; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
; RUN: opt %s -S -riscv-codegenprepare -mtriple=riscv64 -mattr=+v | FileCheck %s

declare i64 @llvm.vscale.i64()
declare float @llvm.vector.reduce.fadd.nxv4f32(float, <vscale x 4 x float>)

define float @reduce_fadd(ptr %f) {
; CHECK-LABEL: define float @reduce_fadd(
; CHECK-SAME: ptr [[F:%.*]]) #[[ATTR2:[0-9]+]] {
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[VSCALE:%.*]] = tail call i64 @llvm.vscale.i64()
; CHECK-NEXT:    [[VECSIZE:%.*]] = shl nuw nsw i64 [[VSCALE]], 2
; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
; CHECK:       vector.body:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = phi <vscale x 4 x float> [ insertelement (<vscale x 4 x float> poison, float 0.000000e+00, i64 0), [[ENTRY]] ], [ [[TMP2:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[F]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[GEP]], align 4
; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <vscale x 4 x float> [[TMP0]], i64 0
; CHECK-NEXT:    [[ACC:%.*]] = tail call float @llvm.vector.reduce.fadd.nxv4f32(float [[TMP1]], <vscale x 4 x float> [[WIDE_LOAD]])
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[VECSIZE]]
; CHECK-NEXT:    [[DONE:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; CHECK-NEXT:    [[TMP2]] = insertelement <vscale x 4 x float> poison, float [[ACC]], i64 0
; CHECK-NEXT:    br i1 [[DONE]], label [[EXIT:%.*]], label [[VECTOR_BODY]]
; CHECK:       exit:
; CHECK-NEXT:    ret float [[ACC]]
;

entry:
  %vscale = tail call i64 @llvm.vscale.i64()
  %vecsize = shl nuw nsw i64 %vscale, 2
  br label %vector.body

vector.body:
  %index = phi i64 [ 0, %entry ], [ %index.next, %vector.body ]
  %vec.phi = phi float [ 0.000000e+00, %entry ], [ %acc, %vector.body ]
  %gep = getelementptr inbounds float, ptr %f, i64 %index
  %wide.load = load <vscale x 4 x float>, ptr %gep, align 4
  %acc = tail call float @llvm.vector.reduce.fadd.nxv4f32(float %vec.phi, <vscale x 4 x float> %wide.load)
  %index.next = add nuw i64 %index, %vecsize
  %done = icmp eq i64 %index.next, 1024
  br i1 %done, label %exit, label %vector.body

exit:
  ret float %acc
}