YJIT: Optimize putobject+opt_ltlt for integers #10401
Conversation
|
Come to think of it, we might even want to remove the other implementation that don't know that rhs is constant for sure and only keep this one 🤔 |
|
This PR is pretty cool. I'm amazed that it makes that much of a performance difference on optcarrot. Waaaaaaaaat!? Might get us over the 4x faster threshold on speed.yjit.org 😆
There are some benchmarks that use a variable rhs in the shift. I think that ideally, we should implement variable-shift support in our backend/cross-platform-assembler. Now that you have support for generating code for multiple YARV instructions at once, it could make sense to use this to tackle comparison + branch insns, especially for the most common cases. Could be a significant win on most benchmarks. |
In `jit_rb_int_lshift()`, we guard against the right hand side changing since we want to avoid generating variable length shifts. When control reaches a `putobject` and `opt_ltlt` pair, though, we know that the right hand side never changes. This commit detects this situation and substitutes an implementation that does not guard against the right hand side changing, saving that work. Deleted some `putobject` Rust tests since they aren't that valuable and cause linking issues. Nice boost to `optcarrot` and `protoboeuf`: ``` ---------- ------------------ bench yjit-pre/yjit-post optcarrot 1.09 protoboeuf 1.12 ---------- ------------------ ```
In
jit_rb_int_lshift(), we guard against the right hand side changingsince we want to avoid generating variable length shifts. When control
reaches a putobject and
opt_ltltpair, though, we know that the righthand side never changes.
This commit detects this situation and substitutes an implementation
that does not guard against the right hand side changing, saving that
work.
Deleted some putobject Rust tests since they aren't that valuable and
cause linking issues.
Nice boost to
optcarrotandprotoboeuf: