v3.15.6

README.txt

@@ -59,7 +59,7 @@ Notes about included DLL files:
 
 Downloading DLL files from alternative sources presents an inherent
 security risk if their source is unknown. All DLL files included have
-been copied from the Ubuntu-20.04 instalation or compiled by me from
+been copied from the Ubuntu-20.04 installation or compiled by me from
 source code obtained from the author's official repository. The exact
 procedure is documented in the build instructions for Windows:
 https://github.com/JayDDee/cpuminer-opt/wiki/Compiling-from-source

RELEASE_NOTES

@@ -65,6 +65,19 @@ If not what makes it happen or not happen?
 Change Log
 ----------
 
+v3.15.6
+
+Implement keccak pre-hash optimization for x16* algos.
+Move conditional mining test to before get_new_work in miner thread.
+Add test for share reject reason when solo mining.
+Add support for floating point, as well as integer, "networkhasps" in
+RPC getmininginfo method.
+
+v3.15.5
+
+Fix stratum jobs lost if 2 jobs received in less than one second.
+ 
+
 v3.15.4
 
 Fixed yescryptr16 broken in v3.15.3.

algo-gate-api.c

@@ -419,7 +419,6 @@ void exec_hash_function( int algo, void *output, const void *pdata )
 const char* const algo_alias_map[][2] =
 {
 //   alias                proper
-  { "argon2d-crds",      "argon2d250"     },
   { "argon2d-dyn",       "argon2d500"     },
   { "argon2d-uis",       "argon2d4096"    },
   { "bcd",               "x13bcd"         },

algo-gate-api.h

@@ -162,7 +162,7 @@ bool ( *ready_to_mine )         ( struct work*, struct stratum_ctx*, int );
 bool ( *do_this_thread )        ( int );
 
 // After do_this_thread
-void ( *resync_threads )        ( struct work* );
+void ( *resync_threads )        ( int, struct work* );
 
 // No longer needed
 json_t* (*longpoll_rpc_call)      ( CURL*, int*, char* );

algo/hodl/hodl-gate.c

@@ -99,13 +99,13 @@ void hodl_build_block_header( struct work* g_work, uint32_t version,
 // called only by thread 0, saves a backup of g_work
 void hodl_get_new_work( struct work* work, struct work* g_work)
 {
-   pthread_rwlock_rdlock( &g_work_lock );
+//   pthread_rwlock_rdlock( &g_work_lock );
 
    work_free( &hodl_work );
    work_copy( &hodl_work, g_work );
    hodl_work.data[ algo_gate.nonce_index ] = ( clock() + rand() ) % 9999;
 
-   pthread_rwlock_unlock( &g_work_lock );
+//   pthread_rwlock_unlock( &g_work_lock );
 }
 
 json_t *hodl_longpoll_rpc_call( CURL *curl, int *err, char* lp_url )
@@ -125,7 +125,7 @@ json_t *hodl_longpoll_rpc_call( CURL *curl, int *err, char* lp_url )
 }
 
 // called by every thread, copies the backup to each thread's work.
-void hodl_resync_threads( struct work* work )
+void hodl_resync_threads( int thr_id, struct work* work )
 {
    int nonce_index = algo_gate.nonce_index;
    pthread_barrier_wait( &hodl_barrier );
@@ -135,6 +135,7 @@ void hodl_resync_threads( struct work* work )
       work_copy( work, &hodl_work );
    }
    work->data[ nonce_index ] = swab32( hodl_work.data[ nonce_index ] );
+   work_restart[thr_id].restart = 0;
 }
 
 bool hodl_do_this_thread( int thr_id )

algo/keccak/keccak-macros.c

@@ -201,6 +201,7 @@
 #define IOTA(r)   XOR64_IOTA(a00, a00, r)
 
 #ifdef P0
+#undef P0
 #undef P1
 #undef P2
 #undef P3

algo/nist5/zr5.c

@@ -156,7 +156,7 @@ int scanhash_zr5( struct work *work, uint32_t max_nonce,
 void zr5_get_new_work( struct work* work, struct work* g_work, int thr_id,
                        uint32_t* end_nonce_ptr )
 {
-   pthread_rwlock_rdlock( &g_work_lock );
+//   pthread_rwlock_rdlock( &g_work_lock );
 
