[llvm] 3a8b488 - [LoopPeel] Add tests for peeling last iteration with loop guards.
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 3 06:30:07 PDT 2025
Author: Florian Hahn
Date: 2025-06-03T14:29:44+01:00
New Revision: 3a8b48862ad69cb07b53b6f386442b3bef4dae40
URL: https://github.com/llvm/llvm-project/commit/3a8b48862ad69cb07b53b6f386442b3bef4dae40
DIFF: https://github.com/llvm/llvm-project/commit/3a8b48862ad69cb07b53b6f386442b3bef4dae40.diff
LOG: [LoopPeel] Add tests for peeling last iteration with loop guards.
Add additional test coverage for peeling the last iteration where
information from loop guards is needed.
Added:
llvm/test/Transforms/LoopUnroll/peel-last-iteration-with-guards.ll
Modified:
llvm/test/Transforms/LoopUnroll/peel-last-iteration-with-variable-trip-count.ll
Removed:
################################################################################
diff --git a/llvm/test/Transforms/LoopUnroll/peel-last-iteration-with-guards.ll b/llvm/test/Transforms/LoopUnroll/peel-last-iteration-with-guards.ll
new file mode 100644
index 0000000000000..af07a97131322
--- /dev/null
+++ b/llvm/test/Transforms/LoopUnroll/peel-last-iteration-with-guards.ll
@@ -0,0 +1,191 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -p loop-unroll -unroll-full-max-count=0 -S %s | FileCheck %s
+
+declare void @foo(i32)
+
+define void @peel_with_guard_known_nonnegative_1(i32 %n) {
+; CHECK-LABEL: define void @peel_with_guard_known_nonnegative_1(
+; CHECK-SAME: i32 [[N:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK-NEXT: [[PRECOND:%.*]] = icmp slt i32 [[N]], 0
+; CHECK-NEXT: br i1 [[PRECOND]], label %[[EXIT:.*]], label %[[PH:.*]]
+; CHECK: [[PH]]:
+; CHECK-NEXT: [[N_EXT:%.*]] = zext i32 [[N]] to i64
+; CHECK-NEXT: [[N_1:%.*]] = add i32 [[N]], 1
+; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N_1]] to i64
+; CHECK-NEXT: br label %[[LOOP:.*]]
+; CHECK: [[LOOP]]:
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT: [[C:%.*]] = icmp eq i64 [[IV]], [[N_EXT]]
+; CHECK-NEXT: [[SPEC_SELECT:%.*]] = select i1 [[C]], i32 10, i32 20
+; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[WIDE_TRIP_COUNT]]
+; CHECK-NEXT: br i1 [[EC]], label %[[EXIT_LOOPEXIT:.*]], label %[[LOOP]]
+; CHECK: [[EXIT_LOOPEXIT]]:
+; CHECK-NEXT: br label %[[EXIT]]
+; CHECK: [[EXIT]]:
+; CHECK-NEXT: ret void
+;
+entry:
+ %precond = icmp slt i32 %n, 0
+ br i1 %precond, label %exit, label %ph
+
+ph:
+ %n.ext = zext i32 %n to i64
+ %n.1 = add i32 %n, 1
+ %wide.trip.count = zext i32 %n.1 to i64
+ br label %loop
+
+loop:
+ %iv = phi i64 [ 0, %ph ], [ %iv.next, %loop ]
+ %c = icmp eq i64 %iv, %n.ext
+ %spec.select = select i1 %c, i32 10, i32 20
+ %iv.next = add i64 %iv, 1
+ %ec = icmp eq i64 %iv.next, %wide.trip.count
+ br i1 %ec, label %exit, label %loop
+
+exit:
+ ret void
+}
+
+define i32 @peel_with_guard_known_nonnegative_2(ptr %x, i32 %w) {
+; CHECK-LABEL: define i32 @peel_with_guard_known_nonnegative_2(
+; CHECK-SAME: ptr [[X:%.*]], i32 [[W:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*]]:
+; CHECK-NEXT: [[PRECOND:%.*]] = icmp sgt i32 [[W]], 0
+; CHECK-NEXT: br i1 [[PRECOND]], label %[[PH:.*]], label %[[EXIT:.*]]
+; CHECK: [[PH]]:
