diff options
Diffstat (limited to 'APEX_1.4/module/fieldsampler/cuda/include')
3 files changed, 225 insertions, 0 deletions
diff --git a/APEX_1.4/module/fieldsampler/cuda/include/common.h b/APEX_1.4/module/fieldsampler/cuda/include/common.h new file mode 100644 index 00000000..02332daa --- /dev/null +++ b/APEX_1.4/module/fieldsampler/cuda/include/common.h @@ -0,0 +1,31 @@ +/* + * Copyright (c) 2008-2015, NVIDIA CORPORATION. All rights reserved. + * + * NVIDIA CORPORATION and its licensors retain all intellectual property + * and proprietary rights in and to this software, related documentation + * and any modifications thereto. Any use, reproduction, disclosure or + * distribution of this software and related documentation without an express + * license agreement from NVIDIA CORPORATION is strictly prohibited. + */ + + +#ifndef __COMMON_H__ +#define __COMMON_H__ + +#define APEX_CUDA_MODULE_PREFIX FieldSampler_ + +#include "ApexCuda.h" +#include "InplaceTypes.h" + +#include "../include/FieldSamplerCommon.h" + + +namespace physx +{ +namespace apex +{ + +} +} // namespace physx::apex + +#endif diff --git a/APEX_1.4/module/fieldsampler/cuda/include/fieldsampler.h b/APEX_1.4/module/fieldsampler/cuda/include/fieldsampler.h new file mode 100644 index 00000000..28e4896c --- /dev/null +++ b/APEX_1.4/module/fieldsampler/cuda/include/fieldsampler.h @@ -0,0 +1,39 @@ +/* + * Copyright (c) 2008-2015, NVIDIA CORPORATION. All rights reserved. + * + * NVIDIA CORPORATION and its licensors retain all intellectual property + * and proprietary rights in and to this software, related documentation + * and any modifications thereto. Any use, reproduction, disclosure or + * distribution of this software and related documentation without an express + * license agreement from NVIDIA CORPORATION is strictly prohibited. + */ + + +APEX_CUDA_SURFACE_3D(surfRefGridAccum) + + +APEX_CUDA_BOUND_KERNEL(APEX_CUDA_KERNEL_DEFAULT_CONFIG, clearKernel, + ((APEX_MEM_BLOCK(float4), g_accumField))((APEX_MEM_BLOCK(float4), g_accumVelocity)) + ) + +APEX_CUDA_BOUND_KERNEL(APEX_CUDA_KERNEL_DEFAULT_CONFIG, composeKernel, + ((APEX_MEM_BLOCK(float4), g_accumField))((APEX_MEM_BLOCK(const float4), g_accumVelocity))((APEX_MEM_BLOCK(const float4), g_velocity))((float, timestep)) + ) + +APEX_CUDA_FREE_KERNEL_3D(APEX_CUDA_KERNEL_DEFAULT_CONFIG, clearGridKernel, + ((uint32_t, numX))((uint32_t, numY))((uint32_t, numZ)) + ) + +APEX_CUDA_BOUND_KERNEL(APEX_CUDA_KERNEL_DEFAULT_CONFIG, applyParticlesKernel, + ((float4*, g_velocity))((const float4*, g_outField)) + ) + +#ifdef APEX_TEST + +APEX_CUDA_BOUND_KERNEL((), testParticleKernel, + ((float4*, g_position))((float4*, g_velocity)) + ((uint32_t*, g_flag)) + ((const float4*, g_initPosition))((const float4*, g_initVelocity)) + ) + +#endif
