reduce compiler warnings and avoid clash with KOKKOS (5a3a5d86) · Commits · 郑智淋 / lammps

lib/gpu/cudpp_mini/cudpp_maximal_launch.cpp

+5 −14

Original line number	Diff line number	Diff line
		@@ -8,16 +8,7 @@
		// in the root directory of this source distribution.
		// -------------------------------------------------------------
		#include "cudpp_maximal_launch.h"

		inline size_t min(size_t x, size_t y)
		{
		return (x <= y) ? x : y;
		}

		inline size_t max(size_t x, size_t y)
		{
		return (x >= y) ? x : y;
		}
		#include <algorithm>

		// computes next highest multiple of f from x
		inline size_t multiple(size_t x, size_t f)
		@@ -65,7 +56,7 @@ size_t maxBlocks(cudaFuncAttributes &attribs,
		size_t ctaLimitSMem = smemPerCTA > 0 ? devprop.sharedMemPerBlock / smemPerCTA : maxBlocksPerSM;
		size_t ctaLimitThreads = maxThreadsPerSM / threadsPerBlock;

		return devprop.multiProcessorCount * min(ctaLimitRegs, min(ctaLimitSMem, min(ctaLimitThreads, maxBlocksPerSM)));
		return devprop.multiProcessorCount * std::min(ctaLimitRegs, std::min(ctaLimitSMem, std::min(ctaLimitThreads, (size_t)maxBlocksPerSM)));
		}

		extern "C"
		@@ -80,15 +71,15 @@ size_t maxBlocksFromPointer(void* kernel,
		{
		err = cudaGetDeviceProperties(&devprop, deviceID);
		if (err != cudaSuccess)
		return -1;
		return (size_t)-1;

		cudaFuncAttributes attr;
		err = cudaFuncGetAttributes(&attr, (const char*)kernel);
		if (err != cudaSuccess)
		return -1;
		return (size_t)-1;

		return maxBlocks(attr, devprop, bytesDynamicSharedMem, threadsPerBlock);
		}

		return -1;
		return (size_t)-1;
		}

+1 −1

Original line number	Diff line number	Diff line
		@@ -244,7 +244,7 @@ class UCL_Kernel {
		template <class dtype>
		inline void add_arg(const dtype* const arg) {
		#if CUDA_VERSION >= 4000
		_kernel_args[_num_args]=const_cast<dtype * const>(arg);
		_kernel_args[_num_args]=const_cast<dtype *>(arg);
		#else
		_param_size = (_param_size+__alignof(dtype)-1) & ~(__alignof(dtype)-1);
		CU_SAFE_CALL(cuParamSetv(_kernel,_param_size,(void*)arg,sizeof(dtype)));