Merge remote-tracking branch 'origin/master' into release/0.13

Author: chfast

README.md

@@ -179,7 +179,7 @@ All bug reports, pull requests and code reviews are very much welcome.
 
 8. Can I CPU Mine?
 
-  No, use geth, the go program made for ethereum by ethereum.
+   No, use geth, the go program made for ethereum by ethereum.
 
 
 

ethminer/MinerAux.h

@@ -239,7 +239,7 @@ class MinerCLI
 		{
 			m_worktimeout = atoi(argv[++i]);
 		}
-		else if ((arg == "-RH" || arg == "--report-hashrate") && i + 1 < argc)
+		else if ((arg == "-RH" || arg == "--report-hashrate"))
 		{
 			m_report_stratum_hashrate = true;
 		}
@@ -531,7 +531,7 @@ class MinerCLI
 			exit(1);
 #endif
 		}
-		else if (m_minerType == MinerType::CUDA || m_minerType == MinerType::Mixed)
+		if (m_minerType == MinerType::CUDA || m_minerType == MinerType::Mixed)
 		{
 #if ETH_ETHASHCUDA
 			if (m_cudaDeviceCount > 0)
@@ -626,9 +626,9 @@ class MinerCLI
 #endif
 #if ETH_ETHASHCUDA
 			<< " CUDA configuration:" << endl
-			<< "    --cuda-block-size Set the CUDA block work size. Default is " << toString(ethash_cuda_miner::c_defaultBlockSize) << endl
-			<< "    --cuda-grid-size Set the CUDA grid size. Default is " << toString(ethash_cuda_miner::c_defaultGridSize) << endl
-			<< "    --cuda-streams Set the number of CUDA streams. Default is " << toString(ethash_cuda_miner::c_defaultNumStreams) << endl
+			<< "    --cuda-block-size Set the CUDA block work size. Default is " << toString(CUDAMiner::c_defaultBlockSize) << endl
+			<< "    --cuda-grid-size Set the CUDA grid size. Default is " << toString(CUDAMiner::c_defaultGridSize) << endl
+			<< "    --cuda-streams Set the number of CUDA streams. Default is " << toString(CUDAMiner::c_defaultNumStreams) << endl
 			<< "    --cuda-schedule <mode> Set the schedule mode for CUDA threads waiting for CUDA devices to finish work. Default is 'sync'. Possible values are:" << endl
 			<< "        auto  - Uses a heuristic based on the number of active CUDA contexts in the process C and the number of logical processors in the system P. If C > P, then yield else spin." << endl
 			<< "        spin  - Instruct CUDA to actively spin when waiting for results from the device." << endl
@@ -1083,11 +1083,11 @@ class MinerCLI
 #endif
 #endif
 #if ETH_ETHASHCUDA
-	unsigned m_globalWorkSizeMultiplier = ethash_cuda_miner::c_defaultGridSize;
-	unsigned m_localWorkSize = ethash_cuda_miner::c_defaultBlockSize;
+	unsigned m_globalWorkSizeMultiplier = CUDAMiner::c_defaultGridSize;
+	unsigned m_localWorkSize = CUDAMiner::c_defaultBlockSize;
 	unsigned m_cudaDeviceCount = 0;
 	unsigned m_cudaDevices[16];
-	unsigned m_numStreams = ethash_cuda_miner::c_defaultNumStreams;
+	unsigned m_numStreams = CUDAMiner::c_defaultNumStreams;
 	unsigned m_cudaSchedule = 4; // sync
 #endif
 	unsigned m_dagLoadMode = 0; // parallel

libethash-cl/CLMiner.cpp

@@ -271,7 +271,7 @@ void CLMiner::report(uint64_t _nonce, WorkPackage const& _w)
 	// TODO: Why re-evaluating?
 	Result r = EthashAux::eval(_w.seed, _w.header, _nonce);
 	if (r.value < _w.boundary)
-		farm.submitProof(Solution{_nonce, r.mixHash, _w.header, _w.seed, _w.boundary, _w.job, false});
+		farm.submitProof(Solution{_nonce, r.mixHash, _w.header, _w.seed, _w.boundary, _w.job, _w.job_len, false});
 	else {
 		farm.failedSolution();
 		cwarn << "FAILURE: GPU gave incorrect result!";
@@ -351,7 +351,6 @@ void CLMiner::workLoop()
 				else
 					startNonce = randomNonce();
 
