Google Git
Sign in
llvm / llvm-project / llvm / 3bc8af8352c3783ae4d3cd3ca6bc4237aecf79a1 / . / test / Transforms / IRCE / stride_more_than_1.ll
blob: da6b5603e92600bf50bbf1808768a3f6829e29c3 [file] [log] [blame]
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
; RUN: opt -verify-loop-info -irce-print-changed-loops -passes=irce -S < %s 2>&1 | FileCheck %s
; RUN: opt -verify-loop-info -irce-print-changed-loops -passes='require<branch-prob>,irce' -S < %s 2>&1 | FileCheck %s
; CHECK: irce: in function test_01: constrained Loop at depth 1 containing: %loop<header><exiting>,%in.bounds<latch><exiting>
; CHECK: irce: in function test_02: constrained Loop at depth 1 containing: %loop<header><exiting>,%in.bounds<latch><exiting>
; CHECK: irce: in function test_03: constrained Loop at depth 1 containing: %loop<header><exiting>,%in.bounds<latch><exiting>
; CHECK: irce: in function test_04: constrained Loop at depth 1 containing: %loop<header><exiting>,%in.bounds<latch><exiting>
; CHECK: irce: in function test_05: constrained Loop at depth 1 containing: %loop<header><exiting>,%in.bounds<latch><exiting>
; CHECK: irce: in function test_06: constrained Loop at depth 1 containing: %loop<header><exiting>,%in.bounds<latch><exiting>
; CHECK-NOT: irce: in function test_07: constrained Loop at depth 1 containing: %loop<header><exiting>,%in.bounds<latch><exiting>
; CHECK: irce: in function test_08: constrained Loop at depth 1 containing: %loop<header><exiting>,%in.bounds<latch><exiting>
; IV = 0; IV <s 100; IV += 7; 0 <= Len <= 50. IRCE is allowed.
define void @test_01(ptr %arr, ptr %a_len_ptr) {
; CHECK-LABEL: define void @test_01
; CHECK-SAME: (ptr [[ARR:%.*]], ptr [[A_LEN_PTR:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[EXIT_MAINLOOP_AT:%.*]] = load i32, ptr [[A_LEN_PTR]], align 4, !range [[RNG0:![0-9]+]]
; CHECK-NEXT: [[TMP0:%.*]] = icmp slt i32 0, [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP0]], label [[LOOP_PREHEADER:%.*]], label [[MAIN_PSEUDO_EXIT:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IDX:%.*]] = phi i32 [ [[IDX_NEXT:%.*]], [[IN_BOUNDS:%.*]] ], [ 0, [[LOOP_PREHEADER]] ]
; CHECK-NEXT: [[IDX_NEXT]] = add nsw i32 [[IDX]], 7
; CHECK-NEXT: [[ABC:%.*]] = icmp slt i32 [[IDX]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 true, label [[IN_BOUNDS]], label [[OUT_OF_BOUNDS_LOOPEXIT1:%.*]]
; CHECK: in.bounds:
; CHECK-NEXT: [[ADDR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX]]
; CHECK-NEXT: store i32 0, ptr [[ADDR]], align 4
; CHECK-NEXT: [[NEXT:%.*]] = icmp slt i32 [[IDX_NEXT]], 100
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt i32 [[IDX_NEXT]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP1]], label [[LOOP]], label [[MAIN_EXIT_SELECTOR:%.*]]
; CHECK: main.exit.selector:
; CHECK-NEXT: [[IDX_NEXT_LCSSA:%.*]] = phi i32 [ [[IDX_NEXT]], [[IN_BOUNDS]] ]
; CHECK-NEXT: [[TMP2:%.*]] = icmp slt i32 [[IDX_NEXT_LCSSA]], 100
; CHECK-NEXT: br i1 [[TMP2]], label [[MAIN_PSEUDO_EXIT]], label [[EXIT:%.*]]
; CHECK: main.pseudo.exit:
; CHECK-NEXT: [[IDX_COPY:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IDX_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: [[INDVAR_END:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[IDX_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: br label [[POSTLOOP:%.*]]
; CHECK: out.of.bounds.loopexit:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS:%.*]]
; CHECK: out.of.bounds.loopexit1:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS]]
; CHECK: out.of.bounds:
; CHECK-NEXT: ret void
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT]]
; CHECK: exit:
; CHECK-NEXT: ret void
; CHECK: postloop:
; CHECK-NEXT: br label [[LOOP_POSTLOOP:%.*]]
; CHECK: loop.postloop:
; CHECK-NEXT: [[IDX_POSTLOOP:%.*]] = phi i32 [ [[IDX_COPY]], [[POSTLOOP]] ], [ [[IDX_NEXT_POSTLOOP:%.*]], [[IN_BOUNDS_POSTLOOP:%.*]] ]
; CHECK-NEXT: [[IDX_NEXT_POSTLOOP]] = add i32 [[IDX_POSTLOOP]], 7
; CHECK-NEXT: [[ABC_POSTLOOP:%.*]] = icmp slt i32 [[IDX_POSTLOOP]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[ABC_POSTLOOP]], label [[IN_BOUNDS_POSTLOOP]], label [[OUT_OF_BOUNDS_LOOPEXIT:%.*]]
; CHECK: in.bounds.postloop:
; CHECK-NEXT: [[ADDR_POSTLOOP:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX_POSTLOOP]]
; CHECK-NEXT: store i32 0, ptr [[ADDR_POSTLOOP]], align 4
; CHECK-NEXT: [[NEXT_POSTLOOP:%.*]] = icmp slt i32 [[IDX_NEXT_POSTLOOP]], 100
; CHECK-NEXT: br i1 [[NEXT_POSTLOOP]], label [[LOOP_POSTLOOP]], label [[EXIT_LOOPEXIT:%.*]], !llvm.loop [[LOOP1:![0-9]+]], !irce.loop.clone !6
;
entry:
%len = load i32, ptr %a_len_ptr, !range !0
br label %loop
loop:
%idx = phi i32 [ 0, %entry ], [ %idx.next, %in.bounds ]
%idx.next = add i32 %idx, 7
%abc = icmp slt i32 %idx, %len
br i1 %abc, label %in.bounds, label %out.of.bounds
in.bounds:
%addr = getelementptr i32, ptr %arr, i32 %idx
store i32 0, ptr %addr
%next = icmp slt i32 %idx.next, 100
br i1 %next, label %loop, label %exit
out.of.bounds:
ret void
exit:
ret void
}
; IV = 0; IV <s MAX_INT - 7; IV += 7; 0 <= Len <= 50. IRCE is allowed.
