miner-burst-1.170820

* Optimization of the algorithm
- Removed support for burstdev pool (poolV2)
- "ShowMsg" parameter was deleted
- "ShowUpdates" parameter was deleted
- "SendBestOnly" parameter was deleted
- "UseFastRcv" parameter was deleted
+ Added statistics file stat.csv (height; baseTarget; deadline)
* bugs fixed

miner.h

@@ -19,6 +19,7 @@ using namespace rapidjson;
 
 #pragma comment(lib,"Ws2_32.lib")
 #include <ws2tcpip.h>
+#include <mswsock.h> // Need for SO_UPDATE_CONNECT_CONTEXT
 
 #include "curses.h" 
 //#include "panel.h" 
@@ -52,12 +53,12 @@ HANDLE hHeap;
 
 bool exit_flag = false;
 #ifdef __AVX2__
-	char const *const version = "v1.170603_AVX2";
+	char const *const version = "v1.170820_AVX2";
 #else
 	#ifdef __AVX__
-		char const *const version = "v1.170603_AVX";
+		char const *const version = "v1.170820_AVX";
 	#else
-		char const *const version = "v1.170603";
+		char const *const version = "v1.170820";
 	#endif
 #endif 
 
@@ -91,22 +92,22 @@ volatile int stopThreads = 0;
 size_t miner_mode = 0;				// ðåæèì ìàéíåðà. 0=ñîëî, 1=ïóë
 size_t cache_size = 100000;			// ðàçìåð êýøà ÷òåíèÿ ïëîòîâ
 std::vector<std::string> paths_dir; // ïóòè
-bool show_msg = false;				// Ïîêàçàòü îáùåíèå ñ ñåðâåðîì â îòïðàâùèêå
-bool show_updates = false;			// Ïîêàçàòü îáùåíèå ñ ñåðâåðîì â àïäåéòåðå
+//bool show_msg = false;				// Ïîêàçàòü îáùåíèå ñ ñåðâåðîì â îòïðàâùèêå
+//bool show_updates = false;			// Ïîêàçàòü îáùåíèå ñ ñåðâåðîì â àïäåéòåðå
 FILE * fp_Log = nullptr;			// óêàçàòåëü íà ëîã-ôàéë
 size_t send_interval = 100;			// âðåìÿ îæèäàíèÿ ìåæäó îòïðàâêàìè
 size_t update_interval = 1000;		// âðåìÿ îæèäàíèÿ ìåæäó àïäåéòàìè
 short win_size_x = 80;
 short win_size_y = 60;
-bool use_fast_rcv = false;
+//bool use_fast_rcv = false;
 bool use_debug = false;
 bool enable_proxy = false;
-bool send_best_only = true;
+//bool send_best_only = true;
 bool use_wakeup = false;
 bool use_log = true;				// Âåñòè ëîã
 bool use_boost = false;				// Èñïîëüçîâàòü ïîâûøåííûé ïðèîðèòåò äëÿ ïîòîêîâ
 bool show_winner = false;			// ïîêàçûâàòü ïîáåäèòåëÿ
-//short can_generate = 1;				// 0 - disable; 1 - can start generate; 2 - already run generate
+//short can_generate = 0;				// 0 - disable; 1 - can start generate; 2 - already run generator
 
 
 SYSTEMTIME cur_time;				// Òåêóùåå âðåìÿ

miner.vcxproj

@@ -175,7 +175,7 @@
       <FavorSizeOrSpeed>Speed</FavorSizeOrSpeed>
       <EnableFiberSafeOptimizations>true</EnableFiberSafeOptimizations>
       <EnableParallelCodeGeneration>true</EnableParallelCodeGeneration>
-      <EnableEnhancedInstructionSet>AdvancedVectorExtensions</EnableEnhancedInstructionSet>
+      <EnableEnhancedInstructionSet>StreamingSIMDExtensions</EnableEnhancedInstructionSet>
       <BrowseInformation>true</BrowseInformation>
       <AssemblerOutput>AssemblyAndSourceCode</AssemblerOutput>
       <OmitFramePointers>true</OmitFramePointers>

mshabal256_avx2.c

@@ -44,6 +44,7 @@ extern "C" {
     const unsigned char *buf6, const unsigned char *buf7,
     size_t num)
   {
+    _mm256_zeroupper();
     union {
       u32 words[64 * MSHABAL256_FACTOR];
       __m256i data[16];
@@ -235,6 +236,7 @@ extern "C" {
       _mm256_storeu_si256((__m256i *)sc->state + j + 28, C[j]);
     }
 
