Version 1.5.1 - GTX 20xx support, Cuda 10.1 and some bug fixes

Version 1.5.1 - 14 May 2019
* Support for CUDA 10.1 with drivers 419.x(Windows) and 418.x(Linux)
* Added suppot for Nvidia Turing architecture (GTX 20xx) Untested.
* Fixed : Cannot Disable gpu mining when using config.txton gpu binaries.
  Now you can simply empty the value of "gpu" and "gputhreads" to disable gpu mining on gpu binaries. Ex: "gpu":"" and "gputhreads":""
* Little performance boost on linux binaries (CPU and GPU)
* Added missing screensaver source files
* change line ending in many files

BuildInfo.h

@@ -1,5 +1,6 @@
 #pragma once
 
 #define RH_PROJECT_NAME "rhminer"
-#define RH_PROJECT_VERSION "1.5"
+#define RH_PROJECT_VERSION "1.5.1"
+
 

CHANGELOG.md

@@ -1,5 +1,15 @@
 # rhminer update and bugfix history
 
+Version 1.5.1 - 14 May 2019
+* Support for CUDA 10.1 with drivers 419.x(Windows) and 418.x(Linux)
+* Added suppot for Nvidia Turing architecture (GTX 20xx) Untested.
+* Fixed : Cannot Disable gpu mining when using config.txton gpu binaries. 
+  Now you can simply empty the value of "gpu" and "gputhreads" to disable gpu mining on gpu binaries. Ex: "gpu":"" and "gputhreads":""
+* Little performance boost on linux binaries (CPU and GPU)
+* Added missing screensaver source files
+* change line ending in many files
+
+
 Version 1.5 - 2 April 2019
 * Added ETHman and AwesomeMiner support (miner_getstat1, miner_restart, miner_reboot, miner_file, control_gpu)
   It is strongly recommented to NOT put rhminer in a loop in a script if you intend to use EthMan or AwseomeMiner.

CMakeLists.txt

@@ -28,6 +28,7 @@ set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_SOURCE_DIR}/CMake")
 
 include_directories(./)
 include_directories(..)
+include_directories(/usr/local/cuda/include)
 include_directories(rhminer)
 
 include (CheckCCompilerFlag)

MinersLib/Global.cpp

@@ -287,7 +287,7 @@ bool GlobalMiningPreset::UpdateToDevModeState(string& connectionParams)
 {
     std::lock_guard<std::mutex> g(*devFeeMutex);
 
-    //TODO: change that to lower time, it causes sopt-mining-emails from nanopool
+    // TODO: change that to lower time, it causes sopt-mining-emails from nanopool
 
     if (TimeGetMilliSec() > m_devFeeTimer24hMS)
     {

MinersLib/Pascal/RandomHash_Blake2b.h

@@ -146,7 +146,6 @@ void CUDA_SYM_DECL(RandomHash_blake2b)(RH_StridePtr roundInput, RH_StridePtr out
 	ctx.c = 0;                         // pointer within buffer
 	ctx.outlen = outlen;
 
-    //TODO: optimiz
 	for (i = 0; i < 128; i++)      // zero input block
 		ctx.b[i] = 0;
 
@@ -160,7 +159,6 @@ void CUDA_SYM_DECL(RandomHash_blake2b)(RH_StridePtr roundInput, RH_StridePtr out
 			ctx.c = 0;                 // counter to zero
 		}
 
-        //TODO: optimiz
 		ctx.b[ctx.c++] = ((const uint8_t *) in)[i];
 	}
 

MinersLib/Pascal/RandomHash_Blake2s.h

@@ -62,7 +62,6 @@ inline int CUDA_SYM_DECL(blake2s_compress_SSE2)( blake2s_state *S, const uint8_t
 	uint32_t m[16];
 	uint32_t v[16];
 
-    //TODO: optimiz -> unroll + MACRO
 	for( size_t i = 0; i < 16; ++i )
 		m[i] = load32_SSE2( block + i * sizeof( m[i] ) );
 

