1/*
2 * Mesa 3-D graphics library
3 * Version:  6.0
4 *
5 * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
6 *
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the "Software"),
9 * to deal in the Software without restriction, including without limitation
10 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
11 * and/or sell copies of the Software, and to permit persons to whom the
12 * Software is furnished to do so, subject to the following conditions:
13 *
14 * The above copyright notice and this permission notice shall be included
15 * in all copies or substantial portions of the Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
18 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
20 * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
21 * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
22 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
23 */
24
25/*
26 * PentiumIII-SIMD (SSE) optimizations contributed by
27 * Andre Werthmann <wertmann@cs.uni-potsdam.de>
28 */
29
30#include "main/glheader.h"
31#include "main/context.h"
32#include "math/m_xform.h"
33#include "tnl/t_context.h"
34
35#include "sse.h"
36#include "x86_xform.h"
37
38#ifdef DEBUG_MATH
39#include "math/m_debug.h"
40#endif
41
42
43#ifdef USE_SSE_ASM
44DECLARE_XFORM_GROUP( sse, 2 )
45DECLARE_XFORM_GROUP( sse, 3 )
46
47#if 1
48/* Some functions are not written in SSE-assembly, because the fpu ones are faster */
49extern void _ASMAPI _mesa_sse_transform_normals_no_rot( NORM_ARGS );
50extern void _ASMAPI _mesa_sse_transform_rescale_normals( NORM_ARGS );
51extern void _ASMAPI _mesa_sse_transform_rescale_normals_no_rot( NORM_ARGS );
52
53extern void _ASMAPI _mesa_sse_transform_points4_general( XFORM_ARGS );
54extern void _ASMAPI _mesa_sse_transform_points4_3d( XFORM_ARGS );
55/* XXX this function segfaults, see below */
56extern void _ASMAPI _mesa_sse_transform_points4_identity( XFORM_ARGS );
57/* XXX this one works, see below */
58extern void _ASMAPI _mesa_x86_transform_points4_identity( XFORM_ARGS );
59#else
60DECLARE_NORM_GROUP( sse )
61#endif
62
63
64extern void _ASMAPI
65_mesa_v16_sse_general_xform( GLfloat *first_vert,
66			     const GLfloat *m,
67			     const GLfloat *src,
68			     GLuint src_stride,
69			     GLuint count );
70
71extern void _ASMAPI
72_mesa_sse_project_vertices( GLfloat *first,
73			    GLfloat *last,
74			    const GLfloat *m,
75			    GLuint stride );
76
77extern void _ASMAPI
78_mesa_sse_project_clipped_vertices( GLfloat *first,
79				    GLfloat *last,
80				    const GLfloat *m,
81				    GLuint stride,
82				    const GLubyte *clipmask );
83#endif
84
85
86void _mesa_init_sse_transform_asm( void )
87{
88#ifdef USE_SSE_ASM
89   ASSIGN_XFORM_GROUP( sse, 2 );
90   ASSIGN_XFORM_GROUP( sse, 3 );
91
92#if 1
93   /* TODO: Finish these off.
94    */
95   _mesa_transform_tab[4][MATRIX_GENERAL] =
96      _mesa_sse_transform_points4_general;
97   _mesa_transform_tab[4][MATRIX_3D] =
98      _mesa_sse_transform_points4_3d;
99   /* XXX NOTE: _mesa_sse_transform_points4_identity segfaults with the
100      conformance tests, so use the x86 version.
101   */
102   _mesa_transform_tab[4][MATRIX_IDENTITY] =
103      _mesa_x86_transform_points4_identity;/*_mesa_sse_transform_points4_identity;*/
104
105   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT] =
106      _mesa_sse_transform_normals_no_rot;
107   _mesa_normal_tab[NORM_TRANSFORM | NORM_RESCALE] =
108      _mesa_sse_transform_rescale_normals;
109   _mesa_normal_tab[NORM_TRANSFORM_NO_ROT | NORM_RESCALE] =
110      _mesa_sse_transform_rescale_normals_no_rot;
111#else
112   ASSIGN_XFORM_GROUP( sse, 4 );
113
114   ASSIGN_NORM_GROUP( sse );
115#endif
116
117#ifdef DEBUG_MATH
118   _math_test_all_transform_functions( "SSE" );
119   _math_test_all_normal_transform_functions( "SSE" );
120#endif
121#endif
122}
123
124