llvm/test/CodeGen/AArch64/arm64-neon-vector-list-spill.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc < %s -verify-machineinstrs -mtriple=arm64-none-linux-gnu -mattr=+neon -fp-contract=fast | FileCheck %s

; FIXME: We should not generate ld/st for such register spill/fill, because the
; test case seems very simple and the register pressure is not high. If the
; spill/fill algorithm is optimized, this test case may not be triggered. And
; then we can delete it.
define i32 @spill.DPairReg(ptr %arg1, i32 %arg2) {
; CHECK-LABEL: spill.DPairReg:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ld2 { v0.2s, v1.2s }, [x0]
; CHECK-NEXT:    cbz w1, .LBB0_2
; CHECK-NEXT:  // %bb.1: // %if.end
; CHECK-NEXT:    mov w0, v0.s[1]
; CHECK-NEXT:    ret
; CHECK-NEXT:  .LBB0_2: // %if.then
; CHECK-NEXT:    sub sp, sp, #48
; CHECK-NEXT:    stp x29, x30, [sp, #32] // 16-byte Folded Spill
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset w30, -8
; CHECK-NEXT:    .cfi_offset w29, -16
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    st1 { v0.2d, v1.2d }, [x8] // 32-byte Folded Spill
; CHECK-NEXT:    bl foo
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    ldp x29, x30, [sp, #32] // 16-byte Folded Reload
; CHECK-NEXT:    ld1 { v0.2d, v1.2d }, [x8] // 32-byte Folded Reload
; CHECK-NEXT:    add sp, sp, #48
; CHECK-NEXT:    mov w0, v0.s[1]
; CHECK-NEXT:    ret
entry:
  %vld = tail call { <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld2.v2i32.p0(ptr %arg1)
  %cmp = icmp eq i32 %arg2, 0
  br i1 %cmp, label %if.then, label %if.end

if.then:
  tail call void @foo()
  br label %if.end

if.end:
  %vld.extract = extractvalue { <2 x i32>, <2 x i32> } %vld, 0
  %res = extractelement <2 x i32> %vld.extract, i32 1
  ret i32 %res
}

define i16 @spill.DTripleReg(ptr %arg1, i32 %arg2) {
; CHECK-LABEL: spill.DTripleReg:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ld3 { v0.4h, v1.4h, v2.4h }, [x0]
; CHECK-NEXT:    cbz w1, .LBB1_2
; CHECK-NEXT:  // %bb.1: // %if.end
; CHECK-NEXT:    umov w0, v0.h[1]
; CHECK-NEXT:    ret
; CHECK-NEXT:  .LBB1_2: // %if.then
; CHECK-NEXT:    sub sp, sp, #64
; CHECK-NEXT:    stp x29, x30, [sp, #48] // 16-byte Folded Spill
; CHECK-NEXT:    .cfi_def_cfa_offset 64
; CHECK-NEXT:    .cfi_offset w30, -8
; CHECK-NEXT:    .cfi_offset w29, -16
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    st1 { v0.2d, v1.2d, v2.2d }, [x8] // 48-byte Folded Spill
; CHECK-NEXT:    bl foo
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    ldp x29, x30, [sp, #48] // 16-byte Folded Reload
; CHECK-NEXT:    ld1 { v0.2d, v1.2d, v2.2d }, [x8] // 48-byte Folded Reload
; CHECK-NEXT:    add sp, sp, #64
; CHECK-NEXT:    umov w0, v0.h[1]
; CHECK-NEXT:    ret
entry:
  %vld = tail call { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld3.v4i16.p0(ptr %arg1)
  %cmp = icmp eq i32 %arg2, 0
  br i1 %cmp, label %if.then, label %if.end

if.then:
  tail call void @foo()
  br label %if.end

if.end:
  %vld.extract = extractvalue { <4 x i16>, <4 x i16>, <4 x i16> } %vld, 0
  %res = extractelement <4 x i16> %vld.extract, i32 1
  ret i16 %res
}

