; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx %s -o - | FileCheck %s --check-prefix=LA32
; RUN: llc --mtriple=loongarch64 --mattr=+lsx %s -o - | FileCheck %s --check-prefix=LA64

define void @vec_reduce_add_v16i8(ptr %src, ptr %dst) nounwind {
; LA32-LABEL: vec_reduce_add_v16i8:
; LA32:       # %bb.0:
; LA32-NEXT:    vld $vr0, $a0, 0
; LA32-NEXT:    vhaddw.h.b $vr0, $vr0, $vr0
; LA32-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA32-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA32-NEXT:    vhaddw.q.d $vr0, $vr0, $vr0
; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
; LA32-NEXT:    st.b $a0, $a1, 0
; LA32-NEXT:    ret
;
; LA64-LABEL: vec_reduce_add_v16i8:
; LA64:       # %bb.0:
; LA64-NEXT:    vld $vr0, $a0, 0
; LA64-NEXT:    vhaddw.h.b $vr0, $vr0, $vr0
; LA64-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA64-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA64-NEXT:    vhaddw.q.d $vr0, $vr0, $vr0
; LA64-NEXT:    vpickve2gr.d $a0, $vr0, 0
; LA64-NEXT:    st.b $a0, $a1, 0
; LA64-NEXT:    ret
  %v = load <16 x i8>, ptr %src
  %res = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %v)
  store i8 %res, ptr %dst
  ret void
}

define void @vec_reduce_add_v8i8(ptr %src, ptr %dst) nounwind {
; LA32-LABEL: vec_reduce_add_v8i8:
; LA32:       # %bb.0:
; LA32-NEXT:    ld.w $a2, $a0, 0
; LA32-NEXT:    ld.w $a0, $a0, 4
; LA32-NEXT:    vinsgr2vr.w $vr0, $a2, 0
; LA32-NEXT:    vinsgr2vr.w $vr0, $a0, 1
; LA32-NEXT:    vhaddw.h.b $vr0, $vr0, $vr0
; LA32-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA32-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
; LA32-NEXT:    st.b $a0, $a1, 0
; LA32-NEXT:    ret
;
; LA64-LABEL: vec_reduce_add_v8i8:
; LA64:       # %bb.0:
; LA64-NEXT:    ld.d $a0, $a0, 0
; LA64-NEXT:    vinsgr2vr.d $vr0, $a0, 0
; LA64-NEXT:    vhaddw.h.b $vr0, $vr0, $vr0
; LA64-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA64-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA64-NEXT:    vpickve2gr.d $a0, $vr0, 0
; LA64-NEXT:    st.b $a0, $a1, 0
; LA64-NEXT:    ret
  %v = load <8 x i8>, ptr %src
  %res = call i8 @llvm.vector.reduce.add.v8i8(<8 x i8> %v)
  store i8 %res, ptr %dst
  ret void
}

define void @vec_reduce_add_v4i8(ptr %src, ptr %dst) nounwind {
; LA32-LABEL: vec_reduce_add_v4i8:
; LA32:       # %bb.0:
; LA32-NEXT:    ld.w $a0, $a0, 0
; LA32-NEXT:    vinsgr2vr.w $vr0, $a0, 0
; LA32-NEXT:    vhaddw.h.b $vr0, $vr0, $vr0
; LA32-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
; LA32-NEXT:    st.b $a0, $a1, 0
; LA32-NEXT:    ret
;
; LA64-LABEL: vec_reduce_add_v4i8:
; LA64:       # %bb.0:
; LA64-NEXT:    ld.w $a0, $a0, 0
; LA64-NEXT:    vinsgr2vr.w $vr0, $a0, 0
; LA64-NEXT:    vhaddw.h.b $vr0, $vr0, $vr0
; LA64-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA64-NEXT:    vpickve2gr.d $a0, $vr0, 0
; LA64-NEXT:    st.b $a0, $a1, 0
; LA64-NEXT:    ret
  %v = load <4 x i8>, ptr %src
  %res = call i8 @llvm.vector.reduce.add.v4i8(<4 x i8> %v)
  store i8 %res, ptr %dst
  ret void
}

