Stream: git-wasmtime

Topic: wasmtime / Issue #2322 Encode modules with variable-lengt...


view this post on Zulip Wasmtime GitHub notifications bot (Oct 26 2020 at 02:02):

github-actions[bot] commented on Issue #2322:

Subscribe to Label Action

cc @peterhuene

<details>
This issue or pull request has been labeled: "wasmtime:api"

Thus the following users have been cc'd because of the following labels:

To subscribe or unsubscribe from this label, edit the <code>.github/subscribe-to-label.json</code> configuration file.

Learn more.
</details>

view this post on Zulip Wasmtime GitHub notifications bot (Nov 04 2020 at 18:52):

fitzgen commented on Issue #2322:

To be clear this is just enabling variable length encoding for integers, right? It is not switching to a delta encoding (which would ensure that essentially all addresses are 1-byte encoded) correct?

view this post on Zulip Wasmtime GitHub notifications bot (Nov 04 2020 at 19:26):

alexcrichton commented on Issue #2322:

Ah yes indeed, a delta encoding might make it even more further compact.

I've so far shied away from big changes like that though. I think we may want to do it eventually, but it trades-off in-memory footprint for lookup speed since we can no longer quickly do a random lookup for a particular trapping pc. Not that we necessarily need that to be too too fast.

view this post on Zulip Wasmtime GitHub notifications bot (Nov 04 2020 at 19:31):

fitzgen commented on Issue #2322:

I'm only advocating for delta encoding in the serialized format, not the in-memory representation.

view this post on Zulip Wasmtime GitHub notifications bot (Nov 05 2020 at 11:49):

tschneidereit commented on Issue #2322:

I'm only advocating for delta encoding in the serialized format, not the in-memory representation.

I think it'd be good to try not to have those two deviate too much: ideally we should at some point be able to mmap the on-disk representation entirely or almost entirely instead of having to read and deserialize it.

There clearly are size/speed tradeoffs to be had here, so that might not be the only thing we support, but I hope it can become a thing we support :)


Last updated: Oct 23 2024 at 20:03 UTC