https://bugs.kde.org/show_bug.cgi?id=404076
--- Comment #5 from Julian Seward <jsew...@acm.org> ---
(In reply to Andreas Arnez from comment #4)
> Created attachment 132244 [details]
> s390x: Support for z14 (arch12) vector instructions
This all looks totally reasonable to me; OK to land as-is.
My only observation is:
@@ -2219,6 +2256,11 @@ static IRExpr *
get_vr(UInt archreg, IRType type, UChar index)
{
UInt offset = s390_vr_offset_by_index(archreg, type, index);
+ if (type == Ity_F128) {
+ return binop(Iop_F64HLtoF128,
+ IRExpr_Get(offset, Ity_F64),
+ IRExpr_Get(offset + 8, Ity_F64));
+ }
Can you use simply IRExpr_Get(offset, Ity_F128) ? If you can, it would give
better performance than doing the two F64 Gets and then joining the results
together.
--
You are receiving this mail because:
You are watching all bug changes.
