Skip to content

[InstCombine] Canonicalize Bit Testing by Shifting to Sign Bit #101822

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 2 commits into from
Closed

[InstCombine] Canonicalize Bit Testing by Shifting to Sign Bit #101822

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ddd1a4b
[InstCombine] Add Tests for Testing Bits; NFC
mskamp Jul 23, 2024
bee6b50
[InstCombine] Canonicalize Bit Testing by Shifting to Sign Bit
mskamp Jul 1, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 18 additions & 4 deletions llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2305,19 +2305,33 @@ Instruction *InstCombinerImpl::foldICmpShlConstant(ICmpInst &Cmp,
if (C.isZero() || (Pred == ICmpInst::ICMP_SGT ? C.isAllOnes() : C.isOne()))
return new ICmpInst(Pred, Shl->getOperand(0), Cmp.getOperand(1));

unsigned TypeBits = C.getBitWidth();
Value *X = Shl->getOperand(0);
Type *ShType = Shl->getType();

// (icmp slt (shl X, (sub bw-1, Y)), 0) --> (icmp ne (and X, (shl 1, Y)), 0)
// (icmp sgt (shl X, (sub bw-1, Y)), -1) --> (icmp eq (and X, (shl 1, Y)), 0)
Value *Y;
if (Shl->hasOneUse() &&
((Pred == ICmpInst::ICMP_SLT && C.isZero()) ||
(Pred == ICmpInst::ICMP_SGT && C.isAllOnes())) &&
match(Shl->getOperand(1),
m_OneUse(m_Sub(m_SpecificInt(TypeBits - 1), m_Value(Y)))))
return new ICmpInst(
Pred == ICmpInst::ICMP_SLT ? ICmpInst::ICMP_NE : ICmpInst::ICMP_EQ,
Builder.CreateAnd(X, Builder.CreateShl(ConstantInt::get(ShType, 1), Y,
"", /*HasNUW=*/true)),
ConstantInt::get(ShType, 0));

const APInt *ShiftAmt;
if (!match(Shl->getOperand(1), m_APInt(ShiftAmt)))
return foldICmpShlOne(Cmp, Shl, C);

// Check that the shift amount is in range. If not, don't perform undefined
// shifts. When the shift is visited, it will be simplified.
unsigned TypeBits = C.getBitWidth();
if (ShiftAmt->uge(TypeBits))
return nullptr;

Value *X = Shl->getOperand(0);
Type *ShType = Shl->getType();

// NSW guarantees that we are only shifting out sign bits from the high bits,
// so we can ASHR the compare constant without needing a mask and eliminate
// the shift.
Expand Down
187 changes: 187 additions & 0 deletions llvm/test/Transforms/InstCombine/icmp-and-shift.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -738,3 +738,190 @@ define i1 @test_const_shr_and_1_ne_0_multi_use_and_negative(i32 %b) {
%ret = and i1 %cmp1, %cmp2
ret i1 %ret
}

define i1 @test_shl_sub_bw_minus_1_slt_0(i32 %a, i32 %b) {
; CHECK-LABEL: @test_shl_sub_bw_minus_1_slt_0(
; CHECK-NEXT: [[TMP1:%.*]] = shl nuw i32 1, [[B:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = and i32 [[TMP1]], [[A:%.*]]
; CHECK-NEXT: [[CMP:%.*]] = icmp ne i32 [[TMP2]], 0
; CHECK-NEXT: ret i1 [[CMP]]
;
%sub = sub i32 31, %b
%shl = shl i32 %a, %sub
%cmp = icmp slt i32 %shl, 0
ret i1 %cmp
}

define i1 @test_const_shl_sub_bw_minus_1_slt_0(i32 %b) {
; CHECK-LABEL: @test_const_shl_sub_bw_minus_1_slt_0(
; CHECK-NEXT: [[TMP1:%.*]] = shl nuw i32 1, [[B:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = and i32 [[TMP1]], 42
; CHECK-NEXT: [[CMP:%.*]] = icmp ne i32 [[TMP2]], 0
; CHECK-NEXT: ret i1 [[CMP]]
;
%sub = sub i32 31, %b
%shl = shl i32 42, %sub
%cmp = icmp slt i32 %shl, 0
ret i1 %cmp
}

define i1 @test_not_shl_sub_bw_minus_1_slt_0(i32 %a, i32 %b) {
; CHECK-LABEL: @test_not_shl_sub_bw_minus_1_slt_0(
; CHECK-NEXT: [[TMP1:%.*]] = shl nuw i32 1, [[B:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = and i32 [[TMP1]], [[A:%.*]]
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP2]], 0
; CHECK-NEXT: ret i1 [[CMP]]
;
%sub = sub i32 31, %b
%shl = shl i32 %a, %sub
%cmp = icmp sge i32 %shl, 0
ret i1 %cmp
}

