PaddlePaddle / Anakin

High performance Cross-platform Inference-engine, you could run Anakin on x86-cpu,arm, nv-gpu, amd-gpu,bitmain and cambricon devices.
https://anakin.baidu.com/
Apache License 2.0
531 stars 135 forks source link

How to use sass_func for computing convolution? #523

Open peyer opened 5 years ago

peyer commented 5 years ago

Thanks for your great work! I just wanna to compare the latency of running convolution layer between anakin_sass and cudnn, and I have successfully ran convolution layer with cudnn. However, when I call api of direct_conv_bias_Kdivis4 in sass_funcs.h, I found that I just did not understand the meaning of parameters like img_in_channel_stride, img_in_height_stride and so on. So, How should I set the img_in_channel_stride parameter properly?

throneclay commented 5 years ago

please take https://github.com/PaddlePaddle/Anakin/blob/master/saber/funcs/impl/cuda/saber_conv_direct.cpp as an example