1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4 2; RUN: opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s | FileCheck %s 3 4target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128" 5 6@c = global i32 1, align 4 7@a = global i32 0, align 4 8 9define void @pr75298_store_reduction_value_in_folded_loop(i64 %iv.start) optsize { 10; CHECK-LABEL: define void @pr75298_store_reduction_value_in_folded_loop( 11; CHECK-SAME: i64 [[IV_START:%.*]]) #[[ATTR0:[0-9]+]] { 12; CHECK-NEXT: entry: 13; CHECK-NEXT: [[CMP3:%.*]] = icmp slt i64 [[IV_START]], 7 14; CHECK-NEXT: br i1 [[CMP3]], label [[PH:%.*]], label [[EXIT:%.*]] 15; CHECK: ph: 16; CHECK-NEXT: [[TMP0:%.*]] = sub i64 7, [[IV_START]] 17; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 18; CHECK: vector.ph: 19; CHECK-NEXT: [[N_RND_UP:%.*]] = add i64 [[TMP0]], 3 20; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 4 21; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]] 22; CHECK-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[TMP0]], 1 23; CHECK-NEXT: [[IND_END:%.*]] = add i64 [[IV_START]], [[N_VEC]] 24; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0 25; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer 26; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 27; CHECK: vector.body: 28; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 29; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ] 30; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i64> poison, i64 [[INDEX]], i64 0 31; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT1]], <4 x i64> poison, <4 x i32> zeroinitializer 32; CHECK-NEXT: [[VEC_IV:%.*]] = add <4 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3> 33; CHECK-NEXT: [[TMP1:%.*]] = icmp ule <4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT]] 34; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @c, align 4 35; CHECK-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <4 x i32> poison, i32 [[TMP2]], i64 0 36; CHECK-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT3]], <4 x i32> poison, <4 x i32> zeroinitializer 37; CHECK-NEXT: [[TMP3]] = xor <4 x i32> [[VEC_PHI]], [[BROADCAST_SPLAT4]] 38; CHECK-NEXT: [[TMP4:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[TMP3]], <4 x i32> [[VEC_PHI]] 39; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 4 40; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 41; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 42; CHECK: middle.block: 43; CHECK-NEXT: [[TMP6:%.*]] = call i32 @llvm.vector.reduce.xor.v4i32(<4 x i32> [[TMP4]]) 44; CHECK-NEXT: store i32 [[TMP6]], ptr @a, align 4 45; CHECK-NEXT: br i1 true, label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]] 46; CHECK: scalar.ph: 47; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[IV_START]], [[PH]] ] 48; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ 0, [[PH]] ] 49; CHECK-NEXT: br label [[LOOP:%.*]] 50; CHECK: loop: 51; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ] 52; CHECK-NEXT: [[RED:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[RED_NEXT:%.*]], [[LOOP]] ] 53; CHECK-NEXT: [[L:%.*]] = load i32, ptr @c, align 4 54; CHECK-NEXT: [[RED_NEXT]] = xor i32 [[RED]], [[L]] 55; CHECK-NEXT: store i32 [[RED_NEXT]], ptr @a, align 4 56; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 57; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 7 58; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[EXIT_LOOPEXIT]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]] 59; CHECK: exit.loopexit: 60; CHECK-NEXT: br label [[EXIT]] 61; CHECK: exit: 62; CHECK-NEXT: ret void 63; 64entry: 65 %cmp3 = icmp slt i64 %iv.start, 7 66 br i1 %cmp3, label %ph, label %exit 67 68ph: 69 br label %loop 70 71loop: 72 %iv = phi i64 [ %iv.start, %ph ], [ %iv.next, %loop ] 73 %red = phi i32 [ 0, %ph ], [ %red.next, %loop ] 74 %l = load i32, ptr @c, align 4 75 %red.next = xor i32 %red, %l 76 store i32 %red.next, ptr @a, align 4 77 %iv.next = add i64 %iv, 1 78 %exitcond.not = icmp eq i64 %iv.next, 7 79 br i1 %exitcond.not, label %exit, label %loop 80 81exit: 82 ret void 83} 84;. 85; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]} 86; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1} 87; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"} 88; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]} 89;. 90