1# RUN: llc -o /dev/null %s -mtriple=arm-eabi -mcpu=cortex-r52 -run-pass  machine-scheduler  -enable-misched -debug-only=machine-scheduler -misched-topdown 2>&1 | FileCheck %s --check-prefix=CHECK --check-prefix=TOPDOWN
2# RUN: llc -o /dev/null %s -mtriple=arm-eabi -mcpu=cortex-r52 -run-pass  machine-scheduler  -enable-misched -debug-only=machine-scheduler -misched-bottomup 2>&1 | FileCheck %s --check-prefix=CHECK --check-prefix=BOTTOMUP
3# REQUIRES: asserts
4--- |
5  ; ModuleID = 'foo.ll'
6  source_filename = "foo.ll"
7  target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64"
8  target triple = "arm---eabi"
9
10  %struct.__neon_int8x8x4_t = type { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }
11  ; Function Attrs: nounwind
12  define <8 x i8> @foo(i8* %A) {
13    %tmp1 = call %struct.__neon_int8x8x4_t @llvm.arm.neon.vld4.v8i8.p0i8(i8* %A, i32 8)
14    %tmp2 = extractvalue %struct.__neon_int8x8x4_t %tmp1, 0
15    %tmp3 = extractvalue %struct.__neon_int8x8x4_t %tmp1, 1
16    %tmp4 = add <8 x i8> %tmp2, %tmp3
17    ret <8 x i8> %tmp4
18  }
19  declare %struct.__neon_int8x8x4_t @llvm.arm.neon.vld4.v8i8.p0i8(i8*, i32)
20
21# CHECK: ********** MI Scheduling **********
22# CHECK: ScheduleDAGMILive::schedule starting
23# CHECK: SU(1):   %1:qqpr = VLD4d8Pseudo %0:gpr, 8, 14, $noreg :: (load 32 from %ir.A, align 8)
24# CHECK: Latency            : 8
25# CHECK: Single Issue       : true;
26# CHECK: SU(2):   %4:dpr = VADDv8i8 %1.dsub_0:qqpr, %1.dsub_1:qqpr, 14, $noreg
27# CHECK: Latency            : 5
28# CHECK: Single Issue       : false;
29# CHECK: SU(3):   %5:gpr, %6:gpr = VMOVRRD %4:dpr, 14, $noreg
30# CHECK: Latency            : 4
31# CHECK: Single Issue       : false;
32
33# TOPDOWN: Scheduling SU(1) %1:qqpr = VLD4d8Pseudo
34# TOPDOWN: Bump cycle to end group
35# TOPDOWN: Scheduling SU(2) %4:dpr = VADDv8i8
36
37# BOTTOMUP: Scheduling SU(2) %4:dpr = VADDv8i8
38# BOTTOMUP: Scheduling SU(1) %1:qqpr = VLD4d8Pseudo
39# BOTTOMUP: Bump cycle to begin group
40
41...
42---
43name:            foo
44alignment:       2
45exposesReturnsTwice: false
46legalized:       false
47regBankSelected: false
48selected:        false
49tracksRegLiveness: true
50registers:
51  - { id: 0, class: gpr }
52  - { id: 1, class: qqpr }
53  - { id: 2, class: dpr }
54  - { id: 3, class: dpr }
55  - { id: 4, class: dpr }
56  - { id: 5, class: gpr }
57  - { id: 6, class: gpr }
58liveins:
59  - { reg: '$r0', virtual-reg: '%0' }
60frameInfo:
61  isFrameAddressTaken: false
62  isReturnAddressTaken: false
63  hasStackMap:     false
64  hasPatchPoint:   false
65  stackSize:       0
66  offsetAdjustment: 0
67  maxAlignment:    0
68  adjustsStack:    false
69  hasCalls:        false
70  maxCallFrameSize: 0
71  hasOpaqueSPAdjustment: false
72  hasVAStart:      false
73  hasMustTailInVarArgFunc: false
74body:             |
75  bb.0 (%ir-block.0):
76    liveins: $r0
77
78    %0 = COPY $r0
79    %1 = VLD4d8Pseudo %0, 8, 14, $noreg :: (load 32 from %ir.A, align 8)
80    %4 = VADDv8i8 %1.dsub_0, %1.dsub_1, 14, $noreg
81    %5, %6 = VMOVRRD %4, 14, $noreg
82    $r0 = COPY %5
83    $r1 = COPY %6
84    BX_RET 14, $noreg, implicit $r0, implicit killed $r1
85
86...
87