Is there a way to make device memory persist between CUDA MEX calls

2013 12 25
2 个回答
回答已采纳
更新时间：2014 6 18
5 次查看（30 天）

0 个投票

I'm wondering if there is a graceful way to allocate data on or transfer data to the GPU in one MEX file (basically MATLAB interfaces to cudaMalloc or cudaMemcpy) and then process that data on the GPU with a different MEX file. I'm wondering if it is possible to do this without the Parallel Computing Toolbox.
When I do the memory transfer/allocation on the GPU, I will need to keep the pointer to the device memory and have it reside in the MATLAB workspace in some form until it is ready to be passed to the data processing MEX file. I'm wondering what the best way to do that is. Would I just convert the pointer value (not dereferenced) to a MATLAB integer and then convert it back again in the data processing MEX file when needed?

请先登录，再回答此问题。

 采纳的回答

Oliver Woodford
Oliver Woodford 2014-1-2
编辑：Oliver Woodford 2014-1-2

2 个投票

Yes, you can reinterpret_cast the pointer to an integer of a sufficient bit length, e.g. uint64, and return this to MATLAB. Then pass the integer to the new mex file, where it is reinterpreted as the pointer to the GPU memory again.
If you want to use this approach, and minimize chances of leaking memory, then wrap the memory allocation in a C++ class, and use the approach outlined in this FEX submission.

4 个评论
显示 2更早的评论 隐藏 2更早的评论

Oliver Woodford
Oliver Woodford 2014-1-2
I have used this approach (passing CUDA memory pointers, wrapped in a C++ class, between multiple mex files) myself, and can confirm it works fine.
Matt J
Matt J 2014-1-3
OK, looks promising. I'll have to study that too, though.
Matt J
Matt J 2014-6-18
I never got a chance to test it, but I accept the answer anyway.
I'm starting to think that the cudaKernel class in the Optimization Toolbox really is the better investment, though. Its interface is a good replacement for a lot of the awkward things that you have to do outside the kernel itself in C/C++.

请先登录，再进行评论。

更多回答（1 个）

Eric Sampson
Eric Sampson 2014-1-2

0 个投票

Matt, see if the methods discussed in the following thread could be adapted for your use: http://www.mathworks.com/matlabcentral/newsreader/view_thread/278243

1 个评论
显示 -1更早的评论 隐藏 -1更早的评论

Matt J
Matt J 2014-1-2
Thanks for the reference, Eric. I'll need to study it for a bit, but it looks like it could be applicable.

请先登录，再进行评论。

类别

在 帮助中心 和 File Exchange 中查找有关 GPU Computing 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by