Reapply [FastISel][X86] Emit more efficient instructions for integer constant materia...

Note: This was originally reverted to track down a buildbot error. Reapply
without any modifications.

Original commit message:
This mostly affects the i64 value type, which always resulted in an 15byte
mobavsq instruction to materialize any constant. The custom code checks the
value of the immediate and tries to use a different and smaller mov
instruction when possible.

This fixes <rdar://problem/17420988>.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@216010 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86FastISel.cpp b/lib/Target/X86/X86FastISel.cpp
index 2166b8a057cb244aec6f750bc5e62d7538b6188b..6426d27ae1a152699e08c4e097708ffa02214af8 100644 (file)
--- a/lib/Target/X86/X86FastISel.cpp
+++ b/lib/Target/X86/X86FastISel.cpp
@@ -3105,7 +3105,34 @@ X86FastISel::TargetSelectInstruction(const Instruction *I)  {
 unsigned X86FastISel::X86MaterializeInt(const ConstantInt *CI, MVT VT) {
   if (VT > MVT::i64)
     return 0;
-  return FastEmit_i(VT, VT, ISD::Constant, CI->getZExtValue());
+
+  uint64_t Imm = CI->getZExtValue();
+  unsigned Opc = 0;
+  switch (VT.SimpleTy) {
+  default: llvm_unreachable("Unexpected value type");
+  case MVT::i1:  VT = MVT::i8; // fall-through
+  case MVT::i8:  Opc = X86::MOV8ri;  break;
+  case MVT::i16: Opc = X86::MOV16ri; break;
+  case MVT::i32: Opc = X86::MOV32ri; break;
+  case MVT::i64: {
+    if (isUInt<32>(Imm))
+      Opc = X86::MOV32ri;
+    else if (isInt<32>(Imm))
+      Opc = X86::MOV64ri32;
+    else
+      Opc = X86::MOV64ri;
+    break;
+  }
+  }
+  if (VT == MVT::i64 && Opc == X86::MOV32ri) {
+    unsigned SrcReg = FastEmitInst_i(Opc, &X86::GR32RegClass, Imm);
+    unsigned ResultReg = createResultReg(&X86::GR64RegClass);
+    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
+            TII.get(TargetOpcode::SUBREG_TO_REG), ResultReg)
+      .addImm(0).addReg(SrcReg).addImm(X86::sub_32bit);
+    return ResultReg;
+  }
+  return FastEmitInst_i(Opc, TLI.getRegClassFor(VT), Imm);
 }
 
 unsigned X86FastISel::X86MaterializeFP(const ConstantFP *CFP, MVT VT) {

diff --git a/test/CodeGen/X86/object-size.ll b/test/CodeGen/X86/object-size.ll
index ec35d2981a1633a9cad4a89f246b42671e1cba41..0610f0b6de2eaca8d1fc65ed6a0e0b6294b256d3 100644 (file)
--- a/test/CodeGen/X86/object-size.ll
+++ b/test/CodeGen/X86/object-size.ll
@@ -1,4 +1,4 @@
-; RUN: llc -O0 < %s -march=x86-64 | FileCheck %s -check-prefix=X64
+; RUN: llc -O0 < %s -march=x86-64 | FileCheck %s
 
 ; ModuleID = 'ts.c'
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
@@ -12,8 +12,8 @@ entry:
   %tmp = load i8** @p                             ; <i8*> [#uses=1]
   %0 = call i64 @llvm.objectsize.i64.p0i8(i8* %tmp, i1 0) ; <i64> [#uses=1]
   %cmp = icmp ne i64 %0, -1                       ; <i1> [#uses=1]
-; X64: movabsq $-1, [[RAX:%r..]]
-; X64: cmpq    $-1, [[RAX]]
+; CHECK: movq $-1, [[RAX:%r..]]
+; CHECK: cmpq $-1, [[RAX]]
   br i1 %cmp, label %cond.true, label %cond.false
 
