Skip to content

[Runtime] Use preferred host memory (pinned memory) in KV cache#17036

Merged
tqchen merged 1 commit intoapache:mainfrom
MasterJH5574:tvm-dev/2024-05-28-kv-cache-pinned-mem
May 29, 2024
Merged

[Runtime] Use preferred host memory (pinned memory) in KV cache#17036
tqchen merged 1 commit intoapache:mainfrom
MasterJH5574:tvm-dev/2024-05-28-kv-cache-pinned-mem

Conversation

@MasterJH5574
Copy link
Copy Markdown
Contributor

This PR updates the PagedKVCache with the pinned memory support, which can reduce the copy overhead between CPU and GPU.

This PR also bumps FlashInfer version, which now supports

  • specifying kernels to build via cmake,
  • pinned memory as host memory.

We also update CMakeLists.txt and config.cmake to include the FlashInfer compile options. Prior to this PR, the kernels being built is hardcoded in FlashInfer header files.

This PR updates the PagedKVCache with the pinned memory support,
which can reduce the copy overhead between CPU and GPU.

This PR also bumps FlashInfer version, which now supports
* specifying kernels to build via cmake,
* pinned memory as host memory.

We also update CMakeLists.txt and config.cmake to include the
FlashInfer compile options. Prior to this PR, the kernels being
built is hardcoded in FlashInfer header files.
@tqchen
Copy link
Copy Markdown
Member

tqchen commented May 29, 2024

@tvm-bot rerun

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