v3.16.2

2025-09-17 23:44:27 +00:00 · 2021-04-08 18:09:31 -04:00
parent 902ec046dd
commit f3333b0070
17 changed files with 826 additions and 336 deletions
--- a/Makefile.am
+++ b/Makefile.am
@@ -196,6 +196,7 @@ cpuminer_SOURCES = \
  algo/verthash/Verthash.c \
  algo/verthash/fopen_utf8.c \
  algo/verthash/tiny_sha3/sha3.c \
+  algo/verthash/tiny_sha3/sha3-4way.c \
  algo/whirlpool/sph_whirlpool.c \
  algo/whirlpool/whirlpool-hash-4way.c \
  algo/whirlpool/whirlpool-gate.c \
--- a/README.md
+++ b/README.md
@@ -135,7 +135,7 @@ Supported Algorithms
                          x14           X14
                          x15           X15
                          x16r          
-                          x16rv2        Ravencoin (RVN)
+                          x16rv2        
                          x16rt         Gincoin (GIN)
                          x16rt-veil    Veil (VEIL)
                          x16s          Pigeoncoin (PGN)
--- a/12
+++ b/12
@@ -65,6 +65,14 @@ If not what makes it happen or not happen?
 Change Log
 ----------

+v3.16.2
+
+Verthash: midstate prehash optimization for all architectures.
+Verthash: AVX2 optimization.
+GBT: added support for Bech32 addresses, untested.
+Linux: added CPU frequency to benchmark log.
+Fixed integer overflow in time calculations.
+
 v3.16.1

 New options for verthash:
@@ -72,16 +80,12 @@ New options for verthash:
              data file, default is "verthash.dat" in the current directory.
  --verify to perform the data file integrity check at startup, default is
           not to verify data file integrity.
-
 Support for creation of default verthash data file if:
   1) --data-file option is not used,
   2) no default data file is found in the current directory, and,
   3) --verify option is used.
-
 More detailed logs related to verthash data file.
-
 Small verthash performance improvement.
-
 Fixed detection of corrupt stats caused by networking issues.

 v3.16.0
--- a/algo/verthash/Verthash.c
+++ b/algo/verthash/Verthash.c
@@ -134,87 +134,117 @@ static inline uint32_t fnv1a(const uint32_t a, const uint32_t b)
    return (a ^ b) * 0x1000193;
 }

-void verthash_hash(const unsigned char* blob_bytes,
-                   const size_t blob_size,
-                   const unsigned char(*input)[VH_HEADER_SIZE],
-                   unsigned char(*output)[VH_HASH_OUT_SIZE])
+void verthash_hash( const unsigned char* blob_bytes,
+                    const size_t blob_size,
+                    const unsigned char(*input)[VH_HEADER_SIZE],
+                    unsigned char(*output)[VH_HASH_OUT_SIZE] )
 {
-    unsigned char p1[VH_HASH_OUT_SIZE] __attribute__ ((aligned (64)));
-    sha3(&input[0], VH_HEADER_SIZE, &p1[0], VH_HASH_OUT_SIZE);
-
-    unsigned char p0[VH_N_SUBSET];
-
-    unsigned char input_header[VH_HEADER_SIZE] __attribute__ ((aligned (64)));
-    memcpy(input_header, input, VH_HEADER_SIZE);
-
-    for (size_t i = 0; i < VH_N_ITER; ++i)
-    {
-        input_header[0] += 1;
-        sha3(&input_header[0], VH_HEADER_SIZE, p0 + i * VH_P0_SIZE, VH_P0_SIZE);
-    }
-
-    uint32_t* p0_index = (uint32_t*)p0;
+    unsigned char p1[ VH_HASH_OUT_SIZE ] __attribute__ ((aligned (64)));
+    unsigned char p0[ VH_N_SUBSET ] __attribute__ ((aligned (64)));
    uint32_t seek_indexes[VH_N_INDEXES] __attribute__ ((aligned (64)));
+    uint32_t* p0_index = (uint32_t*)p0;
+
+    verthash_sha3_512_final_8( p0, ( (uint64_t*)input )[ 9 ] );
    
    for ( size_t x = 0; x < VH_N_ROT; ++x )
    {
        memcpy( seek_indexes + x * (VH_N_SUBSET / sizeof(uint32_t)),
                p0, VH_N_SUBSET);

-//#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)        
-// 512 bit vector processing is actually slower because it reduces the CPU
-// clock significantly, which also slows mem access. The AVX512 rol instruction
-// is still available for smaller vectors.
-
-//        for ( size_t y = 0; y < VH_N_SUBSET / sizeof(uint32_t); y += 16 )
-//        {
-//            __m512i *p0_v = (__m512i*)( p0_index + y );
-//            *p0_v = mm512_rol_32( *p0_v, 1 );
-//        }
-
 #if defined(__AVX2__)

-        for ( size_t y = 0; y < VH_N_SUBSET / sizeof(uint32_t); y += 8 )
+        for ( size_t y = 0; y < VH_N_SUBSET / sizeof(__m256i); y += 8)
        {
-            __m256i *p0_v = (__m256i*)( p0_index + y );
-            *p0_v = mm256_rol_32( *p0_v, 1 );
+           casti_m256i( p0_index, y   ) = mm256_rol_32(
+                                            casti_m256i( p0_index, y   ), 1 );
+           casti_m256i( p0_index, y+1 ) = mm256_rol_32( 
+                                            casti_m256i( p0_index, y+1 ), 1 );
+           casti_m256i( p0_index, y+2 ) = mm256_rol_32(
+                                            casti_m256i( p0_index, y+2 ), 1 );
+           casti_m256i( p0_index, y+3 ) = mm256_rol_32(
+                                            casti_m256i( p0_index, y+3 ), 1 );
+           casti_m256i( p0_index, y+4 ) = mm256_rol_32(
+                                            casti_m256i( p0_index, y+4 ), 1 );
+           casti_m256i( p0_index, y+5 ) = mm256_rol_32(
+                                            casti_m256i( p0_index, y+5 ), 1 );
+           casti_m256i( p0_index, y+6 ) = mm256_rol_32(
+                                            casti_m256i( p0_index, y+6 ), 1 );
+           casti_m256i( p0_index, y+7 ) = mm256_rol_32(
+                                            casti_m256i( p0_index, y+7 ), 1 );
        }

 #else

-        for ( size_t y = 0; y < VH_N_SUBSET / sizeof(uint32_t); y += 4 )
+        for ( size_t y = 0; y < VH_N_SUBSET / sizeof(__m128i); y += 8)
        {
-            __m128i *p0_v = (__m128i*)( p0_index + y );
-            *p0_v = mm128_rol_32( *p0_v, 1 );
+           casti_m128i( p0_index, y   ) = mm128_rol_32(
+                                            casti_m128i( p0_index, y   ), 1 );
+           casti_m128i( p0_index, y+1 ) = mm128_rol_32(
+                                            casti_m128i( p0_index, y+1 ), 1 );
+           casti_m128i( p0_index, y+2 ) = mm128_rol_32(
+                                            casti_m128i( p0_index, y+2 ), 1 );
+           casti_m128i( p0_index, y+3 ) = mm128_rol_32(
+                                            casti_m128i( p0_index, y+3 ), 1 );
+           casti_m128i( p0_index, y+4 ) = mm128_rol_32(
+                                            casti_m128i( p0_index, y+4 ), 1 );
+           casti_m128i( p0_index, y+5 ) = mm128_rol_32(
+                                            casti_m128i( p0_index, y+5 ), 1 );
+           casti_m128i( p0_index, y+6 ) = mm128_rol_32(
+                                            casti_m128i( p0_index, y+6 ), 1 );
+           casti_m128i( p0_index, y+7 ) = mm128_rol_32(
+                                            casti_m128i( p0_index, y+7 ), 1 );
        }
        
