Merge pull request #1 from fireice-uk/dev

Dev

23 files changed, 129 insertions, 57 deletions

diff --git a/xmrstak/backend/amd/amd_gpu/gpu.cpp b/xmrstak/backend/amd/amd_gpu/gpu.cpp
index 79e80bd..9a4ba73 100644
--- a/xmrstak/backend/amd/amd_gpu/gpu.cpp
+++ b/xmrstak/backend/amd/amd_gpu/gpu.cpp
@@ -675,11 +675,18 @@ std::vector<GpuContext> getAMDDevices(int index)
 		}
 
 		std::string devVendor(devVendorVec.data());
-		if( devVendor.find("Advanced Micro Devices") != std::string::npos || devVendor.find("AMD") != std::string::npos)
+
+		bool isAMDDevice = devVendor.find("Advanced Micro Devices") != std::string::npos || devVendor.find("AMD") != std::string::npos;
+		bool isNVIDIADevice = devVendor.find("NVIDIA Corporation") != std::string::npos || devVendor.find("NVIDIA") != std::string::npos;
+
+		std::string selectedOpenCLVendor = xmrstak::params::inst().openCLVendor;
+		if((isAMDDevice && selectedOpenCLVendor == "AMD") || (isNVIDIADevice && selectedOpenCLVendor == "NVIDIA"))
 		{
 			GpuContext ctx;
 			std::vector<char> devNameVec(1024);
 			size_t maxMem;
+			if( devVendor.find("NVIDIA Corporation") != std::string::npos)
+				ctx.isNVIDIA = true;
 
 			if((clStatus = clGetDeviceInfo(device_list[k], CL_DEVICE_MAX_COMPUTE_UNITS, sizeof(int), &(ctx.computeUnits), NULL)) != CL_SUCCESS)
 			{
@@ -699,6 +706,10 @@ std::vector<GpuContext> getAMDDevices(int index)
 				continue;
 			}
 
+			// the allocation for NVIDIA OpenCL is not limited to 1/4 of the GPU memory per allocation
+			if(ctx.isNVIDIA)
+				maxMem = ctx.freeMem;
+
 			if((clStatus = clGetDeviceInfo(device_list[k], CL_DEVICE_NAME, devNameVec.size(), devNameVec.data(), NULL)) != CL_SUCCESS)
 			{
 				printer::inst()->print_msg(L1,"WARNING: %s when calling clGetDeviceInfo to get CL_DEVICE_NAME for device %u.", err_to_str(clStatus), k);
@@ -747,13 +758,15 @@ int getAMDPlatformIdx()
 
 			clGetPlatformInfo(platforms[i], CL_PLATFORM_VENDOR, infoSize, platformNameVec.data(), NULL);
 			std::string platformName(platformNameVec.data());
-			if( platformName.find("Advanced Micro Devices") != std::string::npos ||
+
+			bool isAMDOpenCL = platformName.find("Advanced Micro Devices") != std::string::npos ||
 				platformName.find("Apple") != std::string::npos ||
-				platformName.find("Mesa") != std::string::npos
-			)
+				platformName.find("Mesa") != std::string::npos;
+			bool isNVIDIADevice = platformName.find("NVIDIA Corporation") != std::string::npos || platformName.find("NVIDIA") != std::string::npos;
+			std::string selectedOpenCLVendor = xmrstak::params::inst().openCLVendor;
+			if((isAMDOpenCL && selectedOpenCLVendor == "AMD") || (isNVIDIADevice && selectedOpenCLVendor == "NVIDIA"))
 			{
-
-				printer::inst()->print_msg(L0,"Found AMD platform index id = %i, name = %s",i , platformName.c_str());
+				printer::inst()->print_msg(L0,"Found %s platform index id = %i, name = %s", selectedOpenCLVendor.c_str(), i , platformName.c_str());
 				if(platformName.find("Mesa") != std::string::npos)
 					mesaPlatform = i;
 				else
@@ -819,7 +832,7 @@ size_t InitOpenCL(GpuContext* ctx, size_t num_gpus, size_t platform_idx)
 	std::vector<char> platformNameVec(infoSize);
 	clGetPlatformInfo(PlatformIDList[platform_idx], CL_PLATFORM_VENDOR, infoSize, platformNameVec.data(), NULL);
 	std::string platformName(platformNameVec.data());
-	if( platformName.find("Advanced Micro Devices") == std::string::npos)
+	if(xmrstak::params::inst().openCLVendor == "AMD" && platformName.find("Advanced Micro Devices") == std::string::npos)
 	{
 		printer::inst()->print_msg(L1,"WARNING: using non AMD device: %s", platformName.c_str());
 	}
@@ -907,7 +920,8 @@ size_t InitOpenCL(GpuContext* ctx, size_t num_gpus, size_t platform_idx)
 		{
 			size_t reduced_intensity = (ctx[i].rawIntensity / ctx[i].workSize) * ctx[i].workSize;
 			ctx[i].rawIntensity = reduced_intensity;
-			printer::inst()->print_msg(L0, "WARNING AMD: gpu %d intensity is not a multiple of 'worksize', auto reduce intensity to %d", ctx[i].deviceIdx, int(reduced_intensity));
+			const std::string backendName = xmrstak::params::inst().openCLVendor;
+			printer::inst()->print_msg(L0, "WARNING %s: gpu %d intensity is not a multiple of 'worksize', auto reduce intensity to %d", backendName.c_str(), ctx[i].deviceIdx, int(reduced_intensity));
 		}
 
