ability to set local and global work per device (OpenCL)

Author: S74nk0

ethminer/MinerAux.h

@@ -281,8 +281,16 @@ class MinerCLI
 				std::vector<std::string> strs;
 				boost::split(strs, argv[++i], boost::is_any_of(","));
 				int i = 0;
-				for (std::vector<std::string>::iterator it = strs.begin(); it != strs.end() && i != 16; ++it, ++i)
-					m_globalWorkSizeMultiplier[i] = stol(*it);
+				for (std::vector<std::string>::iterator it = strs.begin(); it != strs.end() && i != 16; ++it, ++i) {
+					unsigned a = stol(*it);
+					if (a > 0) {
+						m_globalWorkSizeMultiplier[i] = a;
+					}
+					else if (a == 0 && arg == "--cl-global-work") { // cuda set derfault
+						// set default here
+						m_globalWorkSizeMultiplier[i] = ethash_cl_miner::c_defaultGlobalWorkSizeMultiplier;
+					}
+				}
 			}
 			catch (...)
 			{
@@ -294,8 +302,16 @@ class MinerCLI
 				std::vector<std::string> strs;
 				boost::split(strs, argv[++i], boost::is_any_of(","));
 				int i = 0;
-				for (std::vector<std::string>::iterator it = strs.begin(); it != strs.end() && i != 16; ++it, ++i)
-					m_localWorkSize[i] = stol(*it);
+				for (std::vector<std::string>::iterator it = strs.begin(); it != strs.end() && i != 16; ++it, ++i) {
+					unsigned a = stol(*it);
+					if (a > 0) {
+						m_localWorkSize[i] = a;
+					}
+					else if (a == 0 && arg == "--cl-local-work") { // cuda set derfault
+						// set default here
+						m_localWorkSize[i] = ethash_cl_miner::c_defaultLocalWorkSize;
+					}
+				}
 			}
 			catch (...)
 			{

libethash-cl/ethash_cl_miner.cpp

@@ -201,8 +201,8 @@ bool ethash_cl_miner::configureGPU(
 	uint64_t _currentBlock
 )
 {
-	s_workgroupSize = _localWorkSize[0];
-	s_initialGlobalWorkSize = _globalWorkSize[0];
+	s_workgroupSize = _localWorkSize;
+	s_initialGlobalWorkSize = _globalWorkSize;
 	s_allowCPU = _allowCPU;
 	s_extraRequiredGPUMem = _extraGPUMemory;
 
@@ -234,8 +234,10 @@ bool ethash_cl_miner::configureGPU(
 
 bool ethash_cl_miner::s_allowCPU = false;
 unsigned ethash_cl_miner::s_extraRequiredGPUMem;
-unsigned ethash_cl_miner::s_workgroupSize = ethash_cl_miner::c_defaultLocalWorkSize;
-unsigned ethash_cl_miner::s_initialGlobalWorkSize = ethash_cl_miner::c_defaultGlobalWorkSizeMultiplier * ethash_cl_miner::c_defaultLocalWorkSize;
+//unsigned ethash_cl_miner::s_workgroupSize = ethash_cl_miner::c_defaultLocalWorkSize;
+//unsigned ethash_cl_miner::s_initialGlobalWorkSize = ethash_cl_miner::c_defaultGlobalWorkSizeMultiplier * ethash_cl_miner::c_defaultLocalWorkSize;
+unsigned *ethash_cl_miner::s_workgroupSize = nullptr;
+unsigned *ethash_cl_miner::s_initialGlobalWorkSize = nullptr;
 
 bool ethash_cl_miner::searchForAllDevices(function<bool(cl::Device const&)> _callback)
 {
@@ -335,6 +337,7 @@ bool ethash_cl_miner::init(
 	void** hostDAG
 )
 {
+	m_device_id = _deviceId;
 	// get all platforms
 	try
 	{
@@ -399,9 +402,9 @@ bool ethash_cl_miner::init(
 		m_queue = cl::CommandQueue(m_context, device);
 
