RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-26 04:32:44 +01:00
Code Issues Projects Releases Wiki Activity
7563d8dfe5
llvm-mirror/test/Transforms/LoopDeletion/noop-loops-with-subloops.ll

430 lines
13 KiB
LLVM
Raw Normal View History

[LoopDeletion] Add test case where outer loop needs to be deleted. In the test case @test1, the inner loop cannot be removed, because it has a live-out value. But the outer loop is a no-op and can be removed.
2020-12-22 18:48:01 +01:00
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -loop-deletion -verify-dom-info -S | FileCheck %s
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64"
%pair_t = type { i64, i64 }
; The loop %inner cannot be removed, because it has outgoing values. But the
; outer loop is a no-op and can be removed.
define void @test1(i64 %N, i64 %M, %pair_t* %ptr) willreturn {
; CHECK-LABEL: @test1(
; CHECK-NEXT: entry:
[LoopDeletion] Also consider loops with subloops for deletion. Currently, LoopDeletion does skip loops that have sub-loops, but this means we currently fail to remove some no-op loops. One example are inner loops with live-out values. Those cannot be removed by itself. But the containing loop may itself be a no-op and the whole loop-nest can be deleted. The legality checks do not seem to rely on analyzing inner-loops only for correctness. With LoopDeletion being a LoopPass, the change means that we now unfortunately need to do some extra work in parent loops, by checking some conditions we already checked. But there appears to be no noticeable compile time impact: http://llvm-compile-time-tracker.com/compare.php?from=02d11f3cda2ab5b8bf4fc02639fd1f4b8c45963e&to=843201e9cf3b6871e18c52aede5897a22994c36c&stat=instructions This changes patch leads to ~10 more loops being deleted on MultiSource, SPEC2000, SPEC2006 with -O3 & LTO This patch is also required (together with a few others) to eliminate a no-op loop in omnetpp as discussed on llvm-dev 'LoopDeletion / removal of empty loops.' (http://lists.llvm.org/pipermail/llvm-dev/2020-December/147462.html) This change becomes relevant after removing potentially infinite loops is made possible in 'must-progress' loops (D86844). Note that I added a function call with side-effects to an outer loop in `llvm/test/Transforms/LoopDeletion/update-scev.ll` to preserve the original spirit of the test. Reviewed By: reames Differential Revision: https://reviews.llvm.org/D93716
2021-01-06 12:51:52 +01:00
; CHECK-NEXT: br label [[EXIT:%.*]]
[LoopDeletion] Add test case where outer loop needs to be deleted. In the test case @test1, the inner loop cannot be removed, because it has a live-out value. But the outer loop is a no-op and can be removed.
2020-12-22 18:48:01 +01:00
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %outer.header
outer.header:
%outer.iv = phi i64 [ 0, %entry ], [ %outer.iv.next, %outer.latch ]
br label %inner
inner:
%inner.iv = phi i64 [ 0, %outer.header ], [ %inner.iv.next, %inner ]
%gep = getelementptr %pair_t, %pair_t* %ptr, i64 %inner.iv
%p = load %pair_t, %pair_t* %gep
%v.0 = extractvalue %pair_t %p, 0
%v.1 = extractvalue %pair_t %p, 1
%inner.ec = icmp ult i64 %v.0, %v.1
%inner.iv.next = add i64 %inner.iv, 1
br i1 %inner.ec, label %outer.latch, label %inner
outer.latch:
%lcssa = phi i64 [ %v.1, %inner ]
%outer.ec = icmp ult i64 %outer.iv, %lcssa
%outer.iv.next = add i64 %outer.iv, 1
br i1 %outer.ec, label %exit, label %outer.header
exit:
ret void
}
declare void @sideeffect()
; Same as @test1, but with a call in the outer loop. Nothing can be deleted.
define void @test2_sideeffect_in_outer(i64 %N, i64 %M, %pair_t* %ptr) willreturn {
; CHECK-LABEL: @test2_sideeffect_in_outer(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[OUTER_HEADER:%.*]]
; CHECK: outer.header:
Revert "[LoopDeletion] Break backedge if we can prove that the loop is exited on 1st iteration" This reverts commit 2531fd70d19aa5d61feb533bbdeee7717a4129eb due to performance regression on the PPC buildbot.