 		if((ret = InitOpenCLGpu(opencl_ctx, &ctx[i], source_code.c_str())) != ERR_SUCCESS)
diff --git a/xmrstak/backend/amd/amd_gpu/gpu.hpp b/xmrstak/backend/amd/amd_gpu/gpu.hpp
index 0db6c90..5ab80b8 100644
--- a/xmrstak/backend/amd/amd_gpu/gpu.hpp
+++ b/xmrstak/backend/amd/amd_gpu/gpu.hpp
@@ -27,6 +27,7 @@ struct GpuContext
 	size_t workSize;
 	int stridedIndex;
 	int memChunk;
+	bool isNVIDIA = false;
 	int compMode;
 
 	/*Output vars*/
diff --git a/xmrstak/backend/amd/autoAdjust.hpp b/xmrstak/backend/amd/autoAdjust.hpp
index 6df0eea..e7e98d4 100644
--- a/xmrstak/backend/amd/autoAdjust.hpp
+++ b/xmrstak/backend/amd/autoAdjust.hpp
@@ -91,6 +91,7 @@ private:
 		std::string conf;
 		for(auto& ctx : devVec)
 		{
+			size_t minFreeMem = 128u * byteToMiB;
 			/* 1000 is a magic selected limit, the reason is that more than 2GiB memory
 			 * sowing down the memory performance because of TLB cache misses
 			 */
@@ -112,12 +113,26 @@ private:
 				 */
 				maxThreads = 2024u;
 			}
+
+			// NVIDIA optimizations
+			if(
+				ctx.isNVIDIA && (
+					ctx.name.find("P100") != std::string::npos ||
+				    ctx.name.find("V100") != std::string::npos
+				)
+			)
+			{
+				// do not limit the number of threads
+				maxThreads = 40000u;
+				minFreeMem = 512u * byteToMiB;
+			}
+
 			// increase all intensity limits by two for aeon
 			if(::jconf::inst()->GetMiningAlgo() == cryptonight_lite)
 				maxThreads *= 2u;
 
 			// keep 128MiB memory free (value is randomly chosen)
-			size_t availableMem = ctx.freeMem - (128u * byteToMiB);
+			size_t availableMem = ctx.freeMem - minFreeMem;
 			// 224byte extra memory is used per thread for meta data
 			size_t perThread = hashMemSize + 224u;
 			size_t maxIntensity = availableMem / perThread;
@@ -138,7 +153,7 @@ private:
 				// set 8 threads per block (this is a good value for the most gpus)
 				conf += std::string("  { \"index\" : ") + std::to_string(ctx.deviceIdx) + ",\n" +
 					"    \"intensity\" : " + std::to_string(intensity) + ", \"worksize\" : " + std::to_string(8) + ",\n" +
-					"    \"affine_to_cpu\" : false, \"strided_index\" : 1, \"mem_chunk\" : 2,\n"
+					"    \"affine_to_cpu\" : false, \"strided_index\" : " + (ctx.isNVIDIA ? "0" : "1") + ", \"mem_chunk\" : 2,\n"
 					"    \"comp_mode\" : true\n" +
 					"  },\n";
 			}
@@ -151,7 +166,9 @@ private:
 		configTpl.replace("PLATFORMINDEX",std::to_string(platformIndex));
 		configTpl.replace("GPUCONFIG",conf);
 		configTpl.write(params::inst().configFileAMD);
-		printer::inst()->print_msg(L0, "AMD: GPU configuration stored in file '%s'", params::inst().configFileAMD.c_str());
+
+		const std::string backendName = xmrstak::params::inst().openCLVendor;
+		printer::inst()->print_msg(L0, "%s: GPU (OpenCL) configuration stored in file '%s'", backendName.c_str(), params::inst().configFileAMD.c_str());
 	}
 
