Skip to content

Fix uninitialized GEMM parallel bounds #197

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.

Sign up for GitHub

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

Merged
shinaoka merged 1 commit into from
Jul 31, 2026
Merged

Fix uninitialized GEMM parallel bounds #197

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 10 additions & 0 deletions docs/2026-07-31-issue-198-worklog.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,10 @@
# Issue #198 worklog

## 2026-07-31

- Reproduced the `strided-einsum2` `faer + parallel` feature combination failure
after the merged uninitialized GEMM work (#188/#196).
- Added the existing `MaybeSendSync` contract to the uninitialized contiguous
output wrapper and its preparation entry point. This is a type-bound fix only;
it does not change kernel behavior or threading policy.
- Verification: `cargo fmt --all -- --check`; `cargo test -p strided-einsum2`.
5 changes: 3 additions & 2 deletions strided-einsum2/src/contiguous.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@ use std::any::{Any, TypeId};
use std::cell::RefCell;
use std::collections::HashMap;
use std::mem::MaybeUninit;
use strided_kernel::MaybeSendSync;
use strided_view::{RawStridedMut, RawStridedRef, StridedArray, StridedView, StridedViewMut};

/// GEMM-ready input operand with contiguous data.
Expand Down Expand Up @@ -777,7 +778,7 @@ pub fn prepare_output_raw<T: ScalarBase + 'static>(
}

#[allow(dead_code)]
impl<'a, 'b, T: Copy + 'static> UninitContiguousOperand<'a, 'b, T> {
impl<'a, 'b, T: Copy + MaybeSendSync + 'static> UninitContiguousOperand<'a, 'b, T> {
#[inline]
pub(crate) fn ptr(&self) -> *mut MaybeUninit<T> {
self.ptr
Expand Down Expand Up @@ -836,7 +837,7 @@ impl<'a, 'b, T: Copy + 'static> UninitContiguousOperand<'a, 'b, T> {
/// Prepare an overwrite-only C operand. A non-fusable destination receives a
/// dense `MaybeUninit` temporary and a compiled uninitialized writeback plan.
#[allow(dead_code)]
pub(crate) fn prepare_output_raw_uninit<'a, 'b, T: ScalarBase + 'static>(
pub(crate) fn prepare_output_raw_uninit<'a, 'b, T: ScalarBase + MaybeSendSync + 'static>(
destination: &'a mut RawStridedMut<'b, MaybeUninit<T>>,
n_group1: usize,
n_group2: usize,
Expand Down
Loading