52 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 52; Check remark analysis runs successfully.3; TODO: Remove this test when we fully migrate to VPlan-based cost model.4; RUN: opt < %s -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -prefer-inloop-reductions -pass-remarks-analysis=loop-vectorize -S | FileCheck %s5 6target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"7 8define i32 @reduction_sum(ptr noalias nocapture %A, ptr noalias nocapture %B) {9; CHECK-LABEL: define i32 @reduction_sum(10; CHECK-SAME: ptr noalias captures(none) [[A:%.*]], ptr noalias captures(none) [[B:%.*]]) {11; CHECK-NEXT: [[ENTRY:.*:]]12; CHECK-NEXT: br label %[[VECTOR_PH:.*]]13; CHECK: [[VECTOR_PH]]:14; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]15; CHECK: [[VECTOR_BODY]]:16; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]17; CHECK-NEXT: [[VEC_PHI:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]18; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]19; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 420; CHECK-NEXT: [[TMP3:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[WIDE_LOAD]])21; CHECK-NEXT: [[TMP4]] = add i32 [[VEC_PHI]], [[TMP3]]22; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 423; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 25624; CHECK-NEXT: br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]25; CHECK: [[MIDDLE_BLOCK]]:26; CHECK-NEXT: br label %[[EXIT:.*]]27; CHECK: [[EXIT]]:28; CHECK-NEXT: ret i32 [[TMP4]]29;30entry:31 br label %body32 33body: ; preds = %entry, %body34 %indvars.iv = phi i64 [ %indvars.iv.next, %body ], [ 0, %entry ]35 %sum.tmp = phi i32 [ %sum, %body ], [ 0, %entry ]36 %gep0 = getelementptr inbounds i32, ptr %A, i64 %indvars.iv37 %load0 = load i32, ptr %gep0, align 438 %sum = add i32 %sum.tmp, %load039 %indvars.iv.next = add i64 %indvars.iv, 140 %exitcond = icmp eq i64 %indvars.iv.next, 25641 br i1 %exitcond, label %exit, label %body42 43exit: ; preds = %body44 %sum.0.lcssa = phi i32 [ %sum, %body ]45 ret i32 %sum.0.lcssa46}47;.48; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}49; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}50; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}51;.52