1 /**************************************************************************
3 * Copyright 2008 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_PPC)
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_ppc.h"
49 #include "tgsi/tgsi_ppc.h"
50 #include "tgsi/tgsi_parse.h"
54 typedef void (PIPE_CDECL
*codegen_function
) (float (*inputs
)[4][4],
55 float (*outputs
)[4][4],
59 const float *builtins
);
62 struct draw_ppc_vertex_shader
{
63 struct draw_vertex_shader base
;
64 struct ppc_function ppc_program
;
66 codegen_function func
;
71 vs_ppc_prepare( struct draw_vertex_shader
*base
,
72 struct draw_context
*draw
)
79 * Simplified vertex shader interface for the pt paths. Given the
80 * complexity of code-generating all the above operations together,
81 * it's time to try doing all the other stuff separately.
84 vs_ppc_run_linear( struct draw_vertex_shader
*base
,
85 const float (*input
)[4],
87 const float (*constants
)[4],
89 unsigned input_stride
,
90 unsigned output_stride
)
92 struct draw_ppc_vertex_shader
*shader
= (struct draw_ppc_vertex_shader
*)base
;
95 #define MAX_VERTICES 4
98 for (i
= 0; i
< count
; i
+= MAX_VERTICES
) {
99 const uint max_vertices
= MIN2(MAX_VERTICES
, count
- i
);
100 float inputs_soa
[PIPE_MAX_SHADER_INPUTS
][4][4] ALIGN16_ATTRIB
;
101 float outputs_soa
[PIPE_MAX_SHADER_OUTPUTS
][4][4] ALIGN16_ATTRIB
;
102 float temps_soa
[TGSI_EXEC_NUM_TEMPS
][4][4] ALIGN16_ATTRIB
;
105 /* convert (up to) four input verts to SoA format */
106 for (attr
= 0; attr
< base
->info
.num_inputs
; attr
++) {
107 const float *vIn
= (const float *) input
;
109 for (vert
= 0; vert
< max_vertices
; vert
++) {
112 printf("Input v%d a%d: %f %f %f %f\n",
113 vert
, attr
, vIn
[0], vIn
[1], vIn
[2], vIn
[3]);
115 inputs_soa
[attr
][0][vert
] = vIn
[attr
* 4 + 0];
116 inputs_soa
[attr
][1][vert
] = vIn
[attr
* 4 + 1];
117 inputs_soa
[attr
][2][vert
] = vIn
[attr
* 4 + 2];
118 inputs_soa
[attr
][3][vert
] = vIn
[attr
* 4 + 3];
119 vIn
+= input_stride
/ 4;
123 /* run compiled shader
125 shader
->func(inputs_soa
, outputs_soa
, temps_soa
,
126 (float (*)[4]) shader
->base
.immediates
,
127 (float (*)[4]) constants
,
128 ppc_builtin_constants
);
130 /* convert (up to) four output verts from SoA back to AoS format */
131 for (attr
= 0; attr
< base
->info
.num_outputs
; attr
++) {
132 float *vOut
= (float *) output
;
134 for (vert
= 0; vert
< max_vertices
; vert
++) {
135 vOut
[attr
* 4 + 0] = outputs_soa
[attr
][0][vert
];
136 vOut
[attr
* 4 + 1] = outputs_soa
[attr
][1][vert
];
137 vOut
[attr
* 4 + 2] = outputs_soa
[attr
][2][vert
];
138 vOut
[attr
* 4 + 3] = outputs_soa
[attr
][3][vert
];
141 printf("Output v%d a%d: %f %f %f %f\n",
142 vert
, attr
, vOut
[0], vOut
[1], vOut
[2], vOut
[3]);
144 vOut
+= output_stride
/ 4;
148 /* advance to next group of four input/output verts */
149 input
= (const float (*)[4])((const char *)input
+ input_stride
* max_vertices
);
150 output
= (float (*)[4])((char *)output
+ output_stride
* max_vertices
);
156 vs_ppc_delete( struct draw_vertex_shader
*base
)
158 struct draw_ppc_vertex_shader
*shader
= (struct draw_ppc_vertex_shader
*)base
;
160 ppc_release_func( &shader
->ppc_program
);
162 align_free( (void *) shader
->base
.immediates
);
164 FREE( (void*) shader
->base
.state
.tokens
);
169 struct draw_vertex_shader
*
170 draw_create_vs_ppc(struct draw_context
*draw
,
171 const struct pipe_shader_state
*templ
)
173 struct draw_ppc_vertex_shader
*vs
;
175 vs
= CALLOC_STRUCT( draw_ppc_vertex_shader
);
179 /* we make a private copy of the tokens */
180 vs
->base
.state
.tokens
= tgsi_dup_tokens(templ
->tokens
);
181 if (!vs
->base
.state
.tokens
)
184 tgsi_scan_shader(templ
->tokens
, &vs
->base
.info
);
186 vs
->base
.draw
= draw
;
189 vs
->base
.create_varient
= draw_vs_varient_aos_ppc
;
192 vs
->base
.create_varient
= draw_vs_varient_generic
;
193 vs
->base
.prepare
= vs_ppc_prepare
;
194 vs
->base
.run_linear
= vs_ppc_run_linear
;
195 vs
->base
.delete = vs_ppc_delete
;
197 vs
->base
.immediates
= align_malloc(TGSI_EXEC_NUM_IMMEDIATES
* 4 *
200 ppc_init_func( &vs
->ppc_program
);
203 ppc_print_code(&vs
->ppc_program
, TRUE
);
204 ppc_indent(&vs
->ppc_program
, 8);
207 if (!tgsi_emit_ppc( (struct tgsi_token
*) vs
->base
.state
.tokens
,
209 (float (*)[4]) vs
->base
.immediates
,
213 vs
->func
= (codegen_function
) ppc_get_func( &vs
->ppc_program
);
222 debug_error("tgsi_emit_ppc() failed, falling back to interpreter\n");
225 ppc_release_func( &vs
->ppc_program
);
233 #else /* PIPE_ARCH_PPC */
236 struct draw_vertex_shader
*
237 draw_create_vs_ppc( struct draw_context
*draw
,
238 const struct pipe_shader_state
*templ
)
244 #endif /* PIPE_ARCH_PPC */