define i16 @spill.DQuadReg(ptr %arg1, i32 %arg2) {
; CHECK-LABEL: spill.DQuadReg:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ld4 { v0.4h, v1.4h, v2.4h, v3.4h }, [x0]
; CHECK-NEXT:    cbz w1, .LBB2_2
; CHECK-NEXT:  // %bb.1: // %if.end
; CHECK-NEXT:    umov w0, v0.h[0]
; CHECK-NEXT:    ret
; CHECK-NEXT:  .LBB2_2: // %if.then
; CHECK-NEXT:    sub sp, sp, #80
; CHECK-NEXT:    stp x29, x30, [sp, #64] // 16-byte Folded Spill
; CHECK-NEXT:    .cfi_def_cfa_offset 80
; CHECK-NEXT:    .cfi_offset w30, -8
; CHECK-NEXT:    .cfi_offset w29, -16
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    st1 { v0.2d, v1.2d, v2.2d, v3.2d }, [x8] // 64-byte Folded Spill
; CHECK-NEXT:    bl foo
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    ldp x29, x30, [sp, #64] // 16-byte Folded Reload
; CHECK-NEXT:    ld1 { v0.2d, v1.2d, v2.2d, v3.2d }, [x8] // 64-byte Folded Reload
; CHECK-NEXT:    add sp, sp, #80
; CHECK-NEXT:    umov w0, v0.h[0]
; CHECK-NEXT:    ret
entry:
  %vld = tail call { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld4.v4i16.p0(ptr %arg1)
  %cmp = icmp eq i32 %arg2, 0
  br i1 %cmp, label %if.then, label %if.end

if.then:
  tail call void @foo()
  br label %if.end

if.end:
  %vld.extract = extractvalue { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } %vld, 0
  %res = extractelement <4 x i16> %vld.extract, i32 0
  ret i16 %res
}

define i32 @spill.QPairReg(ptr %arg1, i32 %arg2) {
; CHECK-LABEL: spill.QPairReg:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ld2 { v0.4s, v1.4s }, [x0]
; CHECK-NEXT:    cbz w1, .LBB3_2
; CHECK-NEXT:  // %bb.1: // %if.end
; CHECK-NEXT:    mov w0, v0.s[1]
; CHECK-NEXT:    ret
; CHECK-NEXT:  .LBB3_2: // %if.then
; CHECK-NEXT:    sub sp, sp, #48
; CHECK-NEXT:    stp x29, x30, [sp, #32] // 16-byte Folded Spill
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset w30, -8
; CHECK-NEXT:    .cfi_offset w29, -16
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    st1 { v0.2d, v1.2d }, [x8] // 32-byte Folded Spill
; CHECK-NEXT:    bl foo
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    ldp x29, x30, [sp, #32] // 16-byte Folded Reload
; CHECK-NEXT:    ld1 { v0.2d, v1.2d }, [x8] // 32-byte Folded Reload
; CHECK-NEXT:    add sp, sp, #48
; CHECK-NEXT:    mov w0, v0.s[1]
; CHECK-NEXT:    ret
entry:
  %vld = tail call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0(ptr %arg1)
  %cmp = icmp eq i32 %arg2, 0
  br i1 %cmp, label %if.then, label %if.end

if.then:
  tail call void @foo()
  br label %if.end

if.end:
  %vld.extract = extractvalue { <4 x i32>, <4 x i32> } %vld, 0
  %res = extractelement <4 x i32> %vld.extract, i32 1
  ret i32 %res
}

define float @spill.QTripleReg(ptr %arg1, i32 %arg2) {
; CHECK-LABEL: spill.QTripleReg:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ld3 { v0.4s, v1.4s, v2.4s }, [x0]
; CHECK-NEXT:    cbz w1, .LBB4_2
; CHECK-NEXT:  // %bb.1: // %if.end
; CHECK-NEXT:    mov s0, v0.s[1]
; CHECK-NEXT:    ret
; CHECK-NEXT:  .LBB4_2: // %if.then
; CHECK-NEXT:    sub sp, sp, #64
; CHECK-NEXT:    stp x29, x30, [sp, #48] // 16-byte Folded Spill
; CHECK-NEXT:    .cfi_def_cfa_offset 64
; CHECK-NEXT:    .cfi_offset w30, -8
; CHECK-NEXT:    .cfi_offset w29, -16
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    st1 { v0.2d, v1.2d, v2.2d }, [x8] // 48-byte Folded Spill
; CHECK-NEXT:    bl foo
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    ldp x29, x30, [sp, #48] // 16-byte Folded Reload
; CHECK-NEXT:    ld1 { v0.2d, v1.2d, v2.2d }, [x8] // 48-byte Folded Reload
; CHECK-NEXT:    add sp, sp, #64
; CHECK-NEXT:    mov s0, v0.s[1]
; CHECK-NEXT:    ret
entry:
  %vld3 = tail call { <4 x float>, <4 x float>, <4 x float> } @llvm.aarch64.neon.ld3.v4f32.p0(ptr %arg1)
  %cmp = icmp eq i32 %arg2, 0
  br i1 %cmp, label %if.then, label %if.end

