1; REQUIRES: aarch64-registered-target
2
3; This test needs to be target specific due to the cost estimate in the output.
4
5; RUN: opt -lower-matrix-intrinsics -pass-remarks=lower-matrix-intrinsics -mtriple=arm64-apple-iphoneos -S < %s 2>&1 | FileCheck  %s
6
7; Test the propagation of matrix expressions along to inlined-at chain. The IR
8; in the test roughly corresponds to the C++ code below, with the IR containing
9; references to a few more functions.
10
11; matrix.h
12; template <typename Ty, unsigned R, unsigned C>
13; struct Matrix {
14;   using matrix_t = Ty __attribute__((matrix_type(R, C)));
15;
16;   matrix_t value;
17; };
18;
19; ; add.h
20; template <typename Ty, unsigned R, unsigned C>
21; Matrix<Ty, R, C> add(Matrix<Ty, R, C> M1, Matrix<Ty, R, C> M2) {
22;   Matrix<Ty, R, C> Result;
23;   Result.value = __builtin_matrix_add(M1.value, M2.value);
24;   return Result;
25; }
26;
27; load.h:
28; template <typename Ty, unsigned R, unsigned C>
29; Matrix<Ty, R, C> load(Ty *Ptr) {
30;   Matrix<Ty, R, C> Result;
31;   Result.value = *reinterpret_cast <typename Matrix<Ty, R, C>::matrix_t *>(Ptr);
32;   return Result;
33; }
34;
35; store.h:
36; template <typename Ty, unsigned R, unsigned C>
37; void store(Matrix<Ty, R, C> M1, Ty *Ptr) {
38;   *reinterpret_cast<typename decltype(M1)::matrix_t *>(Ptr) = M1.value;
39; }
40;
41; toplevel.cpp
42; void test(double *A, double *B, double *C) {
43;   store(add(load<double, 3, 5>(A), load<double, 3, 5>(B)), C);
44; }
45;
46
47target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
48target triple = "aarch64-apple-ios"
49
50; CHECK-LABEL: remark: load.h:41:43: Lowered with 0 stores, 10 loads, 0 compute ops
51; CHECK-NEXT:  load(addr %A)
52
53; CHECK-LABEL: remark: load.h:41:43: Lowered with 0 stores, 10 loads, 0 compute ops
54; CHECK-NEXT:  column.major.load.3x5.double(addr %B, 5)
55
56; CHECK-LABEL: remark: load.h:41:11: Lowered with 0 stores, 1 loads, 0 compute ops
57; CHECK-NEXT: load(addr %D)
58
59; CHECK-LABEL: remark: assign.h:32:43: Lowered with 0 stores, 10 loads, 0 compute ops
60; CHECK-NEXT:  load(addr %A)
61
62; CHECK-LABEL: remark: assign.h:32:43: Lowered with 0 stores, 10 loads, 0 compute ops
63; CHECK-NEXT:  column.major.load.3x5.double(addr %B, 5)
64
65; CHECK-LABEL: remark: toplevel.c:410:0: Lowered with 10 stores, 20 loads, 10 compute ops
66; CHECK-NEXT:  store(
67; CHECK-NEXT:   fadd(
68; CHECK-NEXT:    load(addr %A),
69; CHECK-NEXT:    column.major.load.3x5.double(addr %B, 5)),
70; CHECK-NEXT:   addr %C)
71
72; CHECK-LABEL: remark: toplevel.c:510:0: Lowered with 1 stores, 1 loads, 8 compute ops
73; CHECK-NEXT:  store(
74; CHECK-NEXT:   transpose.1x2.float(transpose.2x1.float(load(addr %D))),
75; CHECK-NEXT:   addr %D)
76
77; CHECK-LABEL: remark: add.h:66:11: Lowered with 0 stores, 0 loads, 10 compute ops
78; CHECK-NEXT:  fadd(
79; CHECK-NEXT:   addr %A,
80; CHECK-NEXT:   scalar)
81
82; CHECK-LABEL: remark: store.h:10:11: Lowered with 10 stores, 0 loads, 0 compute ops
83; CHECK-NEXT:  store(
84; CHECK-NEXT:   scalar,
85; CHECK-NEXT:   addr %C)
86
87; CHECK-LABEL: remark: store.h:66:11: Lowered with 1 stores, 0 loads, 0 compute ops
88; CHECK-NEXT:  store(
89; CHECK-NEXT:  scalar,
90; CHECK-NEXT:  addr %D)
91
92; CHECK-LABEL: remark: transpose.h:13:11: Lowered with 0 stores, 0 loads, 8 compute ops
93; CHECK-NEXT:  transpose.1x2.float(transpose.2x1.float(addr %D))
94
95define void @toplevel(<15 x double>* %A, double* %B, <15 x double>* %C, <2 x float>* %D) !dbg !16 {
96entry:
97  %a = load <15 x double>, <15 x double> *%A, align 16, !dbg !3791
98  %b = call <15 x double> @llvm.matrix.column.major.load(double* %B, i64 5, i1 false, i32 3, i32 5), !dbg !3793
