KellenSunderland opened a new issue #14684: When setting MXNET_CUDA_TENSOR_OP_MATH_ALLOW_CONVERSION no speedup observed URL: https://github.com/apache/incubator-mxnet/issues/14684 Doing a quick debugging session on why I don't see any speedup when enabling MXNET_CUDA_TENSOR_OP_MATH_ALLOW_CONVERSION. So far interesting observations: Enabling CuDNN API logging and observing forward calls I see repeated examples of these calls with default math type. Debugging I have verified we do set correct mathtype during cudnn conv setup here: https://github.com/apache/incubator-mxnet/blob/master/src/operator/nn/cudnn/cudnn_convolution-inl.h#L588 but when running a quick inference sample from the resnet50v2 zoo I get many of these outputs: ``` I! CuDNN (v7301) function cudnnConvolutionForward() called: i! handle: type=cudnnHandle_t; streamId=0x7feef87f4260; i! alpha: type=CUDNN_DATA_FLOAT; val=1.000000; i! xDesc: type=cudnnTensorDescriptor_t: i! dataType: type=cudnnDataType_t; val=CUDNN_DATA_FLOAT (0); i! nbDims: type=int; val=4; i! dimA: type=int; val=[32,512,15,20]; i! strideA: type=int; val=[153600,300,20,1]; i! xData: location=dev; addr=0x7fed72000000; i! wDesc: type=cudnnFilterDescriptor_t: i! dataType: type=cudnnDataType_t; val=CUDNN_DATA_FLOAT (0); i! vect: type=int; val=0; i! nbDims: type=int; val=4; i! dimA: type=int; val=[2048,512,1,1]; i! format: type=cudnnTensorFormat_t; val=CUDNN_TENSOR_NCHW (0); i! wData: location=dev; addr=0x7fee01c00000; i! convDesc: type=cudnnConvolutionDescriptor_t: i! mode: type=cudnnConvolutionMode_t; val=CUDNN_CROSS_CORRELATION (1); i! dataType: type=cudnnDataType_t; val=CUDNN_DATA_FLOAT (0); i! mathType: type=cudnnMathType_t; val=CUDNN_DEFAULT_MATH (0); i! arrayLength: type=int; val=2; i! padA: type=int; val=[0,0]; i! strideA: type=int; val=[1,1]; i! dilationA: type=int; val=[1,1]; i! groupCount: type=int; val=1; i! algo: type=cudnnConvolutionFwdAlgo_t; val=CUDNN_CONVOLUTION_FWD_ALGO_IMPLICIT_PRECOMP_GEMM (1); i! workSpace: location=dev; addr=0x7fed76000000; i! workSpaceSizeInBytes: type=size_t; val=1808; i! beta: type=CUDNN_DATA_FLOAT; val=0.000000; i! yDesc: type=cudnnTensorDescriptor_t: i! dataType: type=cudnnDataType_t; val=CUDNN_DATA_FLOAT (0); i! nbDims: type=int; val=4; i! dimA: type=int; val=[32,2048,15,20]; i! strideA: type=int; val=[614400,300,20,1]; i! yData: location=dev; addr=0x7fed8c000000; i! Time: 2019-04-12T10:46:34.090769 (0d+0h+0m+9s since start) i! Process=7357; Thread=7411; GPU=0; Handle=0x7feeaea83090; StreamId=0x7feef87f4260. ```
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