 cond.true:                                        ; preds = %entry

diff --git a/test/CodeGen/X86/xaluo.ll b/test/CodeGen/X86/xaluo.ll
index bc1d0ce77c6f2c5e5fbf8727d642dde4e728b900..6a98037f29d67e48fd9672bf3476e60b0dba5dea 100644 (file)
--- a/test/CodeGen/X86/xaluo.ll
+++ b/test/CodeGen/X86/xaluo.ll
@@ -1,7 +1,5 @@
-; RUN: llc -mtriple=x86_64-darwin-unknown < %s                             | FileCheck %s --check-prefix=DAG
-; RUN: llc -mtriple=x86_64-darwin-unknown -fast-isel -fast-isel-abort < %s | FileCheck %s --check-prefix=FAST
-; RUN: llc -mtriple=x86_64-darwin-unknown < %s                             | FileCheck %s
-; RUN: llc -mtriple=x86_64-darwin-unknown -fast-isel -fast-isel-abort < %s | FileCheck %s
+; RUN: llc -mtriple=x86_64-darwin-unknown                             < %s | FileCheck %s --check-prefix=CHECK --check-prefix=SDAG
+; RUN: llc -mtriple=x86_64-darwin-unknown -fast-isel -fast-isel-abort < %s | FileCheck %s --check-prefix=CHECK --check-prefix=FAST
 
 ;
 ; Get the actual value of the overflow bit.
@@ -9,12 +7,9 @@
 ; SADDO reg, reg
 define zeroext i1 @saddo.i8(i8 signext %v1, i8 signext %v2, i8* %res) {
 entry:
-; DAG-LABEL:    saddo.i8
-; DAG:          addb %sil, %dil
-; DAG-NEXT:     seto %al
-; FAST-LABEL:   saddo.i8
-; FAST:         addb %sil, %dil
-; FAST-NEXT:    seto %al
+; CHECK-LABEL: saddo.i8
+; CHECK:       addb %sil, %dil
+; CHECK-NEXT:  seto %al
   %t = call {i8, i1} @llvm.sadd.with.overflow.i8(i8 %v1, i8 %v2)
   %val = extractvalue {i8, i1} %t, 0
   %obit = extractvalue {i8, i1} %t, 1
@@ -24,12 +19,9 @@ entry:
 
 define zeroext i1 @saddo.i16(i16 %v1, i16 %v2, i16* %res) {
 entry:
-; DAG-LABEL:    saddo.i16
-; DAG:          addw %si, %di
-; DAG-NEXT:     seto %al
-; FAST-LABEL:   saddo.i16
-; FAST:         addw %si, %di
-; FAST-NEXT:    seto %al
+; CHECK-LABEL: saddo.i16
+; CHECK:       addw %si, %di
+; CHECK-NEXT:  seto %al
   %t = call {i16, i1} @llvm.sadd.with.overflow.i16(i16 %v1, i16 %v2)
   %val = extractvalue {i16, i1} %t, 0
   %obit = extractvalue {i16, i1} %t, 1
@@ -39,12 +31,9 @@ entry:
 
 define zeroext i1 @saddo.i32(i32 %v1, i32 %v2, i32* %res) {
 entry:
-; DAG-LABEL:    saddo.i32
-; DAG:          addl %esi, %edi
-; DAG-NEXT:     seto %al
-; FAST-LABEL:   saddo.i32
-; FAST:         addl %esi, %edi
-; FAST-NEXT:    seto %al
+; CHECK-LABEL: saddo.i32
+; CHECK:       addl %esi, %edi
+; CHECK-NEXT:  seto %al
   %t = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -54,12 +43,9 @@ entry:
 
