[llvm] e911361 - [ARM] Custom Lowering for SADDO_CARRY and SSUBO_CARRY (#154419)
via llvm-commits
llvm-commits at lists.llvm.org
Wed Apr 8 10:20:59 PDT 2026
Author: SiliconA-Z
Date: 2026-04-08T10:20:54-07:00
New Revision: e9113618cba636b194ee08e8d9eb8cb0bafda092
URL: https://github.com/llvm/llvm-project/commit/e9113618cba636b194ee08e8d9eb8cb0bafda092
DIFF: https://github.com/llvm/llvm-project/commit/e9113618cba636b194ee08e8d9eb8cb0bafda092.diff
LOG: [ARM] Custom Lowering for SADDO_CARRY and SSUBO_CARRY (#154419)
To do this, I did refactoring to mirror what goes on with AArch64,
including having the carryFlagToValue do the inversion.
While the patterns are not the best, with pattern matching, I hope to at
make it as good as AArch64 on Thumb2 where we have CSEL.
Added:
Modified:
llvm/lib/Target/ARM/ARMISelLowering.cpp
llvm/test/CodeGen/ARM/sadd_sat.ll
llvm/test/CodeGen/ARM/sadd_sat_plus.ll
llvm/test/CodeGen/ARM/ssub_sat.ll
llvm/test/CodeGen/ARM/ssub_sat_plus.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index 5573413ad4f0e..0ad6c07f36868 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -960,6 +960,8 @@ ARMTargetLowering::ARMTargetLowering(const TargetMachine &TM_,
setOperationAction(ISD::UADDO_CARRY, MVT::i32, Custom);
setOperationAction(ISD::USUBO_CARRY, MVT::i32, Custom);
+ setOperationAction(ISD::SADDO_CARRY, MVT::i32, Custom);
+ setOperationAction(ISD::SSUBO_CARRY, MVT::i32, Custom);
if (Subtarget->hasDSP()) {
setOperationAction(ISD::SADDSAT, MVT::i8, Custom);
setOperationAction(ISD::SSUBSAT, MVT::i8, Custom);
@@ -4721,28 +4723,45 @@ ARMTargetLowering::getARMXALUOOp(SDValue Op, SelectionDAG &DAG,
return std::make_pair(Value, OverflowCmp);
}
-static SDValue ConvertBooleanCarryToCarryFlag(SDValue BoolCarry,
- SelectionDAG &DAG) {
- SDLoc DL(BoolCarry);
- EVT CarryVT = BoolCarry.getValueType();
+static SDValue valueToCarryFlag(SDValue Value, SelectionDAG &DAG, bool Invert) {
+ SDLoc DL(Value);
+ EVT VT = Value.getValueType();
+
+ if (Invert)
+ Value = DAG.getNode(ISD::SUB, DL, MVT::i32,
+ DAG.getConstant(1, DL, MVT::i32), Value);
- // This converts the boolean value carry into the carry flag by doing
- // ARMISD::SUBC Carry, 1
- SDValue Carry = DAG.getNode(ARMISD::SUBC, DL,
- DAG.getVTList(CarryVT, MVT::i32),
- BoolCarry, DAG.getConstant(1, DL, CarryVT));
- return Carry.getValue(1);
+ SDValue Cmp = DAG.getNode(ARMISD::SUBC, DL, DAG.getVTList(VT, MVT::i32),
+ Value, DAG.getConstant(1, DL, VT));
+ return Cmp.getValue(1);
}
-static SDValue ConvertCarryFlagToBooleanCarry(SDValue Flags, EVT VT,
- SelectionDAG &DAG) {
+static SDValue carryFlagToValue(SDValue Flags, EVT VT, SelectionDAG &DAG,
+ bool Invert) {
SDLoc DL(Flags);
+ if (Invert) {
+ // Convert flags to boolean with ADDE 0,0,Carry then compute 1 - bool.
+ SDValue BoolCarry = DAG.getNode(
+ ARMISD::ADDE, DL, DAG.getVTList(VT, MVT::i32),
+ DAG.getConstant(0, DL, VT), DAG.getConstant(0, DL, VT), Flags);
+ return DAG.getNode(ISD::SUB, DL, VT, DAG.getConstant(1, DL, VT), BoolCarry);
+ }
+
// Now convert the carry flag into a boolean carry. We do this
- // using ARMISD:ADDE 0, 0, Carry
+ // using ARMISD::ADDE 0, 0, Carry
return DAG.getNode(ARMISD::ADDE, DL, DAG.getVTList(VT, MVT::i32),
- DAG.getConstant(0, DL, MVT::i32),
- DAG.getConstant(0, DL, MVT::i32), Flags);
+ DAG.getConstant(0, DL, VT), DAG.getConstant(0, DL, VT),
+ Flags);
+}
+
+// Value is 1 if 'V' bit is 1, else 0
+static SDValue overflowFlagToValue(SDValue Flags, EVT VT, SelectionDAG &DAG) {
+ SDLoc DL(Flags);
+ SDValue Zero = DAG.getConstant(0, DL, VT);
+ SDValue One = DAG.getConstant(1, DL, VT);
+ SDValue ARMcc = DAG.getConstant(ARMCC::VS, DL, MVT::i32);
+ return DAG.getNode(ARMISD::CMOV, DL, VT, Zero, One, ARMcc, Flags);
}
SDValue ARMTargetLowering::LowerALUO(SDValue Op, SelectionDAG &DAG) const {
@@ -4762,16 +4781,12 @@ SDValue ARMTargetLowering::LowerALUO(SDValue Op, SelectionDAG &DAG) const {
case ISD::UADDO:
Value = DAG.getNode(ARMISD::ADDC, dl, VTs, LHS, RHS);
// Convert the carry flag into a boolean value.
