[PATCH] D77908: [WebAssembly] Enable nontrapping-fptoint for `default` cpu

[Alon Zakai via Phabricator via cfe-commits]

kripken added a comment.

Is the general plan for LLVM documented somewhere?

It's not obvious to me why something being in the wasm spec means it should be enabled by default in LLVM. (It's also not obvious to me that is wrong! I'm just not sure what the reasoning is here.)

In particular, something being in the wasm spec doesn't mean it is widespread yet. I see https://github.com/emscripten-core/emscripten/pull/10885 proposes to diverge the emscripten defaults from LLVMs. That seems odd to me - why shouldn't those be the same? If they aren't the same it's a potential source of confusion in multiple ways. For example, if the reasoning is "LLVM moves with the spec, emscripten moves with the web" then that means *non*-emscripten toolchains targeting the web have more work to do. Also comparisons between toolchains will get harder to get apples-to-apples.

Again, not opposed to this, just not sure what the big picture is.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D77908/new/

https://reviews.llvm.org/D77908