Apr 09, 2024 06:00:00

AMD is working to further its open source efforts by releasing documentation for its MicroEngine Scheduler and more

AMD, whose open source driver efforts have earned it recognition and expanded its Linux market share , is now working to open source the content of its Micro Engine Scheduler (MES), including documentation.

AMD Working To Release MES Documentation & Source Code - Phoronix
https://www.phoronix.com/news/AMD-MES-Docs-And-Source-Code

AMD announces open-sourcing of its GPU software stack and documentation, including MES | Tom's Hardware
https://www.tomshardware.com/pc-components/gpus/amd-announces-open-sourcing-of-its-gpu-software-stack-and-documentation-including-mes

AMD has been actively involved in 'open source' for some time, and in April 2024, due to growing interest in ROCm (Radeon Open Compute platform) in the Radeon community, AMD revealed that it had created a tracker to obtain feedback and provide updates, and announced that it would soon open source additional parts of its GPU software stack and other hardware documentation.

As community interest grows in ROCm on Radeon, we've created a tracker to capture feedback and provide updates.

Coming soon: Open sourcing additional portions of our software stack and more hardware documentation. https://t.co/huuaA63Kds
— AMD Radeon (@amdradeon) April 2, 2024

In fact, it has become clear that work is underway to release MES documentation and other materials at the end of May 2024.

We are working to release Micro-Engine Scheduler(MES) documentation towards the end of May and will follow up with published source code for external review and feedback. We have also opened a GitHub tracker, which will have the latest status on fixes and release dates. https://t.co/SU5Q1k5tCS
— AMD Radeon (@amdradeon) April 4, 2024

The movement towards making MES documentation public was triggered by an incident where tinycorp, the developer of the neural network framework ' tinygrad ,' experienced the same MES error during training and pointed out that there was a 95% chance that it was a compiler bug.

2nd training run crashed with same MES error, beta driver piece didn't work, and found what I'm 95% sure is a compiler bug. https://t.co/oWowF07Duq

At it stands, I'm not okay with shipping the 7900XTX platform. What should we do?
— the tiny corp (@__tinygrad__) March 5, 2024

Tiny Corp, which was struggling to respond, posted, 'If AMD open-sources their firmware, we'll fix the LLVM spill bug and create a fuzzer for HSA, but it's not worth the effort to fix bugs on a platform I don't own.'

If AMD open sources their firmware, I'll fix their LLVM spilling bug and write a fuzzer for HSA. Otherwise, it's not worth putting tons of effort into fixing bugs on a platform you don't own. https://t.co/c4I2So27YG
— the tiny corp (@__tinygrad__) March 5, 2024

AMD CEO Lisa Su responded to the post, saying, 'Thank you for your cooperation and feedback. We're working hard to provide you with a good solution. Our team is working on it.'