In Thumb1, materialize a move between low registers as a `movs`, if CPSR isn't live.

Summary: Previously, it had always been materialized as a push/pop sequence.

Reviewers: labrinea, jroelofs

Reviewed By: jroelofs

Subscribers: llvm-commits, rengolin

Differential Revision: https://reviews.llvm.org/D30648

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@297134 91177308-0d34-0410-b5e6-96231b3b80d8

6 files changed, 45 insertions, 49 deletions

diff --git a/lib/Target/ARM/Thumb1InstrInfo.cpp b/lib/Target/ARM/Thumb1InstrInfo.cpp
index f1d6fa68d3a..27bff4d75ac 100644
--- a/lib/Target/ARM/Thumb1InstrInfo.cpp
+++ b/lib/Target/ARM/Thumb1InstrInfo.cpp
@@ -54,11 +54,17 @@ void Thumb1InstrInfo::copyPhysReg(MachineBasicBlock &MBB,
         .addReg(SrcReg, getKillRegState(KillSrc))
         .add(predOps(ARMCC::AL));
   else {
-    // FIXME: The performance consequences of this are going to be atrocious.
-    // Some things to try that should be better:
-    //   * 'mov hi, $src; mov $dst, hi', with hi as either r10 or r11
-    //   * 'movs $dst, $src' if cpsr isn't live
-    // See: http://lists.llvm.org/pipermail/llvm-dev/2014-August/075998.html
+    // FIXME: Can also use 'mov hi, $src; mov $dst, hi',
+    // with hi as either r10 or r11.
+
+    const TargetRegisterInfo *RegInfo = st.getRegisterInfo();
+    if (MBB.computeRegisterLiveness(RegInfo, ARM::CPSR, I)
+        == MachineBasicBlock::LQR_Dead) {
+      BuildMI(MBB, I, DL, get(ARM::tMOVSr), DestReg)
+          .addReg(SrcReg, getKillRegState(KillSrc))
+          ->addRegisterDead(ARM::CPSR, RegInfo);
+      return;
+    }
 
     // 'MOV lo, lo' is unpredictable on < v6, so use the stack to do it
     BuildMI(MBB, I, DL, get(ARM::tPUSH))
diff --git a/test/CodeGen/ARM/arm-and-tst-peephole.ll b/test/CodeGen/ARM/arm-and-tst-peephole.ll
index eed21b6b4d4..31691e9468c 100644
--- a/test/CodeGen/ARM/arm-and-tst-peephole.ll
+++ b/test/CodeGen/ARM/arm-and-tst-peephole.ll
@@ -150,16 +150,14 @@ define i32 @test_tst_assessment(i32 %a, i32 %b) {
 ;
 ; THUMB-LABEL: test_tst_assessment:
 ; THUMB:       @ BB#0:
-; THUMB-NEXT:    push {r0}
-; THUMB-NEXT:    pop {r2}
+; THUMB-NEXT:    movs r2, r0
 ; THUMB-NEXT:    movs r0, #1
 ; THUMB-NEXT:    ands r0, r2
 ; THUMB-NEXT:    subs r2, r0, #1
 ; THUMB-NEXT:    lsls r1, r1, #31
 ; THUMB-NEXT:    beq .LBB2_2
 ; THUMB-NEXT:  @ BB#1:
-; THUMB-NEXT:    push {r2}
-; THUMB-NEXT:    pop {r0}
+; THUMB-NEXT:    movs r0, r2
 ; THUMB-NEXT:  .LBB2_2:
 ; THUMB-NEXT:    bx lr
 ;
diff --git a/test/CodeGen/ARM/atomic-cmpxchg.ll b/test/CodeGen/ARM/atomic-cmpxchg.ll
index 8a82af6a69b..e026bae361e 100644
--- a/test/CodeGen/ARM/atomic-cmpxchg.ll
+++ b/test/CodeGen/ARM/atomic-cmpxchg.ll
@@ -24,14 +24,12 @@ entry:
 ; CHECK-THUMB-LABEL: test_cmpxchg_res_i8
 ; CHECK-THUMB: bl __sync_val_compare_and_swap_1
 ; CHECK-THUMB-NOT: mov [[R1:r[0-7]]], r0
-; CHECK-THUMB: push  {r0}
-; CHECK-THUMB: pop {[[R1:r[0-7]]]}
+; CHECK-THUMB: movs [[R1:r[0-7]]], r0
 ; CHECK-THUMB: movs r0, #1
 ; CHECK-THUMB: movs [[R2:r[0-9]+]], #0
 ; CHECK-THUMB: cmp [[R1]], {{r[0-9]+}}
 ; CHECK-THUMB: beq
-; CHECK-THUMB: push  {[[R2]]}
-; CHECK-THUMB: pop {r0}
+; CHECK-THUMB: movs r0, [[R2]]
 
