1 /*---------------------------------------------------------------------------*
2   Project:  Horizon
3   File:     math_Matrix33.cpp
4 
5   Copyright (C)2009-2012 Nintendo Co., Ltd.  All rights reserved.
6 
7   These coded instructions, statements, and computer programs contain
8   proprietary information of Nintendo of America Inc. and/or Nintendo
9   Company Ltd., and are protected by Federal copyright law.  They may
10   not be disclosed to third parties or copied or duplicated in any form,
11   in whole or in part, without the prior written consent of Nintendo.
12 
13   $Rev: 46347 $
14  *---------------------------------------------------------------------------*/
15 
16 #include <nn/math.h>
17 
18 #include <cmath>
19 #include <nn/math/ARMv6/math_Matrix33.h>
20 
21 #if !defined(NN_MATH_AS_INLINE)
22 #include <nn/math/ARMv6/inline/math_Matrix33.ipp>
23 #endif
24 
25 namespace nn {
26 namespace math {
27 namespace ARMv6 {
28 
29 #include <nn/hw/ARM/code32.h>
30 
31 /* Please see man pages for details
32 
33 
34 */
35 
36 NN_FUNC_ATTR_PRIVATE_SECTION
MTX33CopyAsm(MTX33 *,const MTX33 *)37 asm MTX33* MTX33CopyAsm(MTX33*, const MTX33*)
38 {
39     VLDMIA      r1!,{s0-s4}         // Segment and load to shorten stall times due to data hazards
40     MOV         r2, r0              //
41     VLDMIA      r1,{s5-s8}
42     VSTMIA      r2!,{s0-s4}         // Store result
43     VSTMIA      r2,{s5-s8}          // Store result
44     BX          lr                  // Return
45 }
46 
47 NN_FUNC_ATTR_PRIVATE_SECTION
MTX33MAddAsm(MTX33 *,f32,const MTX33 *,const MTX33 *)48 asm MTX33* MTX33MAddAsm(MTX33* , f32, const MTX33*, const MTX33*)
49 {
50     VLDMIA      r2,{s1-s9}          // The entire p2 matrix is put in the [S4-S15] registers
51     VLDMIA      r1!,{s10-s15}       // Matrix p1 is put into the [S16-S23] registers
52 
53     VMLA.F32    s1,s10,s0
54     VMLA.F32    s2,s11,s0
55     VMLA.F32    s3,s12,s0
56 
57     VLDMIA      r1,{s10-s12}        // Continuation of p1
58 
59     VMLA.F32    s4,s13,s0
60     VMLA.F32    s5,s14,s0
61     VMLA.F32    s6,s15,s0
62 
63     VMLA.F32    s7,s10,s0
64     VMLA.F32    s8,s11,s0
65     VMLA.F32    s9,s12,s0
66 
67     VSTMIA      r0,{s1-s9}          // Store result
68     BX          lr                  // Return
69 
70 }
71 
72 NN_FUNC_ATTR_PRIVATE_SECTION
VEC3TransformAsm(VEC3 *,const MTX33 *,const VEC3 *)73 asm VEC3* VEC3TransformAsm(VEC3*, const MTX33*, const VEC3*)
74 {
75     VLDMIA      r1,{s0-s8}          // First line of matrix pM to register [S0-S2]
76     VLDMIA      r2,{s9-s11}         // All vectors are put in the [S10-S12] registers
77 
78     VMUL.F32    s12,s0,s9
79     VMUL.F32    s13,s3,s9
80     VMUL.F32    s14,s6,s9
81 
82     VMLA.F32    s12,s1,s10
83     VMLA.F32    s13,s4,s10
84     VMLA.F32    s14,s7,s10
85 
86     VMLA.F32    s12,s2,s11
87     VMLA.F32    s13,s5,s11
88     VMLA.F32    s14,s8,s11
89 
90     VSTMIA      r0,{s12-s14}        // Store result
91 
92     BX          lr                  // Return
93 
94 }
95 
96 #include <nn/hw/ARM/codereset.h>
97 
98 }  // namespace ARMv6
99 }  // namespace math
100 }  // namespace nn
101