 	std::vector<GpuContext> devVec;
diff --git a/xmrstak/backend/amd/minethd.cpp b/xmrstak/backend/amd/minethd.cpp
index f15b480..9bc3676 100644
--- a/xmrstak/backend/amd/minethd.cpp
+++ b/xmrstak/backend/amd/minethd.cpp
@@ -137,6 +137,8 @@ std::vector<iBackend*>* minethd::thread_starter(uint32_t threadOffset, miner_wor
 	for (i = 0; i < n; i++)
 	{
 		jconf::inst()->GetThreadConfig(i, cfg);
+
+		const std::string backendName = xmrstak::params::inst().openCLVendor;
 		
 		if(cfg.cpu_aff >= 0)
 		{
@@ -144,10 +146,10 @@ std::vector<iBackend*>* minethd::thread_starter(uint32_t threadOffset, miner_wor
 			printer::inst()->print_msg(L1, "WARNING on macOS thread affinity is only advisory.");
 #endif
 
-			printer::inst()->print_msg(L1, "Starting AMD GPU thread %d, affinity: %d.", i, (int)cfg.cpu_aff);
+			printer::inst()->print_msg(L1, "Starting %s GPU (OpenCL) thread %d, affinity: %d.", backendName.c_str(), i, (int)cfg.cpu_aff);
 		}
 		else
-			printer::inst()->print_msg(L1, "Starting AMD GPU thread %d, no affinity.", i);
+			printer::inst()->print_msg(L1, "Starting %s GPU (OpenCL) thread %d, no affinity.", backendName.c_str(), i);
 
 		minethd* thd = new minethd(pWork, i + threadOffset, &vGpuData[i], cfg);
 		pvThreads->push_back(thd);
diff --git a/xmrstak/backend/backendConnector.cpp b/xmrstak/backend/backendConnector.cpp
index acedbd6..6f80a0f 100644
--- a/xmrstak/backend/backendConnector.cpp
+++ b/xmrstak/backend/backendConnector.cpp
@@ -77,11 +77,12 @@ std::vector<iBackend*>* BackendConnector::thread_starter(miner_work& pWork)
 #ifndef CONF_NO_OPENCL
 	if(params::inst().useAMD)
 	{
-		plugin amdplugin("AMD", "xmrstak_opencl_backend");
+		const std::string backendName = xmrstak::params::inst().openCLVendor;
+		plugin amdplugin(backendName, "xmrstak_opencl_backend");
 		std::vector<iBackend*>* amdThreads = amdplugin.startBackend(static_cast<uint32_t>(pvThreads->size()), pWork, environment::inst());
 		pvThreads->insert(std::end(*pvThreads), std::begin(*amdThreads), std::end(*amdThreads));
 		if(amdThreads->size() == 0)
-			printer::inst()->print_msg(L0, "WARNING: backend AMD disabled.");
+			printer::inst()->print_msg(L0, "WARNING: backend %s (OpenCL) disabled.", backendName.c_str());
 	}
 #endif
 
diff --git a/xmrstak/backend/cpu/crypto/c_blake256.c b/xmrstak/backend/cpu/crypto/c_blake256.c
index ff623dd..e5fadfe 100644
--- a/xmrstak/backend/cpu/crypto/c_blake256.c
+++ b/xmrstak/backend/cpu/crypto/c_blake256.c
@@ -124,7 +124,7 @@ void blake224_init(state *S) {
 }
 
