Fix nullptr math & clean segment_csr_cpu.cpp #292
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
When I run this code with LLVM-12's undefined behaviour sanitizer enabled, I see:
```
pytorch/torch-scatter/csrc/cpu/segment_csr_cpu.cpp:60:3: runtime error: applying non-zero offset 8 to null pointer
#0 0x7fa95da15396 in segment_csr_cpu(at::Tensor, at::Tensor, c10::optional<at::Tensor>, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >)::$_0::operator()() const::'lambda2'()::operator()() const::'lambda'()::operator()() const::'lambda'()::operator()() const pytorch/torch-scatter/csrc/cpu/segment_csr_cpu.cpp:60
```
This is because on Line 41 we have `int64_t *arg_out_data = nullptr;`. The value of `arg_out_data` is set conditionally, but `arg_out_data` is used unconditionally within the "segment_csr" kernel. Adding an if-statement gates that.
I've also added `const` and de-shadowed variables in a few places to make the code more readable.
rusty1s
reviewed
May 4, 2022
|
|
||
| for (auto k = 0; k < K; k++) | ||
| Reducer<scalar_t, REDUCE>::write(out_data + n * K + k, vals[k], | ||
| if (reduce2REDUCE.at(reduce) == MIN || reduce2REDUCE.at(reduce) == MAX) { |
Owner
There was a problem hiding this comment.
It looks like we skip writing to the destination here for any other reduction (which is likely unintented - see test failures).
Contributor
Author
|
Thanks; do you happen to know how to fix that? (My current solution is a
bit of a guess since I have almost no familiarity with this code.)
…On Tue, May 3, 2022, 23:48 Matthias Fey ***@***.***> wrote:
***@***.**** commented on this pull request.
Thanks for the PR!
------------------------------
In csrc/cpu/segment_csr_cpu.cpp
<#292 (comment)>
:
>
- for (auto k = 0; k < K; k++)
- Reducer<scalar_t, REDUCE>::write(out_data + n * K + k, vals[k],
+ if (reduce2REDUCE.at(reduce) == MIN || reduce2REDUCE.at(reduce) == MAX) {
It looks like we skip writing to the destination here for any other
reduction (which is likely unintented - see test failures).
—
Reply to this email directly, view it on GitHub
<#292 (review)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAXZHVEGH5ZW5WNXN7KGSZTVIIFTRANCNFSM5U76MKRA>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
Owner
|
I think we could overload |
|
This pull request had no activity for 6 months. It will be closed in 2 weeks unless there is some new activity. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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.
When I run this code with LLVM-12's undefined behaviour sanitizer enabled, I see:
This is because on Line 41 we have
int64_t *arg_out_data = nullptr;. The value ofarg_out_datais set conditionally, butarg_out_datais used unconditionally within the "segment_csr" kernel. Adding an if-statement gates that.I've also added
constand de-shadowed variables in a few places to make the code more readable.