ncnn Execution of Convolution Implementation

waxmsbnn 于 4个月前发布在其他

关注(0)|答案(8)|浏览(46)

Hi NCNN team,

I was exploring the Convolution Implementation found in -> https://github.com/Tencent/ncnn/blob/master/src/layer/convolution.cpp
I want to know on which specific Hardware that Implementation will be executed?

Look forward to your reply.

Thanks

ncnn

来源：https://github.com/Tencent/ncnn/issues/3549

8条答案

按热度按时间

7ivaypg91#

Hi @nihui,

looking forward to your reply.

Thank

赞(0）回复(0）举报 4个月前

7bsow1i62#

Hi @nihui,

Does the Implementation use NCHW or NHCW format?

Thanks

NCHW where N=1

赞(0）回复(0）举报 4个月前

okxuctiv3#

Hi @nihui,

Thanks a lot for your kind reply.

Can you please help me with the script in the repo where I can get the info on selecting the Convolution implementation(either naive or optimized implementation) for execution on X86 CPU?

Look forward to your reply.

Thanks

赞(0）回复(0）举报 4个月前

i7uq4tfw4#

This is the naive implementation.
For all cpu arch other than x86/arm/mips/risc-v, it will be used.

赞(0）回复(0）举报 4个月前

yh2wf1be5#

Hi @nihui,

Thanks for your reply.

If I want to execute the same on x86 what are the changes to be made? Basically, I want to understand the Implementation by executing it.

Thanks

赞(0）回复(0）举报 4个月前