2021-05-25 22:58:13 +02:00
; CHECK-NEXT: [[OUTER_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[OUTER_IV_NEXT:%.*]], [[OUTER_LATCH:%.*]] ]
[LoopDeletion] Add test case where outer loop needs to be deleted. In the test case @test1, the inner loop cannot be removed, because it has a live-out value. But the outer loop is a no-op and can be removed.
2020-12-22 18:48:01 +01:00
; CHECK-NEXT: br label [[INNER:%.*]]
; CHECK: inner:
; CHECK-NEXT: [[INNER_IV:%.*]] = phi i64 [ 0, [[OUTER_HEADER]] ], [ [[INNER_IV_NEXT:%.*]], [[INNER]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr [[PAIR_T:%.*]], %pair_t* [[PTR:%.*]], i64 [[INNER_IV]]
; CHECK-NEXT: [[P:%.*]] = load [[PAIR_T]], %pair_t* [[GEP]], align 4
; CHECK-NEXT: [[V_0:%.*]] = extractvalue [[PAIR_T]] [[P]], 0
; CHECK-NEXT: [[V_1:%.*]] = extractvalue [[PAIR_T]] [[P]], 1
; CHECK-NEXT: [[INNER_EC:%.*]] = icmp ult i64 [[V_0]], [[V_1]]
; CHECK-NEXT: [[INNER_IV_NEXT]] = add i64 [[INNER_IV]], 1
Revert "[LoopDeletion] Break backedge if we can prove that the loop is exited on 1st iteration" This reverts commit 2531fd70d19aa5d61feb533bbdeee7717a4129eb due to performance regression on the PPC buildbot.
2021-05-25 22:58:13 +02:00
; CHECK-NEXT: br i1 [[INNER_EC]], label [[OUTER_LATCH]], label [[INNER]]
[LoopDeletion] Add test case where outer loop needs to be deleted. In the test case @test1, the inner loop cannot be removed, because it has a live-out value. But the outer loop is a no-op and can be removed.
2020-12-22 18:48:01 +01:00
; CHECK: outer.latch:
; CHECK-NEXT: [[LCSSA:%.*]] = phi i64 [ [[V_1]], [[INNER]] ]
; CHECK-NEXT: [[OUTER_EC:%.*]] = icmp ult i64 [[OUTER_IV]], [[LCSSA]]
Revert "[LoopDeletion] Break backedge if we can prove that the loop is exited on 1st iteration" This reverts commit 2531fd70d19aa5d61feb533bbdeee7717a4129eb due to performance regression on the PPC buildbot.
2021-05-25 22:58:13 +02:00
; CHECK-NEXT: [[OUTER_IV_NEXT]] = add i64 [[OUTER_IV]], 1
[LoopDeletion] Add test case where outer loop needs to be deleted. In the test case @test1, the inner loop cannot be removed, because it has a live-out value. But the outer loop is a no-op and can be removed.
2020-12-22 18:48:01 +01:00
; CHECK-NEXT: call void @sideeffect()
Revert "[LoopDeletion] Break backedge if we can prove that the loop is exited on 1st iteration" This reverts commit 2531fd70d19aa5d61feb533bbdeee7717a4129eb due to performance regression on the PPC buildbot.
2021-05-25 22:58:13 +02:00
; CHECK-NEXT: br i1 [[OUTER_EC]], label [[EXIT:%.*]], label [[OUTER_HEADER]]
[LoopDeletion] Add test case where outer loop needs to be deleted. In the test case @test1, the inner loop cannot be removed, because it has a live-out value. But the outer loop is a no-op and can be removed.
2020-12-22 18:48:01 +01:00
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %outer.header
outer.header:
%outer.iv = phi i64 [ 0, %entry ], [ %outer.iv.next, %outer.latch ]
br label %inner
inner:
%inner.iv = phi i64 [ 0, %outer.header ], [ %inner.iv.next, %inner ]
%gep = getelementptr %pair_t, %pair_t* %ptr, i64 %inner.iv
%p = load %pair_t, %pair_t* %gep
%v.0 = extractvalue %pair_t %p, 0
%v.1 = extractvalue %pair_t %p, 1
%inner.ec = icmp ult i64 %v.0, %v.1
%inner.iv.next = add i64 %inner.iv, 1
br i1 %inner.ec, label %outer.latch, label %inner
outer.latch:
%lcssa = phi i64 [ %v.1, %inner ]
%outer.ec = icmp ult i64 %outer.iv, %lcssa
%outer.iv.next = add i64 %outer.iv, 1
call void @sideeffect()
br i1 %outer.ec, label %exit, label %outer.header
exit:
ret void
}
; Same as @test1, but with a call in the inner loop. Nothing can be deleted.
define void @test3_sideeffect_in_inner(i64 %N, i64 %M, %pair_t* %ptr) willreturn {
; CHECK-LABEL: @test3_sideeffect_in_inner(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[OUTER_HEADER:%.*]]
; CHECK: outer.header:
Revert "[LoopDeletion] Break backedge if we can prove that the loop is exited on 1st iteration" This reverts commit 2531fd70d19aa5d61feb533bbdeee7717a4129eb due to performance regression on the PPC buildbot.
2021-05-25 22:58:13 +02:00
; CHECK-NEXT: [[OUTER_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[OUTER_IV_NEXT:%.*]], [[OUTER_LATCH:%.*]] ]
[LoopDeletion] Add test case where outer loop needs to be deleted. In the test case @test1, the inner loop cannot be removed, because it has a live-out value. But the outer loop is a no-op and can be removed.
2020-12-22 18:48:01 +01:00
; CHECK-NEXT: br label [[INNER:%.*]]
; CHECK: inner:
; CHECK-NEXT: [[INNER_IV:%.*]] = phi i64 [ 0, [[OUTER_HEADER]] ], [ [[INNER_IV_NEXT:%.*]], [[INNER]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr [[PAIR_T:%.*]], %pair_t* [[PTR:%.*]], i64 [[INNER_IV]]
; CHECK-NEXT: [[P:%.*]] = load [[PAIR_T]], %pair_t* [[GEP]], align 4
; CHECK-NEXT: [[V_0:%.*]] = extractvalue [[PAIR_T]] [[P]], 0
; CHECK-NEXT: [[V_1:%.*]] = extractvalue [[PAIR_T]] [[P]], 1
; CHECK-NEXT: [[INNER_EC:%.*]] = icmp ult i64 [[V_0]], [[V_1]]
; CHECK-NEXT: [[INNER_IV_NEXT]] = add i64 [[INNER_IV]], 1
; CHECK-NEXT: call void @sideeffect()
Revert "[LoopDeletion] Break backedge if we can prove that the loop is exited on 1st iteration" This reverts commit 2531fd70d19aa5d61feb533bbdeee7717a4129eb due to performance regression on the PPC buildbot.
2021-05-25 22:58:13 +02:00
; CHECK-NEXT: br i1 [[INNER_EC]], label [[OUTER_LATCH]], label [[INNER]]
[LoopDeletion] Add test case where outer loop needs to be deleted. In the test case @test1, the inner loop cannot be removed, because it has a live-out value. But the outer loop is a no-op and can be removed.
2020-12-22 18:48:01 +01:00
; CHECK: outer.latch:
; CHECK-NEXT: [[LCSSA:%.*]] = phi i64 [ [[V_1]], [[INNER]] ]
; CHECK-NEXT: [[OUTER_EC:%.*]] = icmp ult i64 [[OUTER_IV]], [[LCSSA]]
Revert "[LoopDeletion] Break backedge if we can prove that the loop is exited on 1st iteration" This reverts commit 2531fd70d19aa5d61feb533bbdeee7717a4129eb due to performance regression on the PPC buildbot.
2021-05-25 22:58:13 +02:00
; CHECK-NEXT: [[OUTER_IV_NEXT]] = add i64 [[OUTER_IV]], 1
; CHECK-NEXT: br i1 [[OUTER_EC]], label [[EXIT:%.*]], label [[OUTER_HEADER]]
[LoopDeletion] Add test case where outer loop needs to be deleted. In the test case @test1, the inner loop cannot be removed, because it has a live-out value. But the outer loop is a no-op and can be removed.
