Spaces:
Sleeping
Sleeping
| ==16872== NVPROF is profiling process 16872, command: python profile.py -j 3 | |
| ==16872== Profiling application: python profile.py -j 3 | |
| ==16872== Profiling result: | |
| Type Time(%) Time Calls Avg Min Max Name | |
| GPU activities: 24.62% 4.7639ms 36 132.33us 20.640us 279.42us 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) | |
| 14.69% 2.8418ms 84 33.831us 2.2400us 121.82us 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) | |
| 14.66% 2.8366ms 140 20.261us 1.2480us 122.34us 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) | |
| 10.06% 1.9454ms 133 14.626us 864ns 1.8177ms [CUDA memcpy HtoD] | |
| 5.18% 1.0027ms 102 9.8300us 896ns 71.424us 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) | |
| 3.85% 744.51us 36 20.680us 3.0080us 66.720us void CatArrayBatchedCopy<float, unsigned int, int=4>(float*, CatArrInputTensor<float, unsigned int>*, OutputTensorSizeStride<unsigned int, unsigned int=4>, int, unsigned int) | |
| 3.72% 719.45us 24 29.977us 3.3920us 64.416us 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) | |
| 2.86% 554.01us 36 15.389us 1.5040us 42.080us 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) | |
| 2.43% 470.33us 54 8.7090us 1.2470us 32.640us 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) | |
| 2.36% 456.28us 9 50.698us 8.4480us 87.487us 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) | |
| 2.15% 416.35us 27 15.420us 1.4720us 37.856us 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) | |
| 1.86% 360.13us 336 1.0710us 960ns 6.7190us [CUDA memcpy DtoH] | |
| 1.70% 329.79us 84 3.9260us 2.9760us 4.3510us 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*) | |
| 1.64% 317.66us 84 3.7810us 3.0400us 4.1600us 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*) | |
| 1.19% 230.66us 48 4.8050us 1.8240us 8.0640us 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) | |
| 1.17% 227.04us 9 25.226us 2.5600us 59.647us 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) | |
| 1.10% 212.06us 168 1.2620us 960ns 11.808us 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) | |
| 0.86% 166.56us 6 27.759us 4.3200us 59.488us 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) | |
| 0.85% 163.49us 140 1.1670us 800ns 2.0800us 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) | |
| 0.83% 160.41us 168 954ns 768ns 1.2800us void kernelPointwiseApply1<TensorFillOp<long>, long, unsigned int, int=1>(OffsetInfo<TensorFillOp<long>, long, unsigned int>, long, long) | |
| 0.69% 132.67us 84 1.5790us 1.2480us 1.8560us 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) | |
| 0.58% 111.81us 9 12.423us 1.5360us 29.951us 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) | |
| 0.55% 106.08us 3 35.359us 7.2960us 76.896us 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) | |
| 0.40% 77.216us 9 8.5790us 1.5360us 20.704us 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: 99.07% 5.25149s 27 194.50ms 6.1140us 5.24151s cudaMalloc | |
| 0.25% 13.365ms 1360 9.8270us 5.7310us 444.75us cudaLaunch | |
| 0.16% 8.2355ms 469 17.559us 5.4040us 2.1906ms cudaMemcpyAsync | |
| 0.11% 5.9895ms 433 13.832us 1.5650us 1.7388ms cudaStreamSynchronize | |
| 0.10% 5.0736ms 15334 330ns 264ns 17.933us cudaGetDevice | |
| 0.08% 4.1577ms 185 22.474us 208ns 897.48us cuDeviceGetAttribute | |
| 0.07% 3.9718ms 2 1.9859ms 1.9821ms 1.9897ms cudaGetDeviceProperties | |
| 0.05% 2.7476ms 2 1.3738ms 198.55us 2.5491ms cuDeviceTotalMem | |
| 0.04% 1.8782ms 5057 371ns 287ns 16.705us cudaSetDevice | |
| 0.02% 1.0777ms 2 538.85us 15.726us 1.0620ms cudaHostAlloc | |
| 0.02% 1.0019ms 6161 162ns 108ns 16.261us cudaSetupArgument | |
| 0.01% 463.58us 140 3.3110us 2.9140us 13.313us cudaFuncGetAttributes | |
| 0.01% 438.49us 2 219.25us 193.78us 244.71us cuDeviceGetName | |
| 0.01% 291.44us 1360 214ns 144ns 2.5110us cudaConfigureCall | |
| 0.01% 290.38us 1516 191ns 109ns 2.6830us cudaGetLastError | |
| 0.00% 89.031us 37 2.4060us 966ns 4.2400us cudaEventQuery | |
| 0.00% 76.994us 140 549ns 428ns 2.4670us cudaDeviceGetAttribute | |
| 0.00% 46.269us 280 165ns 103ns 1.8560us cudaPeekAtLastError | |
| 0.00% 45.701us 36 1.2690us 874ns 2.1460us cudaEventCreateWithFlags | |
| 0.00% 44.592us 36 1.2380us 907ns 2.4760us cudaEventRecord | |
| 0.00% 42.537us 35 1.2150us 549ns 3.0390us cudaEventDestroy | |
| 0.00% 5.0090us 13 385ns 164ns 1.5380us cudaGetDeviceCount | |
| 0.00% 3.2090us 4 802ns 237ns 2.2260us cuDeviceGetCount | |
| 0.00% 1.7730us 3 591ns 259ns 1.1350us cuDeviceGet | |
| 0.00% 946ns 1 946ns 946ns 946ns cuInit | |
| 0.00% 883ns 1 883ns 883ns 883ns cuDriverGetVersion | |