[X86] Relax an assertion when legalizing vector types.

WidenVSELECTAndMask can fold (and it folds in this case) so we
get a BUILD_VECTOR of constants as mask. convertMask() seems to
work fine when the input is a vector of constants, and we still
need to call it to extend/add elements at the end. but the current
code just asserts on anything but a SETCC or AND/OR/XOR of 2xSETCC.
This change was discussed briefly with Simon Pilgrim, who also
suggests we might consider dropping this assertion in the future.

Fixes PR33715.

llvm-svn: 307508
This commit is contained in:
Davide Italiano 2017-07-09 19:22:48 +00:00
parent b2f8b311d1
commit c4b0ccd049
2 changed files with 20 additions and 0 deletions

View File

@ -2977,7 +2977,11 @@ SDValue DAGTypeLegalizer::convertMask(SDValue InMask, EVT MaskVT,
// Currently a SETCC or a AND/OR/XOR with two SETCCs are handled.
unsigned InMaskOpc = InMask->getOpcode();
// FIXME: This code seems to be too restrictive, we might consider
// generalizing it or dropping it.
assert((InMaskOpc == ISD::SETCC ||
ISD::isBuildVectorOfConstantSDNodes(InMask.getNode()) ||
(isLogicalMaskOp(InMaskOpc) &&
isSETCCorConvertedSETCC(InMask->getOperand(0)) &&
isSETCCorConvertedSETCC(InMask->getOperand(1)))) &&

View File

@ -0,0 +1,16 @@
; Make sure we don't crash with a build vector of integer constants.
; RUN: llc %s -o /dev/null
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"
define i32 @patatino() {
%tmp = insertelement <4 x i32> <i32 1, i32 1, i32 undef, i32 undef>, i32 1, i32 2
%tmp1 = insertelement <4 x i32> %tmp, i32 1, i32 3
%tmp2 = icmp ne <4 x i32> %tmp1, zeroinitializer
%tmp3 = icmp slt <4 x i32> %tmp1, <i32 4, i32 4, i32 4, i32 4>
%tmp4 = or <4 x i1> %tmp2, %tmp3
%tmp5 = select <4 x i1> %tmp4, <4 x i32> zeroinitializer, <4 x i32> <i32 4, i32 4, i32 4, i32 4>
%tmp6 = extractelement <4 x i32> %tmp5, i32 0
ret i32 %tmp6
}