- Overflow = ConvertCarryFlagToBooleanCarry(Value.getValue(1), VT, DAG);
+ Overflow = carryFlagToValue(Value.getValue(1), VT, DAG, false);
break;
case ISD::USUBO:
Value = DAG.getNode(ARMISD::SUBC, dl, VTs, LHS, RHS);
// Convert the carry flag into a boolean value.
- Overflow = ConvertCarryFlagToBooleanCarry(Value.getValue(1), VT, DAG);
- // ARMISD::SUBC returns 0 when we have to borrow, so make it an overflow
- // value. So compute 1 - C.
- Overflow = DAG.getNode(ISD::SUB, dl, MVT::i32,
- DAG.getConstant(1, dl, MVT::i32), Overflow);
+ Overflow = carryFlagToValue(Value.getValue(1), VT, DAG, true);
break;
default: {
// Handle other operations with getARMXALUOOp
@@ -6705,21 +6720,19 @@ static SDValue LowerVSETCC(SDValue Op, SelectionDAG &DAG,
static SDValue LowerSETCCCARRY(SDValue Op, SelectionDAG &DAG) {
SDValue LHS = Op.getOperand(0);
SDValue RHS = Op.getOperand(1);
+
+ assert(LHS.getSimpleValueType().isInteger() && "SETCCCARRY is integer only.");
+
SDValue Carry = Op.getOperand(2);
SDValue Cond = Op.getOperand(3);
SDLoc DL(Op);
- assert(LHS.getSimpleValueType().isInteger() && "SETCCCARRY is integer only.");
-
// ARMISD::SUBE expects a carry not a borrow like ISD::USUBO_CARRY so we
// have to invert the carry first.
- Carry = DAG.getNode(ISD::SUB, DL, MVT::i32,
- DAG.getConstant(1, DL, MVT::i32), Carry);
- // This converts the boolean value carry into the carry flag.
- Carry = ConvertBooleanCarryToCarryFlag(Carry, DAG);
+ SDValue InvCarry = valueToCarryFlag(Carry, DAG, true);
SDVTList VTs = DAG.getVTList(LHS.getValueType(), MVT::i32);
- SDValue Cmp = DAG.getNode(ARMISD::SUBE, DL, VTs, LHS, RHS, Carry);
+ SDValue Cmp = DAG.getNode(ARMISD::SUBE, DL, VTs, LHS, RHS, InvCarry);
SDValue FVal = DAG.getConstant(0, DL, MVT::i32);
SDValue TVal = DAG.getConstant(1, DL, MVT::i32);
@@ -9612,48 +9625,26 @@ static SDValue LowerUDIV(SDValue Op, SelectionDAG &DAG,
return N0;
}
-static SDValue LowerUADDSUBO_CARRY(SDValue Op, SelectionDAG &DAG) {
- SDNode *N = Op.getNode();
- EVT VT = N->getValueType(0);
- SDVTList VTs = DAG.getVTList(VT, MVT::i32);
+static SDValue LowerADDSUBO_CARRY(SDValue Op, SelectionDAG &DAG,
+ unsigned Opcode, bool IsSigned) {
+ EVT VT0 = Op.getValue(0).getValueType();
+ EVT VT1 = Op.getValue(1).getValueType();
- SDValue Carry = Op.getOperand(2);
+ bool InvertCarry = Opcode == ARMISD::SUBE;
+ SDValue OpLHS = Op.getOperand(0);
+ SDValue OpRHS = Op.getOperand(1);
+ SDValue OpCarryIn = valueToCarryFlag(Op.getOperand(2), DAG, InvertCarry);
SDLoc DL(Op);
- SDValue Result;
- if (Op.getOpcode() == ISD::UADDO_CARRY) {
- // This converts the boolean value carry into the carry flag.
- Carry = ConvertBooleanCarryToCarryFlag(Carry, DAG);
-
- // Do the addition proper using the carry flag we wanted.
- Result = DAG.getNode(ARMISD::ADDE, DL, VTs, Op.getOperand(0),
- Op.getOperand(1), Carry);
-
- // Now convert the carry flag into a boolean value.
- Carry = ConvertCarryFlagToBooleanCarry(Result.getValue(1), VT, DAG);
- } else {
- // ARMISD::SUBE expects a carry not a borrow like ISD::USUBO_CARRY so we
- // have to invert the carry first.
