 87,103,119,
-                              135,151,167,183,199,215,231,247 );
-   casti_m512i(d08,n) = mm512_get_32( s,  8, 24, 40, 56, 72, 88,104,120,
-                              136,152,168,184,200,216,232,248 );
-   casti_m512i(d09,n) = mm512_get_32( s,  9, 25, 41, 57, 73, 89,105,121,
-                              137,153,169,185,201,217,233,249 );
-   casti_m512i(d10,n) = mm512_get_32( s, 10, 26, 42, 58, 74, 90,106,122,
-                              138,154,170,186,202,218,234,250 );
-   casti_m512i(d11,n) = mm512_get_32( s, 11, 27, 43, 59, 75, 91,107,123,
-                              139,155,171,187,203,219,235,251 );
-   casti_m512i(d12,n) = mm512_get_32( s, 12, 28, 44, 60, 76, 92,108,124,
-                              140,156,172,188,204,220,236,252 );
-   casti_m512i(d13,n) = mm512_get_32( s, 13, 29, 45, 61, 77, 93,109,125,
-                              141,157,173,189,205,221,237,253 );
-   casti_m512i(d14,n) = mm512_get_32( s, 14, 30, 46, 62, 78, 94,110,126,
-                                   142,158,174,190,206,222,238,254 );
-   casti_m512i(d15,n) = mm512_get_32( s, 15, 31, 47, 63, 79, 95,111,127,
-                                    143,159,175,191,207,223,239,255 );
-}
-
-static inline void mm512_intrlv_8x64x512( void *d, const void *s0,
-                   const void *s1, const void *s2, const void *s3,
-                   const void *s4, const void *s5, const void *s6,
-                   const void *s7 )
-{
-  casti_m512i( d,0 ) = mm512_put_64( s0,    s1,    s2,    s3,
-                                     s4,    s5,    s6,    s7 );
-  casti_m512i( d,1 ) = mm512_put_64( s0+ 8, s1+ 8, s2+ 8, s3+ 8,
-                                     s4+ 8, s5+ 8, s6+ 8, s7+ 8 );
-  casti_m512i( d,2 ) = mm512_put_64( s0+16, s1+16, s2+16, s3+16,
-                                     s4+16, s5+16, s6+16, s7+16 );
-  casti_m512i( d,3 ) = mm512_put_64( s0+24, s1+24, s2+24, s3+24,
-                                     s4+24, s5+24, s6+24, s7+24 );
-  casti_m512i( d,4 ) = mm512_put_64( s0+32, s1+32, s2+32, s3+32,
-                                     s4+32, s5+32, s6+32, s7+32 );
-  casti_m512i( d,5 ) = mm512_put_64( s0+40, s1+40, s2+40, s3+40,
-                                     s4+40, s5+40, s6+40, s7+40 );
-  casti_m512i( d,6 ) = mm512_put_64( s0+48, s1+48, s2+48, s3+48,
-                                     s4+48, s5+48, s6+48, s7+48 );
-  casti_m512i( d,7 ) = mm512_put_64( s0+56, s1+56, s2+56, s3+56,
-                                     s4+56, s5+56, s6+56, s7+56 );
-}
-
-static inline void mm512_intrlv_8x64x256( void *d, const void *s0,
-                   const void *s1, const void *s2, const void *s3,
-                   const void *s4, const void *s5, const void *s6,
-                   const void *s7 )
-{
-  casti_m512i( d,0 ) = mm512_put_64( s0,    s1,    s2,    s3,
-                                     s4,    s5,    s6,    s7 );
-  casti_m512i( d,1 ) = mm512_put_64( s0+ 8, s1+ 8, s2+ 8, s3+ 8,
-                                     s4+ 8, s5+ 8, s6+ 8, s7+ 8 );
-  casti_m512i( d,2 ) = mm512_put_64( s0+16, s1+16, s2+16, s3+16,
-                                     s4+16, s5+16, s6+16, s7+16 );
-  casti_m512i( d,3 ) = mm512_put_64( s0+24, s1+24, s2+24, s3+24,
-                                     s4+24, s5+24, s6+24, s7+24 );
-}
-
-
-// 8 lanes of 512 bits using 64 bit interleaving (typical intermediate hash) 
-static inline void mm512_dintrlv_8x64x512( void *d0, void *d1, void *d2,
-                            void *d3, void *d4, void *d5, void *d6, void *d7,
-             const int n, const void *s )
-{
-   casti_m512i( d0,n ) = mm512_get_64( s, 0,  8, 16, 24, 32, 40, 48, 56 );
-   casti_m512i( d1,n ) = mm512_get_64( s, 1,  9, 17, 25, 33, 41, 49, 57 );
-   casti_m512i( d2,n ) = mm512_get_64( s, 2, 10, 18, 26, 34, 42, 50, 58 );
-   casti_m512i( d3,n ) = mm512_get_64( s, 3, 11, 19, 27, 35, 43, 51, 59 );
-   casti_m512i( d4,n ) = mm512_get_64( s, 4, 12, 20, 28, 36, 44, 52, 60 );
-   casti_m512i( d5,n ) = mm512_get_64( s, 5, 13, 21, 29, 37, 45, 53, 61 );
-   casti_m512i( d6,n ) = mm512_get_64( s, 6, 14, 22, 30, 38, 46, 54, 62 );
-   casti_m512i( d7,n ) = mm512_get_64( s, 7, 15, 23, 31, 39, 47, 55, 63 );
-}
-
-static inline void mm512_dintrlv_4x128x512( void *d0, void *d1, void *d2,
-                                void *d3, const int n, const void *s )
-{
-   casti_m512i( d0,n ) = mm512_get_64( s, 0, 1,  8,  9, 16, 17, 24, 25 );
-   casti_m512i( d1,n ) = mm512_get_64( s, 2, 3, 10, 11, 18, 19, 16, 27 );
-   casti_m512i( d2,n ) = mm512_get_64( s, 4, 5, 12, 13, 20, 21, 28, 29 );
-   casti_m512i( d3,n ) = mm512_get_64( s, 6, 7, 14, 15, 22, 23, 30, 31 );
-}
-
-// AVX-512 user facing functions.
-
-static inline void mm512_intrlv_16x32( void *d, const void *s00,
-    const void *s01, const void *s02, const void *s03, const void *s04,
-    const void *s05, const void *s06, const void *s07, const void *s08,
-    const void *s09, const void *s10, const void *s11, const void *s12,
-    const void *s13, const void *s14, const void *s15, int bit_len )
-{
-   if ( bit_len <= 256 )
-   {
-      mm512_intrlv_16x32x256( d, s00, s01, s02, s03, s04, s05, s06, s07,
-                                     s08, s09, s10, s11, s12, s13, s14, s15 );
-      return;
-   }
-   mm512_intrlv_16x32x512( d, s00, s01, s02, s03, s04, s05, s06, s07,
-                        s08, s09, s10, s11, s12, s13, s14, s15 );
-   if ( bit_len <= 512 ) return;
-   if ( bit_len <= 640 )
-   {
-
-      mm512_intrlv_16x32x128( d+1024, s00+64, s01+64, s02+64, s03+64,
-                s04+64, s05+64, s06+64, s07+64, s08+64, s09+64,
-           s10+64, s11+64, s12+64, s13+64, s14+64, s15+64 );
-      return;
-   }
-   mm512_intrlv_16x32x512( d+1024, s00+64, s01+64, s02+64, s03+64,
-                       s04+64, s05+64, s06+64, s07+64, s08+64, s09+64,
-                       s10+64, s11+64, s12+64, s13+64, s14+64, s15+64 );
-   // bit_len == 1024
-}
-
-// sub-functions can be called directly for 32 & 64 byte hash.
-static inline void mm512_dintrlv_16x32( void *d00, void *d01, void *d02,
-            void *d03, void *d04, void *d05, void *d06, void *d07, void *d08,
-            void *d09, void *d10, void *d11, void *d12, void *d13, void *d14,
-            void *d15, const void *src, const int bit_len )
-{
-   if ( bit_len <= 256 )
-   {
-      mm256_dintrlv_16x32x256( d00, d01, d02, d03, d04, d05, d06, d07,
-                                    d08, d09, d10, d11, d12, d13, d14, d15,
-                                    0,src );
-      return;
-   }
-   mm512_dintrlv_16x32x512( d00, d01, d02, d03, d04, d05, d06, d07,
-                                 d08, d09, d10, d11, d12, d13, d14, d15,
-                                 0, src );
-   if ( bit_len <= 512 ) return;
-   if ( bit_len <= 640 )
-   {
-      // short block, final 16 bytes of input data.
-      mm128_dintrlv_16x32x128( d00, d01, d02, d03, d04, d05, d06, d07,
-                          d08, d09, d10, d11, d12, d13, d14, d15,
-                1, src+1024 );
-      return;
-   }
-   // bit_len == 1024
-   mm512_dintrlv_16x32x512( d00, d01, d02, d03, d04, d05, d06, d07,
-                       d08, d09, d10, d11, d12, d13, d14, d15,
-             1, src+1024 );
-}
-
-static inline void mm512_extr_lane_16x32( void *dst, const void *src,
-                                            const int lane, const int bit_len )
-{
-  if ( bit_len <= 256 )
-  {
-     cast_m256i( dst ) = mm256_get_32( src, lane, lane+16, lane+32, lane+48,
-                               lane+64, lane+80, lane+96, lane+112 );
-     return;
-  }
-  cast_m512i( dst ) = mm512_get_32( src, lane, lane+ 16, lane+ 32, lane+ 48,
-                 lane+ 64, lane+ 80, lane+ 96, lane+112, lane+128, lane+144,
-            lane+160, lane+176, lane+192, lane+208, lane+224, lane+248 );
-}
-
-//
-
-static inline void mm512_intrlv_8x64( void *d, const void *s0,
-                   const void *s1, const void *s2, const void *s3,
-                   const void *s4, const void *s5, const void *s6,
-                   const void *s7, int bit_len )
-{
-   if ( bit_len <= 256 )
-   {
-      mm512_intrlv_8x64x256( d, s0, s1, s2, s3, s4, s5, s6, s7 );
-      return;
-   }
-   mm512_intrlv_8x64x512( d, s0, s1, s2, s3, s4, s5, s6, s7 );
-   if ( bit_len <= 512 ) return;
-   if ( bit_len <= 640 )
-   {
-      casti_m512i( d, 8 ) = mm512_put_64( s7+64, s6+64, s5+64, s4+64,
-                                          s3+64, s2+64, s1+64, s0+64 );
-      casti_m512i( d, 9 ) = mm512_put_64( s7+72, s6+72, s5+72, s4+72,
-                                          s3+72, s2+72, s1+72, s0+72 );
-      return;
-   }
-   // bitlen == 1024
-   mm512_intrlv_8x64x512( d+512, s0+64, s1+64, s2+64, s3+64,
-                           s4+64, s5+64, s6+64, s7+64 );
-}
-
-
-static inline void mm512_dintrlv_8x64( void *d0, void *d1, void *d2,
-                        void *d3, void *d4, void *d5, void *d6, void *d7,
-                        const void *s, const int bit_len )
-{
-   if ( bit_len <= 256 )
-   {
-      mm256_dintrlv_8x64x256( d0, d1, d2, d3, d4, d5, d6, d7, 0, s );
-      return;
-   }
-   mm512_dintrlv_8x64x512( d0, d1, d2, d3, d4, d5, d6, d7, 0, s );
-   if ( bit_len <= 512 ) return;
-   if ( bit_len <= 640 )
-   {
-      // short block, final 16 bytes of input data.
-      mm128_dintrlv_8x64x128( d0, d1, d2, d3, d4, d5, d6, d7, 1, s+512 );
-      return;
-   }
-   // bit_len == 1024
-   mm512_dintrlv_8x64x512( d0, d1, d2, d3, d4, d5, d6, d7, 1, s+512 );
-}
-
-// Extract one lane from 64 bit interleaved data
-static inline void mm512_extr_lane_8x64( void *d, const void *s,
-                                            const int lane, const int bit_len )
-{
-  if ( bit_len <= 256 )
-  {
-     cast_m256i( d ) = mm256_get_64( s, lane, lane+8, lane+16, lane+24 );
-     return;
-  }
-  // else bit_len == 512
-  cast_m512i( d ) = mm512_get_64( s, lane   , lane+ 8, lane+16, lane+24,
-                               lane+32, lane+40, lane+48, lane+56 );
-}
-
-//
-
-static inline void mm512_intrlv_4x128( void *d, const void *s0,
-            const void *s1, const void *s2, const void *s3, const int bit_len )
-{
-  casti_m512i( d, 0 ) = mm512_put_64( s0,    s0+8,  s1,    s1+8,
-                                      s2,    s2+8,  s3,    s3+8 );
-  casti_m512i( d, 1 ) = mm512_put_64( s0+16, s0+24, s1+16, s1+24,
-                                      s2+16, s2+24, s3+16, s3+24 );
-  if ( bit_len <= 256 ) return;
-
-  casti_m512i( d, 2 ) = mm512_put_64( s0+32, s0+40, s1+32, s1+40,
-                                      s2+32, s2+40, s3+32, s3+40 );
-  casti_m512i( d, 3 ) = mm512_put_64( s0+48, s0+56, s1+48, s1+56,
-                                      s2+48, s2+56, s3+48, s3+56 );
-  if ( bit_len <= 512 ) return;
-
-  casti_m512i( d, 4 ) = mm512_put_64( s0+64, s0+72, s1+64, s1+72,
-                                      s2+64, s2+72, s3+64, s3+72 );
-  if ( bit_len <= 640 ) return;
-
-  casti_m512i( d, 5 ) = mm512_put_64( s0+ 80, s0+ 88, s1+ 80, s1+ 88,
-                                      s2+ 80, s2+ 88, s3+ 80, s3+ 88 );
-  casti_m512i( d, 6 ) = mm512_put_64( s0+ 96, s0+104, s1+ 96, s1+104,
-                                      s2+ 96, s2+104, s3+ 96, s3+104 );
-  casti_m512i( d, 7 ) = mm512_put_64( s0+112, s0+120, s1+112, s1+120,
-                                      s2+112, s2+120, s3+112, s3+120 );
-   // bit_len == 1024
-}
-
-static inline void mm512_dintrlv_4x128( void *d0, void *d1, void *d2,
-              void *d3, const void *s, const int bit_len )
-{
-   if ( bit_len <= 256 )
-   {
-      mm256_dintrlv_4x128x256( d0, d1, d2, d3, 0, s );
-      return;
-   }
-   mm512_dintrlv_4x128x512( d0, d1, d2, d3, 0, s );
-   if ( bit_len <= 512 ) return;
-   if ( bit_len <= 640 )
-   {
-      mm128_dintrlv_4x128x128( d0, d1, d2, d3, 1, s+256 );
-      return;
-   }
-   // bit_len == 1024
-   mm512_dintrlv_4x128x512( d0, d1, d2, d3, 1, s+256 );
-}
-
-// input one 8x64 buffer and return 2*4*128
-static inline void mm512_rintrlv_8x64_4x128( void *dst0, void *dst1,
-                                              const void *src, int  bit_len )
-{
-   __m512i* d0 = (__m512i*)dst0;
-   __m512i* d1 = (__m512i*)dst1;
-   uint64_t *s = (uint64_t*)src;
-
-   d0[0] = _mm512_set_epi64( s[ 11], s[  3], s[ 10], s[  2],
-                             s[  9], s[  1], s[  8], s[  0] );
-   d0[1] = _mm512_set_epi64( s[ 27], s[ 19], s[ 26], s[ 18],
-                   s[ 25], s[ 17], s[ 24], s[ 16] );
-   d0[2] = _mm512_set_epi64( s[ 15], s[  7], s[ 14], s[  6],
-                             s[ 13], s[  5], s[ 12], s[  4] );
-   d0[3] = _mm512_set_epi64( s[ 31], s[ 23], s[ 30], s[ 22],
-                             s[ 29], s[ 21], s[ 28], s[ 20] );
-   d1[0] = _mm512_set_epi64( s[ 43], s[ 35], s[ 42], s[ 34],
-                             s[ 41], s[ 33], s[ 40], s[ 32] );
-   d1[1] = _mm512_set_epi64( s[ 59], s[ 51], s[ 58], s[ 50],
-                             s[ 57], s[ 49], s[ 56], s[ 48] );
-   d1[2] = _mm512_set_epi64( s[ 47], s[ 39], s[ 46], s[ 38],
-                             s[ 45], s[ 37], s[ 44], s[ 36] );
-   d1[3] = _mm512_set_epi64( s[ 63], s[ 55], s[ 62], s[ 54],
-                              s[ 61], s[ 53], s[ 60], s[ 52] );
-
-   if ( bit_len <= 512 ) return;
-
-   d0[4] = _mm512_set_epi64( s[ 75], s[ 67], s[ 74], s[ 66],
-                             s[ 73], s[ 65], s[ 72], s[ 64] );
-   d0[5] = _mm512_set_epi64( s[ 91], s[ 83], s[ 90], s[ 82],
-                             s[ 89], s[ 81], s[ 88], s[ 80] );
-   d0[6] = _mm512_set_epi64( s[ 79], s[ 71], s[ 78], s[ 70],
-                             s[ 77], s[ 69], s[ 76], s[ 68] );
-   d0[7] = _mm512_set_epi64( s[ 95], s[ 87], s[ 94], s[ 86],
-                             s[ 93], s[ 85], s[ 92], s[ 84] );
-   d1[4] = _mm512_set_epi64( s[107], s[ 99], s[106], s[ 98],
-                             s[105], s[ 97], s[104], s[ 96] );
-   d1[5] = _mm512_set_epi64( s[123], s[115], s[122], s[114],
-                             s[121], s[113], s[120], s[112] );
-   d1[6] = _mm512_set_epi64( s[111], s[103], s[110], s[102],
-                             s[109], s[101], s[108], s[100] );
-   d1[7] = _mm512_set_epi64( s[127], s[119], s[126], s[118],
-                             s[125], s[117], s[124], s[116] );
-
-}
-
-// input 2 4x128  return 8x64
-static inline void mm512_rintrlv_4x128_8x64( void *dst, const void *src0,
-                                              const void *src1, int  bit_len )
-{
-   __m512i* d = (__m512i*)dst;
-   uint64_t *s0 = (uint64_t*)src0;
-   uint64_t *s1 = (uint64_t*)src1;
-
-   d[0] = _mm512_set_epi64( s1[ 6], s1[ 4], s1[ 2], s1[ 0],
-                            s0[ 6], s0[ 4], s0[ 2], s0[ 0] );
-   d[1] = _mm512_set_epi64( s1[ 7], s1[ 5], s1[ 3], s1[ 1],
-                            s0[ 7], s0[ 5], s0[ 3], s0[ 1] );
-   d[2] = _mm512_set_epi64( s1[14], s1[12], s1[10], s1[ 8],
-                            s0[14], s0[12], s0[10], s0[ 8] );
-   d[3] = _mm512_set_epi64( s1[15], s1[13], s1[11], s1[ 9],
-                            s0[15], s0[13], s0[11], s0[ 9] );
-   d[4] = _mm512_set_epi64( s1[22], s1[20], s1[18], s1[16],
-                            s0[22], s0[20], s0[18], s0[16] );
-   d[5] = _mm512_set_epi64( s1[23], s1[21], s1[19], s1[17],
-                            s0[24], s0[21], s0[19], s0[17] );
-   d[6] = _mm512_set_epi64( s1[22], s1[28], s1[26], s1[24],
-                            s0[22], s0[28], s0[26], s0[24] );
-   d[7] = _mm512_set_epi64( s1[31], s1[29], s1[27], s1[25],
-                            s0[31], s0[29], s0[27], s0[25] );
-
-   if ( bit_len <= 512 ) return;
-
-   d[0] = _mm512_set_epi64( s1[38], s1[36], s1[34], s1[32],
-                            s0[38], s0[36], s0[34], s0[32] );
-   d[1] = _mm512_set_epi64( s1[39], s1[37], s1[35], s1[33],
-                            s0[39], s0[37], s0[35], s0[33] );
-   d[2] = _mm512_set_epi64( s1[46], s1[44], s1[42], s1[40],
-                            s0[46], s0[44], s0[42], s0[40] );
-   d[3] = _mm512_set_epi64( s1[47], s1[45], s1[43], s1[41],
-                            s0[47], s0[45], s0[43], s0[41] );
-   d[4] = _mm512_set_epi64( s1[54], s1[52], s1[50], s1[48],
-                            s0[54], s0[52], s0[50], s0[48] );
-   d[5] = _mm512_set_epi64( s1[55], s1[53], s1[51], s1[49],
-                            s0[55], s0[53], s0[51], s0[49] );
-
-   d[6] = _mm512_set_epi64( s1[62], s1[60], s1[58], s1[56],
-                            s0[62], s0[60], s0[58], s0[56] );
-   d[7] = _mm512_set_epi64( s1[63], s1[61], s1[59], s1[57],
-                            s0[63], s0[61], s0[59], s0[57] );
-
-}
-
-static inline void mm512_extr_lane_4x128( void *d, const void *s,
-                                            const int lane, const int bit_len )
-{
-  int l = lane<<1;
-  if ( bit_len <= 256 )
-  {
-     cast_m256i( d ) = mm256_get_64( s, l, l+1, l+8, l+9 );
-     return;
-  }
-  // else bit_len == 512
-  cast_m512i( d ) = mm512_get_64( s, l   , l+ 1, l+ 8, l+ 9,
-                                l+16, l+17, l+24, l+25 );
-}
-
-#endif // AVX512
-#endif // INTRLV_AVX512_H__
diff --git a/simd-utils/intrlv-mmx.h b/simd-utils/intrlv-mmx.h
deleted file mode 100644
index ed6dae6..0000000
--- a/simd-utils/intrlv-mmx.h
+++ /dev/null
@@ -1,126 +0,0 @@
-#if !defined(INTRLV_MMX_H__)
-#define INTRLV_MMX_H__ 1
-
-#if defined(__MMX__)
-
-//////////////////////////////////////////////////////
-//
-//          MMX 64 bit vectors
-
-#define mm64_put_32( s0, s1 ) \
-  _mm_set_pi32( *((const uint32_t*)(s1)), *((const uint32_t*)(s0)) )
-
-#define mm64_get_32( s, i0, i1 ) \
-  _mm_set_pi32( ((const uint32_t*)(s))[i1], ((const uint32_t*)(s))[i0] )
-
-// 1 MMX block, 8 bytes * 2 lanes
-static inline void mm64_intrlv_2x32( void *d, const void *s0,
-                                     const void *s1, int len )
-{
-  casti_m64( d, 0 ) = mm64_put_32( s0    , s1     );
-  casti_m64( d, 1 ) = mm64_put_32( s0+  4, s1+  4 );
-  casti_m64( d, 2 ) = mm64_put_32( s0+  8, s1+  8 );
-  casti_m64( d, 3 ) = mm64_put_32( s0+ 12, s1+ 12 );
-  casti_m64( d, 4 ) = mm64_put_32( s0+ 16, s1+ 16 );
-  casti_m64( d, 5 ) = mm64_put_32( s0+ 20, s1+ 20 );
-  casti_m64( d, 6 ) = mm64_put_32( s0+ 24, s1+ 24 );
-  casti_m64( d, 7 ) = mm64_put_32( s0+ 28, s1+ 28 );
-
-  if ( len <= 256 ) return;
-
-  casti_m64( d, 8 ) = mm64_put_32( s0+ 32, s1+ 32 );
-  casti_m64( d, 9 ) = mm64_put_32( s0+ 36, s1+ 36 );
-  casti_m64( d,10 ) = mm64_put_32( s0+ 40, s1+ 40 );
-  casti_m64( d,11 ) = mm64_put_32( s0+ 44, s1+ 44 );
-  casti_m64( d,12 ) = mm64_put_32( s0+ 48, s1+ 48 );
-  casti_m64( d,13 ) = mm64_put_32( s0+ 52, s1+ 52 );
-  casti_m64( d,14 ) = mm64_put_32( s0+ 56, s1+ 56 );
-  casti_m64( d,15 ) = mm64_put_32( s0+ 60, s1+ 60 );
-
-  if ( len <= 512 ) return;
-
-  casti_m64( d,16 ) = mm64_put_32( s0+ 64, s1+ 64 );
