https://bugs.kde.org/show_bug.cgi?id=385210
--- Comment #2 from Julian Seward <jsew...@acm.org> ---
diff --git a/VEX/priv/guest_ppc_toIR.c b/VEX/priv/guest_ppc_toIR.c
index 3897b9f2f..79d530d71 100644
--- a/VEX/priv/guest_ppc_toIR.c
+++ b/VEX/priv/guest_ppc_toIR.c
@@ -24297,12 +24297,12 @@ static Bool dis_av_permute ( UInt theInstr )
IRTemp vC_andF = newTemp(Ity_V128);
DIP("vperm v%d,v%d,v%d,v%d\n",
vD_addr, vA_addr, vB_addr, vC_addr);
- /* Limit the Perm8x16 steering values to 0 .. 15 as that is what
+ /* Limit the Perm8x16 steering values to 0 .. 31 as that is what
IR specifies, and also to hide irrelevant bits from
memcheck */
assign( vC_andF,
binop(Iop_AndV128, mkexpr(vC),
- unop(Iop_Dup8x16, mkU8(0xF))) );
+ unop(Iop_Dup8x16, mkU8(0x1F))) );
assign( a_perm,
binop(Iop_Perm8x16, mkexpr(vA), mkexpr(vC_andF)) );
assign( b_perm,
Is this hunk necessary? It seems to be changing the semantics of an
existing instruction, rather than merely re-implementing one.
Otherwise looks OK to me.
--
You are receiving this mail because:
You are watching all bug changes.
