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
|
{ %norun }
{ %cpu=i386,x86_64 }
{ %opt=-Sew -vw }
{$mode objfpc}
{ The test checks that MOVSS instruction assembles without warning.
Running it could be a nice bonus, but it turns out that we have no portable
way to detect SSE4.1 support (for DPPS), so disabled for now. }
uses cpu;
{$asmmode att}
procedure test1; assembler;
var
s: single;
asm
movss s, %xmm6
movss %xmm6, s
{$ifdef cpui386}
movss (%eax, %edx), %xmm7
movss %xmm7, (%eax, %edx)
{$endif}
{$ifdef cpux86_64}
movss (%rax, %rdx), %xmm7
movss %xmm7, (%rax, %rdx)
{$endif}
end;
{$asmmode intel}
procedure test2; assembler;
var
s: single;
asm
movss [s], xmm6
movss xmm6, [s]
{$ifdef cpui386}
movss [eax+edx], xmm7
movss xmm7, [eax+edx]
{$endif}
{$ifdef cpux86_64}
movss [rax+rdx], xmm7
movss xmm7, [rax+rdx]
{$endif}
end;
type
TVector4 = packed record
X, Y, Z, W: Single;
end;
function _VectorDotProductSSE4(Vector1, Vector2: TVector4): Single; assembler;
asm
MOVUPS XMM0, [Vector1]
MOVUPS XMM1, [Vector2]
DPPS XMM0, XMM1, $71 { Only perform calculations on the X, Y and Z coordinates; only store result in the first element }
MOVSS Result, XMM0 { Store result - first element of XMM0 }
end;
var
v: tvector4;
r: single;
begin
v.x:=1;
v.y:=1;
v.z:=1;
v.w:=1;
r:=_vectordotproductSSE4(v,v);
if r<>3 then
halt(1);
writeln('ok');
end.
|