-  casti_m64( d,17 ) = mm64_put_32( s0+ 68, s1+ 68 );
-  casti_m64( d,18 ) = mm64_put_32( s0+ 72, s1+ 72 );
-  casti_m64( d,19 ) = mm64_put_32( s0+ 76, s1+ 76 );
-
-  if ( len <= 640 ) return;
-  casti_m64( d,20 ) = mm64_put_32( s0+ 80, s1+ 80 );
-  casti_m64( d,21 ) = mm64_put_32( s0+ 84, s1+ 84 );
-  casti_m64( d,22 ) = mm64_put_32( s0+ 88, s1+ 88 );
-  casti_m64( d,23 ) = mm64_put_32( s0+ 92, s1+ 92 );
-  casti_m64( d,24 ) = mm64_put_32( s0+ 96, s1+ 96 );
-  casti_m64( d,25 ) = mm64_put_32( s0+100, s1+100 );
-  casti_m64( d,26 ) = mm64_put_32( s0+104, s1+104 );
-  casti_m64( d,27 ) = mm64_put_32( s0+108, s1+108 );
-  casti_m64( d,28 ) = mm64_put_32( s0+112, s1+112 );
-  casti_m64( d,29 ) = mm64_put_32( s0+116, s1+116 );
-  casti_m64( d,30 ) = mm64_put_32( s0+120, s1+120 );
-  casti_m64( d,31 ) = mm64_put_32( s0+124, s1+124 );
-}
-
-static inline void mm64_dintrlv_2x32( void *d00, void *d01, const int n,
-                                      const void *s, int len )
-{
-   casti_m64( d00,0 ) = mm64_get_32( s,  0,  2 );
-   casti_m64( d01,0 ) = mm64_get_32( s,  1,  3 );
-   casti_m64( d00,1 ) = mm64_get_32( s,  4,  6 );
-   casti_m64( d01,1 ) = mm64_get_32( s,  5,  7 );
-   casti_m64( d00,2 ) = mm64_get_32( s,  8, 10 );
-   casti_m64( d01,2 ) = mm64_get_32( s,  9, 11 );
-   casti_m64( d00,3 ) = mm64_get_32( s, 12, 14 );
-   casti_m64( d01,3 ) = mm64_get_32( s, 13, 15 );
-
-   if ( len <= 256 ) return;
-
-   casti_m64( d00,4 ) = mm64_get_32( s, 16, 18 );
-   casti_m64( d01,4 ) = mm64_get_32( s, 17, 19 );
-   casti_m64( d00,5 ) = mm64_get_32( s, 20, 22 );
-   casti_m64( d01,5 ) = mm64_get_32( s, 21, 23 );
-   casti_m64( d00,6 ) = mm64_get_32( s, 24, 26 );
-   casti_m64( d01,6 ) = mm64_get_32( s, 25, 27 );
-   casti_m64( d00,7 ) = mm64_get_32( s, 28, 30 );
-   casti_m64( d01,7 ) = mm64_get_32( s, 29, 31 );
-
-   if ( len <= 512 ) return;
-
-   casti_m64( d00,8 ) = mm64_get_32( s, 32, 34 );
-   casti_m64( d01,8 ) = mm64_get_32( s, 33, 35 );
-   casti_m64( d00,9 ) = mm64_get_32( s, 36, 38 );
-   casti_m64( d01,9 ) = mm64_get_32( s, 37, 39 );
-
-   if ( len <= 640 ) return;
-   casti_m64( d00,10 ) = mm64_get_32( s, 40, 42 );
-   casti_m64( d01,10 ) = mm64_get_32( s, 41, 43 );
-   casti_m64( d00,11 ) = mm64_get_32( s, 44, 46 );
-   casti_m64( d01,11 ) = mm64_get_32( s, 45, 47 );
-   casti_m64( d00,12 ) = mm64_get_32( s, 48, 50 );
-   casti_m64( d01,12 ) = mm64_get_32( s, 49, 51 );
-   casti_m64( d00,13 ) = mm64_get_32( s, 52, 54 );
-   casti_m64( d01,13 ) = mm64_get_32( s, 53, 55 );
-   casti_m64( d00,14 ) = mm64_get_32( s, 56, 58 );
-   casti_m64( d01,14 ) = mm64_get_32( s, 57, 59 );
-   casti_m64( d00,15 ) = mm64_get_32( s, 60, 62 );
-   casti_m64( d01,15 ) = mm64_get_32( s, 61, 63 );
-}
-
-static inline void mm64_extr_lane_2x32( void *d, const void *s,
-                                         const int lane, const int bit_len )
-{
-  casti_m64( d, 0 ) = mm64_get_32( s, lane   , lane+ 4 );
-  casti_m64( d, 1 ) = mm64_get_32( s, lane+ 8, lane+12 );
-  casti_m64( d, 2 ) = mm64_get_32( s, lane+16, lane+20 );
-  casti_m64( d, 3 ) = mm64_get_32( s, lane+24, lane+28 );
-
-  if ( bit_len <= 256 ) return;
-  casti_m64( d, 4 ) = mm64_get_32( s, lane+32, lane+36 );
-  casti_m64( d, 5 ) = mm64_get_32( s, lane+40, lane+44 );
-  casti_m64( d, 6 ) = mm64_get_32( s, lane+48, lane+52 );
-  casti_m64( d, 7 ) = mm64_get_32( s, lane+56, lane+60 );
-  // bit_len == 512
-}
-
-
-
-#endif // MMX
-#endif // INTRLV_MMX_H__
diff --git a/simd-utils/intrlv-selector.h b/simd-utils/intrlv-selector.h
deleted file mode 100644
index 3095b9e..0000000
--- a/simd-utils/intrlv-selector.h
+++ /dev/null
@@ -1,77 +0,0 @@
-#if !defined(INTRLV_SELECTOR_H__)
-#define INTRLV_SELECTOR_H__
-
-//////////////////////////////////////////////////////////////
-//
-//  Generic interface for interleaving data for parallel processing.
-//
-//  Best tech is chosen atomatically.
-
-/*
-#if defined(__AVX512F__)
-
-#define intrlv_4x128      mm512_intrlv_4x128
-#define intrlv_4x128      mm512_intrlv_4x128
-
-#define intrlv_8x64       mm512_intrlv_8x64
-#define dintrlv_8x64      mm512_dintrlv_8x64
-#define extr_lane_8x64    mm512_extr_lane_8x64
-
-#define intrlv_16x32      mm512_intrlv_16x32
-#define dintrlv_16x32     mm512_dintrlv_16x32
-#define extr_lane_16x32    mm512_extr_lane_16x32
-
-#define intrlv_2x128      mm512_intrlv_2x128
-#define dintrlv_2x128     mm512_dintrlv_2x128
-
-#define intrlv_4x64       mm512_intrlv_4x64
-#define dintrlv_4x64      mm512_dintrlv_4x64
-#define extr_lane_4x64    mm512_extr_lane_4x64
-
-#define intrlv_8x32       mm512_intrlv_8x32
-#define dintrlv_8x32      mm512_dintrlv_8x32
-#define extr_lane_8x32    mm512_extr_lane_8x32
-
-#elif defined(__AVX__)
-*/
-#if defined(__AVX__)
-
-#define intrlv_2x128      mm256_intrlv_2x128
-#define dintrlv_2x128     mm256_dintrlv_2x128
-
-#define intrlv_4x64       mm256_intrlv_4x64
-#define dintrlv_4x64      mm256_dintrlv_4x64
-#define extr_lane_4x64    mm256_extr_lane_4x64
-
-#define intrlv_8x32       mm256_intrlv_8x32
-#define dintrlv_8x32      mm256_dintrlv_8x32
-#define extr_lane_8x32    mm256_extr_lane_8x32
-
-#define intrlv_4x32       mm256_intrlv_4x32
-#define dintrlv_4x32      mm256_dintrlv_4x32
-#define extr_lane_4x32    mm256_extr_lane_4x32
-
-#else
-
-#define intrlv_2x128      mm128_intrlv_2x128
-#define dintrlv_2x128     mm128_dintrlv_2x128
-
-#define intrlv_4x64       mm128_intrlv_4x64
-#define dintrlv_4x64      mm128_dintrlv_4x64
-#define extr_lane_4x64    mm128_extr_lane_4x64
-
-#define intrlv_8x32       mm128_intrlv_8x32
-#define dintrlv_8x32      mm128_dintrlv_8x32
-#define extr_lane_8x32    mm128_extr_lane_8x32
-
-#define intrlv_2x64       mm128_intrlv_2x64
-#define dintrlv_2x64      mm128_dintrlv_2x64
-#define extr_lane_2x64    mm128_extr_lane_2x64
-
-#define intrlv_4x32       mm128_intrlv_4x32
-#define dintrlv_4x32      mm128_dintrlv_4x32
-#define extr_lane_4x32    mm128_extr_lane_4x32
-
-#endif
-
-#endif  // INTRLV_SELECTOR_H__
diff --git a/simd-utils/intrlv-sse2.h b/simd-utils/intrlv-sse2.h
deleted file mode 100644
index aeb7cae..0000000
--- a/simd-utils/intrlv-sse2.h
+++ /dev/null
@@ -1,192 +0,0 @@
-#if !defined(INTRLV_SSE2_H__)
-#define INTRLV_SSE2_H__ 1
-
-// Don't call __mm_extract_epi32 directly, it needs SSE4.1.
-// Use mm128_extr_32 wrapper instead, it has both SSE4.1 & SSE2 covered.
-
-#if  defined(__SSE2__)
-
-///////////////////////////////////////////////////////////////
-//
-//           SSE2 128 bit vectors
-
-
-// Macros to abstract typecasting
-
-// Interleave lanes 
-#define mm128_put_64( s0, s1) \
-  _mm_set_epi64x( *((const uint64_t*)(s1)), *((const uint64_t*)(s0)) )
-
-#define mm128_put_32( s0, s1, s2, s3 ) \
-  _mm_set_epi32( *((const uint32_t*)(s3)), *((const uint32_t*)(s2)), \
-                 *((const uint32_t*)(s1)), *((const uint32_t*)(s0)) )
-
-// Deinterleave lanes
-#define mm128_get_64( s, i0, i1 ) \
-  _mm_set_epi64x( ((const uint64_t*)(s))[i1], ((const uint64_t*)(s))[i0] )
-
-#define mm128_get_32( s, i0, i1, i2, i3 ) \
-  _mm_set_epi32( ((const uint32_t*)(s))[i3], ((const uint32_t*)(s))[i2], \
-                 ((const uint32_t*)(s))[i1], ((const uint32_t*)(s))[i0] )
-
-// blend 2 vectors while interleaving: { hi[n], lo[n-1], ... hi[1], lo[0] }
-#define mm128_intrlv_blend_64( hi, lo ) \
-                _mm256_blend_epi16( hi, lo, 0x0f )
-#define mm128_intrlv_blend_32( hi, lo ) \
-                _mm6_blend_epi16( hi, lo, 0x33 )
-
-// 1 sse2 block, 16 x 16 bytes
-
-#define mm128_intrlv_4x32_128( d, s0, s1, s2, s3 )\
-do { \
-   casti_m128i( d,0 ) = _mm_set_epi32( \
-                            mm128_extr_32( s3, 0 ), mm128_extr_32( s2, 0 ), \
-                            mm128_extr_32( s1, 0 ), mm128_extr_32( s0, 0 ) ); \
-   casti_m128i( d,1 ) = _mm_set_epi32( \
-                            mm128_extr_32( s3, 1 ), mm128_extr_32( s2, 1 ), \
-                            mm128_extr_32( s1, 1 ), mm128_extr_32( s0, 1 ) ); \
-   casti_m128i( d,2 ) = _mm_set_epi32( \
-                            mm128_extr_32( s3, 2 ), mm128_extr_32( s2, 2 ), \
-                            mm128_extr_32( s1, 2 ), mm128_extr_32( s0, 2 ) ); \
-   casti_m128i( d,3 ) = _mm_set_epi32( \
-                            mm128_extr_32( s3, 3 ), mm128_extr_32( s2, 3 ), \
-                            mm128_extr_32( s1, 3 ), mm128_extr_32( s0, 3 ) ); \
-} while(0)
-
-static inline void mm128_dintrlv_4x32_128( void *d0, void *d1, void *d2,
-                                           void *d3, const void *src )
-{
-   __m128i s0 = *(__m128i*) src;
-   __m128i s1 = *(__m128i*)(src+16);
-   __m128i s2 = *(__m128i*)(src+32);
-   __m128i s3 = *(__m128i*)(src+48);
-
-   *(__m128i*)d0 = _mm_set_epi32(
-                     mm128_extr_32( s3,0 ), mm128_extr_32( s2,0 ),
-                     mm128_extr_32( s1,0 ), mm128_extr_32( s0,0 ) );
-   *(__m128i*)d1 = _mm_set_epi32(
-                     mm128_extr_32( s3,1 ), mm128_extr_32( s2,1 ),
-                     mm128_extr_32( s1,1 ), mm128_extr_32( s0,1 ) );
-   *(__m128i*)d2 = _mm_set_epi32(
-                     mm128_extr_32( s3,2 ), mm128_extr_32( s2,2 ),
-                     mm128_extr_32( s1,2 ), mm128_extr_32( s0,2 ) );
-   *(__m128i*)d3 = _mm_set_epi32(
-                     mm128_extr_32( s3,3 ), mm128_extr_32( s2,3 ),
-                     mm128_extr_32( s1,3 ), mm128_extr_32( s0,3 ) );
-}
-
-static inline void mm128_intrlv_2x64x128( void *d, const void *s0,
-                       const void *s1 )
-{
-  casti_m128i( d,0 ) = mm128_put_64( s0,    s1    );
-  casti_m128i( d,1 ) = mm128_put_64( s0+ 8, s1+ 8 );
-  casti_m128i( d,2 ) = mm128_put_64( s0+16, s1+16 );
-  casti_m128i( d,3 ) = mm128_put_64( s0+24, s1+24 );
-}
-
-#define mm128_bswap_intrlv_4x32_128( d, src ) \
-do { \
-  __m128i ss = mm128_bswap_32( src );\
-  casti_m128i( d,0 ) = _mm_set1_epi32( mm128_extr_32( ss, 0 ) ); \
-  casti_m128i( d,1 ) = _mm_set1_epi32( mm128_extr_32( ss, 1 ) ); \
-  casti_m128i( d,2 ) = _mm_set1_epi32( mm128_extr_32( ss, 2 ) ); \
-  casti_m128i( d,3 ) = _mm_set1_epi32( mm128_extr_32( ss, 3 ) ); \
-} while(0)
-
-
-//
-// User functions.
-
-// interleave 4 arrays of 32 bit elements for 128 bit processing
-// bit_len must be 256, 512 or 640 bits.
-static inline void mm128_intrlv_4x32( void *d, const void *s0,
-               const void *s1, const void *s2, const void *s3, int bit_len )
-{
-   mm128_intrlv_4x32_128( d    , casti_m128i(s0,0), casti_m128i(s1,0),
-                                 casti_m128i(s2,0), casti_m128i(s3,0) );
-   mm128_intrlv_4x32_128( d+ 64, casti_m128i(s0,1), casti_m128i(s1,1),   
-                                 casti_m128i(s2,1), casti_m128i(s3,1) );   
-   if ( bit_len <= 256 ) return;
-   mm128_intrlv_4x32_128( d+128, casti_m128i(s0,2), casti_m128i(s1,2),
-                                 casti_m128i(s2,2), casti_m128i(s3,2) );
-   mm128_intrlv_4x32_128( d+192, casti_m128i(s0,3), casti_m128i(s1,3),
-                                 casti_m128i(s2,3), casti_m128i(s3,3) );
-   if ( bit_len <= 512 ) return;
-   mm128_intrlv_4x32_128( d+256, casti_m128i(s0,4), casti_m128i(s1,4),
-                                 casti_m128i(s2,4), casti_m128i(s3,4) );
-   if ( bit_len <= 640 ) return;
-   mm128_intrlv_4x32_128( d+320, casti_m128i(s0,5), casti_m128i(s1,5),
-                                 casti_m128i(s2,5), casti_m128i(s3,5) );
-   mm128_intrlv_4x32_128( d+384, casti_m128i(s0,6), casti_m128i(s1,6),
-                                 casti_m128i(s2,6), casti_m128i(s3,6) );
-   mm128_intrlv_4x32_128( d+448, casti_m128i(s0,7), casti_m128i(s1,7),
-                                 casti_m128i(s2,7), casti_m128i(s3,7) );
-   // bit_len == 1024
-}
-
-// Still used by decred due to odd data size: 180 bytes
-// bit_len must be multiple of 32
-static inline void mm128_intrlv_4x32x( void *dst, void *src0, void  *src1,
-                                        void *src2, void *src3, int bit_len )
-{
-   uint32_t *d  = (uint32_t*)dst;
-   uint32_t *s0 = (uint32_t*)src0;
-   uint32_t *s1 = (uint32_t*)src1;
-   uint32_t *s2 = (uint32_t*)src2;
-   uint32_t *s3 = (uint32_t*)src3;
-
-   for ( int i = 0; i < bit_len >> 5; i++, d += 4 )
-   {
-      *d     = *(s0+i);
-      *(d+1) = *(s1+i);
-      *(d+2) = *(s2+i);
-      *(d+3) = *(s3+i);
-   }
-}
-
-static inline void mm128_dintrlv_4x32( void *d0, void *d1, void *d2,
-                                       void *d3, const void *s, int bit_len )
-{
-   mm128_dintrlv_4x32_128( d0    , d1    , d2    , d3    , s     );
-   mm128_dintrlv_4x32_128( d0+ 16, d1+ 16, d2+ 16, d3+ 16, s+ 64 );
-   if ( bit_len <= 256 ) return;
-   mm128_dintrlv_4x32_128( d0+ 32, d1+ 32, d2+ 32, d3+ 32, s+128 );
-   mm128_dintrlv_4x32_128( d0+ 48, d1+ 48, d2+ 48, d3+ 48, s+192 );
-   if ( bit_len <= 512 ) return;
-   mm128_dintrlv_4x32_128( d0+ 64, d1+ 64, d2+ 64, d3+ 64, s+256 );
-   if ( bit_len <= 640 ) return;
-   mm128_dintrlv_4x32_128( d0+ 80, d1+ 80, d2+ 80, d3+ 80, s+320 );
-   mm128_dintrlv_4x32_128( d0+ 96, d1+ 96, d2+ 96, d3+ 96, s+384 );
-   mm128_dintrlv_4x32_128( d0+112, d1+112, d2+112, d3+112, s+448 );
-   // bit_len == 1024
-}
-
-// extract and deinterleave specified lane.
-static inline void mm128_extr_lane_4x32( void *d, const void *s,
-                                         const int lane, const int bit_len )
-{
-  casti_m128i( d, 0 ) =
-             mm128_get_32( s, lane   , lane+ 4, lane+ 8, lane+12 );
-  casti_m128i( d, 1 ) =
-             mm128_get_32( s, lane+16, lane+20, lane+24, lane+28 );
-  if ( bit_len <= 256 ) return;
-  casti_m128i( d, 2 ) =
-             mm128_get_32( s, lane+32, lane+36, lane+40, lane+44 );
-  casti_m128i( d, 3 ) =
-             mm128_get_32( s, lane+48, lane+52, lane+56, lane+60 );
-  // bit_len == 512
-}
-
-// Interleave 80 bytes of 32 bit data for 4 lanes.
-static inline void mm128_bswap_intrlv80_4x32( void *d, const void *s )
-{
-   mm128_bswap_intrlv_4x32_128( d    , casti_m128i( s, 0 ) );
-   mm128_bswap_intrlv_4x32_128( d+ 64, casti_m128i( s, 1 ) );
-   mm128_bswap_intrlv_4x32_128( d+128, casti_m128i( s, 2 ) );
-   mm128_bswap_intrlv_4x32_128( d+192, casti_m128i( s, 3 ) );
-   mm128_bswap_intrlv_4x32_128( d+256, casti_m128i( s, 4 ) );
-}
-
-#endif // SSE2
-#endif // INTRLV_SSE2_H__
-
diff --git a/simd-utils/intrlv.h b/simd-utils/intrlv.h
new file mode 100644
index 0000000..9ab7fa7
--- /dev/null
+++ b/simd-utils/intrlv.h
@@ -0,0 +1,1333 @@
+#if !defined(INTERLEAVE_H__)
+#define INTERLEAVE_H__ 1
+
+// philosophical discussion
+// 
+// transitions:
+//
+//  int32 <-> int64
+//  uint64_t = (uint64_t)int32_lo | ( (uint64_t)int32_hi << 32 )
+//  Efficient transition and post processing, 32 bit granularity is lost.
+//  Not pratical.
+//  
+//  int32 <-> m64
+//  More complex, 32 bit granularity maintained, limited number of mmx regs.
+//  int32 <-> int64 <-> m64 might be more efficient.
+//
+//  int32 <-> m128
+//  Expensive, current implementation.
+//
+//  int32 <-> m256
+//  Very expensive multi stage, current implementation.
+//
+//  int64/m64 <-> m128
+//  Efficient, agnostic to native element size. Common.
+//
+//  m128 <-> m256
+//  Expensive for a single instruction, unavoidable. Common.
+// 
+//  Multi stage options
+//
+//  int32 <-> int64 -> m128
+//  More efficient than insert32, granularity maintained. Common.
+//
+//  int64 <-> m128 -> m256
+//  Unavoidable, reasonably efficient. Common
+//
+//  int32 <-> int64 -> m128 -> m256
+//  Seems inevitable, most efficient despite number of stages. Common.
+//
+//  It seems the best approach is to avoid transitions and use the native type
+//  of the data: 64 & 32 bit use integer, 128 bit use m128i.
+//
+
+////////////////////////////////
+//
+//          32 bit data
+
+// 2x32
+
+static inline void intrlv_2x32( void *dst, const void *src0,
+                                const void *src1, int bit_len )
+{
+   uint32_t *d = (uint32_t*)dst;;
+   const uint32_t *s0 = (const uint32_t*)src0;
+   const uint32_t *s1 = (const uint32_t*)src1;
+   d[ 0] = s0[ 0];    d[ 1] = s1[ 0];   d[ 2] = s0[ 1];    d[ 3] = s1[ 1];
+   d[ 4] = s0[ 2];    d[ 5] = s1[ 2];   d[ 6] = s0[ 3];    d[ 7] = s1[ 3];
+   d[ 8] = s0[ 4];    d[ 9] = s1[ 4];   d[10] = s0[ 5];    d[11] = s1[ 5];
+   d[12] = s0[ 6];    d[13] = s1[ 6];   d[14] = s0[ 7];    d[15] = s1[ 7];
+   if ( bit_len <= 256 ) return;
+   d[16] = s0[ 8];    d[17] = s1[ 8];   d[18] = s0[ 9];    d[19] = s1[ 9];
+   d[20] = s0[10];    d[21] = s1[10];   d[22] = s0[11];    d[23] = s1[11];
+   d[24] = s0[12];    d[25] = s1[12];   d[26] = s0[13];    d[27] = s1[13];
+   d[28] = s0[14];    d[29] = s1[14];   d[30] = s0[15];    d[31] = s1[15];
+   if ( bit_len <= 512 ) return;
+   d[32] = s0[16];    d[33] = s1[16];   d[34] = s0[17];    d[35] = s1[17];
+   d[36] = s0[18];    d[37] = s1[18];   d[38] = s0[19];    d[39] = s1[19];
+   if ( bit_len <= 640 ) return;
+   d[40] = s0[20];    d[41] = s1[20];   d[42] = s0[21];    d[43] = s1[21];
+   d[44] = s0[22];    d[45] = s1[22];   d[46] = s0[23];    d[47] = s1[23];
+   d[48] = s0[24];    d[49] = s1[24];   d[50] = s0[25];    d[51] = s1[25];
+   d[52] = s0[26];    d[53] = s1[26];   d[54] = s0[27];    d[55] = s1[27];
+   d[56] = s0[28];    d[57] = s1[28];   d[58] = s0[29];    d[59] = s1[29];
+   d[60] = s0[30];    d[61] = s1[30];   d[62] = s0[31];    d[63] = s1[31];
+}
+
+static inline void dintrlv_2x32( void *dst0, void *dst1,
+                                 const void *src, int bit_len )
+{
+   uint32_t *d0 = (uint32_t*)dst0;
+   uint32_t *d1 = (uint32_t*)dst1;
+   const uint32_t *s = (const uint32_t*)src;
+
+   d0[ 0] = s[ 0];   d1[ 0] = s[ 1];   d0[ 1] = s[ 2];   d1[ 1] = s[ 3];
+   d0[ 2] = s[ 4];   d1[ 2] = s[ 5];   d0[ 3] = s[ 6];   d1[ 3] = s[ 7];
+   d0[ 4] = s[ 8];   d1[ 4] = s[ 9];   d0[ 5] = s[10];   d1[ 5] = s[11];
+   d0[ 6] = s[12];   d1[ 6] = s[13];   d0[ 7] = s[14];   d1[ 7] = s[15];
+   if ( bit_len <= 256 ) return;
+   d0[ 8] = s[16];   d1[ 8] = s[17];   d0[ 9] = s[18];   d1[ 9] = s[19];
+   d0[10] = s[20];   d1[10] = s[21];   d0[11] = s[22];   d1[11] = s[23];
+   d0[12] = s[24];   d1[12] = s[25];   d0[13] = s[26];   d1[13] = s[27];
+   d0[14] = s[28];   d1[14] = s[29];   d0[15] = s[30];   d1[15] = s[31];
+   if ( bit_len <= 512 ) return;
+   d0[16] = s[32];   d1[16] = s[33];   d0[17] = s[34];   d1[17] = s[35];
+   d0[18] = s[36];   d1[18] = s[37];   d0[19] = s[38];   d1[19] = s[39];
+   if ( bit_len <= 640 ) return;
+   d0[20] = s[40];   d1[20] = s[41];   d0[21] = s[42];   d1[21] = s[43];
+   d0[22] = s[44];   d1[22] = s[45];   d0[23] = s[46];   d1[23] = s[47];
+   d0[24] = s[48];   d1[24] = s[49];   d0[25] = s[50];   d1[25] = s[51];
+   d0[26] = s[52];   d1[26] = s[53];   d0[27] = s[54];   d1[27] = s[55];
+   d0[28] = s[56];   d1[28] = s[57];   d0[29] = s[58];   d1[29] = s[59];
+   d0[30] = s[60];   d1[30] = s[61];   d0[31] = s[61];   d1[31] = s[63];
+}
+
+static inline void extr_lane_2x32( void *dst, const void *src,