 define zeroext i1 @saddo.i64(i64 %v1, i64 %v2, i64* %res) {
 entry:
-; DAG-LABEL:    saddo.i64
-; DAG:          addq %rsi, %rdi
-; DAG-NEXT:     seto %al
-; FAST-LABEL:   saddo.i64
-; FAST:         addq %rsi, %rdi
-; FAST-NEXT:    seto %al
+; CHECK-LABEL: saddo.i64
+; CHECK:       addq %rsi, %rdi
+; CHECK-NEXT:  seto %al
   %t = call {i64, i1} @llvm.sadd.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -120,13 +106,13 @@ entry:
 ; FIXME: DAG doesn't optimize immediates on the LHS.
 define zeroext i1 @saddo.i64imm1(i64 %v1, i64* %res) {
 entry:
-; DAG-LABEL:    saddo.i64imm1
-; DAG:          mov
-; DAG-NEXT:     addq
-; DAG-NEXT:     seto
-; FAST-LABEL:   saddo.i64imm1
-; FAST:         addq $2, %rdi
-; FAST-NEXT:    seto %al
+; SDAG-LABEL: saddo.i64imm1
+; SDAG:       mov
+; SDAG-NEXT:  addq
+; SDAG-NEXT:  seto
+; FAST-LABEL: saddo.i64imm1
+; FAST:       addq $2, %rdi
+; FAST-NEXT:  seto %al
   %t = call {i64, i1} @llvm.sadd.with.overflow.i64(i64 2, i64 %v1)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -137,12 +123,12 @@ entry:
 ; Check boundary conditions for large immediates.
 define zeroext i1 @saddo.i64imm2(i64 %v1, i64* %res) {
 entry:
-; DAG-LABEL:    saddo.i64imm2
-; DAG:          addq $-2147483648, %rdi
-; DAG-NEXT:     seto %al
-; FAST-LABEL:   saddo.i64imm2
-; FAST:         addq $-2147483648, %rdi
-; FAST-NEXT:    seto %al
+; SDAG-LABEL: saddo.i64imm2
+; SDAG:       addq $-2147483648, %rdi
+; SDAG-NEXT:  seto %al
+; FAST-LABEL: saddo.i64imm2
+; FAST:       addq $-2147483648, %rdi
+; FAST-NEXT:  seto %al
   %t = call {i64, i1} @llvm.sadd.with.overflow.i64(i64 %v1, i64 -2147483648)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -152,14 +138,10 @@ entry:
 
 define zeroext i1 @saddo.i64imm3(i64 %v1, i64* %res) {
 entry:
-; DAG-LABEL:    saddo.i64imm3
-; DAG:          movabsq $-21474836489, %[[REG:[a-z]+]]
-; DAG-NEXT:     addq %rdi, %[[REG]]
-; DAG-NEXT:     seto
-; FAST-LABEL:   saddo.i64imm3
-; FAST:         movabsq $-21474836489, %[[REG:[a-z]+]]
-; FAST-NEXT:    addq %rdi, %[[REG]]
-; FAST-NEXT:    seto
+; CHECK-LABEL: saddo.i64imm3
+; CHECK:       movabsq $-21474836489, %[[REG:[a-z]+]]
+; CHECK-NEXT:  addq %rdi, %[[REG]]
+; CHECK-NEXT:  seto
   %t = call {i64, i1} @llvm.sadd.with.overflow.i64(i64 %v1, i64 -21474836489)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -169,12 +151,9 @@ entry:
 
 define zeroext i1 @saddo.i64imm4(i64 %v1, i64* %res) {
 entry:
-; DAG-LABEL:    saddo.i64imm4
-; DAG:          addq $2147483647, %rdi
-; DAG-NEXT:     seto
-; FAST-LABEL:   saddo.i64imm4
-; FAST:         addq $2147483647, %rdi
-; FAST-NEXT:    seto
+; CHECK-LABEL: saddo.i64imm4
+; CHECK:       addq $2147483647, %rdi
+; CHECK-NEXT:  seto
   %t = call {i64, i1} @llvm.sadd.with.overflow.i64(i64 %v1, i64 2147483647)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -182,17 +161,12 @@ entry:
   ret i1 %obit
 }
 