 // datalen = number of bits
-void blake256_update(state *S, const uint8_t *data, uint64_t datalen) {
+void blake256_update(state *S, const uint8_t *data, uint32_t datalen) {
 	int left = S->buflen >> 3;
 	int fill = 64 - left;
 
@@ -155,7 +155,7 @@ void blake256_update(state *S, const uint8_t *data, uint64_t datalen) {
 }
 
 // datalen = number of bits
-void blake224_update(state *S, const uint8_t *data, uint64_t datalen) {
+void blake224_update(state *S, const uint8_t *data, uint32_t datalen) {
 	blake256_update(S, data, datalen);
 }
 
@@ -206,7 +206,7 @@ void blake224_final(state *S, uint8_t *digest) {
 }
 
 // inlen = number of bytes
-void blake256_hash(uint8_t *out, const uint8_t *in, uint64_t inlen) {
+void blake256_hash(uint8_t *out, const uint8_t *in, uint32_t inlen) {
 	state S;
 	blake256_init(&S);
 	blake256_update(&S, in, inlen * 8);
@@ -214,7 +214,7 @@ void blake256_hash(uint8_t *out, const uint8_t *in, uint64_t inlen) {
 }
 
 // inlen = number of bytes
-void blake224_hash(uint8_t *out, const uint8_t *in, uint64_t inlen) {
+void blake224_hash(uint8_t *out, const uint8_t *in, uint32_t inlen) {
 	state S;
 	blake224_init(&S);
 	blake224_update(&S, in, inlen * 8);
@@ -282,13 +282,13 @@ void hmac_blake224_init(hmac_state *S, const uint8_t *_key, uint64_t keylen) {
 }
 
 // datalen = number of bits
-void hmac_blake256_update(hmac_state *S, const uint8_t *data, uint64_t datalen) {
+void hmac_blake256_update(hmac_state *S, const uint8_t *data, uint32_t datalen) {
   // update the inner state
   blake256_update(&S->inner, data, datalen);
 }
 
 // datalen = number of bits
-void hmac_blake224_update(hmac_state *S, const uint8_t *data, uint64_t datalen) {
+void hmac_blake224_update(hmac_state *S, const uint8_t *data, uint32_t datalen) {
   // update the inner state
   blake224_update(&S->inner, data, datalen);
 }
@@ -310,7 +310,7 @@ void hmac_blake224_final(hmac_state *S, uint8_t *digest) {
 }
 
 // keylen = number of bytes; inlen = number of bytes
-void hmac_blake256_hash(uint8_t *out, const uint8_t *key, uint64_t keylen, const uint8_t *in, uint64_t inlen) {
+void hmac_blake256_hash(uint8_t *out, const uint8_t *key, uint64_t keylen, const uint8_t *in, uint32_t inlen) {
 	hmac_state S;
 	hmac_blake256_init(&S, key, keylen);
 	hmac_blake256_update(&S, in, inlen * 8);
@@ -318,7 +318,7 @@ void hmac_blake256_hash(uint8_t *out, const uint8_t *key, uint64_t keylen, const
 }
 
 // keylen = number of bytes; inlen = number of bytes
-void hmac_blake224_hash(uint8_t *out, const uint8_t *key, uint64_t keylen, const uint8_t *in, uint64_t inlen) {
+void hmac_blake224_hash(uint8_t *out, const uint8_t *key, uint64_t keylen, const uint8_t *in, uint32_t inlen) {
 	hmac_state S;
 	hmac_blake224_init(&S, key, keylen);
 	hmac_blake224_update(&S, in, inlen * 8);
diff --git a/xmrstak/backend/cpu/crypto/c_blake256.h b/xmrstak/backend/cpu/crypto/c_blake256.h
index b9c2aad..06c7917 100644
--- a/xmrstak/backend/cpu/crypto/c_blake256.h
+++ b/xmrstak/backend/cpu/crypto/c_blake256.h
@@ -17,27 +17,27 @@ typedef struct {
 void blake256_init(state *);
 void blake224_init(state *);
 
-void blake256_update(state *, const uint8_t *, uint64_t);
-void blake224_update(state *, const uint8_t *, uint64_t);
+void blake256_update(state *, const uint8_t *, uint32_t);
+void blake224_update(state *, const uint8_t *, uint32_t);
 
 void blake256_final(state *, uint8_t *);
 void blake224_final(state *, uint8_t *);
 