+                                   const int lane, const int bit_len )
+{
+   uint32_t *d = (uint32_t*)dst;
+   const uint32_t *s = (const uint32_t*)src;
+   d[ 0] = s[ lane    ];   d[ 1] = s[ lane+ 2 ];
+   d[ 2] = s[ lane+ 4 ];   d[ 3] = s[ lane+ 6 ];
+   d[ 4] = s[ lane+ 8 ];   d[ 5] = s[ lane+10 ];
+   d[ 6] = s[ lane+12 ];   d[ 7] = s[ lane+14 ];
+   if ( bit_len <= 256 ) return;
+   d[ 8] = s[ lane+16 ];   d[ 9] = s[ lane+18 ];
+   d[10] = s[ lane+20 ];   d[11] = s[ lane+22 ];
+   d[12] = s[ lane+24 ];   d[13] = s[ lane+26 ];
+   d[14] = s[ lane+28 ];   d[15] = s[ lane+30 ];
+}
+
+// 4x32
+
+static inline void intrlv_4x32( void *dst, const void *src0, const void *src1,
+                             const void *src2, const void *src3, int bit_len )
+{
+   uint32_t *d = (uint32_t*)dst;
+   const uint32_t *s0 = (const uint32_t*)src0;
+   const uint32_t *s1 = (const uint32_t*)src1;
+   const uint32_t *s2 = (const uint32_t*)src2;
+   const uint32_t *s3 = (const uint32_t*)src3;
+   d[  0] = s0[ 0];   d[  1] = s1[ 0];   d[  2] = s2[ 0];   d[  3] = s3[ 0];
+   d[  4] = s0[ 1];   d[  5] = s1[ 1];   d[  6] = s2[ 1];   d[  7] = s3[ 1];
+   d[  8] = s0[ 2];   d[  9] = s1[ 2];   d[ 10] = s2[ 2];   d[ 11] = s3[ 2];
+   d[ 12] = s0[ 3];   d[ 13] = s1[ 3];   d[ 14] = s2[ 3];   d[ 15] = s3[ 3];
+   d[ 16] = s0[ 4];   d[ 17] = s1[ 4];   d[ 18] = s2[ 4];   d[ 19] = s3[ 4];
+   d[ 20] = s0[ 5];   d[ 21] = s1[ 5];   d[ 22] = s2[ 5];   d[ 23] = s3[ 5];
+   d[ 24] = s0[ 6];   d[ 25] = s1[ 6];   d[ 26] = s2[ 6];   d[ 27] = s3[ 6];
+   d[ 28] = s0[ 7];   d[ 29] = s1[ 7];   d[ 30] = s2[ 7];   d[ 31] = s3[ 7];
+   if ( bit_len <= 256 ) return;
+   d[ 32] = s0[ 8];   d[ 33] = s1[ 8];   d[ 34] = s2[ 8];   d[ 35] = s3[ 8];
+   d[ 36] = s0[ 9];   d[ 37] = s1[ 9];   d[ 38] = s2[ 9];   d[ 39] = s3[ 9];
+   d[ 40] = s0[10];   d[ 41] = s1[10];   d[ 42] = s2[10];   d[ 43] = s3[10];
+   d[ 44] = s0[11];   d[ 45] = s1[11];   d[ 46] = s2[11];   d[ 47] = s3[11];
+   d[ 48] = s0[12];   d[ 49] = s1[12];   d[ 50] = s2[12];   d[ 51] = s3[12];
+   d[ 52] = s0[13];   d[ 53] = s1[13];   d[ 54] = s2[13];   d[ 55] = s3[13];
+   d[ 56] = s0[14];   d[ 57] = s1[14];   d[ 58] = s2[14];   d[ 59] = s3[14];
+   d[ 60] = s0[15];   d[ 61] = s1[15];   d[ 62] = s2[15];   d[ 63] = s3[15];
+   if ( bit_len <= 512 ) return;
+   d[ 64] = s0[16];   d[ 65] = s1[16];   d[ 66] = s2[16];   d[ 67] = s3[16];
+   d[ 68] = s0[17];   d[ 69] = s1[17];   d[ 70] = s2[17];   d[ 71] = s3[17];
+   d[ 72] = s0[18];   d[ 73] = s1[18];   d[ 74] = s2[18];   d[ 75] = s3[18];
+   d[ 76] = s0[19];   d[ 77] = s1[19];   d[ 78] = s2[19];   d[ 79] = s3[19];
+   if ( bit_len <= 640 ) return;
+   d[ 80] = s0[20];   d[ 81] = s1[20];   d[ 82] = s2[20];   d[ 83] = s3[20];
+   d[ 84] = s0[21];   d[ 85] = s1[21];   d[ 86] = s2[21];   d[ 87] = s3[21];
+   d[ 88] = s0[22];   d[ 89] = s1[22];   d[ 90] = s2[22];   d[ 91] = s3[22];
+   d[ 92] = s0[23];   d[ 93] = s1[23];   d[ 94] = s2[23];   d[ 95] = s3[23];
+   d[ 96] = s0[24];   d[ 97] = s1[24];   d[ 98] = s2[24];   d[ 99] = s3[24];
+   d[100] = s0[25];   d[101] = s1[25];   d[102] = s2[25];   d[103] = s3[25];
+   d[104] = s0[26];   d[105] = s1[26];   d[106] = s2[26];   d[107] = s3[26];
+   d[108] = s0[27];   d[109] = s1[27];   d[110] = s2[27];   d[111] = s3[27];
+   d[112] = s0[28];   d[113] = s1[28];   d[114] = s2[28];   d[115] = s3[28];
+   d[116] = s0[29];   d[117] = s1[29];   d[118] = s2[29];   d[119] = s3[29];
+   d[120] = s0[30];   d[121] = s1[30];   d[122] = s2[30];   d[123] = s3[30];
+   d[124] = s0[31];   d[125] = s1[31];   d[126] = s2[31];   d[127] = s3[31];
+}
+
+static inline void intrlv_4x32_512( void *dst, const void *src0,
+                     const void *src1, const void *src2, const void *src3 )
+{
+   uint32_t *d = (uint32_t*)dst;
+   const uint32_t *s0 = (const uint32_t*)src0;
+   const uint32_t *s1 = (const uint32_t*)src1;
+   const uint32_t *s2 = (const uint32_t*)src2;
+   const uint32_t *s3 = (const uint32_t*)src3;
+   d[  0] = s0[ 0];   d[  1] = s1[ 0];   d[  2] = s2[ 0];   d[  3] = s3[ 0];
+   d[  4] = s0[ 1];   d[  5] = s1[ 1];   d[  6] = s2[ 1];   d[  7] = s3[ 1];
+   d[  8] = s0[ 2];   d[  9] = s1[ 2];   d[ 10] = s2[ 2];   d[ 11] = s3[ 2];
+   d[ 12] = s0[ 3];   d[ 13] = s1[ 3];   d[ 14] = s2[ 3];   d[ 15] = s3[ 3];
+   d[ 16] = s0[ 4];   d[ 17] = s1[ 4];   d[ 18] = s2[ 4];   d[ 19] = s3[ 4];
+   d[ 20] = s0[ 5];   d[ 21] = s1[ 5];   d[ 22] = s2[ 5];   d[ 23] = s3[ 5];
+   d[ 24] = s0[ 6];   d[ 25] = s1[ 6];   d[ 26] = s2[ 6];   d[ 27] = s3[ 6];
+   d[ 28] = s0[ 7];   d[ 29] = s1[ 7];   d[ 30] = s2[ 7];   d[ 31] = s3[ 7];
+   d[ 32] = s0[ 8];   d[ 33] = s1[ 8];   d[ 34] = s2[ 8];   d[ 35] = s3[ 8];
+   d[ 36] = s0[ 9];   d[ 37] = s1[ 9];   d[ 38] = s2[ 9];   d[ 39] = s3[ 9];
+   d[ 40] = s0[10];   d[ 41] = s1[10];   d[ 42] = s2[10];   d[ 43] = s3[10];
+   d[ 44] = s0[11];   d[ 45] = s1[11];   d[ 46] = s2[11];   d[ 47] = s3[11];
+   d[ 48] = s0[12];   d[ 49] = s1[12];   d[ 50] = s2[12];   d[ 51] = s3[12];
+   d[ 52] = s0[13];   d[ 53] = s1[13];   d[ 54] = s2[13];   d[ 55] = s3[13];
+   d[ 56] = s0[14];   d[ 57] = s1[14];   d[ 58] = s2[14];   d[ 59] = s3[14];
+   d[ 60] = s0[15];   d[ 61] = s1[15];   d[ 62] = s2[15];   d[ 63] = s3[15];
+}
+
+static inline void dintrlv_4x32( void *dst0, void *dst1, void *dst2,
+                                 void *dst3, const void *src, int bit_len )
+{
+   uint32_t *d0 = (uint32_t*)dst0;
+   uint32_t *d1 = (uint32_t*)dst1;
+   uint32_t *d2 = (uint32_t*)dst2;
+   uint32_t *d3 = (uint32_t*)dst3;
+   const uint32_t *s = (const uint32_t*)src;
+   d0[ 0] = s[  0];   d1[ 0] = s[  1];    d2[ 0] = s[  2];   d3[ 0] = s[  3];
+   d0[ 1] = s[  4];   d1[ 1] = s[  5];    d2[ 1] = s[  6];   d3[ 1] = s[  7];
+   d0[ 2] = s[  8];   d1[ 2] = s[  9];    d2[ 2] = s[ 10];   d3[ 2] = s[ 11];
+   d0[ 3] = s[ 12];   d1[ 3] = s[ 13];    d2[ 3] = s[ 14];   d3[ 3] = s[ 15];
+   d0[ 4] = s[ 16];   d1[ 4] = s[ 17];    d2[ 4] = s[ 18];   d3[ 4] = s[ 19];
+   d0[ 5] = s[ 20];   d1[ 5] = s[ 21];    d2[ 5] = s[ 22];   d3[ 5] = s[ 23];
+   d0[ 6] = s[ 24];   d1[ 6] = s[ 25];    d2[ 6] = s[ 26];   d3[ 6] = s[ 27];
+   d0[ 7] = s[ 28];   d1[ 7] = s[ 29];    d2[ 7] = s[ 30];   d3[ 7] = s[ 31];
+   if ( bit_len <= 256 ) return;
+   d0[ 8] = s[ 32];   d1[ 8] = s[ 33];    d2[ 8] = s[ 34];   d3[ 8] = s[ 35];
+   d0[ 9] = s[ 36];   d1[ 9] = s[ 37];    d2[ 9] = s[ 38];   d3[ 9] = s[ 39];
+   d0[10] = s[ 40];   d1[10] = s[ 41];    d2[10] = s[ 42];   d3[10] = s[ 43];
+   d0[11] = s[ 44];   d1[11] = s[ 45];    d2[11] = s[ 46];   d3[11] = s[ 47];
+   d0[12] = s[ 48];   d1[12] = s[ 49];    d2[12] = s[ 50];   d3[12] = s[ 51];
+   d0[13] = s[ 52];   d1[13] = s[ 53];    d2[13] = s[ 54];   d3[13] = s[ 55];
+   d0[14] = s[ 56];   d1[14] = s[ 57];    d2[14] = s[ 58];   d3[14] = s[ 59];
+   d0[15] = s[ 60];   d1[15] = s[ 61];    d2[15] = s[ 62];   d3[15] = s[ 63];
+   if ( bit_len <= 512 ) return;
+   d0[16] = s[ 64];   d1[16] = s[ 65];    d2[16] = s[ 66];   d3[16] = s[ 67];
+   d0[17] = s[ 68];   d1[17] = s[ 69];    d2[17] = s[ 70];   d3[17] = s[ 71];
+   d0[18] = s[ 72];   d1[18] = s[ 73];    d2[18] = s[ 74];   d3[18] = s[ 75];
+   d0[19] = s[ 76];   d1[19] = s[ 77];    d2[19] = s[ 78];   d3[19] = s[ 79];
+   if ( bit_len <= 640 ) return;
+   d0[20] = s[ 80];   d1[20] = s[ 81];    d2[20] = s[ 82];   d3[20] = s[ 83];
+   d0[21] = s[ 84];   d1[21] = s[ 85];    d2[21] = s[ 86];   d3[21] = s[ 87];
+   d0[22] = s[ 88];   d1[22] = s[ 89];    d2[22] = s[ 90];   d3[22] = s[ 91];
+   d0[23] = s[ 92];   d1[23] = s[ 93];    d2[23] = s[ 94];   d3[23] = s[ 95];
+   d0[24] = s[ 96];   d1[24] = s[ 97];    d2[24] = s[ 98];   d3[24] = s[ 99];
+   d0[25] = s[100];   d1[25] = s[101];    d2[25] = s[102];   d3[25] = s[103];
+   d0[26] = s[104];   d1[26] = s[105];    d2[26] = s[106];   d3[26] = s[107];
+   d0[27] = s[108];   d1[27] = s[109];    d2[27] = s[110];   d3[27] = s[111];
+   d0[28] = s[112];   d1[28] = s[113];    d2[28] = s[114];   d3[28] = s[115];
+   d0[29] = s[116];   d1[29] = s[117];    d2[29] = s[118];   d3[29] = s[119];
+   d0[30] = s[120];   d1[30] = s[121];    d2[30] = s[122];   d3[30] = s[123];
+   d0[31] = s[124];   d1[31] = s[125];    d2[31] = s[126];   d3[31] = s[127];
+}
+
+static inline void dintrlv_4x32_512( void *dst0, void *dst1, void *dst2,
+                                     void *dst3, const void *src )
+{
+   uint32_t *d0 = (uint32_t*)dst0;
+   uint32_t *d1 = (uint32_t*)dst1;
+   uint32_t *d2 = (uint32_t*)dst2;
+   uint32_t *d3 = (uint32_t*)dst3;
+   const uint32_t *s = (const uint32_t*)src;
+   d0[ 0] = s[  0];   d1[ 0] = s[  1];    d2[ 0] = s[  2];   d3[ 0] = s[  3];
+   d0[ 1] = s[  4];   d1[ 1] = s[  5];    d2[ 1] = s[  6];   d3[ 1] = s[  7];
+   d0[ 2] = s[  8];   d1[ 2] = s[  9];    d2[ 2] = s[ 10];   d3[ 2] = s[ 11];
+   d0[ 3] = s[ 12];   d1[ 3] = s[ 13];    d2[ 3] = s[ 14];   d3[ 3] = s[ 15];
+   d0[ 4] = s[ 16];   d1[ 4] = s[ 17];    d2[ 4] = s[ 18];   d3[ 4] = s[ 19];
+   d0[ 5] = s[ 20];   d1[ 5] = s[ 21];    d2[ 5] = s[ 22];   d3[ 5] = s[ 23];
+   d0[ 6] = s[ 24];   d1[ 6] = s[ 25];    d2[ 6] = s[ 26];   d3[ 6] = s[ 27];
+   d0[ 7] = s[ 28];   d1[ 7] = s[ 29];    d2[ 7] = s[ 30];   d3[ 7] = s[ 31];
+   d0[ 8] = s[ 32];   d1[ 8] = s[ 33];    d2[ 8] = s[ 34];   d3[ 8] = s[ 35];
+   d0[ 9] = s[ 36];   d1[ 9] = s[ 37];    d2[ 9] = s[ 38];   d3[ 9] = s[ 39];
+   d0[10] = s[ 40];   d1[10] = s[ 41];    d2[10] = s[ 42];   d3[10] = s[ 43];
+   d0[11] = s[ 44];   d1[11] = s[ 45];    d2[11] = s[ 46];   d3[11] = s[ 47];
+   d0[12] = s[ 48];   d1[12] = s[ 49];    d2[12] = s[ 50];   d3[12] = s[ 51];
+   d0[13] = s[ 52];   d1[13] = s[ 53];    d2[13] = s[ 54];   d3[13] = s[ 55];
+   d0[14] = s[ 56];   d1[14] = s[ 57];    d2[14] = s[ 58];   d3[14] = s[ 59];
+   d0[15] = s[ 60];   d1[15] = s[ 61];    d2[15] = s[ 62];   d3[15] = s[ 63];
+}
+
+#undef DLEAVE_4x32
+
+static inline void extr_lane_4x32( void *d, const void *s,
+                                         const int lane, const int bit_len )
+{
+   ((uint32_t*)d)[ 0] = ((uint32_t*)s)[ lane    ];
+   ((uint32_t*)d)[ 1] = ((uint32_t*)s)[ lane+ 4 ];
+   ((uint32_t*)d)[ 2] = ((uint32_t*)s)[ lane+ 8 ];
+   ((uint32_t*)d)[ 3] = ((uint32_t*)s)[ lane+12 ];
+   ((uint32_t*)d)[ 4] = ((uint32_t*)s)[ lane+16 ];
+   ((uint32_t*)d)[ 5] = ((uint32_t*)s)[ lane+20 ];
+   ((uint32_t*)d)[ 6] = ((uint32_t*)s)[ lane+24 ];
+   ((uint32_t*)d)[ 7] = ((uint32_t*)s)[ lane+28 ];
+   if ( bit_len <= 256 ) return;
+   ((uint32_t*)d)[ 8] = ((uint32_t*)s)[ lane+32 ];
+   ((uint32_t*)d)[ 9] = ((uint32_t*)s)[ lane+36 ];
+   ((uint32_t*)d)[10] = ((uint32_t*)s)[ lane+40 ];
+   ((uint32_t*)d)[11] = ((uint32_t*)s)[ lane+44 ];
+   ((uint32_t*)d)[12] = ((uint32_t*)s)[ lane+48 ];
+   ((uint32_t*)d)[13] = ((uint32_t*)s)[ lane+52 ];
+   ((uint32_t*)d)[14] = ((uint32_t*)s)[ lane+56 ];
+   ((uint32_t*)d)[15] = ((uint32_t*)s)[ lane+60 ];
+}
+
+
+
+// Still used by decred due to odd data size: 180 bytes
+// bit_len must be multiple of 32
+static inline void mm128_intrlv_4x32x( void *dst, void *src0, void  *src1,
+                                        void *src2, void *src3, int bit_len )
+{
+   uint32_t *d  = (uint32_t*)dst;
+   uint32_t *s0 = (uint32_t*)src0;
+   uint32_t *s1 = (uint32_t*)src1;
+   uint32_t *s2 = (uint32_t*)src2;
+   uint32_t *s3 = (uint32_t*)src3;
+
+   for ( int i = 0; i < bit_len >> 5; i++, d += 4 )
+   {
+      *d     = *(s0+i);
+      *(d+1) = *(s1+i);
+      *(d+2) = *(s2+i);
+      *(d+3) = *(s3+i);
+   }
+}
+
+static inline void mm128_bswap32_intrlv80_4x32( void *d, void *src )
+{
+  __m128i sx = mm128_bswap_32( casti_m128i( src,0 ) );
+  __m128i sy = mm128_bswap_32( casti_m128i( src,1 ) );
+  casti_m128i( d, 0 ) = _mm_shuffle_epi32( sx, 0x00 );
+  casti_m128i( d, 1 ) = _mm_shuffle_epi32( sx, 0x55 );
+  casti_m128i( d, 2 ) = _mm_shuffle_epi32( sx, 0xaa );
+  casti_m128i( d, 3 ) = _mm_shuffle_epi32( sx, 0xff );
+  sx = mm128_bswap_32( casti_m128i( src,2 ) );
+  casti_m128i( d, 4 ) = _mm_shuffle_epi32( sy, 0x00 );
+  casti_m128i( d, 5 ) = _mm_shuffle_epi32( sy, 0x55 );
+  casti_m128i( d, 6 ) = _mm_shuffle_epi32( sy, 0xaa );
+  casti_m128i( d, 7 ) = _mm_shuffle_epi32( sy, 0xff );
+  sy = mm128_bswap_32( casti_m128i( src,3 ) );
+  casti_m128i( d, 8 ) = _mm_shuffle_epi32( sx, 0x00 );
+  casti_m128i( d, 9 ) = _mm_shuffle_epi32( sx, 0x55 );
+  casti_m128i( d,10 ) = _mm_shuffle_epi32( sx, 0xaa );
+  casti_m128i( d,11 ) = _mm_shuffle_epi32( sx, 0xff );
+  sx = mm128_bswap_32( casti_m128i( src,4 ) );
+  casti_m128i( d,12 ) = _mm_shuffle_epi32( sy, 0x00 );
+  casti_m128i( d,13 ) = _mm_shuffle_epi32( sy, 0x55 );
+  casti_m128i( d,14 ) = _mm_shuffle_epi32( sy, 0xaa );
+  casti_m128i( d,15 ) = _mm_shuffle_epi32( sy, 0xff );
+  casti_m128i( d,16 ) = _mm_shuffle_epi32( sx, 0x00 );
+  casti_m128i( d,17 ) = _mm_shuffle_epi32( sx, 0x55 );
+  casti_m128i( d,18 ) = _mm_shuffle_epi32( sx, 0xaa );
+  casti_m128i( d,19 ) = _mm_shuffle_epi32( sx, 0xff );
+}
+
+// 8x32
+
+#define ILEAVE_8x32( i ) do \
+{ \
+  uint32_t *d = (uint32_t*)(dst) + ( (i) << 3 ); \
+  d[0] = *( (const uint32_t*)(s0) +(i) ); \
+  d[1] = *( (const uint32_t*)(s1) +(i) ); \
+  d[2] = *( (const uint32_t*)(s2) +(i) ); \
+  d[3] = *( (const uint32_t*)(s3) +(i) ); \
+  d[4] = *( (const uint32_t*)(s4) +(i) ); \
+  d[5] = *( (const uint32_t*)(s5) +(i) ); \
+  d[6] = *( (const uint32_t*)(s6) +(i) ); \
+  d[7] = *( (const uint32_t*)(s7) +(i) ); \
+} while(0)
+  
+static inline void intrlv_8x32( void *dst, const void *s0, const void *s1,
+           const void *s2, const void *s3, const void *s4, const void *s5,
+           const void *s6, const void *s7, int bit_len )
+{
+   ILEAVE_8x32(  0 );   ILEAVE_8x32(  1 );
+   ILEAVE_8x32(  2 );   ILEAVE_8x32(  3 );
+   ILEAVE_8x32(  4 );   ILEAVE_8x32(  5 );
+   ILEAVE_8x32(  6 );   ILEAVE_8x32(  7 );
+   if ( bit_len <= 256 ) return;
+   ILEAVE_8x32(  8 );   ILEAVE_8x32(  9 );
+   ILEAVE_8x32( 10 );   ILEAVE_8x32( 11 );
+   ILEAVE_8x32( 12 );   ILEAVE_8x32( 13 );
+   ILEAVE_8x32( 14 );   ILEAVE_8x32( 15 );
+   if ( bit_len <= 512 ) return;
+   ILEAVE_8x32( 16 );   ILEAVE_8x32( 17 );
+   ILEAVE_8x32( 18 );   ILEAVE_8x32( 19 );
+   if ( bit_len <= 640 ) return;
+   ILEAVE_8x32( 20 );   ILEAVE_8x32( 21 );
+   ILEAVE_8x32( 22 );   ILEAVE_8x32( 23 );
+   ILEAVE_8x32( 24 );   ILEAVE_8x32( 25 );
+   ILEAVE_8x32( 26 );   ILEAVE_8x32( 27 );
+   ILEAVE_8x32( 28 );   ILEAVE_8x32( 29 );
+   ILEAVE_8x32( 30 );   ILEAVE_8x32( 31 );
+}
+
+static inline void intrlv_8x32_512( void *dst, const void *s0, const void *s1,
+               const void *s2, const void *s3, const void *s4, const void *s5,
+               const void *s6, const void *s7 )
+{
+   ILEAVE_8x32(  0 );   ILEAVE_8x32(  1 );
+   ILEAVE_8x32(  2 );   ILEAVE_8x32(  3 );
+   ILEAVE_8x32(  4 );   ILEAVE_8x32(  5 );
+   ILEAVE_8x32(  6 );   ILEAVE_8x32(  7 );
+   ILEAVE_8x32(  8 );   ILEAVE_8x32(  9 );
+   ILEAVE_8x32( 10 );   ILEAVE_8x32( 11 );
+   ILEAVE_8x32( 12 );   ILEAVE_8x32( 13 );
+   ILEAVE_8x32( 14 );   ILEAVE_8x32( 15 );
+}
+
+#undef ILEAVE_8x32
+
+#define DLEAVE_8x32( i ) do \
+{ \
+   const uint32_t *s = (const uint32_t*)(src) + ( (i) << 3 ); \
+   *( (uint32_t*)(d0) +(i) ) = s[0]; \
+   *( (uint32_t*)(d1) +(i) ) = s[1]; \
+   *( (uint32_t*)(d2) +(i) ) = s[2]; \
+   *( (uint32_t*)(d3) +(i) ) = s[3]; \
+   *( (uint32_t*)(d4) +(i) ) = s[4]; \
+   *( (uint32_t*)(d5) +(i) ) = s[5]; \
+   *( (uint32_t*)(d6) +(i) ) = s[6]; \
+   *( (uint32_t*)(d7) +(i) ) = s[7]; \
+} while(0)
+
+static inline void dintrlv_8x32( void *d0, void *d1, void *d2, void *d3,
+        void *d4, void *d5, void *d6, void *d7, const void *src, int bit_len )
+{
+   DLEAVE_8x32(  0 );   DLEAVE_8x32(  1 );
+   DLEAVE_8x32(  2 );   DLEAVE_8x32(  3 );
+   DLEAVE_8x32(  4 );   DLEAVE_8x32(  5 );
+   DLEAVE_8x32(  6 );   DLEAVE_8x32(  7 );
+   if ( bit_len <= 256 ) return;
+   DLEAVE_8x32(  8 );   DLEAVE_8x32(  9 );
+   DLEAVE_8x32( 10 );   DLEAVE_8x32( 11 );
+   DLEAVE_8x32( 12 );   DLEAVE_8x32( 13 );
+   DLEAVE_8x32( 14 );   DLEAVE_8x32( 15 );
+   if ( bit_len <= 512 ) return;
+   DLEAVE_8x32( 16 );   DLEAVE_8x32( 17 );
+   DLEAVE_8x32( 18 );   DLEAVE_8x32( 19 );
+   if ( bit_len <= 640 ) return;
+   DLEAVE_8x32( 20 );   DLEAVE_8x32( 21 );
+   DLEAVE_8x32( 22 );   DLEAVE_8x32( 23 );
+   DLEAVE_8x32( 24 );   DLEAVE_8x32( 25 );
+   DLEAVE_8x32( 26 );   DLEAVE_8x32( 27 );
+   DLEAVE_8x32( 28 );   DLEAVE_8x32( 29 );
+   DLEAVE_8x32( 30 );   DLEAVE_8x32( 31 );
+}
+
+static inline void dintrlv_8x32_512( void *d0, void *d1, void *d2, void *d3,
+                     void *d4, void *d5, void *d6, void *d7, const void *src )
+{
+   DLEAVE_8x32(  0 );   DLEAVE_8x32(  1 );
+   DLEAVE_8x32(  2 );   DLEAVE_8x32(  3 );
+   DLEAVE_8x32(  4 );   DLEAVE_8x32(  5 );
+   DLEAVE_8x32(  6 );   DLEAVE_8x32(  7 );
+   DLEAVE_8x32(  8 );   DLEAVE_8x32(  9 );
+   DLEAVE_8x32( 10 );   DLEAVE_8x32( 11 );
+   DLEAVE_8x32( 12 );   DLEAVE_8x32( 13 );
+   DLEAVE_8x32( 14 );   DLEAVE_8x32( 15 );
+}
+
+#undef DLEAVE_8x32
+
+static inline void extr_lane_8x32( void *d, const void *s,
+                                   const int lane, const int bit_len )
+{
+   ((uint32_t*)d)[ 0] = ((uint32_t*)s)[ lane     ];
+   ((uint32_t*)d)[ 1] = ((uint32_t*)s)[ lane+  8 ];
+   ((uint32_t*)d)[ 2] = ((uint32_t*)s)[ lane+ 16 ];
+   ((uint32_t*)d)[ 3] = ((uint32_t*)s)[ lane+ 24 ];
+   ((uint32_t*)d)[ 4] = ((uint32_t*)s)[ lane+ 32 ];
+   ((uint32_t*)d)[ 5] = ((uint32_t*)s)[ lane+ 40 ];
+   ((uint32_t*)d)[ 6] = ((uint32_t*)s)[ lane+ 48 ];
+   ((uint32_t*)d)[ 7] = ((uint32_t*)s)[ lane+ 56 ];
+   if ( bit_len <= 256 ) return;
+   ((uint32_t*)d)[ 8] = ((uint32_t*)s)[ lane+ 64 ];
+   ((uint32_t*)d)[ 9] = ((uint32_t*)s)[ lane+ 72 ];
+   ((uint32_t*)d)[10] = ((uint32_t*)s)[ lane+ 80 ];
+   ((uint32_t*)d)[11] = ((uint32_t*)s)[ lane+ 88 ];
+   ((uint32_t*)d)[12] = ((uint32_t*)s)[ lane+ 96 ];
+   ((uint32_t*)d)[13] = ((uint32_t*)s)[ lane+104 ];
+   ((uint32_t*)d)[14] = ((uint32_t*)s)[ lane+112 ];
+   ((uint32_t*)d)[15] = ((uint32_t*)s)[ lane+120 ];
+}
+
+#if defined(__AVX2__)
+
+// There a alignment problems with the source buffer on Wwindows,
+// can't use 256 bit bswap. 
+
+static inline void mm256_bswap32_intrlv80_8x32( void *d, void *src )
+{
+   __m256i s0 = mm256_bswap_32( casti_m256i( src,0 ) );
+   __m256i s1 = mm256_bswap_32( casti_m256i( src,1 ) );
+   __m128i s2 = mm128_bswap_32( casti_m128i( src,4 ) );
+  const __m256i zero = m256_zero;
+  const __m256i one  = m256_one_32;
+  const __m256i two  = _mm256_add_epi32( one, one );
+  const __m256i tre  = _mm256_add_epi32( two, one );
+  const __m256i four = _mm256_add_epi32( two, two );
+
+  casti_m256i( d, 0 ) = _mm256_permutevar8x32_epi32( s0, zero );
+  casti_m256i( d, 1 ) = _mm256_permutevar8x32_epi32( s0, one  );
+  casti_m256i( d, 2 ) = _mm256_permutevar8x32_epi32( s0, two  );
+  casti_m256i( d, 3 ) = _mm256_permutevar8x32_epi32( s0, tre  );
