; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=riscv32 -mattr=+f,+v -verify-machineinstrs < %s | FileCheck %s
; RUN: llc -mtriple=riscv64 -mattr=+f,+v -verify-machineinstrs < %s | FileCheck %s

define <vscale x 2 x float> @test_reverse_load_combiner(<vscale x 2 x float>* %ptr, i32 zeroext %evl) {
; CHECK-LABEL: test_reverse_load_combiner:
; CHECK:       # %bb.0:
; CHECK-NEXT:    slli a2, a1, 2
; CHECK-NEXT:    add a0, a2, a0
; CHECK-NEXT:    addi a0, a0, -4
; CHECK-NEXT:    li a2, -4
; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT:    vlse32.v v8, (a0), a2
; CHECK-NEXT:    ret
  %load = call <vscale x 2 x float> @llvm.vp.load.nxv2f32.p0nxv2f32(<vscale x 2 x float>* %ptr, <vscale x 2 x i1> splat (i1 true), i32 %evl)
  %rev = call <vscale x 2 x float> @llvm.experimental.vp.reverse.nxv2f32(<vscale x 2 x float> %load, <vscale x 2 x i1> splat (i1 true), i32 %evl)
  ret <vscale x 2 x float> %rev
}

define <vscale x 2 x float> @test_load_mask_is_vp_reverse(<vscale x 2 x float>* %ptr, <vscale x 2 x i1> %mask, i32 zeroext %evl) {
; CHECK-LABEL: test_load_mask_is_vp_reverse:
; CHECK:       # %bb.0:
; CHECK-NEXT:    slli a2, a1, 2
; CHECK-NEXT:    add a0, a2, a0
; CHECK-NEXT:    addi a0, a0, -4
; CHECK-NEXT:    li a2, -4
; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT:    vlse32.v v8, (a0), a2, v0.t
; CHECK-NEXT:    ret
  %loadmask = call <vscale x 2 x i1> @llvm.experimental.vp.reverse.nxv2i1(<vscale x 2 x i1> %mask, <vscale x 2 x i1> splat (i1 true), i32 %evl)
  %load = call <vscale x 2 x float> @llvm.vp.load.nxv2f32.p0nxv2f32(<vscale x 2 x float>* %ptr, <vscale x 2 x i1> %loadmask, i32 %evl)
  %rev = call <vscale x 2 x float> @llvm.experimental.vp.reverse.nxv2f32(<vscale x 2 x float> %load, <vscale x 2 x i1> splat (i1 true), i32 %evl)
  ret <vscale x 2 x float> %rev
}

define <vscale x 2 x float> @test_load_mask_is_vp_reverse_with_mask(<vscale x 2 x float>* %ptr, <vscale x 2 x i1> %mask, <vscale x 2 x i1> %revmask, i32 zeroext %evl) {
; CHECK-LABEL: test_load_mask_is_vp_reverse_with_mask:
; CHECK:       # %bb.0:
; CHECK-NEXT:    slli a2, a1, 2
; CHECK-NEXT:    add a0, a2, a0
; CHECK-NEXT:    addi a0, a0, -4
; CHECK-NEXT:    li a2, -4
; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT:    vlse32.v v8, (a0), a2, v0.t
; CHECK-NEXT:    ret
  %loadmask = call <vscale x 2 x i1> @llvm.experimental.vp.reverse.nxv2i1(<vscale x 2 x i1> %mask, <vscale x 2 x i1> %revmask, i32 %evl)
  %load = call <vscale x 2 x float> @llvm.vp.load.nxv2f32.p0nxv2f32(<vscale x 2 x float>* %ptr, <vscale x 2 x i1> %loadmask, i32 %evl)
  %rev = call <vscale x 2 x float> @llvm.experimental.vp.reverse.nxv2f32(<vscale x 2 x float> %load, <vscale x 2 x i1> splat (i1 true), i32 %evl)
  ret <vscale x 2 x float> %rev
}

