Stream: git-wasmtime

Topic: wasmtime / Issue #1886 wasmtime: support reference types ...


view this post on Zulip Wasmtime GitHub notifications bot (Jun 16 2020 at 17:02):

fitzgen opened Issue #1886:

I think this should mostly Just Work after https://github.com/bytecodealliance/wasmtime/issues/1617 lands, but we also need to remove these cfgs (and make sure the tests pass, ofc):

view this post on Zulip Wasmtime GitHub notifications bot (Jun 16 2020 at 17:02):

fitzgen labeled Issue #1886:

I think this should mostly Just Work after https://github.com/bytecodealliance/wasmtime/issues/1617 lands, but we also need to remove these cfgs (and make sure the tests pass, ofc):

view this post on Zulip Wasmtime GitHub notifications bot (Jun 16 2020 at 18:45):

cfallin commented on Issue #1886:

For tracking purposes: the initial support for the ref values and opcodes themselves is in the WIP PR #1852, and @julian-seward1 and I are working out the safepoints story from the regalloc.rs and isel sides respectively. We should have at least an initial version done in the next few weeks.

We're pushing things forward for the SpiderMonkey use-case and planning to test there, so I'll be interested to see how we interface with wasmtime's new reftypes support you've been building. The stackmaps interface will likely need a bit of finessing but I'm happy to work that out once we get there.

view this post on Zulip Wasmtime GitHub notifications bot (Jun 16 2020 at 18:45):

cfallin edited a comment on Issue #1886:

For tracking purposes: the initial support for the ref values and opcodes themselves on AArch64 is in the WIP PR #1852, and @julian-seward1 and I are working out the safepoints story from the regalloc.rs and isel sides respectively. We should have at least an initial version done in the next few weeks.

We're pushing things forward for the SpiderMonkey use-case and planning to test there, so I'll be interested to see how we interface with wasmtime's new reftypes support you've been building. The stackmaps interface will likely need a bit of finessing but I'm happy to work that out once we get there.

view this post on Zulip Wasmtime GitHub notifications bot (Jun 16 2020 at 19:18):

fitzgen commented on Issue #1886:

Thanks for the update @cfallin!

FYI we now have test stackmaps support in filetests now too, so that should hopefully make testing in-tree a little easier for you:

The stack maps interfaces definitely need a little finessing (one example off the top of my head would be to have "semantic" methods rather than just exposing the inner bit map directly and expecting callers to figure out how to interpret that; also I suspect we can do better than iterating over every word in the frame and have an iterator over only the live refs instead).

As long as everything remains SP-relative, then I think Wasmtime should be fine (getting SP portably and reliably is much easier than getting the CFA with libunwind).

view this post on Zulip Wasmtime GitHub notifications bot (Nov 13 2020 at 17:52):

cfallin commented on Issue #1886:

Fixed by #2410.

view this post on Zulip Wasmtime GitHub notifications bot (Nov 13 2020 at 17:52):

cfallin closed Issue #1886:

I think this should mostly Just Work after https://github.com/bytecodealliance/wasmtime/issues/1617 lands, but we also need to remove these cfgs (and make sure the tests pass, ofc):


Last updated: Nov 22 2024 at 17:03 UTC