[ARM][FastISel] Use TST #1 instead of CMP #0 for select.

Since r234249, i1 are sext instead of zext; because of that, doing
"CMP rN, #0; IT EQ/NE" isn't correct anymore.

"TST #1" is the conservatively correct alternative - the tradeoff being
that it doesn't have a 16-bit encoding -, so use that instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236569 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/ARM/ARMFastISel.cpp b/lib/Target/ARM/ARMFastISel.cpp
index 9c8d228c1ee714b3df9fbb071db61cf52cc0fc31..dfd8dc5cc2cc51d7af426244a53688b482ccb0d9 100644 (file)
--- a/lib/Target/ARM/ARMFastISel.cpp
+++ b/lib/Target/ARM/ARMFastISel.cpp
@@ -1657,12 +1657,12 @@ bool ARMFastISel::SelectSelect(const Instruction *I) {
     if (Op2Reg == 0) return false;
   }
 
-  unsigned CmpOpc = isThumb2 ? ARM::t2CMPri : ARM::CMPri;
-  CondReg = constrainOperandRegClass(TII.get(CmpOpc), CondReg, 0);
+  unsigned TstOpc = isThumb2 ? ARM::t2TSTri : ARM::TSTri;
+  CondReg = constrainOperandRegClass(TII.get(TstOpc), CondReg, 0);
   AddOptionalDefs(
-      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(CmpOpc))
+      BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(TstOpc))
           .addReg(CondReg)
-          .addImm(0));
+          .addImm(1));
 
   unsigned MovCCOpc;
   const TargetRegisterClass *RC;

diff --git a/test/CodeGen/ARM/fast-isel-select.ll b/test/CodeGen/ARM/fast-isel-select.ll
index 009e7868be5b70b3125dad00d83bd8643be749ab..4eef1d6bb988dfb4511a7e2533213bd7e33fe30d 100644 (file)
--- a/test/CodeGen/ARM/fast-isel-select.ll
+++ b/test/CodeGen/ARM/fast-isel-select.ll
@@ -7,12 +7,12 @@ define i32 @t1(i1 %c) nounwind readnone {
 entry:
 ; ARM: t1
 ; ARM: movw r{{[1-9]}}, #10
-; ARM: cmp r0, #0
+; ARM: tst r0, #1
 ; ARM: moveq r{{[1-9]}}, #20
 ; ARM: mov r0, r{{[1-9]}}
 ; THUMB: t1
 ; THUMB: movs r{{[1-9]}}, #10
-; THUMB: cmp r0, #0
+; THUMB: tst.w r0, #1
 ; THUMB: it eq
 ; THUMB: moveq r{{[1-9]}}, #20
 ; THUMB: mov r0, r{{[1-9]}}
@@ -23,11 +23,11 @@ entry:
 define i32 @t2(i1 %c, i32 %a) nounwind readnone {
 entry:
 ; ARM: t2
-; ARM: cmp r0, #0
+; ARM: tst r0, #1
 ; ARM: moveq r{{[1-9]}}, #20
 ; ARM: mov r0, r{{[1-9]}}
 ; THUMB: t2
-; THUMB: cmp r0, #0
+; THUMB: tst.w r0, #1
 ; THUMB: it eq
 ; THUMB: moveq r{{[1-9]}}, #20
 ; THUMB: mov r0, r{{[1-9]}}
@@ -38,11 +38,11 @@ entry:
 define i32 @t3(i1 %c, i32 %a, i32 %b) nounwind readnone {
 entry:
 ; ARM: t3
-; ARM: cmp r0, #0
+; ARM: tst r0, #1
 ; ARM: movne r2, r1
 ; ARM: add r0, r2, r1
 ; THUMB: t3
-; THUMB: cmp r0, #0
+; THUMB: tst.w r0, #1
 ; THUMB: it ne
 ; THUMB: movne r2, r1
 ; THUMB: add.w r0, r2, r1
@@ -55,12 +55,12 @@ define i32 @t4(i1 %c) nounwind readnone {
 entry:
 ; ARM: t4
 ; ARM: mvn r{{[1-9]}}, #9
-; ARM: cmp r0, #0
+; ARM: tst r0, #1
 ; ARM: mvneq r{{[1-9]}}, #0
 ; ARM: mov r0, r{{[1-9]}}
 ; THUMB-LABEL: t4
 ; THUMB: mvn [[REG:r[1-9]+]], #9
-; THUMB: cmp r0, #0
+; THUMB: tst.w r0, #1
 ; THUMB: it eq
 ; THUMB: mvneq [[REG]], #0
 ; THUMB: mov r0, [[REG]]
@@ -71,11 +71,11 @@ entry:
 define i32 @t5(i1 %c, i32 %a) nounwind readnone {
 entry:
 ; ARM: t5
-; ARM: cmp r0, #0
+; ARM: tst r0, #1
 ; ARM: mvneq r{{[1-9]}}, #1
 ; ARM: mov r0, r{{[1-9]}}
 ; THUMB: t5
-; THUMB: cmp r0, #0
+; THUMB: tst.w r0, #1
 ; THUMB: it eq
 ; THUMB: mvneq r{{[1-9]}}, #1
 ; THUMB: mov r0, r{{[1-9]}}
@@ -87,11 +87,11 @@ entry:
 define i32 @t6(i1 %c, i32 %a) nounwind readnone {
 entry:
 ; ARM: t6
-; ARM: cmp r0, #0
+; ARM: tst r0, #1
 ; ARM: mvneq r{{[1-9]}}, #978944
 ; ARM: mov r0, r{{[1-9]}}
 ; THUMB: t6
-; THUMB: cmp r0, #0
+; THUMB: tst.w r0, #1
 ; THUMB: it eq
 ; THUMB: mvneq r{{[1-9]}}, #978944
 ; THUMB: mov r0, r{{[1-9]}}