-
Notifications
You must be signed in to change notification settings - Fork 2.5k
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.
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
Parallel nvcc/g++ call? #3340
Labels
Comments
I think it is a good idea. We discussed about this while discussion about the related issue gh-3277. |
Also, I think this one is easier to do then gh-3277. |
Is there anyone working on this? Or is there any plan on how to get this working? |
Nobody work on this. Note, we don't call nvcc anymore. we always use g++
now.
There is high level plan, but we don't have detailed plan. There was
discussion that we re-work the cache mechanism to don't request a lock. I'm
not sure we will work on this now. But this would be related.
If you want to work on this, mostly, we need to be able to have this loop
parallelized:
https://github.com/Theano/Theano/blob/27705c7c60d270670a967f66a3f94924de8f0269/theano/gof/vm.py#L1074
Note, the make_thunk call will cause the taking of the lock. Maybe/Probably
you will need to take the lock before that loop. The lock can be taken many
times and we remove it only when we release it as many times it was taken.
Note, we work to compile less c code, but having this is also useful.
…On Fri, Jul 7, 2017 at 4:07 AM khaotik ***@***.***> wrote:
Is there anyone working on this? Or is there any plan on how to get this
working?
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#3340 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AALC-w5WBMRSP4rf71ZtLio6-vdIe5Ziks5sLedSgaJpZM4F0mHi>
.
|
I would be quite busy for July and maybe August. I'll try to tackle it when I got the time. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Is it possible to parallelize the compilation (nvcc)? Something along the line how
make -j
works.The text was updated successfully, but these errors were encountered: