poly_Rq_mul() uses ymm registers, so vzeroupper needs to be executed before returning in order to avoid slowing down subsequent SSE code. Change-Id: Id85e4ede05c612e0edf4c92a298531dd4c358bf4 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/77229 Reviewed-by: David Benjamin <davidben@google.com> Commit-Queue: David Benjamin <davidben@google.com>