define void @test_02(ptr %arr, ptr %a_len_ptr) {
; CHECK-LABEL: define void @test_02
; CHECK-SAME: (ptr [[ARR:%.*]], ptr [[A_LEN_PTR:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[EXIT_MAINLOOP_AT:%.*]] = load i32, ptr [[A_LEN_PTR]], align 4, !range [[RNG0]]
; CHECK-NEXT: [[TMP0:%.*]] = icmp slt i32 0, [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP0]], label [[LOOP_PREHEADER:%.*]], label [[MAIN_PSEUDO_EXIT:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IDX:%.*]] = phi i32 [ [[IDX_NEXT:%.*]], [[IN_BOUNDS:%.*]] ], [ 0, [[LOOP_PREHEADER]] ]
; CHECK-NEXT: [[IDX_NEXT]] = add nsw i32 [[IDX]], 7
; CHECK-NEXT: [[ABC:%.*]] = icmp slt i32 [[IDX]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 true, label [[IN_BOUNDS]], label [[OUT_OF_BOUNDS_LOOPEXIT1:%.*]]
; CHECK: in.bounds:
; CHECK-NEXT: [[ADDR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX]]
; CHECK-NEXT: store i32 0, ptr [[ADDR]], align 4
; CHECK-NEXT: [[NEXT:%.*]] = icmp slt i32 [[IDX_NEXT]], 2147483640
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt i32 [[IDX_NEXT]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP1]], label [[LOOP]], label [[MAIN_EXIT_SELECTOR:%.*]]
; CHECK: main.exit.selector:
; CHECK-NEXT: [[IDX_NEXT_LCSSA:%.*]] = phi i32 [ [[IDX_NEXT]], [[IN_BOUNDS]] ]
; CHECK-NEXT: [[TMP2:%.*]] = icmp slt i32 [[IDX_NEXT_LCSSA]], 2147483640
; CHECK-NEXT: br i1 [[TMP2]], label [[MAIN_PSEUDO_EXIT]], label [[EXIT:%.*]]
; CHECK: main.pseudo.exit:
; CHECK-NEXT: [[IDX_COPY:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IDX_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: [[INDVAR_END:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[IDX_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: br label [[POSTLOOP:%.*]]
; CHECK: out.of.bounds.loopexit:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS:%.*]]
; CHECK: out.of.bounds.loopexit1:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS]]
; CHECK: out.of.bounds:
; CHECK-NEXT: ret void
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT]]
; CHECK: exit:
; CHECK-NEXT: ret void
; CHECK: postloop:
; CHECK-NEXT: br label [[LOOP_POSTLOOP:%.*]]
; CHECK: loop.postloop:
; CHECK-NEXT: [[IDX_POSTLOOP:%.*]] = phi i32 [ [[IDX_COPY]], [[POSTLOOP]] ], [ [[IDX_NEXT_POSTLOOP:%.*]], [[IN_BOUNDS_POSTLOOP:%.*]] ]
; CHECK-NEXT: [[IDX_NEXT_POSTLOOP]] = add i32 [[IDX_POSTLOOP]], 7
; CHECK-NEXT: [[ABC_POSTLOOP:%.*]] = icmp slt i32 [[IDX_POSTLOOP]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[ABC_POSTLOOP]], label [[IN_BOUNDS_POSTLOOP]], label [[OUT_OF_BOUNDS_LOOPEXIT:%.*]]
; CHECK: in.bounds.postloop:
; CHECK-NEXT: [[ADDR_POSTLOOP:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX_POSTLOOP]]
; CHECK-NEXT: store i32 0, ptr [[ADDR_POSTLOOP]], align 4
; CHECK-NEXT: [[NEXT_POSTLOOP:%.*]] = icmp slt i32 [[IDX_NEXT_POSTLOOP]], 2147483640
; CHECK-NEXT: br i1 [[NEXT_POSTLOOP]], label [[LOOP_POSTLOOP]], label [[EXIT_LOOPEXIT:%.*]], !llvm.loop [[LOOP7:![0-9]+]], !irce.loop.clone !6
;
entry:
%len = load i32, ptr %a_len_ptr, !range !0
br label %loop
loop:
%idx = phi i32 [ 0, %entry ], [ %idx.next, %in.bounds ]
%idx.next = add i32 %idx, 7
%abc = icmp slt i32 %idx, %len
br i1 %abc, label %in.bounds, label %out.of.bounds
in.bounds:
%addr = getelementptr i32, ptr %arr, i32 %idx
store i32 0, ptr %addr
%next = icmp slt i32 %idx.next, 2147483640
br i1 %next, label %loop, label %exit
out.of.bounds:
ret void
exit:
ret void
}
; IV = 0; IV <s MAX_INT; IV += 7; 0 <= Len <= MAX_INT - 7. This is the greatest
; value of Len for which IRCE is allowed.
define void @test_03(ptr %arr, ptr %a_len_ptr) {
; CHECK-LABEL: define void @test_03
; CHECK-SAME: (ptr [[ARR:%.*]], ptr [[A_LEN_PTR:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[EXIT_MAINLOOP_AT:%.*]] = load i32, ptr [[A_LEN_PTR]], align 4, !range [[RNG8:![0-9]+]]
; CHECK-NEXT: [[TMP0:%.*]] = icmp slt i32 0, [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP0]], label [[LOOP_PREHEADER:%.*]], label [[MAIN_PSEUDO_EXIT:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IDX:%.*]] = phi i32 [ [[IDX_NEXT:%.*]], [[IN_BOUNDS:%.*]] ], [ 0, [[LOOP_PREHEADER]] ]
; CHECK-NEXT: [[IDX_NEXT]] = add nsw i32 [[IDX]], 7
; CHECK-NEXT: [[ABC:%.*]] = icmp slt i32 [[IDX]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 true, label [[IN_BOUNDS]], label [[OUT_OF_BOUNDS_LOOPEXIT1:%.*]]
; CHECK: in.bounds:
; CHECK-NEXT: [[ADDR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX]]
; CHECK-NEXT: store i32 0, ptr [[ADDR]], align 4
; CHECK-NEXT: [[NEXT:%.*]] = icmp slt i32 [[IDX_NEXT]], 2147483647
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt i32 [[IDX_NEXT]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP1]], label [[LOOP]], label [[MAIN_EXIT_SELECTOR:%.*]]
; CHECK: main.exit.selector:
; CHECK-NEXT: [[IDX_NEXT_LCSSA:%.*]] = phi i32 [ [[IDX_NEXT]], [[IN_BOUNDS]] ]
; CHECK-NEXT: [[TMP2:%.*]] = icmp slt i32 [[IDX_NEXT_LCSSA]], 2147483647
; CHECK-NEXT: br i1 [[TMP2]], label [[MAIN_PSEUDO_EXIT]], label [[EXIT:%.*]]
; CHECK: main.pseudo.exit:
; CHECK-NEXT: [[IDX_COPY:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IDX_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: [[INDVAR_END:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[IDX_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: br label [[POSTLOOP:%.*]]
; CHECK: out.of.bounds.loopexit:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS:%.*]]
; CHECK: out.of.bounds.loopexit1:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS]]
; CHECK: out.of.bounds:
; CHECK-NEXT: ret void
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT]]
; CHECK: exit:
; CHECK-NEXT: ret void
; CHECK: postloop:
; CHECK-NEXT: br label [[LOOP_POSTLOOP:%.*]]
; CHECK: loop.postloop:
; CHECK-NEXT: [[IDX_POSTLOOP:%.*]] = phi i32 [ [[IDX_COPY]], [[POSTLOOP]] ], [ [[IDX_NEXT_POSTLOOP:%.*]], [[IN_BOUNDS_POSTLOOP:%.*]] ]
; CHECK-NEXT: [[IDX_NEXT_POSTLOOP]] = add i32 [[IDX_POSTLOOP]], 7
; CHECK-NEXT: [[ABC_POSTLOOP:%.*]] = icmp slt i32 [[IDX_POSTLOOP]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[ABC_POSTLOOP]], label [[IN_BOUNDS_POSTLOOP]], label [[OUT_OF_BOUNDS_LOOPEXIT:%.*]]
; CHECK: in.bounds.postloop:
; CHECK-NEXT: [[ADDR_POSTLOOP:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX_POSTLOOP]]
; CHECK-NEXT: store i32 0, ptr [[ADDR_POSTLOOP]], align 4
; CHECK-NEXT: [[NEXT_POSTLOOP:%.*]] = icmp slt i32 [[IDX_NEXT_POSTLOOP]], 2147483647
; CHECK-NEXT: br i1 [[NEXT_POSTLOOP]], label [[LOOP_POSTLOOP]], label [[EXIT_LOOPEXIT:%.*]], !llvm.loop [[LOOP9:![0-9]+]], !irce.loop.clone !6
;
entry:
%len = load i32, ptr %a_len_ptr, !range !1
br label %loop
loop:
%idx = phi i32 [ 0, %entry ], [ %idx.next, %in.bounds ]
%idx.next = add i32 %idx, 7
%abc = icmp slt i32 %idx, %len
br i1 %abc, label %in.bounds, label %out.of.bounds
in.bounds:
%addr = getelementptr i32, ptr %arr, i32 %idx
store i32 0, ptr %addr
%next = icmp slt i32 %idx.next, 2147483647
br i1 %next, label %loop, label %exit
out.of.bounds:
ret void
exit:
ret void
}
; IV = 0; IV <s MAX_INT; IV += 7; 0 <= Len <= MAX_INT - 6. IRCE is allowed
; because the branch would fail once idx.next == MAX_INT - 1 keeping the
; access in bounds.