+  casti_m256i( d, 4 ) = _mm256_permutevar8x32_epi32( s0, four );
+  casti_m256i( d, 5 ) = _mm256_permutevar8x32_epi32( s0,
+                                       _mm256_add_epi32( four, one ) );
+  casti_m256i( d, 6 ) = _mm256_permutevar8x32_epi32( s0,
+                                       _mm256_add_epi32( four, two ) );
+  casti_m256i( d, 7 ) = _mm256_permutevar8x32_epi32( s0,
+                                       _mm256_add_epi32( four, tre ) );
+  casti_m256i( d, 8 ) = _mm256_permutevar8x32_epi32( s1, zero );
+  casti_m256i( d, 9 ) = _mm256_permutevar8x32_epi32( s1, one  );
+  casti_m256i( d,10 ) = _mm256_permutevar8x32_epi32( s1, two  );
+  casti_m256i( d,11 ) = _mm256_permutevar8x32_epi32( s1, tre  );
+  casti_m256i( d,12 ) = _mm256_permutevar8x32_epi32( s1, four );
+  casti_m256i( d,13 ) = _mm256_permutevar8x32_epi32( s1,
+                                       _mm256_add_epi32( four, one ) );
+  casti_m256i( d,14 ) = _mm256_permutevar8x32_epi32( s1,
+                                       _mm256_add_epi32( four, two  ) );
+  casti_m256i( d,15 ) = _mm256_permutevar8x32_epi32( s1,
+                                       _mm256_add_epi32( four, tre ) );
+  casti_m256i( d,16 ) = _mm256_permutevar8x32_epi32(
+                             _mm256_castsi128_si256( s2 ), zero );
+  casti_m256i( d,17 ) = _mm256_permutevar8x32_epi32(
+                             _mm256_castsi128_si256( s2 ), one  );
+  casti_m256i( d,18 ) = _mm256_permutevar8x32_epi32(
+                             _mm256_castsi128_si256( s2 ), two  );
+  casti_m256i( d,19 ) = _mm256_permutevar8x32_epi32( 
+                             _mm256_castsi128_si256( s2 ), tre  );
+}
+
+#endif   // AVX2
+
+// 16x32
+
+#define ILEAVE_16x32( i ) do \
+{ \
+  uint32_t *d = (uint32_t*)(dst) + ( (i) << 4 ); \
+  d[ 0] = *( (const uint32_t*)(s00) +(i) ); \
+  d[ 1] = *( (const uint32_t*)(s01) +(i) ); \
+  d[ 2] = *( (const uint32_t*)(s02) +(i) ); \
+  d[ 3] = *( (const uint32_t*)(s03) +(i) ); \
+  d[ 4] = *( (const uint32_t*)(s04) +(i) ); \
+  d[ 5] = *( (const uint32_t*)(s05) +(i) ); \
+  d[ 6] = *( (const uint32_t*)(s06) +(i) ); \
+  d[ 7] = *( (const uint32_t*)(s07) +(i) ); \
+  d[ 8] = *( (const uint32_t*)(s08) +(i) ); \
+  d[ 9] = *( (const uint32_t*)(s09) +(i) ); \
+  d[10] = *( (const uint32_t*)(s10) +(i) ); \
+  d[11] = *( (const uint32_t*)(s11) +(i) ); \
+  d[12] = *( (const uint32_t*)(s12) +(i) ); \
+  d[13] = *( (const uint32_t*)(s13) +(i) ); \
+  d[14] = *( (const uint32_t*)(s14) +(i) ); \
+  d[15] = *( (const uint32_t*)(s15) +(i) ); \
+} while(0)
+
+static inline void intrlv_16x32( void *dst, const void *s00,
+        const void *s01, const void *s02, const void *s03, const void *s04,
+        const void *s05, const void *s06, const void *s07, const void *s08,
+        const void *s09, const void *s10, const void *s11, const void *s12,
+        const void *s13, const void *s14, const void *s15, int bit_len )
+{
+   ILEAVE_16x32(  0 );   ILEAVE_16x32(  1 );
+   ILEAVE_16x32(  2 );   ILEAVE_16x32(  3 );
+   ILEAVE_16x32(  4 );   ILEAVE_16x32(  5 );
+   ILEAVE_16x32(  6 );   ILEAVE_16x32(  7 );
+   if ( bit_len <= 256 ) return;
+   ILEAVE_16x32(  8 );   ILEAVE_16x32(  9 );
+   ILEAVE_16x32( 10 );   ILEAVE_16x32( 11 );
+   ILEAVE_16x32( 12 );   ILEAVE_16x32( 13 );
+   ILEAVE_16x32( 14 );   ILEAVE_16x32( 15 );
+   if ( bit_len <= 512 ) return;
+   ILEAVE_16x32( 16 );   ILEAVE_16x32( 17 );
+   ILEAVE_16x32( 18 );   ILEAVE_16x32( 19 );
+   if ( bit_len <= 640 ) return;
+   ILEAVE_16x32( 20 );   ILEAVE_16x32( 21 );
+   ILEAVE_16x32( 22 );   ILEAVE_16x32( 23 );
+   ILEAVE_16x32( 24 );   ILEAVE_16x32( 25 );
+   ILEAVE_16x32( 26 );   ILEAVE_16x32( 27 );
+   ILEAVE_16x32( 28 );   ILEAVE_16x32( 29 );
+   ILEAVE_16x32( 30 );   ILEAVE_16x32( 31 );
+}
+
+static inline void intrlv_16x32_512( void *dst, const void *s00,
+        const void *s01, const void *s02, const void *s03, const void *s04,
+        const void *s05, const void *s06, const void *s07, const void *s08,
+        const void *s09, const void *s10, const void *s11, const void *s12,
+        const void *s13, const void *s14, const void *s15 )
+{
+   ILEAVE_16x32(  0 );   ILEAVE_16x32(  1 );
+   ILEAVE_16x32(  2 );   ILEAVE_16x32(  3 );
+   ILEAVE_16x32(  4 );   ILEAVE_16x32(  5 );
+   ILEAVE_16x32(  6 );   ILEAVE_16x32(  7 );
+   ILEAVE_16x32(  8 );   ILEAVE_16x32(  9 );
+   ILEAVE_16x32( 10 );   ILEAVE_16x32( 11 );
+   ILEAVE_16x32( 12 );   ILEAVE_16x32( 13 );
+   ILEAVE_16x32( 14 );   ILEAVE_16x32( 15 );
+}
+
+#undef ILEAVE_16x32
+
+#define DLEAVE_16x32( i ) do \
+{ \
+   const uint32_t *s = (const uint32_t*)(src) + ( (i) << 4 ); \
+   *( (uint32_t*)(d00) +(i) ) = s[ 0]; \
+   *( (uint32_t*)(d01) +(i) ) = s[ 1]; \
+   *( (uint32_t*)(d02) +(i) ) = s[ 2]; \
+   *( (uint32_t*)(d03) +(i) ) = s[ 3]; \
+   *( (uint32_t*)(d04) +(i) ) = s[ 4]; \
+   *( (uint32_t*)(d05) +(i) ) = s[ 5]; \
+   *( (uint32_t*)(d06) +(i) ) = s[ 6]; \
+   *( (uint32_t*)(d07) +(i) ) = s[ 7]; \
+   *( (uint32_t*)(d08) +(i) ) = s[ 8]; \
+   *( (uint32_t*)(d09) +(i) ) = s[ 0]; \
+   *( (uint32_t*)(d10) +(i) ) = s[10]; \
+   *( (uint32_t*)(d11) +(i) ) = s[11]; \
+   *( (uint32_t*)(d12) +(i) ) = s[12]; \
+   *( (uint32_t*)(d13) +(i) ) = s[13]; \
+   *( (uint32_t*)(d14) +(i) ) = s[14]; \
+   *( (uint32_t*)(d15) +(i) ) = s[15]; \
+} while(0)
+
+static inline void dintrlv_16x32( void *d00, void *d01, void *d02, void *d03,
+            void *d04, void *d05, void *d06, void *d07, void *d08, void *d09,
+            void *d10, void *d11, void *d12, void *d13, void *d14, void *d15,
+            const void *src, int bit_len )
+{
+   DLEAVE_16x32(  0 );   DLEAVE_16x32(  1 );
+   DLEAVE_16x32(  2 );   DLEAVE_16x32(  3 );
+   DLEAVE_16x32(  4 );   DLEAVE_16x32(  5 );
+   DLEAVE_16x32(  6 );   DLEAVE_16x32(  7 );
+   if ( bit_len <= 256 ) return;
+   DLEAVE_16x32(  8 );   DLEAVE_16x32(  9 );
+   DLEAVE_16x32( 10 );   DLEAVE_16x32( 11 );
+   DLEAVE_16x32( 12 );   DLEAVE_16x32( 13 );
+   DLEAVE_16x32( 14 );   DLEAVE_16x32( 15 );
+   if ( bit_len <= 512 ) return;
+   DLEAVE_16x32( 16 );   DLEAVE_16x32( 17 );
+   DLEAVE_16x32( 18 );   DLEAVE_16x32( 19 );
+   if ( bit_len <= 640 ) return;
+   DLEAVE_16x32( 20 );   DLEAVE_16x32( 21 );
+   DLEAVE_16x32( 22 );   DLEAVE_16x32( 23 );
+   DLEAVE_16x32( 24 );   DLEAVE_16x32( 25 );
+   DLEAVE_16x32( 26 );   DLEAVE_16x32( 27 );
+   DLEAVE_16x32( 28 );   DLEAVE_16x32( 29 );
+   DLEAVE_16x32( 30 );   DLEAVE_16x32( 31 );
+}
+
+static inline void dintrlv_16x32_512( void *d00, void *d01, void *d02,
+                void *d03, void *d04, void *d05, void *d06, void *d07,
+                void *d08, void *d09, void *d10, void *d11, void *d12,
+                void *d13, void *d14, void *d15, const void *src )
+{
+   DLEAVE_16x32(  0 );   DLEAVE_16x32(  1 );
+   DLEAVE_16x32(  2 );   DLEAVE_16x32(  3 );
+   DLEAVE_16x32(  4 );   DLEAVE_16x32(  5 );
+   DLEAVE_16x32(  6 );   DLEAVE_16x32(  7 );
+   DLEAVE_16x32(  8 );   DLEAVE_16x32(  9 );
+   DLEAVE_16x32( 10 );   DLEAVE_16x32( 11 );
+   DLEAVE_16x32( 12 );   DLEAVE_16x32( 13 );
+   DLEAVE_16x32( 14 );   DLEAVE_16x32( 15 );
+}
+
+#undef DLEAVE_16x32
+
+static inline void extr_lane_16x32( void *d, const void *s,
+                                   const int lane, const int bit_len )
+{
+   ((uint32_t*)d)[ 0] = ((uint32_t*)s)[ lane    ];
+   ((uint32_t*)d)[ 1] = ((uint32_t*)s)[ lane+16 ];
+   ((uint32_t*)d)[ 2] = ((uint32_t*)s)[ lane+32 ];
+   ((uint32_t*)d)[ 3] = ((uint32_t*)s)[ lane+48 ];
+   ((uint32_t*)d)[ 4] = ((uint32_t*)s)[ lane+64 ];
+   ((uint32_t*)d)[ 5] = ((uint32_t*)s)[ lane+80 ];
+   ((uint32_t*)d)[ 6] = ((uint32_t*)s)[ lane+96 ];
+   ((uint32_t*)d)[ 7] = ((uint32_t*)s)[ lane+112 ];
+   if ( bit_len <= 256 ) return;
+   ((uint32_t*)d)[ 8] = ((uint32_t*)s)[ lane+128 ];
+   ((uint32_t*)d)[ 9] = ((uint32_t*)s)[ lane+144 ];
+   ((uint32_t*)d)[10] = ((uint32_t*)s)[ lane+160 ];
+   ((uint32_t*)d)[11] = ((uint32_t*)s)[ lane+176 ];
+   ((uint32_t*)d)[12] = ((uint32_t*)s)[ lane+192 ];
+   ((uint32_t*)d)[13] = ((uint32_t*)s)[ lane+208 ];
+   ((uint32_t*)d)[14] = ((uint32_t*)s)[ lane+224 ];
+   ((uint32_t*)d)[15] = ((uint32_t*)s)[ lane+240 ];
+}
+
+#if defined(__AVX512F__) && defined(__AVX512VL__)
+
+static inline void mm512_bswap32_intrlv80_16x32( void *d, void *src )
+{
+  __m512i s0 = mm512_bswap_32( casti_m512i( src, 0 ) );
+  __m128i s1 = mm128_bswap_32( casti_m128i( src, 4 ) );
+  const __m512i zero   = m512_zero;
+  const __m512i one    = m512_one_32;
+  const __m512i two    = _mm512_add_epi32( one,   one  );
+  const __m512i tre    = _mm512_add_epi32( two,   one  );
+  const __m512i four   = _mm512_add_epi32( two,   two  );
+  const __m512i eight  = _mm512_add_epi32( four,  four );
+  const __m512i eleven = _mm512_add_epi32( eight, tre  );
+
+  casti_m512i( d, 0 ) = _mm512_permutexvar_epi32( s0, zero );
+  casti_m512i( d, 1 ) = _mm512_permutexvar_epi32( s0, one  );
+  casti_m512i( d, 2 ) = _mm512_permutexvar_epi32( s0, two  );
+  casti_m512i( d, 3 ) = _mm512_permutexvar_epi32( s0, tre  );
+  casti_m512i( d, 4 ) = _mm512_permutexvar_epi32( s0, four );
+  casti_m512i( d, 5 ) = _mm512_permutexvar_epi32( s0,
+                                    _mm512_add_epi32( four, one ) );
+  casti_m512i( d, 6 ) = _mm512_permutexvar_epi32( s0, 
+                                    _mm512_add_epi32( four, two ) );
+  casti_m512i( d, 7 ) = _mm512_permutexvar_epi32( s0,
+                                    _mm512_add_epi32( four, tre ) );
+  casti_m512i( d, 8 ) = _mm512_permutexvar_epi32( s0, eight );
+  casti_m512i( d, 9 ) = _mm512_permutexvar_epi32( s0, 
+                                    _mm512_add_epi32( eight, one ) );
+  casti_m512i( d,10 ) = _mm512_permutexvar_epi32( s0,
+                                    _mm512_add_epi32( eight, two ) );
+  casti_m512i( d,11 ) = _mm512_permutexvar_epi32( s0, eleven ); 
+  casti_m512i( d,12 ) = _mm512_permutexvar_epi32( s0, 
+                                    _mm512_add_epi32( eleven, one ) );
+  casti_m512i( d,13 ) = _mm512_permutexvar_epi32( s0, 
+                                    _mm512_add_epi32( eleven, two ) );
+  casti_m512i( d,14 ) = _mm512_permutexvar_epi32( s0, 
+                                    _mm512_add_epi32( eleven, tre ) );
+  casti_m512i( d,15 ) = _mm512_permutexvar_epi32( s0,
+                                    _mm512_add_epi32( eleven, four ) );
+  casti_m512i( d,16 ) = _mm512_permutexvar_epi32(
+                          _mm512_castsi128_si512( s1 ), zero );
+  casti_m512i( d,17 ) = _mm512_permutexvar_epi32(
+                          _mm512_castsi128_si512( s1 ), one  );
+  casti_m512i( d,18 ) = _mm512_permutexvar_epi32(
+                          _mm512_castsi128_si512( s1 ), two  );
+  casti_m512i( d,19 ) = _mm512_permutexvar_epi32(
+                          _mm512_castsi128_si512( s1 ), tre  );
+}
+
+#endif    // AVX512
+
+///////////////////////////
+//
+//     64 bit data
+
+// 2x64    (SSE2)
+
+static inline void intrlv_2x64( void *dst, const void *src0,
+                                const void *src1, int bit_len )
+{
+   uint64_t *d = (uint64_t*)dst;;
+   const uint64_t *s0 = (const uint64_t*)src0;
+   const uint64_t *s1 = (const uint64_t*)src1;
+   d[ 0] = s0[ 0];    d[ 1] = s1[ 0];   d[ 2] = s0[ 1];    d[ 3] = s1[ 1];
+   d[ 4] = s0[ 2];    d[ 5] = s1[ 2];   d[ 6] = s0[ 3];    d[ 7] = s1[ 3];
+   if ( bit_len <= 256 ) return;
+   d[ 8] = s0[ 4];    d[ 9] = s1[ 4];   d[10] = s0[ 5];    d[11] = s1[ 5];
+   d[12] = s0[ 6];    d[13] = s1[ 6];   d[14] = s0[ 7];    d[15] = s1[ 7];
+   if ( bit_len <= 512 ) return;
+   d[16] = s0[ 8];    d[17] = s1[ 8];   d[18] = s0[ 9];    d[19] = s1[ 9];
+   if ( bit_len <= 640 ) return;
+   d[20] = s0[10];    d[21] = s1[10];   d[22] = s0[11];    d[23] = s1[11];
+   d[24] = s0[12];    d[25] = s1[12];   d[26] = s0[13];    d[27] = s1[13];
+   d[28] = s0[14];    d[29] = s1[14];   d[30] = s0[15];    d[31] = s1[15];
+}
+
+static inline void dintrlv_2x64( void *dst0, void *dst1,
+                                 const void *src, int bit_len )
+{
+   uint64_t *d0 = (uint64_t*)dst0;
+   uint64_t *d1 = (uint64_t*)dst1;
+   const uint64_t *s = (const uint64_t*)src;
+
+   d0[ 0] = s[ 0];   d1[ 0] = s[ 1];   d0[ 1] = s[ 2];   d1[ 1] = s[ 3];
+   d0[ 2] = s[ 4];   d1[ 2] = s[ 5];   d0[ 3] = s[ 6];   d1[ 3] = s[ 7];
+   if ( bit_len <= 256 ) return;
+   d0[ 4] = s[ 8];   d1[ 4] = s[ 9];   d0[ 5] = s[10];   d1[ 5] = s[11];
+   d0[ 6] = s[12];   d1[ 6] = s[13];   d0[ 7] = s[14];   d1[ 7] = s[15];
+   if ( bit_len <= 512 ) return;
+   d0[ 8] = s[16];   d1[ 8] = s[17];   d0[ 9] = s[18];   d1[ 9] = s[19];
+   if ( bit_len <= 640 ) return;
+   d0[10] = s[20];   d1[10] = s[21];   d0[11] = s[22];   d1[11] = s[23];
+   d0[12] = s[24];   d1[12] = s[25];   d0[13] = s[26];   d1[13] = s[27];
+   d0[14] = s[28];   d1[14] = s[29];   d0[15] = s[30];   d1[15] = s[31];
+}
+
+// 4x64   (AVX2)
+
+static inline void intrlv_4x64( void *dst, const void *src0,
+           const void *src1, const void *src2, const void *src3, int bit_len )
+{
+   uint64_t *d = (uint64_t*)dst;
+   const uint64_t *s0 = (const uint64_t*)src0;
+   const uint64_t *s1 = (const uint64_t*)src1;
+   const uint64_t *s2 = (const uint64_t*)src2;
+   const uint64_t *s3 = (const uint64_t*)src3;
+   d[  0] = s0[ 0];   d[  1] = s1[ 0];   d[  2] = s2[ 0];   d[  3] = s3[ 0];
+   d[  4] = s0[ 1];   d[  5] = s1[ 1];   d[  6] = s2[ 1];   d[  7] = s3[ 1];
+   d[  8] = s0[ 2];   d[  9] = s1[ 2];   d[ 10] = s2[ 2];   d[ 11] = s3[ 2];
+   d[ 12] = s0[ 3];   d[ 13] = s1[ 3];   d[ 14] = s2[ 3];   d[ 15] = s3[ 3];
+   if ( bit_len <= 256 ) return;
+   d[ 16] = s0[ 4];   d[ 17] = s1[ 4];   d[ 18] = s2[ 4];   d[ 19] = s3[ 4];
+   d[ 20] = s0[ 5];   d[ 21] = s1[ 5];   d[ 22] = s2[ 5];   d[ 23] = s3[ 5];
+   d[ 24] = s0[ 6];   d[ 25] = s1[ 6];   d[ 26] = s2[ 6];   d[ 27] = s3[ 6];
+   d[ 28] = s0[ 7];   d[ 29] = s1[ 7];   d[ 30] = s2[ 7];   d[ 31] = s3[ 7];
+   if ( bit_len <= 512 ) return;
+   d[ 32] = s0[ 8];   d[ 33] = s1[ 8];   d[ 34] = s2[ 8];   d[ 35] = s3[ 8];
+   d[ 36] = s0[ 9];   d[ 37] = s1[ 9];   d[ 38] = s2[ 9];   d[ 39] = s3[ 9];
+   if ( bit_len <= 640 ) return;
+   d[ 40] = s0[10];   d[ 41] = s1[10];   d[ 42] = s2[10];   d[ 43] = s3[10];
+   d[ 44] = s0[11];   d[ 45] = s1[11];   d[ 46] = s2[11];   d[ 47] = s3[11];
+   d[ 48] = s0[12];   d[ 49] = s1[12];   d[ 50] = s2[12];   d[ 51] = s3[12];
+   d[ 52] = s0[13];   d[ 53] = s1[13];   d[ 54] = s2[13];   d[ 55] = s3[13];
+   d[ 56] = s0[14];   d[ 57] = s1[14];   d[ 58] = s2[14];   d[ 59] = s3[14];
+   d[ 60] = s0[15];   d[ 61] = s1[15];   d[ 62] = s2[15];   d[ 63] = s3[15];
+}
+
+static inline void intrlv_4x64_512( void *dst, const void *src0,
+                      const void *src1, const void *src2, const void *src3 )
+{
+   uint64_t *d = (uint64_t*)dst;
+   const uint64_t *s0 = (const uint64_t*)src0;
+   const uint64_t *s1 = (const uint64_t*)src1;
+   const uint64_t *s2 = (const uint64_t*)src2;
+   const uint64_t *s3 = (const uint64_t*)src3;
+   d[  0] = s0[ 0];   d[  1] = s1[ 0];   d[  2] = s2[ 0];   d[  3] = s3[ 0];
+   d[  4] = s0[ 1];   d[  5] = s1[ 1];   d[  6] = s2[ 1];   d[  7] = s3[ 1];
+   d[  8] = s0[ 2];   d[  9] = s1[ 2];   d[ 10] = s2[ 2];   d[ 11] = s3[ 2];
+   d[ 12] = s0[ 3];   d[ 13] = s1[ 3];   d[ 14] = s2[ 3];   d[ 15] = s3[ 3];
+   d[ 16] = s0[ 4];   d[ 17] = s1[ 4];   d[ 18] = s2[ 4];   d[ 19] = s3[ 4];
+   d[ 20] = s0[ 5];   d[ 21] = s1[ 5];   d[ 22] = s2[ 5];   d[ 23] = s3[ 5];
+   d[ 24] = s0[ 6];   d[ 25] = s1[ 6];   d[ 26] = s2[ 6];   d[ 27] = s3[ 6];
+   d[ 28] = s0[ 7];   d[ 29] = s1[ 7];   d[ 30] = s2[ 7];   d[ 31] = s3[ 7];
+}
+
+static inline void dintrlv_4x64( void *dst0, void *dst1, void *dst2,
+                                 void *dst3, const void *src, int bit_len )
+{
+   uint64_t *d0 = (uint64_t*)dst0;
+   uint64_t *d1 = (uint64_t*)dst1;
+   uint64_t *d2 = (uint64_t*)dst2;
+   uint64_t *d3 = (uint64_t*)dst3;
+   const uint64_t *s = (const uint64_t*)src;
+   d0[ 0] = s[ 0];   d1[ 0] = s[ 1];    d2[ 0] = s[ 2];   d3[ 0] = s[ 3];
+   d0[ 1] = s[ 4];   d1[ 1] = s[ 5];    d2[ 1] = s[ 6];   d3[ 1] = s[ 7];
+   d0[ 2] = s[ 8];   d1[ 2] = s[ 9];    d2[ 2] = s[10];   d3[ 2] = s[11];
+   d0[ 3] = s[12];   d1[ 3] = s[13];    d2[ 3] = s[14];   d3[ 3] = s[15];
+   if ( bit_len <= 256 ) return;
+   d0[ 4] = s[16];   d1[ 4] = s[17];    d2[ 4] = s[18];   d3[ 4] = s[19];
+   d0[ 5] = s[20];   d1[ 5] = s[21];    d2[ 5] = s[22];   d3[ 5] = s[23];
+   d0[ 6] = s[24];   d1[ 6] = s[25];    d2[ 6] = s[26];   d3[ 6] = s[27];
+   d0[ 7] = s[28];   d1[ 7] = s[29];    d2[ 7] = s[30];   d3[ 7] = s[31];
+   if ( bit_len <= 512 ) return;
+   d0[ 8] = s[32];   d1[ 8] = s[33];    d2[ 8] = s[34];   d3[ 8] = s[35];
+   d0[ 9] = s[36];   d1[ 9] = s[37];    d2[ 9] = s[38];   d3[ 9] = s[39];
+   if ( bit_len <= 640 ) return;
+   d0[10] = s[40];   d1[10] = s[41];    d2[10] = s[42];   d3[10] = s[43];
+   d0[11] = s[44];   d1[11] = s[45];    d2[11] = s[46];   d3[11] = s[47];
+   d0[12] = s[48];   d1[12] = s[49];    d2[12] = s[50];   d3[12] = s[51];
+   d0[13] = s[52];   d1[13] = s[53];    d2[13] = s[54];   d3[13] = s[55];
+   d0[14] = s[56];   d1[14] = s[57];    d2[14] = s[58];   d3[14] = s[59];
+   d0[15] = s[60];   d1[15] = s[61];    d2[15] = s[62];   d3[15] = s[63];
+}
+
+static inline void dintrlv_4x64_512( void *dst0, void *dst1, void *dst2,
+                                     void *dst3, const void *src )
+{
+   uint64_t *d0 = (uint64_t*)dst0;
+   uint64_t *d1 = (uint64_t*)dst1;
+   uint64_t *d2 = (uint64_t*)dst2;
+   uint64_t *d3 = (uint64_t*)dst3;
+   const uint64_t *s = (const uint64_t*)src;
+   d0[ 0] = s[ 0];   d1[ 0] = s[ 1];    d2[ 0] = s[ 2];   d3[ 0] = s[ 3];
+   d0[ 1] = s[ 4];   d1[ 1] = s[ 5];    d2[ 1] = s[ 6];   d3[ 1] = s[ 7];
+   d0[ 2] = s[ 8];   d1[ 2] = s[ 9];    d2[ 2] = s[10];   d3[ 2] = s[11];
+   d0[ 3] = s[12];   d1[ 3] = s[13];    d2[ 3] = s[14];   d3[ 3] = s[15];
+   d0[ 4] = s[16];   d1[ 4] = s[17];    d2[ 4] = s[18];   d3[ 4] = s[19];
+   d0[ 5] = s[20];   d1[ 5] = s[21];    d2[ 5] = s[22];   d3[ 5] = s[23];
+   d0[ 6] = s[24];   d1[ 6] = s[25];    d2[ 6] = s[26];   d3[ 6] = s[27];
+   d0[ 7] = s[28];   d1[ 7] = s[29];    d2[ 7] = s[30];   d3[ 7] = s[31];
+}
+
+static inline void extr_lane_4x64( void *d, const void *s,
+                                   const int lane, const int bit_len )
+{
+   ((uint64_t*)d)[ 0] = ((uint64_t*)s)[ lane    ];
+   ((uint64_t*)d)[ 1] = ((uint64_t*)s)[ lane+ 4 ];
+   ((uint64_t*)d)[ 2] = ((uint64_t*)s)[ lane+ 8 ];
+   ((uint64_t*)d)[ 3] = ((uint64_t*)s)[ lane+12 ];
+   ((uint64_t*)d)[ 4] = ((uint64_t*)s)[ lane+16 ];
+   ((uint64_t*)d)[ 5] = ((uint64_t*)s)[ lane+20 ];
+   ((uint64_t*)d)[ 6] = ((uint64_t*)s)[ lane+24 ];
+   ((uint64_t*)d)[ 7] = ((uint64_t*)s)[ lane+28 ];
+   if ( bit_len <= 256 ) return;
+   ((uint64_t*)d)[ 8] = ((uint64_t*)s)[ lane+32 ];
