| /dports/math/faiss/faiss-1.7.1/faiss/gpu/utils/ |
| H A D | BlockSelectKernel.cuh | 21 int ThreadsPerBlock>
29 constexpr int kNumWarps = ThreadsPerBlock / kWarpSize; in blockSelect()
41 ThreadsPerBlock> in blockSelect()
53 for (; i < limit; i += ThreadsPerBlock) { in blockSelect()
55 inStart += ThreadsPerBlock; in blockSelect()
77 int ThreadsPerBlock>
86 constexpr int kNumWarps = ThreadsPerBlock / kWarpSize; in blockSelectPair()
98 ThreadsPerBlock> in blockSelectPair()
111 for (; i < limit; i += ThreadsPerBlock) { in blockSelectPair()
113 inKStart += ThreadsPerBlock; in blockSelectPair()
[all …]
|
| H A D | Select.cuh | 146 int ThreadsPerBlock>
148 static constexpr int kNumWarps = ThreadsPerBlock / kWarpSize;
165 utils::isPowerOf2(ThreadsPerBlock), in BlockSelect()
344 int ThreadsPerBlock>
345 struct BlockSelect<K, V, Dir, Comp, 1, NumThreadQ, ThreadsPerBlock> {
346 static constexpr int kNumWarps = ThreadsPerBlock / kWarpSize;
438 int ThreadsPerBlock>
449 utils::isPowerOf2(ThreadsPerBlock), in WarpSelect()
595 int ThreadsPerBlock>
596 struct WarpSelect<K, V, Dir, Comp, 1, NumThreadQ, ThreadsPerBlock> {
[all …]
|
| H A D | WarpSelectKernel.cuh | 21 int ThreadsPerBlock>
29 constexpr int kNumWarps = ThreadsPerBlock / kWarpSize; in warpSelect()
38 ThreadsPerBlock> in warpSelect()
|
| /dports/devel/llvm80/llvm-8.0.1.src/projects/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 103 std::vector<int> ThreadsPerBlock; member in RTLDeviceInfoTy
208 ThreadsPerBlock.resize(NumberOfDevices); in RTLDeviceInfoTy()
311 DeviceInfo.ThreadsPerBlock[device_id] = RTLDeviceInfoTy::DefaultNumThreads; in __tgt_rtl_init_device()
313 DeviceInfo.ThreadsPerBlock[device_id] = maxBlockDimX; in __tgt_rtl_init_device()
316 DeviceInfo.ThreadsPerBlock[device_id] = RTLDeviceInfoTy::HardThreadLimit; in __tgt_rtl_init_device()
365 DeviceInfo.ThreadsPerBlock[device_id]) { in __tgt_rtl_init_device()
366 DeviceInfo.NumTeams[device_id] = DeviceInfo.ThreadsPerBlock[device_id]; in __tgt_rtl_init_device()
368 DeviceInfo.ThreadsPerBlock[device_id]); in __tgt_rtl_init_device()
669 if (cudaThreadsPerBlock > DeviceInfo.ThreadsPerBlock[device_id]) { in __tgt_rtl_run_target_team_region()
670 cudaThreadsPerBlock = DeviceInfo.ThreadsPerBlock[device_id]; in __tgt_rtl_run_target_team_region()
[all …]
|
| /dports/devel/llvm10/llvm-10.0.1.src/projects/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 100 std::vector<int> ThreadsPerBlock; member in RTLDeviceInfoTy
208 ThreadsPerBlock.resize(NumberOfDevices); in RTLDeviceInfoTy()
320 DeviceInfo.ThreadsPerBlock[device_id] = RTLDeviceInfoTy::DefaultNumThreads; in __tgt_rtl_init_device()
322 DeviceInfo.ThreadsPerBlock[device_id] = maxBlockDimX; in __tgt_rtl_init_device()
325 DeviceInfo.ThreadsPerBlock[device_id] = RTLDeviceInfoTy::HardThreadLimit; in __tgt_rtl_init_device()
374 DeviceInfo.ThreadsPerBlock[device_id]) { in __tgt_rtl_init_device()
375 DeviceInfo.NumTeams[device_id] = DeviceInfo.ThreadsPerBlock[device_id]; in __tgt_rtl_init_device()
377 DeviceInfo.ThreadsPerBlock[device_id]); in __tgt_rtl_init_device()
698 if (cudaThreadsPerBlock > DeviceInfo.ThreadsPerBlock[device_id]) { in __tgt_rtl_run_target_team_region()
699 cudaThreadsPerBlock = DeviceInfo.ThreadsPerBlock[device_id]; in __tgt_rtl_run_target_team_region()
[all …]
|
| /dports/devel/tinygo/tinygo-0.14.1/llvm-project/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 100 std::vector<int> ThreadsPerBlock; member in RTLDeviceInfoTy
208 ThreadsPerBlock.resize(NumberOfDevices); in RTLDeviceInfoTy()
