1 /**************************************************************************
3 * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
29 #include "sp_context.h"
32 #include "sp_headers.h"
35 #include "pipe/p_state.h"
36 #include "pipe/p_defines.h"
37 #include "pipe/p_util.h"
38 #include "pipe/p_inlines.h"
39 #include "tgsi/exec/tgsi_exec.h"
40 #include "tgsi/exec/tgsi_sse2.h"
43 #if defined(__i386__) || defined(__386__)
45 #include "x86/rtasm/x86sse.h"
47 /* Surely this should be defined somewhere in a tgsi header:
49 typedef void (XSTDCALL
*codegen_function
)(
50 const struct tgsi_exec_vector
*input
,
51 struct tgsi_exec_vector
*output
,
53 struct tgsi_exec_vector
*temporary
,
54 const struct tgsi_interp_coef
*coef
55 //, const struct tgsi_exec_vector *quadPos
59 struct sp_sse_fragment_shader
{
60 struct sp_fragment_shader base
;
61 struct x86_function sse2_program
;
62 codegen_function func
;
67 * Compute quad X,Y,Z,W for the four fragments in a quad.
69 * This should really be part of the compiled shader.
72 sp_setup_pos_vector(const struct tgsi_interp_coef
*coef
,
74 struct tgsi_exec_vector
*quadpos
)
78 quadpos
->xyzw
[0].f
[0] = x
;
79 quadpos
->xyzw
[0].f
[1] = x
+ 1;
80 quadpos
->xyzw
[0].f
[2] = x
;
81 quadpos
->xyzw
[0].f
[3] = x
+ 1;
84 quadpos
->xyzw
[1].f
[0] = y
;
85 quadpos
->xyzw
[1].f
[1] = y
;
86 quadpos
->xyzw
[1].f
[2] = y
+ 1;
87 quadpos
->xyzw
[1].f
[3] = y
+ 1;
89 /* do Z and W for all fragments in the quad */
90 for (chan
= 2; chan
< 4; chan
++) {
91 const float dadx
= coef
->dadx
[chan
];
92 const float dady
= coef
->dady
[chan
];
93 const float a0
= coef
->a0
[chan
] + dadx
* x
+ dady
* y
;
94 quadpos
->xyzw
[chan
].f
[0] = a0
;
95 quadpos
->xyzw
[chan
].f
[1] = a0
+ dadx
;
96 quadpos
->xyzw
[chan
].f
[2] = a0
+ dady
;
97 quadpos
->xyzw
[chan
].f
[3] = a0
+ dadx
+ dady
;
103 fs_sse_prepare( struct sp_fragment_shader
*base
,
104 struct tgsi_exec_machine
*machine
,
105 struct tgsi_sampler
*samplers
)
110 /* TODO: codegenerate the whole run function, skip this wrapper.
111 * TODO: break dependency on tgsi_exec_machine struct
112 * TODO: push Position calculation into the generated shader
113 * TODO: process >1 quad at a time
116 fs_sse_run( struct sp_fragment_shader
*base
,
117 struct tgsi_exec_machine
*machine
,
118 struct quad_header
*quad
)
120 struct sp_sse_fragment_shader
*shader
= (struct sp_sse_fragment_shader
*) base
;
122 /* Compute X, Y, Z, W vals for this quad -- place in temp[0] for now */
123 sp_setup_pos_vector(quad
->posCoef
,
124 (float)quad
->x0
, (float)quad
->y0
,
127 shader
->func( machine
->Inputs
,
132 // , &machine->QuadPos
135 return ~(machine
->Temps
[TGSI_EXEC_TEMP_KILMASK_I
].xyzw
[TGSI_EXEC_TEMP_KILMASK_C
].u
[0]);
140 fs_sse_delete( struct sp_fragment_shader
*base
)
142 struct sp_sse_fragment_shader
*shader
= (struct sp_sse_fragment_shader
*) base
;
144 x86_release_func( &shader
->sse2_program
);
149 struct sp_fragment_shader
*
150 softpipe_create_fs_sse(struct softpipe_context
*softpipe
,
151 const struct pipe_shader_state
*templ
)
153 struct sp_sse_fragment_shader
*shader
;
155 if (!softpipe
->use_sse
)
158 shader
= CALLOC_STRUCT(sp_sse_fragment_shader
);
162 x86_init_func( &shader
->sse2_program
);
164 if (!tgsi_emit_sse2_fs( templ
->tokens
, &shader
->sse2_program
)) {
169 shader
->func
= (codegen_function
) x86_get_func( &shader
->sse2_program
);
170 assert(shader
->func
);
172 shader
->base
.shader
= *templ
;
173 shader
->base
.prepare
= fs_sse_prepare
;
174 shader
->base
.run
= fs_sse_run
;
175 shader
->base
.delete = fs_sse_delete
;
177 return &shader
->base
;
183 /* Maybe put this varient in the header file.
185 struct sp_fragment_shader
*
186 softpipe_create_fs_sse(struct softpipe_context
*softpipe
,
187 const struct pipe_shader_state
*templ
)