define <vscale x 2 x float> @test_load_mask_not_all_one(<vscale x 2 x float>* %ptr, <vscale x 2 x i1> %notallones, i32 zeroext %evl) {
; CHECK-LABEL: test_load_mask_not_all_one:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT:    vle32.v v9, (a0), v0.t
; CHECK-NEXT:    vid.v v8, v0.t
; CHECK-NEXT:    addi a1, a1, -1
; CHECK-NEXT:    vrsub.vx v10, v8, a1, v0.t
; CHECK-NEXT:    vrgather.vv v8, v9, v10, v0.t
; CHECK-NEXT:    ret
  %load = call <vscale x 2 x float> @llvm.vp.load.nxv2f32.p0nxv2f32(<vscale x 2 x float>* %ptr, <vscale x 2 x i1> %notallones, i32 %evl)
  %rev = call <vscale x 2 x float> @llvm.experimental.vp.reverse.nxv2f32(<vscale x 2 x float> %load, <vscale x 2 x i1> %notallones, i32 %evl)
  ret <vscale x 2 x float> %rev
}

define <vscale x 2 x float> @test_load_reverse_mask_not_all_one(ptr %ptr, <vscale x 2 x i1> %notallones, i32 zeroext %evl) {
; CHECK-LABEL: test_load_reverse_mask_not_all_one:
; CHECK:       # %bb.0:
; CHECK-NEXT:    slli a2, a1, 2
; CHECK-NEXT:    add a0, a2, a0
; CHECK-NEXT:    addi a0, a0, -4
; CHECK-NEXT:    li a2, -4
; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT:    vlse32.v v8, (a0), a2
; CHECK-NEXT:    ret
  %load = call <vscale x 2 x float> @llvm.vp.load.nxv2f32.p0nxv2f32(ptr %ptr, <vscale x 2 x i1> splat (i1 true), i32 %evl)
  %rev = call <vscale x 2 x float> @llvm.experimental.vp.reverse.nxv2f32(<vscale x 2 x float> %load, <vscale x 2 x i1> %notallones, i32 %evl)
  ret <vscale x 2 x float> %rev
}

define <vscale x 2 x float> @test_different_evl(<vscale x 2 x float>* %ptr, <vscale x 2 x i1> %mask, i32 zeroext %evl1, i32 zeroext %evl2) {
; CHECK-LABEL: test_different_evl:
; CHECK:       # %bb.0:
; CHECK-NEXT:    addi a3, a1, -1
; CHECK-NEXT:    vsetvli zero, a1, e16, mf2, ta, ma
; CHECK-NEXT:    vid.v v8
; CHECK-NEXT:    vsetvli zero, zero, e8, mf4, ta, ma
; CHECK-NEXT:    vmv.v.i v9, 0
; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
; CHECK-NEXT:    vrsub.vx v8, v8, a3
; CHECK-NEXT:    vsetvli zero, zero, e8, mf4, ta, ma
; CHECK-NEXT:    vmerge.vim v9, v9, 1, v0
; CHECK-NEXT:    vrgatherei16.vv v10, v9, v8
; CHECK-NEXT:    vmsne.vi v0, v10, 0
; CHECK-NEXT:    vsetvli zero, a2, e32, m1, ta, ma
; CHECK-NEXT:    vle32.v v9, (a0), v0.t
; CHECK-NEXT:    addi a2, a2, -1
; CHECK-NEXT:    vid.v v8
; CHECK-NEXT:    vrsub.vx v10, v8, a2
; CHECK-NEXT:    vrgather.vv v8, v9, v10
; CHECK-NEXT:    ret
  %loadmask = call <vscale x 2 x i1> @llvm.experimental.vp.reverse.nxv2i1(<vscale x 2 x i1> %mask, <vscale x 2 x i1> splat (i1 true), i32 %evl1)
  %load = call <vscale x 2 x float> @llvm.vp.load.nxv2f32.p0nxv2f32(<vscale x 2 x float>* %ptr, <vscale x 2 x i1> %loadmask, i32 %evl2)
  %rev = call <vscale x 2 x float> @llvm.experimental.vp.reverse.nxv2f32(<vscale x 2 x float> %load, <vscale x 2 x i1> splat (i1 true), i32 %evl2)
  ret <vscale x 2 x float> %rev
}