define void @vec_reduce_add_v2i8(ptr %src, ptr %dst) nounwind {
; LA32-LABEL: vec_reduce_add_v2i8:
; LA32:       # %bb.0:
; LA32-NEXT:    ld.h $a0, $a0, 0
; LA32-NEXT:    vinsgr2vr.h $vr0, $a0, 0
; LA32-NEXT:    vhaddw.h.b $vr0, $vr0, $vr0
; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
; LA32-NEXT:    st.b $a0, $a1, 0
; LA32-NEXT:    ret
;
; LA64-LABEL: vec_reduce_add_v2i8:
; LA64:       # %bb.0:
; LA64-NEXT:    ld.h $a0, $a0, 0
; LA64-NEXT:    vinsgr2vr.h $vr0, $a0, 0
; LA64-NEXT:    vhaddw.h.b $vr0, $vr0, $vr0
; LA64-NEXT:    vpickve2gr.d $a0, $vr0, 0
; LA64-NEXT:    st.b $a0, $a1, 0
; LA64-NEXT:    ret
  %v = load <2 x i8>, ptr %src
  %res = call i8 @llvm.vector.reduce.add.v2i8(<2 x i8> %v)
  store i8 %res, ptr %dst
  ret void
}

define void @vec_reduce_add_v8i16(ptr %src, ptr %dst) nounwind {
; LA32-LABEL: vec_reduce_add_v8i16:
; LA32:       # %bb.0:
; LA32-NEXT:    vld $vr0, $a0, 0
; LA32-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA32-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA32-NEXT:    vhaddw.q.d $vr0, $vr0, $vr0
; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
; LA32-NEXT:    st.h $a0, $a1, 0
; LA32-NEXT:    ret
;
; LA64-LABEL: vec_reduce_add_v8i16:
; LA64:       # %bb.0:
; LA64-NEXT:    vld $vr0, $a0, 0
; LA64-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA64-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA64-NEXT:    vhaddw.q.d $vr0, $vr0, $vr0
; LA64-NEXT:    vpickve2gr.d $a0, $vr0, 0
; LA64-NEXT:    st.h $a0, $a1, 0
; LA64-NEXT:    ret
  %v = load <8 x i16>, ptr %src
  %res = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> %v)
  store i16 %res, ptr %dst
  ret void
}

define void @vec_reduce_add_v4i16(ptr %src, ptr %dst) nounwind {
; LA32-LABEL: vec_reduce_add_v4i16:
; LA32:       # %bb.0:
; LA32-NEXT:    ld.w $a2, $a0, 0
; LA32-NEXT:    ld.w $a0, $a0, 4
; LA32-NEXT:    vinsgr2vr.w $vr0, $a2, 0
; LA32-NEXT:    vinsgr2vr.w $vr0, $a0, 1
; LA32-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA32-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
; LA32-NEXT:    st.h $a0, $a1, 0
; LA32-NEXT:    ret
;
; LA64-LABEL: vec_reduce_add_v4i16:
; LA64:       # %bb.0:
; LA64-NEXT:    ld.d $a0, $a0, 0
; LA64-NEXT:    vinsgr2vr.d $vr0, $a0, 0
; LA64-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA64-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA64-NEXT:    vpickve2gr.d $a0, $vr0, 0
; LA64-NEXT:    st.h $a0, $a1, 0
; LA64-NEXT:    ret
  %v = load <4 x i16>, ptr %src
  %res = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> %v)
  store i16 %res, ptr %dst
  ret void
}

define void @vec_reduce_add_v2i16(ptr %src, ptr %dst) nounwind {
; LA32-LABEL: vec_reduce_add_v2i16:
; LA32:       # %bb.0:
; LA32-NEXT:    ld.w $a0, $a0, 0
; LA32-NEXT:    vinsgr2vr.w $vr0, $a0, 0
; LA32-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
; LA32-NEXT:    st.h $a0, $a1, 0
; LA32-NEXT:    ret
;
; LA64-LABEL: vec_reduce_add_v2i16:
; LA64:       # %bb.0:
; LA64-NEXT:    ld.w $a0, $a0, 0
; LA64-NEXT:    vinsgr2vr.w $vr0, $a0, 0
; LA64-NEXT:    vhaddw.w.h $vr0, $vr0, $vr0
; LA64-NEXT:    vpickve2gr.d $a0, $vr0, 0
; LA64-NEXT:    st.h $a0, $a1, 0
; LA64-NEXT:    ret
  %v = load <2 x i16>, ptr %src
  %res = call i16 @llvm.vector.reduce.add.v2i16(<2 x i16> %v)
  store i16 %res, ptr %dst
  ret void
}

define void @vec_reduce_add_v4i32(ptr %src, ptr %dst) nounwind {
; LA32-LABEL: vec_reduce_add_v4i32:
; LA32:       # %bb.0:
; LA32-NEXT:    vld $vr0, $a0, 0
; LA32-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA32-NEXT:    vhaddw.q.d $vr0, $vr0, $vr0
; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
; LA32-NEXT:    st.w $a0, $a1, 0
; LA32-NEXT:    ret
;
; LA64-LABEL: vec_reduce_add_v4i32:
; LA64:       # %bb.0:
; LA64-NEXT:    vld $vr0, $a0, 0
; LA64-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA64-NEXT:    vhaddw.q.d $vr0, $vr0, $vr0
; LA64-NEXT:    vpickve2gr.d $a0, $vr0, 0
; LA64-NEXT:    st.w $a0, $a1, 0
; LA64-NEXT:    ret
  %v = load <4 x i32>, ptr %src
  %res = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %v)
  store i32 %res, ptr %dst
  ret void
}

define void @vec_reduce_add_v2i32(ptr %src, ptr %dst) nounwind {
; LA32-LABEL: vec_reduce_add_v2i32:
; LA32:       # %bb.0:
; LA32-NEXT:    ld.w $a2, $a0, 0
; LA32-NEXT:    ld.w $a0, $a0, 4
; LA32-NEXT:    vinsgr2vr.w $vr0, $a2, 0
; LA32-NEXT:    vinsgr2vr.w $vr0, $a0, 1
; LA32-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
; LA32-NEXT:    st.w $a0, $a1, 0
; LA32-NEXT:    ret
;
; LA64-LABEL: vec_reduce_add_v2i32:
; LA64:       # %bb.0:
; LA64-NEXT:    ld.d $a0, $a0, 0
; LA64-NEXT:    vinsgr2vr.d $vr0, $a0, 0
; LA64-NEXT:    vhaddw.d.w $vr0, $vr0, $vr0
; LA64-NEXT:    vpickve2gr.d $a0, $vr0, 0
; LA64-NEXT:    st.w $a0, $a1, 0
; LA64-NEXT:    ret
  %v = load <2 x i32>, ptr %src
  %res = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> %v)
  store i32 %res, ptr %dst
  ret void
}

define void @vec_reduce_add_v2i64(ptr %src, ptr %dst) nounwind {
; LA32-LABEL: vec_reduce_add_v2i64:
; LA32:       # %bb.0:
; LA32-NEXT:    vld $vr0, $a0, 0
; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 2
; LA32-NEXT:    vpickve2gr.w $a2, $vr0, 0
; LA32-NEXT:    vpickve2gr.w $a3, $vr0, 3
; LA32-NEXT:    vpickve2gr.w $a4, $vr0, 1
; LA32-NEXT:    add.w $a3, $a4, $a3
; LA32-NEXT:    add.w $a0, $a2, $a0
; LA32-NEXT:    sltu $a2, $a0, $a2
; LA32-NEXT:    add.w $a2, $a3, $a2
; LA32-NEXT:    st.w $a0, $a1, 0
; LA32-NEXT:    st.w $a2, $a1, 4
; LA32-NEXT:    ret
;
; LA64-LABEL: vec_reduce_add_v2i64:
; LA64:       # %bb.0:
; LA64-NEXT:    vld $vr0, $a0, 0
; LA64-NEXT:    vhaddw.q.d $vr0, $vr0, $vr0
; LA64-NEXT:    vstelm.d $vr0, $a1, 0, 0
; LA64-NEXT:    ret
  %v = load <2 x i64>, ptr %src
  %res = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %v)
  store i64 %res, ptr %dst
  ret void
}