define void @test_04(ptr %arr, ptr %a_len_ptr) {
; CHECK-LABEL: define void @test_04
; CHECK-SAME: (ptr [[ARR:%.*]], ptr [[A_LEN_PTR:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[EXIT_MAINLOOP_AT:%.*]] = load i32, ptr [[A_LEN_PTR]], align 4, !range [[RNG10:![0-9]+]]
; CHECK-NEXT: [[TMP0:%.*]] = icmp slt i32 0, [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP0]], label [[LOOP_PREHEADER:%.*]], label [[MAIN_PSEUDO_EXIT:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IDX:%.*]] = phi i32 [ [[IDX_NEXT:%.*]], [[IN_BOUNDS:%.*]] ], [ 0, [[LOOP_PREHEADER]] ]
; CHECK-NEXT: [[IDX_NEXT]] = add nsw i32 [[IDX]], 7
; CHECK-NEXT: [[ABC:%.*]] = icmp slt i32 [[IDX]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 true, label [[IN_BOUNDS]], label [[OUT_OF_BOUNDS_LOOPEXIT1:%.*]]
; CHECK: in.bounds:
; CHECK-NEXT: [[ADDR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX]]
; CHECK-NEXT: store i32 0, ptr [[ADDR]], align 4
; CHECK-NEXT: [[NEXT:%.*]] = icmp slt i32 [[IDX_NEXT]], 2147483647
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt i32 [[IDX_NEXT]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP1]], label [[LOOP]], label [[MAIN_EXIT_SELECTOR:%.*]]
; CHECK: main.exit.selector:
; CHECK-NEXT: [[IDX_NEXT_LCSSA:%.*]] = phi i32 [ [[IDX_NEXT]], [[IN_BOUNDS]] ]
; CHECK-NEXT: [[TMP2:%.*]] = icmp slt i32 [[IDX_NEXT_LCSSA]], 2147483647
; CHECK-NEXT: br i1 [[TMP2]], label [[MAIN_PSEUDO_EXIT]], label [[EXIT:%.*]]
; CHECK: main.pseudo.exit:
; CHECK-NEXT: [[IDX_COPY:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IDX_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: [[INDVAR_END:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[IDX_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: br label [[POSTLOOP:%.*]]
; CHECK: out.of.bounds.loopexit:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS:%.*]]
; CHECK: out.of.bounds.loopexit1:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS]]
; CHECK: out.of.bounds:
; CHECK-NEXT: ret void
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT]]
; CHECK: exit:
; CHECK-NEXT: ret void
; CHECK: postloop:
; CHECK-NEXT: br label [[LOOP_POSTLOOP:%.*]]
; CHECK: loop.postloop:
; CHECK-NEXT: [[IDX_POSTLOOP:%.*]] = phi i32 [ [[IDX_COPY]], [[POSTLOOP]] ], [ [[IDX_NEXT_POSTLOOP:%.*]], [[IN_BOUNDS_POSTLOOP:%.*]] ]
; CHECK-NEXT: [[IDX_NEXT_POSTLOOP]] = add i32 [[IDX_POSTLOOP]], 7
; CHECK-NEXT: [[ABC_POSTLOOP:%.*]] = icmp slt i32 [[IDX_POSTLOOP]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[ABC_POSTLOOP]], label [[IN_BOUNDS_POSTLOOP]], label [[OUT_OF_BOUNDS_LOOPEXIT:%.*]]
; CHECK: in.bounds.postloop:
; CHECK-NEXT: [[ADDR_POSTLOOP:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX_POSTLOOP]]
; CHECK-NEXT: store i32 0, ptr [[ADDR_POSTLOOP]], align 4
; CHECK-NEXT: [[NEXT_POSTLOOP:%.*]] = icmp slt i32 [[IDX_NEXT_POSTLOOP]], 2147483647
; CHECK-NEXT: br i1 [[NEXT_POSTLOOP]], label [[LOOP_POSTLOOP]], label [[EXIT_LOOPEXIT:%.*]], !llvm.loop [[LOOP11:![0-9]+]], !irce.loop.clone !6
;
entry:
%len = load i32, ptr %a_len_ptr, !range !2
br label %loop
loop:
%idx = phi i32 [ 0, %entry ], [ %idx.next, %in.bounds ]
%idx.next = add i32 %idx, 7
%abc = icmp slt i32 %idx, %len
br i1 %abc, label %in.bounds, label %out.of.bounds
in.bounds:
%addr = getelementptr i32, ptr %arr, i32 %idx
store i32 0, ptr %addr
%next = icmp slt i32 %idx.next, 2147483647
br i1 %next, label %loop, label %exit
out.of.bounds:
ret void
exit:
ret void
}
; IV = 100; IV >s -1; IV -= 7; 0 <= Len <= 50. IRCE is allowed.
define void @test_05(ptr %arr, ptr %a_len_ptr) {
; CHECK-LABEL: define void @test_05
; CHECK-SAME: (ptr [[ARR:%.*]], ptr [[A_LEN_PTR:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[LEN:%.*]] = load i32, ptr [[A_LEN_PTR]], align 4, !range [[RNG0]]
; CHECK-NEXT: [[EXIT_PRELOOP_AT:%.*]] = add nsw i32 [[LEN]], -1
; CHECK-NEXT: [[TMP0:%.*]] = icmp sgt i32 100, [[EXIT_PRELOOP_AT]]
; CHECK-NEXT: br i1 [[TMP0]], label [[LOOP_PRELOOP_PREHEADER:%.*]], label [[PRELOOP_PSEUDO_EXIT:%.*]]
; CHECK: loop.preloop.preheader:
; CHECK-NEXT: br label [[LOOP_PRELOOP:%.*]]
; CHECK: mainloop:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IDX:%.*]] = phi i32 [ [[IDX_PRELOOP_COPY:%.*]], [[MAINLOOP:%.*]] ], [ [[IDX_NEXT:%.*]], [[IN_BOUNDS:%.*]] ]
; CHECK-NEXT: [[IDX_NEXT]] = add nsw i32 [[IDX]], -7
; CHECK-NEXT: [[ABC:%.*]] = icmp slt i32 [[IDX]], [[LEN]]
; CHECK-NEXT: br i1 true, label [[IN_BOUNDS]], label [[OUT_OF_BOUNDS_LOOPEXIT1:%.*]]
; CHECK: in.bounds:
; CHECK-NEXT: [[ADDR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX]]
; CHECK-NEXT: store i32 0, ptr [[ADDR]], align 4
; CHECK-NEXT: [[NEXT:%.*]] = icmp sgt i32 [[IDX_NEXT]], -1
; CHECK-NEXT: br i1 [[NEXT]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
; CHECK: out.of.bounds.loopexit:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS:%.*]]
; CHECK: out.of.bounds.loopexit1:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS]]
; CHECK: out.of.bounds:
; CHECK-NEXT: ret void
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
; CHECK: loop.preloop:
; CHECK-NEXT: [[IDX_PRELOOP:%.*]] = phi i32 [ [[IDX_NEXT_PRELOOP:%.*]], [[IN_BOUNDS_PRELOOP:%.*]] ], [ 100, [[LOOP_PRELOOP_PREHEADER]] ]
; CHECK-NEXT: [[IDX_NEXT_PRELOOP]] = add i32 [[IDX_PRELOOP]], -7
; CHECK-NEXT: [[ABC_PRELOOP:%.*]] = icmp slt i32 [[IDX_PRELOOP]], [[LEN]]
; CHECK-NEXT: br i1 [[ABC_PRELOOP]], label [[IN_BOUNDS_PRELOOP]], label [[OUT_OF_BOUNDS_LOOPEXIT:%.*]]
; CHECK: in.bounds.preloop:
; CHECK-NEXT: [[ADDR_PRELOOP:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX_PRELOOP]]
; CHECK-NEXT: store i32 0, ptr [[ADDR_PRELOOP]], align 4
; CHECK-NEXT: [[NEXT_PRELOOP:%.*]] = icmp sgt i32 [[IDX_NEXT_PRELOOP]], -1
; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt i32 [[IDX_NEXT_PRELOOP]], [[EXIT_PRELOOP_AT]]
; CHECK-NEXT: br i1 [[TMP1]], label [[LOOP_PRELOOP]], label [[PRELOOP_EXIT_SELECTOR:%.*]], !llvm.loop [[LOOP12:![0-9]+]], !irce.loop.clone !6
; CHECK: preloop.exit.selector:
; CHECK-NEXT: [[IDX_NEXT_PRELOOP_LCSSA:%.*]] = phi i32 [ [[IDX_NEXT_PRELOOP]], [[IN_BOUNDS_PRELOOP]] ]
; CHECK-NEXT: [[TMP2:%.*]] = icmp sgt i32 [[IDX_NEXT_PRELOOP_LCSSA]], -1
; CHECK-NEXT: br i1 [[TMP2]], label [[PRELOOP_PSEUDO_EXIT]], label [[EXIT]]
; CHECK: preloop.pseudo.exit:
; CHECK-NEXT: [[IDX_PRELOOP_COPY]] = phi i32 [ 100, [[ENTRY:%.*]] ], [ [[IDX_NEXT_PRELOOP_LCSSA]], [[PRELOOP_EXIT_SELECTOR]] ]
; CHECK-NEXT: [[INDVAR_END:%.*]] = phi i32 [ 100, [[ENTRY]] ], [ [[IDX_NEXT_PRELOOP_LCSSA]], [[PRELOOP_EXIT_SELECTOR]] ]
; CHECK-NEXT: br label [[MAINLOOP]]
;
entry:
%len = load i32, ptr %a_len_ptr, !range !0
br label %loop
loop:
%idx = phi i32 [ 100, %entry ], [ %idx.next, %in.bounds ]
%idx.next = add i32 %idx, -7
%abc = icmp slt i32 %idx, %len
br i1 %abc, label %in.bounds, label %out.of.bounds
in.bounds:
%addr = getelementptr i32, ptr %arr, i32 %idx
store i32 0, ptr %addr
%next = icmp sgt i32 %idx.next, -1
br i1 %next, label %loop, label %exit
out.of.bounds:
ret void
exit:
ret void
}
; IV = MAX_INT - 7; IV >u 6; IV -= 7; 10 <= Len <= 50. IRCE is allowed.
