[llvm] [TableGen] Reduce stack usage of setTargetRuntimeLibcallSets (PR #162194)

via llvm-commits llvm-commits at lists.llvm.org
Mon Oct 6 21:22:46 PDT 2025


Nadharm wrote:

> Actually we should continue to use a range loop over the libcall entry functions. Fully unrolling this is much larger code. That is, we should roll back this, and [9c361cc](https://github.com/llvm/llvm-project/commit/9c361cc0685a7b4c69994748b2ef159e5ce5e018) and do what it originally did

Ah fair, yeah. As long as [9c361cc](https://github.com/llvm/llvm-project/commit/9c361cc0685a7b4c69994748b2ef159e5ce5e018) is also reverted and we go back to the `static` array implementation, this would also resolve the stack usage regression.

> My understanding is that the stack regression was due to in place lists

Yeah I think this is another case of MSVC not optimizing initializer lists, but I don't think that should be an issue in the pre-[9c361cc](https://github.com/llvm/llvm-project/commit/9c361cc0685a7b4c69994748b2ef159e5ce5e018)  static array implementation.

https://github.com/llvm/llvm-project/pull/162194


More information about the llvm-commits mailing list