build: Update C++ standard from c++17 to c++20 (#128) - #129
Merged
Conversation
* build: Update C++ standard from c++17 to c++20 Upstream PyTorch raised ATen's minimum required C++ standard from 17 to 20 (pytorch/pytorch#178150), which broke the pytorch_backend build. Align this repo's TRITON_MIN_CXX_STANDARD/CMAKE_CXX_STANDARD default with that floor for consistency across the stack. TRI-1877 * fix: rename semaphore.h to avoid colliding with POSIX <semaphore.h> Under C++20, libstdc++'s <condition_variable> transitively needs the POSIX <semaphore.h> to support std::counting_semaphore. This repo's own src/semaphore.h has the same name and sits ahead of the system include path (-I<src> precedes default system dirs), so the compiler resolved the internal #include <semaphore.h> to this project's own header instead of glibc's, which defines no POSIX semaphore symbols. That broke the C++17->20 bump build with a cascade of "sem_t does not name a type" / "condition_variable does not name a type" errors. Rename to triton_semaphore.h so it can no longer shadow the system header; harmless under C++17 since nothing pulled in <semaphore.h> there. TRI-1877 * fix: rename filesystem.h/.cc to avoid colliding with <filesystem> Same class of bug as the semaphore.h rename in this repo: a project header sharing a name with a C++ standard library header, combined with -I<src> preceding the system include path, means any future internal #include <filesystem> from libstdc++ (or a toolchain bump) would silently resolve to this project's own header instead. Not currently triggering a build failure, but proactively renamed given this repo already proved the pattern is live once. TRI-1877 (cherry picked from commit 38b69d0)
|
mc-nv
requested review from
Vinya567,
mattwittwer,
nv-rinig,
whoisj and
yinggeh
September 11, 2026 03:39
yinggeh
approved these changes
Sep 11, 2026
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Cherry-pick of the C++17->C++20 standard bump (TRI-1877) onto
r26.09.Upstream PyTorch raised ATen's minimum required C++ standard from 17
to 20 (pytorch/pytorch#178150), which broke the pytorch_backend build.
Align this repo's TRITON_MIN_CXX_STANDARD/CMAKE_CXX_STANDARD default
with that floor for consistency across the stack.
TRI-1877
Under C++20, libstdc++'s <condition_variable> transitively needs the
POSIX <semaphore.h> to support std::counting_semaphore. This repo's
own src/semaphore.h has the same name and sits ahead of the system
include path (-I precedes default system dirs), so the compiler
resolved the internal #include <semaphore.h> to this project's own
header instead of glibc's, which defines no POSIX semaphore symbols.
That broke the C++17->20 bump build with a cascade of "sem_t does not
name a type" / "condition_variable does not name a type" errors.
Rename to triton_semaphore.h so it can no longer shadow the system
header; harmless under C++17 since nothing pulled in <semaphore.h>
there.
TRI-1877
Same class of bug as the semaphore.h rename in this repo: a project
header sharing a name with a C++ standard library header, combined
with -I preceding the system include path, means any future
internal #include from libstdc++ (or a toolchain bump)
would silently resolve to this project's own header instead. Not
currently triggering a build failure, but proactively renamed given
this repo already proved the pattern is live once.
TRI-1877
(cherry picked from commit 38b69d0)