+	_mm256_zeroupper();
 #undef M
   }
 

mshabal_avx1.c

@@ -222,7 +222,7 @@ extern "C" {
 		  _mm_storeu_si128((__m128i *)sc->state + j + 12, B[j]);
 		  _mm_storeu_si128((__m128i *)sc->state + j + 28, C[j]);
 	  }
-
+	 // _mm256_zeroupper();
 #undef M
   }
 
@@ -265,53 +265,6 @@ extern "C" {
 	  sc->out_size = out_size;
   }
 
-
-
-  static const mshabal_u32 A_init_256[] = {
-	  C32(0x52F84552), C32(0xE54B7999), C32(0x2D8EE3EC), C32(0xB9645191),
-	  C32(0xE0078B86), C32(0xBB7C44C9), C32(0xD2B5C1CA), C32(0xB0D2EB8C),
-	  C32(0x14CE5A45), C32(0x22AF50DC), C32(0xEFFDBC6B), C32(0xEB21B74A)
-  };
-
-  static const mshabal_u32 B_init_256[] = {
-	  C32(0xB555C6EE), C32(0x3E710596), C32(0xA72A652F), C32(0x9301515F),
-	  C32(0xDA28C1FA), C32(0x696FD868), C32(0x9CB6BF72), C32(0x0AFE4002),
-	  C32(0xA6E03615), C32(0x5138C1D4), C32(0xBE216306), C32(0xB38B8890),
-	  C32(0x3EA8B96B), C32(0x3299ACE4), C32(0x30924DD4), C32(0x55CB34A5)
-  };
-
-  static const mshabal_u32 C_init_256[] = {
-	  C32(0xB405F031), C32(0xC4233EBA), C32(0xB3733979), C32(0xC0DD9D55),
-	  C32(0xC51C28AE), C32(0xA327B8E1), C32(0x56C56167), C32(0xED614433),
-	  C32(0x88B59D60), C32(0x60E2CEBA), C32(0x758B4B8B), C32(0x83E82A7F),
-	  C32(0xBC968828), C32(0xE6E00BF7), C32(0xBA839E55), C32(0x9B491C60)
-  };
-
-
-  void  avx1_mshabal_init2(mshabal_context *cc, unsigned out_size)
-  {
-	  //for (unsigned u = 0; u < 176; u++)  cc->state[u] = 0;
-
-	  memset(cc->buf0, 0, sizeof cc->buf0);
-	  memset(cc->buf1, 0, sizeof cc->buf1);
-	  memset(cc->buf2, 0, sizeof cc->buf2);
-	  memset(cc->buf3, 0, sizeof cc->buf3);
-
-	  for (unsigned u = 0; u < 176; u = u + 12 + 16 + 16)
-	  {
-		  memcpy(cc->state + u, A_init_256, sizeof(u32) * 12);
-		  memcpy(cc->state + u + 12, B_init_256, sizeof(u32) * 16);
-		  memcpy(cc->state + u + 12 + 16, C_init_256, sizeof(u32) * 16);
-	  }
-
-	  cc->Wlow = 1;
-	  cc->Whigh = 0;
-	  cc->ptr = 0;
-	  //cc->Whigh = cc->Wlow = C32(0xFFFFFFFF);
-	  cc->out_size = out_size;
-  }
-
-
 