if.then:
  tail call void @foo()
  br label %if.end

if.end:
  %vld3.extract = extractvalue { <4 x float>, <4 x float>, <4 x float> } %vld3, 0
  %res = extractelement <4 x float> %vld3.extract, i32 1
  ret float %res
}

define i8 @spill.QQuadReg(ptr %arg1, i32 %arg2) {
; CHECK-LABEL: spill.QQuadReg:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ld4 { v0.16b, v1.16b, v2.16b, v3.16b }, [x0]
; CHECK-NEXT:    cbz w1, .LBB5_2
; CHECK-NEXT:  // %bb.1: // %if.end
; CHECK-NEXT:    umov w0, v0.b[1]
; CHECK-NEXT:    ret
; CHECK-NEXT:  .LBB5_2: // %if.then
; CHECK-NEXT:    sub sp, sp, #80
; CHECK-NEXT:    stp x29, x30, [sp, #64] // 16-byte Folded Spill
; CHECK-NEXT:    .cfi_def_cfa_offset 80
; CHECK-NEXT:    .cfi_offset w30, -8
; CHECK-NEXT:    .cfi_offset w29, -16
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    st1 { v0.2d, v1.2d, v2.2d, v3.2d }, [x8] // 64-byte Folded Spill
; CHECK-NEXT:    bl foo
; CHECK-NEXT:    mov x8, sp
; CHECK-NEXT:    ldp x29, x30, [sp, #64] // 16-byte Folded Reload
; CHECK-NEXT:    ld1 { v0.2d, v1.2d, v2.2d, v3.2d }, [x8] // 64-byte Folded Reload
; CHECK-NEXT:    add sp, sp, #80
; CHECK-NEXT:    umov w0, v0.b[1]
; CHECK-NEXT:    ret
entry:
  %vld = tail call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld4.v16i8.p0(ptr %arg1)
  %cmp = icmp eq i32 %arg2, 0
  br i1 %cmp, label %if.then, label %if.end

if.then:
  tail call void @foo()
  br label %if.end

if.end:
  %vld.extract = extractvalue { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } %vld, 0
  %res = extractelement <16 x i8> %vld.extract, i32 1
  ret i8 %res
}

declare { <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld2.v2i32.p0(ptr)
declare { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld3.v4i16.p0(ptr)
declare { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld4.v4i16.p0(ptr)
declare { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0(ptr)
declare { <4 x float>, <4 x float>, <4 x float> } @llvm.aarch64.neon.ld3.v4f32.p0(ptr)
declare { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld4.v16i8.p0(ptr)

declare void @foo()