define void @test_06(ptr %arr, ptr %a_len_ptr) {
; CHECK-LABEL: define void @test_06
; CHECK-SAME: (ptr [[ARR:%.*]], ptr [[A_LEN_PTR:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[LEN:%.*]] = load i32, ptr [[A_LEN_PTR]], align 4, !range [[RNG13:![0-9]+]]
; CHECK-NEXT: [[EXIT_PRELOOP_AT:%.*]] = add nsw i32 [[LEN]], -1
; CHECK-NEXT: [[TMP0:%.*]] = icmp ugt i32 2147483640, [[EXIT_PRELOOP_AT]]
; CHECK-NEXT: br i1 [[TMP0]], label [[LOOP_PRELOOP_PREHEADER:%.*]], label [[PRELOOP_PSEUDO_EXIT:%.*]]
; CHECK: loop.preloop.preheader:
; CHECK-NEXT: br label [[LOOP_PRELOOP:%.*]]
; CHECK: mainloop:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IDX:%.*]] = phi i32 [ [[IDX_PRELOOP_COPY:%.*]], [[MAINLOOP:%.*]] ], [ [[IDX_NEXT:%.*]], [[IN_BOUNDS:%.*]] ]
; CHECK-NEXT: [[IDX_NEXT]] = add i32 [[IDX]], -7
; CHECK-NEXT: [[ABC:%.*]] = icmp slt i32 [[IDX]], [[LEN]]
; CHECK-NEXT: br i1 true, label [[IN_BOUNDS]], label [[OUT_OF_BOUNDS_LOOPEXIT1:%.*]]
; CHECK: in.bounds:
; CHECK-NEXT: [[ADDR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX]]
; CHECK-NEXT: store i32 0, ptr [[ADDR]], align 4
; CHECK-NEXT: [[NEXT:%.*]] = icmp ugt i32 [[IDX_NEXT]], 6
; CHECK-NEXT: br i1 [[NEXT]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
; CHECK: out.of.bounds.loopexit:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS:%.*]]
; CHECK: out.of.bounds.loopexit1:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS]]
; CHECK: out.of.bounds:
; CHECK-NEXT: ret void
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
; CHECK: loop.preloop:
; CHECK-NEXT: [[IDX_PRELOOP:%.*]] = phi i32 [ [[IDX_NEXT_PRELOOP:%.*]], [[IN_BOUNDS_PRELOOP:%.*]] ], [ 2147483640, [[LOOP_PRELOOP_PREHEADER]] ]
; CHECK-NEXT: [[IDX_NEXT_PRELOOP]] = add i32 [[IDX_PRELOOP]], -7
; CHECK-NEXT: [[ABC_PRELOOP:%.*]] = icmp slt i32 [[IDX_PRELOOP]], [[LEN]]
; CHECK-NEXT: br i1 [[ABC_PRELOOP]], label [[IN_BOUNDS_PRELOOP]], label [[OUT_OF_BOUNDS_LOOPEXIT:%.*]]
; CHECK: in.bounds.preloop:
; CHECK-NEXT: [[ADDR_PRELOOP:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX_PRELOOP]]
; CHECK-NEXT: store i32 0, ptr [[ADDR_PRELOOP]], align 4
; CHECK-NEXT: [[NEXT_PRELOOP:%.*]] = icmp ugt i32 [[IDX_NEXT_PRELOOP]], 6
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i32 [[IDX_NEXT_PRELOOP]], [[EXIT_PRELOOP_AT]]
; CHECK-NEXT: br i1 [[TMP1]], label [[LOOP_PRELOOP]], label [[PRELOOP_EXIT_SELECTOR:%.*]], !llvm.loop [[LOOP14:![0-9]+]], !irce.loop.clone !6
; CHECK: preloop.exit.selector:
; CHECK-NEXT: [[IDX_NEXT_PRELOOP_LCSSA:%.*]] = phi i32 [ [[IDX_NEXT_PRELOOP]], [[IN_BOUNDS_PRELOOP]] ]
; CHECK-NEXT: [[TMP2:%.*]] = icmp ugt i32 [[IDX_NEXT_PRELOOP_LCSSA]], 6
; CHECK-NEXT: br i1 [[TMP2]], label [[PRELOOP_PSEUDO_EXIT]], label [[EXIT]]
; CHECK: preloop.pseudo.exit:
; CHECK-NEXT: [[IDX_PRELOOP_COPY]] = phi i32 [ 2147483640, [[ENTRY:%.*]] ], [ [[IDX_NEXT_PRELOOP_LCSSA]], [[PRELOOP_EXIT_SELECTOR]] ]
; CHECK-NEXT: [[INDVAR_END:%.*]] = phi i32 [ 2147483640, [[ENTRY]] ], [ [[IDX_NEXT_PRELOOP_LCSSA]], [[PRELOOP_EXIT_SELECTOR]] ]
; CHECK-NEXT: br label [[MAINLOOP]]
;
entry:
%len = load i32, ptr %a_len_ptr, !range !3
br label %loop
loop:
%idx = phi i32 [ 2147483640, %entry ], [ %idx.next, %in.bounds ]
%idx.next = add i32 %idx, -7
%abc = icmp slt i32 %idx, %len
br i1 %abc, label %in.bounds, label %out.of.bounds
in.bounds:
%addr = getelementptr i32, ptr %arr, i32 %idx
store i32 0, ptr %addr
%next = icmp ugt i32 %idx.next, 6
br i1 %next, label %loop, label %exit
out.of.bounds:
ret void
exit:
ret void
}
; IV = MAX_INT - 7; IV >u 5; IV -= 7; 10 <= Len <= 50. IRCE is not allowed,
; because we can cross the 0 border.
define void @test_07(ptr %arr, ptr %a_len_ptr) {
; CHECK-LABEL: define void @test_07
; CHECK-SAME: (ptr [[ARR:%.*]], ptr [[A_LEN_PTR:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[LEN:%.*]] = load i32, ptr [[A_LEN_PTR]], align 4, !range [[RNG13]]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IDX:%.*]] = phi i32 [ 2147483640, [[ENTRY:%.*]] ], [ [[IDX_NEXT:%.*]], [[IN_BOUNDS:%.*]] ]
; CHECK-NEXT: [[IDX_NEXT]] = add i32 [[IDX]], -7
; CHECK-NEXT: [[ABC:%.*]] = icmp slt i32 [[IDX]], [[LEN]]
; CHECK-NEXT: br i1 [[ABC]], label [[IN_BOUNDS]], label [[OUT_OF_BOUNDS:%.*]]
; CHECK: in.bounds:
; CHECK-NEXT: [[ADDR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX]]
; CHECK-NEXT: store i32 0, ptr [[ADDR]], align 4
; CHECK-NEXT: [[NEXT:%.*]] = icmp ugt i32 [[IDX_NEXT]], 5
; CHECK-NEXT: br i1 [[NEXT]], label [[LOOP]], label [[EXIT:%.*]]
; CHECK: out.of.bounds:
; CHECK-NEXT: ret void
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
%len = load i32, ptr %a_len_ptr, !range !3
br label %loop
loop:
%idx = phi i32 [ 2147483640, %entry ], [ %idx.next, %in.bounds ]
%idx.next = add i32 %idx, -7
%abc = icmp slt i32 %idx, %len
br i1 %abc, label %in.bounds, label %out.of.bounds
in.bounds:
%addr = getelementptr i32, ptr %arr, i32 %idx
store i32 0, ptr %addr
%next = icmp ugt i32 %idx.next, 5
br i1 %next, label %loop, label %exit
out.of.bounds:
ret void
exit:
ret void
}
; IV = MAX_INT; IV >u 6; IV -= 7; 10 <= Len <= 50. IRCE is allowed.