+; CHECK-NEXT: [[SUB:%.*]] = add nsw i32 [[W]], -1
+; CHECK-NEXT: [[TMP0:%.*]] = zext nneg i32 [[SUB]] to i64
+; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[W]] to i64
+; CHECK-NEXT: br label %[[LOOP_HEADER:.*]]
+; CHECK: [[LOOP_HEADER]]:
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
+; CHECK-NEXT: [[RED:%.*]] = phi i32 [ 0, %[[PH]] ], [ [[ADD:%.*]], %[[LOOP_LATCH]] ]
+; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i64 [[IV]], [[TMP0]]
+; CHECK-NEXT: br i1 [[CMP1]], label %[[IF_THEN:.*]], label %[[LOOP_LATCH]]
+; CHECK: [[IF_THEN]]:
+; CHECK-NEXT: tail call void @foo()
+; CHECK-NEXT: br label %[[LOOP_LATCH]]
+; CHECK: [[LOOP_LATCH]]:
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds nuw i32, ptr [[X]], i64 [[IV]]
+; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+; CHECK-NEXT: [[ADD]] = add nsw i32 [[TMP1]], [[RED]]
+; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[WIDE_TRIP_COUNT]]
+; CHECK-NEXT: br i1 [[EC]], label %[[EXIT_LOOPEXIT:.*]], label %[[LOOP_HEADER]]
+; CHECK: [[EXIT_LOOPEXIT]]:
+; CHECK-NEXT: [[ADD_LCSSA:%.*]] = phi i32 [ [[ADD]], %[[LOOP_LATCH]] ]
+; CHECK-NEXT: br label %[[EXIT]]
+; CHECK: [[EXIT]]:
+; CHECK-NEXT: [[RED_LCSSA:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[ADD_LCSSA]], %[[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT: ret i32 [[RED_LCSSA]]
+;
+entry:
+ %precond = icmp sgt i32 %w, 0
+ br i1 %precond, label %ph, label %exit
+
+ph:
+ %sub = add nsw i32 %w, -1
+ %0 = zext nneg i32 %sub to i64
+ %wide.trip.count = zext nneg i32 %w to i64
+ br label %loop.header
+
+loop.header:
+ %iv = phi i64 [ 0, %ph ], [ %iv.next, %loop.latch ]
+ %red = phi i32 [ 0, %ph ], [ %add, %loop.latch ]
+ %cmp1 = icmp eq i64 %iv, %0
+ br i1 %cmp1, label %if.then, label %loop.latch
+
+if.then:
+ tail call void @foo()
+ br label %loop.latch
+
+loop.latch:
+ %arrayidx = getelementptr inbounds nuw i32, ptr %x, i64 %iv
+ %1 = load i32, ptr %arrayidx, align 4
+ %add = add nsw i32 %1, %red
+ %iv.next = add nuw nsw i64 %iv, 1
+ %ec = icmp eq i64 %iv.next, %wide.trip.count
+ br i1 %ec, label %exit, label %loop.header
+
+exit:
+ %red.lcssa = phi i32 [ 0, %entry ], [ %add, %loop.latch ]
+ ret i32 %red.lcssa
+}
+
+define void @peel_with_guard2(i32 %n) {
+; CHECK-LABEL: define void @peel_with_guard2(
+; CHECK-SAME: i32 [[N:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK-NEXT: [[SUB:%.*]] = add i32 [[N]], -1
+; CHECK-NEXT: [[PRECOND:%.*]] = icmp eq i32 [[SUB]], 0
+; CHECK-NEXT: br i1 [[PRECOND]], label %[[EXIT:.*]], label %[[LOOP_HEADER_PREHEADER:.*]]
+; CHECK: [[LOOP_HEADER_PREHEADER]]:
+; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -2
+; CHECK-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
+; CHECK-NEXT: br i1 [[TMP1]], label %[[LOOP_HEADER_PREHEADER_SPLIT:.*]], label %[[EXIT_LOOPEXIT_PEEL_BEGIN:.*]]
+; CHECK: [[LOOP_HEADER_PREHEADER_SPLIT]]:
+; CHECK-NEXT: br label %[[LOOP_HEADER:.*]]
+; CHECK: [[LOOP_HEADER]]:
+; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ], [ 1, %[[LOOP_HEADER_PREHEADER_SPLIT]] ]
+; CHECK-NEXT: br i1 false, label %[[THEN:.*]], label %[[LOOP_LATCH]]
+; CHECK: [[THEN]]:
+; CHECK-NEXT: call void @foo()
+; CHECK-NEXT: br label %[[LOOP_LATCH]]