 #endif

-//        for (size_t y = 0; y < VH_N_SUBSET / sizeof(uint32_t); ++y)
-//        {
-//            *(p0_index + y) = ( *(p0_index + y) << 1 )
-//            | ( 1 & (*(p0_index + y) >> 31) );
-//        }
    }

+    sha3( &input[0], VH_HEADER_SIZE, &p1[0], VH_HASH_OUT_SIZE );
+    
    uint32_t* p1_32 = (uint32_t*)p1;
    uint32_t* blob_bytes_32 = (uint32_t*)blob_bytes;
    uint32_t value_accumulator = 0x811c9dc5;
-    const uint32_t mdiv = ((blob_size - VH_HASH_OUT_SIZE) / VH_BYTE_ALIGNMENT) + 1;
-    for (size_t i = 0; i < VH_N_INDEXES; i++)
+    const uint32_t mdiv = ( ( blob_size - VH_HASH_OUT_SIZE )
+                             / VH_BYTE_ALIGNMENT ) + 1;
+#if defined (__AVX2__)        
+    const __m256i k = _mm256_set1_epi32( 0x1000193 );
+#elif defined(__SSE41__)
+    const __m128i k = _mm_set1_epi32( 0x1000193 );
+#endif
+
+    for ( size_t i = 0; i < VH_N_INDEXES; i++ )
    {
-        const uint32_t offset = (fnv1a(seek_indexes[i], value_accumulator) % mdiv) * VH_BYTE_ALIGNMENT / sizeof(uint32_t);
+        const uint32_t offset =
+                      ( fnv1a( seek_indexes[i], value_accumulator) % mdiv )
+                      * ( VH_BYTE_ALIGNMENT / sizeof(uint32_t) );
        const uint32_t *blob_off = blob_bytes_32 + offset;
-        for (size_t i2 = 0; i2 < VH_HASH_OUT_SIZE / sizeof(uint32_t); i2++)
-        {
-            const uint32_t value = *( blob_off + i2 );
-            uint32_t* p1_ptr = p1_32 + i2;
-            *p1_ptr = fnv1a( *p1_ptr, value );
-            value_accumulator = fnv1a( value_accumulator, value );
-        }
+
+        // update value accumulator for next seek index
+        value_accumulator = fnv1a( value_accumulator, blob_off[0] );
+        value_accumulator = fnv1a( value_accumulator, blob_off[1] );
+        value_accumulator = fnv1a( value_accumulator, blob_off[2] );
+        value_accumulator = fnv1a( value_accumulator, blob_off[3] );
+        value_accumulator = fnv1a( value_accumulator, blob_off[4] );
+        value_accumulator = fnv1a( value_accumulator, blob_off[5] );
+        value_accumulator = fnv1a( value_accumulator, blob_off[6] );
+        value_accumulator = fnv1a( value_accumulator, blob_off[7] );
+        
+#if defined (__AVX2__)        
+        *(__m256i*)p1_32 = _mm256_mullo_epi32( _mm256_xor_si256(
+                                  *(__m256i*)p1_32, *(__m256i*)blob_off ), k );
+#elif defined(__SSE41__)
+        casti_m128i( p1_32, 0 ) = _mm_mullo_epi32( _mm_xor_si128( 
+                    casti_m128i( p1_32, 0 ), casti_m128i( blob_off, 0 ) ), k );
+        casti_m128i( p1_32, 1 ) = _mm_mullo_epi32( _mm_xor_si128( 
+                    casti_m128i( p1_32, 1 ), casti_m128i( blob_off, 1 ) ), k );
+#else
+         for ( size_t i2 = 0; i2 < VH_HASH_OUT_SIZE / sizeof(uint32_t); i2++ )
+            p1_32[i2] = fnv1a( p1_32[i2], blob_off[i2] );
+#endif
+
    }

-    memcpy(output, p1, VH_HASH_OUT_SIZE);
+    memcpy( output, p1, VH_HASH_OUT_SIZE );
 }

 //-----------------------------------------------------------------------------
--- a/algo/verthash/Verthash.h
+++ b/algo/verthash/Verthash.h
@@ -52,6 +52,8 @@ void verthash_hash(const unsigned char* blob_bytes,
                   const unsigned char(*input)[VH_HEADER_SIZE],
                   unsigned char(*output)[VH_HASH_OUT_SIZE]);

+void verthash_sha3_512_prehash_72( const void *input );
+void verthash_sha3_512_final_8( void *hash, const uint64_t nonce );

