[llvm] 3251593 - [X86] Place new constant node in topological order in X86DAGToDAGISel::matchBitExtract.

[Craig Topper via llvm-commits]

Author: Craig Topper
Date: 2020-09-14T16:28:37-07:00
New Revision: 32515938901685bcbc438d5f5bb03cb8a9f4c637

URL: https://github.com/llvm/llvm-project/commit/32515938901685bcbc438d5f5bb03cb8a9f4c637
DIFF: https://github.com/llvm/llvm-project/commit/32515938901685bcbc438d5f5bb03cb8a9f4c637.diff

LOG: [X86] Place new constant node in topological order in X86DAGToDAGISel::matchBitExtract.

Fixes PR47525

Added: 
    llvm/test/CodeGen/X86/pr47525.ll

Modified: 
    llvm/lib/Target/X86/X86ISelDAGToDAG.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
index 840f132ec666..3b5a29ef31fc 100644
--- a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
+++ b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
@@ -3502,6 +3502,7 @@ bool X86DAGToDAGISel::matchBitExtract(SDNode *Node) {
   // Shift NBits left by 8 bits, thus producing 'control'.
   // This makes the low 8 bits to be zero.
   SDValue C8 = CurDAG->getConstant(8, DL, MVT::i8);
+  insertDAGNode(*CurDAG, SDValue(Node, 0), C8);
   SDValue Control = CurDAG->getNode(ISD::SHL, DL, MVT::i32, NBits, C8);
   insertDAGNode(*CurDAG, SDValue(Node, 0), Control);
 

diff  --git a/llvm/test/CodeGen/X86/pr47525.ll b/llvm/test/CodeGen/X86/pr47525.ll
new file mode 100644
index 000000000000..e0f01f3c5115
--- /dev/null
+++ b/llvm/test/CodeGen/X86/pr47525.ll
@@ -0,0 +1,42 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=bmi | FileCheck %s
+
+ at a = external local_unnamed_addr global i32, align 4
+ at f = external local_unnamed_addr global i32, align 4
+
+define void @g(i32* %x, i32* %y, i32* %z) {
+; CHECK-LABEL: g:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    movl {{.*}}(%rip), %eax
+; CHECK-NEXT:    #APP
+; CHECK-NEXT:    #NO_APP
+; CHECK-NEXT:    xorl %ecx, %ecx
+; CHECK-NEXT:    testl %eax, %eax
+; CHECK-NEXT:    sete %cl
+; CHECK-NEXT:    addl %ecx, %ecx
+; CHECK-NEXT:    orl (%rdi), %ecx
+; CHECK-NEXT:    movl $0, (%rsi)
+; CHECK-NEXT:    #APP
+; CHECK-NEXT:    #NO_APP
+; CHECK-NEXT:    shll $8, %eax
+; CHECK-NEXT:    bextrl %eax, {{.*}}(%rip), %eax
+; CHECK-NEXT:    orl %ecx, %eax
+; CHECK-NEXT:    movl %eax, (%rdx)
+; CHECK-NEXT:    retq
+entry:
+  %0 = load i32, i32* @a, align 4
+  %1 = tail call i32 asm "", "=r,r,~{dirflag},~{fpsr},~{flags}"(i32 %0)
+  %2 = icmp eq i32 %1, 0
+  %shl1 = select i1 %2, i32 2, i32 0
+  %3 = load i32, i32* %x, align 4
+  %or = or i32 %3, %shl1
+  store i32 0, i32* %y, align 4
+  %4 = tail call i32 asm "", "=r,~{dirflag},~{fpsr},~{flags}"()
+  %notmask = shl nsw i32 -1, %4
+  %sub = xor i32 %notmask, -1
+  %5 = load i32, i32* @f, align 4
+  %and4 = and i32 %5, %sub
+  %or6 = or i32 %and4, %or
+  store i32 %or6, i32* %z, align 4
+  ret void
+}