[SCEV] Use context instruction for SCEVUnknowns in getConstantMultiple. #163260
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
Follow-up to llvm#160941. Even if we don't have a context instruction for the caller, we should be able to provide context instructions for SCEVUnknowns. Unless I am missing something, SCEVUnknown only become available at the point their underlying IR instruction has been defined. If it is an argument, it should be safe to use the first instruction in the entry block or the instruction itself if it wraps an instruction. This allows getConstantMultiple to make better use of alignment assumptions.
llvmbot
added
the
llvm:analysis
|
@llvm/pr-subscribers-llvm-analysis Author: Florian Hahn (fhahn) ChangesFollow-up to #160941. Even if we don't have a context instruction for the caller, we should be able to provide context instructions for SCEVUnknowns. Unless I am missing something, SCEVUnknown only become available at the point their underlying IR instruction has been defined. If it is an argument, it should be safe to use the first instruction in the entry block or the instruction itself if it wraps an instruction. This allows getConstantMultiple to make better use of alignment assumptions. Full diff: https://github.com/llvm/llvm-project/pull/163260.diff 2 Files Affected:
diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index 3fab6b0572cb7..12f58c3f083ac 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -6417,8 +6417,18 @@ APInt ScalarEvolution::getConstantMultipleImpl(const SCEV *S,
case scSequentialUMinExpr:
return GetGCDMultiple(cast<SCEVNAryExpr>(S));
case scUnknown: {
- // ask ValueTracking for known bits
+ // Ask ValueTracking for known bits. SCEVUnknown only become available at
+ // the point their underlying IR instruction has been defined. If CtxI was
+ // not provided, use:
+ // * the first instruction in the entry block if it is an argument
+ // * the instruction itself otherwise.
const SCEVUnknown *U = cast<SCEVUnknown>(S);
+ if (!CtxI) {
+ if (isa<Argument>(U->getValue()))
+ CtxI = &*F.getEntryBlock().begin();
+ else if (auto *I = dyn_cast<Instruction>(U->getValue()))
+ CtxI = I;
+ }
unsigned Known =
computeKnownBits(U->getValue(), getDataLayout(), &AC, CtxI, &DT)
.countMinTrailingZeros();
diff --git a/llvm/test/Analysis/ScalarEvolution/trip-multiple-guard-info.ll b/llvm/test/Analysis/ScalarEvolution/trip-multiple-guard-info.ll
index 7ba422da79ad8..a477465cb0699 100644
--- a/llvm/test/Analysis/ScalarEvolution/trip-multiple-guard-info.ll
+++ b/llvm/test/Analysis/ScalarEvolution/trip-multiple-guard-info.ll
@@ -578,22 +578,22 @@ define void @test_ptr_aligned_by_2_and_4_via_assumption(ptr %start, ptr %end) {
; CHECK-LABEL: 'test_ptr_aligned_by_2_and_4_via_assumption'
; CHECK-NEXT: Classifying expressions for: @test_ptr_aligned_by_2_and_4_via_assumption
; CHECK-NEXT: %iv = phi ptr [ %start, %entry ], [ %iv.next, %loop ]
-; CHECK-NEXT: --> {%start,+,4}<%loop> U: full-set S: full-set Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT: --> {%start,+,4}<%loop> U: [0,-1) S: [-9223372036854775808,9223372036854775807) Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
; CHECK-NEXT: %iv.next = getelementptr i8, ptr %iv, i64 4
-; CHECK-NEXT: --> {(4 + %start),+,4}<%loop> U: full-set S: full-set Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT: --> {(4 + %start),+,4}<%loop> U: [0,-1) S: [-9223372036854775808,9223372036854775807) Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
; CHECK-NEXT: Determining loop execution counts for: @test_ptr_aligned_by_2_and_4_via_assumption
; CHECK-NEXT: Loop %loop: Unpredictable backedge-taken count.
; CHECK-NEXT: Loop %loop: Unpredictable constant max backedge-taken count.
; CHECK-NEXT: Loop %loop: Unpredictable symbolic max backedge-taken count.
