Fix async error reporting and memory type handling in CCLAdaptor and DeviceAdaptor - #266
Conversation
Summary of ChangesHello @MC952-arch, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request addresses several critical bug fixes within the Highlights
Ignored Files
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces several bug fixes. The main fix corrects the signature of flagcxCommGetAsyncError across multiple files, changing an output parameter from pass-by-value to pass-by-pointer. This ensures that the asynchronous error status is correctly propagated to the caller. Additionally, the memory allocation and deallocation logic in several device adaptors has been refactored. While the refactoring is good, I've suggested making the conditional logic more explicit to improve code clarity and robustness against future changes.
| } else { | ||
| if (stream == NULL) { | ||
| DEVCHECK(cudaFree(ptr)); | ||
| } else { | ||
| DEVCHECK(cudaFreeAsync(ptr, stream->base)); | ||
| } | ||
| } else if (type == flagcxMemManaged) { | ||
| DEVCHECK(cudaFree(ptr)); | ||
| } |
There was a problem hiding this comment.
For improved clarity and robustness, it's better to explicitly check for flagcxMemDevice rather than relying on an else block. This makes the code easier to understand and safer if new flagcxMemType_t values are added in the future.
} else if (type == flagcxMemDevice) {
if (stream == NULL) {
DEVCHECK(cudaFree(ptr));
} else {
DEVCHECK(cudaFreeAsync(ptr, stream->base));
}
}| } else { | ||
| if (stream == NULL) { | ||
| DEVCHECK(cudaFree(ptr)); | ||
| } else { | ||
| DEVCHECK(cudaFreeAsync(ptr, stream->base)); | ||
| } | ||
| } else if (type == flagcxMemManaged) { | ||
| DEVCHECK(cudaFree(ptr)); | ||
| } |
There was a problem hiding this comment.
For improved clarity and robustness, it's better to explicitly check for flagcxMemDevice rather than relying on an else block. This makes the code easier to understand and safer if new flagcxMemType_t values are added in the future.
| } else { | |
| if (stream == NULL) { | |
| DEVCHECK(cudaFree(ptr)); | |
| } else { | |
| DEVCHECK(cudaFreeAsync(ptr, stream->base)); | |
| } | |
| } else if (type == flagcxMemManaged) { | |
| DEVCHECK(cudaFree(ptr)); | |
| } | |
| } else if (type == flagcxMemDevice) { | |
| if (stream == NULL) { | |
| DEVCHECK(cudaFree(ptr)); | |
| } else { | |
| DEVCHECK(cudaFreeAsync(ptr, stream->base)); | |
| } | |
| } |
| } else { | ||
| if (stream == NULL) { | ||
| DEVCHECK(mcMalloc(ptr, size)); | ||
| } else { | ||
| DEVCHECK(mcMallocAsync(ptr, size, stream->base)); | ||
| } | ||
| } else if (type == flagcxMemManaged) { | ||
| DEVCHECK(mcMallocManaged(ptr, size, mcMemAttachGlobal)); | ||
| } |
There was a problem hiding this comment.
For improved clarity and robustness, it's better to explicitly check for flagcxMemDevice rather than relying on an else block. This makes the code easier to understand and safer if new flagcxMemType_t values are added in the future.
| } else { | |
| if (stream == NULL) { | |
| DEVCHECK(mcMalloc(ptr, size)); | |
| } else { | |
| DEVCHECK(mcMallocAsync(ptr, size, stream->base)); | |
| } | |
| } else if (type == flagcxMemManaged) { | |
| DEVCHECK(mcMallocManaged(ptr, size, mcMemAttachGlobal)); | |
| } | |
| } else if (type == flagcxMemDevice) { | |
| if (stream == NULL) { | |
| DEVCHECK(mcMalloc(ptr, size)); | |
| } else { | |
| DEVCHECK(mcMallocAsync(ptr, size, stream->base)); | |
| } | |
| } |
| } else { | ||
| if (stream == NULL) { | ||
| DEVCHECK(musaMalloc(ptr, size)); | ||
| } else { | ||
| DEVCHECK(musaMalloc(ptr, size)); | ||
| // MUSA currently does not support async malloc | ||
| // DEVCHECK(musaMallocAsync(ptr, size, stream->base)); | ||
| } | ||
| } else if (type == flagcxMemManaged) { | ||
| DEVCHECK(musaMallocManaged(ptr, size, musaMemAttachGlobal)); | ||
| } |
There was a problem hiding this comment.
For improved clarity and robustness, it's better to explicitly check for flagcxMemDevice rather than relying on an else block. This makes the code easier to understand and safer if new flagcxMemType_t values are added in the future.
| } else { | |
| if (stream == NULL) { | |
| DEVCHECK(musaMalloc(ptr, size)); | |
| } else { | |
| DEVCHECK(musaMalloc(ptr, size)); | |
| // MUSA currently does not support async malloc | |
| // DEVCHECK(musaMallocAsync(ptr, size, stream->base)); | |
| } | |
| } else if (type == flagcxMemManaged) { | |
| DEVCHECK(musaMallocManaged(ptr, size, musaMemAttachGlobal)); | |
| } | |
| } else if (type == flagcxMemDevice) { | |
| if (stream == NULL) { | |
| DEVCHECK(musaMalloc(ptr, size)); | |
| } else { | |
| DEVCHECK(musaMalloc(ptr, size)); | |
| // MUSA currently does not support async malloc | |
| // DEVCHECK(musaMallocAsync(ptr, size, stream->base)); | |
| } | |
| } |
MC952-arch
changed the title
3 participants
This PR addresses two key issues: