[AArch64] Avoid overflow when using shl lower mul #97148
Merged
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
Transforming `(mul x, -(2^(N-M) - 1) * 2^M)` to `(sub (shl x, M), (shl x, N))` will cause overflow when N is 32 and M is 31.
|
@llvm/pr-subscribers-backend-aarch64 Author: DianQK (DianQK) ChangesFixes #97147. Transforming Full diff: https://github.com/llvm/llvm-project/pull/97148.diff 2 Files Affected:
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 0d53f71a4def8..e022646969b62 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -18059,16 +18059,28 @@ static SDValue performMulCombine(SDNode *N, SelectionDAG &DAG,
unsigned ShiftAmt;
auto Shl = [&](SDValue N0, unsigned N1) {
+ if (!N0.getNode())
+ return SDValue();
SDValue RHS = DAG.getConstant(N1, DL, MVT::i64);
- return DAG.getNode(ISD::SHL, DL, VT, N0, RHS);
+ SDValue SHL = DAG.getNode(ISD::SHL, DL, VT, N0, RHS);
+ // If shift causes overflow, ignore this combine.
+ if (SHL->isUndef())
+ return SDValue();
+ return SHL;
};
auto Add = [&](SDValue N0, SDValue N1) {
+ if (!N0.getNode() || !N1.getNode())
+ return SDValue();
return DAG.getNode(ISD::ADD, DL, VT, N0, N1);
};
auto Sub = [&](SDValue N0, SDValue N1) {
+ if (!N0.getNode() || !N1.getNode())
+ return SDValue();
return DAG.getNode(ISD::SUB, DL, VT, N0, N1);
};
auto Negate = [&](SDValue N) {
+ if (!N0.getNode())
+ return SDValue();
SDValue Zero = DAG.getConstant(0, DL, VT);
return DAG.getNode(ISD::SUB, DL, VT, Zero, N);
};
diff --git a/llvm/test/CodeGen/AArch64/mul_pow2.ll b/llvm/test/CodeGen/AArch64/mul_pow2.ll
index c4839175ded5a..16a47c9a49a05 100644
--- a/llvm/test/CodeGen/AArch64/mul_pow2.ll
+++ b/llvm/test/CodeGen/AArch64/mul_pow2.ll
@@ -992,3 +992,22 @@ define <4 x i32> @muladd_demand_commute(<4 x i32> %x, <4 x i32> %y) {
%r = and <4 x i32> %a, <i32 131071, i32 131071, i32 131071, i32 131071>
ret <4 x i32> %r
}
+
+; Transforming `(mul x, -(2^(N-M) - 1) * 2^M)` to `(sub (shl x, M), (shl x, N))`
+; will cause overflow when N is 32 and M is 31.
+define i32 @shift_overflow(i32 %x) {
+; CHECK-LABEL: shift_overflow:
+; CHECK: // %bb.0:
+; CHECK-NEXT: mov w8, #-2147483648 // =0x80000000
+; CHECK-NEXT: mul w0, w0, w8
+; CHECK-NEXT: ret
+;
+; GISEL-LABEL: shift_overflow:
+; GISEL: // %bb.0:
+; GISEL-NEXT: mov w8, #-2147483648 // =0x80000000
+; GISEL-NEXT: mul w0, w0, w8
+; GISEL-NEXT: ret
+ %const = bitcast i32 2147483648 to i32
+ %r = mul i32 %x, %const
+ ret i32 %r
+}
|
|
LGTM, thanks very much for helping fix this issue. |
davemgreen
reviewed
Jun 29, 2024
I guess I should use |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/154/builds/712 Here is the relevant piece of the build log for the reference: |
DianQK
added a commit
to DianQK/llvm-project
that referenced
this pull request
Jun 30, 2024
Fixes llvm#97147. Transforming `(mul x, -(2^(N-M) - 1) * 2^M)` to `(sub (shl x, M), (shl x, N))` will cause overflow when N is 32 and M is 31. I still added checks for all scenarios, even other scenarios, don't seem to cause overflow. (cherry picked from commit 4a96803)
nikic
pushed a commit
to rust-lang/llvm-project
that referenced
this pull request
Jun 30, 2024
Fixes llvm#97147. Transforming `(mul x, -(2^(N-M) - 1) * 2^M)` to `(sub (shl x, M), (shl x, N))` will cause overflow when N is 32 and M is 31. I still added checks for all scenarios, even other scenarios, don't seem to cause overflow. (cherry picked from commit 4a96803)
lravenclaw
pushed a commit
to lravenclaw/llvm-project
that referenced
this pull request
Jul 3, 2024
Fixes llvm#97147. Transforming `(mul x, -(2^(N-M) - 1) * 2^M)` to `(sub (shl x, M), (shl x, N))` will cause overflow when N is 32 and M is 31. I still added checks for all scenarios, even other scenarios, don't seem to cause overflow.
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.
Fixes #97147.
Transforming
(mul x, -(2^(N-M) - 1) * 2^M)to(sub (shl x, M), (shl x, N))will cause overflow when N is 32 and M is 31. I still added checks for all scenarios, even other scenarios don't seem to cause overflow.