    // ignore POK in first word
    const int wkcmp_sz = 72;  // (19-1) * sizeof(uint32_t)
@@ -174,7 +174,7 @@ void zr5_get_new_work( struct work* work, struct work* g_work, int thr_id,
    else
        ++(*nonceptr);
 
-   pthread_rwlock_unlock( &g_work_lock );
+//   pthread_rwlock_unlock( &g_work_lock );
 }
 
 void zr5_display_pok( struct work* work )

algo/quark/hmq1725-4way.c

@@ -912,7 +912,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
    sph_whirlpool512_full( &ctx.whirlpool, hash2, hash2, 64 );
    sph_whirlpool512_full( &ctx.whirlpool, hash3, hash3, 64 );
 
-// A = fugue serial, B = sha512 prarallel
+// A = fugue serial, B = sha512 parallel
    
    intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
 

algo/x16/x16r-4way.c

@@ -16,8 +16,7 @@
 
 #if defined (X16R_8WAY)
 
-// Perform midstate prehash of hash functions with block size <= 64 bytes
-// and interleave 4x64 before nonce insertion for final hash.
+// Perform midstate prehash of hash functions with block size <= 72 bytes.
 
 void x16r_8way_prehash( void *vdata, void *pdata )
 {
@@ -34,6 +33,11 @@ void x16r_8way_prehash( void *vdata, void *pdata )
          jh512_8way_init( &x16r_ctx.jh );
          jh512_8way_update( &x16r_ctx.jh, vdata, 64 );
       break;
+      case KECCAK:
+         mm512_bswap32_intrlv80_8x64( vdata, pdata );
+         keccak512_8way_init( &x16r_ctx.keccak );
+         keccak512_8way_update( &x16r_ctx.keccak, vdata, 72 );
+      break;
       case SKEIN:
          mm512_bswap32_intrlv80_8x64( vdata, pdata );
          skein512_8way_init( &x16r_ctx.skein );
@@ -173,13 +177,13 @@ int x16r_8way_hash_generic( void* output, const void* input, int thrid )
                           hash7, vhash );
          break;
          case KECCAK:
-            keccak512_8way_init( &ctx.keccak );
-            if ( i == 0 )
-               keccak512_8way_update( &ctx.keccak, input, size );
+           if ( i == 0 )
+               keccak512_8way_update( &ctx.keccak, input + (72<<3), 8 );
             else
             {
                intrlv_8x64( vhash, in0, in1, in2, in3, in4, in5, in6, in7, 
                             size<<3 );
+               keccak512_8way_init( &ctx.keccak );
                keccak512_8way_update( &ctx.keccak, vhash, size );
             }
             keccak512_8way_close( &ctx.keccak, vhash );
@@ -490,6 +494,7 @@ int scanhash_x16r_8way( struct work *work, uint32_t max_nonce,
    {
       x16_r_s_getAlgoString( (const uint8_t*)bedata1, x16r_hash_order );
       s_ntime = ntime;
+
       if ( opt_debug && !thr_id )
           applog( LOG_INFO, "hash order %s (%08x)", x16r_hash_order, ntime );
    }
@@ -533,6 +538,11 @@ void x16r_4way_prehash( void *vdata, void *pdata )
          jh512_4way_init( &x16r_ctx.jh );
          jh512_4way_update( &x16r_ctx.jh, vdata, 64 );
       break;
+      case KECCAK:
+         mm256_bswap32_intrlv80_4x64( vdata, pdata );
+         keccak512_4way_init( &x16r_ctx.keccak );
+         keccak512_4way_update( &x16r_ctx.keccak, vdata, 72 );
+      break;
       case SKEIN:
          mm256_bswap32_intrlv80_4x64( vdata, pdata );
          skein512_4way_prehash64( &x16r_ctx.skein, vdata );
@@ -646,12 +656,12 @@ int x16r_4way_hash_generic( void* output, const void* input, int thrid )
             dintrlv_4x64_512( hash0, hash1, hash2, hash3, vhash );
          break;
          case KECCAK:
-            keccak512_4way_init( &ctx.keccak );
-            if ( i == 0 )
-               keccak512_4way_update( &ctx.keccak, input, size );
+           if ( i == 0 )
+               keccak512_4way_update( &ctx.keccak, input + (72<<2), 8 );
             else
             {
                intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
+               keccak512_4way_init( &ctx.keccak );
                keccak512_4way_update( &ctx.keccak, vhash, size );
             }
             keccak512_4way_close( &ctx.keccak, vhash );
@@ -883,7 +893,7 @@ int scanhash_x16r_4way( struct work *work, uint32_t max_nonce,
       x16_r_s_getAlgoString( (const uint8_t*)bedata1, x16r_hash_order );
       s_ntime = ntime;
       if ( opt_debug && !thr_id )
-              applog( LOG_INFO, "hash order %s (%08x)", x16r_hash_order, ntime );
+         applog( LOG_INFO, "hash order %s (%08x)", x16r_hash_order, ntime );
    }
 