320 DeviceInfo.ThreadsPerBlock[device_id] = RTLDeviceInfoTy::DefaultNumThreads; in __tgt_rtl_init_device()
322 DeviceInfo.ThreadsPerBlock[device_id] = maxBlockDimX; in __tgt_rtl_init_device()
325 DeviceInfo.ThreadsPerBlock[device_id] = RTLDeviceInfoTy::HardThreadLimit; in __tgt_rtl_init_device()
374 DeviceInfo.ThreadsPerBlock[device_id]) { in __tgt_rtl_init_device()
375 DeviceInfo.NumTeams[device_id] = DeviceInfo.ThreadsPerBlock[device_id]; in __tgt_rtl_init_device()
377 DeviceInfo.ThreadsPerBlock[device_id]); in __tgt_rtl_init_device()
698 if (cudaThreadsPerBlock > DeviceInfo.ThreadsPerBlock[device_id]) { in __tgt_rtl_run_target_team_region()
699 cudaThreadsPerBlock = DeviceInfo.ThreadsPerBlock[device_id]; in __tgt_rtl_run_target_team_region()
[all …]
|
| /dports/devel/llvm90/llvm-9.0.1.src/projects/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 102 std::vector<int> ThreadsPerBlock; member in RTLDeviceInfoTy
210 ThreadsPerBlock.resize(NumberOfDevices); in RTLDeviceInfoTy()
322 DeviceInfo.ThreadsPerBlock[device_id] = RTLDeviceInfoTy::DefaultNumThreads; in __tgt_rtl_init_device()
324 DeviceInfo.ThreadsPerBlock[device_id] = maxBlockDimX; in __tgt_rtl_init_device()
327 DeviceInfo.ThreadsPerBlock[device_id] = RTLDeviceInfoTy::HardThreadLimit; in __tgt_rtl_init_device()
376 DeviceInfo.ThreadsPerBlock[device_id]) { in __tgt_rtl_init_device()
377 DeviceInfo.NumTeams[device_id] = DeviceInfo.ThreadsPerBlock[device_id]; in __tgt_rtl_init_device()
379 DeviceInfo.ThreadsPerBlock[device_id]); in __tgt_rtl_init_device()
700 if (cudaThreadsPerBlock > DeviceInfo.ThreadsPerBlock[device_id]) { in __tgt_rtl_run_target_team_region()
701 cudaThreadsPerBlock = DeviceInfo.ThreadsPerBlock[device_id]; in __tgt_rtl_run_target_team_region()
[all …]
|
| /dports/math/faiss/faiss-1.7.1/faiss/gpu/impl/ |
| H A D | IVFUtilsSelect1.cu | 24 template <int ThreadsPerBlock, int NumWarpQ, int NumThreadQ, bool Dir>
32 constexpr int kNumWarps = ThreadsPerBlock / kWarpSize; in pass1SelectLists()
45 ThreadsPerBlock> in pass1SelectLists()
|
| H A D | IVFUtilsSelect2.cu | 52 template <int ThreadsPerBlock, int NumWarpQ, int NumThreadQ, bool Dir>
63 constexpr int kNumWarps = ThreadsPerBlock / kWarpSize; in pass2SelectLists()
76 ThreadsPerBlock> in pass2SelectLists()
|
| H A D | IVFInterleaved.cu | 18 template <int ThreadsPerBlock, int NumWarpQ, int NumThreadQ>
31 constexpr int kNumWarps = ThreadsPerBlock / kWarpSize; in ivfInterleavedScan2()
46 ThreadsPerBlock> in ivfInterleavedScan2()
|
| H A D | L2Select.cu | 128 template <typename T, int NumWarpQ, int NumThreadQ, int ThreadsPerBlock>
137 constexpr int kNumWarps = ThreadsPerBlock / kWarpSize; in l2SelectMinK()
149 ThreadsPerBlock> in l2SelectMinK()
|
| H A D | IVFInterleaved.cuh | 36 int ThreadsPerBlock,
54 constexpr int kNumWarps = ThreadsPerBlock / kWarpSize; in ivfInterleavedScan()
91 ThreadsPerBlock> in ivfInterleavedScan()
|
| /dports/devel/wasi-compiler-rt13/llvm-project-13.0.1.src/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 143 int ThreadsPerBlock = 0; member
613 DeviceData[DeviceId].ThreadsPerBlock = MaxBlockDimX; in initDevice()
616 DeviceData[DeviceId].ThreadsPerBlock > EnvTeamThreadLimit) { in initDevice()
619 DeviceData[DeviceId].ThreadsPerBlock, EnvTeamThreadLimit); in initDevice()
620 DeviceData[DeviceId].ThreadsPerBlock = EnvTeamThreadLimit; in initDevice()
700 DeviceData[DeviceId].ThreadsPerBlock) { in initDevice()
702 DeviceData[DeviceId].ThreadsPerBlock); in initDevice()
703 DeviceData[DeviceId].NumTeams = DeviceData[DeviceId].ThreadsPerBlock; in initDevice()