 #endif // !Verthash_INCLUDE_ONCE

--- a/algo/verthash/tiny_sha3/sha3-4way.c
+++ b/algo/verthash/tiny_sha3/sha3-4way.c
@@ -0,0 +1,301 @@
+#if defined(__AVX2__)
+
+// sha3-4way.c
+// 19-Nov-11  Markku-Juhani O. Saarinen <mjos@iki.fi>
+// vectorization by JayDDee 2021-03-27
+//
+// Revised 07-Aug-15 to match with official release of FIPS PUB 202 "SHA3"
+// Revised 03-Sep-15 for portability + OpenSSL - style API
+
+#include "sha3-4way.h"
+
+// constants
+static const uint64_t keccakf_rndc[24] = {
+        0x0000000000000001, 0x0000000000008082, 0x800000000000808a,
+        0x8000000080008000, 0x000000000000808b, 0x0000000080000001,
+        0x8000000080008081, 0x8000000000008009, 0x000000000000008a,
+        0x0000000000000088, 0x0000000080008009, 0x000000008000000a,
+        0x000000008000808b, 0x800000000000008b, 0x8000000000008089,
+        0x8000000000008003, 0x8000000000008002, 0x8000000000000080,
+        0x000000000000800a, 0x800000008000000a, 0x8000000080008081,
+        0x8000000000008080, 0x0000000080000001, 0x8000000080008008
+    };
+
+void sha3_4way_keccakf( __m256i st[25] )
+{
+   int i, j, r;
+   __m256i t, bc[5];
+
+   for ( r = 0; r < KECCAKF_ROUNDS; r++ )
+   {
+      // Theta
+      bc[0] = _mm256_xor_si256( st[0],
+                           mm256_xor4( st[5], st[10], st[15], st[20] ) );
+      bc[1] = _mm256_xor_si256( st[1],
+                           mm256_xor4( st[6], st[11], st[16], st[21] ) );
+      bc[2] = _mm256_xor_si256( st[2],
+                           mm256_xor4( st[7], st[12], st[17], st[22] ) );
+      bc[3] = _mm256_xor_si256( st[3],
+                           mm256_xor4( st[8], st[13], st[18], st[23] ) );
+      bc[4] = _mm256_xor_si256( st[4],
+                           mm256_xor4( st[9], st[14], st[19], st[24] ) );
+
+      for ( i = 0; i < 5; i++ )
+      {
+         t = _mm256_xor_si256( bc[ (i+4) % 5 ],
+                               mm256_rol_64( bc[ (i+1) % 5 ], 1 ) );
+         st[ i    ]  = _mm256_xor_si256( st[ i    ],  t );
+         st[ i+5  ]  = _mm256_xor_si256( st[ i+ 5 ],  t );
+         st[ i+10 ]  = _mm256_xor_si256( st[ i+10 ],  t );
+         st[ i+15 ]  = _mm256_xor_si256( st[ i+15 ],  t );
+         st[ i+20 ]  = _mm256_xor_si256( st[ i+20 ],  t );
+      }
+
+      // Rho Pi
+#define RHO_PI( i, c ) \
+   bc[0] = st[ i ]; \
+   st[ i ] = mm256_rol_64( t, c ); \
+   t = bc[0]
+
+      t = st[1];
+
+      RHO_PI( 10,  1 );
+      RHO_PI(  7,  3 );
+      RHO_PI( 11,  6 );
+      RHO_PI( 17, 10 );
+      RHO_PI( 18, 15 );
+      RHO_PI(  3, 21 );
+      RHO_PI(  5, 28 );
+      RHO_PI( 16, 36 );
+      RHO_PI(  8, 45 );
+      RHO_PI( 21, 55 );
+      RHO_PI( 24,  2 );
+      RHO_PI(  4, 14 );
+      RHO_PI( 15, 27 );
+      RHO_PI( 23, 41 );
+      RHO_PI( 19, 56 );
+      RHO_PI( 13,  8 );
+      RHO_PI( 12, 25 );
+      RHO_PI(  2, 43 );
+      RHO_PI( 20, 62 );
+      RHO_PI( 14, 18 );
+      RHO_PI( 22, 39 );
+      RHO_PI(  9, 61 );
+      RHO_PI(  6, 20 );
+      RHO_PI(  1, 44 );
+
+#undef RHO_PI        
+
+      //  Chi
+      for ( j = 0; j < 25; j += 5 )
+      {
+         memcpy( bc, &st[ j ], 5*32 );
+         st[ j   ] = _mm256_xor_si256( st[ j   ],
+                                       _mm256_andnot_si256( bc[1], bc[2] ) );
+         st[ j+1 ] = _mm256_xor_si256( st[ j+1 ],
+                                       _mm256_andnot_si256( bc[2], bc[3] ) );
+         st[ j+2 ] = _mm256_xor_si256( st[ j+2 ],
+                                       _mm256_andnot_si256( bc[3], bc[4] ) );
+         st[ j+3 ] = _mm256_xor_si256( st[ j+3 ],
+                                       _mm256_andnot_si256( bc[4], bc[0] ) );
+         st[ j+4 ] = _mm256_xor_si256( st[ j+4 ],
+                                       _mm256_andnot_si256( bc[0], bc[1] ) );
+      }
+
+      //  Iota
+      st[0] = _mm256_xor_si256( st[0],
+                                _mm256_set1_epi64x( keccakf_rndc[ r ] ) );
+   }
+}
+
+int sha3_4way_init( sha3_4way_ctx_t *c, int mdlen )
+{
+    for ( int i = 0; i < 25; i++ )  c->st[ i ] = m256_zero;
+    c->mdlen = mdlen;
+    c->rsiz = 200 - 2 * mdlen;
+    c->pt = 0;
+    return 1;
+}
+
+int sha3_4way_update( sha3_4way_ctx_t *c, const void *data, size_t len )
+{
+    size_t i;
+    int j =  c->pt;
+    const int rsiz = c->rsiz / 8;
+    const int l = len / 8;
+
+    for ( i = 0; i < l; i++ )
+    {
+        c->st[ j ] = _mm256_xor_si256( c->st[ j ],
+                                       ( (const __m256i*)data )[i] );
+        j++;
+        if ( j >= rsiz )
+        {
+            sha3_4way_keccakf( c->st );
+            j = 0;
+        }
+    }
+    c->pt = j;
+
+    return 1;
+}
+
+int sha3_4way_final( void *md, sha3_4way_ctx_t *c )
+{
+    c->st[ c->pt ] = _mm256_xor_si256( c->st[ c->pt ],
+                                       m256_const1_64( 6 ) );
+    c->st[ c->rsiz / 8 - 1 ] =
+                       _mm256_xor_si256( c->st[ c->rsiz / 8 - 1 ],
+                                         m256_const1_64( 0x8000000000000000 ) );
+    sha3_4way_keccakf( c->st );
+    memcpy( md, c->st, c->mdlen * 4 );
+    return 1;
+}
+
+void *sha3_4way( const void *in, size_t inlen, void *md, int mdlen )
+{
+    sha3_4way_ctx_t ctx;
+    sha3_4way_init( &ctx, mdlen);
+    sha3_4way_update( &ctx, in, inlen );
+    sha3_4way_final( md, &ctx );
+    return md;
+}
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+void sha3_8way_keccakf( __m512i st[25] )
+{
+    int i, j, r;
+    __m512i t, bc[5];
+
+    // actual iteration
+    for ( r = 0; r < KECCAKF_ROUNDS; r++ )
+    {
+
+        // Theta
+        for ( i = 0; i < 5; i++ )
+           bc[i] = _mm512_xor_si512( st[i], 
+              mm512_xor4( st[ i+5 ], st[ i+10 ], st[ i+15 ], st[i+20 ] ) );
+
+        for ( i = 0; i < 5; i++ )
+        {
+            t = _mm512_xor_si512( bc[(i + 4) % 5],
+                                  _mm512_rol_epi64( bc[(i + 1) % 5], 1 ) );
+            for ( j = 0; j < 25; j += 5 )
+                st[j + i]  = _mm512_xor_si512( st[j + i],  t );
+        }
+
+        // Rho Pi
+#define RHO_PI( i, c ) \
+   bc[0] = st[ i ]; \
+   st[ i ] = _mm512_rol_epi64( t, c ); \
+   t = bc[0]
+
+        t = st[1];
+
+        RHO_PI( 10,  1 );        
+        RHO_PI(  7,  3 );
+        RHO_PI( 11,  6 );
+        RHO_PI( 17, 10 );
+        RHO_PI( 18, 15 );
+        RHO_PI(  3, 21 );
+        RHO_PI(  5, 28 );
+        RHO_PI( 16, 36 );
+        RHO_PI(  8, 45 );
+        RHO_PI( 21, 55 );
+        RHO_PI( 24,  2 );
+        RHO_PI(  4, 14 );
+        RHO_PI( 15, 27 );
+        RHO_PI( 23, 41 );
+        RHO_PI( 19, 56 );
+        RHO_PI( 13,  8 );
+        RHO_PI( 12, 25 );
+        RHO_PI(  2, 43 );
+        RHO_PI( 20, 62 );
+        RHO_PI( 14, 18 );
+        RHO_PI( 22, 39 );
+        RHO_PI(  9, 61 );
+        RHO_PI(  6, 20 );
+        RHO_PI(  1, 44 );
+
+#undef RHO_PI        
+
+        //  Chi
+        for ( j = 0; j < 25; j += 5 )
+        {
+            for ( i = 0; i < 5; i++ )
+                bc[i] = st[j + i];
+            for ( i = 0; i < 5; i++ )
+                st[ j+i ] = _mm512_xor_si512(  st[ j+i ],  _mm512_andnot_si512(
+                                         bc[ (i+1) % 5 ], bc[ (i+2) % 5 ] ) );
+        }
+
+        //  Iota
+        st[0] = _mm512_xor_si512( st[0], _mm512_set1_epi64( keccakf_rndc[r] ) );
+    }
+}
+
+// Initialize the context for SHA3
+
+int sha3_8way_init( sha3_8way_ctx_t *c, int mdlen )
+{
+    for ( int i = 0; i < 25; i++ )  c->st[ i ] = m512_zero;
+    c->mdlen = mdlen;
+    c->rsiz = 200 - 2 * mdlen;
+    c->pt = 0;
+    return 1;
+}
+
+// update state with more data
+
+int sha3_8way_update( sha3_8way_ctx_t *c, const void *data, size_t len )
+{
+    size_t i;
+    int j =  c->pt;
+    const int rsiz = c->rsiz / 8;
+    const int l = len / 8;
+
+    for ( i = 0; i < l; i++ )
+    {
+        c->st[ j ] = _mm512_xor_si512( c->st[ j ],
+                                        ( (const __m512i*)data )[i] );
+        j++;
+        if ( j >= rsiz )
+        {
+            sha3_8way_keccakf( c->st );
+            j = 0;
+        }
+    }
+    c->pt = j;
+
+    return 1;
+}
+
+// finalize and output a hash
+
+int sha3_8way_final( void *md, sha3_8way_ctx_t *c )
+{
+    c->st[ c->pt ] =
+                       _mm512_xor_si512( c->st[ c->pt ],
+                                         m512_const1_64( 6 ) );
+    c->st[ c->rsiz / 8 - 1 ] =
+                       _mm512_xor_si512( c->st[ c->rsiz / 8 - 1 ],
+                                         m512_const1_64( 0x8000000000000000 ) );
+    sha3_8way_keccakf( c->st );
+    memcpy( md, c->st, c->mdlen * 8 );
+    return 1;
+}
+
+// compute a SHA-3 hash (md) of given byte length from "in"
+
+void *sha3_8way( const void *in, size_t inlen, void *md, int mdlen )
+{
+    sha3_8way_ctx_t sha3;
+    sha3_8way_init( &sha3, mdlen);
+    sha3_8way_update( &sha3, in, inlen );
+    sha3_8way_final( md, &sha3 );
+    return md;
+}
+
+#endif  // AVX512
+#endif  // AVX2
--- a/algo/verthash/tiny_sha3/sha3-4way.h
+++ b/algo/verthash/tiny_sha3/sha3-4way.h
@@ -0,0 +1,67 @@
+// sha3.h
+// 19-Nov-11  Markku-Juhani O. Saarinen <mjos@iki.fi>
+// 2021-03-27 JayDDee
+//
+#ifndef SHA3_4WAY_H
+#define SHA3_4WAY_H
+
+#include <stddef.h>
+#include <stdint.h>
+#include "simd-utils.h"
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#ifndef KECCAKF_ROUNDS
+#define KECCAKF_ROUNDS 24
+#endif
+
+#if defined(__AVX2__)
+
+typedef struct
+{
+   __m256i st[25];                     // 64-bit words * 4 lanes
+    int pt, rsiz, mdlen;                    // these don't overflow
+} sha3_4way_ctx_t __attribute__ ((aligned (64)));;
+
+// Compression function.
+void sha3_4way_keccakf( __m256i st[25] );
+
+// OpenSSL - like interfece
+int sha3_4way_init( sha3_4way_ctx_t *c, int mdlen );    // mdlen = hash output in bytes
+int sha3_4way_update( sha3_4way_ctx_t *c, const void *data, size_t len );
+int sha3_4way_final( void *md, sha3_4way_ctx_t *c );    // digest goes to md
+
+// compute a sha3 hash (md) of given byte length from "in"
+void *sha3_4way( const void *in, size_t inlen, void *md, int mdlen );
+
+
+#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+
+// state context
+typedef struct
+{
+   __m512i st[25];                     // 64-bit words * 8 lanes
+    int pt, rsiz, mdlen;                    // these don't overflow
+} sha3_8way_ctx_t __attribute__ ((aligned (64)));;
+
+// Compression function.
+void sha3_8way_keccakf( __m512i st[25] );
+
+// OpenSSL - like interfece
+int sha3_8way_init( sha3_8way_ctx_t *c, int mdlen );    // mdlen = hash output in bytes
+int sha3_8way_update( sha3_8way_ctx_t *c, const void *data, size_t len );
+int sha3_8way_final( void *md, sha3_8way_ctx_t *c );    // digest goes to md
+
+// compute a sha3 hash (md) of given byte length from "in"
+void *sha3_8way( const void *in, size_t inlen, void *md, int mdlen );
+
+#endif // AVX512
+#endif // AVX2
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif
--- a/algo/verthash/tiny_sha3/sha3.c
+++ b/algo/verthash/tiny_sha3/sha3.c
@@ -5,6 +5,7 @@
 // Revised 03-Sep-15 for portability + OpenSSL - style API

