From 54552bfee66e4253d0652c6499c3ca91740f873d Mon Sep 17 00:00:00 2001
From: Nemanja Ivanovic <nemanja.i.ibm@gmail.com>
Date: Thu, 27 Oct 2016 05:17:58 +0000
Subject: [PATCH] [PowerPC] - No SExt/ZExt needed for count trailing zeros

This patch corresponds to review:
https://reviews.llvm.org/D25896

It just eliminates the redundant ZExt after a count trailing zeros instruction.

llvm-svn: 285267
---
 lib/Target/PowerPC/PPCISelDAGToDAG.cpp        |  6 ++-
 .../PowerPC/no-ext-with-count-zeros.ll        | 54 +++++++++++++++++++
 2 files changed, 58 insertions(+), 2 deletions(-)
 create mode 100644 test/CodeGen/PowerPC/no-ext-with-count-zeros.ll

diff --git a/lib/Target/PowerPC/PPCISelDAGToDAG.cpp b/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
index 1395a4b4099..cf8b858c6a7 100644
--- a/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
+++ b/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
@@ -4041,8 +4041,9 @@ static bool PeepholePPC64ZExtGather(SDValue Op32,
     return true;
   }
 
-  // CNTLZW always produces a 64-bit value in [0,32], and so is zero extended.
-  if (Op32.getMachineOpcode() == PPC::CNTLZW) {
+  // CNT[LT]ZW always produce a 64-bit value in [0,32], and so is zero extended.
+  if (Op32.getMachineOpcode() == PPC::CNTLZW ||
+      Op32.getMachineOpcode() == PPC::CNTTZW) {
     ToPromote.insert(Op32.getNode());
     return true;
   }
@@ -4237,6 +4238,7 @@ void PPCDAGToDAGISel::PeepholePPC64ZExt() {
       case PPC::LHBRX:     NewOpcode = PPC::LHBRX8; break;
       case PPC::LWBRX:     NewOpcode = PPC::LWBRX8; break;
       case PPC::CNTLZW:    NewOpcode = PPC::CNTLZW8; break;
+      case PPC::CNTTZW:    NewOpcode = PPC::CNTTZW8; break;
       case PPC::RLWIMI:    NewOpcode = PPC::RLWIMI8; break;
       case PPC::OR:        NewOpcode = PPC::OR8; break;
       case PPC::SELECT_I4: NewOpcode = PPC::SELECT_I8; break;
diff --git a/test/CodeGen/PowerPC/no-ext-with-count-zeros.ll b/test/CodeGen/PowerPC/no-ext-with-count-zeros.ll
new file mode 100644
index 00000000000..4d58e74f7b8
--- /dev/null
+++ b/test/CodeGen/PowerPC/no-ext-with-count-zeros.ll
@@ -0,0 +1,54 @@
+; Function Attrs: nounwind readnone
+; RUN: llc -verify-machineinstrs -mtriple=powerpc64le-unknown-unknown \
+; RUN:   -mcpu=pwr9 < %s | FileCheck %s
+
+define signext i32 @ctw(i32 signext %a) {
+entry:
+  %0 = tail call i32 @llvm.cttz.i32(i32 %a, i1 false)
+  ret i32 %0
+; CHECK-LABEL: ctw
+; CHECK: cnttzw 3, 3
+; CHECK-NEXT: blr
+}
+
+; Function Attrs: nounwind readnone
+declare i32 @llvm.cttz.i32(i32, i1)
+
+; Function Attrs: nounwind readnone
+define signext i32 @clw(i32 signext %a) {
+entry:
+  %0 = tail call i32 @llvm.ctlz.i32(i32 %a, i1 false)
+  ret i32 %0
+; CHECK-LABEL: clw
+; CHECK: cntlzw 3, 3
+; CHECK-NEXT: blr
+}
+
+; Function Attrs: nounwind readnone
+declare i32 @llvm.ctlz.i32(i32, i1)
+
+; Function Attrs: nounwind readnone
+define i64 @ctd(i64 %a) {
+entry:
+  %0 = tail call i64 @llvm.cttz.i64(i64 %a, i1 false)
+  ret i64 %0
+; CHECK-LABEL: ctd
+; CHECK: cnttzd 3, 3
+; CHECK-NEXT: blr
+}
+
+; Function Attrs: nounwind readnone
+declare i64 @llvm.cttz.i64(i64, i1)
+
+; Function Attrs: nounwind readnone
+define i64 @cld(i64 %a) {
+entry:
+  %0 = tail call i64 @llvm.ctlz.i64(i64 %a, i1 false)
+  ret i64 %0
+; CHECK-LABEL: cld
+; CHECK: cntlzd 3, 3
+; CHECK-NEXT: blr
+}
+
+; Function Attrs: nounwind readnone
+declare i64 @llvm.ctlz.i64(i64, i1)