Skip to content

添加paddle在cudnn v6环境下cudnn convolution对double类型的支持 #954

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
ccx0912 opened this issue Dec 20, 2016 · 4 comments
Closed

添加paddle在cudnn v6环境下cudnn convolution对double类型的支持 #954

ccx0912 opened this issue Dec 20, 2016 · 4 comments
Assignees
@hedaoyuan
Labels
design_doc

Comments

@ccx0912
Copy link

ccx0912 commented Dec 20, 2016 •
edited
Loading

在测试paddle的在nvidia tesla P4上运行时，为了尽可能的发挥Pascal架构的优势，需要使用到cudnn v6和cuda 8.0。检查cudnn v5与v6的区别，发现在paddle/cuda/src/hl_cuda_cudnn.cc文件中使用的cudnn函数，只有函数cudnnSetConvolution2dDescriptor有所修改，cudnn v6后cudnn convolution可以支持double类型，所以函数的输入需要添加一个变量cudnnDataType_t变量来指定计算的数据类型，当前的PR #949 中也仅修改了这个函数，而对于cudnn v6中的其他特征并未考虑，如

  • 对更多的数据类型的支持（如CUDNN_DATA_INT8等）；
  • 新添加的函数（如cudnnReduceTensor等）；
  • 原来enum中添加的新成员；
  • 原来结构体中添加的新变量等。
@ccx0912 ccx0912 changed the title 添加paddle对cudnn v6和cuda 8.0的支持 #949 添加paddle对cudnn v6和cuda 8.0的支持 Dec 20, 2016
@reyoung
Copy link
Collaborator

reyoung commented Dec 20, 2016

@hedaoyuan Is this issue already fixed?

@hedaoyuan
Copy link
Contributor

从PR内容看起来主要是，修改cudnnSetConvolution2dDescriptor增加一个data_type参数，所以原来cudnn convolution是不支持double类型，cudnn v6后可以支持double类型了，PR主要是在cudnn v6环境下增加double的能力这样是吗？

如果只是修改convolution支持double的话，修改一下ISSUE的描述，当前标题跟内容不是特别相符，我昨天看到PR的标题的时候一开始也没看明白准备写什么。对cudnn v6的其他特性支持可以另外新建一个ISSUE来描述。

@ccx0912 ccx0912 changed the title 添加paddle对cudnn v6和cuda 8.0的支持 添加paddle在cudnn v6环境下对cudnn convolution double类型的支持 Dec 20, 2016
@ccx0912 ccx0912 changed the title 添加paddle在cudnn v6环境下对cudnn convolution double类型的支持 添加paddle在cudnn v6环境下cudnn convolution对double类型的支持 Dec 20, 2016
@hedaoyuan
Copy link
Contributor

另外，我看你的ISSUE描述应该是测试了Tesla P4，CUDNN V6等，如果有一些测试对比的性能数据的话也可以新建ISSUE贴上来。
可以是自己的Benchmark，也可以是基于paddle/benchmark的。

@fayeshine
Copy link

fayeshine commented Feb 17, 2017 •
edited
Loading

@ccx0912 cudnn v6是在哪出现的？有链接么

zhhsplendid pushed a commit to zhhsplendid/Paddle that referenced this issue Sep 25, 2019
@RaindragonD @shanyi15
update api_cn (PaddlePaddle#954)
9a601dd 
* update api_cn; fix the error of reverting layers_cn to 1.4; include previous changes

* update layers_cn; add code example to elementwise_mod

* delete recordio_writer_cn; revert index_cn

* update api_cn after review

* fix io_cn.rst

* update layers_cn

* update io_cn
wangxicoding pushed a commit to wangxicoding/Paddle that referenced this issue Dec 9, 2021
Fix README typo of SimCSE (PaddlePaddle#954)
b0a9b37
tianyuzhou668 pushed a commit to tianyuzhou668/Paddle that referenced this issue May 12, 2025
@rengar-2012
Aclnn grid sample (PaddlePaddle#954)
13b905a
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@hedaoyuan hedaoyuan

Labels
design_doc
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
4 participants
@reyoung @fayeshine @ccx0912 @hedaoyuan