    x16r_4way_prehash( vdata, pdata );

algo/yespower/yespower-blake2b.c

@@ -259,15 +259,24 @@ static inline void salsa20_simd_unshuffle(const salsa20_blk_t *Bin,
 #define WRITE_X(out) \
     (out).q[0] = X0; (out).q[1] = X1; (out).q[2] = X2; (out).q[3] = X3;
 
-#ifdef __XOP__
+#if defined(__AVX512VL__)
+
+#define ARX(out, in1, in2, s) \
+   out = _mm_xor_si128(out, _mm_rol_epi32(_mm_add_epi32(in1, in2), s));
+
+#elif defined(__XOP__)
+
 #define ARX(out, in1, in2, s) \
     out = _mm_xor_si128(out, _mm_roti_epi32(_mm_add_epi32(in1, in2), s));
+
 #else
+
 #define ARX(out, in1, in2, s) { \
     __m128i tmp = _mm_add_epi32(in1, in2); \
     out = _mm_xor_si128(out, _mm_slli_epi32(tmp, s)); \
     out = _mm_xor_si128(out, _mm_srli_epi32(tmp, 32 - s)); \
 }
+
 #endif
 
 #define SALSA20_2ROUNDS \

algo/yespower/yespower-opt.c

@@ -165,6 +165,7 @@ static inline void salsa20_simd_unshuffle(const salsa20_blk_t *Bin,
 }
 
 #ifdef __SSE2__
+
 #define DECL_X \
 	__m128i X0, X1, X2, X3;
 #define DECL_Y \
@@ -174,15 +175,24 @@ static inline void salsa20_simd_unshuffle(const salsa20_blk_t *Bin,
 #define WRITE_X(out) \
 	(out).q[0] = X0; (out).q[1] = X1; (out).q[2] = X2; (out).q[3] = X3;
 
-#ifdef __XOP__
+#if defined(__AVX512VL__)
+
+#define ARX(out, in1, in2, s) \
+   out = _mm_xor_si128(out, _mm_rol_epi32(_mm_add_epi32(in1, in2), s));
+
+#elif defined(__XOP__)
+
 #define ARX(out, in1, in2, s) \
 	out = _mm_xor_si128(out, _mm_roti_epi32(_mm_add_epi32(in1, in2), s));
+
 #else
+
 #define ARX(out, in1, in2, s) { \
 	__m128i tmp = _mm_add_epi32(in1, in2); \
 	out = _mm_xor_si128(out, _mm_slli_epi32(tmp, s)); \
 	out = _mm_xor_si128(out, _mm_srli_epi32(tmp, 32 - s)); \
 }
+
 #endif
 
 #define SALSA20_2ROUNDS \

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 3.15.4.
+# Generated by GNU Autoconf 2.69 for cpuminer-opt 3.15.6.
 #
 #
 # 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='3.15.4'
-PACKAGE_STRING='cpuminer-opt 3.15.4'
+PACKAGE_VERSION='3.15.6'
+PACKAGE_STRING='cpuminer-opt 3.15.6'
 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 3.15.4 to adapt to many kinds of systems.
+\`configure' configures cpuminer-opt 3.15.6 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 3.15.4:";;
+     short | recursive ) echo "Configuration of cpuminer-opt 3.15.6:";;
    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 3.15.4
+cpuminer-opt configure 3.15.6
 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 3.15.4, which was
+It was created by cpuminer-opt $as_me 3.15.6, 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='3.15.4'
+ VERSION='3.15.6'
 
