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 **************************************************************************/
30 * Keith Whitwell <keith@tungstengraphics.com>
36 #if defined(__i386__) || defined(__386__)
38 #include "pipe/p_util.h"
39 #include "pipe/p_shader_tokens.h"
41 #include "draw_private.h"
42 #include "draw_context.h"
44 #include "rtasm/rtasm_cpu.h"
45 #include "rtasm/rtasm_x86sse.h"
46 #include "tgsi/exec/tgsi_sse2.h"
47 #include "tgsi/util/tgsi_parse.h"
49 #define SSE_MAX_VERTICES 4
51 typedef void (XSTDCALL
*codegen_function
) (
52 const struct tgsi_exec_vector
*input
, /* 1 */
53 struct tgsi_exec_vector
*output
, /* 2 */
54 float (*constant
)[4], /* 3 */
55 struct tgsi_exec_vector
*temporary
, /* 4 */
56 float (*immediates
)[4], /* 5 */
57 const float (*aos_input
)[4], /* 6 */
58 uint num_inputs
, /* 7 */
59 uint input_stride
, /* 8 */
60 float (*aos_output
)[4], /* 9 */
61 uint num_outputs
, /* 10 */
62 uint output_stride
); /* 11 */
64 struct draw_sse_vertex_shader
{
65 struct draw_vertex_shader base
;
66 struct x86_function sse2_program
;
68 codegen_function func
;
70 struct tgsi_exec_machine
*machine
;
75 vs_sse_prepare( struct draw_vertex_shader
*base
,
76 struct draw_context
*draw
)
82 /* Simplified vertex shader interface for the pt paths. Given the
83 * complexity of code-generating all the above operations together,
84 * it's time to try doing all the other stuff separately.
87 vs_sse_run_linear( struct draw_vertex_shader
*base
,
88 const float (*input
)[4],
90 const float (*constants
)[4],
92 unsigned input_stride
,
93 unsigned output_stride
)
95 struct draw_sse_vertex_shader
*shader
= (struct draw_sse_vertex_shader
*)base
;
96 struct tgsi_exec_machine
*machine
= shader
->machine
;
99 for (i
= 0; i
< count
; i
+= MAX_TGSI_VERTICES
) {
100 unsigned int max_vertices
= MIN2(MAX_TGSI_VERTICES
, count
- i
);
102 /* run compiled shader
104 shader
->func(machine
->Inputs
,
106 (float (*)[4])constants
,
108 (float (*)[4])shader
->base
.immediates
,
110 base
->info
.num_inputs
,
113 base
->info
.num_outputs
,
116 input
= (const float (*)[4])((const char *)input
+ input_stride
* max_vertices
);
117 output
= (float (*)[4])((char *)output
+ output_stride
* max_vertices
);
125 vs_sse_delete( struct draw_vertex_shader
*base
)
127 struct draw_sse_vertex_shader
*shader
= (struct draw_sse_vertex_shader
*)base
;
129 x86_release_func( &shader
->sse2_program
);
131 align_free(shader
->base
.immediates
);
133 FREE( (void*) shader
->base
.state
.tokens
);
138 struct draw_vertex_shader
*
139 draw_create_vs_sse(struct draw_context
*draw
,
140 const struct pipe_shader_state
*templ
)
142 struct draw_sse_vertex_shader
*vs
;
144 if (!rtasm_cpu_has_sse2())
147 vs
= CALLOC_STRUCT( draw_sse_vertex_shader
);
151 /* we make a private copy of the tokens */
152 vs
->base
.state
.tokens
= tgsi_dup_tokens(templ
->tokens
);
153 if (!vs
->base
.state
.tokens
)
156 tgsi_scan_shader(templ
->tokens
, &vs
->base
.info
);
158 vs
->base
.draw
= draw
;
159 vs
->base
.create_varient
= draw_vs_varient_aos_sse
;
160 // vs->base.create_varient = draw_vs_varient_generic;
161 vs
->base
.prepare
= vs_sse_prepare
;
162 vs
->base
.run_linear
= vs_sse_run_linear
;
163 vs
->base
.delete = vs_sse_delete
;
165 vs
->base
.immediates
= align_malloc(TGSI_EXEC_NUM_IMMEDIATES
* 4 *
168 vs
->machine
= &draw
->vs
.machine
;
170 x86_init_func( &vs
->sse2_program
);
172 if (!tgsi_emit_sse2( (struct tgsi_token
*) vs
->base
.state
.tokens
,
174 (float (*)[4])vs
->base
.immediates
,
178 vs
->func
= (codegen_function
) x86_get_func( &vs
->sse2_program
);
186 fprintf(stderr
, "tgsi_emit_sse2() failed, falling back to interpreter\n");
188 x86_release_func( &vs
->sse2_program
);
198 struct draw_vertex_shader
*
199 draw_create_vs_sse( struct draw_context
*draw
,
200 const struct pipe_shader_state
*templ
)