1048 if (CudaThreadsPerBlock > DeviceData[DeviceId].ThreadsPerBlock) { in runTargetTeamRegion()
1050 DeviceData[DeviceId].ThreadsPerBlock); in runTargetTeamRegion()
[all …]
|
| /dports/devel/wasi-libcxx/llvm-project-13.0.1.src/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 143 int ThreadsPerBlock = 0; member
613 DeviceData[DeviceId].ThreadsPerBlock = MaxBlockDimX; in initDevice()
616 DeviceData[DeviceId].ThreadsPerBlock > EnvTeamThreadLimit) { in initDevice()
619 DeviceData[DeviceId].ThreadsPerBlock, EnvTeamThreadLimit); in initDevice()
620 DeviceData[DeviceId].ThreadsPerBlock = EnvTeamThreadLimit; in initDevice()
700 DeviceData[DeviceId].ThreadsPerBlock) { in initDevice()
702 DeviceData[DeviceId].ThreadsPerBlock); in initDevice()
703 DeviceData[DeviceId].NumTeams = DeviceData[DeviceId].ThreadsPerBlock; in initDevice()
1048 if (CudaThreadsPerBlock > DeviceData[DeviceId].ThreadsPerBlock) { in runTargetTeamRegion()
1050 DeviceData[DeviceId].ThreadsPerBlock); in runTargetTeamRegion()
[all …]
|
| /dports/devel/llvm13/llvm-project-13.0.1.src/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 143 int ThreadsPerBlock = 0; member
613 DeviceData[DeviceId].ThreadsPerBlock = MaxBlockDimX; in initDevice()
616 DeviceData[DeviceId].ThreadsPerBlock > EnvTeamThreadLimit) { in initDevice()
619 DeviceData[DeviceId].ThreadsPerBlock, EnvTeamThreadLimit); in initDevice()
620 DeviceData[DeviceId].ThreadsPerBlock = EnvTeamThreadLimit; in initDevice()
700 DeviceData[DeviceId].ThreadsPerBlock) { in initDevice()
702 DeviceData[DeviceId].ThreadsPerBlock); in initDevice()
703 DeviceData[DeviceId].NumTeams = DeviceData[DeviceId].ThreadsPerBlock; in initDevice()
1048 if (CudaThreadsPerBlock > DeviceData[DeviceId].ThreadsPerBlock) { in runTargetTeamRegion()
1050 DeviceData[DeviceId].ThreadsPerBlock); in runTargetTeamRegion()
[all …]
|
| /dports/devel/llvm-devel/llvm-project-f05c95f10fc1d8171071735af8ad3a9e87633120/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 182 int ThreadsPerBlock = 0; member
658 DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::DefaultNumThreads; in initDevice()
661 DeviceData[DeviceId].ThreadsPerBlock = MaxBlockDimX; in initDevice()
664 DeviceData[DeviceId].ThreadsPerBlock > EnvTeamThreadLimit) { in initDevice()
667 DeviceData[DeviceId].ThreadsPerBlock, EnvTeamThreadLimit); in initDevice()
668 DeviceData[DeviceId].ThreadsPerBlock = EnvTeamThreadLimit; in initDevice()
748 DeviceData[DeviceId].ThreadsPerBlock) { in initDevice()
750 DeviceData[DeviceId].ThreadsPerBlock); in initDevice()
1099 if (CudaThreadsPerBlock > DeviceData[DeviceId].ThreadsPerBlock) { in runTargetTeamRegion()
1101 DeviceData[DeviceId].ThreadsPerBlock); in runTargetTeamRegion()
[all …]
|
| /dports/devel/llvm-cheri/llvm-project-37c49ff00e3eadce5d8703fdc4497f28458c64a8/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 124 int ThreadsPerBlock = 0; member
478 DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::DefaultNumThreads; in initDevice()
481 DeviceData[DeviceId].ThreadsPerBlock = MaxBlockDimX; in initDevice()
486 DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::HardThreadLimit; in initDevice()
509 DeviceData[DeviceId].BlocksPerGrid, DeviceData[DeviceId].ThreadsPerBlock, in initDevice()
534 DeviceData[DeviceId].ThreadsPerBlock) { in initDevice()
536 DeviceData[DeviceId].ThreadsPerBlock); in initDevice()
537 DeviceData[DeviceId].NumTeams = DeviceData[DeviceId].ThreadsPerBlock; in initDevice()
883 if (CudaThreadsPerBlock > DeviceData[DeviceId].ThreadsPerBlock) { in runTargetTeamRegion()
885 DeviceData[DeviceId].ThreadsPerBlock); in runTargetTeamRegion()
