[REVIEW] Pagerank fix #919
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Contributor
|
Please update the changelog in order to start CI tests. View the gpuCI docs here. |
seunghwak
approved these changes
May 31, 2020
Contributor
seunghwak
left a comment
seunghwak
left a comment
There was a problem hiding this comment.
If we are passing non-owning pointers, they should be marked as const (mark the pointer itself is const, not the values pointed by the pointer, or we can use iterators instead to be more generic as well) to be safe.
Better establish this convention in our codebase, but hope we can retire this code sooner than later.
rlratzel
approved these changes
Jun 1, 2020
raydouglass
suggested changes
Jun 1, 2020
| number: {{ git_revision_count }} | ||
| string: cuda{{ cuda_version }}_{{ git_revision_count }} | ||
| number: {{ GIT_DESCRIBE_NUMBER }} | ||
| string: cuda{{ cuda_version }}_{{ GIT_DESCRIBE_HASH }}_{{ GIT_DESCRIBE_NUMBER }} |
Contributor
There was a problem hiding this comment.
Something is wrong with how you branched or you're targeting the wrong branch. These are v0.15 changes.
seunghwak
reviewed
Jun 1, 2020
| // A thrust::swap would probably be more efficent if the vectors were passed everywhere | ||
| // instead of pointers. std::swap is unsafe though. Just copying for now, as this may soon be | ||
| // replaced by the pattern accelerator. | ||
| copy(n, pr, tmp); |
Contributor
There was a problem hiding this comment.
and just FYI, what you could do instead is something like
pagerank(value_t* pr, ...) {
rmm::device_uvector<value_t> tmp{};
value_t* p_new = pr;
value_t* p_old = tmp.data().get();
...
// at the end of iteration
if (iter % 2 == 0) {
p_new = tmp.data().get();
p_old = pr;
}
else {
p_new = pr;
p_old = tmp.data().get();
}
}
Member
Author
|
#923 made it to 0.15 as well |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
double-free bug found by @rlratzel
pagerankandtmpthrust vectors are both scope managed now, but raw pointers to their device memory are std::swapped inside the solver. By the time the scope ends forpagerankits device pointer may be pointing to the data that was freed bytmp.Just doing a cpy for now as a fix since this may soon be replaced by the pattern accelerator.
A thrust::swap would probably be ideal if the vectors were passed to this section of the code instead of raw pointers.