MinersLib/Pascal/RandomHash_Cpu.cpp

@@ -54,8 +54,6 @@ struct RH_StrideArrayStruct
 };
 #define RH_StrideArrayStruct_GetAccum(strideArray) (&((RH_StrideArrayStruct*)strideArray)->accum)
 
-
-
 #include "corelib/CommonData.h"
 
 //--------------------------------------------------------------------------------------------------
@@ -80,7 +78,6 @@ inline RH_StridePtr CUDA_SYM(RH_StrideArrayGet)(RH_StridePtrArray strideArrayVar
 CUDA_DECL_HOST_AND_DEVICE
 inline RH_StridePtr CUDA_SYM(RH_StrideArrayAllocOutput)(RandomHash_State* state, U32 initialSize) 
 {
-
     if (state->m_isMidStateRound)
     {
         RHMINER_ASSERT(state->m_stridesAllocIndex + initialSize + 8 < state->m_stridesAllocMidstateBarrier);
@@ -98,7 +95,6 @@ inline RH_StridePtr CUDA_SYM(RH_StrideArrayAllocOutput)(RandomHash_State* state,
     state->m_stridesAllocIndex += initialSize + RH_IDEAL_ALIGNMENT;
     RH_ASSERT(state->m_stridesAllocIndex < RH_STRIDE_BANK_SIZE);
 
-
     RH_STRIDE_SET_SIZE(stride, initialSize);
     RH_STRIDE_INIT_INTEGRITY(stride);
 
@@ -121,7 +117,6 @@ inline void CUDA_SYM(RH_StrideArrayGrow)(RandomHash_State* state, RH_StridePtr s
     state->m_stridesAllocIndex += growSize;
     RH_ASSERT(state->m_stridesAllocIndex < RH_STRIDE_BANK_SIZE);
 
-
     RH_STRIDE_SET_SIZE(stride, RH_STRIDE_GET_SIZE(stride) + growSize);
     RH_STRIDE_INIT_INTEGRITY(stride);
 }
@@ -144,7 +139,6 @@ inline void CUDA_SYM(RH_StrideArrayClose)(RandomHash_State* state, RH_StridePtr
     RH_ASSERT(state->m_stridesAllocIndex < RH_STRIDE_BANK_SIZE);
     RH_ASSERT((size_t(state->m_stridesAllocIndex) % 32) == 0);
 
-
     RH_STRIDE_CHECK_INTEGRITY(stride);
 }
 
@@ -198,37 +192,31 @@ inline void CUDA_DECL_HOST_AND_DEVICE CUDA_SYM(RandomHash_Initialize)(RandomHash
     state->m_strideID = 0;
     RH_STRIDEARRAY_RESET(state->m_round5Phase2PrecalcArray);
 
-
     if (state->m_isCachedOutputs)
     {
-
         state->m_isMidStateRound = true;
 
         RH_ASSERT(state->m_stridesAllocIndex);
         if (state->m_stridesAllocMidstateBarrierNext != RH_STRIDE_BANK_SIZE) 
         {
-
             state->m_stridesAllocIndex = 0;
             state->m_stridesAllocMidstateBarrier = state->m_stridesAllocMidstateBarrierNext;
 
 #ifdef RHMINER_DEBUG_STRIDE_INTEGRITY_CHECK
-
             memset(state->m_stridesInstances + 0, (U8)0xBA, state->m_stridesAllocMidstateBarrierNext); 
             U64* check = (U64*)(state->m_stridesInstances + RH_STRIDE_BANK_SIZE);
             RHMINER_ASSERT(*check == 0xFF55AA44BB8800DDLLU);
 #endif            
         }
         else 
         {
-
             const U32 ReqDelta = 4096;
             state->m_stridesAllocIndex = RHMINER_ALIGN(state->m_stridesAllocIndex, 4096) + ReqDelta;
             state->m_stridesAllocMidstateBarrierNext = state->m_stridesAllocMidstateBarrier;
             state->m_stridesAllocMidstateBarrier = RH_STRIDE_BANK_SIZE;
 
 
 #ifdef RHMINER_DEBUG_STRIDE_INTEGRITY_CHECK
-
             memset(state->m_stridesInstances + state->m_stridesAllocIndex, (U8)0xBA, RH_STRIDE_BANK_SIZE - state->m_stridesAllocIndex);   
             U64* check = (U64*)(state->m_stridesInstances + RH_STRIDE_BANK_SIZE);
             RHMINER_ASSERT(*check == 0xFF55AA44BB8800DDLLU);
@@ -289,7 +277,6 @@ void CUDA_SYM(AllocateArray)(U8*& arrayData, int count)
 
     RH_STRIDEARRAY_GET_MAXSIZE(arrayData) = count;
     RH_STRIDEARRAY_GET_EXTRA(arrayData, memoryboost) = g_memoryBoostLevel;
-
     RH_STRIDEARRAY_GET_EXTRA(arrayData, sseoptimization) = g_sseOptimization;
     RH_CUDA_ERROR_CHECK();
 }
@@ -387,7 +374,6 @@ void CUDA_SYM(RandomHash_SetHeader)(RandomHash_State* state, U8* sourceHeader, U
     RHMINER_ASSERT(PascalHeaderSize <= PascalHeaderSize);
     memcpy(targetInput, sourceHeader, PascalHeaderSize);
 
-
 }
 
 
@@ -500,6 +486,7 @@ void CUDA_SYM_DECL(RandomHash_Compress)(RandomHash_State* state, RH_StridePtrArr
 #ifdef RH_ENABLE_OPTIM_STRIDE_ARRAY_MURMUR3
     MurmurHash3_x86_32_State tstate = *RH_StrideArrayStruct_GetAccum(inputs);
     U32 seed = _CM(MurmurHash3_x86_32_Finalize)(&tstate);
+
 #else
     U32 seed = _CM(RandomHash_ChecksumArray)(inputs);
 #endif
@@ -522,9 +509,6 @@ void CUDA_SYM_DECL(RandomHash_Compress)(RandomHash_State* state, RH_StridePtrArr
 inline void CUDA_SYM_DECL(RandomHash_MiddlePoint)(RandomHash_State* state)
 {
     state->m_midStateNonce = *(U32*)(RH_STRIDE_GET_DATA(state->m_roundInput)+PascalHeaderNoncePosV4(PascalHeaderSize));
-
-
-
     if (!state->m_isMidStateRound)
     {
         const U32 ReqDelta = 4096;
@@ -548,7 +532,6 @@ inline void CUDA_SYM_DECL(RandomHash_Phase_1_push)(RandomHash_State* state, int
     {
         if (in_round == 5)
         {
-
             if (state->m_midStateNonce == *(U32*)(RH_STRIDE_GET_DATA(state->m_roundInput) + PascalHeaderNoncePosV4(PascalHeaderSize)))
             {
                 state->m_skipPhase1 = 1;
@@ -559,7 +542,6 @@ inline void CUDA_SYM_DECL(RandomHash_Phase_1_push)(RandomHash_State* state, int
 
     state->m_data[in_round-1].backup_io_results = state->m_data[in_round-1].io_results;
 
-
     if (in_round == RH_N)
         state->m_data[in_round - 1].io_results = state->m_data[RH_N].parenAndNeighbortOutputs;
     else
@@ -576,14 +558,12 @@ inline void CUDA_SYM_DECL(RandomHash_Phase_1_pop)(RandomHash_State* state, int i
         {
             RH_StridePtrArray testCache = state->m_data[RH_N].parenAndNeighbortOutputs;
 
-
             state->m_isCachedOutputs = false;
             state->m_skipPhase1 = 0;
 
             skipLastUpdate = true;
         }
 
-
         pano = state->m_data[RH_N].parenAndNeighbortOutputs;
     }
     else
@@ -594,15 +574,13 @@ inline void CUDA_SYM_DECL(RandomHash_Phase_1_pop)(RandomHash_State* state, int i
     }
 
 #ifdef RH_ENABLE_OPTIM_STRIDE_ARRAY_MURMUR3
-
     U32 seed;
 