define void @test_08(ptr %arr, ptr %a_len_ptr) {
; CHECK-LABEL: define void @test_08
; CHECK-SAME: (ptr [[ARR:%.*]], ptr [[A_LEN_PTR:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[LEN:%.*]] = load i32, ptr [[A_LEN_PTR]], align 4, !range [[RNG13]]
; CHECK-NEXT: [[EXIT_PRELOOP_AT:%.*]] = add nsw i32 [[LEN]], -1
; CHECK-NEXT: [[TMP0:%.*]] = icmp ugt i32 2147483647, [[EXIT_PRELOOP_AT]]
; CHECK-NEXT: br i1 [[TMP0]], label [[LOOP_PRELOOP_PREHEADER:%.*]], label [[PRELOOP_PSEUDO_EXIT:%.*]]
; CHECK: loop.preloop.preheader:
; CHECK-NEXT: br label [[LOOP_PRELOOP:%.*]]
; CHECK: mainloop:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IDX:%.*]] = phi i32 [ [[IDX_PRELOOP_COPY:%.*]], [[MAINLOOP:%.*]] ], [ [[IDX_NEXT:%.*]], [[IN_BOUNDS:%.*]] ]
; CHECK-NEXT: [[IDX_NEXT]] = add i32 [[IDX]], -7
; CHECK-NEXT: [[ABC:%.*]] = icmp slt i32 [[IDX]], [[LEN]]
; CHECK-NEXT: br i1 true, label [[IN_BOUNDS]], label [[OUT_OF_BOUNDS_LOOPEXIT1:%.*]]
; CHECK: in.bounds:
; CHECK-NEXT: [[ADDR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX]]
; CHECK-NEXT: store i32 0, ptr [[ADDR]], align 4
; CHECK-NEXT: [[NEXT:%.*]] = icmp ugt i32 [[IDX_NEXT]], 6
; CHECK-NEXT: br i1 [[NEXT]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
; CHECK: out.of.bounds.loopexit:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS:%.*]]
; CHECK: out.of.bounds.loopexit1:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS]]
; CHECK: out.of.bounds:
; CHECK-NEXT: ret void
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
; CHECK: loop.preloop:
; CHECK-NEXT: [[IDX_PRELOOP:%.*]] = phi i32 [ [[IDX_NEXT_PRELOOP:%.*]], [[IN_BOUNDS_PRELOOP:%.*]] ], [ 2147483647, [[LOOP_PRELOOP_PREHEADER]] ]
; CHECK-NEXT: [[IDX_NEXT_PRELOOP]] = add i32 [[IDX_PRELOOP]], -7
; CHECK-NEXT: [[ABC_PRELOOP:%.*]] = icmp slt i32 [[IDX_PRELOOP]], [[LEN]]
; CHECK-NEXT: br i1 [[ABC_PRELOOP]], label [[IN_BOUNDS_PRELOOP]], label [[OUT_OF_BOUNDS_LOOPEXIT:%.*]]
; CHECK: in.bounds.preloop:
; CHECK-NEXT: [[ADDR_PRELOOP:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[IDX_PRELOOP]]
; CHECK-NEXT: store i32 0, ptr [[ADDR_PRELOOP]], align 4
; CHECK-NEXT: [[NEXT_PRELOOP:%.*]] = icmp ugt i32 [[IDX_NEXT_PRELOOP]], 6
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i32 [[IDX_NEXT_PRELOOP]], [[EXIT_PRELOOP_AT]]
; CHECK-NEXT: br i1 [[TMP1]], label [[LOOP_PRELOOP]], label [[PRELOOP_EXIT_SELECTOR:%.*]], !llvm.loop [[LOOP15:![0-9]+]], !irce.loop.clone !6
; CHECK: preloop.exit.selector:
; CHECK-NEXT: [[IDX_NEXT_PRELOOP_LCSSA:%.*]] = phi i32 [ [[IDX_NEXT_PRELOOP]], [[IN_BOUNDS_PRELOOP]] ]
; CHECK-NEXT: [[TMP2:%.*]] = icmp ugt i32 [[IDX_NEXT_PRELOOP_LCSSA]], 6
; CHECK-NEXT: br i1 [[TMP2]], label [[PRELOOP_PSEUDO_EXIT]], label [[EXIT]]
; CHECK: preloop.pseudo.exit:
; CHECK-NEXT: [[IDX_PRELOOP_COPY]] = phi i32 [ 2147483647, [[ENTRY:%.*]] ], [ [[IDX_NEXT_PRELOOP_LCSSA]], [[PRELOOP_EXIT_SELECTOR]] ]
; CHECK-NEXT: [[INDVAR_END:%.*]] = phi i32 [ 2147483647, [[ENTRY]] ], [ [[IDX_NEXT_PRELOOP_LCSSA]], [[PRELOOP_EXIT_SELECTOR]] ]
; CHECK-NEXT: br label [[MAINLOOP]]
;
entry:
%len = load i32, ptr %a_len_ptr, !range !3
br label %loop
loop:
%idx = phi i32 [ 2147483647, %entry ], [ %idx.next, %in.bounds ]
%idx.next = add i32 %idx, -7
%abc = icmp slt i32 %idx, %len
br i1 %abc, label %in.bounds, label %out.of.bounds
in.bounds:
%addr = getelementptr i32, ptr %arr, i32 %idx
store i32 0, ptr %addr
%next = icmp ugt i32 %idx.next, 6
br i1 %next, label %loop, label %exit
out.of.bounds:
ret void
exit:
ret void
}
; IRCE is legal here.
; Here how it is done if the step was 1: https://godbolt.org/z/jEqWaseWc
; It is also legal for step 4. Proof:
; - Capacity check ensures that iv < limit <= SINT_MAX - 3, meaning that
; iv <= SINT_MAX - 4.
; - Because of this, iv.next is always computed w/o overflow.
; - The loop goes to backedge as long as iv < capacity - 3 && iv < num.elements - 4.
; - So iterating up to smin(capacity - 3, num.elements - 4) should be safe.
; - Proof by alive2: https://alive2.llvm.org/ce/z/vEhMxa
define i32 @test_09(ptr %p, ptr %capacity_p, ptr %num_elements_p) {
; CHECK-LABEL: define i32 @test_09
; CHECK-SAME: (ptr [[P:%.*]], ptr [[CAPACITY_P:%.*]], ptr [[NUM_ELEMENTS_P:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CAPACITY:%.*]] = load i32, ptr [[CAPACITY_P]], align 4, !range [[RNG16:![0-9]+]]
; CHECK-NEXT: [[NUM_ELEMENTS:%.*]] = load i32, ptr [[NUM_ELEMENTS_P]], align 4, !range [[RNG16]]
; CHECK-NEXT: [[LIMIT:%.*]] = sub i32 [[CAPACITY]], 3
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[CAPACITY]], -3
; CHECK-NEXT: [[TMP1:%.*]] = add nuw i32 [[CAPACITY]], 2147483646
; CHECK-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[TMP1]], i32 0)
; CHECK-NEXT: [[TMP2:%.*]] = sub i32 [[TMP0]], [[SMAX]]
; CHECK-NEXT: [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[LIMIT]], i32 0)
; CHECK-NEXT: [[SMAX2:%.*]] = call i32 @llvm.smax.i32(i32 [[SMIN]], i32 -1)
; CHECK-NEXT: [[TMP3:%.*]] = add nsw i32 [[SMAX2]], 1
; CHECK-NEXT: [[TMP4:%.*]] = mul i32 [[TMP2]], [[TMP3]]
; CHECK-NEXT: [[SMIN3:%.*]] = call i32 @llvm.smin.i32(i32 [[NUM_ELEMENTS]], i32 [[TMP4]])
; CHECK-NEXT: [[EXIT_MAINLOOP_AT:%.*]] = call i32 @llvm.smax.i32(i32 [[SMIN3]], i32 0)
; CHECK-NEXT: [[TMP5:%.*]] = icmp slt i32 0, [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP5]], label [[LOOP_PREHEADER:%.*]], label [[MAIN_PSEUDO_EXIT:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[BACKEDGE:%.*]] ], [ 0, [[LOOP_PREHEADER]] ]
; CHECK-NEXT: [[CAPACITY_CHECK:%.*]] = icmp slt i32 [[IV]], [[LIMIT]]
; CHECK-NEXT: br i1 true, label [[BACKEDGE]], label [[OUT_OF_BOUNDS_LOOPEXIT5:%.*]], !prof [[PROF17:![0-9]+]]
; CHECK: backedge:
; CHECK-NEXT: [[IV_WIDE:%.*]] = zext i32 [[IV]] to i64
; CHECK-NEXT: [[EL_PTR:%.*]] = getelementptr i32, ptr [[P]], i64 [[IV_WIDE]]
; CHECK-NEXT: store i32 1, ptr [[EL_PTR]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i32 [[IV]], 4
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV_NEXT]], [[NUM_ELEMENTS]]
; CHECK-NEXT: [[TMP6:%.*]] = icmp slt i32 [[IV_NEXT]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP6]], label [[LOOP]], label [[MAIN_EXIT_SELECTOR:%.*]]
; CHECK: main.exit.selector:
; CHECK-NEXT: [[IV_NEXT_LCSSA:%.*]] = phi i32 [ [[IV_NEXT]], [[BACKEDGE]] ]
; CHECK-NEXT: [[IV_LCSSA:%.*]] = phi i32 [ [[IV]], [[BACKEDGE]] ]
; CHECK-NEXT: [[TMP7:%.*]] = icmp slt i32 [[IV_NEXT_LCSSA]], [[NUM_ELEMENTS]]
; CHECK-NEXT: br i1 [[TMP7]], label [[MAIN_PSEUDO_EXIT]], label [[EXIT:%.*]]
; CHECK: main.pseudo.exit:
; CHECK-NEXT: [[IV_COPY:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: [[INDVAR_END:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[IV_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: br label [[POSTLOOP:%.*]]
; CHECK: exit.loopexit:
; CHECK-NEXT: [[IV_LCSSA1_PH:%.*]] = phi i32 [ [[IV_POSTLOOP:%.*]], [[BACKEDGE_POSTLOOP:%.*]] ]
; CHECK-NEXT: br label [[EXIT]]
; CHECK: exit:
; CHECK-NEXT: [[IV_LCSSA1:%.*]] = phi i32 [ [[IV_LCSSA]], [[MAIN_EXIT_SELECTOR]] ], [ [[IV_LCSSA1_PH]], [[EXIT_LOOPEXIT:%.*]] ]
; CHECK-NEXT: ret i32 [[IV_LCSSA1]]
; CHECK: out_of_bounds.loopexit:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS:%.*]]
; CHECK: out_of_bounds.loopexit5:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS]]
; CHECK: out_of_bounds:
; CHECK-NEXT: ret i32 -1
; CHECK: postloop:
; CHECK-NEXT: br label [[LOOP_POSTLOOP:%.*]]
; CHECK: loop.postloop:
; CHECK-NEXT: [[IV_POSTLOOP]] = phi i32 [ [[IV_COPY]], [[POSTLOOP]] ], [ [[IV_NEXT_POSTLOOP:%.*]], [[BACKEDGE_POSTLOOP]] ]
; CHECK-NEXT: [[CAPACITY_CHECK_POSTLOOP:%.*]] = icmp slt i32 [[IV_POSTLOOP]], [[LIMIT]]
; CHECK-NEXT: br i1 [[CAPACITY_CHECK_POSTLOOP]], label [[BACKEDGE_POSTLOOP]], label [[OUT_OF_BOUNDS_LOOPEXIT:%.*]], !prof [[PROF17]]
; CHECK: backedge.postloop:
; CHECK-NEXT: [[IV_WIDE_POSTLOOP:%.*]] = zext i32 [[IV_POSTLOOP]] to i64
; CHECK-NEXT: [[EL_PTR_POSTLOOP:%.*]] = getelementptr i32, ptr [[P]], i64 [[IV_WIDE_POSTLOOP]]
; CHECK-NEXT: store i32 1, ptr [[EL_PTR_POSTLOOP]], align 4
; CHECK-NEXT: [[IV_NEXT_POSTLOOP]] = add nuw nsw i32 [[IV_POSTLOOP]], 4
; CHECK-NEXT: [[LOOP_COND_POSTLOOP:%.*]] = icmp slt i32 [[IV_NEXT_POSTLOOP]], [[NUM_ELEMENTS]]
; CHECK-NEXT: br i1 [[LOOP_COND_POSTLOOP]], label [[LOOP_POSTLOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP18:![0-9]+]], !irce.loop.clone !6
;
entry:
%capacity = load i32, ptr %capacity_p, !range !4
%num_elements = load i32, ptr %num_elements_p, !range !4
%limit = sub i32 %capacity, 3
br label %loop
loop:
%iv = phi i32 [0, %entry], [%iv.next, %backedge]
%capacity_check = icmp slt i32 %iv, %limit
br i1 %capacity_check, label %backedge, label %out_of_bounds, !prof !5
backedge:
%iv.wide = zext i32 %iv to i64
%el.ptr = getelementptr i32, ptr %p, i64 %iv.wide
store i32 1, ptr %el.ptr
%iv.next = add nuw nsw i32 %iv, 4
%loop_cond = icmp slt i32 %iv.next, %num_elements
br i1 %loop_cond, label %loop, label %exit
exit:
ret i32 %iv
out_of_bounds:
ret i32 -1
}
; Same as test_09 but range check comparison is inversed.
; IRCE is allowed.
define i32 @test_10(ptr %p, ptr %capacity_p, ptr %num_elements_p) {
; CHECK-LABEL: define i32 @test_10
; CHECK-SAME: (ptr [[P:%.*]], ptr [[CAPACITY_P:%.*]], ptr [[NUM_ELEMENTS_P:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CAPACITY:%.*]] = load i32, ptr [[CAPACITY_P]], align 4, !range [[RNG16]]
; CHECK-NEXT: [[NUM_ELEMENTS:%.*]] = load i32, ptr [[NUM_ELEMENTS_P]], align 4, !range [[RNG16]]
; CHECK-NEXT: [[LIMIT:%.*]] = sub i32 [[CAPACITY]], 3
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[CAPACITY]], -3
; CHECK-NEXT: [[TMP1:%.*]] = add nuw i32 [[CAPACITY]], 2147483646
; CHECK-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[TMP1]], i32 0)
; CHECK-NEXT: [[TMP2:%.*]] = sub i32 [[TMP0]], [[SMAX]]
; CHECK-NEXT: [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[LIMIT]], i32 0)
; CHECK-NEXT: [[SMAX2:%.*]] = call i32 @llvm.smax.i32(i32 [[SMIN]], i32 -1)
; CHECK-NEXT: [[TMP3:%.*]] = add nsw i32 [[SMAX2]], 1
; CHECK-NEXT: [[TMP4:%.*]] = mul i32 [[TMP2]], [[TMP3]]
; CHECK-NEXT: [[SMIN3:%.*]] = call i32 @llvm.smin.i32(i32 [[NUM_ELEMENTS]], i32 [[TMP4]])
; CHECK-NEXT: [[EXIT_MAINLOOP_AT:%.*]] = call i32 @llvm.smax.i32(i32 [[SMIN3]], i32 0)
; CHECK-NEXT: [[TMP5:%.*]] = icmp slt i32 0, [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP5]], label [[LOOP_PREHEADER:%.*]], label [[MAIN_PSEUDO_EXIT:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[BACKEDGE:%.*]] ], [ 0, [[LOOP_PREHEADER]] ]
; CHECK-NEXT: [[CAPACITY_CHECK:%.*]] = icmp slt i32 [[IV]], [[LIMIT]]
; CHECK-NEXT: br i1 true, label [[BACKEDGE]], label [[OUT_OF_BOUNDS_LOOPEXIT5:%.*]], !prof [[PROF17]]
; CHECK: backedge:
; CHECK-NEXT: [[IV_WIDE:%.*]] = zext i32 [[IV]] to i64
; CHECK-NEXT: [[EL_PTR:%.*]] = getelementptr i32, ptr [[P]], i64 [[IV_WIDE]]
; CHECK-NEXT: store i32 1, ptr [[EL_PTR]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i32 [[IV]], 4
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV_NEXT]], [[NUM_ELEMENTS]]
; CHECK-NEXT: [[TMP6:%.*]] = icmp slt i32 [[IV_NEXT]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP6]], label [[LOOP]], label [[MAIN_EXIT_SELECTOR:%.*]]
; CHECK: main.exit.selector:
; CHECK-NEXT: [[IV_NEXT_LCSSA:%.*]] = phi i32 [ [[IV_NEXT]], [[BACKEDGE]] ]
; CHECK-NEXT: [[IV_LCSSA:%.*]] = phi i32 [ [[IV]], [[BACKEDGE]] ]
; CHECK-NEXT: [[TMP7:%.*]] = icmp slt i32 [[IV_NEXT_LCSSA]], [[NUM_ELEMENTS]]
; CHECK-NEXT: br i1 [[TMP7]], label [[MAIN_PSEUDO_EXIT]], label [[EXIT:%.*]]
; CHECK: main.pseudo.exit:
; CHECK-NEXT: [[IV_COPY:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: [[INDVAR_END:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[IV_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: br label [[POSTLOOP:%.*]]
; CHECK: exit.loopexit:
; CHECK-NEXT: [[IV_LCSSA1_PH:%.*]] = phi i32 [ [[IV_POSTLOOP:%.*]], [[BACKEDGE_POSTLOOP:%.*]] ]
; CHECK-NEXT: br label [[EXIT]]
; CHECK: exit:
; CHECK-NEXT: [[IV_LCSSA1:%.*]] = phi i32 [ [[IV_LCSSA]], [[MAIN_EXIT_SELECTOR]] ], [ [[IV_LCSSA1_PH]], [[EXIT_LOOPEXIT:%.*]] ]
; CHECK-NEXT: ret i32 [[IV_LCSSA1]]
; CHECK: out_of_bounds.loopexit:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS:%.*]]
; CHECK: out_of_bounds.loopexit5:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS]]
; CHECK: out_of_bounds:
; CHECK-NEXT: ret i32 -1
; CHECK: postloop:
; CHECK-NEXT: br label [[LOOP_POSTLOOP:%.*]]
; CHECK: loop.postloop:
; CHECK-NEXT: [[IV_POSTLOOP]] = phi i32 [ [[IV_COPY]], [[POSTLOOP]] ], [ [[IV_NEXT_POSTLOOP:%.*]], [[BACKEDGE_POSTLOOP]] ]
; CHECK-NEXT: [[CAPACITY_CHECK_POSTLOOP:%.*]] = icmp slt i32 [[IV_POSTLOOP]], [[LIMIT]]
; CHECK-NEXT: br i1 [[CAPACITY_CHECK_POSTLOOP]], label [[BACKEDGE_POSTLOOP]], label [[OUT_OF_BOUNDS_LOOPEXIT:%.*]], !prof [[PROF17]]
; CHECK: backedge.postloop:
; CHECK-NEXT: [[IV_WIDE_POSTLOOP:%.*]] = zext i32 [[IV_POSTLOOP]] to i64
; CHECK-NEXT: [[EL_PTR_POSTLOOP:%.*]] = getelementptr i32, ptr [[P]], i64 [[IV_WIDE_POSTLOOP]]
; CHECK-NEXT: store i32 1, ptr [[EL_PTR_POSTLOOP]], align 4
; CHECK-NEXT: [[IV_NEXT_POSTLOOP]] = add nuw nsw i32 [[IV_POSTLOOP]], 4
; CHECK-NEXT: [[LOOP_COND_POSTLOOP:%.*]] = icmp slt i32 [[IV_NEXT_POSTLOOP]], [[NUM_ELEMENTS]]
; CHECK-NEXT: br i1 [[LOOP_COND_POSTLOOP]], label [[LOOP_POSTLOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP19:![0-9]+]], !irce.loop.clone !6
;
entry:
%capacity = load i32, ptr %capacity_p, !range !4
%num_elements = load i32, ptr %num_elements_p, !range !4
%limit = sub i32 %capacity, 3
br label %loop
loop:
%iv = phi i32 [0, %entry], [%iv.next, %backedge]
%capacity_check = icmp sge i32 %iv, %limit
br i1 %capacity_check, label %out_of_bounds, label %backedge, !prof !6
backedge:
%iv.wide = zext i32 %iv to i64
%el.ptr = getelementptr i32, ptr %p, i64 %iv.wide
store i32 1, ptr %el.ptr
%iv.next = add nuw nsw i32 %iv, 4
%loop_cond = icmp slt i32 %iv.next, %num_elements
br i1 %loop_cond, label %loop, label %exit
exit:
ret i32 %iv
out_of_bounds:
ret i32 -1
}
; Same as test_09 but range check comparison is non-strict:
; IRCE is allowed.
define i32 @test_11(ptr %p, ptr %capacity_p, ptr %num_elements_p) {
; CHECK-LABEL: define i32 @test_11
; CHECK-SAME: (ptr [[P:%.*]], ptr [[CAPACITY_P:%.*]], ptr [[NUM_ELEMENTS_P:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CAPACITY:%.*]] = load i32, ptr [[CAPACITY_P]], align 4, !range [[RNG16]]
; CHECK-NEXT: [[NUM_ELEMENTS:%.*]] = load i32, ptr [[NUM_ELEMENTS_P]], align 4, !range [[RNG16]]
; CHECK-NEXT: [[LIMIT:%.*]] = sub i32 [[CAPACITY]], 4
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[CAPACITY]], -3
; CHECK-NEXT: [[TMP1:%.*]] = add nuw i32 [[CAPACITY]], 2147483646
; CHECK-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[TMP1]], i32 0)
; CHECK-NEXT: [[TMP2:%.*]] = sub i32 [[TMP0]], [[SMAX]]
; CHECK-NEXT: [[TMP3:%.*]] = add nsw i32 [[CAPACITY]], -3
; CHECK-NEXT: [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[TMP3]], i32 0)
; CHECK-NEXT: [[SMAX2:%.*]] = call i32 @llvm.smax.i32(i32 [[SMIN]], i32 -1)
; CHECK-NEXT: [[TMP4:%.*]] = add nsw i32 [[SMAX2]], 1
; CHECK-NEXT: [[TMP5:%.*]] = mul i32 [[TMP2]], [[TMP4]]
; CHECK-NEXT: [[SMIN3:%.*]] = call i32 @llvm.smin.i32(i32 [[NUM_ELEMENTS]], i32 [[TMP5]])
; CHECK-NEXT: [[EXIT_MAINLOOP_AT:%.*]] = call i32 @llvm.smax.i32(i32 [[SMIN3]], i32 0)
; CHECK-NEXT: [[TMP6:%.*]] = icmp slt i32 0, [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP6]], label [[LOOP_PREHEADER:%.*]], label [[MAIN_PSEUDO_EXIT:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[BACKEDGE:%.*]] ], [ 0, [[LOOP_PREHEADER]] ]
; CHECK-NEXT: [[CAPACITY_CHECK:%.*]] = icmp sle i32 [[IV]], [[LIMIT]]
; CHECK-NEXT: br i1 true, label [[BACKEDGE]], label [[OUT_OF_BOUNDS_LOOPEXIT5:%.*]], !prof [[PROF17]]
; CHECK: backedge:
; CHECK-NEXT: [[IV_WIDE:%.*]] = zext i32 [[IV]] to i64
; CHECK-NEXT: [[EL_PTR:%.*]] = getelementptr i32, ptr [[P]], i64 [[IV_WIDE]]
; CHECK-NEXT: store i32 1, ptr [[EL_PTR]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i32 [[IV]], 4
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV_NEXT]], [[NUM_ELEMENTS]]
; CHECK-NEXT: [[TMP7:%.*]] = icmp slt i32 [[IV_NEXT]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP7]], label [[LOOP]], label [[MAIN_EXIT_SELECTOR:%.*]]
; CHECK: main.exit.selector:
; CHECK-NEXT: [[IV_NEXT_LCSSA:%.*]] = phi i32 [ [[IV_NEXT]], [[BACKEDGE]] ]
; CHECK-NEXT: [[IV_LCSSA:%.*]] = phi i32 [ [[IV]], [[BACKEDGE]] ]
; CHECK-NEXT: [[TMP8:%.*]] = icmp slt i32 [[IV_NEXT_LCSSA]], [[NUM_ELEMENTS]]
; CHECK-NEXT: br i1 [[TMP8]], label [[MAIN_PSEUDO_EXIT]], label [[EXIT:%.*]]
; CHECK: main.pseudo.exit:
; CHECK-NEXT: [[IV_COPY:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: [[INDVAR_END:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[IV_NEXT_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: br label [[POSTLOOP:%.*]]
; CHECK: exit.loopexit:
; CHECK-NEXT: [[IV_LCSSA1_PH:%.*]] = phi i32 [ [[IV_POSTLOOP:%.*]], [[BACKEDGE_POSTLOOP:%.*]] ]
; CHECK-NEXT: br label [[EXIT]]
; CHECK: exit:
; CHECK-NEXT: [[IV_LCSSA1:%.*]] = phi i32 [ [[IV_LCSSA]], [[MAIN_EXIT_SELECTOR]] ], [ [[IV_LCSSA1_PH]], [[EXIT_LOOPEXIT:%.*]] ]
; CHECK-NEXT: ret i32 [[IV_LCSSA1]]
; CHECK: out_of_bounds.loopexit:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS:%.*]]
; CHECK: out_of_bounds.loopexit5:
; CHECK-NEXT: br label [[OUT_OF_BOUNDS]]
; CHECK: out_of_bounds:
; CHECK-NEXT: ret i32 -1
; CHECK: postloop:
; CHECK-NEXT: br label [[LOOP_POSTLOOP:%.*]]
; CHECK: loop.postloop:
; CHECK-NEXT: [[IV_POSTLOOP]] = phi i32 [ [[IV_COPY]], [[POSTLOOP]] ], [ [[IV_NEXT_POSTLOOP:%.*]], [[BACKEDGE_POSTLOOP]] ]
; CHECK-NEXT: [[CAPACITY_CHECK_POSTLOOP:%.*]] = icmp sle i32 [[IV_POSTLOOP]], [[LIMIT]]
; CHECK-NEXT: br i1 [[CAPACITY_CHECK_POSTLOOP]], label [[BACKEDGE_POSTLOOP]], label [[OUT_OF_BOUNDS_LOOPEXIT:%.*]], !prof [[PROF17]]
; CHECK: backedge.postloop:
; CHECK-NEXT: [[IV_WIDE_POSTLOOP:%.*]] = zext i32 [[IV_POSTLOOP]] to i64
; CHECK-NEXT: [[EL_PTR_POSTLOOP:%.*]] = getelementptr i32, ptr [[P]], i64 [[IV_WIDE_POSTLOOP]]
; CHECK-NEXT: store i32 1, ptr [[EL_PTR_POSTLOOP]], align 4
; CHECK-NEXT: [[IV_NEXT_POSTLOOP]] = add nuw nsw i32 [[IV_POSTLOOP]], 4
; CHECK-NEXT: [[LOOP_COND_POSTLOOP:%.*]] = icmp slt i32 [[IV_NEXT_POSTLOOP]], [[NUM_ELEMENTS]]
; CHECK-NEXT: br i1 [[LOOP_COND_POSTLOOP]], label [[LOOP_POSTLOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP20:![0-9]+]], !irce.loop.clone !6
;
entry:
%capacity = load i32, ptr %capacity_p, !range !4
%num_elements = load i32, ptr %num_elements_p, !range !4
%limit = sub i32 %capacity, 4
br label %loop
loop:
%iv = phi i32 [0, %entry], [%iv.next, %backedge]
%capacity_check = icmp sle i32 %iv, %limit
br i1 %capacity_check, label %backedge, label %out_of_bounds, !prof !5
backedge:
%iv.wide = zext i32 %iv to i64
%el.ptr = getelementptr i32, ptr %p, i64 %iv.wide
store i32 1, ptr %el.ptr
%iv.next = add nuw nsw i32 %iv, 4
%loop_cond = icmp slt i32 %iv.next, %num_elements
br i1 %loop_cond, label %loop, label %exit
exit:
ret i32 %iv
out_of_bounds:
ret i32 -1
}
; Indvar base is non-overflowing binary 'or':
; check that IRCE isn't trying to add NSW flag on it.
