shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
llvm-project/llvm/test/Transforms/LoopUnroll/peel-last-iteration-with-guards.ll
Florian Hahn e5ff7055be
[LoopPeel] Use loop guards when checking if last iter can be peeled. (#142605) 
Apply loop guards to BTC before checking if the last iteration should be
peeled off. This also adds an assert to make sure applying the guards
does not pessimize the results. I checked on a large test set and it did
not trigger there, but it adds an additional guard to catch potential
cases where loop-guards pessimize results.

Peels ~15% more loops.

PR: https://github.com/llvm/llvm-project/pull/142605
2025-06-10 08:29:42 +01:00

212 lines
8.7 KiB
LLVM
Raw Permalink Blame History

; 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: [[TMP0:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1
; CHECK-NEXT: [[TMP1:%.*]] = icmp ne i64 [[TMP0]], 0
; CHECK-NEXT: br i1 [[TMP1]], label %[[PH_SPLIT:.*]], label %[[EXIT_LOOPEXIT_PEEL_BEGIN:.*]]
; CHECK: [[PH_SPLIT]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ 0, %[[PH_SPLIT]] ], [ [[IV_NEXT1:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
; CHECK-NEXT: [[TMP2:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], 1
; CHECK-NEXT: [[EC1:%.*]] = icmp eq i64 [[IV_NEXT1]], [[TMP2]]
; CHECK-NEXT: br i1 [[EC1]], label %[[EXIT_LOOPEXIT_PEEL_BEGIN_LOOPEXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK: [[EXIT_LOOPEXIT_PEEL_BEGIN_LOOPEXIT]]:
; CHECK-NEXT: [[DOTPH:%.*]] = phi i64 [ [[IV_NEXT1]], %[[LOOP]] ]
; CHECK-NEXT: br label %[[EXIT_LOOPEXIT_PEEL_BEGIN]]
; CHECK: [[EXIT_LOOPEXIT_PEEL_BEGIN]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[PH]] ], [ [[DOTPH]], %[[EXIT_LOOPEXIT_PEEL_BEGIN_LOOPEXIT]] ]
; CHECK-NEXT: br label %[[LOOP_PEEL:.*]]
; CHECK: [[LOOP_PEEL]]:
; 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_PEEL_NEXT:.*]], label %[[EXIT_LOOPEXIT_PEEL_NEXT]]
; CHECK: [[EXIT_LOOPEXIT_PEEL_NEXT]]:
; CHECK-NEXT: br label %[[LOOP_PEEL_NEXT:.*]]
; CHECK: [[LOOP_PEEL_NEXT]]:
; CHECK-NEXT: br label %[[EXIT_LOOPEXIT:.*]]
; 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 [[LOOP2:![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}
; CHECK: [[LOOP2]] = distinct !{[[LOOP2]], [[META1]]}
;.
Reference in New Issue View Git Blame Copy Permalink