     {
         if (skipLastUpdate)
         {
             _CM(RH_STRIDE_ARRAY_UPDATE_MURMUR3)(pano, RH_STRIDEARRAY_GET_SIZE(pano) - 1); 
 
-
             RH_STRIDEARRAY_PUSHBACK_MANY_ALL(state->m_round5Phase2PrecalcArray, pano);
         }
         else
@@ -614,6 +592,7 @@ inline void CUDA_SYM_DECL(RandomHash_Phase_1_pop)(RandomHash_State* state, int i
     MurmurHash3_x86_32_State tstate = *RH_StrideArrayStruct_GetAccum(pano);
     seed = _CM(MurmurHash3_x86_32_Finalize)(&tstate);
 
+
 #else
     U32 seed = _CM(RandomHash_ChecksumArray)(state->m_data[in_round].parenAndNeighbortOutputs);
 #endif
@@ -649,7 +628,6 @@ void CUDA_SYM_DECL(RandomHash_Phase_2_pop)(RandomHash_State* state, int in_round
     {
         pano = state->m_data[RH_N].parenAndNeighbortOutputs;
 
-
         state->m_isCachedOutputs = true;
     }
     else
@@ -661,7 +639,7 @@ void CUDA_SYM_DECL(RandomHash_Phase_2_pop)(RandomHash_State* state, int in_round
 
     if (in_round == 5)
     {
-
+        
         {
             RH_STRIDEARRAY_PUSHBACK(state->m_round5Phase2PrecalcArray, RH_STRIDEARRAY_GET(pano, RH_STRIDEARRAY_GET_SIZE(pano)-1));
 
@@ -675,7 +653,6 @@ void CUDA_SYM_DECL(RandomHash_Phase_2_pop)(RandomHash_State* state, int in_round
         {
             state->m_data[in_round].first_round_consume = true;
 
-
 
             RH_STRIDEARRAY_PUSHBACK(state->m_round5Phase2PrecalcArray, RH_STRIDEARRAY_GET(pano, RH_STRIDEARRAY_GET_SIZE(pano)-1));
             _CM(RH_STRIDE_ARRAY_UPDATE_MURMUR3)(state->m_round5Phase2PrecalcArray, RH_STRIDEARRAY_GET_SIZE(state->m_round5Phase2PrecalcArray) - 1);
@@ -696,12 +673,10 @@ void CUDA_SYM_DECL(RandomHash_Phase_2_pop)(RandomHash_State* state, int in_round
     _CM(RandomHash_Compress)(state, state->m_data[in_round].roundOutputs, state->m_workBytes, in_round);  
     RH_ASSERT(RH_STRIDE_GET_SIZE(state->m_workBytes) <= 100);
 
-
     if (in_round != RH_N)
     {
         if (in_round == 4 && state->m_isMidStateRound)
         {
-
             if (state->m_stridesAllocMidstateBarrier != RH_STRIDE_BANK_SIZE)
                 state->m_stridesAllocMidstateBarrierNext = RH_STRIDE_BANK_SIZE;
         }
@@ -819,10 +794,8 @@ inline void CUDA_SYM_DECL(RandomHash_end)(RandomHash_State* state, int in_round)
     _CM(RandomHash_Expand)(state, output, in_round, RH_N - in_round, state->m_data[in_round].roundOutputs);
     RH_STRIDEARRAY_RESET(state->m_data[in_round].io_results);
 
-
     RH_STRIDEARRAY_PUSHBACK_MANY_ALL(state->m_data[in_round].io_results, state->m_data[in_round].roundOutputs); 
 
-
     if (in_round == 5)
         _CM(RH_STRIDE_ARRAY_UPDATE_MURMUR3_DUO)(state->m_data[5].roundOutputs, RH_STRIDEARRAY_GET_SIZE(state->m_data[5].roundOutputs) - 1, state->m_round5Phase2PrecalcArray);
 
@@ -846,12 +819,8 @@ inline void CUDA_SYM_DECL(RandomHash_FirstCall_push)(RandomHash_State* state, in
 