- Carry = DAG.getNode(ISD::SUB, DL, MVT::i32,
- DAG.getConstant(1, DL, MVT::i32), Carry);
- // This converts the boolean value carry into the carry flag.
- Carry = ConvertBooleanCarryToCarryFlag(Carry, DAG);
-
- // Do the subtraction proper using the carry flag we wanted.
- Result = DAG.getNode(ARMISD::SUBE, DL, VTs, Op.getOperand(0),
- Op.getOperand(1), Carry);
+ SDValue Result = DAG.getNode(Opcode, DL, DAG.getVTList(VT0, MVT::i32), OpLHS,
+ OpRHS, OpCarryIn);
- // Now convert the carry flag into a boolean value.
- Carry = ConvertCarryFlagToBooleanCarry(Result.getValue(1), VT, DAG);
- // But the carry returned by ARMISD::SUBE is not a borrow as expected
- // by ISD::USUBO_CARRY, so compute 1 - C.
- Carry = DAG.getNode(ISD::SUB, DL, MVT::i32,
- DAG.getConstant(1, DL, MVT::i32), Carry);
- }
+ SDValue OutFlag =
+ IsSigned ? overflowFlagToValue(Result.getValue(1), VT1, DAG)
+ : carryFlagToValue(Result.getValue(1), VT1, DAG, InvertCarry);
- // Return both values.
- return DAG.getNode(ISD::MERGE_VALUES, DL, N->getVTList(), Result, Carry);
+ return DAG.getMergeValues({Result, OutFlag}, DL);
}
SDValue ARMTargetLowering::LowerWindowsDIVLibCall(SDValue Op, SelectionDAG &DAG,
@@ -10519,8 +10510,13 @@ SDValue ARMTargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG) const {
return LowerDIV_Windows(Op, DAG, /* Signed */ false);
return LowerUDIV(Op, DAG, Subtarget);
case ISD::UADDO_CARRY:
+ return LowerADDSUBO_CARRY(Op, DAG, ARMISD::ADDE, false /*unsigned*/);
case ISD::USUBO_CARRY:
- return LowerUADDSUBO_CARRY(Op, DAG);
+ return LowerADDSUBO_CARRY(Op, DAG, ARMISD::SUBE, false /*unsigned*/);
+ case ISD::SADDO_CARRY:
+ return LowerADDSUBO_CARRY(Op, DAG, ARMISD::ADDE, true /*signed*/);
+ case ISD::SSUBO_CARRY:
+ return LowerADDSUBO_CARRY(Op, DAG, ARMISD::SUBE, true /*signed*/);
case ISD::UADDO:
case ISD::USUBO:
case ISD::UMULO:
diff --git a/llvm/test/CodeGen/ARM/sadd_sat.ll b/llvm/test/CodeGen/ARM/sadd_sat.ll
index b8f7a2daaeaba..f0366fe459abb 100644
--- a/llvm/test/CodeGen/ARM/sadd_sat.ll
+++ b/llvm/test/CodeGen/ARM/sadd_sat.ll
@@ -64,86 +64,81 @@ define i32 @func(i32 %x, i32 %y) nounwind {
define i64 @func2(i64 %x, i64 %y) nounwind {
; CHECK-T16-LABEL: func2:
; CHECK-T16: @ %bb.0:
-; CHECK-T16-NEXT: .save {r4, lr}
-; CHECK-T16-NEXT: push {r4, lr}
-; CHECK-T16-NEXT: mov r4, r1
-; CHECK-T16-NEXT: eors r1, r3
-; CHECK-T16-NEXT: adds r2, r0, r2
-; CHECK-T16-NEXT: adcs r3, r4
-; CHECK-T16-NEXT: eors r4, r3
-; CHECK-T16-NEXT: bics r4, r1
-; CHECK-T16-NEXT: asrs r0, r3, #31
-; CHECK-T16-NEXT: movs r1, #1
-; CHECK-T16-NEXT: lsls r1, r1, #31
-; CHECK-T16-NEXT: eors r1, r0
-; CHECK-T16-NEXT: cmp r4, #0
-; CHECK-T16-NEXT: bpl .LBB1_3
