Skip to content

CUDA Compiler Refactor #875

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

Jump to bottom
Merged
pvelesko merged 31 commits into from
Jun 29, 2024
Merged

CUDA Compiler Refactor #875

pvelesko merged 31 commits into from
Jun 29, 2024

Conversation

@pvelesko
Copy link
Collaborator

@pvelesko pvelesko commented Jun 18, 2024 •
edited
Loading

Looked into rewriting cucc in C++ but I really don't see any benefit for it. We have a python dependency through unit testing anyways.

Fixes #877
Fixes #874

  • Incorporate vector operation overloading fix from Jenny
  • CUDA compiler drop cmake configuration
  • nvcc symbolink link
  • Implement __shfl_sync variants partially (mask off/off or print error)
  • cuda_bf16.h support
  • cuda_fp16.h support
  • math_constants.h
  • map cudaMallocAsync, cudaFreeAsync to serial versions

Future work:

  • cuda_runtime.h is not C compatible and some HeCBench tests fail to compile for that reason.
  • Implement _sync support for masks other than all 0 or 1

@pvelesko pvelesko mentioned this pull request Jun 18, 2024
Closed
@pvelesko pvelesko mentioned this pull request Jun 21, 2024
Closed
@pvelesko pvelesko marked this pull request as ready for review June 25, 2024 13:30
@pvelesko pvelesko requested review from franz and pjaaskel June 25, 2024 13:47
@Kerilk Kerilk requested a review from jjennychen June 27, 2024 14:49
@pvelesko pvelesko merged commit 4edbcb6 into main Jun 29, 2024
@pvelesko pvelesko deleted the cucc-cpp branch June 29, 2024 08:09
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@franz franz Awaiting requested review from franz

@pjaaskel pjaaskel Awaiting requested review from pjaaskel

@jjennychen jjennychen Awaiting requested review from jjennychen

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

cucc is a fork bomb if a symlink called nvcc pointing to cucc exists Double definitions in vector operator overloading for CUDA source code

3 participants

@pvelesko @jjennychen