Skip to content

Unified memory#157

Merged
simonpintarelli merged 16 commits intomasterfrom
unified_memory
Feb 4, 2026
Merged

Unified memory#157
simonpintarelli merged 16 commits intomasterfrom
unified_memory

Conversation

@mtaillefumier
Copy link
Copy Markdown
Collaborator

Add AMD unified memory support to COSMA

@simonpintarelli
Copy link
Copy Markdown
Member

cscs-ci run gh200

@simonpintarelli simonpintarelli self-requested a review December 4, 2025 08:12
@mtaillefumier
Copy link
Copy Markdown
Collaborator Author

I need to push a patch for fixing a typo

Copy link
Copy Markdown
Member

@simonpintarelli simonpintarelli left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm

Comment thread src/cosma/aligned_allocator.hpp Outdated
#else
}
#endif
}
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think the last closing } shouldn't be there.

Comment thread src/cosma/context.cpp Outdated
// requires memory_pool_.resize(cpu_mem_limit);
#ifdef COSMA_HAVE_GPU
gpu_ctx_ = gpu::make_context<Scalar>(streams, tile_m, tile_n, tile_k);
gpu_ctx_.use_unified_memory_ = cosma::get_unified_memory();
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

gpu_ctx_ is a pointer to gpu::mm_handle<..>, which is a class from tiled-mm.
But there is no corresponding PR to tiled-mm introducing a field use_unified_memory_.
Also probably needs an #ifdef COSMA_USE_UNIFIED_MEMORY here?

@simonpintarelli simonpintarelli merged commit 1343256 into master Feb 4, 2026
1 check passed
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.

3 participants