[InstCombine] Fold ((cst << x) & 1) --> x == 0 when cst is odd #79772
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@llvm/pr-subscribers-llvm-transforms Author: None (elhewaty) Changes
Full diff: https://github.com/llvm/llvm-project/pull/79772.diff 2 Files Affected:
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
index 0620752e3213945..e95cf7eaa940623 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
@@ -2220,10 +2220,13 @@ Instruction *InstCombinerImpl::visitAnd(BinaryOperator &I) {
Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
Value *X, *Y;
- if (match(Op0, m_OneUse(m_LogicalShift(m_One(), m_Value(X)))) &&
+ const APInt *C;
+ if ((match(Op0, m_OneUse(m_LShr(m_One(), m_Value(X)))) ||
+ (match(Op0, m_OneUse(m_Shl(m_APInt(C), m_Value(X)))) &&
+ ((*C & 1) == 1))) &&
match(Op1, m_One())) {
- // (1 << X) & 1 --> zext(X == 0)
// (1 >> X) & 1 --> zext(X == 0)
+ // (C << X) & 1 --> zext(X == 0)
Value *IsZero = Builder.CreateICmpEQ(X, ConstantInt::get(Ty, 0));
return new ZExtInst(IsZero, Ty);
}
@@ -2246,7 +2249,6 @@ Instruction *InstCombinerImpl::visitAnd(BinaryOperator &I) {
isKnownToBeAPowerOfTwo(Y, /*OrZero*/ true, /*Depth*/ 0, &I))
return BinaryOperator::CreateAnd(Builder.CreateNot(X), Y);
- const APInt *C;
if (match(Op1, m_APInt(C))) {
const APInt *XorC;
if (match(Op0, m_OneUse(m_Xor(m_Value(X), m_APInt(XorC))))) {
diff --git a/llvm/test/Transforms/InstCombine/and.ll b/llvm/test/Transforms/InstCombine/and.ll
index 2e37fee07cc47bf..a6833fe9a1b9c6e 100644
--- a/llvm/test/Transforms/InstCombine/and.ll
+++ b/llvm/test/Transforms/InstCombine/and.ll
@@ -7,6 +7,68 @@ declare void @use32(i32)
; There should be no 'and' instructions left in any test.
+define i32 @test_with_1(i32 %x) {
+; CHECK-LABEL: @test_with_1(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[X:%.*]], 0
+; CHECK-NEXT: [[AND:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT: ret i32 [[AND]]
+;
+ %shl = shl i32 1, %x
+ %and = and i32 %shl, 1
+ ret i32 %and
+}
+
+define i32 @test_with_3(i32 %x) {
+; CHECK-LABEL: @test_with_3(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[X:%.*]], 0
+; CHECK-NEXT: [[AND:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT: ret i32 [[AND]]
+;
+ %shl = shl i32 3, %x
+ %and = and i32 %shl, 1
+ ret i32 %and
+}
+
+define i32 @test_with_5(i32 %x) {
+; CHECK-LABEL: @test_with_5(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[X:%.*]], 0
+; CHECK-NEXT: [[AND:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT: ret i32 [[AND]]
+;
+ %shl = shl i32 5, %x
+ %and = and i32 %shl, 1
+ ret i32 %and
+}
+
+define i32 @test_with_neg_5(i32 %x) {
+; CHECK-LABEL: @test_with_neg_5(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[X:%.*]], 0
+; CHECK-NEXT: [[AND:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT: ret i32 [[AND]]
+;
+ %shl = shl i32 -5, %x
+ %and = and i32 %shl, 1
+ ret i32 %and
+}
+
+define i32 @test_with_even(i32 %x) {
+; CHECK-LABEL: @test_with_even(
+; CHECK-NEXT: ret i32 0
+;
+ %shl = shl i32 4, %x
+ %and = and i32 %shl, 1
+ ret i32 %and
+}
+
+define i32 @test_with_neg_even(i32 %x) {
+; CHECK-LABEL: @test_with_neg_even(
+; CHECK-NEXT: ret i32 0
+;
+ %shl = shl i32 -4, %x
+ %and = and i32 %shl, 1
+ ret i32 %and
+}
+
define i32 @test1(i32 %A) {
; CHECK-LABEL: @test1(
; CHECK-NEXT: ret i32 0
|
elhewaty
force-pushed
the
instcombine-shl-and
branch
from
96b93c3
to
0f15cdf
Compare
dtcxzyw
added a commit
to dtcxzyw/llvm-opt-benchmark
that referenced
this pull request
Jan 31, 2024
dtcxzyw
approved these changes
Jan 31, 2024
| const APInt *C; | ||
| if ((match(Op0, m_OneUse(m_LogicalShift(m_One(), m_Value(X)))) || | ||
| (match(Op0, m_OneUse(m_Shl(m_APInt(C), m_Value(X)))) && | ||
| C->countr_zero() == 0)) && |
There was a problem hiding this comment.
Suggested change
| C->countr_zero() == 0)) && | |
| (*C)[0])) && |
| } | ||
|
|
||
| ; help wanted: why the optimization does not work on this case | ||
| ; or when vector elements are different, even if they are odd numbers? |
There was a problem hiding this comment.
m_APInt only captures a constant scalar or a constant splat (i.e., all elements are the same).
You can handle this case with KnownBits. But I don't think it is worth doing that.
There was a problem hiding this comment.
@dtcxzyw, why did the build fail?
There was a problem hiding this comment.
@dtcxzyw, why did the build fail?
Don't worry about the build failure on windows. It is not related to your patch.
elhewaty
force-pushed
the
instcombine-shl-and
branch
from
0f15cdf
to
51fd291
Compare
agozillon
pushed a commit
to agozillon/llvm-project
that referenced
this pull request
Feb 5, 2024
…79772) Fold ((cst << x) & 1) to zext(x == 0) when cst is odd. Fixes: llvm#73384 Alive2: https://alive2.llvm.org/ce/z/5RbaK6
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
((cst << x) & 1)-->x == 0when cst is odd #73384