Accompanying https://reviews.llvm.org/D125728, this updates LLVM Codegen's "generic" CPU to enable the same new features.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Thanks, I think CodeGen/WebAssembly/target-features.ll could be updated to add a -mcpu=generic test?
Update tests to use -mcpu=mvp to avoid breaking their target features when -mcpu=generic changes. And update target-features.ll to for the new -mcpu=generic features.
LGTM, but it would be good to get an approval from someone else too.
llvm/test/CodeGen/WebAssembly/target-features.ll | ||
---|---|---|
57–58 | This comment is now out of date. |
On the emscripten side we still need to come up with a framework that will disable these new defaults when users select older browser support. I'm not sure if we want to push back on landing this until such as frameworks is in place, or just go ahead and land this and deal with the change when it rolls in. @tlively @dschuff @kripken ?
llvm/lib/Target/WebAssembly/WebAssembly.td | ||
---|---|---|
99 | It might be worth explaining what "latest stable" means here: how were the features here chosen, how often this is updated, etc. Maybe we could have a page in the tool-conventions repo with a doc for all that? Ideally the plan there would cover all major wasm toolchains. Then we could all link to there. I think that would really help users in our ecosystem understand how this stuff works. |
- Add a comment about what "generic" means.
llvm/lib/Target/WebAssembly/WebAssembly.td | ||
---|---|---|
99 | I've now updated this to say:
That is somewhat vague, but I believe represents our actual intent here. I'll post to https://github.com/WebAssembly/tool-conventions/issues/158 about this for broader discussion. |
llvm/lib/Target/WebAssembly/WebAssembly.td | ||
---|---|---|
99 | Sounds good, thanks. |
This might have broken check-lld: http://45.33.8.238/linux/89856/step_11.txt
Please take a look and revert for now if it takes a while to fix.
I've now submitted b5d0bf9b9853688d34290fafdd31c95aca58f624 which fixes these failures.
Thanks!
Looks like it breaks check-llvm too: http://45.33.8.238/linux/89859/step_12.txt
Yes, we are seeing these 10 failures on our internal bot as well the public bot https://lab.llvm.org/buildbot/#/builders/109/builds/49346
I tried to fix the remaining tests by adding -mcpu=mvp, but at least 2 of the 10 tests (debug-info.ll and debug-info64.ll) still failed and it wasn't obvious how to fix them, so I ended up reverting your original change and the follow-up commit to get the bots back to green since these tests are failing on a lot of bots.
It might be worth explaining what "latest stable" means here: how were the features here chosen, how often this is updated, etc.
Maybe we could have a page in the tool-conventions repo with a doc for all that? Ideally the plan there would cover all major wasm toolchains. Then we could all link to there. I think that would really help users in our ecosystem understand how this stuff works.