\ No newline at end of file diff --git a/APEX_1.4/module/fieldsampler/cuda/include/fieldsamplerInc.h b/APEX_1.4/module/fieldsampler/cuda/include/fieldsamplerInc.h new file mode 100644 index 00000000..23a9c3fb --- /dev/null +++ b/APEX_1.4/module/fieldsampler/cuda/include/fieldsamplerInc.h @@ -0,0 +1,155 @@ +/* + * Copyright (c) 2008-2015, NVIDIA CORPORATION. All rights reserved. + * + * NVIDIA CORPORATION and its licensors retain all intellectual property + * and proprietary rights in and to this software, related documentation + * and any modifications thereto. Any use, reproduction, disclosure or + * distribution of this software and related documentation without an express + * license agreement from NVIDIA CORPORATION is strictly prohibited. + */ + + +APEX_CUDA_SURFACE_3D(surfRefFieldSamplerGridAccum) + + +#ifdef FIELD_SAMPLER_SEPARATE_KERNELS + +APEX_CUDA_BOUND_KERNEL(FIELD_SAMPLER_POINTS_KERNEL_CONFIG, fieldSamplerPointsKernel, + ((nvidia::fieldsampler::FieldSamplerKernelParams, baseParams)) + ((APEX_MEM_BLOCK(float4), accumField)) + ((APEX_MEM_BLOCK(float4), accumVelocity)) + ((APEX_MEM_BLOCK(const float4), positionMass)) + ((APEX_MEM_BLOCK(const float4), velocity)) + ((nvidia::fieldsampler::FieldSamplerParamsEx, paramsEx)) + ((InplaceHandle<nvidia::fieldsampler::FieldSamplerQueryParams>, queryParamsHandle)) + ((nvidia::fieldsampler::FieldSamplerKernelMode::Enum, kernelMode)) + ) +APEX_CUDA_FREE_KERNEL_3D(FIELD_SAMPLER_GRID_KERNEL_CONFIG, fieldSamplerGridKernel, + ((nvidia::fieldsampler::FieldSamplerKernelParams, baseParams)) + ((nvidia::fieldsampler::FieldSamplerGridKernelParams, gridParams)) + ((nvidia::fieldsampler::FieldSamplerParamsEx, paramsEx)) + ((InplaceHandle<nvidia::fieldsampler::FieldSamplerQueryParams>, queryParamsHandle)) + ((nvidia::fieldsampler::FieldSamplerKernelMode::Enum, kernelMode)) + ) + +#ifndef __CUDACC__ +#define LAUNCH_FIELD_SAMPLER_KERNEL( launchData ) \ + const ApexCudaConstStorage& _storage_ = *getFieldSamplerCudaConstStorage(); \ + InplaceHandle<nvidia::fieldsampler::FieldSamplerQueryParams> queryParamsHandle = _storage_.mappedHandle( launchData.queryParamsHandle ); \ + uint32_t fieldSamplerCount = launchData.fieldSamplerArray->size(); \ + switch( launchData.kernelType ) \ + { \ + case nvidia::fieldsampler::FieldSamplerKernelType::POINTS: \ + { \ + const nvidia::fieldsampler::FieldSamplerPointsKernelLaunchDataIntl& data = static_cast<const nvidia::fieldsampler::FieldSamplerPointsKernelLaunchDataIntl&>(launchData); \ + const nvidia::fieldsampler::FieldSamplerPointsKernelArgs* kernelArgs = static_cast<const nvidia::fieldsampler::FieldSamplerPointsKernelArgs*>(data.kernelArgs); \ + for (uint32_t i = 0, activeIdx = 0; i < fieldSamplerCount; ++i) \ + { \ + const nvidia::fieldsampler::FieldSamplerWrapperGPU* wrapper = static_cast<const nvidia::fieldsampler::FieldSamplerWrapperGPU* >( (*data.fieldSamplerArray)[i].mFieldSamplerWrapper ); \ + if (wrapper->isEnabled()) \ + { \ + nvidia::fieldsampler::FieldSamplerParamsEx paramsEx; \ + paramsEx.paramsHandle = _storage_.mappedHandle( wrapper->getParamsHandle() ); \ + paramsEx.multiplier = (*data.fieldSamplerArray)[i].mMultiplier; \ + nvidia::fieldsampler::FieldSamplerKernelMode::Enum kernelMode = (++activeIdx == data.activeFieldSamplerCount) ? data.kernelMode : nvidia::fieldsampler::FieldSamplerKernelMode::DEFAULT; \ + ON_LAUNCH_FIELD_SAMPLER_KERNEL( wrapper->getInternalFieldSampler(), wrapper->getInternalFieldSamplerDesc() ); \ + SCENE_CUDA_OBJ(this, fieldSamplerPointsKernel)( data.stream, data.threadCount, \ + *static_cast<const nvidia::fieldsampler::FieldSamplerKernelParams*>(kernelArgs), \ + nvidia::apex::createApexCudaMemRef(kernelArgs->accumField, size_t(data.memRefSize)), \ + nvidia::apex::createApexCudaMemRef(kernelArgs->accumVelocity, size_t(data.memRefSize)), \ + nvidia::apex::createApexCudaMemRef(kernelArgs->positionMass, size_t(data.memRefSize), ApexCudaMemFlags::IN), \ + nvidia::apex::createApexCudaMemRef(kernelArgs->velocity, size_t(data.memRefSize), ApexCudaMemFlags::IN), \ + paramsEx, queryParamsHandle, kernelMode ); \ + } \ + } \ + } \ + return true; \ + case nvidia::fieldsampler::FieldSamplerKernelType::GRID: \ + { \ + const nvidia::fieldsampler::FieldSamplerGridKernelLaunchDataIntl& data = static_cast<const nvidia::fieldsampler::FieldSamplerGridKernelLaunchDataIntl&>(launchData); \ + const nvidia::fieldsampler::FieldSamplerGridKernelArgs* kernelArgs = static_cast<const nvidia::fieldsampler::FieldSamplerGridKernelArgs*>(data.kernelArgs); \ + SCENE_CUDA_OBJ(this, surfRefFieldSamplerGridAccum).bindTo(*data.accumArray, ApexCudaMemFlags::IN_OUT); \ + for (uint32_t i = 0, activeIdx = 0; i < fieldSamplerCount; ++i) \ + { \ + const nvidia::fieldsampler::FieldSamplerWrapperGPU* wrapper = static_cast<const nvidia::fieldsampler::FieldSamplerWrapperGPU* >( (*data.fieldSamplerArray)[i].mFieldSamplerWrapper ); \ + if (wrapper->isEnabled() && wrapper->getInternalFieldSamplerDesc().gridSupportType == FieldSamplerGridSupportTypeIntl::VELOCITY_PER_CELL) \ + { \ + nvidia::fieldsampler::FieldSamplerParamsEx paramsEx; \ + paramsEx.paramsHandle = _storage_.mappedHandle( wrapper->getParamsHandle() ); \ + paramsEx.multiplier = (*data.fieldSamplerArray)[i].mMultiplier; \ + nvidia::fieldsampler::FieldSamplerKernelMode::Enum kernelMode = (++activeIdx == data.activeFieldSamplerCount) ? data.kernelMode : nvidia::fieldsampler::FieldSamplerKernelMode::DEFAULT; \ + ON_LAUNCH_FIELD_SAMPLER_KERNEL( wrapper->getInternalFieldSampler(), wrapper->getInternalFieldSamplerDesc() ); \ + SCENE_CUDA_OBJ(this, fieldSamplerGridKernel)( data.stream, data.threadCountX, data.threadCountY, data.threadCountZ, \ + *static_cast<const nvidia::fieldsampler::FieldSamplerKernelParams*>(kernelArgs), \ + *static_cast<const nvidia::fieldsampler::FieldSamplerGridKernelParams*>(kernelArgs), \ + paramsEx, queryParamsHandle, kernelMode ); \ + } \ + } \ + SCENE_CUDA_OBJ(this, surfRefFieldSamplerGridAccum).unbind(); \ + } \ + return true; \ + default: \ + PX_ALWAYS_ASSERT(); \ + return false; \ + }; +#endif + +#else + +APEX_CUDA_BOUND_KERNEL(FIELD_SAMPLER_POINTS_KERNEL_CONFIG, fieldSamplerPointsKernel, + ((nvidia::fieldsampler::FieldSamplerKernelParams, baseParams)) + ((APEX_MEM_BLOCK(float4), accumField)) + ((APEX_MEM_BLOCK(float4), accumVelocity)) + ((APEX_MEM_BLOCK(const float4), positionMass)) + ((APEX_MEM_BLOCK(const