 CUDA_DECL_KERNEL void CUDA_SYM(RandomHash_Block0)(RandomHash_State* allStates)
 {
-#ifdef RH_FULLDEBUG_CPU
-    _hash++;
-#endif
-
     CUDA_DECLARE_STATE();
-
+    /*#define RH_B0*/     
     RandomHash_FirstCall_push(state, 5);
     RandomHash_Phase_init(state, 5);
     RandomHash_Phase_1_push(state, 5);
@@ -1123,6 +1092,11 @@ CUDA_DECL_KERNEL void CUDA_SYM(RandomHash_Init)(RandomHash_State* allStates, U8*
 
     if (state->m_isCachedOutputs)
         startNonce = state->m_midStateNonce;
+
+#ifdef RH_SCREEN_SAVER_MODE
+    extern void ScreensaverFeed(U32 nonce);
+    ScreensaverFeed(startNonce);
+#endif
 
     state->m_startNonce = startNonce;
     *(U32*)(RH_STRIDE_GET_DATA(state->m_roundInput) + PascalHeaderNoncePosV4(PascalHeaderSize)) = startNonce;
@@ -1164,9 +1138,6 @@ void RandomHash_Alloc(void** out_ptr, size_t size)
 
 void RandomHash_Search(RandomHash_State* in_state, U8* out_hash, U32 startNonce)
 {
-#ifdef RH_FULLDEBUG_CPU
-    _n = 0;
-#endif
     RandomHash_State* allStates = in_state;
     RandomHash_Init(allStates, out_hash, startNonce);
     RH_CALL_ALL_KERNEL_BLOCKS

MinersLib/Pascal/RandomHash_MD5.h

@@ -30,7 +30,6 @@ error
 
 /* MD5_F, MD5_G and MD5_H are basic MD5 functions: selection, majority, parity */
 
-//todo optimiz -> intrinsics
 #define MD5_F(x, y, z) (((x) & (y)) | ((~x) & (z)))
 #define MD5_G(x, y, z) (((x) & (z)) | ((y) & (~z)))
 #define MD5_H(x, y, z) ((x) ^ (y) ^ (z))

MinersLib/Pascal/RandomHash_MurMur3_32.h

@@ -552,9 +552,12 @@ uint32_t CUDA_SYM_DECL(MurmurHash3_x86_32_Fast)(const U8* key, int len)
     while (key != keyEnd)
     {
         r0 = *(U64*)(key);
-        MURMUR3_BODY((U32)(r0));
+		key += sizeof(U64);
+#if defined(RH_USE_CUDA_MEM_BOOST)
+		RH_PREFETCH_MEM((const char*)key);
+#endif
+		MURMUR3_BODY((U32)(r0));
         MURMUR3_BODY((U32)(r0 >> 32));
-        key += sizeof(U64);
     }
 
     if (m >= 4)

MinersLib/Pascal/RandomHash_core.h

@@ -129,7 +129,6 @@ struct RH_StrideStruct
 
 #else //!CPU
 
-//TODO: Optmiz - Test memcpy
 #define RH_INPLACE_MEMCPY_128(pDst, pSrc, byteCount)                    \
     {U8* end = pDst + byteCount;                                        \
     while(pDst < end)                                                   \

MinersLib/Pascal/RandomHash_def.h

@@ -148,7 +148,10 @@ inline U32 _mm_extract_epi32_( __m128i V)
             m_tmp = _mm_srli_epi32(m_tmp,(32-count)); \
             m = _mm_or_si128(m,m_tmp);}
 
-    #define RH_PREFETCH_MEM(addr) _mm_prefetch((char*)addr,_MM_HINT_T0);
+    //#define RH_PREFETCH_MEM(addr) _mm_prefetch((char*)addr,_MM_HINT_T0);
+	#define RH_PREFETCH_MEM(addr) _mm_prefetch((char*)addr,_MM_HINT_NTA); 
+
+
     #define BIG_CONSTANT(x) (x)
     #define KERNEL_LOG(...) PrintOutCritical(__VA_ARGS__)
     #define KERNEL0_LOG(...) PrintOutCritical(__VA_ARGS__)