; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
; RUN: llc -mtriple=aarch64-none-linux-gnu %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
; RUN: llc -mtriple=aarch64-none-linux-gnu -global-isel %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI

; ===== Legal Scalars =====

define i8 @load_i8(ptr %ptr){
; CHECK-LABEL: load_i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldrb w0, [x0]
; CHECK-NEXT:    ret
    %a = load i8 , ptr %ptr
    ret i8 %a
}

define i16 @load_i16(ptr %ptr){
; CHECK-LABEL: load_i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldrh w0, [x0]
; CHECK-NEXT:    ret
    %a = load i16 , ptr %ptr
    ret i16 %a
}

define i32 @load_i32(ptr %ptr){
; CHECK-LABEL: load_i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr w0, [x0]
; CHECK-NEXT:    ret
    %a = load i32 , ptr %ptr
    ret i32 %a
}

define i64 @load_i64(ptr %ptr){
; CHECK-LABEL: load_i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr x0, [x0]
; CHECK-NEXT:    ret
    %a = load i64 , ptr %ptr
    ret i64 %a
}

; ===== Legal Vector Types =====

define <8 x i8> @load_v8i8(ptr %ptr){
; CHECK-LABEL: load_v8i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    ret
    %a = load <8 x i8>, ptr %ptr
    ret <8 x i8> %a
}

define <16 x i8> @load_v16i8(ptr %ptr){
; CHECK-LABEL: load_v16i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    ret
    %a = load <16 x i8>, ptr %ptr
    ret <16 x i8> %a
}

define <4 x i16> @load_v4i16(ptr %ptr){
; CHECK-LABEL: load_v4i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    ret
    %a = load <4 x i16>, ptr %ptr
    ret <4 x i16> %a
}

define <8 x i16> @load_v8i16(ptr %ptr){
; CHECK-LABEL: load_v8i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    ret
    %a = load <8 x i16>, ptr %ptr
    ret <8 x i16> %a
}

define <2 x i32> @load_v2i32(ptr %ptr){
; CHECK-LABEL: load_v2i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    ret
    %a = load <2 x i32>, ptr %ptr
    ret <2 x i32> %a
}

define <4 x i32> @load_v4i32(ptr %ptr){
; CHECK-LABEL: load_v4i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    ret
    %a = load <4 x i32>, ptr %ptr
    ret <4 x i32> %a
}

define <2 x i64> @load_v2i64(ptr %ptr){
; CHECK-LABEL: load_v2i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    ret
    %a = load <2 x i64>, ptr %ptr
    ret <2 x i64> %a
}

; ===== Smaller/Larger Width Vectors with Legal Element Sizes =====

define <2 x i8> @load_v2i8(ptr %ptr, <2 x i8> %b){
; CHECK-SD-LABEL: load_v2i8:
; CHECK-SD:       // %bb.0:
; CHECK-SD-NEXT:    ld1 { v0.b }[0], [x0]
; CHECK-SD-NEXT:    add x8, x0, #1
; CHECK-SD-NEXT:    ld1 { v0.b }[4], [x8]
; CHECK-SD-NEXT:    // kill: def $d0 killed $d0 killed $q0
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: load_v2i8:
; CHECK-GI:       // %bb.0:
; CHECK-GI-NEXT:    ld1 { v0.b }[0], [x0]
; CHECK-GI-NEXT:    ldr b1, [x0, #1]
; CHECK-GI-NEXT:    mov v0.s[1], v1.s[0]
; CHECK-GI-NEXT:    // kill: def $d0 killed $d0 killed $q0
; CHECK-GI-NEXT:    ret
    %a = load <2 x i8>, ptr %ptr
    ret <2 x i8> %a
}

define i32 @load_v4i8(ptr %ptr, <4 x i8> %b){
; CHECK-LABEL: load_v4i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr w0, [x0]
; CHECK-NEXT:    ret
    %a = load <4 x i8>, ptr %ptr
    %c = bitcast <4 x i8> %a to i32
    ret i32 %c
}

define <32 x i8> @load_v32i8(ptr %ptr){
; CHECK-LABEL: load_v32i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q0, q1, [x0]
; CHECK-NEXT:    ret
    %a = load <32 x i8>, ptr %ptr
    ret <32 x i8> %a
}