-; TODO: FastISel shouldn't use movabsq.
 define zeroext i1 @saddo.i64imm5(i64 %v1, i64* %res) {
 entry:
-; DAG-LABEL:    saddo.i64imm5
-; DAG:          movl $2147483648, %ecx
-; DAG:          addq %rdi, %rcx
-; DAG-NEXT:     seto
-; FAST-LABEL:   saddo.i64imm5
-; FAST:         movabsq $2147483648, %[[REG:[a-z]+]]
-; FAST:         addq %rdi, %[[REG]]
-; FAST-NEXT:     seto
+; CHECK-LABEL: saddo.i64imm5
+; CHECK:       movl $2147483648
+; CHECK:       addq %rdi
+; CHECK-NEXT:  seto
   %t = call {i64, i1} @llvm.sadd.with.overflow.i64(i64 %v1, i64 2147483648)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -203,12 +177,9 @@ entry:
 ; UADDO
 define zeroext i1 @uaddo.i32(i32 %v1, i32 %v2, i32* %res) {
 entry:
-; DAG-LABEL:    uaddo.i32
-; DAG:          addl %esi, %edi
-; DAG-NEXT:     setb %al
-; FAST-LABEL:   uaddo.i32
-; FAST:         addl %esi, %edi
-; FAST-NEXT:    setb %al
+; CHECK-LABEL: uaddo.i32
+; CHECK:       addl %esi, %edi
+; CHECK-NEXT:  setb %al
   %t = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -218,12 +189,9 @@ entry:
 
 define zeroext i1 @uaddo.i64(i64 %v1, i64 %v2, i64* %res) {
 entry:
-; DAG-LABEL:    uaddo.i64
-; DAG:          addq %rsi, %rdi
-; DAG-NEXT:     setb %al
-; FAST-LABEL:   uaddo.i64
-; FAST:         addq %rsi, %rdi
-; FAST-NEXT:    setb %al
+; CHECK-LABEL: uaddo.i64
+; CHECK:       addq %rsi, %rdi
+; CHECK-NEXT:  setb %al
   %t = call {i64, i1} @llvm.uadd.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -279,12 +247,9 @@ entry:
 ; SSUBO
 define zeroext i1 @ssubo.i32(i32 %v1, i32 %v2, i32* %res) {
 entry:
-; DAG-LABEL:    ssubo.i32
-; DAG:          subl %esi, %edi
-; DAG-NEXT:     seto %al
-; FAST-LABEL:   ssubo.i32
-; FAST:         subl %esi, %edi
-; FAST-NEXT:    seto %al
+; CHECK-LABEL: ssubo.i32
+; CHECK:       subl %esi, %edi
+; CHECK-NEXT:  seto %al
   %t = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -294,12 +259,9 @@ entry:
 
 define zeroext i1 @ssubo.i64(i64 %v1, i64 %v2, i64* %res) {
 entry:
-; DAG-LABEL:    ssubo.i64
-; DAG:          subq %rsi, %rdi
-; DAG-NEXT:     seto %al
-; FAST-LABEL:   ssubo.i64
-; FAST:         subq %rsi, %rdi
-; FAST-NEXT:    seto %al
+; CHECK-LABEL: ssubo.i64
+; CHECK:       subq %rsi, %rdi
+; CHECK-NEXT:  seto %al
   %t = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -310,12 +272,9 @@ entry:
 ; USUBO
 define zeroext i1 @usubo.i32(i32 %v1, i32 %v2, i32* %res) {
 entry:
-; DAG-LABEL:    usubo.i32
-; DAG:          subl %esi, %edi
-; DAG-NEXT:     setb %al
-; FAST-LABEL:   usubo.i32
-; FAST:         subl %esi, %edi
-; FAST-NEXT:    setb %al
+; CHECK-LABEL: usubo.i32
+; CHECK:       subl %esi, %edi
+; CHECK-NEXT:  setb %al
   %t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -325,12 +284,9 @@ entry:
 
 define zeroext i1 @usubo.i64(i64 %v1, i64 %v2, i64* %res) {
 entry:
-; DAG-LABEL:    usubo.i64
-; DAG:          subq %rsi, %rdi
-; DAG-NEXT:     setb %al
-; FAST-LABEL:   usubo.i64
-; FAST:         subq %rsi, %rdi
-; FAST-NEXT:    setb %al
+; CHECK-LABEL: usubo.i64
+; CHECK:       subq %rsi, %rdi
+; CHECK-NEXT:  setb %al
   %t = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -354,12 +310,9 @@ entry:
 
