[webkit-reviews] review requested: [Bug 196091] B3::Value should have different kinds of adjacency lists : [Attachment 366116] Patch

[bugzilla-daemon at webkit.org]

Robin Morisset <rmorisset at apple.com> has asked	for review:
Bug 196091: B3::Value should have different kinds of adjacency lists
https://bugs.webkit.org/show_bug.cgi?id=196091

Attachment 366116: Patch

https://bugs.webkit.org/attachment.cgi?id=366116&action=review




--- Comment #15 from Robin Morisset <rmorisset at apple.com> ---
Created attachment 366116

  --> https://bugs.webkit.org/attachment.cgi?id=366116&action=review

Patch

This uses an inline capacity of 3 (as before) for the VarArgs case. I have an
alternative patch that customizes the inline capacity to the opcode, but the
implementation gets extremely gross, for (I think) little benefit.
A global run on RAMification was inconclusive (dominated by noise in LuaJSFight
among other things), so I ran 4 runs of RAMification on my machine, only
looking at the peak footprint of the wasm benchmarks, after having disabled
their call to main to only measure the compilation footprint.
The result was a rather disappointing 1% reduction in peak footprint.
I'd like to land this since it is still an improvement, and monitor the bots
for any regression/progression.