From 88e57f376eabc375cd9f374369cf9e8e4dcb90af Mon Sep 17 00:00:00 2001
From: Nemanja Ivanovic <nemanja.i.ibm@gmail.com>
Date: Mon, 6 Mar 2017 07:32:13 +0000
Subject: [PATCH 2/3] [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

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@297001 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/PowerPC/PPCISelLowering.cpp |  7 ++--
 test/CodeGen/PowerPC/pr32140.ll        | 59 ++++++++++++++++++++++++++++++++++
 2 files changed, 64 insertions(+), 2 deletions(-)
 create mode 100644 test/CodeGen/PowerPC/pr32140.ll

diff --git a/lib/Target/PowerPC/PPCISelLowering.cpp b/lib/Target/PowerPC/PPCISelLowering.cpp
index c8eb6f1..521bb32 100644
--- a/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -11230,9 +11230,12 @@ SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N,
       // 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[] = {
diff --git a/test/CodeGen/PowerPC/pr32140.ll b/test/CodeGen/PowerPC/pr32140.ll
new file mode 100644
index 0000000..827a904
--- /dev/null
+++ b/test/CodeGen/PowerPC/pr32140.ll
@@ -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
+
+@as = common local_unnamed_addr global i16 0, align 2
+@bs = common local_unnamed_addr global i16 0, align 2
+@ai = common local_unnamed_addr global i32 0, align 4
+@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)
-- 
1.8.3.1