 define zeroext i1 @smulo.i16(i16 %v1, i16 %v2, i16* %res) {
 entry:
-; DAG-LABEL:    smulo.i16
-; DAG:          imulw %si, %di
-; DAG-NEXT:     seto %al
-; FAST-LABEL:   smulo.i16
-; FAST:         imulw %si, %di
-; FAST-NEXT:    seto %al
+; CHECK-LABEL: smulo.i16
+; CHECK:       imulw %si, %di
+; CHECK-NEXT:  seto %al
   %t = call {i16, i1} @llvm.smul.with.overflow.i16(i16 %v1, i16 %v2)
   %val = extractvalue {i16, i1} %t, 0
   %obit = extractvalue {i16, i1} %t, 1
@@ -369,12 +322,9 @@ entry:
 
 define zeroext i1 @smulo.i32(i32 %v1, i32 %v2, i32* %res) {
 entry:
-; DAG-LABEL:    smulo.i32
-; DAG:          imull %esi, %edi
-; DAG-NEXT:     seto %al
-; FAST-LABEL:   smulo.i32
-; FAST:         imull %esi, %edi
-; FAST-NEXT:    seto %al
+; CHECK-LABEL: smulo.i32
+; CHECK:       imull %esi, %edi
+; CHECK-NEXT:  seto %al
   %t = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -384,12 +334,9 @@ entry:
 
 define zeroext i1 @smulo.i64(i64 %v1, i64 %v2, i64* %res) {
 entry:
-; DAG-LABEL:    smulo.i64
-; DAG:          imulq %rsi, %rdi
-; DAG-NEXT:     seto %al
-; FAST-LABEL:   smulo.i64
-; FAST:         imulq %rsi, %rdi
-; FAST-NEXT:    seto %al
+; CHECK-LABEL: smulo.i64
+; CHECK:       imulq %rsi, %rdi
+; CHECK-NEXT:  seto %al
   %t = call {i64, i1} @llvm.smul.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -413,12 +360,9 @@ entry:
 
 define zeroext i1 @umulo.i16(i16 %v1, i16 %v2, i16* %res) {
 entry:
-; DAG-LABEL:    umulo.i16
-; DAG:          mulw %si
-; DAG-NEXT:     seto
-; FAST-LABEL:   umulo.i16
-; FAST:         mulw %si
-; FAST-NEXT:    seto
+; CHECK-LABEL: umulo.i16
+; CHECK:       mulw %si
+; CHECK-NEXT:  seto
   %t = call {i16, i1} @llvm.umul.with.overflow.i16(i16 %v1, i16 %v2)
   %val = extractvalue {i16, i1} %t, 0
   %obit = extractvalue {i16, i1} %t, 1
@@ -428,12 +372,9 @@ entry:
 
 define zeroext i1 @umulo.i32(i32 %v1, i32 %v2, i32* %res) {
 entry:
-; DAG-LABEL:    umulo.i32
-; DAG:          mull %esi
-; DAG-NEXT:     seto
-; FAST-LABEL:   umulo.i32
-; FAST:         mull %esi
-; FAST-NEXT:    seto
+; CHECK-LABEL: umulo.i32
+; CHECK:       mull %esi
+; CHECK-NEXT:  seto
   %t = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -443,12 +384,9 @@ entry:
 
 define zeroext i1 @umulo.i64(i64 %v1, i64 %v2, i64* %res) {
 entry:
-; DAG-LABEL:    umulo.i64
-; DAG:          mulq %rsi
-; DAG-NEXT:     seto
-; FAST-LABEL:   umulo.i64
-; FAST:         mulq %rsi
-; FAST-NEXT:    seto
+; CHECK-LABEL: umulo.i64
+; CHECK:       mulq %rsi
+; CHECK-NEXT:  seto
   %t = call {i64, i1} @llvm.umul.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -461,9 +399,9 @@ entry:
 ;
 define i32 @saddo.select.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    saddo.select.i32
-; CHECK:          addl   %esi, %eax
-; CHECK-NEXT:     cmovol %edi, %esi
+; CHECK-LABEL: saddo.select.i32
+; CHECK:       addl   %esi, %eax
+; CHECK-NEXT:  cmovol %edi, %esi
   %t = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %v1, i32 %v2)
   %obit = extractvalue {i32, i1} %t, 1
   %ret = select i1 %obit, i32 %v1, i32 %v2
@@ -472,9 +410,9 @@ entry:
 
