[llvm] r297001 - [PowerPC] Fix failure with STBRX when store is narrower than the bswap

[Nemanja Ivanovic via llvm-commits]

Author: nemanjai
Date: Mon Mar  6 01:32:13 2017
New Revision: 297001

URL: http://llvm.org/viewvc/llvm-project?rev=297001&view=rev
Log:
[PowerPC] Fix failure with STBRX when store is narrower than the bswap

Fixes a crash caused by r296811 by truncating the input of the STBRX node
when the bswap is wider than i32.

Fixes https://bugs.llvm.org/show_bug.cgi?id=32140

Differential Revision: https://reviews.llvm.org/D30615

Added:
    llvm/trunk/test/CodeGen/PowerPC/pr32140.ll
Modified:
    llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp

Modified: llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp?rev=297001&r1=297000&r2=297001&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp Mon Mar  6 01:32:13 2017
@@ -11395,9 +11395,12 @@ SDValue PPCTargetLowering::PerformDAGCom
       // it need to be shifted to the right side before STBRX.
       EVT mVT = cast<StoreSDNode>(N)->getMemoryVT();
       if (Op1VT.bitsGT(mVT)) {
-        int shift = Op1VT.getSizeInBits() - mVT.getSizeInBits();
+        int Shift = Op1VT.getSizeInBits() - mVT.getSizeInBits();
         BSwapOp = DAG.getNode(ISD::SRL, dl, Op1VT, BSwapOp,
-                              DAG.getConstant(shift, dl, MVT::i32));
+                              DAG.getConstant(Shift, dl, MVT::i32));
+        // Need to truncate if this is a bswap of i64 stored as i32/i16.
+        if (Op1VT == MVT::i64)
+          BSwapOp = DAG.getNode(ISD::TRUNCATE, dl, MVT::i32, BSwapOp);
       }
 
       SDValue Ops[] = {

Added: llvm/trunk/test/CodeGen/PowerPC/pr32140.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/pr32140.ll?rev=297001&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/PowerPC/pr32140.ll (added)
+++ llvm/trunk/test/CodeGen/PowerPC/pr32140.ll Mon Mar  6 01:32:13 2017
@@ -0,0 +1,59 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=powerpc64le-linux-gnu -mcpu=pwr8 < %s | FileCheck %s
+; RUN: llc -mtriple=powerpc64-linux-gnu -mcpu=pwr8 < %s | FileCheck %s
+
+ at as = common local_unnamed_addr global i16 0, align 2
+ at bs = common local_unnamed_addr global i16 0, align 2
+ at ai = common local_unnamed_addr global i32 0, align 4
+ at bi = common local_unnamed_addr global i32 0, align 4
+
+define void @bswapStorei64Toi32() {
+; CHECK-LABEL: bswapStorei64Toi32:
+; CHECK:       # BB#0: # %entry
+; CHECK:         lwa 3, 0(3)
+; CHECK-NEXT:    rldicl 3, 3, 32, 32
+; CHECK-NEXT:    stwbrx 3, 0, 4
+; CHECK-NEXT:    blr
+entry:
+  %0 = load i32, i32* @ai, align 4
+  %conv.i = sext i32 %0 to i64
+  %or26.i = tail call i64 @llvm.bswap.i64(i64 %conv.i)
+  %conv = trunc i64 %or26.i to i32
+  store i32 %conv, i32* @bi, align 4
+  ret void
+}
+
+define void @bswapStorei32Toi16() {
+; CHECK-LABEL: bswapStorei32Toi16:
+; CHECK:       # BB#0: # %entry
+; CHECK:         lha 3, 0(3)
+; CHECK-NEXT:    srwi 3, 3, 16
+; CHECK-NEXT:    sthbrx 3, 0, 4
+; CHECK-NEXT:    blr
+entry:
+  %0 = load i16, i16* @as, align 2
+  %conv.i = sext i16 %0 to i32
+  %or26.i = tail call i32 @llvm.bswap.i32(i32 %conv.i)
+  %conv = trunc i32 %or26.i to i16
+  store i16 %conv, i16* @bs, align 2
+  ret void
+}
+
+define void @bswapStorei64Toi16() {
+; CHECK-LABEL: bswapStorei64Toi16:
+; CHECK:       # BB#0: # %entry
+; CHECK:         lha 3, 0(3)
+; CHECK-NEXT:    rldicl 3, 3, 16, 48
+; CHECK-NEXT:    sthbrx 3, 0, 4
+; CHECK-NEXT:    blr
+entry:
+  %0 = load i16, i16* @as, align 2
+  %conv.i = sext i16 %0 to i64
+  %or26.i = tail call i64 @llvm.bswap.i64(i64 %conv.i)
+  %conv = trunc i64 %or26.i to i16
+  store i16 %conv, i16* @bs, align 2
+  ret void
+}
+
+declare i32 @llvm.bswap.i32(i32)
+declare i64 @llvm.bswap.i64(i64)