Spaces:
Sleeping
Sleeping
| ==17288== NVPROF is profiling process 17288, command: python profile.py -b -j 4 | |
| ==17288== Profiling application: python profile.py -b -j 4 | |
| ==17288== Profiling result: | |
| Type Time(%) Time Calls Avg Min Max Name | |
| GPU activities: 19.68% 2.5049ms 24 104.37us 8.3520us 280.09us void spatialDepthwiseConvolutionUpdateOutput<float, float, unsigned int, int=0>(THCDeviceTensor<float, int=4, int, DefaultPtrTraits>, THCDeviceTensor<float, int=4, int, DefaultPtrTraits>, THCDeviceTensor<float, int=4, int, DefaultPtrTraits>, THCDeviceTensor<float, int=1, int, DefaultPtrTraits>, bool, unsigned int, int, int, int, int, int, int, int, int, int, int, int, int, int, int) | |
| 18.44% 2.3476ms 107 21.939us 864ns 1.6917ms [CUDA memcpy HtoD] | |
| 12.54% 1.5960ms 130 12.277us 1.2480us 122.88us void kernelPointwiseApply3<TensorAddOp<long>, long, long, long, unsigned int, int=1, int=2, int=2>(OffsetInfo<TensorAddOp<long>, long, unsigned int>, OffsetInfo<long, long, int=1>, OffsetInfo<long, long, int=2>, long, long) | |
| 11.22% 1.4280ms 78 18.307us 1.3760us 121.47us void kernelPointwiseApply2<TensorTakeOp<float, int, int=-2>, float, long, unsigned int, int=1, int=1>(OffsetInfo<int, TensorTakeOp<float, int, int=-2>, float>, OffsetInfo<int=-2, TensorTakeOp<float, int, int=-2>, long>, TensorTakeOp<float, int, int=-2>, float) | |
| 4.68% 595.55us 24 24.814us 1.7280us 64.000us void kernelPointwiseApply2<CopyOp<float, float>, float, float, unsigned int, int=1, int=2>(OffsetInfo<float, float, float>, OffsetInfo<CopyOp<float, float>, float, unsigned int>, float, float) | |
| 4.62% 588.34us 96 6.1280us 896ns 70.591us void kernelPointwiseApply2<CopyOp<float, float>, float, float, unsigned int, int=1, int=1>(OffsetInfo<float, float, float>, OffsetInfo<CopyOp<float, float>, float, unsigned int>, float, float) | |
| 3.67% 467.23us 12 38.935us 2.0480us 87.167us void kernelPointwiseApply3<TensorAddOp<float>, float, float, float, unsigned int, int=1, int=1, int=1>(OffsetInfo<TensorAddOp<float>, float, unsigned int>, OffsetInfo<float, float, int=1>, OffsetInfo<float, float, int=1>, float, float) | |
| 3.15% 400.80us 18 22.266us 7.0080us 65.215us void CatArrayBatchedCopy<float, unsigned int, int=4>(float*, CatArrInputTensor<float, unsigned int>*, OutputTensorSizeStride<unsigned int, unsigned int=4>, int, unsigned int) | |
| 2.57% 326.59us 312 1.0460us 960ns 1.3440us [CUDA memcpy DtoH] | |
| 2.19% 278.33us 78 3.5680us 2.9760us 4.2240us void kernelReduceAll<long, unsigned int, long, thrust::identity<long>, ReduceMax<long>, int=1>(TensorInfo<long, unsigned int>, unsigned int, long, long, thrust::identity<long>, long*) | |
| 2.17% 276.80us 24 11.533us 1.1200us 38.207us void kernelPointwiseApply3<TensorAddOp<float>, float, float, float, unsigned int, int=1, int=2, int=2>(OffsetInfo<TensorAddOp<float>, float, unsigned int>, OffsetInfo<float, float, int=1>, OffsetInfo<float, float, int=2>, float, float) | |
| 2.13% 271.10us 78 3.4750us 3.0080us 3.9360us void kernelReduceAll<long, unsigned int, long, thrust::identity<long>, ReduceMin<long>, int=1>(TensorInfo<long, unsigned int>, unsigned int, long, long, thrust::identity<long>, long*) | |
| 1.94% 246.33us 48 5.1310us 928ns 14.624us void kernelPointwiseApply2<CopyOp<float, float>, float, float, unsigned int, int=2, int=1>(OffsetInfo<float, float, float>, OffsetInfo<CopyOp<float, float>, float, unsigned int>, float, float) | |
| 1.43% 181.66us 48 3.7840us 1.1520us 8.1920us void kernelPointwiseApply2<CopyOp<float, float>, float, float, unsigned int, int=2, int=2>(OffsetInfo<float, float, float>, OffsetInfo<CopyOp<float, float>, float, unsigned int>, float, float) | |
| 1.41% 178.85us 156 1.1460us 959ns 1.5360us void kernelPointwiseApply2<TensorMulConstantOp<long>, long, long, unsigned int, int=1, int=1>(OffsetInfo<TensorMulConstantOp<long>, long, unsigned int>, OffsetInfo<long, long, int=1>, long, long) | |