+; CHECK: [[LOOP_LATCH]]:
+; CHECK-NEXT: [[IV_NEXT]] = add nuw i32 [[IV]], 1
+; CHECK-NEXT: [[TMP2:%.*]] = sub i32 [[N]], 1
+; CHECK-NEXT: [[EC:%.*]] = icmp eq i32 [[IV_NEXT]], [[TMP2]]
+; CHECK-NEXT: br i1 [[EC]], label %[[EXIT_LOOPEXIT_PEEL_BEGIN_LOOPEXIT:.*]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK: [[EXIT_LOOPEXIT_PEEL_BEGIN_LOOPEXIT]]:
+; CHECK-NEXT: [[DOTPH:%.*]] = phi i32 [ [[IV_NEXT]], %[[LOOP_LATCH]] ]
+; CHECK-NEXT: br label %[[EXIT_LOOPEXIT_PEEL_BEGIN]]
+; CHECK: [[EXIT_LOOPEXIT_PEEL_BEGIN]]:
+; CHECK-NEXT: [[TMP3:%.*]] = phi i32 [ 1, %[[LOOP_HEADER_PREHEADER]] ], [ [[DOTPH]], %[[EXIT_LOOPEXIT_PEEL_BEGIN_LOOPEXIT]] ]
+; CHECK-NEXT: br label %[[LOOP_HEADER_PEEL:.*]]
+; CHECK: [[LOOP_HEADER_PEEL]]:
+; CHECK-NEXT: [[C_PEEL:%.*]] = icmp eq i32 [[TMP3]], [[SUB]]
+; CHECK-NEXT: br i1 [[C_PEEL]], label %[[THEN_PEEL:.*]], label %[[LOOP_LATCH_PEEL:.*]]
+; CHECK: [[THEN_PEEL]]:
+; CHECK-NEXT: call void @foo()
+; CHECK-NEXT: br label %[[LOOP_LATCH_PEEL]]
+; CHECK: [[LOOP_LATCH_PEEL]]:
+; CHECK-NEXT: [[IV_NEXT_PEEL:%.*]] = add nuw i32 [[TMP3]], 1
+; CHECK-NEXT: [[EC_PEEL:%.*]] = icmp eq i32 [[IV_NEXT_PEEL]], [[N]]
+; CHECK-NEXT: br i1 [[EC_PEEL]], label %[[EXIT_LOOPEXIT_PEEL_NEXT:.*]], label %[[EXIT_LOOPEXIT_PEEL_NEXT]]
+; CHECK: [[EXIT_LOOPEXIT_PEEL_NEXT]]:
+; CHECK-NEXT: br label %[[LOOP_HEADER_PEEL_NEXT:.*]]
+; CHECK: [[LOOP_HEADER_PEEL_NEXT]]:
+; CHECK-NEXT: br label %[[EXIT_LOOPEXIT:.*]]
+; CHECK: [[EXIT_LOOPEXIT]]:
+; CHECK-NEXT: br label %[[EXIT]]
+; CHECK: [[EXIT]]:
+; CHECK-NEXT: ret void
+;
+entry:
+ %sub = add i32 %n, -1
+ %precond = icmp eq i32 %sub, 0
+ br i1 %precond, label %exit, label %loop.header
+
+loop.header:
+ %iv = phi i32 [ 1, %entry ], [ %iv.next, %loop.latch ]
+ %c = icmp eq i32 %iv, %sub
+ br i1 %c, label %then, label %loop.latch
+
+then:
+ call void @foo()
+ br label %loop.latch
+
+loop.latch:
+ %iv.next = add nuw i32 %iv, 1
+ %ec = icmp eq i32 %iv.next, %n
+ br i1 %ec, label %exit, label %loop.header
+
+exit:
+ ret void
+}
+;.
+; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]]}
+; CHECK: [[META1]] = !{!"llvm.loop.peeled.count", i32 1}
+;.
diff --git a/llvm/test/Transforms/LoopUnroll/peel-last-iteration-with-variable-trip-count.ll b/llvm/test/Transforms/LoopUnroll/peel-last-iteration-with-variable-trip-count.ll
index a8712e04d43c1..6346d7d97e826 100644
--- a/llvm/test/Transforms/LoopUnroll/peel-last-iteration-with-variable-trip-count.ll
+++ b/llvm/test/Transforms/LoopUnroll/peel-last-iteration-with-variable-trip-count.ll
@@ -1,7 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt -p loop-unroll -unroll-full-max-count=0 -S %s | FileCheck %s
-
declare void @foo(i32)
define i32 @peel_last_with_trip_count_check_lcssa_phi(i32 %n) {
@@ -235,9 +234,10 @@ then:
inner.latch:
%iv.next = add i32 %iv, 1
- %exitcond.not = icmp eq i32 %iv.next, %n
- br i1 %exitcond.not, label %outer.header, label %inner.header
+ %ec = icmp eq i32 %iv.next, %n
+ br i1 %ec, label %outer.header, label %inner.header
}
+
;.
; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]]}
; CHECK: [[META1]] = !{!"llvm.loop.peeled.count", i32 1}
More information about the llvm-commits
mailing list