• Leszek Swirski's avatar
    [maglev] Make jump gap allocation match move emission · 5a716edd
    Leszek Swirski authored
    Jump gap moves (for phis and register merges) are emitted as a parallel
    move (i.e. treated as a single mapping from registers to registers and
    emitted in a way that they don't clobber each other). However, the phi
    input allocation was updating the register state as if they were
    serialised moves (i.e. a list of moves, one after the other, where each
    move could clobber another move's input).
    
    Now the jump phi initialisation doesn't update register state.
    
    Bug: v8:7700
    Change-Id: Iecf3211d59d9c416a4449aea22fef633717d92d3
    Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3784983Reviewed-by: 's avatarToon Verwaest <verwaest@chromium.org>
    Auto-Submit: Leszek Swirski <leszeks@chromium.org>
    Commit-Queue: Toon Verwaest <verwaest@chromium.org>
    Cr-Commit-Position: refs/heads/main@{#81934}
    5a716edd
maglev-regalloc.cc 57.8 KB