Skip to content

Fix max atomic size for NVHPC#7681

Merged
davebayer merged 2 commits intoNVIDIA:mainfrom
miscco:fix_warpspeed_atomics_nvhpc
Feb 16, 2026
Merged

Fix max atomic size for NVHPC#7681
davebayer merged 2 commits intoNVIDIA:mainfrom
miscco:fix_warpspeed_atomics_nvhpc

Conversation

@miscco
Copy link
Contributor

@miscco miscco commented Feb 16, 2026

We were hardcoding 16 bytes as the hard limit for the lookahead algorithm.

However, that is not correct for NVHPC which only supports 4 or 8 byte atomics

Also be safe and conditionally switch based on SM90 because that is required for 16 byte atomics

Fixes #7650

We were hardcoding 16 bytes as the hard limit for the lookahead algorithm.

However, that is not correct for NVHPC which only supports 4 or 8 byte atomics

Also be safe and conditionally switch based on SM90 because that is required for 16 byte atomics

Fixes NVIDIA#7650
@miscco miscco requested a review from a team as a code owner February 16, 2026 08:41
@miscco miscco requested a review from NaderAlAwar February 16, 2026 08:41
@github-project-automation github-project-automation bot moved this to Todo in CCCL Feb 16, 2026
@cccl-authenticator-app cccl-authenticator-app bot moved this from Todo to In Review in CCCL Feb 16, 2026
@davebayer davebayer enabled auto-merge (squash) February 16, 2026 10:06
@github-actions

This comment has been minimized.

@github-actions
Copy link
Contributor

🥳 CI Workflow Results

🟩 Finished in 10h 15m: Pass: 100%/99 | Total: 5d 15h | Max: 5h 38m | Hits: 72%/113374

See results here.

@davebayer davebayer merged commit eef7ced into NVIDIA:main Feb 16, 2026
223 of 232 checks passed
@github-project-automation github-project-automation bot moved this from In Review to Done in CCCL Feb 16, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Archived in project

Development

Successfully merging this pull request may close these issues.

[BUG]: Static assertion failure in atomic_ref for double scan on sm_120 (<=16 vs <16 size check mismatch)

2 participants