+   ((uint64_t*)d)[ 9] = ((uint64_t*)s)[ lane+36 ];
+   ((uint64_t*)d)[10] = ((uint64_t*)s)[ lane+40 ];
+   ((uint64_t*)d)[11] = ((uint64_t*)s)[ lane+44 ];
+   ((uint64_t*)d)[12] = ((uint64_t*)s)[ lane+48 ];
+   ((uint64_t*)d)[13] = ((uint64_t*)s)[ lane+52 ];
+   ((uint64_t*)d)[14] = ((uint64_t*)s)[ lane+56 ];
+   ((uint64_t*)d)[15] = ((uint64_t*)s)[ lane+60 ];
+}
+
+#if defined(__AVX2__)
+
+// There a alignment problems with the source buffer on Wwindows,
+// can't use 256 bit bswap.
+
+static inline void mm256_bswap32_intrlv80_4x64( void *d, void *src )
+{
+  __m256i s0 = mm256_bswap_32( casti_m256i( src, 0 ) );
+  __m256i s1 = mm256_bswap_32( casti_m256i( src, 1 ) );
+  __m128i s2 = mm128_bswap_32( casti_m128i( src, 4 ) );
+
+  casti_m256i( d, 0 ) = _mm256_permute4x64_epi64( s0, 0x00 );
+  casti_m256i( d, 1 ) = _mm256_permute4x64_epi64( s0, 0x55 );
+  casti_m256i( d, 2 ) = _mm256_permute4x64_epi64( s0, 0xaa );
+  casti_m256i( d, 3 ) = _mm256_permute4x64_epi64( s0, 0xff );
+  casti_m256i( d, 4 ) = _mm256_permute4x64_epi64( s1, 0x00 );
+  casti_m256i( d, 5 ) = _mm256_permute4x64_epi64( s1, 0x55 );
+  casti_m256i( d, 6 ) = _mm256_permute4x64_epi64( s1, 0xaa );
+  casti_m256i( d, 7 ) = _mm256_permute4x64_epi64( s1, 0xff );
+  casti_m256i( d, 8 ) = _mm256_permute4x64_epi64(
+                          _mm256_castsi128_si256( s2 ), 0x00 );
+  casti_m256i( d, 9 ) = _mm256_permute4x64_epi64(
+                          _mm256_castsi128_si256( s2 ), 0x55 );
+}
+
+#endif  // AVX2
+
+// 8x64   (AVX512)
+
+#define ILEAVE_8x64( i ) do \
+{ \
+  uint64_t *d = (uint64_t*)(dst) + ( (i) << 3 ); \
+  d[0] = *( (const uint64_t*)(s0) +(i) ); \
+  d[1] = *( (const uint64_t*)(s1) +(i) ); \
+  d[2] = *( (const uint64_t*)(s2) +(i) ); \
+  d[3] = *( (const uint64_t*)(s3) +(i) ); \
+  d[4] = *( (const uint64_t*)(s4) +(i) ); \
+  d[5] = *( (const uint64_t*)(s5) +(i) ); \
+  d[6] = *( (const uint64_t*)(s6) +(i) ); \
+  d[7] = *( (const uint64_t*)(s7) +(i) ); \
+} while(0)
+
+static inline void intrlv_8x64( void *dst, const void *s0,
+        const void *s1, const void *s2, const void *s3, const void *s4,
+        const void *s5, const void *s6, const void *s7, int bit_len )
+{
+   ILEAVE_8x64(  0 );   ILEAVE_8x64(  1 );
+   ILEAVE_8x64(  2 );   ILEAVE_8x64(  3 );
+   if ( bit_len <= 256 ) return;
+   ILEAVE_8x64(  4 );   ILEAVE_8x64(  5 );
+   ILEAVE_8x64(  6 );   ILEAVE_8x64(  7 );
+   if ( bit_len <= 512 ) return;
+   ILEAVE_8x64(  8 );   ILEAVE_8x64(  9 );
+   if ( bit_len <= 640 ) return;
+   ILEAVE_8x64( 10 );   ILEAVE_8x64( 11 );
+   ILEAVE_8x64( 12 );   ILEAVE_8x64( 13 );
+   ILEAVE_8x64( 14 );   ILEAVE_8x64( 15 );
+}
+
+#undef ILEAVE_8x64
+
+#define DLEAVE_8x64( i ) do \
+{ \
+   const uint64_t *s = (const uint64_t*)(src) + ( (i) << 3 ); \
+   *( (uint64_t*)(d0) +(i) ) = s[0]; \
+   *( (uint64_t*)(d1) +(i) ) = s[1]; \
+   *( (uint64_t*)(d2) +(i) ) = s[2]; \
+   *( (uint64_t*)(d3) +(i) ) = s[3]; \
+   *( (uint64_t*)(d4) +(i) ) = s[4]; \
+   *( (uint64_t*)(d5) +(i) ) = s[5]; \
+   *( (uint64_t*)(d6) +(i) ) = s[6]; \
+   *( (uint64_t*)(d7) +(i) ) = s[7]; \
+} while(0)
+
+static inline void dintrlv_8x64( void *d0, void *d1, void *d2, void *d3,
+      void *d4, void *d5, void *d6, void *d7, const void *src, int bit_len )
+{
+   DLEAVE_8x64(  0 );   DLEAVE_8x64(  1 );
+   DLEAVE_8x64(  2 );   DLEAVE_8x64(  3 );
+   if ( bit_len <= 256 ) return;
+   DLEAVE_8x64(  4 );   DLEAVE_8x64(  5 );
+   DLEAVE_8x64(  6 );   DLEAVE_8x64(  7 );
+   if ( bit_len <= 512 ) return;
+   DLEAVE_8x64(  8 );   DLEAVE_8x64(  9 );
+   if ( bit_len <= 640 ) return;
+   DLEAVE_8x64( 10 );   DLEAVE_8x64( 11 );
+   DLEAVE_8x64( 12 );   DLEAVE_8x64( 13 );
+   DLEAVE_8x64( 14 );   DLEAVE_8x64( 15 );
+}
+
+#undef DLEAVE_8x64
+
+static inline void extr_lane_8x64( void *d, const void *s,
+                                   const int lane, const int bit_len )
+{
+   ((uint64_t*)d)[ 0] = ((uint64_t*)s)[ lane     ];
+   ((uint64_t*)d)[ 1] = ((uint64_t*)s)[ lane+  8 ];
+   ((uint64_t*)d)[ 2] = ((uint64_t*)s)[ lane+ 16 ];
+   ((uint64_t*)d)[ 3] = ((uint64_t*)s)[ lane+ 24 ];
+   ((uint64_t*)d)[ 4] = ((uint64_t*)s)[ lane+ 32 ];
+   ((uint64_t*)d)[ 5] = ((uint64_t*)s)[ lane+ 40 ];
+   ((uint64_t*)d)[ 6] = ((uint64_t*)s)[ lane+ 48 ];
+   ((uint64_t*)d)[ 7] = ((uint64_t*)s)[ lane+ 56 ];
+   if ( bit_len <= 256 ) return;
+   ((uint64_t*)d)[ 8] = ((uint64_t*)s)[ lane+ 64 ];
+   ((uint64_t*)d)[ 9] = ((uint64_t*)s)[ lane+ 72 ];
+   ((uint64_t*)d)[10] = ((uint64_t*)s)[ lane+ 80 ];
+   ((uint64_t*)d)[11] = ((uint64_t*)s)[ lane+ 88 ];
+   ((uint64_t*)d)[12] = ((uint64_t*)s)[ lane+ 96 ];
+   ((uint64_t*)d)[13] = ((uint64_t*)s)[ lane+104 ];
+   ((uint64_t*)d)[14] = ((uint64_t*)s)[ lane+112 ];
+   ((uint64_t*)d)[15] = ((uint64_t*)s)[ lane+120 ];
+}
+
+#if defined(__AVX512F__) && defined(__AVX512VL__)
+
+static inline void mm512_bswap32_intrlv80_8x64( void *dst, void *src )
+{
+   __m512i *d = (__m512i*)dst;
+   __m512i s0 = mm512_bswap_32( casti_m512i(src, 0 ) );
+   __m128i s1 = mm128_bswap_32( casti_m128i(src, 4 ) );
+  const __m512i zero = m512_zero;
+  const __m512i one  = m512_one_64;
+  const __m512i two  = _mm512_add_epi64( one, one );
+  const __m512i tre  = _mm512_add_epi64( two, one );
+  const __m512i four = _mm512_add_epi64( two, two );
+
+  d[0] = _mm512_permutexvar_epi64( s0, zero );
+  d[1] = _mm512_permutexvar_epi64( s0, one  );
+  d[2] = _mm512_permutexvar_epi64( s0, two  );
+  d[3] = _mm512_permutexvar_epi64( s0, tre  );
+  d[4] = _mm512_permutexvar_epi64( s0, four );
+  d[5] = _mm512_permutexvar_epi64( s0, _mm512_add_epi64( four, one ) );
+  d[6] = _mm512_permutexvar_epi64( s0, _mm512_add_epi64( four, two ) );
+  d[7] = _mm512_permutexvar_epi64( s0, _mm512_add_epi64( four, tre ) );
+  d[8] = _mm512_permutexvar_epi64(
+           _mm512_castsi128_si512( s1 ), zero );
+  d[9] = _mm512_permutexvar_epi64(
+           _mm512_castsi128_si512( s1 ), one  );
+}
+
+#endif  // AVX512
+
+//////////////////////////
+//
+//      128 bit data
+
+// 2x128  (AVX2)
+
+static inline void intrlv_2x128( void *dst, const void *src0,
+                                 const void *src1, int bit_len )
+{
+   __m128i *d = (__m128i*)dst;
+   const __m128i *s0 = (const __m128i*)src0;
+   const __m128i *s1 = (const __m128i*)src1;
+   d[ 0] = s0[0];   d[ 1] = s1[0];
+   d[ 2] = s0[1];   d[ 3] = s1[1];
+   if ( bit_len <= 256 ) return;
+   d[ 4] = s0[2];   d[ 5] = s1[2];
+   d[ 6] = s0[3];   d[ 7] = s1[3];
+   if ( bit_len <= 512 ) return;
+   d[ 8] = s0[4];   d[ 9] = s1[4];
+   if ( bit_len <= 640 ) return;
+   d[10] = s0[5];   d[11] = s1[5];
+   d[12] = s0[6];   d[13] = s1[6];
+   d[14] = s0[7];   d[15] = s1[7];
+}
+
+static inline void intrlv_2x128_512( void *dst, const void *src0,
+                                     const void *src1 )
+{
+   __m128i *d = (__m128i*)dst;
+   const __m128i *s0 = (const __m128i*)src0;
+   const __m128i *s1 = (const __m128i*)src1;
+   d[0] = s0[0];   d[1] = s1[0];
+   d[2] = s0[1];   d[3] = s1[1];
+   d[4] = s0[2];   d[5] = s1[2];
+   d[6] = s0[3];   d[7] = s1[3];
+}
+
+static inline void dintrlv_2x128( void *dst0, void *dst1,
+                                  const void *src, int bit_len )
+{
+   __m128i *d0 = (__m128i*)dst0;
+   __m128i *d1 = (__m128i*)dst1;
+   const __m128i *s = (const __m128i*)src;
+
+   d0[0] = s[ 0];   d1[0] = s[ 1];
+   d0[1] = s[ 2];   d1[1] = s[ 3];
+   if ( bit_len <= 256 ) return;
+   d0[2] = s[ 4];   d1[2] = s[ 5];
+   d0[3] = s[ 6];   d1[3] = s[ 7];
+   if ( bit_len <= 512 ) return;
+   d0[4] = s[ 8];   d1[4] = s[ 9];
+   if ( bit_len <= 640 ) return;
+   d0[5] = s[10];   d1[5] = s[11];
+   d0[6] = s[12];   d1[6] = s[13];
+   d0[7] = s[14];   d1[7] = s[15];
+}
+
+static inline void dintrlv_2x128_512( void *dst0, void *dst1, const void *src )
+{
+   __m128i *d0 = (__m128i*)dst0;   
+   __m128i *d1 = (__m128i*)dst1;
+   const __m128i *s = (const __m128i*)src;
+
+   d0[0] = s[0];   d1[0] = s[1];
+   d0[1] = s[2];   d1[1] = s[3];
+   d0[2] = s[4];   d1[2] = s[5];
+   d0[3] = s[6];   d1[3] = s[7];
+}
+
+// 4x128  (AVX512)
+
+static inline void intrlv_4x128( void *dst, const void *src0,
+           const void *src1, const void *src2, const void *src3, int bit_len )
+{
+   __m128i *d = (__m128i*)dst;
+   const __m128i *s0 = (const __m128i*)src0;
+   const __m128i *s1 = (const __m128i*)src1;
+   const __m128i *s2 = (const __m128i*)src2;
+   const __m128i *s3 = (const __m128i*)src3;
+   d[ 0] = s0[0];    d[ 1] = s1[0];    d[ 2] = s2[0];    d[ 3] = s3[0];
+   d[ 4] = s0[1];    d[ 5] = s1[1];    d[ 6] = s2[1];    d[ 7] = s3[1];
+   if ( bit_len <= 256 ) return;
+   d[ 8] = s0[2];    d[ 9] = s1[2];    d[10] = s2[2];    d[11] = s3[2];
+   d[12] = s0[3];    d[13] = s1[3];    d[14] = s2[3];    d[15] = s3[3];
+   if ( bit_len <= 512 ) return;
+   d[16] = s0[4];    d[17] = s1[4];    d[18] = s2[4];    d[19] = s3[4];
+   if ( bit_len <= 640 ) return;
+   d[20] = s0[5];    d[21] = s1[5];    d[22] = s2[5];    d[23] = s3[5];
+   d[24] = s0[6];    d[25] = s1[6];    d[26] = s2[6];    d[27] = s3[6];
+   d[28] = s0[7];    d[29] = s1[7];    d[30] = s2[7];    d[31] = s3[7];
+}
+
+static inline void intrlv_4x128_512( void *dst, const void *src0,
+                      const void *src1, const void *src2, const void *src3 )
+{
+   __m128i *d = (__m128i*)dst;
+   const __m128i *s0 = (const __m128i*)src0;
+   const __m128i *s1 = (const __m128i*)src1;
+   const __m128i *s2 = (const __m128i*)src2;
+   const __m128i *s3 = (const __m128i*)src3; 
+   d[ 0] = s0[0];    d[ 1] = s1[0];    d[ 2] = s2[0];    d[ 3] = s3[0];
+   d[ 4] = s0[1];    d[ 5] = s1[1];    d[ 6] = s2[1];    d[ 7] = s3[1];
+   d[ 8] = s0[2];    d[ 9] = s1[2];    d[10] = s2[2];    d[11] = s3[2];
+   d[12] = s0[3];    d[13] = s1[3];    d[14] = s2[3];    d[15] = s3[3];
+}
+
+static inline void dintrlv_4x128( void *dst0, void *dst1, void *dst2,
+                                  void *dst3, const void *src, int bit_len )
+{
+   __m128i *d0 = (__m128i*)dst0;
+   __m128i *d1 = (__m128i*)dst1;
+   __m128i *d2 = (__m128i*)dst2;
+   __m128i *d3 = (__m128i*)dst3;
+   const __m128i *s = (const __m128i*)src;
+   d0[0] = s[ 0];   d1[0] = s[ 1];    d2[0] = s[ 2];   d3[0] = s[ 3];
+   d0[1] = s[ 4];   d1[1] = s[ 5];    d2[1] = s[ 6];   d3[1] = s[ 7];
+   if ( bit_len <= 256 ) return;
+   d0[2] = s[ 8];   d1[2] = s[ 9];    d2[2] = s[10];   d3[2] = s[11];
+   d0[3] = s[12];   d1[3] = s[13];    d2[3] = s[14];   d3[3] = s[15];
+   if ( bit_len <= 512 ) return;
+   d0[4] = s[16];   d1[4] = s[17];    d2[4] = s[18];   d3[4] = s[19];
+   if ( bit_len <= 640 ) return;
+   d0[5] = s[20];   d1[5] = s[21];    d2[5] = s[22];   d3[5] = s[23];
+   d0[6] = s[24];   d1[6] = s[25];    d2[6] = s[26];   d3[6] = s[27];
+   d0[7] = s[28];   d1[7] = s[29];    d2[7] = s[30];   d3[7] = s[31];
+}
+
+static inline void dintrlv_4x128_512( void *dst0, void *dst1, void *dst2,
+                                      void *dst3, const void *src )
+{
+   __m128i *d0 = (__m128i*)dst0;
+   __m128i *d1 = (__m128i*)dst1;
+   __m128i *d2 = (__m128i*)dst2;
+   __m128i *d3 = (__m128i*)dst3;
+   const __m128i *s = (const __m128i*)src;
+   d0[0] = s[ 0];   d1[0] = s[ 1];    d2[0] = s[ 2];   d3[0] = s[ 3];
+   d0[1] = s[ 4];   d1[1] = s[ 5];    d2[1] = s[ 6];   d3[1] = s[ 7];
+   d0[2] = s[ 8];   d1[2] = s[ 9];    d2[2] = s[10];   d3[2] = s[11];
+   d0[3] = s[12];   d1[3] = s[13];    d2[3] = s[14];   d3[3] = s[15];
+}
+
+
+///////////////////////////
+//
+// Re-intereleaving
+
+// 4x64 -> 4x32
+
+#define RLEAVE_4x64_4x32( i ) do \
+{ \
+   uint32_t *d = (uint32_t*)dst + (i); \
+   const uint32_t *s = (const uint32_t*)src + (i); \
+   d[0] = s[0];   d[1] = s[2]; \
+   d[2] = s[4];   d[3] = s[6]; \
+   d[4] = s[1];   d[5] = s[3]; \
+   d[6] = s[5];   d[7] = s[7]; \
+} while(0)
+
+
+// Convert 4x64 byte (256 bit) vectors to 4x32 (128 bit) vectors for AVX
+// bit_len must be multiple of 64
+static inline void rintrlv_4x64_4x32( void *dst, void *src,
+                                            int  bit_len )
+{
+   RLEAVE_4x64_4x32(   0 );   RLEAVE_4x64_4x32(   8 );
+   RLEAVE_4x64_4x32(  16 );   RLEAVE_4x64_4x32(  24 );
+   if ( bit_len <= 256 ) return;
+   RLEAVE_4x64_4x32(  32 );   RLEAVE_4x64_4x32(  40 );
+   RLEAVE_4x64_4x32(  48 );   RLEAVE_4x64_4x32(  56 );
+   if ( bit_len <= 512 ) return;
+   RLEAVE_4x64_4x32(  64 );   RLEAVE_4x64_4x32(  72 );
+   RLEAVE_4x64_4x32(  80 );   RLEAVE_4x64_4x32(  88 );
+   RLEAVE_4x64_4x32(  96 );   RLEAVE_4x64_4x32( 104 );
+   RLEAVE_4x64_4x32( 112 );   RLEAVE_4x64_4x32( 120 );
+}
+
+#undef RLEAVE_4x64_4x32
+
+
+// 4x32 -> 4x64
+
+#define RLEAVE_4x32_4x64(i) do \
+{ \
+ uint32_t *d = (uint32_t*)dst + (i); \
+ const uint32_t *s = (const uint32_t*)src + (i); \
+ d[0] = s[0];  d[1] = s[4]; \
+ d[2] = s[1];  d[3] = s[5]; \
+ d[4] = s[2];  d[5] = s[6]; \
+ d[6] = s[3];  d[7] = s[7]; \
+} while(0)
+
+static inline void rintrlv_4x32_4x64( void *dst,
+                                      const void *src, int  bit_len )
+{
+  RLEAVE_4x32_4x64(   0 );  RLEAVE_4x32_4x64(   8 );
+  RLEAVE_4x32_4x64(  16 );  RLEAVE_4x32_4x64(  24 );
+  if ( bit_len <= 256 ) return;
+  RLEAVE_4x32_4x64(  32 );  RLEAVE_4x32_4x64(  40 );
+  RLEAVE_4x32_4x64(  48 );  RLEAVE_4x32_4x64(  56 );
+  if ( bit_len <= 512 ) return;
+  RLEAVE_4x32_4x64(  64 );  RLEAVE_4x32_4x64(  72 );
+  RLEAVE_4x32_4x64(  80 );  RLEAVE_4x32_4x64(  88 );
+  RLEAVE_4x32_4x64(  96 );  RLEAVE_4x32_4x64( 104 );
+  RLEAVE_4x32_4x64( 112 );  RLEAVE_4x32_4x64( 120 );
+}
+
+#undef RLEAVE_4x32_4x64
+
+
+// 2x128 -> 4x64
+
+#define RLEAVE_2x128_4x64( i ) do \
+{ \
+   uint64_t *d = (uint64_t*)dst + ((i)<<1); \
+   const uint64_t *s0 = (const uint64_t*)src0 + (i); \
+   const uint64_t *s1 = (const uint64_t*)src1 + (i); \
+   d[0] = s0[0];    d[1] = s0[2]; \
+   d[2] = s1[0];    d[3] = s1[2]; \
+   d[4] = s0[1];    d[5] = s0[3]; \
+   d[6] = s1[1];    d[7] = s1[3]; \
+} while(0)
+
+static inline void rintrlv_2x128_4x64( void *dst, const void *src0,
+                                         const void *src1, int  bit_len )
+{
+   RLEAVE_2x128_4x64(  0 );   RLEAVE_2x128_4x64(  4 );
+   if ( bit_len <= 256 ) return;
+   RLEAVE_2x128_4x64(  8 );   RLEAVE_2x128_4x64( 12 );
+   if ( bit_len <= 512 ) return;
+   RLEAVE_2x128_4x64( 16 );   RLEAVE_2x128_4x64( 20 );
+   RLEAVE_2x128_4x64( 24 );   RLEAVE_2x128_4x64( 28 );
+}
+
+#undef RLEAVE_2x128_4x64
+
+
+// 4x64 -> 2x128
+
+#define RLEAVE_4x64_2x128( i ) do \
+{ \
+   uint64_t *d0 = (uint64_t*)dst0 + (i); \
+   uint64_t *d1 = (uint64_t*)dst1 + (i); \
+   const uint64_t *s = (const uint64_t*)src + ((i)<<1); \
+   d0[0] = s[0];   d0[1] = s[4]; \
+   d0[2] = s[1];   d0[3] = s[5]; \
+   d1[0] = s[2];   d1[1] = s[6]; \
+   d1[2] = s[3];   d1[3] = s[7]; \
+} while(0)
+
+static inline void rintrlv_4x64_2x128( void *dst0, void *dst1,
+                                       const void *src, int bit_len )
+{
+   RLEAVE_4x64_2x128(  0 );   RLEAVE_4x64_2x128(  4 );
+   if ( bit_len <= 256 ) return;
+   RLEAVE_4x64_2x128(  8 );   RLEAVE_4x64_2x128( 12 );
+   if ( bit_len <= 512 ) return;
+   RLEAVE_4x64_2x128( 16 );   RLEAVE_4x64_2x128( 20 );
+   RLEAVE_4x64_2x128( 24 );   RLEAVE_4x64_2x128( 28 );
+}
+
+
+//
+// Some functions customized for mining.
+
+// 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 )
+
+#endif   // SSE4_1
+
+#if defined(__AVX2__)
+
+#define mm256_intrlv_blend_128( hi, lo ) \
+                _mm256_blend_epi32( hi, lo, 0x0f )
+
+#define mm256_intrlv_blend_64( hi, lo ) \
+                _mm256_blend_epi32( hi, lo, 0x33 )
+
+#define mm256_intrlv_blend_32( hi, lo ) \
+           _mm256_blend_epi32( hi, lo, 0x55 )
+
+// Blend 32 byte lanes of hash from 2 sources according to control mask.
+// macro due to 256 bit value arg.
+#define mm256_blend_hash_4x64( dst, a, b, mask ) \
+do { \
+    dst[0] = _mm256_blendv_epi8( a[0], b[0], mask ); \
+    dst[1] = _mm256_blendv_epi8( a[1], b[1], mask ); \
+    dst[2] = _mm256_blendv_epi8( a[2], b[2], mask ); \
+    dst[3] = _mm256_blendv_epi8( a[3], b[3], mask ); \
+    dst[4] = _mm256_blendv_epi8( a[4], b[4], mask ); \
+    dst[5] = _mm256_blendv_epi8( a[5], b[5], mask ); \
+    dst[6] = _mm256_blendv_epi8( a[6], b[6], mask ); \
+    dst[7] = _mm256_blendv_epi8( a[7], b[7], mask ); \
+} while(0)
+
+#endif  // AVX2
+
+#endif // INTERLEAVE_H__
diff --git a/simd-utils/simd-128.h b/simd-utils/simd-128.h
index a1d3cd4..82c0dc5 100644
--- a/simd-utils/simd-128.h
+++ b/simd-utils/simd-128.h
@@ -44,7 +44,17 @@
 // repeatedly. It may be better for the application to reimplement the
 // utility to better suit its usage.
 //
-
+// More tips:
+//
+// Conversions from integer to vector should be avoided whenever possible.
+// Extract, insert and set and set1 instructions should be avoided.
+// In addition to the issues with constants set is also very inefficient with
+// variables.
+// Converting integer data to perform a couple of vector operations
+// then converting back to integer should be avoided. Converting data in
+// registers should also be avoided. Conversion should be limited to buffers
+// in memory where the data is loaded directly to vector registers, bypassing
+// the integer to vector conversion.
 //
 // Pseudo constants.
 //
@@ -71,7 +81,7 @@ static inline __m128i m128_one_64_fn()
   asm( "pxor %0, %0\n\t"
        "pcmpeqd %%xmm1, %%xmm1\n\t"
        "psubq %%xmm1, %0\n\t"
-       :"=x"(a)
+       : "=x"(a)
        :
        : "xmm1" );
   return a;
@@ -84,7 +94,7 @@ static inline __m128i m128_one_32_fn()
   asm( "pxor %0, %0\n\t"
        "pcmpeqd %%xmm1, %%xmm1\n\t"
        "psubd %%xmm1, %0\n\t"
-       :"=x"(a)
+       : "=x"(a)
        :
        : "xmm1" );
   return a;
@@ -97,7 +107,7 @@ static inline __m128i m128_one_16_fn()
   asm( "pxor %0, %0\n\t"
        "pcmpeqd %%xmm1, %%xmm1\n\t"
        "psubw %%xmm1, %0\n\t"
-       :"=x"(a)
+       : "=x"(a)
        :
        : "xmm1" );
   return a;
@@ -110,7 +120,7 @@ static inline __m128i m128_one_8_fn()
   asm( "pxor %0, %0\n\t"
        "pcmpeqd %%xmm1, %%xmm1\n\t"
        "psubb %%xmm1, %0\n\t"
-       :"=x"(a)
+       : "=x"(a)
        :
        : "xmm1" );
   return a;
@@ -121,7 +131,7 @@ static inline __m128i m128_neg1_fn()
 {
    __m128i a;
    asm( "pcmpeqd %0, %0\n\t"
-        :"=x"(a) );
+        : "=x"(a) );
    return a;
 }
 #define m128_neg1    m128_neg1_fn()
@@ -133,7 +143,7 @@ static inline __m128i m128_one_128_fn()
    __m128i a;
    asm( "pinsrq $0, $1, %0\n\t"
         "pinsrq $1, $0, %0\n\t"
-        :"=x"(a) );
+        : "=x"(a) );
    return a;
 }
 #define m128_one_128    m128_one_128_fn()
@@ -145,8 +155,8 @@ static inline __m128i m128_const_64( uint64_t hi, uint64_t lo )
    __m128i a;
    asm( "pinsrq $0, %2, %0\n\t"
         "pinsrq $1, %1, %0\n\t"
-        :"=x"(a)
-        :"r"(hi),"r"(lo) );
+        : "=x"(a)
+        : "r"(hi), "r"(lo) );
    return a;
 } 
 