| 1.34% 170.62us 8 21.327us 1.2480us 60.159us void kernelPointwiseApply2<TensorDivConstantOp<float>, float, float, unsigned int, int=1, int=2>(OffsetInfo<TensorDivConstantOp<float>, float, unsigned int>, OffsetInfo<float, float, int=1>, float, float) | |
| 1.27% 161.18us 156 1.0330us 928ns 1.2480us void kernelPointwiseApply1<TensorFillOp<long>, long, unsigned int, int=1>(OffsetInfo<TensorFillOp<long>, long, unsigned int>, long, long) | |
| 1.13% 143.30us 130 1.1020us 800ns 2.0480us void thrust::cuda_cub::core::_kernel_agent<thrust::cuda_cub::__parallel_for::ParallelForAgent<thrust::cuda_cub::__tabulate::functor<thrust::device_ptr<long>, LinspaceOp<long, long>, long>, long>, thrust::cuda_cub::__tabulate::functor<thrust::device_ptr<long>, LinspaceOp<long, long>, long>, long>(thrust::device_ptr<long>, long) | |
| 1.10% 140.09us 12 11.674us 1.2480us 36.127us void kernelPointwiseApply3<TensorSubOp<float>, float, float, float, unsigned int, int=1, int=2, int=2>(OffsetInfo<TensorSubOp<float>, float, unsigned int>, OffsetInfo<float, float, int=1>, OffsetInfo<float, float, int=2>, float, float) | |
| 0.93% 118.56us 12 9.8790us 1.2160us 29.983us void kernelPointwiseApply3<TensorAddOp<float>, float, float, float, unsigned int, int=1, int=1, int=2>(OffsetInfo<TensorAddOp<float>, float, unsigned int>, OffsetInfo<float, float, int=1>, OffsetInfo<float, float, int=1>, float, float) | |
| 0.91% 115.26us 78 1.4770us 1.2160us 1.7920us void kernelPointwiseApply2<TensorRemainderOp<long>, long, long, unsigned int, int=1, int=1>(OffsetInfo<TensorRemainderOp<long>, long, unsigned int>, OffsetInfo<long, long, int=1>, long, long) | |
| 0.86% 109.12us 4 27.279us 3.3600us 76.703us void kernelPointwiseApply2<TensorDivConstantOp<float>, float, float, unsigned int, int=1, int=-1>(OffsetInfo<TensorDivConstantOp<float>, float, unsigned int>, OffsetInfo<float, float, int=1>, float, float) | |
| 0.65% 82.752us 12 6.8960us 1.1840us 20.896us void kernelPointwiseApply2<Tensor_neg_Float_Op, float, float, unsigned int, int=1, int=2>(OffsetInfo<float, unsigned int, int=1>, OffsetInfo<float, unsigned int, int=2>, unsigned int, Tensor_neg_Float_Op) | |
| API calls: 99.14% 5.41033s 35 154.58ms 7.4430us 5.35626s cudaMalloc | |
| 0.22% 11.895ms 1226 9.7010us 5.0900us 82.151us cudaLaunch | |
| 0.16% 8.7230ms 185 47.151us 308ns 4.0620ms cuDeviceGetAttribute | |
| 0.15% 8.1108ms 419 19.357us 6.0580us 1.6921ms cudaMemcpyAsync | |
| 0.09% 5.1797ms 13735 377ns 265ns 382.93us cudaGetDevice | |
| 0.08% 4.2560ms 2 2.1280ms 2.1218ms 2.1342ms cudaGetDeviceProperties | |
| 0.05% 2.7402ms 401 6.8330us 1.6790us 292.86us cudaStreamSynchronize | |
| 0.03% 1.7312ms 4508 384ns 288ns 20.415us cudaSetDevice | |
| 0.02% 1.2230ms 1 1.2230ms 1.2230ms 1.2230ms cudaHostAlloc | |
| 0.02% 851.30us 5392 157ns 110ns 13.164us cudaSetupArgument | |
| 0.01% 548.09us 2 274.04us 273.09us 275.00us cuDeviceTotalMem | |
| 0.01% 449.36us 130 3.4560us 2.9700us 6.8560us cudaFuncGetAttributes | |
| 0.01% 416.63us 2 208.32us 208.18us 208.46us cuDeviceGetName | |
| 0.00% 239.25us 1340 178ns 116ns 496ns cudaGetLastError | |
| 0.00% 231.17us 1226 188ns 139ns 1.0700us cudaConfigureCall | |
| 0.00% 71.746us 130 551ns 406ns 1.4920us cudaDeviceGetAttribute | |
| 0.00% 47.977us 17 2.8220us 2.4690us 3.5120us cudaEventQuery | |
| 0.00% 44.070us 260 169ns 107ns 411ns cudaPeekAtLastError | |
| 0.00% 27.922us 18 1.5510us 1.4070us 2.4460us cudaEventCreateWithFlags | |
| 0.00% 26.123us 17 1.5360us 1.3210us 2.0040us cudaEventDestroy | |
| 0.00% 25.594us 18 1.4210us 1.3320us 1.9000us cudaEventRecord | |
| 0.00% 6.9690us 13 536ns 235ns 1.6330us cudaGetDeviceCount | |
| 0.00% 3.3440us 4 836ns 356ns 2.1080us cuDeviceGetCount | |
| 0.00% 2.0700us 3 690ns 363ns 1.1930us cuDeviceGet | |
| 0.00% 990ns 1 990ns 990ns 990ns cuDriverGetVersion | |
| 0.00% 978ns 1 978ns 978ns 978ns cuInit | |