+; CHECK-T16-NEXT: .save {r4, r5, r7, lr}
+; CHECK-T16-NEXT: push {r4, r5, r7, lr}
+; CHECK-T16-NEXT: mov r5, r0
+; CHECK-T16-NEXT: adds r0, r0, r2
+; CHECK-T16-NEXT: mov r0, r1
+; CHECK-T16-NEXT: adcs r0, r3
+; CHECK-T16-NEXT: asrs r0, r0, #31
+; CHECK-T16-NEXT: movs r4, #1
+; CHECK-T16-NEXT: lsls r4, r4, #31
+; CHECK-T16-NEXT: eors r4, r0
+; CHECK-T16-NEXT: adds r2, r5, r2
+; CHECK-T16-NEXT: adcs r1, r3
+; CHECK-T16-NEXT: bvc .LBB1_3
; CHECK-T16-NEXT: @ %bb.1:
-; CHECK-T16-NEXT: bpl .LBB1_4
+; CHECK-T16-NEXT: bvc .LBB1_4
; CHECK-T16-NEXT: .LBB1_2:
-; CHECK-T16-NEXT: pop {r4, pc}
+; CHECK-T16-NEXT: mov r1, r4
+; CHECK-T16-NEXT: pop {r4, r5, r7, pc}
; CHECK-T16-NEXT: .LBB1_3:
; CHECK-T16-NEXT: mov r0, r2
-; CHECK-T16-NEXT: bmi .LBB1_2
+; CHECK-T16-NEXT: bvs .LBB1_2
; CHECK-T16-NEXT: .LBB1_4:
-; CHECK-T16-NEXT: mov r1, r3
-; CHECK-T16-NEXT: pop {r4, pc}
+; CHECK-T16-NEXT: mov r4, r1
+; CHECK-T16-NEXT: mov r1, r4
+; CHECK-T16-NEXT: pop {r4, r5, r7, pc}
;
; CHECK-T2-LABEL: func2:
; CHECK-T2: @ %bb.0:
; CHECK-T2-NEXT: adds r0, r0, r2
-; CHECK-T2-NEXT: eor.w r12, r1, r3
-; CHECK-T2-NEXT: adc.w r2, r1, r3
-; CHECK-T2-NEXT: eors r1, r2
-; CHECK-T2-NEXT: bics.w r1, r1, r12
-; CHECK-T2-NEXT: it mi
-; CHECK-T2-NEXT: asrmi r0, r2, #31
-; CHECK-T2-NEXT: mov.w r1, #-2147483648
-; CHECK-T2-NEXT: it mi
-; CHECK-T2-NEXT: eormi.w r2, r1, r2, asr #31
-; CHECK-T2-NEXT: mov r1, r2
+; CHECK-T2-NEXT: mov.w r2, #-2147483648
+; CHECK-T2-NEXT: adcs r1, r3
+; CHECK-T2-NEXT: it vs
+; CHECK-T2-NEXT: asrvs r0, r1, #31
+; CHECK-T2-NEXT: it vs
+; CHECK-T2-NEXT: eorvs.w r1, r2, r1, asr #31
; CHECK-T2-NEXT: bx lr
;
; CHECK-ARM-LABEL: func2:
; CHECK-ARM: @ %bb.0:
; CHECK-ARM-NEXT: adds r0, r0, r2
-; CHECK-ARM-NEXT: eor r12, r1, r3
-; CHECK-ARM-NEXT: adc r2, r1, r3
-; CHECK-ARM-NEXT: eor r1, r1, r2
-; CHECK-ARM-NEXT: bics r1, r1, r12
-; CHECK-ARM-NEXT: asrmi r0, r2, #31
-; CHECK-ARM-NEXT: mov r1, #-2147483648
-; CHECK-ARM-NEXT: eormi r2, r1, r2, asr #31
-; CHECK-ARM-NEXT: mov r1, r2
+; CHECK-ARM-NEXT: mov r2, #-2147483648
+; CHECK-ARM-NEXT: adcs r1, r1, r3
+; CHECK-ARM-NEXT: asrvs r0, r1, #31
+; CHECK-ARM-NEXT: eorvs r1, r2, r1, asr #31
; CHECK-ARM-NEXT: bx lr
;
; CHECK-T15TE-LABEL: func2:
; CHECK-T15TE: @ %bb.0:
-; CHECK-T15TE-NEXT: .save {r4, lr}
-; CHECK-T15TE-NEXT: push {r4, lr}
-; CHECK-T15TE-NEXT: movs r4, r1
-; CHECK-T15TE-NEXT: eors r1, r3
-; CHECK-T15TE-NEXT: adds r2, r0, r2
-; CHECK-T15TE-NEXT: adcs r3, r4
-; CHECK-T15TE-NEXT: eors r4, r3
-; CHECK-T15TE-NEXT: bics r4, r1
-; CHECK-T15TE-NEXT: asrs r0, r3, #31
-; CHECK-T15TE-NEXT: movs r1, #1
-; CHECK-T15TE-NEXT: lsls r1, r1, #31
-; CHECK-T15TE-NEXT: eors r1, r0
-; CHECK-T15TE-NEXT: cmp r4, #0
