Chris Sullivan

Support using GPU via CUDA

12

I am using standard gdb with code that has CUDA library calls. I don't need to debug the CUDA specific code, but would like the functionality of rr for the...

Method overloading via interfaces and type mistmatch

5

First off, allow me to compliment you on this nice python library and f2py extension. After some work, I got it working for my library, but I did run into...

[TIR][CUDA] Add native FP8 support to codegen

1

Adds native FP8 type support for CUDA. The e4m3/e5m2 struct types provide explicit type conversions that target hardware native conversion ops. \* Conditionally run Storage and Compute legalization for targets...

[Disco] Add MSCCLPP initialization along side NCCL

1

* Use CCL type traits to share common code between NCCL and MSCCLPP API invocations in disco * Add bench to validate results and compare various supported CCL approaches for...

[Blackwell] Support DescriptorLoadOp when deciding to use shared memory for scales This ensures scales produced by TMA are elligible for transfer to tensor memory in later lowering. #### [PR chain](https://github.com/jlebar/git-pr-chain)...

[Bench][Blackwell] Fix warp specialization for fp8 x mxfp4 bench Applies a bug fix for padded scale loads in fp8xmxfp4 mode ensuring TMA load requirements are met when using the unpacked...

[Bench][Blackwell] Support optional scale TMAs in warp specialization for tl.dot_scaled This enables automatic warp specialization for block scaled workloads. #### [PR chain](https://github.com/jlebar/git-pr-chain) 1. #6535 1. 👉 #6536 👈 **YOU ARE...