From d138a7bb3f49cc92d0ab81ac68c63284adce2fbf Mon Sep 17 00:00:00 2001
From: Hal Finkel <hfinkel@anl.gov>
Date: Sun, 4 Jan 2015 12:35:03 +0000
Subject: [PATCH] [PowerPC] Materialize i64 constants using bit inversion

Materializing full 64-bit constants on PPC64 can be expensive, requiring up to
5 instructions depending on the locations of the non-zero bits. Sometimes
materializing the bit-reversed constant, and then flipping the bits, requires
fewer instructions than the direct method. If so, do that instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225132 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/PowerPC/PPCISelDAGToDAG.cpp | 32 ++++++++++++++++++++++++--
 test/CodeGen/PowerPC/constants-i64.ll  | 20 ++++++++++++++++
 2 files changed, 50 insertions(+), 2 deletions(-)
 create mode 100644 test/CodeGen/PowerPC/constants-i64.ll

diff --git a/lib/Target/PowerPC/PPCISelDAGToDAG.cpp b/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
index c6dc50143c0..2e1c1abf5b6 100644
--- a/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
+++ b/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
@@ -544,7 +544,7 @@ SDNode *PPCDAGToDAGISel::SelectBitfieldInsert(SDNode *N) {
 
 // Predict the number of instructions that would be generated by calling
 // SelectInt64(N).
-static unsigned SelectInt64Count(int64_t Imm) {
+static unsigned SelectInt64CountDirect(int64_t Imm) {
   // Assume no remaining bits.
   unsigned Remainder = 0;
   // Assume no shift required.
@@ -602,9 +602,21 @@ static unsigned SelectInt64Count(int64_t Imm) {
   return Result;
 }
 
+static unsigned SelectInt64Count(int64_t Imm) {
+  unsigned DirectCount = SelectInt64CountDirect(Imm);
+
+  // If might be cheaper to materialize the bit-inverted constant, and then
+  // flip the bits (which takes one nor instruction).
+  unsigned NotDirectCount = SelectInt64CountDirect(~(uint64_t) Imm) + 1;
+  if (NotDirectCount < DirectCount)
+    return NotDirectCount;
+
+  return DirectCount;
+}
+
 // Select a 64-bit constant. For cost-modeling purposes, SelectInt64Count
 // (above) needs to be kept in sync with this function.
-static SDNode *SelectInt64(SelectionDAG *CurDAG, SDLoc dl, int64_t Imm) {
+static SDNode *SelectInt64Direct(SelectionDAG *CurDAG, SDLoc dl, int64_t Imm) {
   // Assume no remaining bits.
   unsigned Remainder = 0;
   // Assume no shift required.
@@ -678,6 +690,22 @@ static SDNode *SelectInt64(SelectionDAG *CurDAG, SDLoc dl, int64_t Imm) {
   return Result;
 }
 
+static SDNode *SelectInt64(SelectionDAG *CurDAG, SDLoc dl, int64_t Imm) {
+  unsigned DirectCount = SelectInt64CountDirect(Imm);
+
+  // If might be cheaper to materialize the bit-inverted constant, and then
+  // flip the bits (which takes one nor instruction).
+  unsigned NotDirectCount = SelectInt64CountDirect(~(uint64_t) Imm) + 1;
+  if (NotDirectCount < DirectCount) {
+    SDValue NotDirectVal =
+      SDValue(SelectInt64Direct(CurDAG, dl, ~(uint64_t) Imm), 0);
+    return CurDAG->getMachineNode(PPC::NOR8, dl, MVT::i64, NotDirectVal,
+                                  NotDirectVal);
+  }
+
+  return SelectInt64Direct(CurDAG, dl, Imm);
+}
+
 // Select a 64-bit constant.
 static SDNode *SelectInt64(SelectionDAG *CurDAG, SDNode *N) {
   SDLoc dl(N);
diff --git a/test/CodeGen/PowerPC/constants-i64.ll b/test/CodeGen/PowerPC/constants-i64.ll
new file mode 100644
index 00000000000..6d9ef0888a5
--- /dev/null
+++ b/test/CodeGen/PowerPC/constants-i64.ll
@@ -0,0 +1,20 @@
+; RUN: llc -mcpu=ppc64 < %s | FileCheck %s
+target datalayout = "E-m:e-i64:64-n32:64"
+target triple = "powerpc64-unknown-linux-gnu"
+
+; Function Attrs: nounwind readnone
+define i64 @cn1() #0 {
+entry:
+  ret i64 281474976710655
+
+; CHECK-LABEL: @cn1
+; CHECK: li [[REG1:[0-9]+]], 0
+; CHECK: ori [[REG2:[0-9]+]], [[REG1]], 65535
+; CHECK: sldi [[REG3:[0-9]+]], [[REG2]], 48
+; CHECK: nor 3, [[REG3]], [[REG3]]
+; CHECK: blr
+}
+
+attributes #0 = { nounwind readnone }
+
+