Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[wamr] Meeting notes for 09-02 #410

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
42 changes: 42 additions & 0 deletions wamr/2024/wamr-09-02.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,42 @@
# WAMR community Meeting Monthly

**See the [instructions](./README.md) for details on how to attend**

## Agenda

1. Opening, welcome and roll call
2. Announce that the meeting will be recorded and hit record as agreed upon in inaugural meeting.
3. Announcements
a. _Submit a PR to add your announcement here_
4. Other agenda items
- Reduce AOT code size when stack trace is enabled

## Attendees

- Jun Xu (Intel)
- Marcin Kolny (Amazon)
- Tianlong Liang (Intel)
- Wenyong Huang (Intel)
- Xin Wang (Intel)
- yamamoto (Midokura/Sony)

## Notes

- Wenyong Huang shared some updates of WAMR:
- The WAMR-2.1.2 was released and the multi-memory feature was merged to main branch after the releasing
- The PR of WASI support for file io and socket on Zephyr platform is under review and most of the comments were addressed
- Intel is adding llama.cpp as a new backend of wasi-nn and successfully ran several LLM models downloaded from huggingface
- Xiaomi started to implement the shared feature and the first version is mainly to support the interpreter mode
- Intel open-sourced the wasm2native project: https://github.com/web-devkits/wasm2native, which compiles WebAssembly to native binary and supports two compilation modes: sandbox mode and no-sandbox mode.

- Discussed how to reduce the generated code size in AOT file when stack trace (or dump-call-stack) is enabled (RFC https://github.com/bytecodealliance/wasm-micro-runtime/issues/3758):
- Basically agree to enable the option 2 which allocates/frees the stack frame at the beginning/returning of a wasm function's AOT code and happens one time so as to reduce the code size in most cases. The current implementation allocates/frees the frame before/after calling a function, since calling a same function may occur in many places, the AOT code of frame allocation/freeing may duplicate many times.
- To enable option 2, we may use another bit in the feature flag of AOT file, which enables the AOT time to keep the backward compatibility.
- Basically agree to add option `--dump-call-stack-features=` for wamrc to set which features/fields to store into the stack frame, e.g., "func-idx", "ip", "sp", "locals" and so on. The feature strings are supposed to be simple and short.

- yamamoto asked why Intel decided to make wasm2native an independent project from wamr?
Wenyong Huang: though wasm2native is ported from wamr project, it is difficult to share the same code base, since it removes the AOT runtime dependency and changes a lot in the AOT compilation process: wamr's aot function's first argument is exec_env and aot function accesses the wasm instance data through exec_env, while wasm2native aot function hasn't this argument and all the instance data is defined/created in the LLVM module (instead of created by AOT runtime). The LLVM IR generated by wasm2native compiler is very different from the LLVM IR generated by wamr's aot compiler.

## Action Items

* [ ] TODO
24 changes: 0 additions & 24 deletions wamr/2024/wamr-09-09.md

This file was deleted.