xref: /llvm-project/llvm/test/Analysis/CostModel/AArch64/sve-ldst.ll (revision 2dd4167a09cfe1a1e09852e99b4984ddb3e89cf7) 
1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 3
2; RUN: opt  -passes="print<cost-model>" 2>&1 -disable-output -mtriple=aarch64--linux-gnu -mattr=+sve < %s | FileCheck %s
3
4target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
5
6define void @scalable_loads() {
7; CHECK-LABEL: 'scalable_loads'
8; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %res.nxv8i8 = load <vscale x 8 x i8>, ptr undef, align 8
9; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %res.nxv16i8 = load <vscale x 16 x i8>, ptr undef, align 16
10; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %res.nxv32i8 = load <vscale x 32 x i8>, ptr undef, align 32
11; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %res.nxv1i64 = load <vscale x 1 x i64>, ptr undef, align 8
12; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %res.nxv32i1 = load <vscale x 32 x i1>, ptr undef, align 4
13; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %res.nxv16i1 = load <vscale x 16 x i1>, ptr undef, align 2
14; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %res.nxv4i1 = load <vscale x 4 x i1>, ptr undef, align 1
15; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
16;
17  %res.nxv8i8 = load <vscale x 8 x i8>, ptr undef
18  %res.nxv16i8 = load <vscale x 16 x i8>, ptr undef
19  %res.nxv32i8 = load <vscale x 32 x i8>, ptr undef
20  %res.nxv1i64 = load <vscale x 1 x i64>, ptr undef
21  %res.nxv32i1 = load <vscale x 32 x i1>, ptr undef
22  %res.nxv16i1 = load <vscale x 16 x i1>, ptr undef
23  %res.nxv4i1 = load <vscale x 4 x i1>, ptr undef
24  ret void
25}
26
27define void @scalable_stores() {
28; CHECK-LABEL: 'scalable_stores'
29; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: store <vscale x 8 x i8> undef, ptr undef, align 8
30; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: store <vscale x 16 x i8> undef, ptr undef, align 16
31; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: store <vscale x 32 x i8> undef, ptr undef, align 32
32; CHECK-NEXT:  Cost Model: Invalid cost for instruction: store <vscale x 1 x i64> undef, ptr undef, align 8
33; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: store <vscale x 32 x i1> undef, ptr undef, align 4
34; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: store <vscale x 16 x i1> undef, ptr undef, align 2
35; CHECK-NEXT:  Cost Model: Invalid cost for instruction: store <vscale x 4 x i1> undef, ptr undef, align 1
36; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
37;
38  store <vscale x 8 x i8> undef, ptr undef
39  store <vscale x 16 x i8> undef, ptr undef
40  store <vscale x 32 x i8> undef, ptr undef
41  store <vscale x 1 x i64> undef, ptr undef
42  store <vscale x 32 x i1> undef, ptr undef
43  store <vscale x 16 x i1> undef, ptr undef
44  store <vscale x 4 x i1> undef, ptr undef
45  ret void
46}
47
48define void @scalable_ext_loads() {
49; CHECK-LABEL: 'scalable_ext_loads'
50; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv16i8 = load <vscale x 16 x i8>, ptr undef, align 16
51; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %zext.nxv16i8to16 = zext <vscale x 16 x i8> %load.nxv16i8 to <vscale x 16 x i16>
52; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv8i8 = load <vscale x 8 x i8>, ptr undef, align 8
53; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %zext.nxv8i8to16 = zext <vscale x 8 x i8> %load.nxv8i8 to <vscale x 8 x i16>
54; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv4i8 = load <vscale x 4 x i8>, ptr undef, align 4
55; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %zext.nxv4i8to32 = zext <vscale x 4 x i8> %load.nxv4i8 to <vscale x 4 x i32>
56; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv2i8 = load <vscale x 2 x i8>, ptr undef, align 2
57; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %zext.nxv2i8to64 = zext <vscale x 2 x i8> %load.nxv2i8 to <vscale x 2 x i64>
58; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv8i16 = load <vscale x 8 x i16>, ptr undef, align 16
59; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %zext.nxv8i16to32 = zext <vscale x 8 x i16> %load.nxv8i16 to <vscale x 8 x i32>
60; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv4i16 = load <vscale x 4 x i16>, ptr undef, align 8
61; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %zext.nxv4i16to32 = zext <vscale x 4 x i16> %load.nxv4i16 to <vscale x 4 x i32>
62; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv2i16 = load <vscale x 2 x i16>, ptr undef, align 4
63; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %zext.nxv2i16to64 = zext <vscale x 2 x i16> %load.nxv2i16 to <vscale x 2 x i64>
64; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv4i32 = load <vscale x 4 x i32>, ptr undef, align 16
65; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %zext.nxv4i32to64 = zext <vscale x 4 x i32> %load.nxv4i32 to <vscale x 4 x i64>
66; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv2i32 = load <vscale x 2 x i32>, ptr undef, align 8
67; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %zext.nxv2i32to64 = zext <vscale x 2 x i32> %load.nxv2i32 to <vscale x 2 x i64>
68; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv4i8.2 = load <vscale x 4 x i8>, ptr undef, align 4
69; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zext.nxv4i8to64 = zext <vscale x 4 x i8> %load.nxv4i8.2 to <vscale x 4 x i64>
70; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv4i16.2 = load <vscale x 4 x i16>, ptr undef, align 8
71; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zext.nxv4i16to64 = zext <vscale x 4 x i16> %load.nxv4i16.2 to <vscale x 4 x i64>
72; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv8i8.2 = load <vscale x 8 x i8>, ptr undef, align 8
73; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zext.nxv8i8to32 = zext <vscale x 8 x i8> %load.nxv8i8.2 to <vscale x 8 x i32>
74; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load.nxv8i8.3 = load <vscale x 8 x i8>, ptr undef, align 8
75; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %zext.nxv8i8to64 = zext <vscale x 8 x i8> %load.nxv8i8.3 to <vscale x 8 x i64>
76; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv16i8 = load <vscale x 16 x i8>, ptr undef, align 16
77; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sext.nxv16i8to16 = sext <vscale x 16 x i8> %load2.nxv16i8 to <vscale x 16 x i16>
78; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv8i8 = load <vscale x 8 x i8>, ptr undef, align 8
79; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %sext.nxv8i8to16 = sext <vscale x 8 x i8> %load2.nxv8i8 to <vscale x 8 x i16>
80; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv4i8 = load <vscale x 4 x i8>, ptr undef, align 4
81; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %sext.nxv4i8to32 = sext <vscale x 4 x i8> %load2.nxv4i8 to <vscale x 4 x i32>
82; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv2i8 = load <vscale x 2 x i8>, ptr undef, align 2
83; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %sext.nxv2i8to64 = sext <vscale x 2 x i8> %load2.nxv2i8 to <vscale x 2 x i64>
84; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv8i16 = load <vscale x 8 x i16>, ptr undef, align 16
85; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sext.nxv8i16to32 = sext <vscale x 8 x i16> %load2.nxv8i16 to <vscale x 8 x i32>
86; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv4i16 = load <vscale x 4 x i16>, ptr undef, align 8
87; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %sext.nxv4i16to32 = sext <vscale x 4 x i16> %load2.nxv4i16 to <vscale x 4 x i32>
88; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv2i16 = load <vscale x 2 x i16>, ptr undef, align 4
89; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %sext.nxv2i16to64 = sext <vscale x 2 x i16> %load2.nxv2i16 to <vscale x 2 x i64>
90; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv4i32 = load <vscale x 4 x i32>, ptr undef, align 16
91; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sext.nxv4i32to64 = sext <vscale x 4 x i32> %load2.nxv4i32 to <vscale x 4 x i64>
92; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv2i32 = load <vscale x 2 x i32>, ptr undef, align 8
93; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %sext.nxv2i32to64 = sext <vscale x 2 x i32> %load2.nxv2i32 to <vscale x 2 x i64>
94; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv4i8.2 = load <vscale x 4 x i8>, ptr undef, align 4
95; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sext.nxv4i8to64 = sext <vscale x 4 x i8> %load2.nxv4i8.2 to <vscale x 4 x i64>
96; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv4i16.2 = load <vscale x 4 x i16>, ptr undef, align 8
97; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sext.nxv4i16to64 = sext <vscale x 4 x i16> %load2.nxv4i16.2 to <vscale x 4 x i64>
98; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv8i8.2 = load <vscale x 8 x i8>, ptr undef, align 8
99; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sext.nxv8i8to32 = sext <vscale x 8 x i8> %load2.nxv8i8.2 to <vscale x 8 x i32>
100; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load2.nxv8i8.3 = load <vscale x 8 x i8>, ptr undef, align 8
101; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %sext.nxv8i8to64 = sext <vscale x 8 x i8> %load2.nxv8i8.3 to <vscale x 8 x i64>
102; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
103;
104
105
106  %load.nxv16i8 = load <vscale x 16 x i8>, ptr undef
107  %zext.nxv16i8to16 = zext <vscale x 16 x i8> %load.nxv16i8 to <vscale x 16 x i16>
108  %load.nxv8i8 = load <vscale x 8 x i8>, ptr undef
109  %zext.nxv8i8to16 = zext <vscale x 8 x i8> %load.nxv8i8 to <vscale x 8 x i16>
110  %load.nxv4i8 = load <vscale x 4 x i8>, ptr undef
111  %zext.nxv4i8to32 = zext <vscale x 4 x i8> %load.nxv4i8 to <vscale x 4 x i32>
112  %load.nxv2i8 = load <vscale x 2 x i8>, ptr undef
113  %zext.nxv2i8to64 = zext <vscale x 2 x i8> %load.nxv2i8 to <vscale x 2 x i64>
114  %load.nxv8i16 = load <vscale x 8 x i16>, ptr undef
115  %zext.nxv8i16to32 = zext <vscale x 8 x i16> %load.nxv8i16 to <vscale x 8 x i32>
116  %load.nxv4i16 = load <vscale x 4 x i16>, ptr undef
117  %zext.nxv4i16to32 = zext <vscale x 4 x i16> %load.nxv4i16 to <vscale x 4 x i32>
118  %load.nxv2i16 = load <vscale x 2 x i16>, ptr undef
119  %zext.nxv2i16to64 = zext <vscale x 2 x i16> %load.nxv2i16 to <vscale x 2 x i64>
120  %load.nxv4i32 = load <vscale x 4 x i32>, ptr undef
121  %zext.nxv4i32to64 = zext <vscale x 4 x i32> %load.nxv4i32 to <vscale x 4 x i64>
122  %load.nxv2i32 = load <vscale x 2 x i32>, ptr undef
123  %zext.nxv2i32to64 = zext <vscale x 2 x i32> %load.nxv2i32 to <vscale x 2 x i64>
124  %load.nxv4i8.2 = load <vscale x 4 x i8>, ptr undef
125  %zext.nxv4i8to64 = zext <vscale x 4 x i8> %load.nxv4i8.2 to <vscale x 4 x i64>
126  %load.nxv4i16.2 = load <vscale x 4 x i16>, ptr undef
127  %zext.nxv4i16to64 = zext <vscale x 4 x i16> %load.nxv4i16.2 to <vscale x 4 x i64>
128  %load.nxv8i8.2 = load <vscale x 8 x i8>, ptr undef
129  %zext.nxv8i8to32 = zext <vscale x 8 x i8> %load.nxv8i8.2 to <vscale x 8 x i32>
130  %load.nxv8i8.3 = load <vscale x 8 x i8>, ptr undef
131  %zext.nxv8i8to64 = zext <vscale x 8 x i8> %load.nxv8i8.3 to <vscale x 8 x i64>
132
133  %load2.nxv16i8 = load <vscale x 16 x i8>, ptr undef
134  %sext.nxv16i8to16 = sext <vscale x 16 x i8> %load2.nxv16i8 to <vscale x 16 x i16>
135  %load2.nxv8i8 = load <vscale x 8 x i8>, ptr undef
136  %sext.nxv8i8to16 = sext <vscale x 8 x i8> %load2.nxv8i8 to <vscale x 8 x i16>
137  %load2.nxv4i8 = load <vscale x 4 x i8>, ptr undef
138  %sext.nxv4i8to32 = sext <vscale x 4 x i8> %load2.nxv4i8 to <vscale x 4 x i32>
139  %load2.nxv2i8 = load <vscale x 2 x i8>, ptr undef
140  %sext.nxv2i8to64 = sext <vscale x 2 x i8> %load2.nxv2i8 to <vscale x 2 x i64>
141  %load2.nxv8i16 = load <vscale x 8 x i16>, ptr undef
142  %sext.nxv8i16to32 = sext <vscale x 8 x i16> %load2.nxv8i16 to <vscale x 8 x i32>
143  %load2.nxv4i16 = load <vscale x 4 x i16>, ptr undef
144  %sext.nxv4i16to32 = sext <vscale x 4 x i16> %load2.nxv4i16 to <vscale x 4 x i32>
145  %load2.nxv2i16 = load <vscale x 2 x i16>, ptr undef
146  %sext.nxv2i16to64 = sext <vscale x 2 x i16> %load2.nxv2i16 to <vscale x 2 x i64>
147  %load2.nxv4i32 = load <vscale x 4 x i32>, ptr undef
148  %sext.nxv4i32to64 = sext <vscale x 4 x i32> %load2.nxv4i32 to <vscale x 4 x i64>
149  %load2.nxv2i32 = load <vscale x 2 x i32>, ptr undef
150  %sext.nxv2i32to64 = sext <vscale x 2 x i32> %load2.nxv2i32 to <vscale x 2 x i64>
151  %load2.nxv4i8.2 = load <vscale x 4 x i8>, ptr undef
152  %sext.nxv4i8to64 = sext <vscale x 4 x i8> %load2.nxv4i8.2 to <vscale x 4 x i64>
153  %load2.nxv4i16.2 = load <vscale x 4 x i16>, ptr undef
154  %sext.nxv4i16to64 = sext <vscale x 4 x i16> %load2.nxv4i16.2 to <vscale x 4 x i64>
155  %load2.nxv8i8.2 = load <vscale x 8 x i8>, ptr undef
156  %sext.nxv8i8to32 = sext <vscale x 8 x i8> %load2.nxv8i8.2 to <vscale x 8 x i32>
157  %load2.nxv8i8.3 = load <vscale x 8 x i8>, ptr undef
158  %sext.nxv8i8to64 = sext <vscale x 8 x i8> %load2.nxv8i8.3 to <vscale x 8 x i64>
159
160  ret void
161}
162
163