 ; CHECK-ARMV6-LABEL: test_cmpxchg_res_i8:
 ; CHECK-ARMV6-NEXT:  .fnstart
diff --git a/test/CodeGen/ARM/smml.ll b/test/CodeGen/ARM/smml.ll
index aa093192f2b..4788644cf19 100644
--- a/test/CodeGen/ARM/smml.ll
+++ b/test/CodeGen/ARM/smml.ll
@@ -1,20 +1,15 @@
-; RUN: llc -mtriple=arm-eabi %s -o - | FileCheck %s
-; RUN: llc -mtriple=armv6-eabi %s -o - | FileCheck %s -check-prefix=CHECK-V6
-; RUN: llc -mtriple=armv7-eabi %s -o - | FileCheck %s -check-prefix=CHECK-V7
-; RUN: llc -mtriple=thumb-eabi %s -o - | FileCheck %s -check-prefix=CHECK-THUMB
-; RUN: llc -mtriple=thumbv6-eabi %s -o - | FileCheck %s -check-prefix=CHECK-THUMB
-; RUN: llc -mtriple=thumbv6t2-eabi %s -o - | FileCheck %s -check-prefix=CHECK-THUMBV6T2
-; RUN: llc -mtriple=thumbv7-eabi %s -o - | FileCheck %s -check-prefix=CHECK-THUMBV7
+; RUN: llc -mtriple=arm-eabi %s -o - | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-V4
+; RUN: llc -mtriple=armv6-eabi %s -o - | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-V6
+; RUN: llc -mtriple=armv7-eabi %s -o - | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-V6
+; RUN: llc -mtriple=thumb-eabi %s -o - | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-THUMB
+; RUN: llc -mtriple=thumbv6-eabi %s -o - | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-THUMBV6
+; RUN: llc -mtriple=thumbv6t2-eabi %s -o - | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-THUMBV6T2
+; RUN: llc -mtriple=thumbv7-eabi %s -o - | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-THUMBV6T2
 
 define i32 @Test0(i32 %a, i32 %b, i32 %c) nounwind readnone ssp {
 entry:
 ; CHECK-LABEL: Test0
 ; CHECK-NOT: smmls
-; CHECK-V6-NOT: smmls
-; CHECK-V7-NOT: smmls
-; CHECK_THUMB-NOT: smmls
-; CHECK-THUMBV6T2-NOT: smmls
-; CHECK-THUMBV7-NOT: smmls
   %conv4 = zext i32 %a to i64
   %conv1 = sext i32 %b to i64
   %conv2 = sext i32 %c to i64
@@ -27,12 +22,11 @@ entry:
 
 define i32 @Test1(i32 %a, i32 %b, i32 %c) {
 ;CHECK-LABEL: Test1
-;CHECK-NOT: smmls
+;CHECK-V4-NOT: smmls
 ;CHECK-THUMB-NOT: smmls
+;CHECK-THUMBV6-NOT: smmls
 ;CHECK-V6: smmls r0, [[Rn:r[1-2]]], [[Rm:r[1-2]]], r0
-;CHECK-V7: smmls r0, [[Rn:r[1-2]]], [[Rm:r[1-2]]], r0
 ;CHECK-THUMBV6T2: smmls r0, [[Rn:r[1-2]]], [[Rm:r[1-2]]], r0
-;CHECK-THUMBV7: smmls r0, [[Rn:r[1-2]]], [[Rm:r[1-2]]], r0
 entry:
   %conv = sext i32 %b to i64
   %conv1 = sext i32 %c to i64
@@ -47,10 +41,21 @@ entry:
 
