From ca62919afb90788dd251a7f45c88e2bda3ef549f Mon Sep 17 00:00:00 2001
From: Tim Northover <tnorthover@apple.com>
Date: Mon, 6 Mar 2017 19:04:17 +0000
Subject: [PATCH] GlobalISel: don't emit degenerate G_INSERT instructions.

Before, we were producing G_INSERT instructions that were actually closer to a
cast or even a COPY when both input and output sizes are the same. This doesn't
really make sense and means that everything interpreting a G_INSERT also has to
handle all these kinds of casts.

So now we detect these degenerate cases and emit real casts instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@297051 91177308-0d34-0410-b5e6-96231b3b80d8
---
 .../CodeGen/GlobalISel/MachineIRBuilder.h     |  3 +++
 lib/CodeGen/GlobalISel/MachineIRBuilder.cpp   | 25 +++++++++++++++++++
 .../AArch64/GlobalISel/arm64-irtranslator.ll  | 20 +++++++++++++++
 3 files changed, 48 insertions(+)

diff --git a/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h b/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h
index 4eafe88831a..e3880f0834b 100644
--- a/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h
+++ b/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h
@@ -338,6 +338,9 @@ public:
   /// \return The newly created instruction.
   MachineInstrBuilder buildZExtOrTrunc(unsigned Res, unsigned Op);
 
+  /// Build and insert an appropriate cast between two registers of equal size.
+  MachineInstrBuilder buildCast(unsigned Dst, unsigned Src);
+
   /// Build and insert G_BR \p Dest
   ///
   /// G_BR is an unconditional branch to \p Dest.
diff --git a/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp b/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
index 08f29f4621c..1b7c7a6e52b 100644
--- a/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
+++ b/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
@@ -362,6 +362,26 @@ MachineInstrBuilder MachineIRBuilder::buildZExtOrTrunc(unsigned Res,
   return buildInstr(Opcode).addDef(Res).addUse(Op);
 }
 
+
+MachineInstrBuilder MachineIRBuilder::buildCast(unsigned Dst, unsigned Src) {
+  LLT SrcTy = MRI->getType(Src);
+  LLT DstTy = MRI->getType(Dst);
+  if (SrcTy == DstTy)
+    return buildCopy(Dst, Src);
+
+  unsigned Opcode;
+  if (SrcTy.isPointer() && DstTy.isScalar())
+    Opcode = TargetOpcode::G_PTRTOINT;
+  else if (DstTy.isPointer() && SrcTy.isScalar())
+    Opcode = TargetOpcode::G_INTTOPTR;
+  else {
+    assert(!SrcTy.isPointer() && !DstTy.isPointer() && "n G_ADDRCAST yet");
+    Opcode = TargetOpcode::G_BITCAST;
+  }
+
+  return buildInstr(Opcode).addDef(Dst).addUse(Src);
+}
+
 MachineInstrBuilder MachineIRBuilder::buildExtract(ArrayRef<unsigned> Results,
                                                    ArrayRef<uint64_t> Indices,
                                                    unsigned Src) {
@@ -462,6 +482,11 @@ MachineInstrBuilder MachineIRBuilder::buildUnmerge(ArrayRef<unsigned> Res,
 
 MachineInstrBuilder MachineIRBuilder::buildInsert(unsigned Res, unsigned Src,
                                                   unsigned Op, unsigned Index) {
+  if (MRI->getType(Res).getSizeInBits() == MRI->getType(Op).getSizeInBits()) {
+    assert(Index == 0 && "insertion past the end of a register");
+    return buildCast(Res, Op);
+  }
+
   return buildInstr(TargetOpcode::G_INSERT)
       .addDef(Res)
       .addUse(Src)
diff --git a/test/CodeGen/AArch64/GlobalISel/arm64-irtranslator.ll b/test/CodeGen/AArch64/GlobalISel/arm64-irtranslator.ll
index a05c4a8fe41..65a1422c360 100644
--- a/test/CodeGen/AArch64/GlobalISel/arm64-irtranslator.ll
+++ b/test/CodeGen/AArch64/GlobalISel/arm64-irtranslator.ll
@@ -890,6 +890,26 @@ define void @test_insertvalue(%struct.nested* %addr, i32 %val) {
   ret void
 }
 
+define [1 x i64] @test_trivial_insert([1 x i64] %s, i64 %val) {
+; CHECK-LABEL: name: test_trivial_insert
+; CHECK: [[STRUCT:%[0-9]+]](s64) = COPY %x0
+; CHECK: [[VAL:%[0-9]+]](s64) = COPY %x1
+; CHECK: [[RES:%[0-9]+]](s64) = COPY [[VAL]](s64)
+; CHECK: %x0 = COPY [[RES]]
+  %res = insertvalue [1 x i64] %s, i64 %val, 0
+  ret [1 x i64] %res
+}
+
+define [1 x i8*] @test_trivial_insert_ptr([1 x i8*] %s, i8* %val) {
+; CHECK-LABEL: name: test_trivial_insert_ptr
+; CHECK: [[STRUCT:%[0-9]+]](s64) = COPY %x0
+; CHECK: [[VAL:%[0-9]+]](p0) = COPY %x1
+; CHECK: [[RES:%[0-9]+]](s64) = G_PTRTOINT [[VAL]](p0)
+; CHECK: %x0 = COPY [[RES]]
+  %res = insertvalue [1 x i8*] %s, i8* %val, 0
+  ret [1 x i8*] %res
+}
+
 ; CHECK-LABEL: name: test_insertvalue_agg
 ; CHECK: [[SMALLSTRUCT:%[0-9]+]](s64) = G_LOAD
 ; CHECK: [[STRUCT:%[0-9]+]](s128) = G_LOAD