-				current = w;
 				auto switchEnd = std::chrono::high_resolution_clock::now();
 				auto globalSwitchTime = std::chrono::duration_cast<std::chrono::milliseconds>(switchEnd - workSwitchStart).count();
 				auto localSwitchTime = std::chrono::duration_cast<std::chrono::microseconds>(switchEnd - localSwitchStart).count();
@@ -367,14 +366,11 @@ void CLMiner::workLoop()
 			if (results[0] > 0)
 			{
 				// Ignore results except the first one.
-				nonce = startNonce + results[1];
+				nonce = current.startNonce + results[1];
 				// Reset search buffer if any solution found.
 				m_queue.enqueueWriteBuffer(m_searchBuffer, CL_FALSE, 0, sizeof(c_zero), &c_zero);
 			}
 
-			// Increase start nonce for following kernel execution.
-			startNonce += m_globalWorkSize;
-
 			// Run the kernel.
 			m_searchKernel.setArg(3, startNonce);
 			m_queue.enqueueNDRangeKernel(m_searchKernel, cl::NullRange, m_globalWorkSize, m_workgroupSize);
@@ -384,6 +380,11 @@ void CLMiner::workLoop()
 			if (nonce != 0)
 				report(nonce, current);
 
+			current = w;        // kernel now processing newest work
+			current.startNonce = startNonce;
+			// Increase start nonce for following kernel execution.
+			startNonce += m_globalWorkSize;
+
 			// Report hash count
 			addHashCount(m_globalWorkSize);
 
@@ -678,6 +679,18 @@ bool CLMiner::init(const h256& seed)
 			return false;
 		}
 
+		//check whether the current dag fits in memory everytime we recreate the DAG
+		cl_ulong result = 0;
+		device.getInfo(CL_DEVICE_GLOBAL_MEM_SIZE, &result);
+		if (result < dagSize)
+		{
+			cnote <<
+			"OpenCL device " << device.getInfo<CL_DEVICE_NAME>()
+							 << " has insufficient GPU memory." << result <<
+							 " bytes of memory found < " << dagSize << " bytes of memory required";	
+			return false;
+		}
+
 		// create buffer for dag
 		try
 		{
@@ -708,8 +721,6 @@ bool CLMiner::init(const h256& seed)
 		ETHCL_LOG("Creating mining buffer");
 		m_searchBuffer = cl::Buffer(m_context, CL_MEM_WRITE_ONLY, (c_maxSearchResults + 1) * sizeof(uint32_t));
 
-		cllog << "Generating DAG";
-
 		uint32_t const work = (uint32_t)(dagSize / sizeof(node));
 		uint32_t fullRuns = work / m_globalWorkSize;
 		uint32_t const restWork = work % m_globalWorkSize;
@@ -719,14 +730,18 @@ bool CLMiner::init(const h256& seed)
 		m_dagKernel.setArg(2, m_dag);
 		m_dagKernel.setArg(3, ~0u);
 
+		auto startDAG = std::chrono::steady_clock::now();
 		for (uint32_t i = 0; i < fullRuns; i++)
 		{
 			m_dagKernel.setArg(0, i * m_globalWorkSize);
 			m_queue.enqueueNDRangeKernel(m_dagKernel, cl::NullRange, m_globalWorkSize, m_workgroupSize);
 			m_queue.finish();
-			cllog << "DAG" << int(100.0f * i / fullRuns) << '%';
 		}
+		auto endDAG = std::chrono::steady_clock::now();
 
+		auto dagTime = std::chrono::duration_cast<std::chrono::milliseconds>(endDAG-startDAG);
+		float gb = (float)dagSize / (1024 * 1024 * 1024);
+		cnote << gb << " GB of DAG data generated in" << dagTime.count() << "ms.";
 	}
 	catch (cl::Error const& err)
 	{