[llvm] 0afbf17 - [LAA,LV] Add tests with early-exit loops and deref assumptions.
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Thu Jul 3 11:41:41 PDT 2025
Author: Florian Hahn
Date: 2025-07-03T19:41:13+01:00
New Revision: 0afbf1721346faca3dbcd659acf2c558c26146fb
URL: https://github.com/llvm/llvm-project/commit/0afbf1721346faca3dbcd659acf2c558c26146fb
DIFF: https://github.com/llvm/llvm-project/commit/0afbf1721346faca3dbcd659acf2c558c26146fb.diff
LOG: [LAA,LV] Add tests with early-exit loops and deref assumptions.
Adds additional test coverage for early-exit loops with deref
assumptions, as suggested in
https://github.com/llvm/llvm-project/pull/128436.
Added:
llvm/test/Transforms/LoopVectorize/single-early-exit-deref-assumptions.ll
Modified:
llvm/test/Analysis/LoopAccessAnalysis/early-exit-runtime-checks.ll
Removed:
################################################################################
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/early-exit-runtime-checks.ll b/llvm/test/Analysis/LoopAccessAnalysis/early-exit-runtime-checks.ll
index d4cd61bfde2f4..1dc8d4a7e73f8 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/early-exit-runtime-checks.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/early-exit-runtime-checks.ll
@@ -504,3 +504,123 @@ e.0:
e.1:
ret i32 1
}
+
+define void @all_exits_dominate_latch_countable_exits_at_most_500_iterations_known_deref_via_assumption(ptr %A, ptr %B) {
+; CHECK-LABEL: 'all_exits_dominate_latch_countable_exits_at_most_500_iterations_known_deref_via_assumption'
+; CHECK-NEXT: loop.header:
+; CHECK-NEXT: Memory dependences are safe with run-time checks
+; CHECK-NEXT: Dependences:
+; CHECK-NEXT: Run-time memory checks:
+; CHECK-NEXT: Check 0:
+; CHECK-NEXT: Comparing group GRP0:
+; CHECK-NEXT: %gep.B = getelementptr inbounds i32, ptr %B, i64 %iv
+; CHECK-NEXT: Against group GRP1:
+; CHECK-NEXT: %gep.A = getelementptr inbounds i32, ptr %A, i64 %iv
+; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group GRP0:
+; CHECK-NEXT: (Low: %B High: inttoptr (i64 -1 to ptr))
+; CHECK-NEXT: Member: {%B,+,4}<nuw><%loop.header>
+; CHECK-NEXT: Group GRP1:
+; CHECK-NEXT: (Low: %A High: inttoptr (i64 -1 to ptr))
+; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop.header>
+; CHECK-EMPTY:
+; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
+; CHECK-NEXT: SCEV assumptions:
+; CHECK-EMPTY:
+; CHECK-NEXT: Expressions re-written:
+;
+entry:
+ call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %A, i64 2000) ]
+ call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %B, i64 2000) ]
+ br label %loop.header
+
+loop.header:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %latch ]
+ %gep.A = getelementptr inbounds i32, ptr %A, i64 %iv
+ %gep.B = getelementptr inbounds i32, ptr %B, i64 %iv
+ %l = load i32, ptr %gep.A, align 4
+ store i32 0, ptr %gep.B, align 4
+ %cntable.c.1 = icmp ult i64 %iv, 1000
+ %iv.next = add nuw nsw i64 %iv, 1
+ br i1 %cntable.c.1, label %b2, label %e.1
+
+b2:
+ %uncntable.c.0 = icmp eq i32 %l, 0
+ br i1 %uncntable.c.0, label %e.2, label %b3
+
+b3:
+ %cntable.c.2 = icmp eq i64 %iv.next, 500
+ br i1 %cntable.c.2, label %cleanup4, label %latch
+
+latch:
+ br label %loop.header
+
+cleanup4:
+ ret void
+
+e.1:
+ ret void
+
+e.2:
+ ret void
+}
+
+define void @all_exits_dominate_latch_countable_exits_at_most_500_iterations_deref_via_assumption_too_small(ptr %A, ptr %B) {
