InstCombine/Select: remove redundant code (NFC) #112388
Merged
Conversation
This file contains hidden or 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
InstCombinerImpl::foldSelectInstWithICmp has some inlined code for select-icmp-xor simplification, but this simplification is already done by other code. Cover the cases that it claims to simplify, and demonstrate that stripping it doesn't cause test changes.
|
@llvm/pr-subscribers-llvm-transforms Author: Ramkumar Ramachandra (artagnon) ChangesInstCombinerImpl::foldSelectInstWithICmp has some inlined code for select-icmp-xor simplification, but this simplification is already done by other code. Cover the cases that it claims to simplify, and demonstrate that stripping it doesn't cause test changes. Full diff: https://github.com/llvm/llvm-project/pull/112388.diff 2 Files Affected:
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
index 820d3608c8dc49..39a6e8e191eb28 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -1953,56 +1953,6 @@ Instruction *InstCombinerImpl::foldSelectInstWithICmp(SelectInst &SI,
return &SI;
}
- // FIXME: This code is nearly duplicated in InstSimplify. Using/refactoring
- // decomposeBitTestICmp() might help.
- if (TrueVal->getType()->isIntOrIntVectorTy()) {
- unsigned BitWidth =
- DL.getTypeSizeInBits(TrueVal->getType()->getScalarType());
- APInt MinSignedValue = APInt::getSignedMinValue(BitWidth);
- Value *X;
- const APInt *Y, *C;
- bool TrueWhenUnset;
- bool IsBitTest = false;
- if (ICmpInst::isEquality(Pred) &&
- match(CmpLHS, m_And(m_Value(X), m_Power2(Y))) &&
- match(CmpRHS, m_Zero())) {
- IsBitTest = true;
- TrueWhenUnset = Pred == ICmpInst::ICMP_EQ;
- } else if (Pred == ICmpInst::ICMP_SLT && match(CmpRHS, m_Zero())) {
- X = CmpLHS;
- Y = &MinSignedValue;
- IsBitTest = true;
- TrueWhenUnset = false;
- } else if (Pred == ICmpInst::ICMP_SGT && match(CmpRHS, m_AllOnes())) {
- X = CmpLHS;
- Y = &MinSignedValue;
- IsBitTest = true;
- TrueWhenUnset = true;
- }
- if (IsBitTest) {
- Value *V = nullptr;
- // (X & Y) == 0 ? X : X ^ Y --> X & ~Y
- if (TrueWhenUnset && TrueVal == X &&
- match(FalseVal, m_Xor(m_Specific(X), m_APInt(C))) && *Y == *C)
- V = Builder.CreateAnd(X, ~(*Y));
- // (X & Y) != 0 ? X ^ Y : X --> X & ~Y
- else if (!TrueWhenUnset && FalseVal == X &&
- match(TrueVal, m_Xor(m_Specific(X), m_APInt(C))) && *Y == *C)
- V = Builder.CreateAnd(X, ~(*Y));
- // (X & Y) == 0 ? X ^ Y : X --> X | Y
- else if (TrueWhenUnset && FalseVal == X &&
- match(TrueVal, m_Xor(m_Specific(X), m_APInt(C))) && *Y == *C)
- V = Builder.CreateOr(X, *Y);
- // (X & Y) != 0 ? X : X ^ Y --> X | Y
- else if (!TrueWhenUnset && TrueVal == X &&
- match(FalseVal, m_Xor(m_Specific(X), m_APInt(C))) && *Y == *C)
- V = Builder.CreateOr(X, *Y);
-
- if (V)
- return replaceInstUsesWith(SI, V);
- }
- }
-
if (Instruction *V =
foldSelectICmpAndAnd(SI.getType(), ICI, TrueVal, FalseVal, Builder))
return V;
diff --git a/llvm/test/Transforms/InstCombine/select-icmp-xor.ll b/llvm/test/Transforms/InstCombine/select-icmp-xor.ll
new file mode 100644
index 00000000000000..c8ce114a683eee
--- /dev/null
+++ b/llvm/test/Transforms/InstCombine/select-icmp-xor.ll
@@ -0,0 +1,190 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -passes=instcombine -S %s | FileCheck %s
+
+define i8 @select_icmp_eq_pow2(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_eq_pow2(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[SEL:%.*]] = and i8 [[X]], -5
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %and = and i8 %x, 4
