15#ifndef GPURECONSTRUCTIONCUDA_H
16#define GPURECONSTRUCTIONCUDA_H
30struct GPUReconstructionCUDAInternals;
43 template <
class T, int32_t I = 0,
typename... Args>
45 template <
class T, int32_t I = 0,
typename... Args>
50 template <
class T,
class S>
66 std::unique_ptr<gpu_reconstruction_kernels::threadContext>
GetThreadContext()
override;
81 void GetITSTraits(std::unique_ptr<o2::its::TrackerTraits>* trackerTraits, std::unique_ptr<o2::its::VertexerTraits>* vertexerTraits, std::unique_ptr<o2::its::TimeFrame>* timeFrame)
override;
93 int32_t genRTC(std::string&
filename, uint32_t& nCompile);
95 void loadKernelModules(
bool perKernel);
96 const char *mRtcSrcExtension =
".src", *mRtcBinExtension =
".o";
o2::gpu::GPUReconstruction * GPUReconstruction_Create_CUDA(const o2::gpu::GPUSettingsDeviceBackend &cfg)
void runKernelBackendInternal(const krnlSetupTime &_xyz, const Args &... args)
~GPUReconstructionCUDABackend() override
void getRTCKernelCalls(std::vector< std::string > &kernels)
friend GPUh() void GPUCommonAlgorithm GPUReconstructionCUDAInternals * mInternals
GPUReconstructionCUDABackend(const GPUSettingsDeviceBackend &cfg)
virtual int32_t GPUChkErrInternal(const int64_t error, const char *file, int32_t line) const override
void runKernelBackend(const krnlSetupArgs< T, I, Args... > &args)
void PrintKernelOccupancies() override
int32_t ExitDevice_Runtime() override
int32_t unregisterMemoryForGPU_internal(const void *ptr) override
bool IsEventDone(deviceEvent *evList, int32_t nEvents=1) override
GPUReconstructionCUDA(const GPUSettingsDeviceBackend &cfg)
size_t WriteToConstantMemory(size_t offset, const void *src, size_t size, int32_t stream=-1, deviceEvent *ev=nullptr) override
void UpdateAutomaticProcessingSettings() override
void startGPUProfiling() override
int32_t InitDevice_Runtime() override
void StreamWaitForEvents(int32_t stream, deviceEvent *evList, int32_t nEvents=1) override
int32_t PrepareTextures() override
void RecordMarker(deviceEvent *ev, int32_t stream) override
bool CanQueryMaxMemory() override
void SynchronizeEvents(deviceEvent *evList, int32_t nEvents=1) override
size_t GPUMemCpy(void *dst, const void *src, size_t size, int32_t stream, int32_t toGPU, deviceEvent *ev=nullptr, deviceEvent *evList=nullptr, int32_t nEvents=1) override
void endGPUProfiling() override
int32_t registerMemoryForGPU_internal(const void *ptr, size_t size) override
int32_t GPUDebug(const char *state="UNKNOWN", int32_t stream=-1, bool force=false) override
void SynchronizeStream(int32_t stream) override
void GetITSTraits(std::unique_ptr< o2::its::TrackerTraits > *trackerTraits, std::unique_ptr< o2::its::VertexerTraits > *vertexerTraits, std::unique_ptr< o2::its::TimeFrame > *timeFrame) override
std::unique_ptr< gpu_reconstruction_kernels::threadContext > GetThreadContext() override
void SynchronizeGPU() override
~GPUReconstructionCUDA() override
void ReleaseEvent(deviceEvent ev) override
virtual void * getGPUPointer(void *ptr)
typedef void(APIENTRYP PFNGLCULLFACEPROC)(GLenum mode)