2020-12-22 18:48:01 +01:00
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %outer.header
outer.header:
%outer.iv = phi i64 [ 0, %entry ], [ %outer.iv.next, %outer.latch ]
br label %inner
inner:
%inner.iv = phi i64 [ 0, %outer.header ], [ %inner.iv.next, %inner ]
%gep = getelementptr %pair_t, %pair_t* %ptr, i64 %inner.iv
%p = load %pair_t, %pair_t* %gep
%v.0 = extractvalue %pair_t %p, 0
%v.1 = extractvalue %pair_t %p, 1
%inner.ec = icmp ult i64 %v.0, %v.1
%inner.iv.next = add i64 %inner.iv, 1
call void @sideeffect()
br i1 %inner.ec, label %outer.latch, label %inner
outer.latch:
%lcssa = phi i64 [ %v.1, %inner ]
%outer.ec = icmp ult i64 %outer.iv, %lcssa
%outer.iv.next = add i64 %outer.iv, 1
br i1 %outer.ec, label %exit, label %outer.header
exit:
ret void
}
[LoopDeletion] Add test with potentially infinite sub-loop. Tests for PR50511.
2021-05-28 17:58:59 +02:00
[LoopDeletion] Add more tests with infinite sub-loops & mustprogress. A couple of additional tests inspired by PR50511.
2021-05-30 17:34:30 +02:00
; The inner loop may not terminate, so we cannot remove it, unless the
[LoopDeletion] Add test with potentially infinite sub-loop. Tests for PR50511.
2021-05-28 17:58:59 +02:00
; function/loop is mustprogress. Test case from PR50511.
define void @inner_loop_may_be_infinite(i1 %c1, i1 %c2) {
; CHECK-LABEL: @inner_loop_may_be_infinite(
[LoopDeletion] Consider infinite loops alive, unless mustprogress. The current loop or any of its sub-loops may be infinite. Unless the function or the loops are marked as mustprogress, this in itself makes the loop *not* dead. This patch moves the logic to check whether the current loop is finite or mustprogress to `isLoopDead` and also extends it to check the sub-loops. This should fix PR50511. Reviewed By: nikic Differential Revision: https://reviews.llvm.org/D103382
2021-06-01 13:20:46 +02:00
; CHECK-NEXT: br label [[LOOP1:%.*]]
; CHECK: loop1:
; CHECK-NEXT: br i1 [[C1:%.*]], label [[LOOP1_LATCH:%.*]], label [[LOOP2_PREHEADER:%.*]]
; CHECK: loop2.preheader:
; CHECK-NEXT: br label [[LOOP2:%.*]]
; CHECK: loop2:
; CHECK-NEXT: br i1 [[C2:%.*]], label [[LOOP1_LATCH_LOOPEXIT:%.*]], label [[LOOP2]]
; CHECK: loop1.latch.loopexit:
; CHECK-NEXT: br label [[LOOP1_LATCH]]
; CHECK: loop1.latch:
; CHECK-NEXT: br i1 false, label [[LOOP1_LATCH_LOOP1_CRIT_EDGE:%.*]], label [[EXIT:%.*]]
; CHECK: loop1.latch.loop1_crit_edge:
; CHECK-NEXT: unreachable
[LoopDeletion] Add test with potentially infinite sub-loop. Tests for PR50511.
2021-05-28 17:58:59 +02:00
; CHECK: exit:
; CHECK-NEXT: ret void
;
br label %loop1
loop1:
br i1 %c1, label %loop1.latch, label %loop2
loop2:
br i1 %c2, label %loop1.latch, label %loop2
loop1.latch:
br i1 false, label %loop1, label %exit
exit:
ret void
}
[LoopDeletion] Add more tests with infinite sub-loops & mustprogress. A couple of additional tests inspired by PR50511.