; CHECK-NEXT: Loop %loop: Predicated backedge-taken count is ((-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64)) /u 4)
; CHECK-NEXT: Predicates:
-; CHECK-NEXT: Equal predicate: (zext i2 ((trunc i64 (ptrtoint ptr %end to i64) to i2) + (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2))) to i64) == 0
+; CHECK-NEXT: Equal predicate: (zext i2 (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2)) to i64) == 0
; CHECK-NEXT: Loop %loop: Predicated constant max backedge-taken count is i64 4611686018427387903
; CHECK-NEXT: Predicates:
-; CHECK-NEXT: Equal predicate: (zext i2 ((trunc i64 (ptrtoint ptr %end to i64) to i2) + (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2))) to i64) == 0
+; CHECK-NEXT: Equal predicate: (zext i2 (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2)) to i64) == 0
; CHECK-NEXT: Loop %loop: Predicated symbolic max backedge-taken count is ((-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64)) /u 4)
; CHECK-NEXT: Predicates:
-; CHECK-NEXT: Equal predicate: (zext i2 ((trunc i64 (ptrtoint ptr %end to i64) to i2) + (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2))) to i64) == 0
+; CHECK-NEXT: Equal predicate: (zext i2 (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2)) to i64) == 0
;
entry:
call void @llvm.assume(i1 true) [ "align"(ptr %start, i64 2) ]
@@ -615,9 +615,9 @@ define void @test_ptrs_aligned_by_4_via_assumption(ptr %start, ptr %end) {
; CHECK-LABEL: 'test_ptrs_aligned_by_4_via_assumption'
; CHECK-NEXT: Classifying expressions for: @test_ptrs_aligned_by_4_via_assumption
; CHECK-NEXT: %iv = phi ptr [ %start, %entry ], [ %iv.next, %loop ]
-; CHECK-NEXT: --> {%start,+,4}<%loop> U: full-set S: full-set Exits: ((4 * ((-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64)) /u 4))<nuw> + %start) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT: --> {%start,+,4}<%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: (-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64) + %start) LoopDispositions: { %loop: Computable }
; CHECK-NEXT: %iv.next = getelementptr i8, ptr %iv, i64 4
-; CHECK-NEXT: --> {(4 + %start),+,4}<%loop> U: full-set S: full-set Exits: (4 + (4 * ((-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64)) /u 4))<nuw> + %start) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT: --> {(4 + %start),+,4}<%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: ((-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64) + %start) LoopDispositions: { %loop: Computable }
; CHECK-NEXT: Determining loop execution counts for: @test_ptrs_aligned_by_4_via_assumption
; CHECK-NEXT: Loop %loop: backedge-taken count is ((-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64)) /u 4)
; CHECK-NEXT: Loop %loop: constant max backedge-taken count is i64 4611686018427387903
@@ -644,9 +644,9 @@ define void @test_ptrs_aligned_by_8_via_assumption(ptr %start, ptr %end) {
; CHECK-LABEL: 'test_ptrs_aligned_by_8_via_assumption'
; CHECK-NEXT: Classifying expressions for: @test_ptrs_aligned_by_8_via_assumption
; CHECK-NEXT: %iv = phi ptr [ %start, %entry ], [ %iv.next, %loop ]
-; CHECK-NEXT: --> {%start,+,4}<%loop> U: full-set S: full-set Exits: ((4 * ((-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64)) /u 4))<nuw> + %start) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT: --> {%start,+,4}<%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: (-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64) + %start) LoopDispositions: { %loop: Computable }
; CHECK-NEXT: %iv.next = getelementptr i8, ptr %iv, i64 4
-; CHECK-NEXT: --> {(4 + %start),+,4}<%loop> U: full-set S: full-set Exits: (4 + (4 * ((-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64)) /u 4))<nuw> + %start) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT: --> {(4 + %start)<nuw><nsw>,+,4}<%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: ((-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64) + %start) LoopDispositions: { %loop: Computable }
; CHECK-NEXT: Determining loop execution counts for: @test_ptrs_aligned_by_8_via_assumption
; CHECK-NEXT: Loop %loop: backedge-taken count is ((-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64)) /u 4)
; CHECK-NEXT: Loop %loop: constant max backedge-taken count is i64 4611686018427387903
@@ -677,22 +677,22 @@ define void @test_ptr_aligned_by_4_via_assumption_multiple_loop_predecessors(ptr
; CHECK-NEXT: %c = call i1 @cond()
; CHECK-NEXT: --> %c U: full-set S: full-set
; CHECK-NEXT: %iv = phi ptr [ %start, %then ], [ %start, %else ], [ %iv.next, %loop ]
-; CHECK-NEXT: --> {%start,+,4}<%loop> U: full-set S: full-set Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT: --> {%start,+,4}<%loop> U: [0,-1) S: [-9223372036854775808,9223372036854775807) Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
; CHECK-NEXT: %iv.next = getelementptr i8, ptr %iv, i64 4
-; CHECK-NEXT: --> {(4 + %start),+,4}<%loop> U: full-set S: full-set Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT: --> {(4 + %start),+,4}<%loop> U: [0,-1) S: [-9223372036854775808,9223372036854775807) Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
; CHECK-NEXT: Determining loop execution counts for: @test_ptr_aligned_by_4_via_assumption_multiple_loop_predecessors
; CHECK-NEXT: Loop %loop: Unpredictable backedge-taken count.