+ %icmp = icmp eq i8 %and, 0
+ %xor = xor i8 %x, 4
+ %sel = select i1 %icmp, i8 %x, i8 %xor
+ ret i8 %sel
+}
+
+define i8 @select_icmp_eq_pow2_flipped(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_eq_pow2_flipped(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[SEL:%.*]] = or i8 [[X]], 4
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %and = and i8 %x, 4
+ %icmp = icmp eq i8 %and, 0
+ %xor = xor i8 %x, 4
+ %sel = select i1 %icmp, i8 %xor, i8 %x
+ ret i8 %sel
+}
+
+define i8 @select_icmp_eq_not_pow2(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_eq_not_pow2(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[AND:%.*]] = and i8 [[X]], 5
+; CHECK-NEXT: [[ICMP:%.*]] = icmp eq i8 [[AND]], 0
+; CHECK-NEXT: [[XOR:%.*]] = xor i8 [[X]], 5
+; CHECK-NEXT: [[SEL:%.*]] = select i1 [[ICMP]], i8 [[X]], i8 [[XOR]]
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %and = and i8 %x, 5
+ %icmp = icmp eq i8 %and, 0
+ %xor = xor i8 %x, 5
+ %sel = select i1 %icmp, i8 %x, i8 %xor
+ ret i8 %sel
+}
+
+define i8 @select_icmp_ne_pow2(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_ne_pow2(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[SEL:%.*]] = and i8 [[X]], -5
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %and = and i8 %x, 4
+ %icmp = icmp ne i8 %and, 0
+ %xor = xor i8 %x, 4
+ %sel = select i1 %icmp, i8 %xor, i8 %x
+ ret i8 %sel
+}
+
+define i8 @select_icmp_ne_pow2_flipped(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_ne_pow2_flipped(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[SEL:%.*]] = or i8 [[X]], 4
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %and = and i8 %x, 4
+ %icmp = icmp ne i8 %and, 0
+ %xor = xor i8 %x, 4
+ %sel = select i1 %icmp, i8 %x, i8 %xor
+ ret i8 %sel
+}
+
+define i8 @select_icmp_ne_not_pow2(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_ne_not_pow2(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[AND:%.*]] = and i8 [[X]], 5
+; CHECK-NEXT: [[ICMP_NOT:%.*]] = icmp eq i8 [[AND]], 0
+; CHECK-NEXT: [[XOR:%.*]] = xor i8 [[X]], 5
+; CHECK-NEXT: [[SEL:%.*]] = select i1 [[ICMP_NOT]], i8 [[X]], i8 [[XOR]]
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %and = and i8 %x, 5
+ %icmp = icmp ne i8 %and, 0
+ %xor = xor i8 %x, 5
+ %sel = select i1 %icmp, i8 %xor, i8 %x
+ ret i8 %sel
+}
+
+define i8 @select_icmp_slt_zero_smin(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_slt_zero_smin(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[SEL:%.*]] = or i8 [[X]], -128
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %icmp = icmp slt i8 %x, 0
+ %xor = xor i8 %x, -128
+ %sel = select i1 %icmp, i8 %x, i8 %xor
+ ret i8 %sel
+}
+
+define i8 @select_icmp_slt_zero_smin_flipped(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_slt_zero_smin_flipped(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[SEL:%.*]] = and i8 [[X]], 127
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %icmp = icmp slt i8 %x, 0
+ %xor = xor i8 %x, -128
+ %sel = select i1 %icmp, i8 %xor, i8 %x
+ ret i8 %sel
+}
+
+define i8 @select_icmp_slt_not_zero(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_slt_not_zero(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[ICMP:%.*]] = icmp slt i8 [[X]], 1
+; CHECK-NEXT: [[XOR:%.*]] = xor i8 [[X]], -128
+; CHECK-NEXT: [[SEL:%.*]] = select i1 [[ICMP]], i8 [[X]], i8 [[XOR]]
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %icmp = icmp slt i8 %x, 1
+ %xor = xor i8 %x, -128
+ %sel = select i1 %icmp, i8 %x, i8 %xor