 define i64 @saddo.select.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    saddo.select.i64
-; CHECK:          addq   %rsi, %rax
-; CHECK-NEXT:     cmovoq %rdi, %rsi
+; CHECK-LABEL: saddo.select.i64
+; CHECK:       addq   %rsi, %rax
+; CHECK-NEXT:  cmovoq %rdi, %rsi
   %t = call {i64, i1} @llvm.sadd.with.overflow.i64(i64 %v1, i64 %v2)
   %obit = extractvalue {i64, i1} %t, 1
   %ret = select i1 %obit, i64 %v1, i64 %v2
@@ -483,9 +421,9 @@ entry:
 
 define i32 @uaddo.select.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    uaddo.select.i32
-; CHECK:          addl   %esi, %eax
-; CHECK-NEXT:     cmovbl %edi, %esi
+; CHECK-LABEL: uaddo.select.i32
+; CHECK:       addl   %esi, %eax
+; CHECK-NEXT:  cmovbl %edi, %esi
   %t = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %v1, i32 %v2)
   %obit = extractvalue {i32, i1} %t, 1
   %ret = select i1 %obit, i32 %v1, i32 %v2
@@ -494,9 +432,9 @@ entry:
 
 define i64 @uaddo.select.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    uaddo.select.i64
-; CHECK:          addq   %rsi, %rax
-; CHECK-NEXT:     cmovbq %rdi, %rsi
+; CHECK-LABEL: uaddo.select.i64
+; CHECK:       addq   %rsi, %rax
+; CHECK-NEXT:  cmovbq %rdi, %rsi
   %t = call {i64, i1} @llvm.uadd.with.overflow.i64(i64 %v1, i64 %v2)
   %obit = extractvalue {i64, i1} %t, 1
   %ret = select i1 %obit, i64 %v1, i64 %v2
@@ -505,9 +443,9 @@ entry:
 
 define i32 @ssubo.select.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    ssubo.select.i32
-; CHECK:          cmpl   %esi, %edi
-; CHECK-NEXT:     cmovol %edi, %esi
+; CHECK-LABEL: ssubo.select.i32
+; CHECK:       cmpl   %esi, %edi
+; CHECK-NEXT:  cmovol %edi, %esi
   %t = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %v1, i32 %v2)
   %obit = extractvalue {i32, i1} %t, 1
   %ret = select i1 %obit, i32 %v1, i32 %v2
@@ -516,9 +454,9 @@ entry:
 
 define i64 @ssubo.select.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    ssubo.select.i64
-; CHECK:          cmpq   %rsi, %rdi
-; CHECK-NEXT:     cmovoq %rdi, %rsi
+; CHECK-LABEL: ssubo.select.i64
+; CHECK:       cmpq   %rsi, %rdi
+; CHECK-NEXT:  cmovoq %rdi, %rsi
   %t = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 %v1, i64 %v2)
   %obit = extractvalue {i64, i1} %t, 1
   %ret = select i1 %obit, i64 %v1, i64 %v2
@@ -527,9 +465,9 @@ entry:
 
 define i32 @usubo.select.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    usubo.select.i32
-; CHECK:          cmpl   %esi, %edi
-; CHECK-NEXT:     cmovbl %edi, %esi
+; CHECK-LABEL: usubo.select.i32
+; CHECK:       cmpl   %esi, %edi
+; CHECK-NEXT:  cmovbl %edi, %esi
   %t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %v1, i32 %v2)
   %obit = extractvalue {i32, i1} %t, 1
   %ret = select i1 %obit, i32 %v1, i32 %v2
@@ -538,9 +476,9 @@ entry:
 
