[SCEV] Fold ((-1 * C1) * D / C1) -> -1 * D. #157555

fhahn · 2025-09-08T21:09:29Z

Treat negative constants C as -1 * abs(C1) when folding multiplies and udivs.

Alive2 Proof: https://alive2.llvm.org/ce/z/bdj9W2

llvmbot · 2025-09-08T21:10:01Z

@llvm/pr-subscribers-llvm-analysis

Author: Florian Hahn (fhahn)

Changes

Treat negative constants C as -1 * abs(C1) when folding multiplies and udivs.

Alive2 Proof: https://alive2.llvm.org/ce/z/bdj9W2

Full diff: https://github.com/llvm/llvm-project/pull/157555.diff

2 Files Affected:

(modified) llvm/lib/Analysis/ScalarEvolution.cpp (+9-4)
(modified) llvm/test/Analysis/ScalarEvolution/mul-udiv-folds.ll (+1-1)

diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index bd57d1192eb94..93f20bedc75e3 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -3219,10 +3219,15 @@ const SCEV *ScalarEvolution::getMulExpr(SmallVectorImpl<const SCEV *> &Ops,
       // Try to fold (C * D /u C) -> D, if C is a power-of-2 and D is a multiple
       //  of C.
       const SCEV *D;
-      if (match(Ops[1], m_scev_UDiv(m_SCEV(D), m_scev_Specific(LHSC))) &&
-          LHSC->getAPInt().isPowerOf2() &&
-          LHSC->getAPInt().logBase2() <= getMinTrailingZeros(D)) {
-        return D;
+      APInt LHSV = LHSC->getAPInt();
+      auto *C2 =
+          cast<SCEVConstant>(LHSV.isNegative() && !LHSV.isMinSignedValue()
+                                 ? getAbsExpr(LHSC, false)
+                                 : LHSC);
+      if (C2 && match(Ops[1], m_scev_UDiv(m_SCEV(D), m_scev_Specific(C2))) &&
+          C2->getAPInt().isPowerOf2() &&
+          C2->getAPInt().logBase2() <= getMinTrailingZeros(D)) {
+        return C2 == LHSC ? D : getNegativeSCEV(D);
       }
     }
   }
diff --git a/llvm/test/Analysis/ScalarEvolution/mul-udiv-folds.ll b/llvm/test/Analysis/ScalarEvolution/mul-udiv-folds.ll
index b5d976ba3acda..9f5ed91fbec10 100644
--- a/llvm/test/Analysis/ScalarEvolution/mul-udiv-folds.ll
+++ b/llvm/test/Analysis/ScalarEvolution/mul-udiv-folds.ll
@@ -23,7 +23,7 @@ define void @udiv4_and_udiv2(i1 %c, ptr %A) {
 ; CHECK-NEXT:    %gep.16 = getelementptr i16, ptr %A, i64 %iv
 ; CHECK-NEXT:    --> {((2 * ((zext i32 %start to i64) /u 4))<nuw><nsw> + %A),+,2}<%loop> U: full-set S: full-set Exits: ((zext i32 %start to i64) + %A) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %gep.32 = getelementptr i32, ptr %A, i64 %iv
-; CHECK-NEXT:    --> {((zext i32 %start to i64) + %A),+,4}<%loop> U: full-set S: full-set Exits: ((zext i32 %start to i64) + (4 * ((zext i32 %start to i64) /u 2))<nuw><nsw> + (-4 * ((zext i32 %start to i64) /u 4))<nsw> + %A) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {((zext i32 %start to i64) + %A),+,4}<%loop> U: full-set S: full-set Exits: ((4 * ((zext i32 %start to i64) /u 2))<nuw><nsw> + %A) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %gep.40 = getelementptr <{ i32, i8 }>, ptr %A, i64 %iv
 ; CHECK-NEXT:    --> {((5 * ((zext i32 %start to i64) /u 4))<nuw><nsw> + %A),+,5}<%loop> U: full-set S: full-set Exits: ((5 * ((zext i32 %start to i64) /u 2))<nuw><nsw> + %A) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %gep.48 = getelementptr <{ i32, i16 }>, ptr %A, i64 %iv

