When the FFT multiplication algorithm invokes itself for the
recursive steps, the input is "mod Fn"-normalized, i.e. it is
at most of the shape (1 << N), but we only read N bits of it,
so in the rare case where it was exactly 1 << N, that lone top
bit was ignored, leading to an incorrect result of the overall
multiplication.
Fixed: chromium:1228267
Change-Id: I7b245fc3701696d95e5d75fb970f02d72ce40ff8
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3032081
Reviewed-by: Maya Lekova <mslekova@chromium.org>
Commit-Queue: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/master@{#75755}
cdb882518d