Investigate LLVM Codegen: Atomic Variables and Atomic Instructions

[skyzh]

From my previous experience of RISC-V programming with Rust, I found that Rust won't generate amoswap instruction for spin lock. That's why I wrote __sync_lock_release and __sync_lock_test_and_set in arch.rs. On that, spin crate will generate non-atomic instruction for implementing the lock, which may cause significant slowdown. This affects all Atomic type in Rust.

[skyzh]

Related pages: https://rcore-os.github.io/rCore_tutorial_doc/chapter3/part4.html

Maybe we should add '+a' option somewhere in Rust.

[skyzh]

It seems that rustc has already enabled atomic instruction.

riscv64gc target: https://github.com/rust-lang/rust/blob/master/src/librustc_target/spec/riscv64gc_unknown_none_elf.rs
riscv64imac target: https://github.com/rust-lang/rust/blob/master/src/librustc_target/spec/riscv64imac_unknown_none_elf.rs

Therefore, this should be a issue related to LLVM.

Author of one project similar to core-os proposed that: LLVM will generate lr/sc for atomic instructions. https://github.com/Jaic1/xv6-riscv-rust

Maybe we should wait for LLVM support for RISC-V atomic instruction.