Skip to content

Commit

Permalink
Adding missed optimisation
Browse files Browse the repository at this point in the history
  • Loading branch information
@jf-botto
jf-botto committed Oct 2, 2024
1 parent 0ca5720 commit 1ccc6c8
Show file tree
Hide file tree
Showing 3 changed files with 57 additions and 47 deletions.
26 changes: 26 additions & 0 deletions llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4301,6 +4301,29 @@ static SDValue LowerPREFETCH(SDValue Op, SelectionDAG &DAG) {
Op.getOperand(1));
}

// Converts SETCC (AND X Y) Z ULT -> SETCC (AND X (Y & ~(Z - 1)) 0 EQ when Y is
// a power of 2. This is then lowered to ANDS X (Y & ~(Z - 1)) instead of SUBS
// (AND X Y) Z which produces a better opt with EmitComparison
static void simplifySetCCIntoEq(ISD::CondCode &CC, SDValue &LHS, SDValue &RHS,
SelectionDAG &DAG, const SDLoc dl) {
if (CC == ISD::SETULT && LHS.getOpcode() == ISD::AND && LHS->hasOneUse()) {
ConstantSDNode *LHSConstOp = dyn_cast<ConstantSDNode>(LHS.getOperand(1));
ConstantSDNode *RHSConst = dyn_cast<ConstantSDNode>(RHS);
if (LHSConstOp && RHSConst) {
uint64_t lhsConstValue = LHSConstOp->getZExtValue();
uint64_t rhsConstant = RHSConst->getZExtValue();
if (isPowerOf2_64(rhsConstant)) {
uint64_t newMaskValue = lhsConstValue & ~(rhsConstant - 1);
LHS =
DAG.getNode(ISD::AND, dl, LHS.getValueType(), LHS.getOperand(0),
DAG.getConstant(newMaskValue, dl, LHS.getValueType()));
RHS = DAG.getConstant(0, dl, RHS.getValueType());
CC = ISD::SETEQ;
}
}
}
}

SDValue AArch64TargetLowering::LowerFP_EXTEND(SDValue Op,
SelectionDAG &DAG) const {
EVT VT = Op.getValueType();
Expand Down Expand Up @@ -10589,6 +10612,9 @@ SDValue AArch64TargetLowering::LowerSETCC(SDValue Op, SelectionDAG &DAG) const {
}

if (LHS.getValueType().isInteger()) {

simplifySetCCIntoEq(CC, LHS, RHS, DAG, dl);