2021-05-30 17:34:30 +02:00
; Similar to @inner_loop_may_be_infinite, but the function is marked as
; mustprogress. The loops can be removed.
define void @inner_loop_may_be_infinite_but_fn_mustprogress(i1 %c1, i1 %c2) mustprogress {
; CHECK-LABEL: @inner_loop_may_be_infinite_but_fn_mustprogress(
[LoopDeletion] Add test with potentially infinite sub-loop. Tests for PR50511.
2021-05-28 17:58:59 +02:00
; CHECK-NEXT: br label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
br label %loop1
loop1:
br i1 %c1, label %loop1.latch, label %loop2
loop2:
br i1 %c2, label %loop1.latch, label %loop2
loop1.latch:
br i1 false, label %loop1, label %exit
exit:
ret void
}
[LoopDeletion] Add more tests with infinite sub-loops & mustprogress. A couple of additional tests inspired by PR50511.
2021-05-30 17:34:30 +02:00
; Similar to @inner_loop_may_be_infinite, but the parent loop loop1 is marked
; as mustprogress. The loops can be removed.
[LoopDeletion] Add additional test cases with more nested loops. Also remove mustprogress function attribute from one of the tests Extends test coverage for D103382.
2021-05-31 21:23:43 +02:00
define void @inner_loop_may_be_infinite_but_top_loop_mustprogress(i1 %c1, i1 %c2) {
[LoopDeletion] Add more tests with infinite sub-loops & mustprogress. A couple of additional tests inspired by PR50511.
2021-05-30 17:34:30 +02:00
; CHECK-LABEL: @inner_loop_may_be_infinite_but_top_loop_mustprogress(
; CHECK-NEXT: br label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
br label %loop1
loop1:
br i1 %c1, label %loop1.latch, label %loop2
loop2:
br i1 %c2, label %loop1.latch, label %loop2
loop1.latch:
br i1 false, label %loop1, label %exit, !llvm.loop !3
exit:
ret void
}
; loop3 and loop1 are not marked as mustprogress, but loop2 (parent of loop3)
; is. The loops can be removed.
define void @loop2_mustprogress_but_not_child_loop_loop3(i1 %c1, i1 %c2, i1 %c3) {
; CHECK-LABEL: @loop2_mustprogress_but_not_child_loop_loop3(
; CHECK-NEXT: br label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
br label %loop1
loop1:
br i1 %c1, label %loop1.latch, label %loop2
loop2:
br label %loop3
loop3:
br i1 %c2, label %loop2.latch, label %loop3
loop2.latch:
br i1 %c3, label %loop1.latch, label %loop2, !llvm.loop !3
loop1.latch:
br i1 false, label %loop1, label %exit
exit:
ret void
}
; Cannot remove loop3 or loop1, as they are not mustprogress. loop2 is
; mustprogress and can be removed.
define void @loop2_mustprogress_but_not_sibling_loop(i1 %c1, i1 %c2, i1 %c3) {
; CHECK-LABEL: @loop2_mustprogress_but_not_sibling_loop(
[LoopDeletion] Consider infinite loops alive, unless mustprogress. The current loop or any of its sub-loops may be infinite. Unless the function or the loops are marked as mustprogress, this in itself makes the loop *not* dead. This patch moves the logic to check whether the current loop is finite or mustprogress to `isLoopDead` and also extends it to check the sub-loops. This should fix PR50511. Reviewed By: nikic Differential Revision: https://reviews.llvm.org/D103382
2021-06-01 13:20:46 +02:00
; CHECK-NEXT: br label [[LOOP1:%.*]]
; CHECK: loop1:
; CHECK-NEXT: br i1 [[C1:%.*]], label [[LOOP1_LATCH:%.*]], label [[LOOP2_PREHEADER:%.*]]
; CHECK: loop2.preheader:
; CHECK-NEXT: br label [[LOOP3_PREHEADER:%.*]]
; CHECK: loop3.preheader:
; CHECK-NEXT: br label [[LOOP3:%.*]]
; CHECK: loop3:
; CHECK-NEXT: br i1 [[C3:%.*]], label [[LOOP1_LATCH_LOOPEXIT:%.*]], label [[LOOP3]]
; CHECK: loop1.latch.loopexit:
; CHECK-NEXT: br label [[LOOP1_LATCH]]
; CHECK: loop1.latch:
; CHECK-NEXT: br i1 false, label [[LOOP1_LATCH_LOOP1_CRIT_EDGE:%.*]], label [[EXIT:%.*]]
; CHECK: loop1.latch.loop1_crit_edge:
; CHECK-NEXT: unreachable
[LoopDeletion] Add more tests with infinite sub-loops & mustprogress. A couple of additional tests inspired by PR50511.
2021-05-30 17:34:30 +02:00
; CHECK: exit:
; CHECK-NEXT: ret void
;
br label %loop1
loop1:
br i1 %c1, label %loop1.latch, label %loop2
loop2:
br i1 %c2, label %loop3, label %loop2, !llvm.loop !4
loop3:
[LoopDeletion] Add additional test cases with more nested loops. Also remove mustprogress function attribute from one of the tests Extends test coverage for D103382.
2021-05-31 21:23:43 +02:00
br i1 %c3, label %loop1.latch, label %loop3
[LoopDeletion] Add more tests with infinite sub-loops & mustprogress. A couple of additional tests inspired by PR50511.
2021-05-30 17:34:30 +02:00
loop1.latch:
br i1 false, label %loop1, label %exit
exit:
ret void
}
[LoopDeletion] Add additional test cases with more nested loops. Also remove mustprogress function attribute from one of the tests Extends test coverage for D103382.
2021-05-31 21:23:43 +02:00
define void @loop2_finite_but_child_is_not(i1 %c1, i1 %c2, i1 %c3) {
; CHECK-LABEL: @loop2_finite_but_child_is_not(
; CHECK-NEXT: entry:
[LoopDeletion] Consider infinite loops alive, unless mustprogress. The current loop or any of its sub-loops may be infinite. Unless the function or the loops are marked as mustprogress, this in itself makes the loop *not* dead. This patch moves the logic to check whether the current loop is finite or mustprogress to `isLoopDead` and also extends it to check the sub-loops. This should fix PR50511. Reviewed By: nikic Differential Revision: https://reviews.llvm.org/D103382
2021-06-01 13:20:46 +02:00
; CHECK-NEXT: br label [[LOOP1:%.*]]
; CHECK: loop1:
; CHECK-NEXT: [[IV1:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV1_NEXT:%.*]], [[LOOP1_LATCH:%.*]] ]
; CHECK-NEXT: br i1 [[C1:%.*]], label [[LOOP1_LATCH]], label [[LOOP2_PREHEADER:%.*]]
; CHECK: loop2.preheader:
; CHECK-NEXT: br label [[LOOP2:%.*]]
; CHECK: loop2:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[LOOP2_LATCH:%.*]] ], [ 0, [[LOOP2_PREHEADER]] ]
; CHECK-NEXT: br label [[LOOP3:%.*]]
; CHECK: loop3:
; CHECK-NEXT: br i1 [[C2:%.*]], label [[LOOP2_LATCH]], label [[LOOP3]]
; CHECK: loop2.latch:
; CHECK-NEXT: [[IV_NEXT]] = add nuw i32 [[IV]], 1
; CHECK-NEXT: [[C:%.*]] = icmp ugt i32 [[IV]], 200
; CHECK-NEXT: br i1 [[C]], label [[LOOP1_LATCH_LOOPEXIT:%.*]], label [[LOOP2]]
; CHECK: loop1.latch.loopexit:
; CHECK-NEXT: br label [[LOOP1_LATCH]]
; CHECK: loop1.latch:
; CHECK-NEXT: [[IV1_NEXT]] = add nuw i32 [[IV1]], 1
; CHECK-NEXT: [[C4:%.*]] = icmp ult i32 [[IV1_NEXT]], 200
; CHECK-NEXT: br i1 [[C4]], label [[LOOP1]], label [[EXIT:%.*]]
[LoopDeletion] Add additional test cases with more nested loops. Also remove mustprogress function attribute from one of the tests Extends test coverage for D103382.