 
 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 3.15.4, which was
+This file was extended by cpuminer-opt $as_me 3.15.6, 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 3.15.4
+cpuminer-opt config.status 3.15.6
 configured by $0, generated by GNU Autoconf 2.69,
   with options \\"\$ac_cs_config\\"
 

configure.ac

@@ -1,4 +1,4 @@
-AC_INIT([cpuminer-opt], [3.15.4])
+AC_INIT([cpuminer-opt], [3.15.6])
 
 AC_PREREQ([2.59c])
 AC_CANONICAL_SYSTEM

cpu-miner.c

@@ -490,8 +490,13 @@ static bool get_mininginfo( CURL *curl, struct work *work )
 	   }
 
       key = json_object_get( res, "networkhashps" );
-	   if ( key && json_is_integer( key ) )
-		   net_hashrate = (double) json_integer_value( key );
+      if ( key )
+      {
+         if ( json_is_integer( key ) )
+            net_hashrate = (double) json_integer_value( key );
+         else if ( json_is_real( key ) )
+            net_hashrate = (double) json_real_value( key );
+      }
 
       key = json_object_get( res, "blocks" );
 	   if ( key && json_is_integer( key ) )
@@ -506,26 +511,7 @@ static bool get_mininginfo( CURL *curl, struct work *work )
 	      // complete missing data from getwork
 	      work->height = (uint32_t) net_blocks + 1;
 	      if ( work->height > g_work.height )
-         {
             restart_threads();
-
-/* redundant with new block log
-            if ( !opt_quiet )
-            {
-		         char netinfo[64] = { 0 };
-		         char srate[32] = { 0 };
-		         sprintf( netinfo, "diff %.2f", net_diff );
-		         if ( net_hashrate )
-               {
-	               format_hashrate( net_hashrate, srate );
-                  strcat( netinfo, ", net " );
-			         strcat( netinfo, srate );
-		         }
-		         applog( LOG_BLUE, "%s block %d, %s",
-			                algo_names[opt_algo], work->height, netinfo );
-		      }
-*/
-         } 
 	   }  // res
 	}
 	json_decref( val );
@@ -920,12 +906,12 @@ static bool gbt_work_decode( const json_t *val, struct work *work )
    tmp = json_object_get( val, "workid" );
    if ( tmp )
    {
-   if ( !json_is_string( tmp ) )
-   {
-      applog( LOG_ERR, "JSON invalid workid" );
-      goto out;
-   }
-   work->workid = strdup( json_string_value( tmp ) );
+      if ( !json_is_string( tmp ) )
+      {
+         applog( LOG_ERR, "JSON invalid workid" );
+         goto out;
+      }
+      work->workid = strdup( json_string_value( tmp ) );
    }
 
    rc = true;
@@ -1078,13 +1064,12 @@ 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.
-                  : exp32 * last_targetdiff
-                      * (double)(submitted_share_count - accepted_share_count )
-                    / (double)uptime.tv_sec;
+                : ltd * (double)(submitted_share_count - accepted_share_count )
+                  / (double)uptime.tv_sec;
       double lost_shrate = share_time == 0. ? 0.
-               : exp32 * last_targetdiff  * (double)(submits - accepts )
-                / share_time;
+               : ltd  * (double)(submits - accepts ) / share_time;
       char lshr_units[4] = {0};
       char lghr_units[4] = {0};
       scale_hash_for_display( &lost_shrate, lshr_units );
@@ -1190,9 +1175,11 @@ static int share_result( int result, struct work *work,
    {
      sprintf( ares, "A%d", accepted_share_count );
      sprintf( bres, "B%d", solved_block_count );
-     stale = work ? work->data[ algo_gate.ntime_index ]
-                 != g_work.data[ algo_gate.ntime_index ] : false; 
-     if ( reason ) stale = stale || strstr( reason, "Invalid job id" );
+     if ( reason )
+        stale = strstr( reason, "job" );
+     else if ( work )
+        stale =  work->data[ algo_gate.ntime_index ]
+             != g_work.data[ algo_gate.ntime_index ];
      if ( stale )
      {
         stale_share_count++;
@@ -1260,14 +1247,13 @@ static int share_result( int result, struct work *work,
    if ( unlikely( !( opt_quiet || result || stale ) ) )
    {
       uint32_t str[8];
+      uint32_t *targ;
 