diff --git a/simd-utils/simd-512.h b/simd-utils/simd-512.h
index 41c9e3f..01e30bd 100644
--- a/simd-utils/simd-512.h
+++ b/simd-utils/simd-512.h
@@ -41,79 +41,6 @@
 // Experimental, not fully tested.
 
 
-//
-// Compile time vector constants and initializers.
-//
-// The following macro constants and functions should only be used
-// for compile time initialization of constant and variable vector
-// arrays. These constants use memory, use set instruction or pseudo
-// constants at run time to avoid using memory.
-
-// Constant initializers
-#define mm512_const_64( x7, x6, x5, x4, x3, x2, x1, x0 ) \
-                     {{ x7, x6, x5, x4, x3, x2, x1, x0 }}
-
-#define mm512_const1_64( x ) {{ x,x,x,x,x,x,x }}
-
-#define mm512_const_32( x15, x14, x13, x12, x11, x10, x09, x08, \
-                        x07, x06, x05, x04, x03, x02, x01, x00 ) \
-                     {{ x15, x14, x13, x12, x11, x10, x09, x08, }} \
-                        x07, x06, x05, x04, x03, x02, x01, x00 }}
-
-#define mm512_const1_32( x ) {{ x,x,x,x, x,x,x,x, x,x,x,x, x,x,x,x }}
-
-#define mm512_const_16( x31, x30, x29, x28, x27, x26, x25, x24, \
-                        x23, x22, x21, x20, x19, x18, x17, x16, \
-                        x15, x14, x13, x12, x11, x10, x09, x08, \
-                        x07, x06, x05, x04, x03, x02, x01, x00 ) \
-                     {{ x31, x30, x29, x28, x27, x26, x25, x24, \
-                        x23, x22, x21, x20, x19, x18, x17, x16, \
-                        x15, x14, x13, x12, x11, x10, x09, x08, \
-                        x07, x06, x05, x04, x03, x02, x01, x00 }}
-
-#define mm512_const1_16( x ) {{ x,x,x,x, x,x,x,x, x,x,x,x, x,x,x,x, \
-                                x,x,x,x, x,x,x,x, x,x,x,x, x,x,x,x }}
-
-#define mm512_const_8( x63, x62, x61, x60, x59, x58, x57, x56, \
-                       x55, x54, x53, x52, x51, x50, x49, x48, \
-                       x47, x46, x45, x44, x43, x42, x41, x40, \
-                       x39, x38, x37, x36, x35, x34, x33, x32, \
-                       x31, x30, x29, x28, x27, x26, x25, x24, \
-                       x23, x22, x21, x20, x19, x18, x17, x16, \
-                       x15, x14, x13, x12, x11, x10, x09, x08, \
-                       x07, x06, x05, x04, x03, x02, x01, x00 ) \
-                    {{ x63, x62, x61, x60, x59, x58, x57, x56, \
-                       x55, x54, x53, x52, x51, x50, x49, x48, \
-                       x47, x46, x45, x44, x43, x42, x41, x40, \
-                       x39, x38, x37, x36, x35, x34, x33, x32, \
-                       x31, x30, x29, x28, x27, x26, x25, x24, \
-                       x23, x22, x21, x20, x19, x18, x17, x16, \
-                       x15, x14, x13, x12, x11, x10, x09, x08, \
-                       x07, x06, x05, x04, x03, x02, x01, x00 }}
-
-#define mm512_const1_8( x ) {{ x,x,x,x, x,x,x,x, x,x,x,x, x,x,x,x, \
-                               x,x,x,x, x,x,x,x, x,x,x,x, x,x,x,x, \
-                               x,x,x,x, x,x,x,x, x,x,x,x, x,x,x,x, \
-                               x,x,x,x, x,x,x,x, x,x,x,x, x,x,x,x }}
-
-// Predefined compile time constant vectors.
-#define c512_zero      mm512_const1_64(   0ULL )
-#define c512_neg1      mm512_const1_64(   0xFFFFFFFFFFFFFFFFULL )
-#define c512_one_512   mm512_const_epi64( 0ULL, 0ULL, 0ULL, 0ULL, \
-                                          0ULL, 0ULL, 0ULL, 1ULL )
-#define c512_one_256   mm512_const_64(    0ULL, 0ULL, 0ULL, 1ULL, \
-                                          0ULL, 0ULL, 0ULL, 1ULL )
-#define c512_one_128   mm512_const_64(    0ULL, 1ULL, 0ULL, 1ULL, \
-                                          0ULL, 1ULL, 0ULL, 1ULL )
-#define c512_one_64    mm512_const1_64(   1ULL )
-#define c512_one_32    mm512_const1_32(   1UL )
-#define c512_one_16    mm512_const1_16(   1U )
-#define c512_one_8     mm512_const1_8(    1U )
-#define c512_neg1_64   mm512_const1_64( 0xFFFFFFFFFFFFFFFFULL )
-#define c512_neg1_32   mm512_const1_32( 0xFFFFFFFFUL )
-#define c512_neg1_16   mm512_const1_32( 0xFFFFU )
-#define c512_neg1_8    mm512_const1_32( 0xFFU )
-
 //
 // Pseudo constants.
 
