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)
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"
51 #include "tgsi/tgsi_exec.h"
53 #define SSE_MAX_VERTICES 4
55 typedef void (PIPE_CDECL
*codegen_function
) (
56 const struct tgsi_exec_vector
*input
, /* 1 */
57 struct tgsi_exec_vector
*output
, /* 2 */
58 float (*constant
)[4], /* 3 */
59 struct tgsi_exec_vector
*temporary
, /* 4 */
60 float (*immediates
)[4], /* 5 */
61 const float (*aos_input
)[4], /* 6 */
62 uint num_inputs
, /* 7 */
63 uint input_stride
, /* 8 */
64 float (*aos_output
)[4], /* 9 */
65 uint num_outputs
, /* 10 */
66 uint output_stride
); /* 11 */
68 struct draw_sse_vertex_shader
{
69 struct draw_vertex_shader base
;
70 struct x86_function sse2_program
;
72 codegen_function func
;
74 struct tgsi_exec_machine
*machine
;
79 vs_sse_prepare( struct draw_vertex_shader
*base
,
80 struct draw_context
*draw
)
86 /* Simplified vertex shader interface for the pt paths. Given the
87 * complexity of code-generating all the above operations together,
88 * it's time to try doing all the other stuff separately.
91 vs_sse_run_linear( struct draw_vertex_shader
*base
,
92 const float (*input
)[4],
94 const float (*constants
)[4],
96 unsigned input_stride
,
97 unsigned output_stride
)
99 struct draw_sse_vertex_shader
*shader
= (struct draw_sse_vertex_shader
*)base
;
100 struct tgsi_exec_machine
*machine
= shader
->machine
;
103 /* By default, execute all channels. XXX move this inside the loop
104 * below when we support shader conditionals/loops.
106 tgsi_set_exec_mask(machine
, 1, 1, 1, 1);
108 for (i
= 0; i
< count
; i
+= MAX_TGSI_VERTICES
) {
109 unsigned int max_vertices
= MIN2(MAX_TGSI_VERTICES
, count
- i
);
111 if (max_vertices
< 4) {
112 /* disable the unused execution channels */
113 tgsi_set_exec_mask(machine
,
120 /* run compiled shader
122 shader
->func(machine
->Inputs
,
124 (float (*)[4])constants
,
126 (float (*)[4])shader
->base
.immediates
,
128 base
->info
.num_inputs
,
131 base
->info
.num_outputs
,
134 input
= (const float (*)[4])((const char *)input
+ input_stride
* max_vertices
);
135 output
= (float (*)[4])((char *)output
+ output_stride
* max_vertices
);
143 vs_sse_delete( struct draw_vertex_shader
*base
)
145 struct draw_sse_vertex_shader
*shader
= (struct draw_sse_vertex_shader
*)base
;
147 x86_release_func( &shader
->sse2_program
);
149 align_free( (void *) shader
->base
.immediates
);
151 FREE( (void*) shader
->base
.state
.tokens
);
156 struct draw_vertex_shader
*
157 draw_create_vs_sse(struct draw_context
*draw
,
158 const struct pipe_shader_state
*templ
)
160 struct draw_sse_vertex_shader
*vs
;
162 if (!rtasm_cpu_has_sse2())
165 vs
= CALLOC_STRUCT( draw_sse_vertex_shader
);
169 /* we make a private copy of the tokens */
170 vs
->base
.state
.tokens
= tgsi_dup_tokens(templ
->tokens
);
171 if (!vs
->base
.state
.tokens
)
174 tgsi_scan_shader(templ
->tokens
, &vs
->base
.info
);
176 vs
->base
.draw
= draw
;
178 vs
->base
.create_varient
= draw_vs_varient_aos_sse
;
180 vs
->base
.create_varient
= draw_vs_varient_generic
;
181 vs
->base
.prepare
= vs_sse_prepare
;
182 vs
->base
.run_linear
= vs_sse_run_linear
;
183 vs
->base
.delete = vs_sse_delete
;
185 vs
->base
.immediates
= align_malloc(TGSI_EXEC_NUM_IMMEDIATES
* 4 *
188 vs
->machine
= draw
->vs
.machine
;
190 x86_init_func( &vs
->sse2_program
);
192 if (!tgsi_emit_sse2( (struct tgsi_token
*) vs
->base
.state
.tokens
,
194 (float (*)[4])vs
->base
.immediates
,
198 vs
->func
= (codegen_function
) x86_get_func( &vs
->sse2_program
);
206 debug_error("tgsi_emit_sse2() failed, falling back to interpreter\n");
208 x86_release_func( &vs
->sse2_program
);
218 struct draw_vertex_shader
*
219 draw_create_vs_sse( struct draw_context
*draw
,
220 const struct pipe_shader_state
*templ
)