-void blake256_hash(uint8_t *, const uint8_t *, uint64_t);
-void blake224_hash(uint8_t *, const uint8_t *, uint64_t);
+void blake256_hash(uint8_t *, const uint8_t *, uint32_t);
+void blake224_hash(uint8_t *, const uint8_t *, uint32_t);
 
 /* HMAC functions: */
 
 void hmac_blake256_init(hmac_state *, const uint8_t *, uint64_t);
 void hmac_blake224_init(hmac_state *, const uint8_t *, uint64_t);
 
-void hmac_blake256_update(hmac_state *, const uint8_t *, uint64_t);
-void hmac_blake224_update(hmac_state *, const uint8_t *, uint64_t);
+void hmac_blake256_update(hmac_state *, const uint8_t *, uint32_t);
+void hmac_blake224_update(hmac_state *, const uint8_t *, uint32_t);
 
 void hmac_blake256_final(hmac_state *, uint8_t *);
 void hmac_blake224_final(hmac_state *, uint8_t *);
 
-void hmac_blake256_hash(uint8_t *, const uint8_t *, uint64_t, const uint8_t *, uint64_t);
-void hmac_blake224_hash(uint8_t *, const uint8_t *, uint64_t, const uint8_t *, uint64_t);
+void hmac_blake256_hash(uint8_t *, const uint8_t *, uint64_t, const uint8_t *, uint32_t);
+void hmac_blake224_hash(uint8_t *, const uint8_t *, uint64_t, const uint8_t *, uint32_t);
 
 #endif /* _BLAKE256_H_ */
diff --git a/xmrstak/backend/cpu/crypto/c_skein.h b/xmrstak/backend/cpu/crypto/c_skein.h
index 6165a2a..86dbc08 100644
--- a/xmrstak/backend/cpu/crypto/c_skein.h
+++ b/xmrstak/backend/cpu/crypto/c_skein.h
@@ -37,7 +37,7 @@ typedef enum
 }
 SkeinHashReturn;
 
-typedef size_t   SkeinDataLength;                /* bit count  type */
+typedef uint32_t SkeinDataLength;                /* bit count  type */
 typedef u08b_t   SkeinBitSequence;               /* bit stream type */
 
 /* "all-in-one" call */
diff --git a/xmrstak/backend/cpu/crypto/cryptonight_aesni.h b/xmrstak/backend/cpu/crypto/cryptonight_aesni.h
index 5203de8..7562de1 100644
--- a/xmrstak/backend/cpu/crypto/cryptonight_aesni.h
+++ b/xmrstak/backend/cpu/crypto/cryptonight_aesni.h
@@ -43,7 +43,7 @@ extern "C"
 {
 	void keccak(const uint8_t *in, int inlen, uint8_t *md, int mdlen);
 	void keccakf(uint64_t st[25], int rounds);
-	extern void(*const extra_hashes[4])(const void *, size_t, char *);
+	extern void(*const extra_hashes[4])(const void *, uint32_t, char *);
 }
 
 // This will shift and xor tmp1 into itself as 4 32-bit vals such as
@@ -429,7 +429,7 @@ inline void cryptonight_monero_tweak(uint64_t* mem_out, __m128i tmp)
 	tmp = _mm_castps_si128(_mm_movehl_ps(_mm_castsi128_ps(tmp), _mm_castsi128_ps(tmp)));
 	uint64_t vh = _mm_cvtsi128_si64(tmp);
 
-	uint8_t x = vh >> 24;
+	uint8_t x = static_cast<uint8_t>(vh >> 24);
 	static const uint16_t table = 0x7531;
 	const uint8_t index = (((x >> 3) & 6) | (x & 1)) << 1;
 	vh ^= ((table >> index) & 0x3) << 28;
diff --git a/xmrstak/backend/cpu/crypto/cryptonight_common.cpp b/xmrstak/backend/cpu/crypto/cryptonight_common.cpp
index ac696dd..3ff3cb9 100644
--- a/xmrstak/backend/cpu/crypto/cryptonight_common.cpp
+++ b/xmrstak/backend/cpu/crypto/cryptonight_common.cpp
@@ -56,23 +56,23 @@ extern "C"
 #include <string.h>
 #endif // _WIN32
 