Treat negative constants C as -1 * abs(C1) when folding multiplies and udivs. Alive2 Proof: https://alive2.llvm.org/ce/z/bdj9W2

preames · 2025-09-09T15:32:04Z

llvm/lib/Analysis/ScalarEvolution.cpp

+      // D is a multiple of C2, and C1 is a multiple of C2.
      const SCEV *D;
+      APInt C1V = LHSC->getAPInt();
+      // If C1 is negative, try (-1 * abs(C1)) instead.


Adjust the comment to something like:
(C1 * D /u C2) == -1 * -C1 * D /u C2 when C1 != INT_MIN

Though maybe that's not quite right because that equality still holds (just not in a useful way) when C1 is INT_MIN.

Basically, we need something here with more of indication of why it's correct.

Updated, thanks!

preames

LGTM w/prior comment addressed.

nikic · 2025-09-10T08:46:58Z

llvm/lib/Analysis/ScalarEvolution.cpp

      const SCEV *D;
+      APInt C1V = LHSC->getAPInt();
+      // (C1 * D /u C2) == -1 * -C1 * D /u C2 when C1 != INT_MIN.
+      if (C1V.isNegative() && !C1V.isMinSignedValue())


Why is the INT_MIN check here needed? Isn't abs() just a no-op for that?

~~Ah, I guess this avoid doing an unnecessary getNegativeSCEV() lateron.~~ No, that's based on the equality comparison, so it shouldn't matter if we do an extra abs() here.

Yeah it's a no-op, we would end up with (-INT_MIN * D /u C2) == (-1 * -INT_MIN * D /u C2). Can change to use abs() if you prefer and remove the check?

I'll leave it as-is for now, when this gets generalized to non-constants we can adjust it

Treat negative constants C as -1 * abs(C1) when folding multiplies and udivs. Alive2 Proof: https://alive2.llvm.org/ce/z/bdj9W2 PR: llvm/llvm-project#157555

If C2 >u C1 and C1 >u 1, fold to A /u (C2 /u C1). Depends on llvm#157555. Alive2 Proof: https://alive2.llvm.org/ce/z/BWvQYN

If C2 >u C1 and C1 >u 1, fold to A /u (C2 /u C1). Depends on #157555. Alive2 Proof: https://alive2.llvm.org/ce/z/BWvQYN PR: #157656

… (#157656) If C2 >u C1 and C1 >u 1, fold to A /u (C2 /u C1). Depends on llvm/llvm-project#157555. Alive2 Proof: https://alive2.llvm.org/ce/z/BWvQYN PR: llvm/llvm-project#157656

…158328) This reverts commit fd58f23. The recommit contains an extra check to make sure that D is a multiple of C2, if C2 > C1. This fixes the issue causing the revert fd58f23. Tests have been added in 6a726e9. Original message: If C2 >u C1 and C1 >u 1, fold to A /u (C2 /u C1). Depends on #157555. Alive2 Proof: https://alive2.llvm.org/ce/z/BWvQYN PR: #157656

… C2 > C1." (#158328) This reverts commit fd58f23. The recommit contains an extra check to make sure that D is a multiple of C2, if C2 > C1. This fixes the issue causing the revert fd58f23. Tests have been added in 6a726e9. Original message: If C2 >u C1 and C1 >u 1, fold to A /u (C2 /u C1). Depends on llvm/llvm-project#157555. Alive2 Proof: https://alive2.llvm.org/ce/z/BWvQYN PR: llvm/llvm-project#157656

fhahn requested review from nikic, preames and efriedma-quic September 8, 2025 21:09

llvmbot added the llvm:analysis Includes value tracking, cost tables and constant folding label Sep 8, 2025

fhahn mentioned this pull request Sep 8, 2025

[SCEV] Generalize (C * A /u C) -> A fold to (C1 * A /u C2) -> C1/C2 * A. #157159

Merged

[SCEV] Fold ((-1 * C1) * D / C1) -> -1 * D.