 #include "sha3.h"
+#include <string.h>

 // update the state with given number of rounds

@@ -21,6 +22,7 @@ void sha3_keccakf(uint64_t st[25])
        0x000000000000800a, 0x800000008000000a, 0x8000000080008081,
        0x8000000000008080, 0x0000000080000001, 0x8000000080008008
    };
+/*
    const int keccakf_rotc[24] = {
        1,  3,  6,  10, 15, 21, 28, 36, 45, 55, 2,  14,
        27, 41, 56, 8,  25, 43, 62, 18, 39, 61, 20, 44
@@ -29,6 +31,7 @@ void sha3_keccakf(uint64_t st[25])
        10, 7,  11, 17, 18, 3, 5,  16, 8,  21, 24, 4,
        15, 23, 19, 13, 12, 2, 20, 14, 22, 9,  6,  1
    };
+*/

    // variables
    int i, j, r;
@@ -60,14 +63,50 @@ void sha3_keccakf(uint64_t st[25])
                st[j + i] ^= t;
        }

+        
        // Rho Pi
+#define RHO_PI( i, c ) \
+   bc[0] = st[ i ]; \
+   st[ i ] = ROTL64( t, c ); \
+   t = bc[0]
+
        t = st[1];
+
+        RHO_PI( 10,  1 );
+        RHO_PI(  7,  3 );
+        RHO_PI( 11,  6 );
+        RHO_PI( 17, 10 );
+        RHO_PI( 18, 15 );
+        RHO_PI(  3, 21 );
+        RHO_PI(  5, 28 );
+        RHO_PI( 16, 36 );
+        RHO_PI(  8, 45 );
+        RHO_PI( 21, 55 );
+        RHO_PI( 24,  2 );
+        RHO_PI(  4, 14 );
+        RHO_PI( 15, 27 );
+        RHO_PI( 23, 41 );
+        RHO_PI( 19, 56 );
+        RHO_PI( 13,  8 );
+        RHO_PI( 12, 25 );
+        RHO_PI(  2, 43 );
+        RHO_PI( 20, 62 );
+        RHO_PI( 14, 18 );
+        RHO_PI( 22, 39 );
+        RHO_PI(  9, 61 );
+        RHO_PI(  6, 20 );
+        RHO_PI(  1, 44 );
+
+#undef RHO_PI        
+
+/*        
        for (i = 0; i < 24; i++) {
            j = keccakf_piln[i];
            bc[0] = st[j];
            st[j] = ROTL64(t, keccakf_rotc[i]);
            t = bc[0];
        }
+*/

        //  Chi
        for (j = 0; j < 25; j += 5) {
@@ -118,17 +157,20 @@ int sha3_init(sha3_ctx_t *c, int mdlen)
 int sha3_update(sha3_ctx_t *c, const void *data, size_t len)
 {
    size_t i;
-    int j;
+    int j = c->pt / 8;
+    const int rsiz = c->rsiz / 8;
+    const int l = len / 8;

-    j = c->pt;
-    for (i = 0; i < len; i++) {
-        c->st.b[j++] ^= ((const uint8_t *) data)[i];
-        if (j >= c->rsiz) {
-            sha3_keccakf(c->st.q);
+    for ( i = 0; i < l; i++ )
+    {
+        c->st.q[ j++ ] ^= ( ((const uint64_t *) data) [i] );
+        if ( j >= rsiz )
+        {
+            sha3_keccakf( c->st.q );
            j = 0;
        }
    }
-    c->pt = j;
+    c->pt = j*8;

    return 1;
 }
@@ -137,16 +179,10 @@ int sha3_update(sha3_ctx_t *c, const void *data, size_t len)

 int sha3_final(void *md, sha3_ctx_t *c)
 {
-    int i;
-
-    c->st.b[c->pt] ^= 0x06;
-    c->st.b[c->rsiz - 1] ^= 0x80;
+    c->st.q[ c->pt / 8 ] ^= 6;
+    c->st.q[ c->rsiz / 8 - 1 ] ^= 0x8000000000000000;
    sha3_keccakf(c->st.q);
-
-    for (i = 0; i < c->mdlen; i++) {
-        ((uint8_t *) md)[i] = c->st.b[i];
-    }
-
+    memcpy( md, c->st.q, c->mdlen );
    return 1;
 }