-; CHECK-T15TE-NEXT: bpl .LBB1_3
+; CHECK-T15TE-NEXT: .save {r4, r5, r7, lr}
+; CHECK-T15TE-NEXT: push {r4, r5, r7, lr}
+; CHECK-T15TE-NEXT: movs r5, r0
+; CHECK-T15TE-NEXT: adds r0, r0, r2
+; CHECK-T15TE-NEXT: mov r12, r1
+; CHECK-T15TE-NEXT: mov r0, r12
+; CHECK-T15TE-NEXT: adcs r0, r3
+; CHECK-T15TE-NEXT: asrs r0, r0, #31
+; CHECK-T15TE-NEXT: movs r4, #1
+; CHECK-T15TE-NEXT: lsls r4, r4, #31
+; CHECK-T15TE-NEXT: eors r4, r0
+; CHECK-T15TE-NEXT: adds r2, r5, r2
+; CHECK-T15TE-NEXT: adcs r1, r3
+; CHECK-T15TE-NEXT: bvc .LBB1_3
; CHECK-T15TE-NEXT: @ %bb.1:
-; CHECK-T15TE-NEXT: bpl .LBB1_4
+; CHECK-T15TE-NEXT: bvc .LBB1_4
; CHECK-T15TE-NEXT: .LBB1_2:
-; CHECK-T15TE-NEXT: pop {r4, pc}
+; CHECK-T15TE-NEXT: movs r1, r4
+; CHECK-T15TE-NEXT: pop {r4, r5, r7, pc}
; CHECK-T15TE-NEXT: .LBB1_3:
; CHECK-T15TE-NEXT: mov r12, r2
; CHECK-T15TE-NEXT: mov r0, r12
-; CHECK-T15TE-NEXT: bmi .LBB1_2
+; CHECK-T15TE-NEXT: bvs .LBB1_2
; CHECK-T15TE-NEXT: .LBB1_4:
-; CHECK-T15TE-NEXT: movs r1, r3
-; CHECK-T15TE-NEXT: pop {r4, pc}
+; CHECK-T15TE-NEXT: movs r4, r1
+; CHECK-T15TE-NEXT: movs r1, r4
+; CHECK-T15TE-NEXT: pop {r4, r5, r7, pc}
%tmp = call i64 @llvm.sadd.sat.i64(i64 %x, i64 %y)
ret i64 %tmp
}
diff --git a/llvm/test/CodeGen/ARM/sadd_sat_plus.ll b/llvm/test/CodeGen/ARM/sadd_sat_plus.ll
index 0ddb64fc3f2d1..d54c97cca9eb0 100644
--- a/llvm/test/CodeGen/ARM/sadd_sat_plus.ll
+++ b/llvm/test/CodeGen/ARM/sadd_sat_plus.ll
@@ -53,65 +53,54 @@ define i32 @func32(i32 %x, i32 %y, i32 %z) nounwind {
define i64 @func64(i64 %x, i64 %y, i64 %z) nounwind {
; CHECK-T1-LABEL: func64:
; CHECK-T1: @ %bb.0:
-; CHECK-T1-NEXT: .save {r4, lr}
-; CHECK-T1-NEXT: push {r4, lr}
-; CHECK-T1-NEXT: ldr r3, [sp, #12]
-; CHECK-T1-NEXT: mov r2, r1
-; CHECK-T1-NEXT: eors r2, r3
-; CHECK-T1-NEXT: ldr r4, [sp, #8]
-; CHECK-T1-NEXT: adds r4, r0, r4
-; CHECK-T1-NEXT: adcs r3, r1
-; CHECK-T1-NEXT: eors r1, r3
-; CHECK-T1-NEXT: bics r1, r2
-; CHECK-T1-NEXT: asrs r0, r3, #31
+; CHECK-T1-NEXT: .save {r4, r5, r7, lr}
+; CHECK-T1-NEXT: push {r4, r5, r7, lr}
+; CHECK-T1-NEXT: mov r3, r0
+; CHECK-T1-NEXT: ldr r4, [sp, #20]
+; CHECK-T1-NEXT: ldr r5, [sp, #16]
+; CHECK-T1-NEXT: adds r0, r0, r5
+; CHECK-T1-NEXT: mov r0, r1
+; CHECK-T1-NEXT: adcs r0, r4
+; CHECK-T1-NEXT: asrs r0, r0, #31
; CHECK-T1-NEXT: movs r2, #1
; CHECK-T1-NEXT: lsls r2, r2, #31
; CHECK-T1-NEXT: eors r2, r0
-; CHECK-T1-NEXT: cmp r1, #0
-; CHECK-T1-NEXT: bpl .LBB1_3
+; CHECK-T1-NEXT: adds r3, r3, r5
+; CHECK-T1-NEXT: adcs r1, r4
+; CHECK-T1-NEXT: bvc .LBB1_3
; CHECK-T1-NEXT: @ %bb.1:
-; CHECK-T1-NEXT: bpl .LBB1_4
+; CHECK-T1-NEXT: bvc .LBB1_4
; CHECK-T1-NEXT: .LBB1_2:
; CHECK-T1-NEXT: mov r1, r2
-; CHECK-T1-NEXT: pop {r4, pc}