   /* see shabal_small.h */
   void  avx1_mshabal(mshabal_context *sc, const void *data0, const void *data1,
@@ -435,7 +388,6 @@ extern "C" {
 		  out = (u32*)dst3;
 		  for (z = 0; z < out_size_w32; z++)  out[(size_t)z] = sc->state[off + (size_t)(z << 2) + 3];
 	  }
-	  _mm256_zeroupper();
   }
   //#pragma optimize("", on)
 

rapidjson/allocators.h

@@ -179,9 +179,10 @@ class MemoryPoolAllocator {
 
         size = RAPIDJSON_ALIGN(size);
         if (chunkHead_ == 0 || chunkHead_->size + size > chunkHead_->capacity)
-            AddChunk(chunk_capacity_ > size ? chunk_capacity_ : size);
+            if (!AddChunk(chunk_capacity_ > size ? chunk_capacity_ : size))
+                return NULL;
 
-        void *buffer = reinterpret_cast<char *>(chunkHead_ + 1) + chunkHead_->size;
+        void *buffer = reinterpret_cast<char *>(chunkHead_) + RAPIDJSON_ALIGN(sizeof(ChunkHeader)) + chunkHead_->size;
         chunkHead_->size += size;
         return buffer;
     }
@@ -194,26 +195,30 @@ class MemoryPoolAllocator {
         if (newSize == 0)
             return NULL;
 
+        originalSize = RAPIDJSON_ALIGN(originalSize);
+        newSize = RAPIDJSON_ALIGN(newSize);
+
         // Do not shrink if new size is smaller than original
         if (originalSize >= newSize)
             return originalPtr;
 
         // Simply expand it if it is the last allocation and there is sufficient space
-        if (originalPtr == (char *)(chunkHead_ + 1) + chunkHead_->size - originalSize) {
+        if (originalPtr == reinterpret_cast<char *>(chunkHead_) + RAPIDJSON_ALIGN(sizeof(ChunkHeader)) + chunkHead_->size - originalSize) {
             size_t increment = static_cast<size_t>(newSize - originalSize);
-            increment = RAPIDJSON_ALIGN(increment);
             if (chunkHead_->size + increment <= chunkHead_->capacity) {
                 chunkHead_->size += increment;
                 return originalPtr;
             }
         }
 
         // Realloc process: allocate and copy memory, do not free original buffer.
-        void* newBuffer = Malloc(newSize);
-        RAPIDJSON_ASSERT(newBuffer != 0);   // Do not handle out-of-memory explicitly.
-        if (originalSize)
-            std::memcpy(newBuffer, originalPtr, originalSize);
-        return newBuffer;
+        if (void* newBuffer = Malloc(newSize)) {
+            if (originalSize)
+                std::memcpy(newBuffer, originalPtr, originalSize);
+            return newBuffer;
+        }
+        else
+            return NULL;
     }
 
     //! Frees a memory block (concept Allocator)
@@ -227,15 +232,20 @@ class MemoryPoolAllocator {
 
     //! Creates a new chunk.
     /*! \param capacity Capacity of the chunk in bytes.
+        \return true if success.
     */
-    void AddChunk(size_t capacity) {
+    bool AddChunk(size_t capacity) {
         if (!baseAllocator_)
-            ownBaseAllocator_ = baseAllocator_ = RAPIDJSON_NEW(BaseAllocator());
-        ChunkHeader* chunk = reinterpret_cast<ChunkHeader*>(baseAllocator_->Malloc(sizeof(ChunkHeader) + capacity));
-        chunk->capacity = capacity;
-        chunk->size = 0;
-        chunk->next = chunkHead_;
-        chunkHead_ =  chunk;
+            ownBaseAllocator_ = baseAllocator_ = RAPIDJSON_NEW(BaseAllocator)();
+        if (ChunkHeader* chunk = reinterpret_cast<ChunkHeader*>(baseAllocator_->Malloc(RAPIDJSON_ALIGN(sizeof(ChunkHeader)) + capacity))) {
+            chunk->capacity = capacity;
+            chunk->size = 0;
+            chunk->next = chunkHead_;
+            chunkHead_ =  chunk;
+            return true;
+        }
+        else
+            return false;
     }
 
     static const int kDefaultChunkCapacity = 64 * 1024; //!< Default chunk capacity.