Skip to content

[SCEV] Rewrite more SCEVAddExpr when applying guards. #159942

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 4 commits into
base: main
Choose a base branch
from
Open

[SCEV] Rewrite more SCEVAddExpr when applying guards. #159942

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
956297a
[SCEV] Improve handling of divisibility information from loop guards.
fhahn Oct 11, 2025
040a73a
[SCEV] Rewrite more SCEVAddExpr when applying guards.
fhahn Sep 19, 2025
0f40aa4
!fixup getMinsSCEV -> getAddExpr + getMinusOne.
fhahn Oct 9, 2025
0710344
!fixup undo test changes, remove stray period.
fhahn Oct 11, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions llvm/include/llvm/Analysis/ScalarEvolution.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -763,6 +763,10 @@ class ScalarEvolution {
getUMinFromMismatchedTypes(SmallVectorImpl<const SCEV *> &Ops,
bool Sequential = false);

/// Try to match the pattern generated by getURemExpr(A, B). If successful,
/// Assign A and B to LHS and RHS, respectively.
LLVM_ABI bool matchURem(const SCEV *Expr, const SCEV *&LHS, const SCEV *&RHS);

/// Transitively follow the chain of pointer-type operands until reaching a
/// SCEV that does not have a single pointer operand. This returns a
/// SCEVUnknown pointer for well-formed pointer-type expressions, but corner
Expand Down Expand Up @@ -2316,10 +2320,6 @@ class ScalarEvolution {
/// an add rec on said loop.
void getUsedLoops(const SCEV *S, SmallPtrSetImpl<const Loop *> &LoopsUsed);

/// Try to match the pattern generated by getURemExpr(A, B). If successful,
/// Assign A and B to LHS and RHS, respectively.
LLVM_ABI bool matchURem(const SCEV *Expr, const SCEV *&LHS, const SCEV *&RHS);

/// Look for a SCEV expression with type `SCEVType` and operands `Ops` in
/// `UniqueSCEVs`. Return if found, else nullptr.
SCEV *findExistingSCEVInCache(SCEVTypes SCEVType, ArrayRef<const SCEV *> Ops);
Expand Down
324 changes: 190 additions & 134 deletions llvm/lib/Analysis/ScalarEvolution.cpp
View file
Open in desktop

Large diffs are not rendered by default.

