Skip to content

Commit 2d02726

Browse files
authored
[SCEV] Collect guard info for ICMP NE w/o constants. (llvm#160500)
When collecting information from loop guards, use UMax(1, %b - %a) for ICMP NE %a, %b, if neither are constant. This improves results in some cases, and will be even more useful together with * llvm#160012 * llvm#159942 https://alive2.llvm.org/ce/z/YyBvoT PR: llvm#160500
1 parent 8c6b499 commit 2d02726

File tree

4 files changed

+27
-13
lines changed

4 files changed

+27
-13
lines changed

llvm/lib/Analysis/ScalarEvolution.cpp

Lines changed: 15 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -15761,6 +15761,21 @@ void ScalarEvolution::LoopGuards::collectFromBlock(
1576115761
const SCEV *OneAlignedUp =
1576215762
GetNextSCEVDividesByDivisor(One, DividesBy);
1576315763
To = SE.getUMaxExpr(FromRewritten, OneAlignedUp);
15764+
} else {
15765+
if (LHS->getType()->isPointerTy()) {
15766+
LHS = SE.getLosslessPtrToIntExpr(LHS);
15767+
RHS = SE.getLosslessPtrToIntExpr(RHS);
15768+
if (isa<SCEVCouldNotCompute>(LHS) || isa<SCEVCouldNotCompute>(RHS))
15769+
break;
15770+
}
15771+
auto AddSubRewrite = [&](const SCEV *A, const SCEV *B) {
15772+
const SCEV *Sub = SE.getMinusSCEV(A, B);
15773+
AddRewrite(Sub, Sub,
15774+
SE.getUMaxExpr(Sub, SE.getOne(From->getType())));
15775+
};
15776+
AddSubRewrite(LHS, RHS);
15777+
AddSubRewrite(RHS, LHS);
15778+
continue;
1576415779
}
1576515780
break;
1576615781
default:

llvm/test/Analysis/ScalarEvolution/ptrtoint.ll

Lines changed: 10 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -382,7 +382,7 @@ define void @pr46786_c26_char(ptr %arg, ptr %arg1, ptr %arg2) {
382382
; X64-NEXT: %i9 = ptrtoint ptr %i7 to i64
383383
; X64-NEXT: --> {(ptrtoint ptr %arg to i64),+,1}<nuw><%bb6> U: full-set S: full-set Exits: (-1 + (ptrtoint ptr %arg1 to i64)) LoopDispositions: { %bb6: Computable }
384384
; X64-NEXT: %i10 = sub i64 %i9, %i4
385-
; X64-NEXT: --> {0,+,1}<nuw><%bb6> U: full-set S: full-set Exits: (-1 + (-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64)) LoopDispositions: { %bb6: Computable }
385+
; X64-NEXT: --> {0,+,1}<nuw><%bb6> U: [0,-1) S: [0,-1) Exits: (-1 + (-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64)) LoopDispositions: { %bb6: Computable }
386386
; X64-NEXT: %i11 = getelementptr inbounds i8, ptr %arg2, i64 %i10
387387
; X64-NEXT: --> {%arg2,+,1}<nw><%bb6> U: full-set S: full-set Exits: (-1 + (-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64) + %arg2) LoopDispositions: { %bb6: Computable }
388388
; X64-NEXT: %i12 = load i8, ptr %i11, align 1
@@ -393,7 +393,7 @@ define void @pr46786_c26_char(ptr %arg, ptr %arg1, ptr %arg2) {
393393
; X64-NEXT: --> {(1 + %arg),+,1}<nuw><%bb6> U: full-set S: full-set Exits: ((-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64) + %arg) LoopDispositions: { %bb6: Computable }
394394
; X64-NEXT: Determining loop execution counts for: @pr46786_c26_char
395395
; X64-NEXT: Loop %bb6: backedge-taken count is (-1 + (-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64))
396-
; X64-NEXT: Loop %bb6: constant max backedge-taken count is i64 -1
396+
; X64-NEXT: Loop %bb6: constant max backedge-taken count is i64 -2
397397
; X64-NEXT: Loop %bb6: symbolic max backedge-taken count is (-1 + (-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64))
398398
; X64-NEXT: Loop %bb6: Trip multiple is 1
399399
;
@@ -406,9 +406,9 @@ define void @pr46786_c26_char(ptr %arg, ptr %arg1, ptr %arg2) {
406406
; X32-NEXT: %i8 = load i8, ptr %i7, align 1
407407
; X32-NEXT: --> %i8 U: full-set S: full-set Exits: <<Unknown>> LoopDispositions: { %bb6: Variant }
408408
; X32-NEXT: %i9 = ptrtoint ptr %i7 to i64
409-
; X32-NEXT: --> {(zext i32 (ptrtoint ptr %arg to i32) to i64),+,1}<nuw><%bb6> U: [0,8589934591) S: [0,8589934591) Exits: ((zext i32 (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32)) to i64) + (zext i32 (ptrtoint ptr %arg to i32) to i64)) LoopDispositions: { %bb6: Computable }
409+
; X32-NEXT: --> {(zext i32 (ptrtoint ptr %arg to i32) to i64),+,1}<nuw><%bb6> U: [0,8589934590) S: [0,8589934590) Exits: ((zext i32 (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32)) to i64) + (zext i32 (ptrtoint ptr %arg to i32) to i64)) LoopDispositions: { %bb6: Computable }
410410
; X32-NEXT: %i10 = sub i64 %i9, %i4
411-
; X32-NEXT: --> {0,+,1}<nuw><%bb6> U: [0,4294967296) S: [0,4294967296) Exits: (zext i32 (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32)) to i64) LoopDispositions: { %bb6: Computable }
411+
; X32-NEXT: --> {0,+,1}<nuw><%bb6> U: [0,4294967295) S: [0,4294967295) Exits: (zext i32 (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32)) to i64) LoopDispositions: { %bb6: Computable }
412412
; X32-NEXT: %i11 = getelementptr inbounds i8, ptr %arg2, i64 %i10
413413
; X32-NEXT: --> {%arg2,+,1}<%bb6> U: full-set S: full-set Exits: (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32) + %arg2) LoopDispositions: { %bb6: Computable }
414414
; X32-NEXT: %i12 = load i8, ptr %i11, align 1
@@ -419,7 +419,7 @@ define void @pr46786_c26_char(ptr %arg, ptr %arg1, ptr %arg2) {
419419
; X32-NEXT: --> {(1 + %arg),+,1}<nuw><%bb6> U: full-set S: full-set Exits: ((-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32) + %arg) LoopDispositions: { %bb6: Computable }
420420
; X32-NEXT: Determining loop execution counts for: @pr46786_c26_char
421421
; X32-NEXT: Loop %bb6: backedge-taken count is (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32))
422-
; X32-NEXT: Loop %bb6: constant max backedge-taken count is i32 -1
422+
; X32-NEXT: Loop %bb6: constant max backedge-taken count is i32 -2
423423
; X32-NEXT: Loop %bb6: symbolic max backedge-taken count is (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32))
424424
; X32-NEXT: Loop %bb6: Trip multiple is 1
425425
;
@@ -459,7 +459,7 @@ define void @pr46786_c26_char_cmp_ops_swapped(ptr %arg, ptr %arg1, ptr %arg2) {
459459
; X64-NEXT: %i9 = ptrtoint ptr %i7 to i64
460460
; X64-NEXT: --> {(ptrtoint ptr %arg to i64),+,1}<nuw><%bb6> U: full-set S: full-set Exits: (-1 + (ptrtoint ptr %arg1 to i64)) LoopDispositions: { %bb6: Computable }
461461
; X64-NEXT: %i10 = sub i64 %i9, %i4
462-
; X64-NEXT: --> {0,+,1}<nuw><%bb6> U: full-set S: full-set Exits: (-1 + (-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64)) LoopDispositions: { %bb6: Computable }
462+
; X64-NEXT: --> {0,+,1}<nuw><%bb6> U: [0,-1) S: [0,-1) Exits: (-1 + (-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64)) LoopDispositions: { %bb6: Computable }
463463
; X64-NEXT: %i11 = getelementptr inbounds i8, ptr %arg2, i64 %i10
464464
; X64-NEXT: --> {%arg2,+,1}<nw><%bb6> U: full-set S: full-set Exits: (-1 + (-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64) + %arg2) LoopDispositions: { %bb6: Computable }
465465
; X64-NEXT: %i12 = load i8, ptr %i11, align 1
@@ -470,7 +470,7 @@ define void @pr46786_c26_char_cmp_ops_swapped(ptr %arg, ptr %arg1, ptr %arg2) {
470470
; X64-NEXT: --> {(1 + %arg),+,1}<nuw><%bb6> U: full-set S: full-set Exits: ((-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64) + %arg) LoopDispositions: { %bb6: Computable }
471471
; X64-NEXT: Determining loop execution counts for: @pr46786_c26_char_cmp_ops_swapped
472472
; X64-NEXT: Loop %bb6: backedge-taken count is (-1 + (-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64))
473-
; X64-NEXT: Loop %bb6: constant max backedge-taken count is i64 -1
473+
; X64-NEXT: Loop %bb6: constant max backedge-taken count is i64 -2
474474
; X64-NEXT: Loop %bb6: symbolic max backedge-taken count is (-1 + (-1 * (ptrtoint ptr %arg to i64)) + (ptrtoint ptr %arg1 to i64))
475475
; X64-NEXT: Loop %bb6: Trip multiple is 1
476476
;
@@ -483,9 +483,9 @@ define void @pr46786_c26_char_cmp_ops_swapped(ptr %arg, ptr %arg1, ptr %arg2) {
483483
; X32-NEXT: %i8 = load i8, ptr %i7, align 1
484484
; X32-NEXT: --> %i8 U: full-set S: full-set Exits: <<Unknown>> LoopDispositions: { %bb6: Variant }
485485
; X32-NEXT: %i9 = ptrtoint ptr %i7 to i64
486-
; X32-NEXT: --> {(zext i32 (ptrtoint ptr %arg to i32) to i64),+,1}<nuw><%bb6> U: [0,8589934591) S: [0,8589934591) Exits: ((zext i32 (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32)) to i64) + (zext i32 (ptrtoint ptr %arg to i32) to i64)) LoopDispositions: { %bb6: Computable }
486+
; X32-NEXT: --> {(zext i32 (ptrtoint ptr %arg to i32) to i64),+,1}<nuw><%bb6> U: [0,8589934590) S: [0,8589934590) Exits: ((zext i32 (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32)) to i64) + (zext i32 (ptrtoint ptr %arg to i32) to i64)) LoopDispositions: { %bb6: Computable }
487487
; X32-NEXT: %i10 = sub i64 %i9, %i4
488-
; X32-NEXT: --> {0,+,1}<nuw><%bb6> U: [0,4294967296) S: [0,4294967296) Exits: (zext i32 (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32)) to i64) LoopDispositions: { %bb6: Computable }
488+
; X32-NEXT: --> {0,+,1}<nuw><%bb6> U: [0,4294967295) S: [0,4294967295) Exits: (zext i32 (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32)) to i64) LoopDispositions: { %bb6: Computable }
489489
; X32-NEXT: %i11 = getelementptr inbounds i8, ptr %arg2, i64 %i10
490490
; X32-NEXT: --> {%arg2,+,1}<%bb6> U: full-set S: full-set Exits: (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32) + %arg2) LoopDispositions: { %bb6: Computable }
491491
; X32-NEXT: %i12 = load i8, ptr %i11, align 1
@@ -496,7 +496,7 @@ define void @pr46786_c26_char_cmp_ops_swapped(ptr %arg, ptr %arg1, ptr %arg2) {
496496
; X32-NEXT: --> {(1 + %arg),+,1}<nuw><%bb6> U: full-set S: full-set Exits: ((-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32) + %arg) LoopDispositions: { %bb6: Computable }
497497
; X32-NEXT: Determining loop execution counts for: @pr46786_c26_char_cmp_ops_swapped
498498
; X32-NEXT: Loop %bb6: backedge-taken count is (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32))
499-
; X32-NEXT: Loop %bb6: constant max backedge-taken count is i32 -1
499+
; X32-NEXT: Loop %bb6: constant max backedge-taken count is i32 -2
500500
; X32-NEXT: Loop %bb6: symbolic max backedge-taken count is (-1 + (-1 * (ptrtoint ptr %arg to i32)) + (ptrtoint ptr %arg1 to i32))
501501
; X32-NEXT: Loop %bb6: Trip multiple is 1
502502
;