+ ret i8 %sel
+}
+
+define i8 @select_icmp_slt_not_smin(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_slt_not_smin(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[ICMP:%.*]] = icmp slt i8 [[X]], 0
+; CHECK-NEXT: [[XOR:%.*]] = xor i8 [[X]], -127
+; CHECK-NEXT: [[SEL:%.*]] = select i1 [[ICMP]], i8 [[X]], i8 [[XOR]]
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %icmp = icmp slt i8 %x, 0
+ %xor = xor i8 %x, -127
+ %sel = select i1 %icmp, i8 %x, i8 %xor
+ ret i8 %sel
+}
+
+define i8 @select_icmp_sgt_allones_smin(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_sgt_allones_smin(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[SEL:%.*]] = and i8 [[X]], 127
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %icmp = icmp sgt i8 %x, 255
+ %xor = xor i8 %x, -128
+ %sel = select i1 %icmp, i8 %x, i8 %xor
+ ret i8 %sel
+}
+
+define i8 @select_icmp_sgt_allones_smin_flipped(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_sgt_allones_smin_flipped(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[SEL:%.*]] = or i8 [[X]], -128
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %icmp = icmp sgt i8 %x, 255
+ %xor = xor i8 %x, -128
+ %sel = select i1 %icmp, i8 %xor, i8 %x
+ ret i8 %sel
+}
+
+define i8 @select_icmp_sgt_not_allones(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_sgt_not_allones(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[ICMP:%.*]] = icmp sgt i8 [[X]], -2
+; CHECK-NEXT: [[XOR:%.*]] = xor i8 [[X]], -128
+; CHECK-NEXT: [[SEL:%.*]] = select i1 [[ICMP]], i8 [[X]], i8 [[XOR]]
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %icmp = icmp sgt i8 %x, 254
+ %xor = xor i8 %x, -128
+ %sel = select i1 %icmp, i8 %x, i8 %xor
+ ret i8 %sel
+}
+
+define i8 @select_icmp_sgt_not_smin(i8 %x) {
+; CHECK-LABEL: define i8 @select_icmp_sgt_not_smin(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT: [[XOR:%.*]] = xor i8 [[X]], -127
+; CHECK-NEXT: [[ICMP1:%.*]] = icmp slt i8 [[X]], 0
+; CHECK-NEXT: [[SEL:%.*]] = select i1 [[ICMP1]], i8 [[XOR]], i8 [[X]]
+; CHECK-NEXT: ret i8 [[SEL]]
+;
+ %icmp = icmp sgt i8 %x, 255
+ %xor = xor i8 %x, -127
+ %sel = select i1 %icmp, i8 %x, i8 %xor
+ ret i8 %sel
+}
|
dtcxzyw
reviewed
Oct 15, 2024
There was a problem hiding this comment.
FYI this piece of code still has (small) code coverage on my dataset.
See also https://dtcxzyw.github.io/llvm-opt-benchmark/coverage/data/zyw/opt/actions-runner/_work/llvm-opt-benchmark/llvm-opt-benchmark/llvm/llvm-project/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp.html#L1961.
Of course we can just drop them/make the logic simpler :)
This was referenced Oct 15, 2024
I don't doubt that the code has coverage, but I think a later and more powerful simplification subsumes it. |
|
Could you please point to which simplification is the subsuming one? Is it foldSelectICmpEq? |
I'm not sure: I just realized that the code wasn't covered by tests, so I wrote the tests covering it; I removed the code and the tests were still passing. Will investigate more in the morning. |
dtcxzyw
approved these changes
Oct 16, 2024
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.
InstCombinerImpl::foldSelectInstWithICmp has some inlined code for select-icmp-xor simplification, but this simplification is already done by other code, via another path:
(X & Y) == 0 ? X : X ^ Y ->
((X & Y) == 0 ? 0 : Y) ^ X ->
(X & Y) ^ X ->
X & ~Y
Cover the cases that it claims to simplify, and demonstrate that stripping it doesn't cause test changes.