Skip to content

Start on GPU extensions#276

Closed
kshyatt wants to merge 81 commits intoQuantumKitHub:matrixalgebrafrom
kshyatt:ksh/gpu
Closed

Start on GPU extensions#276
kshyatt wants to merge 81 commits intoQuantumKitHub:matrixalgebrafrom
kshyatt:ksh/gpu

Conversation

@kshyatt
Copy link
Copy Markdown
Member

@kshyatt kshyatt commented Aug 26, 2025

Lots of stuff still broken but some things are working

@kshyatt
Copy link
Copy Markdown
Member Author

kshyatt commented Aug 29, 2025

cc @Jutho and @lkdvos

Where we're at now:

  • Most tests are passing for CUDA except for the @constinferred (haven't tried super recently) and a correctness test in rightorth with Polar
  • CUDA extension is currently using CUDA.DeviceMemory hardcoded because this dramatically simplifies getting all the types to play nicely. My goal is to get to a "minimum passing extension" for the tests and then experiment with what can be simplified and stripped out.
  • Once I get the CUDA extension shaped up and working nicely, I'll try to graft as much as I can to AMD.
  • We're still waiting on some PRs I made to CUDA.jl itself to fix interface issues to merge and for a new version to tag, that will also help with tests passing.

@lkdvos lkdvos force-pushed the matrixalgebra branch 4 times, most recently from 46f8360 to fe980c0 Compare October 3, 2025 02:00
@lkdvos lkdvos deleted the branch QuantumKitHub:matrixalgebra October 3, 2025 12:19
@lkdvos lkdvos closed this Oct 3, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants