https://bugs.kde.org/show_bug.cgi?id=523097
--- Comment #3 from [email protected] --- Created attachment 195469 --> https://bugs.kde.org/attachment.cgi?id=195469&action=edit proposed patch Rebased on top of master(In reply to Mark Wielaard from comment #2) > (In reply to mcermak from comment #1) > > Created attachment 194912 [details] > > proposed patch > > I am probably reviewing this in the wrong order. > The patch needs a bit of rebasing, Rebased in branch users/mcermak/try-bug523097-extractps-pextrb-pextrw . > the usage of breakupV128to32s in guest_x86_toIR.c should be breakup128to32s, I've merged them into breakupV128to32s() here: https://sourceware.org/git/?p=valgrind.git;a=commitdiff;h=44b56c971dacfa5092dd629238f8df19d31d6194 > and the sse4-x86.stdout.exp needs regeneration. Done. > > But with that updated things look good. > > One oddity I saw reading the code. > The amd64 variant has a comment for PEXTRW > /* The memory case is disallowed, apparently. */ and fails. > But the x86 just handles this (like the PEXTRB case). I think for this bug we need to look at the SSE4.1 one: VEX/priv/guest_amd64_toIR.c:18376 (that's dis_PEXTRW(), not dis_PEXTRW_128_EregOnly_toG()). That one was used for modelling the x86 decoder. Rebased patch attached. -- You are receiving this mail because: You are watching all bug changes.
