75 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 42; RUN: opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s | FileCheck %s3 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 47@a = global i32 0, align 48 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]], 714; 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 label [[VECTOR_PH:%.*]]18; CHECK: vector.ph:19; CHECK-NEXT: [[N_RND_UP:%.*]] = add i64 [[TMP0]], 320; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 421; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]22; CHECK-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[TMP0]], 123; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 024; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer25; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]26; CHECK: vector.body:27; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]28; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]29; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i64> poison, i64 [[INDEX]], i64 030; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT1]], <4 x i64> poison, <4 x i32> zeroinitializer31; CHECK-NEXT: [[VEC_IV:%.*]] = add <4 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3>32; CHECK-NEXT: [[TMP1:%.*]] = icmp ule <4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT]]33; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @c, align 434; CHECK-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <4 x i32> poison, i32 [[TMP2]], i64 035; CHECK-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT3]], <4 x i32> poison, <4 x i32> zeroinitializer36; CHECK-NEXT: [[TMP3]] = xor <4 x i32> [[VEC_PHI]], [[BROADCAST_SPLAT4]]37; CHECK-NEXT: [[TMP4:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[TMP3]], <4 x i32> [[VEC_PHI]]38; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 439; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]40; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]41; CHECK: middle.block:42; CHECK-NEXT: [[TMP6:%.*]] = call i32 @llvm.vector.reduce.xor.v4i32(<4 x i32> [[TMP4]])43; CHECK-NEXT: store i32 [[TMP6]], ptr @a, align 444; CHECK-NEXT: br label [[LOOP:%.*]]45; CHECK: exit.loopexit:46; CHECK-NEXT: br label [[EXIT]]47; CHECK: exit:48; CHECK-NEXT: ret void49;50entry:51 %cmp3 = icmp slt i64 %iv.start, 752 br i1 %cmp3, label %ph, label %exit53 54ph:55 br label %loop56 57loop:58 %iv = phi i64 [ %iv.start, %ph ], [ %iv.next, %loop ]59 %red = phi i32 [ 0, %ph ], [ %red.next, %loop ]60 %l = load i32, ptr @c, align 461 %red.next = xor i32 %red, %l62 store i32 %red.next, ptr @a, align 463 %iv.next = add i64 %iv, 164 %exitcond.not = icmp eq i64 %iv.next, 765 br i1 %exitcond.not, label %exit, label %loop66 67exit:68 ret void69}70;.71; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}72; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}73; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}74;.75