6 changes: 3 additions & 3 deletions llvm/test/Analysis/ScalarEvolution/backedge-taken-count-guard-info-apply-to-adds.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,9 +33,9 @@ declare void @clobber()
define void @test_add_sub_1_guard(ptr %src, i32 %n) {
; CHECK-LABEL: 'test_add_sub_1_guard'
; CHECK-NEXT: Determining loop execution counts for: @test_add_sub_1_guard
; CHECK-NEXT: Loop %loop: backedge-taken count is (zext i32 (-1 + (%n /u 2))<nsw> to i64)
; CHECK-NEXT: Loop %loop: constant max backedge-taken count is i64 4294967295
; CHECK-NEXT: Loop %loop: symbolic max backedge-taken count is (zext i32 (-1 + (%n /u 2))<nsw> to i64)
; CHECK-NEXT: Loop %loop: backedge-taken count is i64 0
; CHECK-NEXT: Loop %loop: constant max backedge-taken count is i64 0
; CHECK-NEXT: Loop %loop: symbolic max backedge-taken count is i64 0
; CHECK-NEXT: Loop %loop: Trip multiple is 1
;
entry:
Expand Down
12 changes: 6 additions & 6 deletions llvm/test/Analysis/ScalarEvolution/trip-count-minmax.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,12 +102,12 @@ define void @umax(i32 noundef %a, i32 noundef %b) {
; CHECK-NEXT: %cond = select i1 %cmp, i32 %mul, i32 %mul1
; CHECK-NEXT: --> ((2 * %a) umax (4 * %b)) U: [0,-1) S: [-2147483648,2147483647)
; CHECK-NEXT: %i.011 = phi i32 [ %inc, %for.body ], [ 0, %entry ]
; CHECK-NEXT: --> {0,+,1}<nuw><nsw><%for.body> U: [0,-2147483648) S: [0,-2147483648) Exits: (-1 + ((2 * %a) umax (4 * %b))) LoopDispositions: { %for.body: Computable }
; CHECK-NEXT: --> {0,+,1}<nuw><nsw><%for.body> U: [0,2147483647) S: [0,2147483647) Exits: (-1 + ((2 * %a) umax (4 * %b))) LoopDispositions: { %for.body: Computable }
; CHECK-NEXT: %inc = add nuw nsw i32 %i.011, 1
; CHECK-NEXT: --> {1,+,1}<nuw><%for.body> U: [1,-1) S: [1,-1) Exits: ((2 * %a) umax (4 * %b)) LoopDispositions: { %for.body: Computable }
; CHECK-NEXT: --> {1,+,1}<nuw><nsw><%for.body> U: [1,-2147483648) S: [1,-2147483648) Exits: ((2 * %a) umax (4 * %b)) LoopDispositions: { %for.body: Computable }
; CHECK-NEXT: Determining loop execution counts for: @umax
; CHECK-NEXT: Loop %for.body: backedge-taken count is (-1 + ((2 * %a) umax (4 * %b)))
; CHECK-NEXT: Loop %for.body: constant max backedge-taken count is i32 -3
; CHECK-NEXT: Loop %for.body: constant max backedge-taken count is i32 2147483646
; CHECK-NEXT: Loop %for.body: symbolic max backedge-taken count is (-1 + ((2 * %a) umax (4 * %b)))
; CHECK-NEXT: Loop %for.body: Trip multiple is 2
;
Expand Down Expand Up @@ -197,12 +197,12 @@ define void @smax(i32 noundef %a, i32 noundef %b) {
; CHECK-NEXT: %cond = select i1 %cmp, i32 %mul, i32 %mul1
; CHECK-NEXT: --> ((2 * %a)<nsw> smax (4 * %b)<nsw>) U: [0,-1) S: [-2147483648,2147483647)
; CHECK-NEXT: %i.011 = phi i32 [ %inc, %for.body ], [ 0, %entry ]
; CHECK-NEXT: --> {0,+,1}<nuw><nsw><%for.body> U: [0,-2147483648) S: [0,-2147483648) Exits: (-1 + ((2 * %a)<nsw> smax (4 * %b)<nsw>)) LoopDispositions: { %for.body: Computable }
; CHECK-NEXT: --> {0,+,1}<nuw><nsw><%for.body> U: [0,2147483647) S: [0,2147483647) Exits: (-1 + ((2 * %a)<nsw> smax (4 * %b)<nsw>)) LoopDispositions: { %for.body: Computable }
; CHECK-NEXT: %inc = add nuw nsw i32 %i.011, 1
; CHECK-NEXT: --> {1,+,1}<nuw><%for.body> U: [1,-1) S: [1,-1) Exits: ((2 * %a)<nsw> smax (4 * %b)<nsw>) LoopDispositions: { %for.body: Computable }
; CHECK-NEXT: --> {1,+,1}<nuw><nsw><%for.body> U: [1,-2147483648) S: [1,-2147483648) Exits: ((2 * %a)<nsw> smax (4 * %b)<nsw>) LoopDispositions: { %for.body: Computable }
; CHECK-NEXT: Determining loop execution counts for: @smax
; CHECK-NEXT: Loop %for.body: backedge-taken count is (-1 + ((2 * %a)<nsw> smax (4 * %b)<nsw>))
; CHECK-NEXT: Loop %for.body: constant max backedge-taken count is i32 -3
; CHECK-NEXT: Loop %for.body: constant max backedge-taken count is i32 2147483646
; CHECK-NEXT: Loop %for.body: symbolic max backedge-taken count is (-1 + ((2 * %a)<nsw> smax (4 * %b)<nsw>))
; CHECK-NEXT: Loop %for.body: Trip multiple is 2
;
Expand Down
6 changes: 2 additions & 4 deletions llvm/test/Transforms/IndVarSimplify/canonicalize-cmp.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -343,14 +343,13 @@ define void @slt_no_smax_needed(i64 %n, ptr %dst) {
; CHECK-NEXT: [[PRE:%.*]] = icmp ult i32 [[ADD_1]], 8
; CHECK-NEXT: br i1 [[PRE]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[SHR]], i32 1)
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i32 [[IV]]
; CHECK-NEXT: store i8 0, ptr [[GEP]], align 1
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[IV_NEXT]], [[SMAX]]
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[IV_NEXT]], [[SHR]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT]]
Expand Down Expand Up @@ -385,14 +384,13 @@ define void @ult_no_umax_needed(i64 %n, ptr %dst) {
; CHECK-NEXT: [[PRE:%.*]] = icmp ult i32 [[ADD_1]], 8
; CHECK-NEXT: br i1 [[PRE]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[SHR]], i32 1)
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i32 [[IV]]
; CHECK-NEXT: store i8 0, ptr [[GEP]], align 1
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[IV_NEXT]], [[UMAX]]
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[IV_NEXT]], [[SHR]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT]]
Expand Down
5 changes: 2 additions & 3 deletions llvm/test/Transforms/IndVarSimplify/loop-guard-order.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,7 +114,7 @@ define i32 @urem_order1(i32 %n) {
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
; CHECK-NEXT: call void @foo()
; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 3
; CHECK-NEXT: [[IV_NEXT]] = add nuw i32 [[IV]], 3
; CHECK-NEXT: [[EC:%.*]] = icmp eq i32 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT_LOOPEXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT_LOOPEXIT]]:
Expand Down Expand Up @@ -205,13 +205,12 @@ define i64 @test_loop_with_div_order_1(i64 %n) {
; CHECK-NEXT: [[PARITY_CHECK:%.*]] = icmp eq i64 [[IS_ODD]], 0
; CHECK-NEXT: br i1 [[PARITY_CHECK]], label %[[LOOP_PREHEADER:.*]], label %[[EXIT]]
; CHECK: [[LOOP_PREHEADER]]:
; CHECK-NEXT: [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[UPPER_BOUND]], i64 1)
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
; CHECK-NEXT: [[DUMMY:%.*]] = load volatile i64, ptr null, align 8
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[IV_NEXT]], [[UMAX]]
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[IV_NEXT]], [[UPPER_BOUND]]
; CHECK-NEXT: br i1 [[EXITCOND]], label %[[LOOP]], label %[[EXIT_LOOPEXIT:.*]]
; CHECK: [[EXIT_LOOPEXIT]]:
; CHECK-NEXT: br label %[[EXIT]]
Expand Down
6 changes: 3 additions & 3 deletions llvm/test/Transforms/LoopIdiom/add-nsw-zext-fold.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,9 +61,9 @@ define void @test_memset_size_can_use_info_from_guards(i32 %x, ptr %dst) {
; CHECK: [[LOOP1_BACKEDGE]]:
; CHECK-NEXT: br label %[[LOOP1]]
; CHECK: [[LOOP2_PREHEADER]]:
; CHECK-NEXT: [[TMP0:%.*]] = zext i32 [[SUB]] to i64
; CHECK-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 1
; CHECK-NEXT: [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[TMP1]], i64 1)
; CHECK-NEXT: [[TMP0:%.*]] = add nsw i32 [[SHR]], -1
; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; CHECK-NEXT: [[UMAX:%.*]] = add nuw nsw i64 [[TMP1]], 1
Comment on lines +64 to +66
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is missing a fold end-to-end from https://github.com/dtcxzyw/llvm-opt-benchmark/pull/2845/files/ed058c0c689f0dadf267033c74e0960c3593baff#r2367574398