 define i64 @usubo.select.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    usubo.select.i64
-; CHECK:          cmpq   %rsi, %rdi
-; CHECK-NEXT:     cmovbq %rdi, %rsi
+; CHECK-LABEL: usubo.select.i64
+; CHECK:       cmpq   %rsi, %rdi
+; CHECK-NEXT:  cmovbq %rdi, %rsi
   %t = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %v1, i64 %v2)
   %obit = extractvalue {i64, i1} %t, 1
   %ret = select i1 %obit, i64 %v1, i64 %v2
@@ -549,9 +487,9 @@ entry:
 
 define i32 @smulo.select.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    smulo.select.i32
-; CHECK:          imull  %esi, %eax
-; CHECK-NEXT:     cmovol %edi, %esi
+; CHECK-LABEL: smulo.select.i32
+; CHECK:       imull  %esi, %eax
+; CHECK-NEXT:  cmovol %edi, %esi
   %t = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %v1, i32 %v2)
   %obit = extractvalue {i32, i1} %t, 1
   %ret = select i1 %obit, i32 %v1, i32 %v2
@@ -560,9 +498,9 @@ entry:
 
 define i64 @smulo.select.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    smulo.select.i64
-; CHECK:          imulq  %rsi, %rax
-; CHECK-NEXT:     cmovoq %rdi, %rsi
+; CHECK-LABEL: smulo.select.i64
+; CHECK:       imulq  %rsi, %rax
+; CHECK-NEXT:  cmovoq %rdi, %rsi
   %t = call {i64, i1} @llvm.smul.with.overflow.i64(i64 %v1, i64 %v2)
   %obit = extractvalue {i64, i1} %t, 1
   %ret = select i1 %obit, i64 %v1, i64 %v2
@@ -571,9 +509,9 @@ entry:
 
 define i32 @umulo.select.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    umulo.select.i32
-; CHECK:          mull   %esi
-; CHECK-NEXT:     cmovol %edi, %esi
+; CHECK-LABEL: umulo.select.i32
+; CHECK:       mull   %esi
+; CHECK-NEXT:  cmovol %edi, %esi
   %t = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %v1, i32 %v2)
   %obit = extractvalue {i32, i1} %t, 1
   %ret = select i1 %obit, i32 %v1, i32 %v2
@@ -582,9 +520,9 @@ entry:
 
 define i64 @umulo.select.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    umulo.select.i64
-; CHECK:          mulq   %rsi
-; CHECK-NEXT:     cmovoq %rdi, %rsi
+; CHECK-LABEL: umulo.select.i64
+; CHECK:       mulq   %rsi
+; CHECK-NEXT:  cmovoq %rdi, %rsi
   %t = call {i64, i1} @llvm.umul.with.overflow.i64(i64 %v1, i64 %v2)
   %obit = extractvalue {i64, i1} %t, 1
   %ret = select i1 %obit, i64 %v1, i64 %v2
@@ -597,9 +535,9 @@ entry:
 ;
 define zeroext i1 @saddo.br.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    saddo.br.i32
-; CHECK:          addl   %esi, %edi
-; CHECK-NEXT:     jo
+; CHECK-LABEL: saddo.br.i32
+; CHECK:       addl   %esi, %edi
+; CHECK-NEXT:  jo
   %t = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -614,9 +552,9 @@ continue:
 
 define zeroext i1 @saddo.br.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    saddo.br.i64
-; CHECK:          addq   %rsi, %rdi
-; CHECK-NEXT:     jo
+; CHECK-LABEL: saddo.br.i64
+; CHECK:       addq   %rsi, %rdi
+; CHECK-NEXT:  jo
   %t = call {i64, i1} @llvm.sadd.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -631,9 +569,9 @@ continue:
 
 define zeroext i1 @uaddo.br.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    uaddo.br.i32
-; CHECK:          addl   %esi, %edi
-; CHECK-NEXT:     jb
+; CHECK-LABEL: uaddo.br.i32
+; CHECK:       addl   %esi, %edi
+; CHECK-NEXT:  jb
   %t = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -648,9 +586,9 @@ continue:
 