define i1 @test_shl_nuw_sub_bw_minus_1_slt_0(i32 %a, i32 %b) {
; CHECK-LABEL: @test_shl_nuw_sub_bw_minus_1_slt_0(
; CHECK-NEXT: [[TMP1:%.*]] = shl nuw i32 1, [[B:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = and i32 [[TMP1]], [[A:%.*]]
; CHECK-NEXT: [[CMP:%.*]] = icmp ne i32 [[TMP2]], 0
; CHECK-NEXT: ret i1 [[CMP]]
;
%sub = sub i32 31, %b
%shl = shl nuw i32 %a, %sub
%cmp = icmp slt i32 %shl, 0
ret i1 %cmp
}

define i1 @test_not_const_shl_sub_bw_minus_1_slt_0(i32 %b) {
; CHECK-LABEL: @test_not_const_shl_sub_bw_minus_1_slt_0(
; CHECK-NEXT: [[TMP1:%.*]] = shl nuw i32 1, [[B:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = and i32 [[TMP1]], 42
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP2]], 0
; CHECK-NEXT: ret i1 [[CMP]]
;
%sub = sub i32 31, %b
%shl = shl i32 42, %sub
%cmp = icmp sge i32 %shl, 0
ret i1 %cmp
}

define <8 x i1> @test_shl_sub_bw_minus_1_slt_0_v8i8(<8 x i8> %a, <8 x i8> %b) {
; CHECK-LABEL: @test_shl_sub_bw_minus_1_slt_0_v8i8(
; CHECK-NEXT: [[TMP1:%.*]] = shl nuw <8 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>, [[B:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = and <8 x i8> [[TMP1]], [[A:%.*]]
; CHECK-NEXT: [[CMP:%.*]] = icmp ne <8 x i8> [[TMP2]], zeroinitializer
; CHECK-NEXT: ret <8 x i1> [[CMP]]
;
%sub = sub <8 x i8> <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>, %b
%shl = shl <8 x i8> %a, %sub
%cmp = icmp slt <8 x i8> %shl, zeroinitializer
ret <8 x i1> %cmp
}

define <8 x i1> @test_const_shl_sub_bw_minus_1_slt_0_v8i8_splat(<8 x i8> %b) {
; CHECK-LABEL: @test_const_shl_sub_bw_minus_1_slt_0_v8i8_splat(
; CHECK-NEXT: [[TMP1:%.*]] = shl nuw <8 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>, [[B:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = and <8 x i8> [[TMP1]], <i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 42>
; CHECK-NEXT: [[CMP:%.*]] = icmp ne <8 x i8> [[TMP2]], zeroinitializer
; CHECK-NEXT: ret <8 x i1> [[CMP]]
;
%sub = sub <8 x i8> <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>, %b
%shl = shl <8 x i8> <i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 42>, %sub
%cmp = icmp slt <8 x i8> %shl, zeroinitializer
ret <8 x i1> %cmp
}

define <8 x i1> @test_const_shl_sub_bw_minus_1_slt_0_v8i8_splat_poison_1(<8 x i8> %b) {
; CHECK-LABEL: @test_const_shl_sub_bw_minus_1_slt_0_v8i8_splat_poison_1(
; CHECK-NEXT: [[SUB:%.*]] = sub <8 x i8> <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 poison>, [[B:%.*]]
; CHECK-NEXT: [[SHL:%.*]] = shl <8 x i8> <i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 42>, [[SUB]]
; CHECK-NEXT: [[CMP:%.*]] = icmp slt <8 x i8> [[SHL]], zeroinitializer
; CHECK-NEXT: ret <8 x i1> [[CMP]]
;
%sub = sub <8 x i8> <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 poison>, %b
%shl = shl <8 x i8> <i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 42>, %sub
%cmp = icmp slt <8 x i8> %shl, zeroinitializer
ret <8 x i1> %cmp
}

define <8 x i1> @test_const_shl_sub_bw_minus_1_slt_0_v8i8_splat_poison_2(<8 x i8> %b) {
; CHECK-LABEL: @test_const_shl_sub_bw_minus_1_slt_0_v8i8_splat_poison_2(
; CHECK-NEXT: [[TMP1:%.*]] = shl nuw <8 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>, [[B:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = and <8 x i8> [[TMP1]], <i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 poison>
; CHECK-NEXT: [[CMP:%.*]] = icmp ne <8 x i8> [[TMP2]], zeroinitializer
; CHECK-NEXT: ret <8 x i1> [[CMP]]
;
%sub = sub <8 x i8> <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>, %b
%shl = shl <8 x i8> <i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 42, i8 poison>, %sub
%cmp = icmp slt <8 x i8> %shl, zeroinitializer
ret <8 x i1> %cmp
}