https://alive2.llvm.org/ce/z/BuHtNm

; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 1 [[DST]], i8 0, i64 [[UMAX]], i1 false)
; CHECK-NEXT: br label %[[LOOP2:.*]]
; CHECK: [[LOOP2]]:
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/LoopVectorize/dont-fold-tail-for-divisible-TC.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -193,7 +193,7 @@ define dso_local void @cannotProveAlignedTC(ptr noalias nocapture %A, i32 %p, i3
; CHECK-NEXT: store i32 13, ptr [[TMP12]], align 1
; CHECK-NEXT: br label [[PRED_STORE_CONTINUE6]]
; CHECK: pred.store.continue6:
; CHECK-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
; CHECK-NEXT: [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
Expand Down
7 changes: 1 addition & 6 deletions llvm/test/Transforms/LoopVectorize/runtime-checks-difference.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -465,12 +465,7 @@ define void @remove_diff_checks_via_guards(i32 %x, i32 %y, ptr %A) {
; CHECK-NEXT: [[TMP13:%.*]] = icmp ugt i64 [[SMAX]], 4294967295
; CHECK-NEXT: [[TMP14:%.*]] = or i1 [[TMP12]], [[TMP13]]
; CHECK-NEXT: [[TMP15:%.*]] = or i1 [[TMP9]], [[TMP14]]
; CHECK-NEXT: br i1 [[TMP15]], [[SCALAR_PH]], label %[[VECTOR_MEMCHECK:.*]]
; CHECK: [[VECTOR_MEMCHECK]]:
; CHECK-NEXT: [[TMP16:%.*]] = sext i32 [[OFFSET]] to i64
; CHECK-NEXT: [[TMP17:%.*]] = shl nsw i64 [[TMP16]], 2
; CHECK-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP17]], 16
; CHECK-NEXT: br i1 [[DIFF_CHECK]], [[SCALAR_PH]], [[VECTOR_PH1:label %.*]]
; CHECK-NEXT: br i1 [[TMP15]], [[SCALAR_PH]], [[VECTOR_PH:label %.*]]
;
entry:
%offset = sub i32 %x, %y
Expand Down
Loading