You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We could implement something similar to help identify optimizations that we are missing in v8 that already exist in gcc and LLVM. This could be implemented by compiling C code, generated by a fuzzer, into WASM, then running it through v8 and dumping the generated assembly code and comparing it to the assembly code which gets generated by LLVM/gcc for the same C code.
The text was updated successfully, but these errors were encountered:
This repo has the code used for the above mentioned poster. I found that it is easy to modify it to work with a different compiler. Of course, our situation is a bit different, since V8 is not just a compiler, but we can still dump generated assembly code, so we should be able to use this tool.
Migrated from riscv-collab/v8#291:
This idea is based on this poster from this year's LLVM Developers' Meeting: https://www.lowrisc.org/blog/2020/10/how-we-used-differential-testing-to-rapidly-find-and-fix-missed-optimisation-opportunities-in-llvms-risc-v-backend/llvm-dev-meeting-oct-2020-poster.pdf
We could implement something similar to help identify optimizations that we are missing in v8 that already exist in gcc and LLVM. This could be implemented by compiling C code, generated by a fuzzer, into WASM, then running it through v8 and dumping the generated assembly code and comparing it to the assembly code which gets generated by LLVM/gcc for the same C code.
The text was updated successfully, but these errors were encountered: