; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -verify-machineinstrs -mattr=+simd128 | FileCheck %s

; Test SIMD v128.load{32,64}_zero instructions

target triple = "wasm32-unknown-unknown"

;===----------------------------------------------------------------------------
; v128.load32_zero
;===----------------------------------------------------------------------------

define <4 x i32> @load_zero_i32_no_offset(ptr %p) {
; CHECK-LABEL: load_zero_i32_no_offset:
; CHECK:         .functype load_zero_i32_no_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    v128.load32_zero 0
; CHECK-NEXT:    # fallthrough-return
  %x = load i32, ptr %p
  %v = insertelement <4 x i32> zeroinitializer, i32 %x, i32 0
  ret <4 x i32> %v
}

define <4 x i32> @load_zero_i32_with_folded_offset(ptr %p) {
; CHECK-LABEL: load_zero_i32_with_folded_offset:
; CHECK:         .functype load_zero_i32_with_folded_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    v128.load32_zero 24
; CHECK-NEXT:    # fallthrough-return
  %q = ptrtoint ptr %p to i32
  %r = add nuw i32 %q, 24
  %s = inttoptr i32 %r to ptr
  %x = load i32, ptr %s
  %t = insertelement <4 x i32> zeroinitializer, i32 %x, i32 0
  ret <4 x i32> %t
}

define <4 x i32> @load_zero_i32_with_folded_gep_offset(ptr %p) {
; CHECK-LABEL: load_zero_i32_with_folded_gep_offset:
; CHECK:         .functype load_zero_i32_with_folded_gep_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    v128.load32_zero 24
; CHECK-NEXT:    # fallthrough-return
  %s = getelementptr inbounds i32, ptr %p, i32 6
  %x = load i32, ptr %s
  %t = insertelement <4 x i32> zeroinitializer, i32 %x, i32 0
  ret <4 x i32> %t
}

define <4 x i32> @load_zero_i32_with_unfolded_gep_negative_offset(ptr %p) {
; CHECK-LABEL: load_zero_i32_with_unfolded_gep_negative_offset:
; CHECK:         .functype load_zero_i32_with_unfolded_gep_negative_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32.const -24
; CHECK-NEXT:    i32.add
; CHECK-NEXT:    v128.load32_zero 0
; CHECK-NEXT:    # fallthrough-return
  %s = getelementptr inbounds i32, ptr %p, i32 -6
  %x = load i32, ptr %s
  %t = insertelement <4 x i32> zeroinitializer, i32 %x, i32 0
  ret <4 x i32> %t
}

define <4 x i32> @load_zero_i32_with_unfolded_offset(ptr %p) {
; CHECK-LABEL: load_zero_i32_with_unfolded_offset:
; CHECK:         .functype load_zero_i32_with_unfolded_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32.const 24
; CHECK-NEXT:    i32.add
; CHECK-NEXT:    v128.load32_zero 0
; CHECK-NEXT:    # fallthrough-return
  %q = ptrtoint ptr %p to i32
  %r = add nsw i32 %q, 24
  %s = inttoptr i32 %r to ptr
  %x = load i32, ptr %s
  %t = insertelement <4 x i32> zeroinitializer, i32 %x, i32 0
  ret <4 x i32> %t
}

define <4 x i32> @load_zero_i32_with_unfolded_gep_offset(ptr %p) {
; CHECK-LABEL: load_zero_i32_with_unfolded_gep_offset:
; CHECK:         .functype load_zero_i32_with_unfolded_gep_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32.const 24
; CHECK-NEXT:    i32.add
; CHECK-NEXT:    v128.load32_zero 0
; CHECK-NEXT:    # fallthrough-return
  %s = getelementptr i32, ptr %p, i32 6
  %x = load i32, ptr %s
  %t = insertelement <4 x i32> zeroinitializer, i32 %x, i32 0
  ret <4 x i32> %t
}

define <4 x i32> @load_zero_i32_from_numeric_address() {
; CHECK-LABEL: load_zero_i32_from_numeric_address:
; CHECK:         .functype load_zero_i32_from_numeric_address () -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    i32.const 0
; CHECK-NEXT:    v128.load32_zero 42
; CHECK-NEXT:    # fallthrough-return
  %s = inttoptr i32 42 to ptr
  %x = load i32, ptr %s
  %t = insertelement <4 x i32> zeroinitializer, i32 %x, i32 0
  ret <4 x i32> %t
}

@gv_i32 = global i32 0
define <4 x i32> @load_zero_i32_from_global_address() {
; CHECK-LABEL: load_zero_i32_from_global_address:
; CHECK:         .functype load_zero_i32_from_global_address () -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    i32.const 0
; CHECK-NEXT:    v128.load32_zero gv_i32
; CHECK-NEXT:    # fallthrough-return
  %x = load i32, ptr @gv_i32
  %t = insertelement <4 x i32> zeroinitializer, i32 %x, i32 0
  ret <4 x i32> %t
}

;===----------------------------------------------------------------------------
; v128.load64_zero
;===----------------------------------------------------------------------------

define <2 x i64> @load_zero_i64_no_offset(ptr %p) {
; CHECK-LABEL: load_zero_i64_no_offset:
; CHECK:         .functype load_zero_i64_no_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    v128.load64_zero 0
; CHECK-NEXT:    # fallthrough-return
  %x = load i64, ptr %p
  %v = insertelement <2 x i64> zeroinitializer, i64 %x, i32 0
  ret <2 x i64> %v
}

define <2 x i64> @load_zero_i64_with_folded_offset(ptr %p) {
; CHECK-LABEL: load_zero_i64_with_folded_offset:
; CHECK:         .functype load_zero_i64_with_folded_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    v128.load64_zero 24
; CHECK-NEXT:    # fallthrough-return
  %q = ptrtoint ptr %p to i32
  %r = add nuw i32 %q, 24
  %s = inttoptr i32 %r to ptr
  %x = load i64, ptr %s
  %t = insertelement <2 x i64> zeroinitializer, i64 %x, i32 0
  ret <2 x i64> %t
}

define <2 x i64> @load_zero_i64_with_folded_gep_offset(ptr %p) {
; CHECK-LABEL: load_zero_i64_with_folded_gep_offset:
; CHECK:         .functype load_zero_i64_with_folded_gep_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    v128.load64_zero 48
; CHECK-NEXT:    # fallthrough-return
  %s = getelementptr inbounds i64, ptr %p, i64 6
  %x = load i64, ptr %s
  %t = insertelement <2 x i64> zeroinitializer, i64 %x, i32 0
  ret <2 x i64> %t
}

define <2 x i64> @load_zero_i64_with_unfolded_gep_negative_offset(ptr %p) {
; CHECK-LABEL: load_zero_i64_with_unfolded_gep_negative_offset:
; CHECK:         .functype load_zero_i64_with_unfolded_gep_negative_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32.const -48
; CHECK-NEXT:    i32.add
; CHECK-NEXT:    v128.load64_zero 0
; CHECK-NEXT:    # fallthrough-return
  %s = getelementptr inbounds i64, ptr %p, i64 -6
  %x = load i64, ptr %s
  %t = insertelement <2 x i64> zeroinitializer, i64 %x, i32 0
  ret <2 x i64> %t
}

define <2 x i64> @load_zero_i64_with_unfolded_offset(ptr %p) {
; CHECK-LABEL: load_zero_i64_with_unfolded_offset:
; CHECK:         .functype load_zero_i64_with_unfolded_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32.const 24
; CHECK-NEXT:    i32.add
; CHECK-NEXT:    v128.load64_zero 0
; CHECK-NEXT:    # fallthrough-return
  %q = ptrtoint ptr %p to i32
  %r = add nsw i32 %q, 24
  %s = inttoptr i32 %r to ptr
  %x = load i64, ptr %s
  %t = insertelement <2 x i64> zeroinitializer, i64 %x, i32 0
  ret <2 x i64> %t
}

define <2 x i64> @load_zero_i64_with_unfolded_gep_offset(ptr %p) {
; CHECK-LABEL: load_zero_i64_with_unfolded_gep_offset:
; CHECK:         .functype load_zero_i64_with_unfolded_gep_offset (i32) -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    local.get 0
; CHECK-NEXT:    i32.const 48
; CHECK-NEXT:    i32.add
; CHECK-NEXT:    v128.load64_zero 0
; CHECK-NEXT:    # fallthrough-return
  %s = getelementptr i64, ptr %p, i64 6
  %x = load i64, ptr %s
  %t = insertelement <2 x i64> zeroinitializer, i64 %x, i32 0
  ret <2 x i64> %t
}

define <2 x i64> @load_zero_i64_from_numeric_address() {
; CHECK-LABEL: load_zero_i64_from_numeric_address:
; CHECK:         .functype load_zero_i64_from_numeric_address () -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    i32.const 0
; CHECK-NEXT:    v128.load64_zero 42
; CHECK-NEXT:    # fallthrough-return
  %s = inttoptr i32 42 to ptr
  %x = load i64, ptr %s
  %t = insertelement <2 x i64> zeroinitializer, i64 %x, i32 0
  ret <2 x i64> %t
}

@gv_i64 = global i64 0
define <2 x i64> @load_zero_i64_from_global_address() {
; CHECK-LABEL: load_zero_i64_from_global_address:
; CHECK:         .functype load_zero_i64_from_global_address () -> (v128)
; CHECK-NEXT:  # %bb.0:
; CHECK-NEXT:    i32.const 0
; CHECK-NEXT:    v128.load64_zero gv_i64
; CHECK-NEXT:    # fallthrough-return
  %x = load i64, ptr @gv_i64
  %t = insertelement <2 x i64> zeroinitializer, i64 %x, i32 0
  ret <2 x i64> %t
}