SDValue CCVal;
SDValue Cmp = getAArch64Cmp(
LHS, RHS, ISD::getSetCCInverse(CC, LHS.getValueType()), CCVal, DAG, dl);
Expand Down
53 changes: 21 additions & 32 deletions llvm/test/CodeGen/AArch64/icmp-ult-eq-fold.ll
View file
Original file line number Diff line number Diff line change
Expand Up @@ -5,9 +5,8 @@
define i1 @lt8_u8(i8 %0) {
; CHECK-LABEL: lt8_u8:
; CHECK: // %bb.0:
; CHECK-NEXT: and w8, w0, #0xff
; CHECK-NEXT: cmp w8, #8
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: tst w0, #0xf8
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%2 = icmp ult i8 %0, 8
ret i1 %2
Expand All @@ -16,9 +15,8 @@ define i1 @lt8_u8(i8 %0) {
define i1 @lt32_u8(i8 %0) {
; CHECK-LABEL: lt32_u8:
; CHECK: // %bb.0:
; CHECK-NEXT: and w8, w0, #0xff
; CHECK-NEXT: cmp w8, #32
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: tst w0, #0xe0
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%2 = icmp ult i8 %0, 32
ret i1 %2
Expand All @@ -27,9 +25,8 @@ define i1 @lt32_u8(i8 %0) {
define i1 @lt64_u8(i8 %0) {
; CHECK-LABEL: lt64_u8:
; CHECK: // %bb.0:
; CHECK-NEXT: and w8, w0, #0xff
; CHECK-NEXT: cmp w8, #64
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: tst w0, #0xc0
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%2 = icmp ult i8 %0, 64
ret i1 %2
Expand Down Expand Up @@ -98,10 +95,9 @@ define i1 @lt64_u64(i64 %0) {
define i1 @lt8_u16_and_5(i8 %0) {
; CHECK-LABEL: lt8_u16_and_5:
; CHECK: // %bb.0:
; CHECK-NEXT: mov w8, #5 // =0x5
; CHECK-NEXT: and w8, w0, w8
; CHECK-NEXT: cmp w8, #16
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: mov w8, wzr
; CHECK-NEXT: cmp w8, #0
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%2 = and i8 %0, 5
%3 = icmp ult i8 %2, 16
Expand All @@ -111,10 +107,8 @@ define i1 @lt8_u16_and_5(i8 %0) {
define i1 @lt8_u16_and_19(i8 %0) {
; CHECK-LABEL: lt8_u16_and_19:
; CHECK: // %bb.0:
; CHECK-NEXT: mov w8, #19 // =0x13
; CHECK-NEXT: and w8, w0, w8
; CHECK-NEXT: cmp w8, #16
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: tst w0, #0x10
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%2 = and i8 %0, 19
%3 = icmp ult i8 %2, 16
Expand All @@ -124,9 +118,9 @@ define i1 @lt8_u16_and_19(i8 %0) {
define i1 @lt32_u16_and_7(i32 %0) {
; CHECK-LABEL: lt32_u16_and_7:
; CHECK: // %bb.0:
; CHECK-NEXT: and w8, w0, #0x7
; CHECK-NEXT: cmp w8, #16
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: mov w8, wzr
; CHECK-NEXT: cmp w8, #0
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%2 = and i32 %0, 7
%3 = icmp ult i32 %2, 16
Expand All @@ -136,10 +130,8 @@ define i1 @lt32_u16_and_7(i32 %0) {
define i1 @lt32_u16_and_21(i32 %0) {
; CHECK-LABEL: lt32_u16_and_21:
; CHECK: // %bb.0:
; CHECK-NEXT: mov w8, #21 // =0x15
; CHECK-NEXT: and w8, w0, w8
; CHECK-NEXT: cmp w8, #16
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: tst w0, #0x10
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%2 = and i32 %0, 21
%3 = icmp ult i32 %2, 16
Expand All @@ -149,10 +141,9 @@ define i1 @lt32_u16_and_21(i32 %0) {
define i1 @lt64_u16_and_9(i64 %0) {
; CHECK-LABEL: lt64_u16_and_9:
; CHECK: // %bb.0:
; CHECK-NEXT: mov w8, #9 // =0x9
; CHECK-NEXT: and x8, x0, x8
; CHECK-NEXT: cmp x8, #16
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: mov x8, xzr
; CHECK-NEXT: cmp x8, #0
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%2 = and i64 %0, 9
%3 = icmp ult i64 %2, 16
Expand All @@ -162,10 +153,8 @@ define i1 @lt64_u16_and_9(i64 %0) {
define i1 @lt64_u16_and_23(i64 %0) {
; CHECK-LABEL: lt64_u16_and_23:
; CHECK: // %bb.0:
; CHECK-NEXT: mov w8, #23 // =0x17
; CHECK-NEXT: and x8, x0, x8
; CHECK-NEXT: cmp x8, #16
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: tst x0, #0x10
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%2 = and i64 %0, 23
%3 = icmp ult i64 %2, 16
Expand Down
25 changes: 10 additions & 15 deletions llvm/test/CodeGen/AArch64/signed-truncation-check.ll
View file
Original file line number Diff line number Diff line change
Expand Up @@ -287,9 +287,8 @@ define i1 @add_ultcmp_bad_i16_i8_add(i16 %x, i16 %y) nounwind {
; CHECK-LABEL: add_ultcmp_bad_i16_i8_add:
; CHECK: // %bb.0:
; CHECK-NEXT: add w8, w0, w1
; CHECK-NEXT: and w8, w8, #0xffff
; CHECK-NEXT: cmp w8, #256
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: tst w8, #0xff00
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%tmp0 = add i16 %x, %y
%tmp1 = icmp ult i16 %tmp0, 256 ; 1U << 8
Expand Down Expand Up @@ -328,9 +327,8 @@ define i1 @add_ultcmp_bad_i16_i8_c0notpoweroftwo(i16 %x) nounwind {
; CHECK-LABEL: add_ultcmp_bad_i16_i8_c0notpoweroftwo:
; CHECK: // %bb.0:
; CHECK-NEXT: add w8, w0, #192
; CHECK-NEXT: and w8, w8, #0xffff
; CHECK-NEXT: cmp w8, #256
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: tst w8, #0xff00
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%tmp0 = add i16 %x, 192 ; (1U << (8-1)) + (1U << (8-1-1))
%tmp1 = icmp ult i16 %tmp0, 256 ; 1U << 8
Expand All @@ -356,9 +354,8 @@ define i1 @add_ultcmp_bad_i16_i8_magic(i16 %x) nounwind {
; CHECK-LABEL: add_ultcmp_bad_i16_i8_magic:
; CHECK: // %bb.0:
; CHECK-NEXT: add w8, w0, #64
; CHECK-NEXT: and w8, w8, #0xffff
; CHECK-NEXT: cmp w8, #256
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: tst w8, #0xff00
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%tmp0 = add i16 %x, 64 ; 1U << (8-1-1)
%tmp1 = icmp ult i16 %tmp0, 256 ; 1U << 8
Expand All @@ -370,9 +367,8 @@ define i1 @add_ultcmp_bad_i16_i4(i16 %x) nounwind {
; CHECK-LABEL: add_ultcmp_bad_i16_i4:
; CHECK: // %bb.0:
; CHECK-NEXT: add w8, w0, #8
; CHECK-NEXT: and w8, w8, #0xffff
; CHECK-NEXT: cmp w8, #16
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: tst w8, #0xfff0
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%tmp0 = add i16 %x, 8 ; 1U << (4-1)
%tmp1 = icmp ult i16 %tmp0, 16 ; 1U << 4
Expand All @@ -384,9 +380,8 @@ define i1 @add_ultcmp_bad_i24_i8(i24 %x) nounwind {
; CHECK-LABEL: add_ultcmp_bad_i24_i8:
; CHECK: // %bb.0:
; CHECK-NEXT: add w8, w0, #128
; CHECK-NEXT: and w8, w8, #0xffffff
; CHECK-NEXT: cmp w8, #256
; CHECK-NEXT: cset w0, lo
; CHECK-NEXT: tst w8, #0xffff00
; CHECK-NEXT: cset w0, eq
; CHECK-NEXT: ret
%tmp0 = add i24 %x, 128 ; 1U << (8-1)
%tmp1 = icmp ult i24 %tmp0, 256 ; 1U << 8
Expand Down

0 comments on commit 1ccc6c8

Please sign in to comment.