-      if ( reason )
-         applog( LOG_WARNING, "Reject reason: %s", reason );
+      if ( reason ) applog( LOG_WARNING, "Reject reason: %s", reason );
          
-      // display share hash and target for troubleshooting
       diff_to_hash( str, my_stats.share_diff );
       applog2( LOG_INFO, "Hash:   %08x%08x%08x...", str[7], str[6], str[5] );
-      uint32_t *targ;
+
       if ( work )
          targ = work->target;
       else
@@ -1580,6 +1566,7 @@ start:
          {
             double miner_hr = 0.;
             double net_hr = net_hashrate;
+            double nd = net_diff * exp32;
             char net_hr_units[4] = {0};
             char miner_hr_units[4] = {0};
             char net_ttf[32];
@@ -1594,11 +1581,11 @@ start:
             pthread_mutex_unlock( &stats_lock );
 
             if ( net_hr > 0. )
-               sprintf_et( net_ttf, ( net_diff * exp32 ) / net_hr );
+               sprintf_et( net_ttf, nd / net_hr );
             else
                sprintf( net_ttf, "NA" );
             if ( miner_hr > 0. )
-               sprintf_et( miner_ttf, ( net_diff * exp32 ) / miner_hr );
+               sprintf_et( miner_ttf, nd / miner_hr );
             else
                sprintf( miner_ttf, "NA" );
 
@@ -1848,10 +1835,19 @@ bool submit_solution( struct work *work, const void *hash,
                    work->data[ algo_gate.ntime_index ] );
      }
 
-     if ( unlikely( lowdiff_debug ) )
+     if ( opt_debug )
      {
         uint32_t* h = (uint32_t*)hash;
         uint32_t* t = (uint32_t*)work->target;
+        uint32_t* d = (uint32_t*)work->data;
+
+        unsigned char *xnonce2str = abin2hex( work->xnonce2,
+                                              work->xnonce2_len );
+        applog(LOG_INFO,"Thread %d, Nonce %08x, Xnonce2 %s", thr->id,
+                       work->data[ algo_gate.nonce_index ], xnonce2str );
+        free( xnonce2str );
+        applog(LOG_INFO,"Data[0:19]: %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x", d[0],d[1],d[2],d[3],d[4],d[5],d[6],d[7],d[8],d[9] );
+        applog(LOG_INFO,"          : %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x", d[10],d[11],d[12],d[13],d[14],d[15],d[16],d[17],d[18],d[19]);
         applog(LOG_INFO,"Hash[7:0]: %08x %08x %08x %08x %08x %08x %08x %08x",
                                     h[7],h[6],h[5],h[4],h[3],h[2],h[1],h[0]);
         applog(LOG_INFO,"Targ[7:0]: %08x %08x %08x %08x %08x %08x %08x %08x",
@@ -1956,8 +1952,6 @@ void std_get_new_work( struct work* work, struct work* g_work, int thr_id,
    uint32_t *nonceptr = work->data + algo_gate.nonce_index;
    bool force_new_work = false; 
 
-   pthread_rwlock_rdlock( &g_work_lock );
-   
    if ( have_stratum ) 
       force_new_work = work->job_id ?    strtoul(   work->job_id, NULL, 16 )
                                       != strtoul( g_work->job_id, NULL, 16 )
@@ -1973,8 +1967,6 @@ void std_get_new_work( struct work* work, struct work* g_work, int thr_id,
    }
    else
        ++(*nonceptr);
-
-   pthread_rwlock_unlock( &g_work_lock );
 }
 
 bool std_ready_to_mine( struct work* work, struct stratum_ctx* stratum,
@@ -1990,13 +1982,14 @@ bool std_ready_to_mine( struct work* work, struct stratum_ctx* stratum,
 
 static void stratum_gen_work( struct stratum_ctx *sctx, struct work *g_work )
 {
-   // Safer than testing the job id
-   bool new_job =  *get_stratum_job_ntime()
-                   != g_work->data[ algo_gate.ntime_index ];
+   bool new_job;
 
    pthread_rwlock_wrlock( &g_work_lock );
    pthread_mutex_lock( &sctx->work_lock );
 