@@ -155,7 +191,6 @@ int sha3_final(void *md, sha3_ctx_t *c)
 void *sha3(const void *in, size_t inlen, void *md, int mdlen)
 {
    sha3_ctx_t sha3;
-
    sha3_init(&sha3, mdlen);
    sha3_update(&sha3, in, inlen);
    sha3_final(md, &sha3);
--- a/algo/verthash/verthash-gate.c
+++ b/algo/verthash/verthash-gate.c
@@ -1,6 +1,7 @@
 #include "algo-gate-api.h"
 #include "algo/sha/sph_sha2.h"
 #include "Verthash.h"
+#include "tiny_sha3/sha3-4way.h"

 static verthash_info_t verthashInfo;

@@ -12,6 +13,82 @@ static const uint8_t verthashDatFileHash_bytes[32] =
  0x29, 0xec, 0xf8, 0x8f, 0x8a, 0xd4, 0x76, 0x39,
  0xb6, 0xed, 0xed, 0xaf, 0xd7, 0x21, 0xaa, 0x48 };

+#if defined(__AVX2__)
+
+static __thread sha3_4way_ctx_t sha3_mid_ctxA;
+static __thread sha3_4way_ctx_t sha3_mid_ctxB;
+
+#else
+
+static __thread sha3_ctx_t sha3_mid_ctx[8];
+
+#endif
+
+void verthash_sha3_512_prehash_72( const void *input )
+{
+#if defined(__AVX2__)
+   
+   __m256i vin[10];
+   mm256_intrlv80_4x64( vin, input );
+
+   sha3_4way_init( &sha3_mid_ctxA, 64 );
+   sha3_4way_init( &sha3_mid_ctxB, 64 );
+
+   vin[0] = _mm256_add_epi8( vin[0], _mm256_set_epi64x( 4,3,2,1 ) );
+   sha3_4way_update( &sha3_mid_ctxA, vin, 72 );
+
+   vin[0] = _mm256_add_epi8( vin[0], _mm256_set1_epi64x( 4 ) );
+   sha3_4way_update( &sha3_mid_ctxB, vin, 72 );
+
+#else
+
+   char in[80] __attribute__ ((aligned (64)));
+   memcpy( in, input, 80 );   
+   for ( int i = 0; i < 8; i++ )
+   {
+      in[0] += 1;
+      sha3_init( &sha3_mid_ctx[i], 64 );
+      sha3_update( &sha3_mid_ctx[i], in, 72 );
+   }
+
+#endif
+}
+
+void verthash_sha3_512_final_8( void *hash, const uint64_t nonce )
+{
+#if defined(__AVX2__)
+
+    __m256i vhashA[ 10 ] __attribute__ ((aligned (64)));
+    __m256i vhashB[ 10 ] __attribute__ ((aligned (64)));
+
+   sha3_4way_ctx_t ctx;
+   __m256i vnonce = _mm256_set1_epi64x( nonce );
+
+   memcpy( &ctx, &sha3_mid_ctxA, sizeof ctx );
+   sha3_4way_update( &ctx, &vnonce, 8 );
+   sha3_4way_final( vhashA, &ctx );
+
+   memcpy( &ctx, &sha3_mid_ctxB, sizeof ctx );
+   sha3_4way_update( &ctx, &vnonce, 8 );
+   sha3_4way_final( vhashB, &ctx );
+
+   dintrlv_4x64( hash,     hash+64,  hash+128, hash+192, vhashA, 512 );
+   dintrlv_4x64( hash+256, hash+320, hash+384, hash+448, vhashB, 512 );
+   
+#else
+
+   for ( int i = 0; i < 8; i++ )
+   {
+      sha3_ctx_t ctx;
+      memcpy( &ctx, &sha3_mid_ctx[i], sizeof ctx );
+      sha3_update( &ctx, &nonce, 8 );
+      sha3_final( hash + i*64, &ctx );
+   }
+   
+#endif
+}
+
+
 int scanhash_verthash( struct work *work, uint32_t max_nonce,
                      uint64_t *hashes_done, struct thr_info *mythr )
 {
@@ -26,6 +103,8 @@ int scanhash_verthash( struct work *work, uint32_t max_nonce,
   const bool bench = opt_benchmark;

   mm128_bswap32_80( edata, pdata );
+   verthash_sha3_512_prehash_72( edata );
+
   do
   {
      edata[19] = n;
@@ -51,15 +130,14 @@ bool register_verthash_algo( algo_gate_t* gate )

  opt_target_factor = 256.0;
  gate->scanhash  = (void*)&scanhash_verthash;
+  gate->optimizations = AVX2_OPT;
   
-  // verthash data file
  char *verthash_data_file = opt_data_file ? opt_data_file
                                           : default_verthash_data_file;
  
   int vhLoadResult = verthash_info_init( &verthashInfo, verthash_data_file );
   if (vhLoadResult == 0) // No Error
   {
-      //  and verify data file(if it was enabled)
      if ( opt_verify )
      {
         uint8_t vhDataFileHash[32] = { 0 };
@@ -78,7 +156,6 @@ bool register_verthash_algo( algo_gate_t* gate )
      }
   }
   else
-
   {
      // Handle Verthash error codes
      if ( vhLoadResult == 1 )
--- a/20
+++ b/20
@@ -1,6 +1,6 @@
 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.69 for cpuminer-opt 3.16.1.
+# Generated by GNU Autoconf 2.69 for cpuminer-opt 3.16.2.
 #
 #
 # 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='3.16.1'
-PACKAGE_STRING='cpuminer-opt 3.16.1'
+PACKAGE_VERSION='3.16.2'
+PACKAGE_STRING='cpuminer-opt 3.16.2'
 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 3.16.1 to adapt to many kinds of systems.
+\`configure' configures cpuminer-opt 3.16.2 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 3.16.1:";;
+     short | recursive ) echo "Configuration of cpuminer-opt 3.16.2:";;
   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 3.16.1
+cpuminer-opt configure 3.16.2
 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 3.16.1, which was
+It was created by cpuminer-opt $as_me 3.16.2, 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='3.16.1'
+ VERSION='3.16.2'


 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 3.16.1, which was
+This file was extended by cpuminer-opt $as_me 3.16.2, 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 3.16.1
+cpuminer-opt config.status 3.16.2
 configured by $0, generated by GNU Autoconf 2.69,
  with options \\"\$ac_cs_config\\"

--- a/configure.ac
+++ b/configure.ac
@@ -1,4 +1,4 @@
-AC_INIT([cpuminer-opt], [3.16.1])
+AC_INIT([cpuminer-opt], [3.16.2])

 AC_PREREQ([2.59c])
 AC_CANONICAL_SYSTEM
--- a/cpu-miner.c
+++ b/cpu-miner.c
@@ -555,7 +555,11 @@ static bool gbt_work_decode( const json_t *val, struct work *work )
         if ( !s )
            continue;
         if ( !strcmp( s, "segwit" ) || !strcmp( s, "!segwit" ) )
+         {
            segwit = true;
+            if ( opt_debug )
+               applog( LOG_INFO, "GBT: SegWit is enabled" );
+         }
      }
   }
 // Segwit END
@@ -954,25 +958,25 @@ void scale_hash_for_display ( double* hashrate, char* prefix )
  else                          { *prefix = 'Y';  *hashrate /= 1e24; }
 }