define i32 @binop_or_is_iv_base(ptr %p, i32 %end) {
; CHECK-LABEL: define i32 @binop_or_is_iv_base
; CHECK-SAME: (ptr [[P:%.*]], i32 [[END:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[N:%.*]] = load atomic i32, ptr [[P]] unordered, align 8, !range [[RNG8]]
; CHECK-NEXT: [[CMP0:%.*]] = icmp sgt i32 [[END]], 7
; CHECK-NEXT: br i1 [[CMP0]], label [[LOOP_HEADER_PREHEADER:%.*]], label [[COMMON_RET:%.*]]
; CHECK: loop.header.preheader:
; CHECK-NEXT: [[TMP0:%.*]] = add nuw nsw i32 [[N]], 7
; CHECK-NEXT: [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[END]], i32 [[TMP0]])
; CHECK-NEXT: [[EXIT_MAINLOOP_AT:%.*]] = call i32 @llvm.smax.i32(i32 [[SMIN]], i32 7)
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt i32 7, [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP1]], label [[LOOP_HEADER_PREHEADER1:%.*]], label [[MAIN_PSEUDO_EXIT:%.*]]
; CHECK: loop.header.preheader1:
; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
; CHECK: loop.header:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_ADD:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_HEADER_PREHEADER1]] ]
; CHECK-NEXT: [[CHECK:%.*]] = icmp ult i32 [[IV]], [[N]]
; CHECK-NEXT: br i1 true, label [[GUARDED]], label [[DEOPT_LOOPEXIT2:%.*]]
; CHECK: guarded:
; CHECK-NEXT: [[IV_ADD]] = add i32 [[IV]], 8
; CHECK-NEXT: [[IV_OR:%.*]] = or i32 [[IV_ADD]], 7
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[IV_OR]], [[END]]
; CHECK-NEXT: [[TMP2:%.*]] = icmp slt i32 [[IV_OR]], [[EXIT_MAINLOOP_AT]]
; CHECK-NEXT: br i1 [[TMP2]], label [[LOOP_HEADER]], label [[MAIN_EXIT_SELECTOR:%.*]]
; CHECK: main.exit.selector:
; CHECK-NEXT: [[IV_ADD_LCSSA:%.*]] = phi i32 [ [[IV_ADD]], [[GUARDED]] ]
; CHECK-NEXT: [[IV_OR_LCSSA:%.*]] = phi i32 [ [[IV_OR]], [[GUARDED]] ]
; CHECK-NEXT: [[TMP3:%.*]] = icmp slt i32 [[IV_OR_LCSSA]], [[END]]
; CHECK-NEXT: br i1 [[TMP3]], label [[MAIN_PSEUDO_EXIT]], label [[COMMON_RET_LOOPEXIT:%.*]]
; CHECK: main.pseudo.exit:
; CHECK-NEXT: [[IV_COPY:%.*]] = phi i32 [ 0, [[LOOP_HEADER_PREHEADER]] ], [ [[IV_ADD_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: [[INDVAR_END:%.*]] = phi i32 [ 7, [[LOOP_HEADER_PREHEADER]] ], [ [[IV_OR_LCSSA]], [[MAIN_EXIT_SELECTOR]] ]
; CHECK-NEXT: br label [[POSTLOOP:%.*]]
; CHECK: deopt.loopexit:
; CHECK-NEXT: br label [[DEOPT:%.*]]
; CHECK: deopt.loopexit2:
; CHECK-NEXT: br label [[DEOPT]]
; CHECK: deopt:
; CHECK-NEXT: [[RV:%.*]] = call i32 (...) @llvm.experimental.deoptimize.i32(i32 13) [ "deopt"() ]
; CHECK-NEXT: ret i32 [[RV]]
; CHECK: common.ret.loopexit.loopexit:
; CHECK-NEXT: br label [[COMMON_RET_LOOPEXIT]]
; CHECK: common.ret.loopexit:
; CHECK-NEXT: br label [[COMMON_RET]]
; CHECK: common.ret:
; CHECK-NEXT: ret i32 [[END]]
; CHECK: postloop:
; CHECK-NEXT: br label [[LOOP_HEADER_POSTLOOP:%.*]]
; CHECK: loop.header.postloop:
; CHECK-NEXT: [[IV_POSTLOOP:%.*]] = phi i32 [ [[IV_ADD_POSTLOOP:%.*]], [[GUARDED_POSTLOOP:%.*]] ], [ [[IV_COPY]], [[POSTLOOP]] ]
; CHECK-NEXT: [[CHECK_POSTLOOP:%.*]] = icmp ult i32 [[IV_POSTLOOP]], [[N]]
; CHECK-NEXT: br i1 [[CHECK_POSTLOOP]], label [[GUARDED_POSTLOOP]], label [[DEOPT_LOOPEXIT:%.*]]
; CHECK: guarded.postloop:
; CHECK-NEXT: [[IV_ADD_POSTLOOP]] = add i32 [[IV_POSTLOOP]], 8
; CHECK-NEXT: [[IV_OR_POSTLOOP:%.*]] = or i32 [[IV_ADD_POSTLOOP]], 7
; CHECK-NEXT: [[CMP_POSTLOOP:%.*]] = icmp slt i32 [[IV_OR_POSTLOOP]], [[END]]
; CHECK-NEXT: br i1 [[CMP_POSTLOOP]], label [[LOOP_HEADER_POSTLOOP]], label [[COMMON_RET_LOOPEXIT_LOOPEXIT:%.*]], !llvm.loop [[LOOP21:![0-9]+]], !irce.loop.clone !6
;
entry:
%n = load atomic i32, ptr %p unordered, align 8, !range !1
%cmp0 = icmp sgt i32 %end, 7
br i1 %cmp0, label %loop.header, label %common.ret
loop.header:
%iv = phi i32 [ %iv.add, %guarded ], [ 0, %entry ]
%check = icmp ult i32 %iv, %n
br i1 %check, label %guarded, label %deopt
guarded:
%iv.add = add i32 %iv, 8
%iv.or = or i32 %iv.add, 7
%cmp = icmp slt i32 %iv.or, %end
br i1 %cmp, label %loop.header, label %common.ret
deopt:
%rv = call i32 (...) @llvm.experimental.deoptimize.i32(i32 13) [ "deopt"() ]
ret i32 %rv
common.ret:
ret i32 %end
}
declare i32 @llvm.experimental.deoptimize.i32(...)
!0 = !{i32 0, i32 50}
!1 = !{i32 0, i32 2147483640}
!2 = !{i32 0, i32 2147483641}
!3 = !{i32 10, i32 50}
!4 = !{i32 1, i32 2147483648}
!5 = !{!"branch_weights", i32 1000, i32 1}
!6 = !{!"branch_weights", i32 1, i32 1000}