-void do_blake_hash(const void* input, size_t len, char* output) {
+void do_blake_hash(const void* input, uint32_t len, char* output) {
 	blake256_hash((uint8_t*)output, (const uint8_t*)input, len);
 }
 
-void do_groestl_hash(const void* input, size_t len, char* output) {
+void do_groestl_hash(const void* input, uint32_t len, char* output) {
 	groestl((const uint8_t*)input, len * 8, (uint8_t*)output);
 }
 
-void do_jh_hash(const void* input, size_t len, char* output) {
+void do_jh_hash(const void* input, uint32_t len, char* output) {
 	jh_hash(32 * 8, (const uint8_t*)input, 8 * len, (uint8_t*)output);
 }
 
-void do_skein_hash(const void* input, size_t len, char* output) {
+void do_skein_hash(const void* input, uint32_t len, char* output) {
 	skein_hash(8 * 32, (const uint8_t*)input, 8 * len, (uint8_t*)output);
 }
 
-void (* const extra_hashes[4])(const void *, size_t, char *) = {do_blake_hash, do_groestl_hash, do_jh_hash, do_skein_hash};
+void (* const extra_hashes[4])(const void *, uint32_t, char *) = {do_blake_hash, do_groestl_hash, do_jh_hash, do_skein_hash};
 
 #ifdef _WIN32
 #include "xmrstak/misc/uac.hpp"
diff --git a/xmrstak/backend/cpu/crypto/hash.h b/xmrstak/backend/cpu/crypto/hash.h
index c12d355..2af3309 100644
--- a/xmrstak/backend/cpu/crypto/hash.h
+++ b/xmrstak/backend/cpu/crypto/hash.h
@@ -1,5 +1,7 @@
 #pragma once
 
+#include <stdint.h>
+
 typedef unsigned char BitSequence;
-typedef unsigned long long DataLength;
+typedef uint32_t DataLength;
 typedef enum {SUCCESS = 0, FAIL = 1, BAD_HASHLEN = 2} HashReturn;
diff --git a/xmrstak/backend/cpu/hwlocMemory.cpp b/xmrstak/backend/cpu/hwlocMemory.cpp
index 94d2b53..089570f 100644
--- a/xmrstak/backend/cpu/hwlocMemory.cpp
+++ b/xmrstak/backend/cpu/hwlocMemory.cpp
@@ -30,7 +30,7 @@ void bindMemoryToNUMANode( size_t puId )
 
 	depth = hwloc_get_type_depth(topology, HWLOC_OBJ_PU);
 
-	for( size_t i = 0;
+	for( uint32_t i = 0;
 		i < hwloc_get_nbobjs_by_depth(topology, depth);
 		i++ )
 	{
diff --git a/xmrstak/backend/cpu/minethd.cpp b/xmrstak/backend/cpu/minethd.cpp
index ec3e145..263c83a 100644
--- a/xmrstak/backend/cpu/minethd.cpp
+++ b/xmrstak/backend/cpu/minethd.cpp
@@ -627,22 +627,22 @@ minethd::cn_hash_fun_multi minethd::func_multi_selector(size_t N, bool bHaveAes,
 
 void minethd::double_work_main()
 {
-	multiway_work_main<2>();
+	multiway_work_main<2u>();
 }
 
 void minethd::triple_work_main()
 {
-	multiway_work_main<3>();
+	multiway_work_main<3u>();
 }
 
 void minethd::quad_work_main()
 {
-	multiway_work_main<4>();
+	multiway_work_main<4u>();
 }
 
 void minethd::penta_work_main()
 {
-	multiway_work_main<5>();
+	multiway_work_main<5u>();
 }
 
 template<size_t N>
@@ -656,7 +656,7 @@ void minethd::prep_multiway_work(uint8_t *bWorkBlob, uint32_t **piNonce)
 	}
 }
 
-template<size_t N>
+template<uint32_t N>
 void minethd::multiway_work_main()
 {
 	if(affinity >= 0) //-1 means no affinity
diff --git a/xmrstak/backend/cpu/minethd.hpp b/xmrstak/backend/cpu/minethd.hpp
index 59583b6..85a95d1 100644
--- a/xmrstak/backend/cpu/minethd.hpp
+++ b/xmrstak/backend/cpu/minethd.hpp
@@ -35,7 +35,7 @@ private:
 
 	minethd(miner_work& pWork, size_t iNo, int iMultiway, bool no_prefetch, int64_t affinity);
 
