Skip to content
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

fix hcc linking error caused by __fdividef #25

Merged
merged 2 commits into from
Jun 14, 2018
Merged

Conversation

sunway513
Copy link

This commit fixes linking error with recent HIP updates on, thanks to @AlexVlx .

@ROCm-Apps-Test
Copy link

Can one of the admins verify this patch?

@sunway513 sunway513 requested a review from whchung June 13, 2018 19:47
Copy link
Collaborator

@whchung whchung left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would the proposed change be applicable on CUDA path?

@whchung
Copy link
Collaborator

whchung commented Jun 13, 2018

ok to test

@sunway513
Copy link
Author

Not tested yet -- this is a perfect need from NV CI..

@whchung
Copy link
Collaborator

whchung commented Jun 13, 2018

@parallelo It seems "ok to test" doesn't really trigger the test.... guess I'm not admin?

@parallelo
Copy link

Jenkins: ok to test

@parallelo
Copy link

Let me check the Jenkins logs for the blocker..

@parallelo
Copy link

Jenkins: add to whitelist

@parallelo
Copy link

Jenkins logs show:

Pull request #25s author has been whitelisted

Jenkins: test this please

@parallelo
Copy link

Okay, all set. Build triggered for merge commit.

Next: figure out why the group whitelist isn't working properly with this Jenkins plugin.

@whchung whchung mentioned this pull request Jun 13, 2018
@@ -50,7 +50,7 @@ __global__ void InvVarianceToVarianceKernel(int nthreads, double epsilon,
int sample_size, T* variance) {
GPU_1D_KERNEL_LOOP(index, nthreads) {
T inv_var = variance[index];
T var = __fdividef(1, inv_var * inv_var) - T(epsilon);
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

__fdividef() has some magic tricks in CUDA as it appears. I'd recommend using #if / #else to switch logic between CUDA and ROCm before we figure this out.

Copy link
Collaborator

@whchung whchung left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

__fdividef() has some magic tricks in CUDA as it appears. I'd recommend using #if / #else to switch logic between CUDA and ROCm before we figure this out.

Copy link
Collaborator

@whchung whchung left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

// instead of #?

@whchung whchung merged commit 5dafc62 into develop-upstream Jun 14, 2018
sunway513 pushed a commit that referenced this pull request Jun 20, 2018
fix hcc linking error caused by __fdividef
@whchung whchung deleted the fix_fdividef branch August 27, 2018 23:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

4 participants