e8801addac3401b3e6c6b835fe03786a7dd6259a
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 "pipe/p_util.h"
35 #include "draw_private.h"
36 #include "draw_context.h"
37 #include "draw_vertex.h"
39 #include "x86/rtasm/x86sse.h"
41 #include "pipe/tgsi/exec/tgsi_core.h"
42 #include "pipe/llvm/gallivm.h"
48 static INLINE
unsigned
49 compute_clipmask(const float *clip
, /*const*/ float plane
[][4], unsigned nr
)
54 for (i
= 0; i
< nr
; i
++) {
55 if (dot4(clip
, plane
[i
]) < 0)
63 typedef void (XSTDCALL
*codegen_function
) (
64 const struct tgsi_exec_vector
*input
,
65 struct tgsi_exec_vector
*output
,
67 struct tgsi_exec_vector
*temporary
);
71 * Transform vertices with the current vertex program/shader
72 * Up to four vertices can be shaded at a time.
73 * \param vbuffer the input vertex data
74 * \param elts indexes of four input vertices
75 * \param count number of vertices to shade [1..4]
76 * \param vOut array of pointers to four output vertices
79 run_vertex_program(struct draw_context
*draw
,
80 unsigned elts
[4], unsigned count
,
81 struct vertex_header
*vOut
[])
83 struct tgsi_exec_machine
*machine
= &draw
->machine
;
86 ALIGN16_DECL(struct tgsi_exec_vector
, inputs
, PIPE_ATTRIB_MAX
);
87 ALIGN16_DECL(struct tgsi_exec_vector
, outputs
, PIPE_ATTRIB_MAX
);
88 const float *scale
= draw
->viewport
.scale
;
89 const float *trans
= draw
->viewport
.translate
;
92 assert(draw
->vertex_shader
->state
->output_semantic_name
[0]
93 == TGSI_SEMANTIC_POSITION
);
95 /* Consts does not require 16 byte alignment. */
96 machine
->Consts
= (float (*)[4]) draw
->user
.constants
;
98 machine
->Inputs
= ALIGN16_ASSIGN(inputs
);
99 machine
->Outputs
= ALIGN16_ASSIGN(outputs
);
101 draw_vertex_fetch( draw
, machine
, elts
, count
);
104 #if defined(__i386__) || defined(__386__)
107 codegen_function func
= (codegen_function
) x86_get_func( &draw
->vertex_shader
->sse2_program
);
118 tgsi_exec_machine_run( machine
);
121 /* store machine results */
122 for (j
= 0; j
< count
; j
++) {
126 /* Handle attr[0] (position) specially:
128 * XXX: Computing the clipmask should be done in the vertex
129 * program as a set of DP4 instructions appended to the
130 * user-provided code.
132 x
= vOut
[j
]->clip
[0] = machine
->Outputs
[0].xyzw
[0].f
[j
];
133 y
= vOut
[j
]->clip
[1] = machine
->Outputs
[0].xyzw
[1].f
[j
];
134 z
= vOut
[j
]->clip
[2] = machine
->Outputs
[0].xyzw
[2].f
[j
];
135 w
= vOut
[j
]->clip
[3] = machine
->Outputs
[0].xyzw
[3].f
[j
];
137 vOut
[j
]->clipmask
= compute_clipmask(vOut
[j
]->clip
, draw
->plane
, draw
->nr_planes
);
138 vOut
[j
]->edgeflag
= 1;
146 /* Viewport mapping */
147 vOut
[j
]->data
[0][0] = x
* scale
[0] + trans
[0];
148 vOut
[j
]->data
[0][1] = y
* scale
[1] + trans
[1];
149 vOut
[j
]->data
[0][2] = z
* scale
[2] + trans
[2];
150 vOut
[j
]->data
[0][3] = w
;
153 printf("output[%d]win: %f %f %f %f\n", j
,
157 vOut
[j
]->data
[0][3]);
159 /* Remaining attributes are packed into sequential post-transform
160 * vertex attrib slots.
161 * Skip 0 since we just did it above.
162 * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs.
164 for (slot
= 1; slot
< draw
->vertex_info
.num_attribs
- 2; slot
++) {
165 vOut
[j
]->data
[slot
][0] = machine
->Outputs
[slot
].xyzw
[0].f
[j
];
166 vOut
[j
]->data
[slot
][1] = machine
->Outputs
[slot
].xyzw
[1].f
[j
];
167 vOut
[j
]->data
[slot
][2] = machine
->Outputs
[slot
].xyzw
[2].f
[j
];
168 vOut
[j
]->data
[slot
][3] = machine
->Outputs
[slot
].xyzw
[3].f
[j
];
170 printf("output[%d][%d]: %f %f %f %f\n", j
, slot
,
171 vOut
[j
]->data
[slot
][0],
172 vOut
[j
]->data
[slot
][1],
173 vOut
[j
]->data
[slot
][2],
174 vOut
[j
]->data
[slot
][3]);
177 } /* loop over vertices */
182 * Run the vertex shader on all vertices in the vertex queue.
183 * Called by the draw module when the vertx cache needs to be flushed.
185 void draw_vertex_shader_queue_flush( struct draw_context
*draw
)
189 // fprintf(stderr, " q(%d) ", draw->vs.queue_nr );
191 if (draw
->vertex_shader
->llvm_prog
) {
192 draw_vertex_shader_queue_flush_llvm(draw
);
197 /* run vertex shader on vertex cache entries, four per invokation */
198 for (i
= 0; i
< draw
->vs
.queue_nr
; i
+= 4) {
199 struct vertex_header
*dests
[4];
203 for (j
= 0; j
< 4; j
++) {
204 elts
[j
] = draw
->vs
.queue
[i
+ j
].elt
;
205 dests
[j
] = draw
->vs
.queue
[i
+ j
].dest
;
208 n
= MIN2(4, draw
->vs
.queue_nr
- i
);
212 run_vertex_program(draw
, elts
, n
, dests
);
215 draw
->vs
.queue_nr
= 0;
220 draw_create_vertex_shader(struct draw_context
*draw
,
221 const struct pipe_shader_state
*shader
)
223 struct draw_vertex_shader
*vs
;
225 vs
= CALLOC_STRUCT( draw_vertex_shader
);
232 #if defined(__i386__) || defined(__386__)
234 /* cast-away const */
235 struct pipe_shader_state
*sh
= (struct pipe_shader_state
*) shader
;
237 x86_init_func( &vs
->sse2_program
);
238 tgsi_emit_sse2( sh
->tokens
, &vs
->sse2_program
);
242 vs
->llvm_prog
= gallivm_from_tgsi(shader
->tokens
, GALLIVM_VS
);
243 draw
->engine
= gallivm_global_cpu_engine();
245 draw
->engine
= gallivm_cpu_engine_create(vs
->llvm_prog
);
248 gallivm_cpu_jit_compile(draw
->engine
, vs
->llvm_prog
);
254 void draw_bind_vertex_shader(struct draw_context
*draw
,
258 draw
->vertex_shader
= (struct draw_vertex_shader
*)(vcso
);
260 /* specify the fragment program to interpret/execute */
261 tgsi_exec_machine_init(&draw
->machine
,
262 draw
->vertex_shader
->state
->tokens
,
267 void draw_delete_vertex_shader(struct draw_context
*draw
,
270 struct draw_vertex_shader
*vs
;
272 vs
= (struct draw_vertex_shader
*) vcso
;
274 #if defined(__i386__) || defined(__386__)
275 x86_release_func( (struct x86_function
*) &vs
->sse2_program
);