-static inline void sprintf_et( char *str, int seconds )
+static inline void sprintf_et( char *str, long unsigned int seconds )
 {
-   // sprintf doesn't like uint64_t, Linux thinks it's long, Windows long long.
-   unsigned int min = seconds / 60;
-   unsigned int sec = seconds % 60;
-   unsigned int hrs = min / 60;
+   long unsigned int min = seconds / 60;
+   long unsigned int sec = seconds % 60;
+   long unsigned int hrs = min / 60;
+   
   if ( unlikely( hrs ) )   
   {
-      unsigned int years = hrs / (24*365);
-      unsigned int days = hrs / 24;
-      if ( years )
-         sprintf( str, "%uy%ud", years, years % 365 );
-      else if ( days )  //0d00h
-         sprintf( str, "%ud%02uh", days, hrs % 24 );
+      long unsigned int days = hrs / 24;
+      long unsigned int years = days / 365;
+      if ( years )      // 0y000d
+         sprintf( str, "%luy%lud", years, years % 365 );
+      else if ( days )  // 0d00h
+         sprintf( str, "%lud%02luh", days, hrs % 24 );
      else         // 0h00m  
-         sprintf( str, "%uh%02um", hrs, min % 60 );
+         sprintf( str, "%luh%02lum", hrs, min % 60 );
   }
   else         // 0m00s
-      sprintf( str, "%um%02us", min, sec );
+      sprintf( str, "%lum%02lus", min, sec );
 }

 const long double exp32 = EXP32;                                  // 2**32
@@ -1071,7 +1075,8 @@ void report_summary_log( bool force )
   
   double share_time = (double)et.tv_sec + (double)et.tv_usec / 1e6;
   double ghrate = global_hashrate;
-   double shrate = safe_div( exp32 * last_targetdiff * (double)(accepts),
+   double target_diff = exp32 * last_targetdiff;
+   double shrate = safe_div( target_diff * (double)(accepts),
                             share_time, 0. );
   double sess_hrate = safe_div( exp32 * norm_diff_sum,
                                 (double)uptime.tv_sec, 0. );
@@ -1099,12 +1104,12 @@ void report_summary_log( bool force )

   if ( accepted_share_count < submitted_share_count )
   {
-      double ltd = exp32 * last_targetdiff;
      double lost_ghrate = uptime.tv_sec == 0 ? 0.
-                : ltd * (double)(submitted_share_count - accepted_share_count )
+                : target_diff
+                       * (double)(submitted_share_count - accepted_share_count )
                  / (double)uptime.tv_sec;
      double lost_shrate = share_time == 0. ? 0.
-               : ltd  * (double)(submits - accepts ) / share_time;
+               : target_diff  * (double)(submits - accepts ) / share_time;
      char lshr_units[4] = {0};
      char lghr_units[4] = {0};
      scale_hash_for_display( &lost_shrate, lshr_units );
@@ -2437,10 +2442,14 @@ static void *miner_thread( void *userdata )
 #if ((defined(_WIN64) || defined(__WINDOWS__)) || defined(_WIN32))
             applog( LOG_NOTICE, "Total: %s %sH/s", hr, hr_units );
 #else
-             applog( LOG_NOTICE, "Total: %s %sH/s, CPU temp: %dC",
-                                  hr, hr_units, (uint32_t)cpu_temp(0) );
+             float lo_freq = 0., hi_freq = 0.;
+             linux_cpu_hilo_freq( &lo_freq, &hi_freq );
+             applog( LOG_NOTICE,
+                     "Total: %s %sH/s, Temp: %dC, Freq: %.3f/%.3f GHz",
+                     hr, hr_units, (uint32_t)cpu_temp(0), lo_freq / 1e6,
+                     hi_freq / 1e6 );
 #endif
-	       }
+          }
       }  // benchmark

       // conditional mining
--- a/miner.h
+++ b/miner.h
@@ -900,7 +900,7 @@ Options:\n\
      --benchmark       run in offline benchmark mode\n\
      --cpu-affinity    set process affinity to cpu core(s), mask 0x3 for cores 0 and 1\n\
      --cpu-priority    set process priority (default: 0 idle, 2 normal to 5 highest)\n\
-  -b, --api-bind        IP/Port for the miner API (default: 127.0.0.1:4048)\n\
+  -b, --api-bind=address[:port]   IP address for the miner API, default port is 4048)\n\
      --api-remote      Allow remote control\n\
      --max-temp=N      Only mine if cpu temp is less than specified value (linux)\n\
      --max-rate=N[KMG] Only mine if net hashrate is less than specified value\n\
--- a/simd-utils/intrlv.h
+++ b/simd-utils/intrlv.h
@@ -1225,37 +1225,6 @@ static inline void intrlv_4x64( void *dst, const void *src0,
   d[31] = _mm_unpackhi_epi64( s2[7], s3[7] );
 }

-/*
-static inline void intrlv_4x64( void *dst, void *src0,
-           void *src1, void *src2, void *src3, int bit_len )
-{
-   uint64_t *d = (uint64_t*)dst;
-   uint64_t *s0 = (uint64_t*)src0;
-   uint64_t *s1 = (uint64_t*)src1;
-   uint64_t *s2 = (uint64_t*)src2;
-   uint64_t *s3 = (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 )
 {
@@ -1282,26 +1251,6 @@ static inline void intrlv_4x64_512( void *dst, const void *src0,
   d[15] = _mm_unpackhi_epi64( s2[3], s3[3] );
 }

-/*
-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, const int bit_len )
 {
@@ -1347,38 +1296,6 @@ static inline void dintrlv_4x64( void *dst0, void *dst1, void *dst2,
   d3[7] = _mm_unpackhi_epi64( s[29], s[31] );
 }

-
-/*
-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 )
 {
@@ -1405,26 +1322,6 @@ static inline void dintrlv_4x64_512( void *dst0, void *dst1, void *dst2,
   d3[3] = _mm_unpackhi_epi64( s[13], s[15] );
 }

-/*
-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 )
 {
@@ -1440,9 +1337,41 @@ static inline void extr_lane_4x64( void *d, const void *s,
 }

 #if defined(__AVX2__)
+// Doesn't really need AVX2, just SSSE3, but is only used with AVX2 code.

-// There a alignment problems with the source buffer on Wwindows,
-// can't use 256 bit bswap.
+static inline void mm256_intrlv80_4x64( void *d, const void *src )
+{
+  __m128i s0 = casti_m128i( src,0 );
+  __m128i s1 = casti_m128i( src,1 );
+  __m128i s2 = casti_m128i( src,2 );
+  __m128i s3 = casti_m128i( src,3 );
+  __m128i s4 = casti_m128i( src,4 );
+
+  casti_m128i( d,  0 ) =
+  casti_m128i( d,  1 ) = _mm_shuffle_epi32( s0, 0x44 );
+  casti_m128i( d,  2 ) =
+  casti_m128i( d,  3 ) = _mm_shuffle_epi32( s0, 0xee );
+
+  casti_m128i( d,  4 ) =
+  casti_m128i( d,  5 ) = _mm_shuffle_epi32( s1, 0x44 );
+  casti_m128i( d,  6 ) =
+  casti_m128i( d,  7 ) = _mm_shuffle_epi32( s1, 0xee );
+
+  casti_m128i( d,  8 ) =
+  casti_m128i( d,  9 ) = _mm_shuffle_epi32( s2, 0x44 );
+  casti_m128i( d, 10 ) =
+  casti_m128i( d, 11 ) = _mm_shuffle_epi32( s2, 0xee );
+
+  casti_m128i( d, 12 ) =
+  casti_m128i( d, 13 ) = _mm_shuffle_epi32( s3, 0x44 );
+  casti_m128i( d, 14 ) =
+  casti_m128i( d, 15 ) = _mm_shuffle_epi32( s3, 0xee );
+
+  casti_m128i( d, 16 ) =
+  casti_m128i( d, 17 ) = _mm_shuffle_epi32( s4, 0x44 );
+  casti_m128i( d, 18 ) =
+  casti_m128i( d, 19 ) = _mm_shuffle_epi32( s4, 0xee );
+}

 static inline void mm256_bswap32_intrlv80_4x64( void *d, const void *src )
 {
@@ -1636,40 +1565,6 @@ static inline void intrlv_8x64_512( void *dst, const void *src0,
   d[31] = _mm_unpackhi_epi64( s6[3], s7[3] );
 }