-	template<size_t N>
+	template<uint32_t N>
 	void multiway_work_main();
 
 	template<size_t N>
diff --git a/xmrstak/cli/cli-miner.cpp b/xmrstak/cli/cli-miner.cpp
index c425f04..6118682 100644
--- a/xmrstak/cli/cli-miner.cpp
+++ b/xmrstak/cli/cli-miner.cpp
@@ -80,6 +80,8 @@ void help()
 #ifndef CONF_NO_OPENCL
 	cout<<"  --noAMD                    disable the AMD miner backend"<<endl;
 	cout<<"  --noAMDCache               disable the AMD(OpenCL) cache for precompiled binaries"<<endl;
+	cout<<"  --openCLVendor VENDOR      use OpenCL driver of VENDOR and devices [AMD,NVIDIA]"<<endl;
+	cout<<"                             default: AMD"<<endl;
 	cout<<"  --amd FILE                 AMD backend miner config file"<<endl;
 #endif
 #ifndef CONF_NO_CUDA
@@ -450,6 +452,24 @@ int main(int argc, char *argv[])
 		{
 			params::inst().useAMD = false;
 		}
+		else if(opName.compare("--openCLVendor") == 0)
+		{
+			++i;
+			if( i >=argc )
+			{
+				printer::inst()->print_msg(L0, "No argument for parameter '--openCLVendor' given");
+				win_exit();
+				return 1;
+			}
+			std::string vendor(argv[i]);
+			params::inst().openCLVendor = vendor;
+			if(vendor != "AMD" && vendor != "NVIDIA")
+			{
+				printer::inst()->print_msg(L0, "'--openCLVendor' must be 'AMD' or 'NVIDIA'");
+				win_exit();
+				return 1;
+			}
+		}
 		else if(opName.compare("--noAMDCache") == 0)
 		{
 			params::inst().AMDCache = false;
diff --git a/xmrstak/jconf.cpp b/xmrstak/jconf.cpp
index 5a2d33a..6fad9b4 100644
--- a/xmrstak/jconf.cpp
+++ b/xmrstak/jconf.cpp
@@ -100,6 +100,9 @@ xmrstak_coin_algo coin_algos[] = {
 	{ "croat", cryptonight, cryptonight, 0u, nullptr },
 	{ "cryptonight", cryptonight, cryptonight, 0u, nullptr },
 	{ "cryptonight_lite", cryptonight_lite, cryptonight_lite, 0u, nullptr },
+	{ "cryptonight_heavy", cryptonight_heavy, cryptonight_heavy, 0u, nullptr },
+	{ "cryptonight_lite_v7", cryptonight_aeon, cryptonight_aeon, 0u, nullptr },
+	{ "cryptonight_v7", cryptonight_monero, cryptonight_monero, 0u, nullptr },
 	{ "edollar", cryptonight, cryptonight, 0u, nullptr },
 	{ "electroneum", cryptonight, cryptonight, 0u, nullptr },
 	{ "graft", cryptonight_monero, cryptonight, 8u, nullptr },
diff --git a/xmrstak/misc/console.cpp b/xmrstak/misc/console.cpp
index de5eed3..d961b71 100644
--- a/xmrstak/misc/console.cpp
+++ b/xmrstak/misc/console.cpp
@@ -225,7 +225,7 @@ void printer::print_str(const char* str)
 
 //Do a press any key for the windows folk. *insert any key joke here*
 #ifdef _WIN32
-void win_exit(size_t code)
+void win_exit(int code)
 {
 	size_t envSize = 0;
 	getenv_s(&envSize, nullptr, 0, "XMRSTAK_NOWAIT");
@@ -238,7 +238,7 @@ void win_exit(size_t code)
 }
 
 #else
-void win_exit(size_t code) 
+void win_exit(int code)
 { 
 	std::exit(code);
 }
diff --git a/xmrstak/misc/console.hpp b/xmrstak/misc/console.hpp
index cfbeddd..6717631 100644
--- a/xmrstak/misc/console.hpp
+++ b/xmrstak/misc/console.hpp
@@ -49,4 +49,4 @@ private:
 	FILE* logfile;
 };
 
