Skip to content

Instantly share code, notes, and snippets.

@soumith
Created December 18, 2014 02:44
Show Gist options
  • Save soumith/e6297e93dd2fe3751562 to your computer and use it in GitHub Desktop.
Save soumith/e6297e93dd2fe3751562 to your computer and use it in GitHub Desktop.
benchmarks prelim
All timings averaged over 10 runs and in miliseconds on a Tesla K40m with ECC off.
Time combines forward + backwardInputs + backwardGradients
###############################################################################
ModelType: AlexNet Kernels: cudnn Input shape: 128x3x224x224
cudnn :updateOutput(): 147.32
cudnn :updateGradInput(): 167.79
cudnn :accGradParameters(): 153.96
cudnn :TOTAL: 469.07
ModelType: AlexNet Kernels: nn Input shape: 128x3x224x224
nn :updateOutput(): 201.53
nn :updateGradInput(): 165.09
nn :accGradParameters(): 193.94
nn :TOTAL: 560.56
ModelType: AlexNet Kernels: fbcufft Input shape: 128x3x224x224
fbcufft :updateOutput(): 94.34
fbcufft :updateGradInput(): 96.69
fbcufft :accGradParameters(): 93.20
fbcufft :TOTAL: 284.23
ModelType: AlexNet Kernels: cuda-convnet2 Input shape: 3x224x224x128
cuda-convnet2 :updateOutput(): 99.03
cuda-convnet2 :updateGradInput(): 104.59
cuda-convnet2 :accGradParameters(): 103.29
cuda-convnet2 :TOTAL: 306.91
###################################################################################
ModelType: OverFeat[fast] Kernels: cudnn Input shape: 128x3x231x231
cudnn :updateOutput(): 459.06
cudnn :updateGradInput(): 634.26
cudnn :accGradParameters(): 508.02
cudnn :TOTAL: 1601.35
ModelType: OverFeat[fast] Kernels: nn Input shape: 128x3x231x231
nn :updateOutput(): 710.33
nn :updateGradInput(): 433.29
nn :accGradParameters(): 548.13
nn :TOTAL: 1691.74
ModelType: OverFeat[fast] Kernels: fbcufft Input shape: 128x3x231x231
fbcufft :updateOutput(): 375.65
fbcufft :updateGradInput(): 460.48
fbcufft :accGradParameters(): 397.85
fbcufft :TOTAL: 1233.98
ModelType: OverFeat[fast] Kernels: cuda-convnet2 Input shape: 3x231x231x128
cuda-convnet2 :updateOutput(): 433.11
cuda-convnet2 :updateGradInput(): 398.87
cuda-convnet2 :accGradParameters(): 450.82
cuda-convnet2 :TOTAL: 1282.80
######################################################################################
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment