[llvm] r302733 - [InstCombine] remove fold that swaps xor/or with constants; NFCI
Sanjay Patel via llvm-commits
llvm-commits at lists.llvm.org
Wed May 10 14:33:55 PDT 2017
Author: spatel
Date: Wed May 10 16:33:55 2017
New Revision: 302733
URL: http://llvm.org/viewvc/llvm-project?rev=302733&view=rev
Log:
[InstCombine] remove fold that swaps xor/or with constants; NFCI
// (X ^ C1) | C2 --> (X | C2) ^ (C1&~C2)
This canonicalization was added at:
https://reviews.llvm.org/rL7264
By moving xors out/down, we can more easily combine constants. I'm adding
tests that do not change with this patch, so we can verify that those kinds
of transforms are still happening.
This is no-functional-change-intended because there's a later fold:
// (X^C)|Y -> (X|Y)^C iff Y&C == 0
...and demanded-bits appears to guarantee that any fold that would have
hit the fold we're removing here would be caught by that 2nd fold.
Similar reasoning was used in:
https://reviews.llvm.org/rL299384
The larger motivation for removing this code is that it could interfere with
the fix for PR32706:
https://bugs.llvm.org/show_bug.cgi?id=32706
Ie, we're not checking if the 'xor' is actually a 'not', so we could reverse
a 'not' optimization and cause an infinite loop by altering an 'xor X, -1'.
Differential Revision: https://reviews.llvm.org/D33050
Modified:
llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
llvm/trunk/test/Transforms/InstCombine/or-xor.ll
Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp?rev=302733&r1=302732&r2=302733&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp Wed May 10 16:33:55 2017
@@ -1986,18 +1986,6 @@ Instruction *InstCombiner::visitOr(Binar
if (Value *V = SimplifyBSwap(I))
return replaceInstUsesWith(I, V);
- if (ConstantInt *RHS = dyn_cast<ConstantInt>(Op1)) {
- ConstantInt *C1 = nullptr; Value *X = nullptr;
- // (X ^ C1) | C2 --> (X | C2) ^ (C1&~C2)
- if (match(Op0, m_Xor(m_Value(X), m_ConstantInt(C1))) &&
- Op0->hasOneUse()) {
- Value *Or = Builder->CreateOr(X, RHS);
- Or->takeName(Op0);
- return BinaryOperator::CreateXor(Or,
- Builder->getInt(C1->getValue() & ~RHS->getValue()));
- }
- }
-
if (isa<Constant>(Op1))
if (Instruction *FoldedLogic = foldOpWithConstantIntoOperand(I))
return FoldedLogic;
Modified: llvm/trunk/test/Transforms/InstCombine/or-xor.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/or-xor.ll?rev=302733&r1=302732&r2=302733&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/InstCombine/or-xor.ll (original)
+++ llvm/trunk/test/Transforms/InstCombine/or-xor.ll Wed May 10 16:33:55 2017
@@ -230,3 +230,73 @@ define i32 @test16(i32 %a, i32 %b) {
%xor = or i32 %and1, %and2
ret i32 %xor
}
+
+define i8 @not_or(i8 %x) {
+; CHECK-LABEL: @not_or(
+; CHECK-NEXT: [[NOTX:%.*]] = or i8 %x, 7
+; CHECK-NEXT: [[OR:%.*]] = xor i8 [[NOTX]], -8
+; CHECK-NEXT: ret i8 [[OR]]
+;
+ %notx = xor i8 %x, -1
+ %or = or i8 %notx, 7
+ ret i8 %or
+}
+
+define i8 @not_or_xor(i8 %x) {
+; CHECK-LABEL: @not_or_xor(
+; CHECK-NEXT: [[NOTX:%.*]] = or i8 %x, 7
+; CHECK-NEXT: [[XOR:%.*]] = xor i8 [[NOTX]], -12
+; CHECK-NEXT: ret i8 [[XOR]]
+;
+ %notx = xor i8 %x, -1
+ %or = or i8 %notx, 7
+ %xor = xor i8 %or, 12
+ ret i8 %xor
+}
+
+define i8 @xor_or(i8 %x) {
+; CHECK-LABEL: @xor_or(
+; CHECK-NEXT: [[XOR:%.*]] = or i8 %x, 7
+; CHECK-NEXT: [[OR:%.*]] = xor i8 [[XOR]], 32
+; CHECK-NEXT: ret i8 [[OR]]
+;
+ %xor = xor i8 %x, 32
+ %or = or i8 %xor, 7
+ ret i8 %or
+}
+
+define i8 @xor_or2(i8 %x) {
+; CHECK-LABEL: @xor_or2(
+; CHECK-NEXT: [[XOR:%.*]] = or i8 %x, 7
+; CHECK-NEXT: [[OR:%.*]] = xor i8 [[XOR]], 32
+; CHECK-NEXT: ret i8 [[OR]]
+;
+ %xor = xor i8 %x, 33
+ %or = or i8 %xor, 7
+ ret i8 %or
+}
+
+define i8 @xor_or_xor(i8 %x) {
+; CHECK-LABEL: @xor_or_xor(
+; CHECK-NEXT: [[XOR1:%.*]] = or i8 %x, 7
+; CHECK-NEXT: [[XOR2:%.*]] = xor i8 [[XOR1]], 44
+; CHECK-NEXT: ret i8 [[XOR2]]
+;
+ %xor1 = xor i8 %x, 33
+ %or = or i8 %xor1, 7
+ %xor2 = xor i8 %or, 12
+ ret i8 %xor2
+}
+
+define i8 @or_xor_or(i8 %x) {
+; CHECK-LABEL: @or_xor_or(
+; CHECK-NEXT: [[XOR:%.*]] = or i8 %x, 39
+; CHECK-NEXT: [[OR2:%.*]] = xor i8 [[XOR]], 8
+; CHECK-NEXT: ret i8 [[OR2]]
+;
+ %or1 = or i8 %x, 33
+ %xor = xor i8 %or1, 12
+ %or2 = or i8 %xor, 7
+ ret i8 %or2
+}
+
More information about the llvm-commits
mailing list