+; CHECK-T1-NEXT: pop {r4, r5, r7, pc}
; CHECK-T1-NEXT: .LBB1_3:
-; CHECK-T1-NEXT: mov r0, r4
-; CHECK-T1-NEXT: bmi .LBB1_2
+; CHECK-T1-NEXT: mov r0, r3
+; CHECK-T1-NEXT: bvs .LBB1_2
; CHECK-T1-NEXT: .LBB1_4:
-; CHECK-T1-NEXT: mov r2, r3
+; CHECK-T1-NEXT: mov r2, r1
; CHECK-T1-NEXT: mov r1, r2
-; CHECK-T1-NEXT: pop {r4, pc}
+; CHECK-T1-NEXT: pop {r4, r5, r7, pc}
;
; CHECK-T2-LABEL: func64:
; CHECK-T2: @ %bb.0:
-; CHECK-T2-NEXT: ldr r2, [sp]
-; CHECK-T2-NEXT: ldr.w r12, [sp, #4]
-; CHECK-T2-NEXT: adds r0, r0, r2
-; CHECK-T2-NEXT: adc.w r2, r1, r12
-; CHECK-T2-NEXT: eor.w r3, r1, r12
-; CHECK-T2-NEXT: eors r1, r2
-; CHECK-T2-NEXT: bics r1, r3
-; CHECK-T2-NEXT: it mi
-; CHECK-T2-NEXT: asrmi r0, r2, #31
-; CHECK-T2-NEXT: mov.w r1, #-2147483648
-; CHECK-T2-NEXT: it mi
-; CHECK-T2-NEXT: eormi.w r2, r1, r2, asr #31
-; CHECK-T2-NEXT: mov r1, r2
+; CHECK-T2-NEXT: ldrd r3, r2, [sp]
+; CHECK-T2-NEXT: adds r0, r0, r3
+; CHECK-T2-NEXT: adcs r1, r2
+; CHECK-T2-NEXT: it vs
+; CHECK-T2-NEXT: asrvs r0, r1, #31
+; CHECK-T2-NEXT: mov.w r2, #-2147483648
+; CHECK-T2-NEXT: it vs
+; CHECK-T2-NEXT: eorvs.w r1, r2, r1, asr #31
; CHECK-T2-NEXT: bx lr
;
; CHECK-ARM-LABEL: func64:
; CHECK-ARM: @ %bb.0:
-; CHECK-ARM-NEXT: ldr r12, [sp]
-; CHECK-ARM-NEXT: ldr r2, [sp, #4]
-; CHECK-ARM-NEXT: adds r0, r0, r12
-; CHECK-ARM-NEXT: eor r3, r1, r2
-; CHECK-ARM-NEXT: adc r2, r1, r2
-; CHECK-ARM-NEXT: eor r1, r1, r2
-; CHECK-ARM-NEXT: bics r1, r1, r3
-; CHECK-ARM-NEXT: asrmi r0, r2, #31
-; CHECK-ARM-NEXT: mov r1, #-2147483648
-; CHECK-ARM-NEXT: eormi r2, r1, r2, asr #31
-; CHECK-ARM-NEXT: mov r1, r2
+; CHECK-ARM-NEXT: ldm sp, {r2, r3}
+; CHECK-ARM-NEXT: adds r0, r0, r2
+; CHECK-ARM-NEXT: mov r2, #-2147483648
+; CHECK-ARM-NEXT: adcs r1, r1, r3
+; CHECK-ARM-NEXT: asrvs r0, r1, #31
+; CHECK-ARM-NEXT: eorvs r1, r2, r1, asr #31
; CHECK-ARM-NEXT: bx lr
%a = mul i64 %y, %z
%tmp = call i64 @llvm.sadd.sat.i64(i64 %x, i64 %z)
diff --git a/llvm/test/CodeGen/ARM/ssub_sat.ll b/llvm/test/CodeGen/ARM/ssub_sat.ll
index 0978bfd1f0140..8cce3a76a50db 100644
--- a/llvm/test/CodeGen/ARM/ssub_sat.ll
+++ b/llvm/test/CodeGen/ARM/ssub_sat.ll
@@ -64,56 +64,48 @@ define i64 @func2(i64 %x, i64 %y) nounwind {
; CHECK-T1: @ %bb.0:
; CHECK-T1-NEXT: .save {r4, r5, r7, lr}
; CHECK-T1-NEXT: push {r4, r5, r7, lr}
-; CHECK-T1-NEXT: mov r4, r1
-; CHECK-T1-NEXT: eors r1, r3
-; CHECK-T1-NEXT: subs r5, r0, r2
-; CHECK-T1-NEXT: mov r2, r4
-; CHECK-T1-NEXT: sbcs r2, r3
-; CHECK-T1-NEXT: eors r4, r2
-; CHECK-T1-NEXT: ands r4, r1
-; CHECK-T1-NEXT: asrs r0, r2, #31
-; CHECK-T1-NEXT: movs r1, #1
-; CHECK-T1-NEXT: lsls r1, r1, #31
-; CHECK-T1-NEXT: eors r1, r0
-; CHECK-T1-NEXT: cmp r4, #0
-; CHECK-T1-NEXT: bpl .LBB1_3