-void win_exit(size_t code = 1);
+void win_exit(int code = 1);
diff --git a/xmrstak/misc/executor.hpp b/xmrstak/misc/executor.hpp
index 9e145c3..be5ee6c 100644
--- a/xmrstak/misc/executor.hpp
+++ b/xmrstak/misc/executor.hpp
@@ -64,7 +64,7 @@ private:
 	inline bool is_dev_time()
 	{
 		//Add 2 seconds to compensate for connect
-		constexpr size_t dev_portion = double(iDevDonatePeriod) * fDevDonationLevel + 2;
+		constexpr size_t dev_portion = static_cast<size_t>(double(iDevDonatePeriod) * fDevDonationLevel + 2.);
 
 		if(dev_portion < 12) //No point in bothering with less than 10s
 			return false;
diff --git a/xmrstak/misc/telemetry.cpp b/xmrstak/misc/telemetry.cpp
index 28fb522..a3a2a12 100644
--- a/xmrstak/misc/telemetry.cpp
+++ b/xmrstak/misc/telemetry.cpp
@@ -89,8 +89,8 @@ double telemetry::calc_telemetry_data(size_t iLastMilisec, size_t iThread)
 		return nan("");
 
 	double fHashes, fTime;
-	fHashes = iLastestHashCnt - iEarliestHashCnt;
-	fTime = iLastestStamp - iEarliestStamp;
+	fHashes = static_cast<double>(iLastestHashCnt - iEarliestHashCnt);
+	fTime = static_cast<double>(iLastestStamp - iEarliestStamp);
 	fTime /= 1000.0;
 
 	return fHashes / fTime;
diff --git a/xmrstak/net/socket.cpp b/xmrstak/net/socket.cpp
index 7c58a8e..9bc608f 100644
--- a/xmrstak/net/socket.cpp
+++ b/xmrstak/net/socket.cpp
@@ -156,7 +156,8 @@ int plain_socket::recv(char* buf, unsigned int len)
 
 bool plain_socket::send(const char* buf)
 {
-	int pos = 0, slen = strlen(buf);
+	size_t pos = 0;
+	size_t slen = strlen(buf);
 	while (pos != slen)
 	{
 		int ret = ::send(hSocket, buf + pos, slen - pos, 0);
diff --git a/xmrstak/params.hpp b/xmrstak/params.hpp
index 0171f28..6e676cf 100644
--- a/xmrstak/params.hpp
+++ b/xmrstak/params.hpp
@@ -24,6 +24,8 @@ struct params
 	bool AMDCache;
 	bool useNVIDIA;
 	bool useCPU;
+	// user selected OpenCL vendor
+	std::string openCLVendor;
 
 	bool poolUseTls = false;
 	std::string poolURL;
@@ -60,6 +62,7 @@ struct params
 		AMDCache(true),
 		useNVIDIA(true),
 		useCPU(true),
+		openCLVendor("AMD"),
 		configFile("config.txt"),
 		configFilePools("pools.txt"),
 		configFileAMD("amd.txt"),
diff --git a/xmrstak/pools.tpl b/xmrstak/pools.tpl
index 7973f7c..9c38fd6 100644
--- a/xmrstak/pools.tpl
+++ b/xmrstak/pools.tpl
@@ -22,8 +22,6 @@ POOLCONF],
  *
  *    aeon7 (use this for Aeon's new PoW)
  *    croat
- *    cryptonight (try this if your coin is not listed)
- *    cryptonight_lite
  *    edollar
  *    electroneum
  *    graft
@@ -31,8 +29,18 @@ POOLCONF],
  *    intense
  *    karbo
  *    monero7 (use this for Monero's new PoW)
- *    sumokoin
+ *    sumokoin (automatic switch with block version 3 to cryptonight_heavy)
+ *
+ * Native algorithms which not depends on any block versions:
  *
+ *    # 1MiB scratchpad memory
+ *    cryptonight_lite
+ *    cryptonight_lite_v7
+ *    # 2MiB scratchpad memory
+ *    cryptonight
+ *    cryptonight_v7
+ *    # 4MiB scratchpad memory
+ *    cryptonight_heavy
  */
 
 "currency" : "CURRENCY",