define <vscale x 2 x float> @test_reverse_load_combiner_splice(ptr %ptr, i32 zeroext %evl) {
; CHECK-LABEL: test_reverse_load_combiner_splice:
; CHECK:       # %bb.0:
; CHECK-NEXT:    slli a2, a1, 2
; CHECK-NEXT:    add a0, a2, a0
; CHECK-NEXT:    addi a0, a0, -4
; CHECK-NEXT:    li a2, -4
; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT:    vlse32.v v8, (a0), a2
; CHECK-NEXT:    ret
  %load = call <vscale x 2 x float> @llvm.vp.load(ptr %ptr, <vscale x 2 x i1> splat (i1 true), i32 %evl)
  %rev = call <vscale x 2 x float> @llvm.vector.reverse(<vscale x 2 x float> %load)
  %splice = call <vscale x 2 x float> @llvm.vector.splice.right(<vscale x 2 x float> %rev, <vscale x 2 x float> poison, i32 %evl)
  ret <vscale x 2 x float> %splice
}

define <vscale x 2 x float> @test_load_mask_is_reverse_splice(ptr %ptr, <vscale x 2 x i1> %mask, i32 zeroext %evl) {
; CHECK-LABEL: test_load_mask_is_reverse_splice:
; CHECK:       # %bb.0:
; CHECK-NEXT:    slli a2, a1, 2
; CHECK-NEXT:    add a0, a2, a0
; CHECK-NEXT:    addi a0, a0, -4
; CHECK-NEXT:    li a2, -4
; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT:    vlse32.v v8, (a0), a2, v0.t
; CHECK-NEXT:    ret
  %loadmask.rev = call <vscale x 2 x i1> @llvm.vector.reverse(<vscale x 2 x i1> %mask)
  %loadmask.splice = call <vscale x 2 x i1> @llvm.vector.splice.right(<vscale x 2 x i1> %loadmask.rev, <vscale x 2 x i1> poison, i32 %evl)
  %load = call <vscale x 2 x float> @llvm.vp.load(ptr %ptr, <vscale x 2 x i1> %loadmask.splice, i32 %evl)
  %rev = call <vscale x 2 x float> @llvm.vector.reverse(<vscale x 2 x float> %load)
  %splice = call <vscale x 2 x float> @llvm.vector.splice.right(<vscale x 2 x float> %rev, <vscale x 2 x float> poison, i32 %evl)
  ret <vscale x 2 x float> %splice
}

define <vscale x 2 x float> @test_different_evl_splice(ptr %ptr, i32 zeroext %evl1, i32 zeroext %evl2) {
; CHECK-LABEL: test_different_evl_splice:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT:    vle32.v v8, (a0)
; CHECK-NEXT:    csrr a0, vlenb
; CHECK-NEXT:    vsetvli a1, zero, e32, m1, ta, ma
; CHECK-NEXT:    vid.v v9
; CHECK-NEXT:    srli a0, a0, 2
; CHECK-NEXT:    addi a1, a0, -1
; CHECK-NEXT:    vrsub.vx v9, v9, a1
; CHECK-NEXT:    vrgather.vv v10, v8, v9
; CHECK-NEXT:    sub a0, a0, a2
; CHECK-NEXT:    vslidedown.vx v8, v10, a0
; CHECK-NEXT:    ret
  %load = call <vscale x 2 x float> @llvm.vp.load(ptr %ptr, <vscale x 2 x i1> splat (i1 true), i32 %evl1)
  %rev = call <vscale x 2 x float> @llvm.vector.reverse(<vscale x 2 x float> %load)
  %splice = call <vscale x 2 x float> @llvm.vector.splice.right(<vscale x 2 x float> %rev, <vscale x 2 x float> poison, i32 %evl2)
  ret <vscale x 2 x float> %splice
}