99  %c  = fadd <15 x double> %a, %b, !dbg !100
100  store <15 x double> %c, <15 x double> *%C, align 16, !dbg !102
101
102  %load = load <2 x float>, <2 x float>* %D, !dbg !104
103  %t1 = call <2 x float> @llvm.matrix.transpose(<2 x float> %load, i32 2, i32 1), !dbg !106
104  %t2 = call <2 x float> @llvm.matrix.transpose(<2 x float> %t1, i32 1, i32 2), !dbg !106
105  store <2 x float> %t2, <2 x float>* %D, !dbg !108
106  ret void
107}
108
109declare <15 x double> @llvm.matrix.column.major.load(double*, i64, i1, i32, i32)
110declare <2 x float> @llvm.matrix.transpose(<2 x float>, i32, i32)
111
112!llvm.dbg.cu = !{!0}
113!llvm.module.flags = !{!3, !4}
114
115!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2)
116!1 = !DIFile(filename: "load.h", directory: "/test")
117!2 = !{}
118!3 = !{i32 2, !"Dwarf Version", i32 4}
119!4 = !{i32 2, !"Debug Info Version", i32 3}
120!5 = distinct !DISubprogram(name: "load_fn", scope: !1, file: !1, line: 1, type: !6, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !12)
121!17 = !DIFile(filename: "toplevel.c", directory: "/test")
122!16 = distinct !DISubprogram(name: "toplevel", scope: !1, file: !17, line: 1, type: !6, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !12)
123!18 = !DIFile(filename: "assign.h", directory: "/test")
124!19 = distinct !DISubprogram(name: "assign", scope: !1, file: !18, line: 1, type: !6, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !12)
125
126!20 = !DIFile(filename: "add.h", directory: "/test")
127!21 = distinct !DISubprogram(name: "add_fn", scope: !1, file: !20, line: 1, type: !6, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !12)
128
129!22 = !DIFile(filename: "store.h", directory: "/test")
130!23 = distinct !DISubprogram(name: "store_fn", scope: !1, file: !22, line: 1, type: !6, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !12)
131
132!24 = !DIFile(filename: "transpose.h", directory: "/test")
133!25 = distinct !DISubprogram(name: "transpose", scope: !1, file: !24, line: 1, type: !6, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !12)
134
135
136!6 = !DISubroutineType(types: !7)
137!7 = !{null, !8, !8, !11}
138!8 = !DIDerivedType(tag: DW_TAG_restrict_type, baseType: !9)
139!9 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !10, size: 32, align: 32)
140!10 = !DIBasicType(name: "float", size: 32, align: 32, encoding: DW_ATE_float)
141!11 = !DIBasicType(name: "int", size: 32, align: 32, encoding: DW_ATE_signed)
142!12 = !{!13}
143!13 = !DILocalVariable(name: "a", arg: 1, scope: !5, file: !1, line: 1, type: !8)
144!14 = !DILocation(line: 1, column: 27, scope: !5)
145
146!3791 = !DILocation(line: 41, column: 43, scope: !5, inlinedAt: !3795)
147!3792 = !DILocation(line: 405, column: 3, scope: !16)
148!3793 = !DILocation(line: 41, column: 43, scope: !5, inlinedAt: !3796)
149!3794 = !DILocation(line: 406, column: 11, scope: !16)
150!3795 = !DILocation(line: 32, column: 43, scope: !19, inlinedAt: !3792)
151!3796 = !DILocation(line: 32, column: 43, scope: !19, inlinedAt: !3794)
152
153!100 = !DILocation(line: 66, column: 11, scope: !21, inlinedAt: !101)
154!101 = !DILocation(line: 410, column: 11, scope: !16)
155
156!102 = !DILocation(line: 10, column: 11, scope: !23, inlinedAt: !103)
157!103 = !DILocation(line: 410, column: 0, scope: !16)
158
159!104 = !DILocation(line: 41, column: 11, scope: !5, inlinedAt: !101)
160!105 = !DILocation(line: 500, column: 11, scope: !16)
161
162!106 = !DILocation(line: 13, column: 11, scope: !25, inlinedAt: !101)
163!107 = !DILocation(line: 510, column: 11, scope: !16)
164
165!108 = !DILocation(line: 66, column: 11, scope: !23, inlinedAt: !109)
166!109 = !DILocation(line: 510, column: 0, scope: !16)
167