+   new_job =  sctx->new_job;
+   sctx->new_job = false;
+   
    free( g_work->job_id );
    g_work->job_id = strdup( sctx->job.job_id );
    g_work->xnonce2_len = sctx->xnonce2_size;
@@ -2009,8 +2002,12 @@ static void stratum_gen_work( struct stratum_ctx *sctx, struct work *g_work )
    g_work->targetdiff = sctx->job.diff
                            / ( opt_target_factor * opt_diff_factor );
    diff_to_hash( g_work->target, g_work->targetdiff );
+
    // Increment extranonce2
-   for ( int t = 0; t < sctx->xnonce2_size && !( ++sctx->job.xnonce2[t] ); t++ );
+   for ( int t = 0;
+         t < sctx->xnonce2_size && !( ++sctx->job.xnonce2[t] );
+         t++ );
+
    g_work_time = time(NULL);
    restart_threads();
 
@@ -2032,7 +2029,7 @@ static void stratum_gen_work( struct stratum_ctx *sctx, struct work *g_work )
    else if ( last_block_height != sctx->block_height )
       applog( LOG_BLUE, "New Block %d, Job %s",
                         sctx->block_height, g_work->job_id );
-   else if ( new_job && g_work->job_id )
+   else if ( g_work->job_id && new_job )
       applog( LOG_BLUE, "New Work: Block %d, Net diff %.5g, Job %s",
                          sctx->block_height, net_diff, g_work->job_id );
    else if ( !opt_quiet )
@@ -2065,11 +2062,12 @@ static void stratum_gen_work( struct stratum_ctx *sctx, struct work *g_work )
 
          if ( likely( hr > 0. ) )
          {
+            double nd = net_diff * exp32;
             char hr_units[4] = {0};
             char block_ttf[32];
             char share_ttf[32];
 
-            sprintf_et( block_ttf, ( net_diff * exp32 ) /  hr );
+            sprintf_et( block_ttf, nd /  hr );
             sprintf_et( share_ttf, ( g_work->targetdiff * exp32 ) / hr );
             scale_hash_for_display ( &hr, hr_units );
             applog2( LOG_INFO, "TTF @ %.2f %sh/s: Block %s, Share %s",
@@ -2085,7 +2083,7 @@ static void stratum_gen_work( struct stratum_ctx *sctx, struct work *g_work )
                     : et.tv_sec / ( last_block_height - session_first_block );
                if ( net_diff && net_ttf )
                {
-                  double net_hr = net_diff * exp32 / net_ttf;
+                  double net_hr = nd / net_ttf;
                   char net_hr_units[4] = {0};
 
                   scale_hash_for_display ( &net_hr, net_hr_units );
@@ -2240,19 +2238,18 @@ static void *miner_thread( void *userdata )
              pthread_rwlock_unlock( &g_work_lock );
           }
 
+          pthread_rwlock_rdlock( &g_work_lock );
+
           algo_gate.get_new_work( &work, &g_work, thr_id, &end_nonce );
+          work_restart[thr_id].restart = 0;
+
+          pthread_rwlock_unlock( &g_work_lock );
 
        } // do_this_thread
-       algo_gate.resync_threads( &work );
+       algo_gate.resync_threads( thr_id, &work );
 
        if ( unlikely( !algo_gate.ready_to_mine( &work, &stratum, thr_id ) ) )
           continue;
-       // conditional mining
-       if ( unlikely( !wanna_mine( thr_id ) ) )
-       {
-          sleep(5);
-	       continue;
-       }
 
 // LP_SCANTIME overrides opt_scantime option, is this right?
 