; FIXME: We should not generate ld/st for such register spill/fill, because the
; test case seems very simple and the register pressure is not high. If the
; spill/fill algorithm is optimized, this test case may not be triggered. And
; then we can delete it.
; check the spill for Register Class QPair_with_qsub_0_in_FPR128Lo
define <8 x i16> @test_2xFPR128Lo(i64 %got, ptr %ptr, <1 x i64> %a) {
; CHECK-LABEL: test_2xFPR128Lo:
; CHECK:       // %bb.0:
; CHECK-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    .cfi_offset w30, -16
; CHECK-NEXT:    movi v0.2d, #0000000000000000
; CHECK-NEXT:    mov v1.16b, v0.16b
; CHECK-NEXT:    st2 { v0.d, v1.d }[0], [x1]
; CHECK-NEXT:    bl foo
; CHECK-NEXT:    movi v0.2d, #0000000000000000
; CHECK-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
  tail call void @llvm.aarch64.neon.st2lane.v1i64.p0(<1 x i64> zeroinitializer, <1 x i64> zeroinitializer, i64 0, ptr %ptr)
  tail call void @foo()
  %sv = shufflevector <1 x i64> zeroinitializer, <1 x i64> %a, <2 x i32> <i32 0, i32 1>
  %1 = bitcast <2 x i64> %sv to <8 x i16>
  %2 = shufflevector <8 x i16> %1, <8 x i16> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
  %3 = mul <8 x i16> %2, %2
  ret <8 x i16> %3
}

; check the spill for Register Class QTriple_with_qsub_0_in_FPR128Lo
define <8 x i16> @test_3xFPR128Lo(i64 %got, ptr %ptr, <1 x i64> %a) {
; CHECK-LABEL: test_3xFPR128Lo:
; CHECK:       // %bb.0:
; CHECK-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    .cfi_offset w30, -16
; CHECK-NEXT:    movi v0.2d, #0000000000000000
; CHECK-NEXT:    mov v1.16b, v0.16b
; CHECK-NEXT:    mov v2.16b, v0.16b
; CHECK-NEXT:    st3 { v0.d, v1.d, v2.d }[0], [x1]
; CHECK-NEXT:    bl foo
; CHECK-NEXT:    movi v0.2d, #0000000000000000
; CHECK-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
  tail call void @llvm.aarch64.neon.st3lane.v1i64.p0(<1 x i64> zeroinitializer, <1 x i64> zeroinitializer, <1 x i64> zeroinitializer, i64 0, ptr %ptr)
  tail call void @foo()
  %sv = shufflevector <1 x i64> zeroinitializer, <1 x i64> %a, <2 x i32> <i32 0, i32 1>
  %1 = bitcast <2 x i64> %sv to <8 x i16>
  %2 = shufflevector <8 x i16> %1, <8 x i16> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
  %3 = mul <8 x i16> %2, %2
  ret <8 x i16> %3
}

; check the spill for Register Class QQuad_with_qsub_0_in_FPR128Lo
define <8 x i16> @test_4xFPR128Lo(i64 %got, ptr %ptr, <1 x i64> %a) {
; CHECK-LABEL: test_4xFPR128Lo:
; CHECK:       // %bb.0:
; CHECK-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    .cfi_offset w30, -16
; CHECK-NEXT:    movi v0.2d, #0000000000000000
; CHECK-NEXT:    mov v1.16b, v0.16b
; CHECK-NEXT:    mov v2.16b, v0.16b
; CHECK-NEXT:    mov v3.16b, v0.16b
; CHECK-NEXT:    st4 { v0.d, v1.d, v2.d, v3.d }[0], [x1]
; CHECK-NEXT:    bl foo
; CHECK-NEXT:    movi v0.2d, #0000000000000000
; CHECK-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
  tail call void @llvm.aarch64.neon.st4lane.v1i64.p0(<1 x i64> zeroinitializer, <1 x i64> zeroinitializer, <1 x i64> zeroinitializer, <1 x i64> zeroinitializer, i64 0, ptr %ptr)
  tail call void @foo()
  %sv = shufflevector <1 x i64> zeroinitializer, <1 x i64> %a, <2 x i32> <i32 0, i32 1>
  %1 = bitcast <2 x i64> %sv to <8 x i16>
  %2 = shufflevector <8 x i16> %1, <8 x i16> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
  %3 = mul <8 x i16> %2, %2
  ret <8 x i16> %3
}

declare void @llvm.aarch64.neon.st2lane.v1i64.p0(<1 x i64>, <1 x i64>, i64, ptr)
declare void @llvm.aarch64.neon.st3lane.v1i64.p0(<1 x i64>, <1 x i64>, <1 x i64>, i64, ptr)
declare void @llvm.aarch64.neon.st4lane.v1i64.p0(<1 x i64>, <1 x i64>, <1 x i64>, <1 x i64>, i64, ptr)