float4), velocity)) + ((InplaceHandle<nvidia::fieldsampler::FieldSamplerParamsExArray>, paramsExArrayHandle)) + ((InplaceHandle<nvidia::fieldsampler::FieldSamplerQueryParams>, queryParamsHandle)) + ((nvidia::fieldsampler::FieldSamplerKernelMode::Enum, kernelMode)) + ) +APEX_CUDA_FREE_KERNEL_3D(FIELD_SAMPLER_GRID_KERNEL_CONFIG, fieldSamplerGridKernel, + ((nvidia::fieldsampler::FieldSamplerKernelParams, baseParams)) + ((nvidia::fieldsampler::FieldSamplerGridKernelParams, gridParams)) + ((InplaceHandle<nvidia::fieldsampler::FieldSamplerParamsExArray>, paramsExArrayHandle)) + ((InplaceHandle<nvidia::fieldsampler::FieldSamplerQueryParams>, queryParamsHandle)) + ((nvidia::fieldsampler::FieldSamplerKernelMode::Enum, kernelMode)) + ) + +#ifndef __CUDACC__ +#define LAUNCH_FIELD_SAMPLER_KERNEL( launchData ) \ + const ApexCudaConstStorage& _storage_ = *getFieldSamplerCudaConstStorage(); \ + InplaceHandle<nvidia::fieldsampler::FieldSamplerParamsExArray> paramsExArrayHandle = _storage_.mappedHandle( launchData.paramsExArrayHandle ); \ + InplaceHandle<nvidia::fieldsampler::FieldSamplerQueryParams> queryParamsHandle = _storage_.mappedHandle( launchData.queryParamsHandle ); \ + switch( launchData.kernelType ) \ + { \ + case nvidia::fieldsampler::FieldSamplerKernelType::POINTS: \ + { \ + const nvidia::fieldsampler::FieldSamplerPointsKernelLaunchDataIntl& data = static_cast<const nvidia::fieldsampler::FieldSamplerPointsKernelLaunchDataIntl&>(launchData); \ + const nvidia::fieldsampler::FieldSamplerPointsKernelArgs* kernelArgs = static_cast<const nvidia::fieldsampler::FieldSamplerPointsKernelArgs*>(data.kernelArgs); \ + SCENE_CUDA_OBJ(this, fieldSamplerPointsKernel)( data.stream, data.threadCount, \ + *static_cast<const nvidia::fieldsampler::FieldSamplerKernelParams*>(kernelArgs), \ + nvidia::apex::createApexCudaMemRef(kernelArgs->accumField, size_t(data.memRefSize)), \ + nvidia::apex::createApexCudaMemRef(kernelArgs->accumVelocity, size_t(data.memRefSize)), \ + nvidia::apex::createApexCudaMemRef(kernelArgs->positionMass, size_t(data.memRefSize), ApexCudaMemFlags::IN), \ + nvidia::apex::createApexCudaMemRef(kernelArgs->velocity, size_t(data.memRefSize), ApexCudaMemFlags::IN), \ + paramsExArrayHandle, queryParamsHandle, data.kernelMode ); \ + } \ + return true; \ + case nvidia::fieldsampler::FieldSamplerKernelType::GRID: \ + { \ + const nvidia::fieldsampler::FieldSamplerGridKernelLaunchDataIntl& data = static_cast<const nvidia::fieldsampler::FieldSamplerGridKernelLaunchDataIntl&>(launchData); \ + const nvidia::fieldsampler::FieldSamplerGridKernelArgs* kernelArgs = static_cast<const nvidia::fieldsampler::FieldSamplerGridKernelArgs*>(data.kernelArgs); \ + SCENE_CUDA_OBJ(this, surfRefFieldSamplerGridAccum).bindTo(*data.accumArray, ApexCudaMemFlags::IN_OUT); \ + SCENE_CUDA_OBJ(this, fieldSamplerGridKernel)( data.stream, data.threadCountX, data.threadCountY, data.threadCountZ, \ + *static_cast<const nvidia::fieldsampler::FieldSamplerKernelParams*>(kernelArgs), \ + *static_cast<const nvidia::fieldsampler::FieldSamplerGridKernelParams*>(kernelArgs), \ + paramsExArrayHandle, queryParamsHandle, data.kernelMode ); \ + SCENE_CUDA_OBJ(this, surfRefFieldSamplerGridAccum).unbind(); \ + } \ + return true; \ + default: \ + PX_ALWAYS_ASSERT(); \ + return false; \ + }; +#endif + +#endif //FIELD_SAMPLER_SEPARATE_KERNELS |