2021-05-31 21:23:43 +02:00
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %loop1
loop1:
%iv1 = phi i32 [ 0, %entry ], [ %iv1.next, %loop1.latch ]
br i1 %c1, label %loop1.latch, label %loop2
loop2:
%iv = phi i32 [ 0, %loop1 ], [ %iv.next, %loop2.latch ]
br label %loop3
loop3:
br i1 %c2, label %loop2.latch, label %loop3
loop2.latch:
%iv.next = add nuw i32 %iv, 1
%c = icmp ugt i32 %iv, 200
br i1 %c, label %loop1.latch, label %loop2
loop1.latch:
%iv1.next = add nuw i32 %iv1, 1
%c4 = icmp ult i32 %iv1.next, 200
br i1 %c4, label %loop1, label %exit
exit:
ret void
}
define void @loop2_finite_and_child_is_mustprogress(i1 %c1, i1 %c2, i1 %c3) {
; CHECK-LABEL: @loop2_finite_and_child_is_mustprogress(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %loop1
loop1:
%iv1 = phi i32 [ 0, %entry ], [ %iv1.next, %loop1.latch ]
br i1 %c1, label %loop1.latch, label %loop2
loop2:
%iv = phi i32 [ 0, %loop1 ], [ %iv.next, %loop2.latch ]
br label %loop3
loop3:
br i1 %c2, label %loop2.latch, label %loop3, !llvm.loop !5
loop2.latch:
%iv.next = add nuw i32 %iv, 1
%c = icmp ugt i32 %iv, 200
br i1 %c, label %loop1.latch, label %loop2
loop1.latch:
%iv1.next = add nuw i32 %iv1, 1
%c4 = icmp ult i32 %iv1.next, 200
br i1 %c4, label %loop1, label %exit
exit:
ret void
}
[Tests] Add additional tests for incorrect willreturn handling (NFC) Highlight a few of the places that don't handle non-willreturn calls correctly right now.
2021-07-24 17:00:16 +02:00
; Inner infinite loop hidden behind a call.
define void @not_willreturn() {
; CHECK-LABEL: @not_willreturn(
; CHECK-NEXT: entry:
[IR] Consider non-willreturn as side effect (PR50511) This adjusts mayHaveSideEffect() to return true for !willReturn() instructions. Just like other side-effects, non-willreturn calls (aka "divergence") cannot be removed and cannot be reordered relative to other side effects. This fixes a number of bugs where non-willreturn calls are either incorrectly dropped or moved. In particular, it also fixes the last open problem in https://bugs.llvm.org/show_bug.cgi?id=50511. I performed a cursory review of all current mayHaveSideEffect() uses, which convinced me that these are indeed the desired default semantics. Places that do not want to consider non-willreturn as a sideeffect generally do not want mayHaveSideEffect() semantics at all. I identified two such cases, which are addressed by D106591 and D106742. Finally, there is a use in SCEV for which we don't really have an appropriate API right now -- what it wants is basically "would this be considered forward progress". I've just spelled out the previous semantics there. Differential Revision: https://reviews.llvm.org/D106749
2021-07-21 21:23:38 +02:00
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: call void @sideeffect() #[[ATTR2:[0-9]+]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw i32 [[IV]], 1
; CHECK-NEXT: [[C:%.*]] = icmp ult i32 [[IV]], 100
; CHECK-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT:%.*]]
[Tests] Add additional tests for incorrect willreturn handling (NFC) Highlight a few of the places that don't handle non-willreturn calls correctly right now.
2021-07-24 17:00:16 +02:00
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]
call void @sideeffect() nounwind readonly
%iv.next = add nuw i32 %iv, 1
%c = icmp ult i32 %iv, 100
br i1 %c, label %loop, label %exit
exit:
ret void
}
[LoopDeletion] Add more tests with infinite sub-loops & mustprogress. A couple of additional tests inspired by PR50511.
2021-05-30 17:34:30 +02:00
!1 = !{!"llvm.loop.mustprogress"}
!2 = distinct !{!2, !1}
!3 = distinct !{!3, !1}
!4 = distinct !{!4, !1}
[LoopDeletion] Add additional test cases with more nested loops. Also remove mustprogress function attribute from one of the tests Extends test coverage for D103382.
2021-05-31 21:23:43 +02:00
!5 = distinct !{!5, !1}
Reference in New Issue Copy Permalink