1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
|
SME FMOPA outer product instruction gives incorrect result
Description of problem:
The SME outer product instructions operate on tiles of elements. In the
below example we are performing an outer product of a vector of 1.0
by itself. This naturally should produce a matrix filled with 1.0
values, however if we read the values of the tile and printf them we
instead observe 0.0 values.
Without digging into the underlying QEMU code this appears to be a bug
in how elements are set based on the tile number, since the same code
using za0.s rather than za1.s correctly reports all 1.0 values as output
as expected.
main.c
```
#include <stdio.h>
void foo(float *dst);
int main() {
float dst[16];
foo(dst);
// This should print:
// >>> 1.000000 1.000000 1.000000 1.000000
// >>> 1.000000 1.000000 1.000000 1.000000
// >>> 1.000000 1.000000 1.000000 1.000000
// >>> 1.000000 1.000000 1.000000 1.000000
for (int i=0; i<4; ++i) {
printf(">>> ");
for (int j=0; j<4; ++j) {
printf("%lf ", (double)dst[i * 4 + j]);
}
printf("\n");
}
}
```
foo.S
```
.global foo
foo:
stp x29, x30, [sp, -80]!
mov x29, sp
stp d8, d9, [sp, 16]
stp d10, d11, [sp, 32]
stp d12, d13, [sp, 48]
stp d14, d15, [sp, 64]
smstart
ptrue p0.s, vl4
fmov z0.s, #1.0
// An outer product of a vector of 1.0 by itself should be a matrix of 1.0.
// Note that we are using tile 1 here (za1.s) rather than tile 0.
zero {za}
fmopa za1.s, p0/m, p0/m, z0.s, z0.s
// Read the first 4x4 sub-matrix of elements from tile 1:
// Note that za1h should be interchangable here.
mov w12, #0
mova z0.s, p0/m, za1v.s[w12, #0]
mova z1.s, p0/m, za1v.s[w12, #1]
mova z2.s, p0/m, za1v.s[w12, #2]
mova z3.s, p0/m, za1v.s[w12, #3]
// And store them to the input pointer (dst in the C code):
st1w {z0.s}, p0, [x0]
add x0, x0, #16
st1w {z1.s}, p0, [x0]
add x0, x0, #16
st1w {z2.s}, p0, [x0]
add x0, x0, #16
st1w {z3.s}, p0, [x0]
smstop
ldp d8, d9, [sp, 16]
ldp d10, d11, [sp, 32]
ldp d12, d13, [sp, 48]
ldp d14, d15, [sp, 64]
ldp x29, x30, [sp], 80
ret
```
Steps to reproduce:
```
$ clang -target aarch64-linux-gnu -march=armv9-a+sme test.c -O1 -static
$ ~/qemu/build/qemu-aarch64 ./a.out
>>> 0.000000 0.000000 0.000000 0.000000
>>> 0.000000 0.000000 0.000000 0.000000
>>> 0.000000 0.000000 0.000000 0.000000
>>> 0.000000 0.000000 0.000000 0.000000
```
|