+; CHECK-T1-NEXT: mov r5, r0
+; CHECK-T1-NEXT: subs r0, r0, r2
+; CHECK-T1-NEXT: mov r0, r1
+; CHECK-T1-NEXT: sbcs r0, r3
+; CHECK-T1-NEXT: asrs r0, r0, #31
+; CHECK-T1-NEXT: movs r4, #1
+; CHECK-T1-NEXT: lsls r4, r4, #31
+; CHECK-T1-NEXT: eors r4, r0
+; CHECK-T1-NEXT: subs r2, r5, r2
+; CHECK-T1-NEXT: sbcs r1, r3
+; CHECK-T1-NEXT: bvc .LBB1_3
; CHECK-T1-NEXT: @ %bb.1:
-; CHECK-T1-NEXT: bpl .LBB1_4
+; CHECK-T1-NEXT: bvc .LBB1_4
; CHECK-T1-NEXT: .LBB1_2:
+; CHECK-T1-NEXT: mov r1, r4
; CHECK-T1-NEXT: pop {r4, r5, r7, pc}
; CHECK-T1-NEXT: .LBB1_3:
-; CHECK-T1-NEXT: mov r0, r5
-; CHECK-T1-NEXT: bmi .LBB1_2
+; CHECK-T1-NEXT: mov r0, r2
+; CHECK-T1-NEXT: bvs .LBB1_2
; CHECK-T1-NEXT: .LBB1_4:
-; CHECK-T1-NEXT: mov r1, r2
+; CHECK-T1-NEXT: mov r4, r1
+; CHECK-T1-NEXT: mov r1, r4
; CHECK-T1-NEXT: pop {r4, r5, r7, pc}
;
; CHECK-T2-LABEL: func2:
; CHECK-T2: @ %bb.0:
; CHECK-T2-NEXT: subs r0, r0, r2
-; CHECK-T2-NEXT: eor.w r12, r1, r3
-; CHECK-T2-NEXT: sbc.w r2, r1, r3
-; CHECK-T2-NEXT: eors r1, r2
-; CHECK-T2-NEXT: ands.w r1, r1, r12
-; CHECK-T2-NEXT: it mi
-; CHECK-T2-NEXT: asrmi r0, r2, #31
-; CHECK-T2-NEXT: mov.w r1, #-2147483648
-; CHECK-T2-NEXT: it mi
-; CHECK-T2-NEXT: eormi.w r2, r1, r2, asr #31
-; CHECK-T2-NEXT: mov r1, r2
+; CHECK-T2-NEXT: mov.w r2, #-2147483648
+; CHECK-T2-NEXT: sbcs r1, r3
+; CHECK-T2-NEXT: it vs
+; CHECK-T2-NEXT: asrvs r0, r1, #31
+; CHECK-T2-NEXT: it vs
+; CHECK-T2-NEXT: eorvs.w r1, r2, r1, asr #31
; CHECK-T2-NEXT: bx lr
;
; CHECK-ARM-LABEL: func2:
; CHECK-ARM: @ %bb.0:
; CHECK-ARM-NEXT: subs r0, r0, r2
-; CHECK-ARM-NEXT: eor r12, r1, r3
-; CHECK-ARM-NEXT: sbc r2, r1, r3
-; CHECK-ARM-NEXT: eor r1, r1, r2
-; CHECK-ARM-NEXT: ands r1, r12, r1
-; CHECK-ARM-NEXT: asrmi r0, r2, #31
-; CHECK-ARM-NEXT: mov r1, #-2147483648
-; CHECK-ARM-NEXT: eormi r2, r1, r2, asr #31
-; CHECK-ARM-NEXT: mov r1, r2
+; CHECK-ARM-NEXT: mov r2, #-2147483648
+; CHECK-ARM-NEXT: sbcs r1, r1, r3
+; CHECK-ARM-NEXT: asrvs r0, r1, #31
+; CHECK-ARM-NEXT: eorvs r1, r2, r1, asr #31
; CHECK-ARM-NEXT: bx lr
%tmp = call i64 @llvm.ssub.sat.i64(i64 %x, i64 %y)
ret i64 %tmp
diff --git a/llvm/test/CodeGen/ARM/ssub_sat_plus.ll b/llvm/test/CodeGen/ARM/ssub_sat_plus.ll
index adf6cafc6ccb8..9690918d1f702 100644
--- a/llvm/test/CodeGen/ARM/ssub_sat_plus.ll
+++ b/llvm/test/CodeGen/ARM/ssub_sat_plus.ll
@@ -56,64 +56,52 @@ define i64 @func64(i64 %x, i64 %y, i64 %z) nounwind {
; CHECK-T1: @ %bb.0:
; CHECK-T1-NEXT: .save {r4, r5, r7, lr}
; CHECK-T1-NEXT: push {r4, r5, r7, lr}
-; CHECK-T1-NEXT: ldr r2, [sp, #20]
-; CHECK-T1-NEXT: mov r5, r1
-; CHECK-T1-NEXT: eors r5, r2
-; CHECK-T1-NEXT: ldr r3, [sp, #16]
-; CHECK-T1-NEXT: subs r4, r0, r3
-; CHECK-T1-NEXT: mov r3, r1