 define zeroext i1 @uaddo.br.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    uaddo.br.i64
-; CHECK:          addq   %rsi, %rdi
-; CHECK-NEXT:     jb
+; CHECK-LABEL: uaddo.br.i64
+; CHECK:       addq   %rsi, %rdi
+; CHECK-NEXT:  jb
   %t = call {i64, i1} @llvm.uadd.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -665,9 +603,9 @@ continue:
 
 define zeroext i1 @ssubo.br.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    ssubo.br.i32
-; CHECK:          cmpl   %esi, %edi
-; CHECK-NEXT:     jo
+; CHECK-LABEL: ssubo.br.i32
+; CHECK:       cmpl   %esi, %edi
+; CHECK-NEXT:  jo
   %t = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -682,9 +620,9 @@ continue:
 
 define zeroext i1 @ssubo.br.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    ssubo.br.i64
-; CHECK:          cmpq   %rsi, %rdi
-; CHECK-NEXT:     jo
+; CHECK-LABEL: ssubo.br.i64
+; CHECK:       cmpq   %rsi, %rdi
+; CHECK-NEXT:  jo
   %t = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -699,9 +637,9 @@ continue:
 
 define zeroext i1 @usubo.br.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    usubo.br.i32
-; CHECK:          cmpl   %esi, %edi
-; CHECK-NEXT:     jb
+; CHECK-LABEL: usubo.br.i32
+; CHECK:       cmpl   %esi, %edi
+; CHECK-NEXT:  jb
   %t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -716,9 +654,9 @@ continue:
 
 define zeroext i1 @usubo.br.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    usubo.br.i64
-; CHECK:          cmpq   %rsi, %rdi
-; CHECK-NEXT:     jb
+; CHECK-LABEL: usubo.br.i64
+; CHECK:       cmpq   %rsi, %rdi
+; CHECK-NEXT:  jb
   %t = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -733,9 +671,9 @@ continue:
 
 define zeroext i1 @smulo.br.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    smulo.br.i32
-; CHECK:          imull  %esi, %edi
-; CHECK-NEXT:     jo
+; CHECK-LABEL: smulo.br.i32
+; CHECK:       imull  %esi, %edi
+; CHECK-NEXT:  jo
   %t = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -750,9 +688,9 @@ continue:
 
 define zeroext i1 @smulo.br.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    smulo.br.i64
-; CHECK:          imulq  %rsi, %rdi
-; CHECK-NEXT:     jo
+; CHECK-LABEL: smulo.br.i64
+; CHECK:       imulq  %rsi, %rdi
+; CHECK-NEXT:  jo
   %t = call {i64, i1} @llvm.smul.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1
@@ -767,9 +705,9 @@ continue:
 
 define zeroext i1 @umulo.br.i32(i32 %v1, i32 %v2) {
 entry:
-; CHECK-LABEL:    umulo.br.i32
-; CHECK:          mull  %esi
-; CHECK-NEXT:     jo
+; CHECK-LABEL: umulo.br.i32
+; CHECK:       mull  %esi
+; CHECK-NEXT:  jo
   %t = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %v1, i32 %v2)
   %val = extractvalue {i32, i1} %t, 0
   %obit = extractvalue {i32, i1} %t, 1
@@ -784,9 +722,9 @@ continue:
 
 define zeroext i1 @umulo.br.i64(i64 %v1, i64 %v2) {
 entry:
-; CHECK-LABEL:    umulo.br.i64
-; CHECK:          mulq  %rsi
-; CHECK-NEXT:     jo
+; CHECK-LABEL: umulo.br.i64
+; CHECK:       mulq  %rsi
+; CHECK-NEXT:  jo
   %t = call {i64, i1} @llvm.umul.with.overflow.i64(i64 %v1, i64 %v2)
   %val = extractvalue {i64, i1} %t, 0
   %obit = extractvalue {i64, i1} %t, 1