@@ -2309,7 +2306,6 @@ static void *miner_thread( void *userdata )
        // init time
        if ( firstwork_time == 0 )
           firstwork_time = time(NULL);
-       work_restart[thr_id].restart = 0;
        hashes_done = 0;
        gettimeofday( (struct timeval *) &tv_start, NULL );
 
@@ -2440,6 +2436,14 @@ static void *miner_thread( void *userdata )
 #endif
 	       }
        }  // benchmark
+
+       // conditional mining
+       if ( unlikely( !wanna_mine( thr_id ) ) )
+       {
+          sleep(5);
+          continue;
+       }
+
    }  // miner_thread loop
 
 out:
@@ -2731,7 +2735,6 @@ static void *stratum_thread(void *userdata )
          pthread_rwlock_wrlock( &g_work_lock );
          g_work_time = 0;
          pthread_rwlock_unlock( &g_work_lock );
-//         restart_threads();
          if ( !stratum_connect( &stratum, stratum.url )
               || !stratum_subscribe( &stratum )
               || !stratum_authorize( &stratum, rpc_user, rpc_pass ) )
@@ -2757,9 +2760,7 @@ static void *stratum_thread(void *userdata )
       report_summary_log( ( stratum_diff != stratum.job.diff )
                        && ( stratum_diff != 0. ) );
       
-      if ( stratum.job.job_id && ( !g_work_time
-                             || ( *get_stratum_job_ntime()
-                                != g_work.data[ algo_gate.ntime_index ] ) ) )
+      if ( stratum.new_job )
          stratum_gen_work( &stratum, &g_work );
 
       if ( likely( stratum_socket_full( &stratum, opt_timeout ) ) )

miner.h

@@ -444,7 +444,8 @@ struct stratum_ctx {
 	struct work work __attribute__ ((aligned (64)));
 	pthread_mutex_t work_lock;
 
-	int block_height;
+   int block_height;
+   bool new_job;  
 } __attribute__ ((aligned (64)));
 
 bool stratum_socket_full(struct stratum_ctx *sctx, int timeout);
@@ -642,7 +643,7 @@ static const char* const algo_names[] = {
         "lyra2z330",
         "m7m",
         "minotaur",
-	"myr-gr",
+        "myr-gr",
         "neoscrypt",
         "nist5",
         "pentablake",
@@ -770,7 +771,7 @@ Options:\n\
                           allium        Garlicoin (GRLC)\n\
                           anime         Animecoin (ANI)\n\
                           argon2        Argon2 Coin (AR2)\n\
-                          argon2d250    argon2d-crds, Credits (CRDS)\n\
+                          argon2d250\n\
                           argon2d500    argon2d-dyn, Dynamic (DYN)\n\
                           argon2d4096   argon2d-uis, Unitus (UIS)\n\
                           axiom         Shabal-256 MemoHash\n\
@@ -795,13 +796,13 @@ Options:\n\
                           lyra2h        Hppcoin\n\
                           lyra2re       lyra2\n\
                           lyra2rev2     lyrav2\n\
-                          lyra2rev3     lyrav2v3, Vertcoin\n\
+                          lyra2rev3     lyrav2v3\n\
                           lyra2z\n\
                           lyra2z330     Lyra2 330 rows\n\
                           m7m           Magi (XMG)\n\
                           myr-gr        Myriad-Groestl\n\
                           minotaur      Ringcoin (RNG)\n\
-			  neoscrypt     NeoScrypt(128, 2, 1)\n\
+                          neoscrypt     NeoScrypt(128, 2, 1)\n\
                           nist5         Nist5\n\
                           pentablake    5 x blake512\n\
                           phi1612       phi\n\
@@ -815,7 +816,7 @@ Options:\n\
                           sha256d       Double SHA-256\n\
                           sha256q       Quad SHA-256, Pyrite (PYE)\n\
                           sha256t       Triple SHA-256, Onecoin (OC)\n\
-			                 sha3d         Double Keccak256 (BSHA3)\n\
+                          sha3d         Double Keccak256 (BSHA3)\n\
                           shavite3      Shavite3\n\
                           skein         Skein+Sha (Skeincoin)\n\
                           skein2        Double Skein (Woodcoin)\n\
@@ -874,7 +875,6 @@ Options:\n\
   -s, --scantime=N      upper bound on time spent scanning current work when\n\
                           long polling is unavailable, in seconds (default: 5)\n\
       --randomize       Randomize scan range start to reduce duplicates\n\
-      --reset-on-stale  Workaround reset stratum if too many stale shares\n\
   -f, --diff-factor     Divide req. difficulty by this factor (std is 1.0)\n\
   -m, --diff-multiplier Multiply difficulty by this factor (std is 1.0)\n\
       --hash-meter      Display thread hash rates\n\

util.c

@@ -1048,53 +1048,51 @@ bool fulltest( const uint32_t *hash, const uint32_t *target )
 	return rc;
 }
 