-; CHECK-T1-NEXT: sbcs r3, r2
-; CHECK-T1-NEXT: eors r1, r3
-; CHECK-T1-NEXT: ands r1, r5
-; CHECK-T1-NEXT: asrs r0, r3, #31
+; CHECK-T1-NEXT: mov r3, r0
+; CHECK-T1-NEXT: ldr r4, [sp, #20]
+; CHECK-T1-NEXT: ldr r5, [sp, #16]
+; CHECK-T1-NEXT: subs r0, r0, r5
+; CHECK-T1-NEXT: mov r0, r1
+; CHECK-T1-NEXT: sbcs r0, r4
+; CHECK-T1-NEXT: asrs r0, r0, #31
; CHECK-T1-NEXT: movs r2, #1
; CHECK-T1-NEXT: lsls r2, r2, #31
; CHECK-T1-NEXT: eors r2, r0
-; CHECK-T1-NEXT: cmp r1, #0
-; CHECK-T1-NEXT: bpl .LBB1_3
+; CHECK-T1-NEXT: subs r3, r3, r5
+; CHECK-T1-NEXT: sbcs r1, r4
+; CHECK-T1-NEXT: bvc .LBB1_3
; CHECK-T1-NEXT: @ %bb.1:
-; CHECK-T1-NEXT: bpl .LBB1_4
+; CHECK-T1-NEXT: bvc .LBB1_4
; CHECK-T1-NEXT: .LBB1_2:
; CHECK-T1-NEXT: mov r1, r2
; CHECK-T1-NEXT: pop {r4, r5, r7, pc}
; CHECK-T1-NEXT: .LBB1_3:
-; CHECK-T1-NEXT: mov r0, r4
-; CHECK-T1-NEXT: bmi .LBB1_2
+; CHECK-T1-NEXT: mov r0, r3
+; CHECK-T1-NEXT: bvs .LBB1_2
; CHECK-T1-NEXT: .LBB1_4:
-; CHECK-T1-NEXT: mov r2, r3
+; CHECK-T1-NEXT: mov r2, r1
; CHECK-T1-NEXT: mov r1, r2
; CHECK-T1-NEXT: pop {r4, r5, r7, pc}
;
; CHECK-T2-LABEL: func64:
; CHECK-T2: @ %bb.0:
-; CHECK-T2-NEXT: ldr r2, [sp]
-; CHECK-T2-NEXT: ldr.w r12, [sp, #4]
-; CHECK-T2-NEXT: subs r0, r0, r2
-; CHECK-T2-NEXT: sbc.w r2, r1, r12
-; CHECK-T2-NEXT: eor.w r3, r1, r12
-; CHECK-T2-NEXT: eors r1, r2
-; CHECK-T2-NEXT: ands r1, r3
-; CHECK-T2-NEXT: it mi
-; CHECK-T2-NEXT: asrmi r0, r2, #31
-; CHECK-T2-NEXT: mov.w r1, #-2147483648
-; CHECK-T2-NEXT: it mi
-; CHECK-T2-NEXT: eormi.w r2, r1, r2, asr #31
-; CHECK-T2-NEXT: mov r1, r2
+; CHECK-T2-NEXT: ldrd r3, r2, [sp]
+; CHECK-T2-NEXT: subs r0, r0, r3
+; CHECK-T2-NEXT: sbcs r1, r2
+; CHECK-T2-NEXT: it vs
+; CHECK-T2-NEXT: asrvs r0, r1, #31
+; CHECK-T2-NEXT: mov.w r2, #-2147483648
+; CHECK-T2-NEXT: it vs
+; CHECK-T2-NEXT: eorvs.w r1, r2, r1, asr #31
; CHECK-T2-NEXT: bx lr
;
; CHECK-ARM-LABEL: func64:
; CHECK-ARM: @ %bb.0:
-; CHECK-ARM-NEXT: ldr r12, [sp]
-; CHECK-ARM-NEXT: ldr r2, [sp, #4]
-; CHECK-ARM-NEXT: subs r0, r0, r12
-; CHECK-ARM-NEXT: eor r3, r1, r2
-; CHECK-ARM-NEXT: sbc r2, r1, r2
-; CHECK-ARM-NEXT: eor r1, r1, r2
-; CHECK-ARM-NEXT: ands r1, r3, r1
-; CHECK-ARM-NEXT: asrmi r0, r2, #31
-; CHECK-ARM-NEXT: mov r1, #-2147483648
-; CHECK-ARM-NEXT: eormi r2, r1, r2, asr #31
-; CHECK-ARM-NEXT: mov r1, r2
+; CHECK-ARM-NEXT: ldm sp, {r2, r3}
+; CHECK-ARM-NEXT: subs r0, r0, r2
+; CHECK-ARM-NEXT: mov r2, #-2147483648
+; CHECK-ARM-NEXT: sbcs r1, r1, r3
+; CHECK-ARM-NEXT: asrvs r0, r1, #31
+; CHECK-ARM-NEXT: eorvs r1, r2, r1, asr #31
; CHECK-ARM-NEXT: bx lr
%a = mul i64 %y, %z
%tmp = call i64 @llvm.ssub.sat.i64(i64 %x, i64 %z)
More information about the llvm-commits
mailing list