define <8 x i1> @test_const_shl_sub_bw_minus_1_slt_0_v8i8_nonsplat(<8 x i8> %b) {
; CHECK-LABEL: @test_const_shl_sub_bw_minus_1_slt_0_v8i8_nonsplat(
; CHECK-NEXT: [[TMP1:%.*]] = shl nuw <8 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>, [[B:%.*]]
; CHECK-NEXT: [[TMP2:%.*]] = and <8 x i8> [[TMP1]], <i8 42, i8 43, i8 44, i8 45, i8 46, i8 47, i8 48, i8 49>
; CHECK-NEXT: [[CMP:%.*]] = icmp ne <8 x i8> [[TMP2]], zeroinitializer
; CHECK-NEXT: ret <8 x i1> [[CMP]]
;
%sub = sub <8 x i8> <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>, %b
%shl = shl <8 x i8> <i8 42, i8 43, i8 44, i8 45, i8 46, i8 47, i8 48, i8 49>, %sub
%cmp = icmp slt <8 x i8> %shl, zeroinitializer
ret <8 x i1> %cmp
}

define i1 @test_shl_sub_non_bw_minus_1_slt_0_negative(i32 %a, i32 %b) {
; CHECK-LABEL: @test_shl_sub_non_bw_minus_1_slt_0_negative(
; CHECK-NEXT: [[SUB:%.*]] = sub i32 32, [[B:%.*]]
; CHECK-NEXT: [[SHL:%.*]] = shl i32 [[A:%.*]], [[SUB]]
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[SHL]], 0
; CHECK-NEXT: ret i1 [[CMP]]
;
%sub = sub i32 32, %b
%shl = shl i32 %a, %sub
%cmp = icmp slt i32 %shl, 0
ret i1 %cmp
}

define i1 @test_shl_sub_bw_minus_1_slt_0_i1_negative(i1 %a, i1 %b) {
; CHECK-LABEL: @test_shl_sub_bw_minus_1_slt_0_i1_negative(
; CHECK-NEXT: ret i1 [[A:%.*]]
;
%sub = sub i1 0, %b
%shl = shl i1 %a, %sub
%cmp = icmp slt i1 %shl, 0
ret i1 %cmp
}

define i1 @test_shl_sub_bw_minus_1_slt_0_multi_use_sub_negative(i32 %a, i32 %b) {
; CHECK-LABEL: @test_shl_sub_bw_minus_1_slt_0_multi_use_sub_negative(
; CHECK-NEXT: [[SUB:%.*]] = sub i32 31, [[B:%.*]]
; CHECK-NEXT: [[SHL:%.*]] = shl i32 [[A:%.*]], [[SUB]]
; CHECK-NEXT: [[CMP1:%.*]] = icmp slt i32 [[SHL]], 0
; CHECK-NEXT: [[CMP2:%.*]] = icmp sgt i32 [[SUB]], [[B]]
; CHECK-NEXT: [[RET:%.*]] = or i1 [[CMP1]], [[CMP2]]
; CHECK-NEXT: ret i1 [[RET]]
;
%sub = sub i32 31, %b
%shl = shl i32 %a, %sub
%cmp1 = icmp slt i32 %shl, 0
%cmp2 = icmp slt i32 %b, %sub
%ret = or i1 %cmp1, %cmp2
ret i1 %ret
}

define i1 @test_shl_sub_bw_minus_1_slt_0_multi_use_shl_negative(i32 %a, i32 %b) {
; CHECK-LABEL: @test_shl_sub_bw_minus_1_slt_0_multi_use_shl_negative(
; CHECK-NEXT: [[SUB:%.*]] = sub i32 31, [[B:%.*]]
; CHECK-NEXT: [[SHL:%.*]] = shl i32 [[A:%.*]], [[SUB]]
; CHECK-NEXT: [[CMP1:%.*]] = icmp slt i32 [[SHL]], 0
; CHECK-NEXT: [[CMP2:%.*]] = icmp eq i32 [[SHL]], [[B]]
; CHECK-NEXT: [[RET:%.*]] = and i1 [[CMP1]], [[CMP2]]
; CHECK-NEXT: ret i1 [[RET]]
;
%sub = sub i32 31, %b
%shl = shl i32 %a, %sub
%cmp1 = icmp slt i32 %shl, 0
%cmp2 = icmp eq i32 %b, %shl
%ret = and i1 %cmp1, %cmp2
ret i1 %ret
}
Loading