; CHECK-NEXT: Loop %loop: Unpredictable constant max backedge-taken count.
; CHECK-NEXT: Loop %loop: Unpredictable symbolic max backedge-taken count.
; CHECK-NEXT: Loop %loop: Predicated backedge-taken count is ((-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64)) /u 4)
; CHECK-NEXT: Predicates:
-; CHECK-NEXT: Equal predicate: (zext i2 ((trunc i64 (ptrtoint ptr %end to i64) to i2) + (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2))) to i64) == 0
+; CHECK-NEXT: Equal predicate: (zext i2 (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2)) to i64) == 0
; CHECK-NEXT: Loop %loop: Predicated constant max backedge-taken count is i64 4611686018427387903
; CHECK-NEXT: Predicates:
-; CHECK-NEXT: Equal predicate: (zext i2 ((trunc i64 (ptrtoint ptr %end to i64) to i2) + (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2))) to i64) == 0
+; CHECK-NEXT: Equal predicate: (zext i2 (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2)) to i64) == 0
; CHECK-NEXT: Loop %loop: Predicated symbolic max backedge-taken count is ((-4 + (-1 * (ptrtoint ptr %start to i64)) + (ptrtoint ptr %end to i64)) /u 4)
; CHECK-NEXT: Predicates:
-; CHECK-NEXT: Equal predicate: (zext i2 ((trunc i64 (ptrtoint ptr %end to i64) to i2) + (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2))) to i64) == 0
+; CHECK-NEXT: Equal predicate: (zext i2 (-1 * (trunc i64 (ptrtoint ptr %start to i64) to i2)) to i64) == 0
;
entry:
call void @llvm.assume(i1 true) [ "align"(ptr %start, i64 2) ]
|
llvm-sync bot
pushed a commit
to arm/arm-toolchain
that referenced
this pull request
Oct 14, 2025
…tantMultiple. (#163260) Follow-up to llvm/llvm-project#160941. Even if we don't have a context instruction for the caller, we should be able to provide context instructions for SCEVUnknowns. Unless I am missing something, SCEVUnknown only become available at the point their underlying IR instruction has been defined. If it is an argument, it should be safe to use the first instruction in the entry block or the instruction itself if it wraps an instruction. This allows getConstantMultiple to make better use of alignment assumptions. PR: llvm/llvm-project#163260
akadutta
pushed a commit
to akadutta/llvm-project
that referenced
this pull request
Oct 14, 2025
…e. (llvm#163260) Follow-up to llvm#160941. Even if we don't have a context instruction for the caller, we should be able to provide context instructions for SCEVUnknowns. Unless I am missing something, SCEVUnknown only become available at the point their underlying IR instruction has been defined. If it is an argument, it should be safe to use the first instruction in the entry block or the instruction itself if it wraps an instruction. This allows getConstantMultiple to make better use of alignment assumptions. PR: llvm#163260
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Follow-up to #160941.
Even if we don't have a context instruction for the caller, we should be able to provide context instructions for SCEVUnknowns. Unless I am missing something, SCEVUnknown only become available at the point their underlying IR instruction has been defined. If it is an argument, it should be safe to use the first instruction in the entry block or the instruction itself if it wraps an instruction.
This allows getConstantMultiple to make better use of alignment assumptions.