b635047

Treat negative constants C as -1 * abs(C1) when folding multiplies and udivs. Alive2 Proof: https://alive2.llvm.org/ce/z/bdj9W2

fhahn force-pushed the scev-mul-div-neg branch from ec99aa1 to b635047 Compare September 9, 2025 10:31

fhahn mentioned this pull request Sep 9, 2025

[SCEV] Fold (C1 * A /u C2) -> A /u (C2 /u C1), if C2 > C1. #157656

Merged

preames reviewed Sep 9, 2025

View reviewed changes

preames approved these changes Sep 9, 2025

View reviewed changes

fhahn added 2 commits September 9, 2025 22:12

!fixup adjust comment, thanks

0c1705c

Merge branch 'main' into scev-mul-div-neg

44844c0

fhahn enabled auto-merge (squash) September 10, 2025 08:45

nikic reviewed Sep 10, 2025

View reviewed changes

fhahn disabled auto-merge September 10, 2025 09:04

Merge branch 'main' into scev-mul-div-neg

ecc0ad5

fhahn enabled auto-merge (squash) September 10, 2025 13:24

fhahn merged commit 6580c91 into llvm:main Sep 10, 2025
9 checks passed

fhahn deleted the scev-mul-div-neg branch September 10, 2025 14:17

fhahn added a commit to fhahn/llvm-project that referenced this pull request Sep 10, 2025

[SCEV] Fold (C1 * A /u C2) -> A /u (C2 /u C1), if C2 > C1.

9780db8

If C2 >u C1 and C1 >u 1, fold to A /u (C2 /u C1). Depends on llvm#157555. Alive2 Proof: https://alive2.llvm.org/ce/z/BWvQYN

fhahn added a commit that referenced this pull request Sep 11, 2025

[SCEV] Fold (C1 * A /u C2) -> A /u (C2 /u C1), if C2 > C1. (#157656)

70012fd

If C2 >u C1 and C1 >u 1, fold to A /u (C2 /u C1). Depends on #157555. Alive2 Proof: https://alive2.llvm.org/ce/z/BWvQYN PR: #157656

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[SCEV] Fold ((-1 * C1) * D / C1) -> -1 * D. #157555

[SCEV] Fold ((-1 * C1) * D / C1) -> -1 * D. #157555

Uh oh!

fhahn commented Sep 8, 2025

Uh oh!

llvmbot commented Sep 8, 2025

Uh oh!

preames Sep 9, 2025

Uh oh!

fhahn Sep 9, 2025

Uh oh!

preames left a comment

Uh oh!

nikic Sep 10, 2025

Uh oh!

nikic Sep 10, 2025 •

edited

Loading

Uh oh!

fhahn Sep 10, 2025

Uh oh!

fhahn Sep 10, 2025

Uh oh!

Uh oh!

Uh oh!

[SCEV] Fold ((-1 * C1) * D / C1) -> -1 * D. #157555

[SCEV] Fold ((-1 * C1) * D / C1) -> -1 * D. #157555

Uh oh!

Conversation

fhahn commented Sep 8, 2025

Uh oh!

llvmbot commented Sep 8, 2025

Uh oh!

preames Sep 9, 2025

Choose a reason for hiding this comment

Uh oh!

fhahn Sep 9, 2025

Choose a reason for hiding this comment

Uh oh!

preames left a comment

Choose a reason for hiding this comment

Uh oh!

nikic Sep 10, 2025

Choose a reason for hiding this comment

Uh oh!

nikic Sep 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fhahn Sep 10, 2025

Choose a reason for hiding this comment

Uh oh!

fhahn Sep 10, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

nikic Sep 10, 2025 •

edited

Loading