-// Mathmatically the difficulty is simply the reciprocal of the hash.
+// Mathmatically the difficulty is simply the reciprocal of the hash: d = 1/h.
 // Both are real numbers but the hash (target) is represented as a 256 bit
-// number with the upper 32 bits representing the whole integer part and the
-// lower 224 bits representing the fractional part:
+// fixed point number with the upper 32 bits representing the whole integer
+// part and the lower 224 bits representing the fractional part:
 //   target[ 255:224 ] = trunc( 1/diff )
 //   target[ 223:  0 ] = frac( 1/diff )
 //
 // The 256 bit hash is exact but any floating point representation is not.
-// Stratum provides the target difficulty as double precision, inexcact, and
+// Stratum provides the target difficulty as double precision, inexcact,
 // which must be converted to a hash target. The converted hash target will
-// likely be less precise to to inexact input and conversion error.
-// converted to 256 bit hash which will also be inexact and likelyless
-// accurate to to error in conversion.
+// likely be less precise due to inexact input and conversion error.
 // On the other hand getwork provides a 256 bit hash target which is exact.
 //
 // How much precision is needed?
 //
-// 128 bit types are implemented in software by the compiler using 64 bit
+// 128 bit types are implemented in software by the compiler on 64 bit
 // hardware resulting in lower performance and more error than would be
-// expected with a hardware 128 bit implementtaion.
+// expected with a hardware 128 bit implementaion.
 // Float80 exploits the internals of the FP unit which provide a 64 bit
 // mantissa in an 80 bit register with hardware rounding. When the destination
 // is double the data is rounded to float64 format. Long double returns all
 // 80 bits without rounding and including any accumulated computation error.
 // Float80 does not fit efficiently in memory.
 //
-// 256 bit hash: 76
+// Significant digits:
+// 256 bit hash: 76     
 // float:         7     (float32, 80 bits with rounding to 32 bits)
 // double:       15     (float64, 80 bits with rounding to 64 bits)
-// long double   19     (float80, 80 bits with no rounding)
-// __float128    33     (128 bits with no rounding)
+// long double:  19     (float80, 80 bits with no rounding)
+// __float128:   33     (128 bits with no rounding)
 // uint32_t:      9
 // uint64_t:     19
 // uint128_t     38
 //
 // The concept of significant digits doesn't apply to the 256 bit hash
-// representation. It's fixed point making leading zeros significant
-// Leading zeros count in the 256 bit 
+// representation. It's fixed point making leading zeros significant,
+// limiting its range and precision due to fewer zon-zero significant digits.
 //
 // Doing calculations with float128 and uint128 increases precision for
 // target_to_diff, but doesn't help with stratum diff being limited to
 // double precision. Is the extra precision really worth the extra cost?
-//
-// With double the error rate is 1/1e15, or one hash in every Petahash
-// with a very low difficulty, not a likely sitiation. Higher difficulty
-// increases the effective precision. Due to the floating nature of the 
-// decimal point leading zeros aren't counted.
+// With float128 the error rate is 1/1e33 compared with 1/1e15 for double.
+// For double that's 1 error in every petahash with a very low difficulty,
+// not a likely situation. With higher difficulty effective precision
+// increases.
 //
 // Unfortunately I can't get float128 to work so long double (float80) is
 // as precise as it gets.
@@ -2172,7 +2170,8 @@ bool stratum_handle_method(struct stratum_ctx *sctx, const char *s)
 
 	if (!strcasecmp(method, "mining.notify")) {
 		ret = stratum_notify(sctx, params);
-		goto out;
+      sctx->new_job = true;
+      goto out;
 	}
 	if (!strcasecmp(method, "mining.ping")) { // cgminer 4.7.1+
 		if (opt_debug) applog(LOG_DEBUG, "Pool ping");