+; CHECK-LABEL: 'all_exits_dominate_latch_countable_exits_at_most_500_iterations_deref_via_assumption_too_small'
+; CHECK-NEXT: loop.header:
+; CHECK-NEXT: Memory dependences are safe with run-time checks
+; CHECK-NEXT: Dependences:
+; CHECK-NEXT: Run-time memory checks:
+; CHECK-NEXT: Check 0:
+; CHECK-NEXT: Comparing group GRP0:
+; CHECK-NEXT: %gep.B = getelementptr inbounds i32, ptr %B, i64 %iv
+; CHECK-NEXT: Against group GRP1:
+; CHECK-NEXT: %gep.A = getelementptr inbounds i32, ptr %A, i64 %iv
+; CHECK-NEXT: Grouped accesses:
+; CHECK-NEXT: Group GRP0:
+; CHECK-NEXT: (Low: %B High: inttoptr (i64 -1 to ptr))
+; CHECK-NEXT: Member: {%B,+,4}<nuw><%loop.header>
+; CHECK-NEXT: Group GRP1:
+; CHECK-NEXT: (Low: %A High: inttoptr (i64 -1 to ptr))
+; CHECK-NEXT: Member: {%A,+,4}<nuw><%loop.header>
+; CHECK-EMPTY:
+; CHECK-NEXT: Non vectorizable stores to invariant address were not found in loop.
+; CHECK-NEXT: SCEV assumptions:
+; CHECK-EMPTY:
+; CHECK-NEXT: Expressions re-written:
+;
+entry:
+ call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %A, i64 1999) ]
+ call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %B, i64 1999) ]
+ br label %loop.header
+
+loop.header:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %latch ]
+ %gep.A = getelementptr inbounds i32, ptr %A, i64 %iv
+ %gep.B = getelementptr inbounds i32, ptr %B, i64 %iv
+ %l = load i32, ptr %gep.A, align 4
+ store i32 0, ptr %gep.B, align 4
+ %cntable.c.1 = icmp ult i64 %iv, 1000
+ %iv.next = add nuw nsw i64 %iv, 1
+ br i1 %cntable.c.1, label %b2, label %e.1
+
+b2:
+ %uncntable.c.0 = icmp eq i32 %l, 0
+ br i1 %uncntable.c.0, label %e.2, label %b3
+
+b3:
+ %cntable.c.2 = icmp eq i64 %iv.next, 500
+ br i1 %cntable.c.2, label %cleanup4, label %latch
+
+latch:
+ br label %loop.header
+
+cleanup4:
+ ret void
+
+e.1:
+ ret void
+
+e.2:
+ ret void
+}
diff --git a/llvm/test/Transforms/LoopVectorize/single-early-exit-deref-assumptions.ll b/llvm/test/Transforms/LoopVectorize/single-early-exit-deref-assumptions.ll
new file mode 100644
index 0000000000000..0fe893abec86c
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/single-early-exit-deref-assumptions.ll
@@ -0,0 +1,96 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 4
+; RUN: opt -S < %s -p loop-vectorize -enable-early-exit-vectorization -force-vector-width=4 | FileCheck %s
+
+define i64 @early_exit_alignment_and_deref_known_via_assumption_with_constant_size(ptr noalias %p1, ptr noalias %p2) nofree nosync {
+; CHECK-LABEL: define i64 @early_exit_alignment_and_deref_known_via_assumption_with_constant_size(
+; CHECK-SAME: ptr noalias [[P1:%.*]], ptr noalias [[P2:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT: entry:
+; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P1]], i64 4), "dereferenceable"(ptr [[P1]], i64 1024) ]
+; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P2]], i64 4), "dereferenceable"(ptr [[P2]], i64 1024) ]
+; CHECK-NEXT: br label [[LOOP:%.*]]
+; CHECK: loop:
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[INDEX]]
+; CHECK-NEXT: [[LD1:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
+; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX]]
+; CHECK-NEXT: [[LD2:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1
+; CHECK-NEXT: [[CMP3:%.*]] = icmp eq i8 [[LD1]], [[LD2]]
+; CHECK-NEXT: br i1 [[CMP3]], label [[LOOP_INC]], label [[LOOP_END:%.*]]
+; CHECK: loop.inc:
+; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 1
+; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDEX_NEXT]], 1024
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[LOOP_END]]
+; CHECK: loop.end:
+; CHECK-NEXT: [[RETVAL:%.*]] = phi i64 [ [[INDEX]], [[LOOP]] ], [ -1, [[LOOP_INC]] ]
+; CHECK-NEXT: ret i64 [[RETVAL]]
+;
+entry:
+ call void @llvm.assume(i1 true) [ "align"(ptr %p1, i64 4), "dereferenceable"(ptr %p1, i64 1024) ]
+ call void @llvm.assume(i1 true) [ "align"(ptr %p2, i64 4), "dereferenceable"(ptr %p2, i64 1024) ]
+ br label %loop
+
+loop:
+ %index = phi i64 [ %index.next, %loop.inc ], [ 0, %entry ]
+ %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
+ %ld1 = load i8, ptr %arrayidx, align 1
+ %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index
+ %ld2 = load i8, ptr %arrayidx1, align 1
+ %cmp3 = icmp eq i8 %ld1, %ld2
+ br i1 %cmp3, label %loop.inc, label %loop.end
+
+loop.inc:
+ %index.next = add i64 %index, 1
+ %exitcond = icmp ne i64 %index.next, 1024
+ br i1 %exitcond, label %loop, label %loop.end
+
+loop.end:
+ %retval = phi i64 [ %index, %loop ], [ -1, %loop.inc ]
+ ret i64 %retval
+}
+
+define i64 @early_exit_alignment_and_deref_via_assumption_with_constant_size_too_small(ptr noalias %p1, ptr noalias %p2) nofree nosync {
+; CHECK-LABEL: define i64 @early_exit_alignment_and_deref_via_assumption_with_constant_size_too_small(
+; CHECK-SAME: ptr noalias [[P1:%.*]], ptr noalias [[P2:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: entry:
+; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P1]], i64 4), "dereferenceable"(ptr [[P1]], i64 1024) ]
+; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P2]], i64 4), "dereferenceable"(ptr [[P2]], i64 1024) ]
+; CHECK-NEXT: br label [[LOOP:%.*]]
+; CHECK: loop:
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[INDEX]]
+; CHECK-NEXT: [[LD1:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
+; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX]]
+; CHECK-NEXT: [[LD2:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1
+; CHECK-NEXT: [[CMP3:%.*]] = icmp eq i8 [[LD1]], [[LD2]]
+; CHECK-NEXT: br i1 [[CMP3]], label [[LOOP_INC]], label [[LOOP_END:%.*]]
+; CHECK: loop.inc:
+; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 1
+; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDEX_NEXT]], 1025
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[LOOP_END]]
+; CHECK: loop.end:
+; CHECK-NEXT: [[RETVAL:%.*]] = phi i64 [ [[INDEX]], [[LOOP]] ], [ -1, [[LOOP_INC]] ]
+; CHECK-NEXT: ret i64 [[RETVAL]]
+;
+entry:
+ call void @llvm.assume(i1 true) [ "align"(ptr %p1, i64 4), "dereferenceable"(ptr %p1, i64 1024) ]
+ call void @llvm.assume(i1 true) [ "align"(ptr %p2, i64 4), "dereferenceable"(ptr %p2, i64 1024) ]
+ br label %loop
+
+loop:
+ %index = phi i64 [ %index.next, %loop.inc ], [ 0, %entry ]
+ %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
+ %ld1 = load i8, ptr %arrayidx, align 1
+ %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index
+ %ld2 = load i8, ptr %arrayidx1, align 1
+ %cmp3 = icmp eq i8 %ld1, %ld2
+ br i1 %cmp3, label %loop.inc, label %loop.end
+
+loop.inc:
+ %index.next = add i64 %index, 1
+ %exitcond = icmp ne i64 %index.next, 1025
+ br i1 %exitcond, label %loop, label %loop.end
+
+loop.end:
+ %retval = phi i64 [ %index, %loop ], [ -1, %loop.inc ]
+ ret i64 %retval
+}
More information about the llvm-commits
mailing list