define <2 x i16> @load_v2i16(ptr %ptr){
; CHECK-SD-LABEL: load_v2i16:
; CHECK-SD:       // %bb.0:
; CHECK-SD-NEXT:    ld1 { v0.h }[0], [x0]
; CHECK-SD-NEXT:    add x8, x0, #2
; CHECK-SD-NEXT:    ld1 { v0.h }[2], [x8]
; CHECK-SD-NEXT:    // kill: def $d0 killed $d0 killed $q0
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: load_v2i16:
; CHECK-GI:       // %bb.0:
; CHECK-GI-NEXT:    ld1 { v0.h }[0], [x0]
; CHECK-GI-NEXT:    ldr h1, [x0, #2]
; CHECK-GI-NEXT:    mov v0.s[1], v1.s[0]
; CHECK-GI-NEXT:    // kill: def $d0 killed $d0 killed $q0
; CHECK-GI-NEXT:    ret
    %a = load <2 x i16>, ptr %ptr
    ret <2 x i16> %a
}

define <16 x i16> @load_v16i16(ptr %ptr){
; CHECK-LABEL: load_v16i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q0, q1, [x0]
; CHECK-NEXT:    ret
    %a = load <16 x i16>, ptr %ptr
    ret <16 x i16> %a
}

define <1 x i32> @load_v1i32(ptr %ptr){
; CHECK-LABEL: load_v1i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr s0, [x0]
; CHECK-NEXT:    ret
    %a = load <1 x i32>, ptr %ptr
    ret <1 x i32> %a
}

define <8 x i32> @load_v8i32(ptr %ptr){
; CHECK-LABEL: load_v8i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q0, q1, [x0]
; CHECK-NEXT:    ret
    %a = load <8 x i32>, ptr %ptr
    ret <8 x i32> %a
}

define <4 x i64> @load_v4i64(ptr %ptr){
; CHECK-LABEL: load_v4i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q0, q1, [x0]
; CHECK-NEXT:    ret
    %a = load <4 x i64>, ptr %ptr
    ret <4 x i64> %a
}

; ===== Vectors with Non-Pow 2 Widths =====

define <3 x i8> @load_v3i8(ptr %ptr){
; CHECK-SD-LABEL: load_v3i8:
; CHECK-SD:       // %bb.0:
; CHECK-SD-NEXT:    ldr s0, [x0]
; CHECK-SD-NEXT:    umov w0, v0.b[0]
; CHECK-SD-NEXT:    umov w1, v0.b[1]
; CHECK-SD-NEXT:    umov w2, v0.b[2]
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: load_v3i8:
; CHECK-GI:       // %bb.0:
; CHECK-GI-NEXT:    ldrb w8, [x0]
; CHECK-GI-NEXT:    ldrb w1, [x0, #1]
; CHECK-GI-NEXT:    ldrb w2, [x0, #2]
; CHECK-GI-NEXT:    mov w0, w8
; CHECK-GI-NEXT:    ret
    %a = load <3 x i8>, ptr %ptr
    ret <3 x i8> %a
}

define <7 x i8> @load_v7i8(ptr %ptr){
; CHECK-SD-LABEL: load_v7i8:
; CHECK-SD:       // %bb.0:
; CHECK-SD-NEXT:    ldr d0, [x0]
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: load_v7i8:
; CHECK-GI:       // %bb.0:
; CHECK-GI-NEXT:    ldr b0, [x0]
; CHECK-GI-NEXT:    ldr b1, [x0, #1]
; CHECK-GI-NEXT:    mov v0.b[0], v0.b[0]
; CHECK-GI-NEXT:    mov v0.b[1], v1.b[0]
; CHECK-GI-NEXT:    ldr b1, [x0, #2]
; CHECK-GI-NEXT:    mov v0.b[2], v1.b[0]
; CHECK-GI-NEXT:    ldr b1, [x0, #3]
; CHECK-GI-NEXT:    mov v0.b[3], v1.b[0]
; CHECK-GI-NEXT:    ldr b1, [x0, #4]
; CHECK-GI-NEXT:    mov v0.b[4], v1.b[0]
; CHECK-GI-NEXT:    ldr b1, [x0, #5]
; CHECK-GI-NEXT:    mov v0.b[5], v1.b[0]
; CHECK-GI-NEXT:    ldr b1, [x0, #6]
; CHECK-GI-NEXT:    mov v0.b[6], v1.b[0]
; CHECK-GI-NEXT:    // kill: def $d0 killed $d0 killed $q0
; CHECK-GI-NEXT:    ret
    %a = load <7 x i8>, ptr %ptr
    ret <7 x i8> %a
}

define <3 x i16> @load_v3i16(ptr %ptr){
; CHECK-SD-LABEL: load_v3i16:
; CHECK-SD:       // %bb.0:
; CHECK-SD-NEXT:    ldr d0, [x0]
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: load_v3i16:
; CHECK-GI:       // %bb.0:
; CHECK-GI-NEXT:    ldr h0, [x0]
; CHECK-GI-NEXT:    add x8, x0, #2
; CHECK-GI-NEXT:    ld1 { v0.h }[1], [x8]
; CHECK-GI-NEXT:    add x8, x0, #4
; CHECK-GI-NEXT:    ld1 { v0.h }[2], [x8]
; CHECK-GI-NEXT:    // kill: def $d0 killed $d0 killed $q0
; CHECK-GI-NEXT:    ret
    %a = load <3 x i16>, ptr %ptr
    ret <3 x i16> %a
}

define <7 x i16> @load_v7i16(ptr %ptr){
; CHECK-SD-LABEL: load_v7i16:
; CHECK-SD:       // %bb.0:
; CHECK-SD-NEXT:    ldr q0, [x0]
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: load_v7i16:
; CHECK-GI:       // %bb.0:
; CHECK-GI-NEXT:    ldr h0, [x0]
; CHECK-GI-NEXT:    add x8, x0, #2
; CHECK-GI-NEXT:    ld1 { v0.h }[1], [x8]
; CHECK-GI-NEXT:    add x8, x0, #4
; CHECK-GI-NEXT:    ld1 { v0.h }[2], [x8]
; CHECK-GI-NEXT:    add x8, x0, #6
; CHECK-GI-NEXT:    ld1 { v0.h }[3], [x8]
; CHECK-GI-NEXT:    add x8, x0, #8
; CHECK-GI-NEXT:    ld1 { v0.h }[4], [x8]
; CHECK-GI-NEXT:    add x8, x0, #10
; CHECK-GI-NEXT:    ld1 { v0.h }[5], [x8]
; CHECK-GI-NEXT:    add x8, x0, #12
; CHECK-GI-NEXT:    ld1 { v0.h }[6], [x8]
; CHECK-GI-NEXT:    ret
    %a = load <7 x i16>, ptr %ptr
    ret <7 x i16> %a
}

define <3 x i32> @load_v3i32(ptr %ptr){
; CHECK-SD-LABEL: load_v3i32:
; CHECK-SD:       // %bb.0:
; CHECK-SD-NEXT:    ldr q0, [x0]
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: load_v3i32:
; CHECK-GI:       // %bb.0:
; CHECK-GI-NEXT:    ldr s0, [x0]
; CHECK-GI-NEXT:    add x8, x0, #4
; CHECK-GI-NEXT:    ld1 { v0.s }[1], [x8]
; CHECK-GI-NEXT:    add x8, x0, #8
; CHECK-GI-NEXT:    ld1 { v0.s }[2], [x8]
; CHECK-GI-NEXT:    ret
    %a = load <3 x i32>, ptr %ptr
    ret <3 x i32> %a
}

define <2 x i128> @load_v2i128(ptr %p) {
; CHECK-SD-LABEL: load_v2i128:
; CHECK-SD:       // %bb.0:
; CHECK-SD-NEXT:    ldp x8, x1, [x0]
; CHECK-SD-NEXT:    ldp x2, x3, [x0, #16]
; CHECK-SD-NEXT:    mov x0, x8
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: load_v2i128:
; CHECK-GI:       // %bb.0:
; CHECK-GI-NEXT:    ldp q0, q1, [x0]
; CHECK-GI-NEXT:    mov d2, v0.d[1]
; CHECK-GI-NEXT:    mov d3, v1.d[1]
; CHECK-GI-NEXT:    fmov x0, d0
; CHECK-GI-NEXT:    fmov x2, d1
; CHECK-GI-NEXT:    fmov x1, d2
; CHECK-GI-NEXT:    fmov x3, d3
; CHECK-GI-NEXT:    ret
    %a = load <2 x i128>, ptr %p
    ret <2 x i128> %a
}

define <2 x fp128> @load_v2f128(ptr %p) {
; CHECK-LABEL: load_v2f128:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q0, q1, [x0]
; CHECK-NEXT:    ret
    %a = load <2 x fp128>, ptr %p
    ret <2 x fp128> %a
}