llvm/test/Transforms/IndVarSimplify/pointer-loop-guards.ll

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -18,7 +18,7 @@ define i64 @test_ptr_compare_guard(ptr %start, ptr %end) {
1818
; CHECK-NEXT: br i1 [[C_1]], label %[[LOOP_LATCH]], label %[[EXIT_LOOPEXIT:.*]]
1919
; CHECK: [[LOOP_LATCH]]:
2020
; CHECK-NEXT: [[PTR_IV_NEXT]] = getelementptr i8, ptr [[PTR_IV]], i64 1
21-
; CHECK-NEXT: [[I64_IV_NEXT]] = add i64 [[I64_IV]], 1
21+
; CHECK-NEXT: [[I64_IV_NEXT]] = add nuw i64 [[I64_IV]], 1
2222
; CHECK-NEXT: [[C_2:%.*]] = icmp eq ptr [[PTR_IV_NEXT]], [[END]]
2323
; CHECK-NEXT: br i1 [[C_2]], label %[[EXIT_LOOPEXIT]], label %[[LOOP_HEADER]]
2424
; CHECK: [[EXIT_LOOPEXIT]]:

llvm/test/Transforms/LoopUnroll/scevunroll.ll

Lines changed: 1 addition & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -465,8 +465,7 @@ define void @peel_int_eq_condition(i32 %start) {
465465
; CHECK-NEXT: br label [[LOOP:%.*]]
466466
; CHECK: loop:
467467
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT_PEEL]], [[ENTRY_PEEL_NEWPH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
468-
; CHECK-NEXT: [[C_0:%.*]] = icmp eq i32 [[IV]], [[START]]
469-
; CHECK-NEXT: br i1 [[C_0]], label [[IF_THEN:%.*]], label [[LOOP_LATCH]]
468+
; CHECK-NEXT: br i1 false, label [[IF_THEN:%.*]], label [[LOOP_LATCH]]
470469
; CHECK: if.then:
471470
; CHECK-NEXT: call void @fn(i32 [[IV]])
472471
; CHECK-NEXT: br label [[LOOP_LATCH]]

0 commit comments

Comments
 (0)