@@ -127,11 +54,77 @@
                                            0ULL, 0ULL, 0ULL, 1ULL )
 #define m512_one_256    _mm512_set4_epi64( 0ULL, 0ULL, 0ULL, 1ULL )
 #define m512_one_128    _mm512_set4_epi64( 0ULL, 1ULL, 0ULL, 1ULL )
-#define m512_one_64     _mm512_set1_epi64( 1ULL )
-#define m512_one_32     _mm512_set1_epi32( 1UL )
-#define m512_one_16     _mm512_set1_epi16( 1U )
-#define m512_one_8      _mm512_set1_epi8(  1U )
-#define m512_neg1       _mm512_set1_epi64( 0xFFFFFFFFFFFFFFFFULL )
+//#define m512_one_64     _mm512_set1_epi64( 1ULL )
+//#define m512_one_32     _mm512_set1_epi32( 1UL )
+//#define m512_one_16     _mm512_set1_epi16( 1U )
+//#define m512_one_8      _mm512_set1_epi8(  1U )
+//#define m512_neg1       _mm512_set1_epi64( 0xFFFFFFFFFFFFFFFFULL )
+
+#define mi512_const_64( i7, i6, i5, i4, i3, i2, i1, i0 ) \
+  _mm512_inserti64x4( _mm512_castsi512_si256( m256_const_64( i3.i2,i1,i0 ) ), \
+                      m256_const_64( i7,i6,i5,i4 ), 1 )
+#define m512_const1_64( i ) m256_const_64( i, i, i, i, i, i, i, i )
+
+static inline __m512i m512_one_64_fn()
+{
+  __m512i a;
+  asm( "vpxorq %0, %0, %0\n\t"
+       "vpcmpeqd %%zmm1, %%zmm1, %%zmm1\n\t"
+       "vpsubq %%zmm1, %0, %0\n\t"
+       :"=x"(a)
+       :
+       : "zmm1" );
+  return a;
+}
+#define m512_one_64    m512_one_64_fn()
+
+static inline __m512i m512_one_32_fn()
+{
+  __m512i a;
+  asm( "vpxord %0, %0, %0\n\t"
+       "vpcmpeqd %%zmm1, %%zmm1, %%zmm1\n\t"
+       "vpsubd %%zmm1, %0, %0\n\t"
+       :"=x"(a)
+       :
+       : "zmm1" );
+  return a;
+}
+#define m512_one_32    m512_one_32_fn()
+
+static inline __m512i m512_one_16_fn()
+{
+  __m512i a;
+  asm( "vpxord %0, %0, %0\n\t"
+       "vpcmpeqd %%zmm1, %%zmm1, %%zmm1\n\t"
+       "vpsubw %%zmm1, %0, %0\n\t"
+       :"=x"(a)
+       :
+       : "zmm1" );
+  return a;
+}
+#define m512_one_16    m512_one_16_fn()
+
+static inline __m512i m512_one_8_fn()
+{
+  __m512i a;
+  asm( "vpxord %0, %0, %0\n\t"
+       "vpcmpeqd %%zmm1, %%zmm1, %%zmm1\n\t"
+       "vpsubb %%zmm1, %0, %0\n\t"
+       :"=x"(a)
+       :
+       : "zmm1" );
+  return a;
+}
+#define m512_one_8    m512_one_8_fn()
+
+static inline __m512i m512_neg1_fn()
+{
+   __m512i a;
+   asm( "vpcmpeqq %0, %0, %0\n\t"
+        :"=x"(a) );
+   return a;
+}
+#define m512_neg1    m512_neg1_fn()
 