 declare void @opaque(i32)
 define void @test_used_flags(i32 %in1, i32 %in2) {
-; CHECK-V7-LABEL: test_used_flags:
-; CHECK-V7: smull [[PROD_LO:r[0-9]+]], [[PROD_HI:r[0-9]+]], r0, r1
-; CHECK-V7: rsbs {{.*}}, [[PROD_LO]], #0
-; CHECK-V7: rscs {{.*}}, [[PROD_HI]], #0
+; CHECK-LABEL: test_used_flags:
+; CHECK-THUMB: cmp  r1, #0
+; CHECK-THUMB: push {r2}
+; CHECK-THUMB: pop  {r3}
+; CHECK-THUMB: ble
+; CHECK-THUMBV6: cmp r1, #0
+; CHECK-THUMBV6: mov r3, r2
+; CHECK-THUMBV6: ble
+; CHECK-V6: smull [[PROD_LO:r[0-9]+]], [[PROD_HI:r[0-9]+]], r0, r1
+; CHECK-V6: rsbs {{.*}}, [[PROD_LO]], #0
+; CHECK-V6: rscs {{.*}}, [[PROD_HI]], #0
+; CHECK-THUMBV6T2: smull [[PROD_LO:r[0-9]+]], [[PROD_HI:r[0-9]+]], r0, r1
+; CHECK-THUMBV6T2: movs	[[ZERO:r[0-9]+]], #0
+; CHECK-THUMBV6T2: rsbs	{{.*}}, [[PROD_LO]], #0
+; CHECK-THUMBV6T2: sbcs.w {{.*}}, [[ZERO]], [[PROD_HI]]
   %in1.64 = sext i32 %in1 to i64
   %in2.64 = sext i32 %in2 to i64
   %mul = mul nsw i64 %in1.64, %in2.64
diff --git a/test/CodeGen/Thumb/copy_thumb.ll b/test/CodeGen/Thumb/copy_thumb.ll
index 528f54bd84e..008c31aba7a 100644
--- a/test/CodeGen/Thumb/copy_thumb.ll
+++ b/test/CodeGen/Thumb/copy_thumb.ll
@@ -16,15 +16,9 @@
 ; RUN: llc -mtriple=thumbv4t-none--eabi < %s | FileCheck %s --check-prefix=CHECK-NOLOLOMOV
 ; RUN: llc -mtriple=thumbv5-none--eabi < %s | FileCheck %s --check-prefix=CHECK-NOLOLOMOV
 ; CHECK-NOLOLOMOV-LABEL: foo
-; CHECK-NOLOLOMOV-NOT:   mov [[TMP:r[0-7]]], [[SRC1:r[01]]]
-; CHECK-NOLOLOMOV:       push  {[[SRC1:r[01]]]}
-; CHECK-NOLOLOMOV-NEXT:  pop {[[TMP:r[0-7]]]}
-; CHECK-NOLOLOMOV-NOT:   mov [[TMP:r[0-7]]], [[SRC1:r[01]]]
-; CHECK-NOLOLOMOV:       push  {[[SRC2:r[01]]]}
-; CHECK-NOLOLOMOV-NEXT:  pop {[[SRC1]]}
-; CHECK-NOLOLOMOV-NOT:   mov [[TMP:r[0-7]]], [[SRC1:r[01]]]
-; CHECK-NOLOLOMOV:       push  {[[TMP]]}
-; CHECK-NOLOLOMOV-NEXT:  pop {[[SRC2]]}
+; CHECK-NOLOLOMOV:       movs [[TMP:r[0-7]]], [[SRC1:r[01]]]
+; CHECK-NOLOLOMOV-NEXT:  movs [[SRC1]], [[SRC2:r[01]]]
+; CHECK-NOLOLOMOV-NEXT:  movs [[SRC2]], [[TMP]]
 ; CHECK-NOLOLOMOV-LABEL: bar
 ; CHECK-NOLOLOMOV-LABEL: fnend
 
diff --git a/test/CodeGen/Thumb/thumb-shrink-wrapping.ll b/test/CodeGen/Thumb/thumb-shrink-wrapping.ll
index f6c137b9e41..c571e351a1e 100644
--- a/test/CodeGen/Thumb/thumb-shrink-wrapping.ll
+++ b/test/CodeGen/Thumb/thumb-shrink-wrapping.ll
@@ -505,14 +505,9 @@ if.end:                                           ; preds = %for.body, %if.else
 ; CHECK-NEXT: str r1, {{\[}}[[TMP_SP]]]
 ; CHECK-NEXT: str r1, {{\[}}[[TMP_SP]], #4]
 ; CHECK-NEXT: str r1, {{\[}}[[TMP_SP]], #8]
-; Thumb has quite a strange way for moving stuff
-; in around. Oh well, match the current sequence.
-; CHECK: push {r1}
-; CHECK-NEXT: pop {r0}
-; CHECK: push {r1}
-; CHECK-NEXT: pop {r2}
-; CHECK: push {r1}
-; CHECK-NEXT: pop {r3}
+; CHECK:      movs r0, r1
+; CHECK-NEXT: movs r2, r1
+; CHECK-NEXT: movs r3, r1
 ; CHECK-NEXT: bl
 ; CHECK-NEXT: lsls r0, r0, #3
 ;