release/19.x: [PowerPC] Fix assert exposed by PR 95931 in LowerBITCAST (#108062) #108834
Conversation
|
@amy-kwan What do you think about merging this PR to the release branch? |
|
@llvm/pr-subscribers-backend-powerpc Author: None (llvmbot) ChangesBackport 22067a8 Requested by: @syzaara Full diff: https://github.com/llvm/llvm-project/pull/108834.diff 2 Files Affected:
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index 21cf4d9eeac173..758de9d732fa7e 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -9338,12 +9338,13 @@ SDValue PPCTargetLowering::LowerBITCAST(SDValue Op, SelectionDAG &DAG) const {
SDLoc dl(Op);
SDValue Op0 = Op->getOperand(0);
+ if (!Subtarget.isPPC64() || (Op0.getOpcode() != ISD::BUILD_PAIR) ||
+ (Op.getValueType() != MVT::f128))
+ return SDValue();
+
SDValue Lo = Op0.getOperand(0);
SDValue Hi = Op0.getOperand(1);
-
- if ((Op.getValueType() != MVT::f128) ||
- (Op0.getOpcode() != ISD::BUILD_PAIR) || (Lo.getValueType() != MVT::i64) ||
- (Hi.getValueType() != MVT::i64) || !Subtarget.isPPC64())
+ if ((Lo.getValueType() != MVT::i64) || (Hi.getValueType() != MVT::i64))
return SDValue();
if (!Subtarget.isLittleEndian())
diff --git a/llvm/test/CodeGen/PowerPC/f128-bitcast.ll b/llvm/test/CodeGen/PowerPC/f128-bitcast.ll
index ffbfbd0c64ff3f..55ba3cb1e05387 100644
--- a/llvm/test/CodeGen/PowerPC/f128-bitcast.ll
+++ b/llvm/test/CodeGen/PowerPC/f128-bitcast.ll
@@ -86,3 +86,25 @@ entry:
ret i64 %1
}
+define <4 x i32> @truncBitcast(i512 %a) {
+; CHECK-LABEL: truncBitcast:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: mtvsrdd v2, r4, r3
+; CHECK-NEXT: blr
+;
+; CHECK-BE-LABEL: truncBitcast:
+; CHECK-BE: # %bb.0: # %entry
+; CHECK-BE-NEXT: mtvsrdd v2, r9, r10
+; CHECK-BE-NEXT: blr
+;
+; CHECK-P8-LABEL: truncBitcast:
+; CHECK-P8: # %bb.0: # %entry
+; CHECK-P8-NEXT: mtfprd f0, r3
+; CHECK-P8-NEXT: mtfprd f1, r4
+; CHECK-P8-NEXT: xxmrghd v2, vs1, vs0
+; CHECK-P8-NEXT: blr
+entry:
+ %0 = trunc i512 %a to i128
+ %1 = bitcast i128 %0 to <4 x i32>
+ ret <4 x i32> %1
+}
|
|
@tru hi Tobias, can you please help to merge this PR? Thank you! |
Hit Assertion failed: Num < NumOperands && "Invalid child # of SDNode!" Fix by checking opcode and value type before calling getOperand. (cherry picked from commit 22067a8)
|
@syzaara (or anyone else). If you would like to add a note about this fix in the release notes (completely optional). Please reply to this comment with a one or two sentence description of the fix. When you are done, please add the release:note label to this PR. |
|
@syzaara, not sure if you were notified, but the cherry-pick of this change to the release branch seems to be causing a test failure. Can you take a look? |
|
I am pretty sure this is not the PR that broke the tests since all the failures are related to LoongArch. I will try to revert some of the loongarch commits locally. |
Backport 22067a8
Requested by: @syzaara