-/*
-#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
-*/
-

 static inline void dintrlv_8x64( void *dst0, void *dst1, void *dst2,
         void *dst3, void *dst4, void *dst5, void *dst6, void *dst7,
@@ -1815,39 +1710,6 @@ static inline void dintrlv_8x64_512( void *dst0, void *dst1, void *dst2,
   d7[3] = _mm_unpackhi_epi64( s[27], s[31] );
 }

-/*
-#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 )
 {
--- a/simd-utils/simd-128.h
+++ b/simd-utils/simd-128.h
@@ -178,7 +178,7 @@ static inline __m128i mm128_mask_32( const __m128i v, const int m )
 // Basic operations without equivalent SIMD intrinsic

 // Bitwise not (~v)  
-#define mm128_not( v )          _mm_xor_si128( (v), m128_neg1 ) 
+#define mm128_not( v )          _mm_xor_si128( v, m128_neg1 ) 

 // Unary negation of elements (-v)
 #define mm128_negate_64( v )    _mm_sub_epi64( m128_zero, v )
@@ -263,7 +263,8 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
   _mm_or_si128( _mm_slli_epi32( v, c ), _mm_srli_epi32( v, 32-(c) ) )


-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+#if defined(__AVX512VL__)
+//#if defined(__AVX512F__) && defined(__AVX512VL__)

 #define mm128_ror_64    _mm_ror_epi64
 #define mm128_rol_64    _mm_rol_epi64
@@ -291,16 +292,13 @@ static inline void memcpy_128( __m128i *dst, const __m128i *src, const int n )
 #define mm128_swap_64( v )    _mm_shuffle_epi32( v, 0x4e )
 #define mm128_ror_1x32( v )   _mm_shuffle_epi32( v, 0x39 )
 #define mm128_rol_1x32( v )   _mm_shuffle_epi32( v, 0x93 )
-//#define mm128_swap_64( v )    _mm_alignr_epi8( v, v,  8 )
-//#define mm128_ror_1x32( v )   _mm_alignr_epi8( v, v,  4 )
-//#define mm128_rol_1x32( v )   _mm_alignr_epi8( v, v, 12 )

 // Swap 32 bit elements in 64 bit lanes
 #define mm128_swap64_32( v )  _mm_shuffle_epi32( v, 0xb1 )

 #if defined(__SSSE3__)

-// Rotate right by c bytes
+// Rotate right by c bytes, no SSE2 equivalent.
 static inline __m128i mm128_ror_x8( const __m128i v, const int c )
 { return _mm_alignr_epi8( v, v, c ); }

--- a/simd-utils/simd-256.h
+++ b/simd-utils/simd-256.h
@@ -18,7 +18,7 @@
 #define mm256_mov64_256( i ) _mm256_castsi128_si256( mm128_mov64_128( i ) )
 #define mm256_mov32_256( i ) _mm256_castsi128_si256( mm128_mov32_128( i ) )

-// Mo0ve low element of vector to integer.
+// Move low element of vector to integer.
 #define mm256_mov256_64( v ) mm128_mov128_64( _mm256_castsi256_si128( v ) )
 #define mm256_mov256_32( v ) mm128_mov128_32( _mm256_castsi256_si128( v ) )

@@ -42,7 +42,7 @@ static inline __m256i m256_const_64( const uint64_t i3, const uint64_t i2,
 // 128 bit vector argument
 #define m256_const1_128( v ) \
   _mm256_permute4x64_epi64( _mm256_castsi128_si256( v ), 0x44 )
-// 64 bit integer argument
+// 64 bit integer argument zero extended to 128 bits.
 #define m256_const1_i128( i ) m256_const1_128( mm128_mov64_128( i ) )
 #define m256_const1_64( i )  _mm256_broadcastq_epi64( mm128_mov64_128( i ) )
 #define m256_const1_32( i )  _mm256_broadcastd_epi32( mm128_mov32_128( i ) )
@@ -168,7 +168,10 @@ static inline void memcpy_256( __m256i *dst, const __m256i *src, const int n )
                    _mm256_srli_epi32( v, 32-(c) ) )


-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
+// The spec says both F & VL are required, but just in case AMD
+// decides to implement ROL/R without AVX512F.
+#if defined(__AVX512VL__)
+//#if defined(__AVX512F__) && defined(__AVX512VL__)

 // AVX512, control must be 8 bit immediate.

@@ -198,21 +201,14 @@ static inline void memcpy_256( __m256i *dst, const __m256i *src, const int n )
 //
 // Rotate elements accross all lanes.
 //
-// AVX2 has no full vector permute for elements less than 32 bits.
-// AVX512 has finer granularity full vector permutes.
-// AVX512 has full vector alignr which might be faster, especially for 32 bit
+// Swap 128 bit elements in 256 bit vector.
+#define mm256_swap_128( v )     _mm256_permute4x64_epi64( v, 0x4e )

+// Rotate 256 bit vector by one 64 bit element
+#define mm256_ror_1x64( v )     _mm256_permute4x64_epi64( v, 0x39 )
+#define mm256_rol_1x64( v )     _mm256_permute4x64_epi64( v, 0x93 )

-#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
-
-static inline __m256i mm256_swap_128( const __m256i v )
-{ return _mm256_alignr_epi64( v, v, 2 ); }
-
-static inline __m256i mm256_ror_1x64( const __m256i v )
-{ return _mm256_alignr_epi64( v, v, 1 ); }
-
-static inline __m256i mm256_rol_1x64( const __m256i v )
-{ return _mm256_alignr_epi64( v, v, 3 ); }
+#if defined(__AVX512F__) && defined(__AVX512VL__)

 static inline __m256i mm256_ror_1x32( const __m256i v )
 { return _mm256_alignr_epi32( v, v, 1 ); }
@@ -220,21 +216,8 @@ static inline __m256i mm256_ror_1x32( const __m256i v )
 static inline __m256i mm256_rol_1x32( const __m256i v )
 { return _mm256_alignr_epi32( v, v, 7 ); }

-static inline __m256i mm256_ror_3x32( const __m256i v )
-{ return _mm256_alignr_epi32( v, v, 3 ); }
-
-static inline __m256i mm256_rol_3x32( const __m256i v )
-{ return _mm256_alignr_epi32( v, v, 5 ); }
-
 #else   // AVX2

-// Swap 128 bit elements in 256 bit vector.
-#define mm256_swap_128( v )     _mm256_permute4x64_epi64( v, 0x4e )
-
-// Rotate 256 bit vector by one 64 bit element
-#define mm256_ror_1x64( v )     _mm256_permute4x64_epi64( v, 0x39 )
-#define mm256_rol_1x64( v )     _mm256_permute4x64_epi64( v, 0x93 )
-
 // Rotate 256 bit vector by one 32 bit element.
 #define mm256_ror_1x32( v ) \
    _mm256_permutevar8x32_epi32( v, \
@@ -246,17 +229,6 @@ static inline __m256i mm256_rol_3x32( const __m256i v )
                     m256_const_64( 0x0000000600000005,  0x0000000400000003, \
                                    0x0000000200000001,  0x0000000000000007 )

-// Rotate 256 bit vector by three 32 bit elements (96 bits).
-#define mm256_ror_3x32( v ) \
-    _mm256_permutevar8x32_epi32( v, \
-                     m256_const_64( 0x0000000200000001, 0x0000000000000007, \
-                                    0x0000000600000005, 0x0000000400000003 ) 
-
-#define mm256_rol_3x32( v ) \
-    _mm256_permutevar8x32_epi32( v, \
-                     m256_const_64( 0x0000000400000003, 0x0000000200000001, \
-                                    0x0000000000000007, 0x0000000600000005 )
-
 #endif    // AVX512 else AVX2

 //
--- a/util.c
+++ b/util.c
@@ -943,6 +943,140 @@ bool jobj_binary(const json_t *obj, const char *key, void *buf, size_t buflen)
 	return true;
 }