[all …]
|
| /dports/devel/llvm11/llvm-11.0.1.src/projects/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 124 int ThreadsPerBlock = 0; member
478 DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::DefaultNumThreads; in initDevice()
481 DeviceData[DeviceId].ThreadsPerBlock = MaxBlockDimX; in initDevice()
486 DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::HardThreadLimit; in initDevice()
509 DeviceData[DeviceId].BlocksPerGrid, DeviceData[DeviceId].ThreadsPerBlock, in initDevice()
534 DeviceData[DeviceId].ThreadsPerBlock) { in initDevice()
536 DeviceData[DeviceId].ThreadsPerBlock); in initDevice()
537 DeviceData[DeviceId].NumTeams = DeviceData[DeviceId].ThreadsPerBlock; in initDevice()
883 if (CudaThreadsPerBlock > DeviceData[DeviceId].ThreadsPerBlock) { in runTargetTeamRegion()
885 DeviceData[DeviceId].ThreadsPerBlock); in runTargetTeamRegion()
[all …]
|
| /dports/www/chromium-legacy/chromium-88.0.4324.182/third_party/llvm/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 123 int ThreadsPerBlock = 0; member
474 DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::DefaultNumThreads; in initDevice()
477 DeviceData[DeviceId].ThreadsPerBlock = MaxBlockDimX; in initDevice()
482 DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::HardThreadLimit; in initDevice()
508 DeviceData[DeviceId].ThreadsPerBlock, DeviceData[DeviceId].WarpSize); in initDevice()
532 DeviceData[DeviceId].ThreadsPerBlock) { in initDevice()
534 DeviceData[DeviceId].ThreadsPerBlock); in initDevice()
535 DeviceData[DeviceId].NumTeams = DeviceData[DeviceId].ThreadsPerBlock; in initDevice()
880 if (CudaThreadsPerBlock > DeviceData[DeviceId].ThreadsPerBlock) { in runTargetTeamRegion()
882 DeviceData[DeviceId].ThreadsPerBlock); in runTargetTeamRegion()
[all …]
|
| /dports/devel/llvm12/llvm-project-12.0.1.src/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 136 int ThreadsPerBlock = 0; member
558 DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::DefaultNumThreads; in initDevice()
561 DeviceData[DeviceId].ThreadsPerBlock = MaxBlockDimX; in initDevice()
566 DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::HardThreadLimit; in initDevice()
592 DeviceData[DeviceId].ThreadsPerBlock, DeviceData[DeviceId].WarpSize); in initDevice()
616 DeviceData[DeviceId].ThreadsPerBlock) { in initDevice()
618 DeviceData[DeviceId].ThreadsPerBlock); in initDevice()
619 DeviceData[DeviceId].NumTeams = DeviceData[DeviceId].ThreadsPerBlock; in initDevice()
952 if (CudaThreadsPerBlock > DeviceData[DeviceId].ThreadsPerBlock) { in runTargetTeamRegion()
954 DeviceData[DeviceId].ThreadsPerBlock); in runTargetTeamRegion()
[all …]
|
| /dports/devel/wasi-compiler-rt12/llvm-project-12.0.1.src/openmp/libomptarget/plugins/cuda/src/ |
| H A D | rtl.cpp | 136 int ThreadsPerBlock = 0; member
558 DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::DefaultNumThreads; in initDevice()
561 DeviceData[DeviceId].ThreadsPerBlock = MaxBlockDimX; in initDevice()
566 DeviceData[DeviceId].ThreadsPerBlock = DeviceRTLTy::HardThreadLimit; in initDevice()
592 DeviceData[DeviceId].ThreadsPerBlock, DeviceData[DeviceId].WarpSize); in initDevice()
616 DeviceData[DeviceId].ThreadsPerBlock) { in initDevice()
618 DeviceData[DeviceId].ThreadsPerBlock); in initDevice()
619 DeviceData[DeviceId].NumTeams = DeviceData[DeviceId].ThreadsPerBlock; in initDevice()
952 if (CudaThreadsPerBlock > DeviceData[DeviceId].ThreadsPerBlock) { in runTargetTeamRegion()
954 DeviceData[DeviceId].ThreadsPerBlock); in runTargetTeamRegion()
[all …]
|