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>
34 #include "util/u_math.h"
35 #include "util/u_memory.h"
36 #include "pipe/p_config.h"
40 #if defined(PIPE_ARCH_X86) && defined(PIPE_ARCH_SSE)
42 #include "pipe/p_shader_tokens.h"
44 #include "draw_private.h"
45 #include "draw_context.h"
47 #include "rtasm/rtasm_cpu.h"
48 #include "rtasm/rtasm_x86sse.h"
49 #include "tgsi/tgsi_sse2.h"
50 #include "tgsi/tgsi_parse.h"
52 #define SSE_MAX_VERTICES 4
54 typedef void (PIPE_CDECL
*codegen_function
) (
55 const struct tgsi_exec_vector
*input
, /* 1 */
56 struct tgsi_exec_vector
*output
, /* 2 */
57 float (*constant
)[4], /* 3 */
58 struct tgsi_exec_vector
*temporary
, /* 4 */
59 float (*immediates
)[4], /* 5 */
60 const float (*aos_input
)[4], /* 6 */
61 uint num_inputs
, /* 7 */
62 uint input_stride
, /* 8 */
63 float (*aos_output
)[4], /* 9 */
64 uint num_outputs
, /* 10 */
65 uint output_stride
); /* 11 */
67 struct draw_sse_vertex_shader
{
68 struct draw_vertex_shader base
;
69 struct x86_function sse2_program
;
71 codegen_function func
;
73 struct tgsi_exec_machine
*machine
;
78 vs_sse_prepare( struct draw_vertex_shader
*base
,
79 struct draw_context
*draw
)
85 /* Simplified vertex shader interface for the pt paths. Given the
86 * complexity of code-generating all the above operations together,
87 * it's time to try doing all the other stuff separately.
90 vs_sse_run_linear( struct draw_vertex_shader
*base
,
91 const float (*input
)[4],
93 const float (*constants
)[4],
95 unsigned input_stride
,
96 unsigned output_stride
)
98 struct draw_sse_vertex_shader
*shader
= (struct draw_sse_vertex_shader
*)base
;
99 struct tgsi_exec_machine
*machine
= shader
->machine
;
102 for (i
= 0; i
< count
; i
+= MAX_TGSI_VERTICES
) {
103 unsigned int max_vertices
= MIN2(MAX_TGSI_VERTICES
, count
- i
);
105 /* run compiled shader
107 shader
->func(machine
->Inputs
,
109 (float (*)[4])constants
,
111 (float (*)[4])shader
->base
.immediates
,
113 base
->info
.num_inputs
,
116 base
->info
.num_outputs
,
119 input
= (const float (*)[4])((const char *)input
+ input_stride
* max_vertices
);
120 output
= (float (*)[4])((char *)output
+ output_stride
* max_vertices
);
128 vs_sse_delete( struct draw_vertex_shader
*base
)
130 struct draw_sse_vertex_shader
*shader
= (struct draw_sse_vertex_shader
*)base
;
132 x86_release_func( &shader
->sse2_program
);
134 align_free( (void *) shader
->base
.immediates
);
136 FREE( (void*) shader
->base
.state
.tokens
);
141 struct draw_vertex_shader
*
142 draw_create_vs_sse(struct draw_context
*draw
,
143 const struct pipe_shader_state
*templ
)
145 struct draw_sse_vertex_shader
*vs
;
147 if (!rtasm_cpu_has_sse2())
150 vs
= CALLOC_STRUCT( draw_sse_vertex_shader
);
154 /* we make a private copy of the tokens */
155 vs
->base
.state
.tokens
= tgsi_dup_tokens(templ
->tokens
);
156 if (!vs
->base
.state
.tokens
)
159 tgsi_scan_shader(templ
->tokens
, &vs
->base
.info
);
161 vs
->base
.draw
= draw
;
163 vs
->base
.create_varient
= draw_vs_varient_aos_sse
;
165 vs
->base
.create_varient
= draw_vs_varient_generic
;
166 vs
->base
.prepare
= vs_sse_prepare
;
167 vs
->base
.run_linear
= vs_sse_run_linear
;
168 vs
->base
.delete = vs_sse_delete
;
170 vs
->base
.immediates
= align_malloc(TGSI_EXEC_NUM_IMMEDIATES
* 4 *
173 vs
->machine
= &draw
->vs
.machine
;
175 x86_init_func( &vs
->sse2_program
);
177 if (!tgsi_emit_sse2( (struct tgsi_token
*) vs
->base
.state
.tokens
,
179 (float (*)[4])vs
->base
.immediates
,
183 vs
->func
= (codegen_function
) x86_get_func( &vs
->sse2_program
);
191 debug_error("tgsi_emit_sse2() failed, falling back to interpreter\n");
193 x86_release_func( &vs
->sse2_program
);
203 struct draw_vertex_shader
*
204 draw_create_vs_sse( struct draw_context
*draw
,
205 const struct pipe_shader_state
*templ
)