+static uint32_t bech32_polymod_step(uint32_t pre) {
+    uint8_t b = pre >> 25;
+    return ((pre & 0x1FFFFFF) << 5) ^
+        (-((b >> 0) & 1) & 0x3b6a57b2UL) ^
+        (-((b >> 1) & 1) & 0x26508e6dUL) ^
+        (-((b >> 2) & 1) & 0x1ea119faUL) ^
+        (-((b >> 3) & 1) & 0x3d4233ddUL) ^
+        (-((b >> 4) & 1) & 0x2a1462b3UL);
+}
+
+static const int8_t bech32_charset_rev[128] = {
+    -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+    -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+    -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+    15, -1, 10, 17, 21, 20, 26, 30,  7,  5, -1, -1, -1, -1, -1, -1,
+    -1, 29, -1, 24, 13, 25,  9,  8, 23, -1, 18, 22, 31, 27, 19, -1,
+     1,  0,  3, 16, 11, 28, 12, 14,  6,  4,  2, -1, -1, -1, -1, -1,
+    -1, 29, -1, 24, 13, 25,  9,  8, 23, -1, 18, 22, 31, 27, 19, -1,
+     1,  0,  3, 16, 11, 28, 12, 14,  6,  4,  2, -1, -1, -1, -1, -1
+};
+
+static bool bech32_decode(char *hrp, uint8_t *data, size_t *data_len, const char *input) {
+    uint32_t chk = 1;
+    size_t i;
+    size_t input_len = strlen(input);
+    size_t hrp_len;
+    int have_lower = 0, have_upper = 0;
+    if (input_len < 8 || input_len > 90) {
+        return false;
+    }
+    *data_len = 0;
+    while (*data_len < input_len && input[(input_len - 1) - *data_len] != '1') {
+        ++(*data_len);
+    }
+    hrp_len = input_len - (1 + *data_len);
+    if (1 + *data_len >= input_len || *data_len < 6) {
+        return false;
+    }
+    *(data_len) -= 6;
+    for (i = 0; i < hrp_len; ++i) {
+        int ch = input[i];
+        if (ch < 33 || ch > 126) {
+            return false;
+        }
+        if (ch >= 'a' && ch <= 'z') {
+            have_lower = 1;
+        } else if (ch >= 'A' && ch <= 'Z') {
+            have_upper = 1;
+            ch = (ch - 'A') + 'a';
+        }
+        hrp[i] = ch;
+        chk = bech32_polymod_step(chk) ^ (ch >> 5);
+    }
+    hrp[i] = 0;
+    chk = bech32_polymod_step(chk);
+    for (i = 0; i < hrp_len; ++i) {
+        chk = bech32_polymod_step(chk) ^ (input[i] & 0x1f);
+    }
+    ++i;
+    while (i < input_len) {
+        int v = (input[i] & 0x80) ? -1 : bech32_charset_rev[(int)input[i]];
+        if (input[i] >= 'a' && input[i] <= 'z') have_lower = 1;
+        if (input[i] >= 'A' && input[i] <= 'Z') have_upper = 1;
+        if (v == -1) {
+            return false;
+        }
+        chk = bech32_polymod_step(chk) ^ v;
+        if (i + 6 < input_len) {
+            data[i - (1 + hrp_len)] = v;
+        }
+        ++i;
+    }
+    if (have_lower && have_upper) {
+        return false;
+    }
+    return chk == 1;
+}
+
+static bool convert_bits(uint8_t *out, size_t *outlen, int outbits, const uint8_t *in, size_t inlen, int inbits, int pad) {
+    uint32_t val = 0;
+    int bits = 0;
+    uint32_t maxv = (((uint32_t)1) << outbits) - 1;
+    while (inlen--) {
+        val = (val << inbits) | *(in++);
+        bits += inbits;
+        while (bits >= outbits) {
+            bits -= outbits;
+            out[(*outlen)++] = (val >> bits) & maxv;
+        }
+    }
+    if (pad) {
+        if (bits) {
+            out[(*outlen)++] = (val << (outbits - bits)) & maxv;
+        }
+    } else if (((val << (outbits - bits)) & maxv) || bits >= inbits) {
+        return false;
+    }
+    return true;
+}
+
+static bool segwit_addr_decode(int *witver, uint8_t *witdata, size_t *witdata_len, const char *addr) {
+    uint8_t data[84];
+    char hrp_actual[84];
+    size_t data_len;
+    if (!bech32_decode(hrp_actual, data, &data_len, addr)) return false;
+    if (data_len == 0 || data_len > 65) return false;
+    if (data[0] > 16) return false;
+    *witdata_len = 0;
+    if (!convert_bits(witdata, witdata_len, 8, data + 1, data_len - 1, 5, 0)) return false;
+    if (*witdata_len < 2 || *witdata_len > 40) return false;
+    if (data[0] == 0 && *witdata_len != 20 && *witdata_len != 32) return false;
+    *witver = data[0];
+    return true;
+}
+
+static size_t bech32_to_script(uint8_t *out, size_t outsz, const char *addr) {
+    uint8_t witprog[40];
+    size_t witprog_len;
+    int witver;
+
+    if (!segwit_addr_decode(&witver, witprog, &witprog_len, addr))
+        return 0;
+    if (outsz < witprog_len + 2)
+        return 0;
+    out[0] = witver ? (0x50 + witver) : 0;
+    out[1] = witprog_len;
+    memcpy(out + 2, witprog, witprog_len);
+
+   if ( opt_debug )
+      applog( LOG_INFO, "Coinbase address uses Bech32 coding");
+
+    return witprog_len + 2;
+}
+
 size_t address_to_script( unsigned char *out, size_t outsz, const char *addr )
 {
 	unsigned char addrbin[ pk_buffer_size_max ];
@@ -950,12 +1084,15 @@ size_t address_to_script( unsigned char *out, size_t outsz, const char *addr )
 	size_t rv;

 	if ( !b58dec( addrbin, outsz, addr ) )
-		return 0;
+		return bech32_to_script( out, outsz, addr );

   addrver = b58check( addrbin, outsz, addr );
   if ( addrver < 0 )
 		return 0;

+   if ( opt_debug )
+      applog( LOG_INFO, "Coinbase address uses B58 coding");
+
   switch ( addrver )
   {
 		case 5:    /* Bitcoin script hash */
@@ -1486,9 +1623,6 @@ static bool stratum_parse_extranonce(struct stratum_ctx *sctx, json_t *params, i
   if ( !opt_quiet ) /* pool dynamic change */
      applog( LOG_INFO, "Stratum extranonce1= %s, extranonce2 size= %d",
         xnonce1, xn2_size);
-//   if (pndx == 0 && opt_debug)
-//		applog(LOG_DEBUG, "Stratum set nonce %s with extranonce2 size=%d",
-//			xnonce1, xn2_size);

 	return true;
 out:
@@ -1638,8 +1772,6 @@ bool stratum_authorize(struct stratum_ctx *sctx, const char *user, const char *p
 		opt_extranonce = false;
      goto out;
 	}
-   if ( !opt_quiet )
-      applog( LOG_INFO, "Extranonce subscription enabled" );

 	sret = stratum_recv_line( sctx );
 	if ( sret )
@@ -1658,8 +1790,8 @@ bool stratum_authorize(struct stratum_ctx *sctx, const char *user, const char *p
 					applog( LOG_WARNING, "Stratum answer id is not correct!" );
 			}
 			res_val = json_object_get( extra, "result" );
-//			if (opt_debug && (!res_val || json_is_false(res_val)))
-//				applog(LOG_DEBUG, "extranonce subscribe not supported");
+			if (opt_debug && (!res_val || json_is_false(res_val)))
+				applog(LOG_DEBUG, "Method extranonce.subscribe is not supported");
 			json_decref( extra );
 		}
 		free(sret);