 
 //
@@ -142,6 +135,15 @@
 #define mm512_negate_32( x ) _mm512_sub_epi32( m512_zero, x )  
 #define mm512_negate_16( x ) _mm512_sub_epi16( m512_zero, x )  
 
+
+
+#define mm256_extr_lo256_512( a ) _mm512_castsi512_si256( a )
+#define mm256_extr_hi256_512( a ) _mm512_extracti64x4_epi64( a, 1 )
+
+#define mm128_extr_lo128_512( a ) _mm512_castsi512_si256( a )
+
+
+
 //
 // Pointer casting
 
@@ -225,6 +227,25 @@
    *((uint32_t*)(d06)) = ((uint32_t*)(s))[14]; \
    *((uint32_t*)(d07)) = ((uint32_t*)(s))[15];
 
+// Add 4 values, fewer dependencies than sequential addition.
+
+
+#define mm512_add4_64( a, b, c, d ) \
+   _mm512_add_epi64( _mm512_add_epi64( a, b ), _mm512_add_epi64( c, d ) )
+
+#define mm512_add4_32( a, b, c, d ) \
+   _mm512_add_epi32( _mm512_add_epi32( a, b ), _mm512_add_epi32( c, d ) )
+
+#define mm512_add4_16( a, b, c, d ) \
+   _mm512_add_epi16( _mm512_add_epi16( a, b ), _mm512_add_epi16( c, d ) )
+
+#define mm512_add4_8( a, b, c, d ) \
+   _mm512_add_epi8( _mm512_add_epi8( a, b ), _mm512_add_epi8( c, d ) )
+
+#define mm512_xor4( a, b, c, d ) \
+   _mm512_xor_si512( _mm512_xor_si256( a, b ), _mm512_xor_si256( c, d ) )
+
+
 
 //
 // Bit rotations.
@@ -263,45 +284,41 @@
 #define mm512_ror_x32( v, n )      _mm512_alignr_epi32( v, v, n )
 
 
-//  Although documented to exist in AVX512F the _mm512_set_epi8 &
-//  _mm512_set_epi16 intrinsics fail to compile. Seems usefull to have
-//  for endian byte swapping. Workaround by using _mm512_set_epi32.
-//  Ugly but it works.
 
 #define mm512_ror_1x16( v ) \
-   _mm512_permutexvar_epi16( v, _mm512_set_epi32( \
-                       0x0000001F, 0x001E001D, 0x001C001B, 0x001A0019, \
-                       0X00180017, 0X00160015, 0X00140013, 0X00120011, \
-                       0X0010000F, 0X000E000D, 0X000C000B, 0X000A0009, \
-                       0X00080007, 0X00060005, 0X00040003, 0X00020001 ) )
+   _mm512_permutexvar_epi16( v, m512_const_64( \
+                       0x0000001F001E001D, 0x001C001B001A0019, \
+                       0X0018001700160015, 0X0014001300120011, \
+                       0X0010000F000E000D, 0X000C000B000A0009, \
+                       0X0008000700060005, 0X0004000300020001 ) )
 
 #define mm512_rol_1x16( v ) \
-   _mm512_permutexvar_epi16( v, _mm512_set_epi32( \
-                       0x001E001D, 0x001C001B, 0x001A0019, 0x00180017, \
-                       0X00160015, 0X00140013, 0X00120011, 0x0010000F, \
-                       0X000E000D, 0X000C000B, 0X000A0009, 0X00080007, \
-                       0X00060005, 0X00040003, 0X00020001, 0x0000001F ) )
+   _mm512_permutexvar_epi16( v, m512_const_64( \
+                       0x001E001D001C001B, 0x001A001900180017, \
+                       0X0016001500140013, 0X001200110010000F, \
+                       0X000E000D000C000B, 0X000A000900080007, \
+                       0X0006000500040003, 0X000200010000001F ) )
 
 
 #define mm512_ror_1x8( v ) \
-   _mm512_permutexvar_epi8( v, _mm512_set_epi32( \
-                       0x003F3E3D, 0x3C3B3A39, 0x38373635, 0x34333231, \
-                       0x302F2E2D, 0x2C2B2A29, 0x28272625, 0x24232221, \
-                       0x201F1E1D, 0x1C1B1A19. 0x18171615, 0x14131211, \
-                       0x100F0E0D, 0x0C0B0A09, 0x08070605, 0x04030201 ) )
+   _mm512_permutexvar_epi8( v, m512_const_64( \
+                       0x003F3E3D3C3B3A39, 0x3837363534333231, \
+                       0x302F2E2D2C2B2A29, 0x2827262524232221, \
+                       0x201F1E1D1C1B1A19. 0x1817161514131211, \
+                       0x100F0E0D0C0B0A09, 0x0807060504030201 ) )
 
 #define mm512_rol_1x8( v ) \
-   _mm512_permutexvar_epi8( v, _mm512_set_epi32( \
-                       0x3E3D3C3B, 0x3A393837, 0x36353433, 0x3231302F. \
-                       0x2E2D2C2B, 0x2A292827, 0x26252423, 0x2221201F, \
-                       0x1E1D1C1B, 0x1A191817, 0x16151413, 0x1211100F, \
-                       0x0E0D0C0B, 0x0A090807, 0x06050403, 0x0201003F ) )
+   _mm512_permutexvar_epi8( v, m512_const_64( \
+                       0x3E3D3C3B3A393837, 0x363534333231302F. \
+                       0x2E2D2C2B2A292827, 0x262524232221201F, \
+                       0x1E1D1C1B1A191817, 0x161514131211100F, \
+                       0x0E0D0C0B0A090807, 0x060504030201003F ) )
 
 // Invert vector: {3,2,1,0} -> {0,1,2,3}
 #define mm512_invert_128( v ) _mm512_permute4f128_epi32( a, 0x1b )
 
 #define mm512_invert_64( v ) \
-   _mm512_permutex_epi64( v, _mm512_set_epi64( 0,1,2,3,4,5,6,7 ) )
+   _mm512_permutex_epi64( v, m512_const_64( 0,1,2,3,4,5,6,7 ) )
 
 #define mm512_invert_32( v ) \
    _mm512_permutexvar_epi32( v, _mm512_set_epi32( \
@@ -378,32 +395,32 @@
 #define mm512_rol1x32_128( v )   _mm512_shuffle_epi32( v, 0x93 )
 
 #define mm512_ror1x16_128( v ) \
-    _mm512_permutexvar_epi16( v, _mm512_set_epi32( \
-                     0x0018001F, 0x001E001D, 0x001C001B, 0x001A0019, \
-                     0x00100017, 0x00160015, 0x00140013, 0x00120011, \
-                     0x0008000F, 0x000E000D, 0x000C000B, 0x000A0009, \
-                     0x00000007, 0x00060005, 0x00040003, 0x00020001 ) )
+    _mm512_permutexvar_epi16( v, m512_const_64( \
+                     0x0018001F001E001D, 0x001C001B001A0019, \
+                     0x0010001700160015, 0x0014001300120011, \
+                     0x0008000F000E000D, 0x000C000B000A0009, \
+                     0x0000000700060005, 0x0004000300020001 ) )
 
 #define mm512_rol1x16_128( v ) \
-    _mm512_permutexvar_epi16( v, _mm512_set_epi32( \
-                     0x001E001D, 0x001C001B, 0x001A0019, 0x0018001F, \
-                     0x00160015, 0x00140013, 0x00120011, 0x00100017, \
-                     0x000E000D, 0x000C000B, 0x000A0009, 0x0008000F, \
-                     0x00060005, 0x00040003, 0x00020001, 0x00000007 ) )
+    _mm512_permutexvar_epi16( v, m512_const_64( \
+                     0x001E001D001C001B, 0x001A00190018001F, \
+                     0x0016001500140013, 0x0012001100100017, \
+                     0x000E000D000C000B, 0x000A00090008000F, \
+                     0x0006000500040003, 0x0002000100000007 ) )
 
 #define mm512_ror1x8_128( v ) \
-    _mm512_permutexvar_epi8( v, _mm512_set_epi32( \
-                     0x303F3E3D, 0x3C3B3A39, 0x38373635, 0x34333231, \
-                     0x202F2E2D, 0x2C2B2A29, 0x28272625, 0x24232221, \
-                     0x101F1E1D, 0x1C1B1A19, 0x18171615, 0x14131211, \
-                     0x000F0E0D, 0x0C0B0A09, 0x08070605, 0x04030201 ) )
+    _mm512_permutexvar_epi8( v, m512_const_64( \
+                     0x303F3E3D3C3B3A39, 0x3837363534333231, \
+                     0x202F2E2D2C2B2A29, 0x2827262524232221, \
+                     0x101F1E1D1C1B1A19, 0x1817161514131211, \
+                     0x000F0E0D0C0B0A09, 0x0807060504030201 ) )
 
 #define mm512_rol1x8_128( v ) \
-    _mm512_permutexvar_epi8( v, _mm512_set_epi32( \
-                     0x3E3D3C3B, 0x3A393837, 0x36353433. 0x3231303F, \
-                     0x2E2D2C2B, 0x2A292827, 0x26252423, 0x2221202F, \
-                     0x1E1D1C1B, 0x1A191817, 0x16151413, 0x1211101F, \
-                     0x0E0D0C0B, 0x0A090807, 0x06050403, 0x0201000F ) )
+    _mm512_permutexvar_epi8( v, m512_const_64( \
+                     0x3E3D3C3B3A393837, 0x363534333231303F, \
+                     0x2E2D2C2B2A292827, 0x262524232221202F, \
+                     0x1E1D1C1B1A191817, 0x161514131211101F, \
+                     0x0E0D0C0B0A090807, 0x060504030201000F ) )
 
 // Rotate 128 bit lanes by c bytes.  
 #define mm512_bror_128( v, c ) \