 		// make sure that global work size is evenly divisible by the local workgroup size
-		m_globalWorkSize = s_initialGlobalWorkSize;
-		if (m_globalWorkSize % s_workgroupSize != 0)
-			m_globalWorkSize = ((m_globalWorkSize / s_workgroupSize) + 1) * s_workgroupSize;
+		m_globalWorkSize = s_initialGlobalWorkSize[_deviceId];
+		if (m_globalWorkSize % s_workgroupSize[_deviceId] != 0)
+			m_globalWorkSize = ((m_globalWorkSize / s_workgroupSize[_deviceId]) + 1) * s_workgroupSize[_deviceId];
 
 		uint64_t dagSize = ethash_get_datasize(_light->block_number);
 		uint32_t dagSize128 = (unsigned)(dagSize / ETHASH_MIX_BYTES);
@@ -411,7 +414,7 @@ bool ethash_cl_miner::init(
 		// note: ETHASH_CL_MINER_KERNEL is simply ethash_cl_miner_kernel.cl compiled
 		// into a byte array by bin2h.cmake. There is no need to load the file by hand in runtime
 		string code(ETHASH_CL_MINER_KERNEL, ETHASH_CL_MINER_KERNEL + ETHASH_CL_MINER_KERNEL_SIZE);
-		addDefinition(code, "GROUP_SIZE", s_workgroupSize);
+		addDefinition(code, "GROUP_SIZE", s_workgroupSize[_deviceId]);
 		addDefinition(code, "DAG_SIZE", dagSize128);
 		addDefinition(code, "LIGHT_SIZE", lightSize64);
 		addDefinition(code, "ACCESSES", ETHASH_ACCESSES);
@@ -495,7 +498,7 @@ bool ethash_cl_miner::init(
 			for (uint32_t i = 0; i < fullRuns; i++)
 			{
 				m_dagKernel.setArg(0, i * m_globalWorkSize);
-				m_queue.enqueueNDRangeKernel(m_dagKernel, cl::NullRange, m_globalWorkSize, s_workgroupSize);
+				m_queue.enqueueNDRangeKernel(m_dagKernel, cl::NullRange, m_globalWorkSize, s_workgroupSize[_deviceId]);
 				m_queue.finish();
 				printf("Generate DAG device OPENCL#%d: %.0f%%\n", _deviceId, 100.0f * (float)i / (float)fullRuns);
 			}
@@ -573,7 +576,7 @@ void ethash_cl_miner::search(uint8_t const* header, uint64_t target, search_hook
 			m_searchKernel.setArg(3, start_nonce);
 
 			// execute it!
-			m_queue.enqueueNDRangeKernel(m_searchKernel, cl::NullRange, m_globalWorkSize, s_workgroupSize);
+			m_queue.enqueueNDRangeKernel(m_searchKernel, cl::NullRange, m_globalWorkSize, s_workgroupSize[m_device_id]);
 
 			pending.push({ start_nonce, buf });
 			buf = (buf + 1) % c_bufferCount;

libethash-cl/ethash_cl_miner.h

@@ -85,10 +85,14 @@ class ethash_cl_miner
 	unsigned m_globalWorkSize;
 	bool m_openclOnePointOne;
 
+	unsigned m_device_id;
+
 	/// The local work size for the search
-	static unsigned s_workgroupSize;
+	// now it is not per group but per device
+	static unsigned *s_workgroupSize; 
 	/// The initial global work size for the searches
-	static unsigned s_initialGlobalWorkSize;
+	// now it is not per group but per device
+	static unsigned *s_initialGlobalWorkSize;
 	/// The target milliseconds per batch for the search. If 0, then no adjustment will happen
 	static unsigned s_msPerBatch;
 	/// Allow CPU to appear as an OpenCL device or not. Default is false