Description[asmjs] Do constant folding for I32Asmjs(Div|Rem)S to avoid checks of constant divisors
This change makes the embenchen/copy benchmark a factor of 2 faster and
brings back the performance loss through graph trimming.
R=titzer@chromium.org
CC=bradnelson@chromium.org
Committed: https://crrev.com/0ddddcb8629d456c678c1eeb3c1d1e27b8c426f2
Cr-Commit-Position: refs/heads/master@{#40628}
Patch Set 1 #Patch Set 2 : Add tests #
Messages
Total messages: 21 (12 generated)
|