1 /**************************************************************************
3 * Copyright 2007-2008 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 **************************************************************************/
31 #include "pipe/p_compiler.h"
32 #include "pipe/p_state.h"
34 #if defined __cplusplus
38 #define MAX_LABELS (4 * 1024) /**< basically, max instructions */
40 #define NUM_CHANNELS 4 /* R,G,B,A */
41 #define QUAD_SIZE 4 /* 4 pixel/quad */
44 * Registers may be treated as float, signed int or unsigned int.
46 union tgsi_exec_channel
50 unsigned u
[QUAD_SIZE
];
54 * A vector[RGBA] of channels[4 pixels]
56 struct tgsi_exec_vector
58 union tgsi_exec_channel xyzw
[NUM_CHANNELS
];
62 * For fragment programs, information for computing fragment input
63 * values from plane equation of the triangle/line.
65 struct tgsi_interp_coef
67 float a0
[NUM_CHANNELS
]; /* in an xyzw layout */
68 float dadx
[NUM_CHANNELS
];
69 float dady
[NUM_CHANNELS
];
73 * Information for sampling textures, which must be implemented
74 * by code outside the TGSI executor.
78 /** Get samples for four fragments in a quad */
79 void (*get_samples
)(struct tgsi_sampler
*sampler
,
80 const float s
[QUAD_SIZE
],
81 const float t
[QUAD_SIZE
],
82 const float p
[QUAD_SIZE
],
84 float rgba
[NUM_CHANNELS
][QUAD_SIZE
]);
88 * For branching/calling subroutines.
90 struct tgsi_exec_labels
92 unsigned labels
[MAX_LABELS
][2];
97 #define TGSI_EXEC_NUM_TEMPS 128
98 #define TGSI_EXEC_NUM_IMMEDIATES 256
101 * Locations of various utility registers (_I = Index, _C = Channel)
103 #define TGSI_EXEC_TEMP_00000000_I (TGSI_EXEC_NUM_TEMPS + 0)
104 #define TGSI_EXEC_TEMP_00000000_C 0
106 #define TGSI_EXEC_TEMP_7FFFFFFF_I (TGSI_EXEC_NUM_TEMPS + 0)
107 #define TGSI_EXEC_TEMP_7FFFFFFF_C 1
109 #define TGSI_EXEC_TEMP_80000000_I (TGSI_EXEC_NUM_TEMPS + 0)
110 #define TGSI_EXEC_TEMP_80000000_C 2
112 #define TGSI_EXEC_TEMP_FFFFFFFF_I (TGSI_EXEC_NUM_TEMPS + 0)
113 #define TGSI_EXEC_TEMP_FFFFFFFF_C 3
115 #define TGSI_EXEC_TEMP_ONE_I (TGSI_EXEC_NUM_TEMPS + 1)
116 #define TGSI_EXEC_TEMP_ONE_C 0
118 #define TGSI_EXEC_TEMP_TWO_I (TGSI_EXEC_NUM_TEMPS + 1)
119 #define TGSI_EXEC_TEMP_TWO_C 1
121 #define TGSI_EXEC_TEMP_128_I (TGSI_EXEC_NUM_TEMPS + 1)
122 #define TGSI_EXEC_TEMP_128_C 2
124 #define TGSI_EXEC_TEMP_MINUS_128_I (TGSI_EXEC_NUM_TEMPS + 1)
125 #define TGSI_EXEC_TEMP_MINUS_128_C 3
127 #define TGSI_EXEC_TEMP_KILMASK_I (TGSI_EXEC_NUM_TEMPS + 2)
128 #define TGSI_EXEC_TEMP_KILMASK_C 0
130 #define TGSI_EXEC_TEMP_OUTPUT_I (TGSI_EXEC_NUM_TEMPS + 2)
131 #define TGSI_EXEC_TEMP_OUTPUT_C 1
133 #define TGSI_EXEC_TEMP_PRIMITIVE_I (TGSI_EXEC_NUM_TEMPS + 2)
134 #define TGSI_EXEC_TEMP_PRIMITIVE_C 2
136 /* NVIDIA condition code (CC) vector
138 #define TGSI_EXEC_CC_GT 0x01
139 #define TGSI_EXEC_CC_EQ 0x02
140 #define TGSI_EXEC_CC_LT 0x04
141 #define TGSI_EXEC_CC_UN 0x08
143 #define TGSI_EXEC_CC_X_MASK 0x000000ff
144 #define TGSI_EXEC_CC_X_SHIFT 0
145 #define TGSI_EXEC_CC_Y_MASK 0x0000ff00
146 #define TGSI_EXEC_CC_Y_SHIFT 8
147 #define TGSI_EXEC_CC_Z_MASK 0x00ff0000
148 #define TGSI_EXEC_CC_Z_SHIFT 16
149 #define TGSI_EXEC_CC_W_MASK 0xff000000
150 #define TGSI_EXEC_CC_W_SHIFT 24
152 #define TGSI_EXEC_TEMP_CC_I (TGSI_EXEC_NUM_TEMPS + 2)
153 #define TGSI_EXEC_TEMP_CC_C 3
155 #define TGSI_EXEC_TEMP_THREE_I (TGSI_EXEC_NUM_TEMPS + 3)
156 #define TGSI_EXEC_TEMP_THREE_C 0
158 #define TGSI_EXEC_TEMP_HALF_I (TGSI_EXEC_NUM_TEMPS + 3)
159 #define TGSI_EXEC_TEMP_HALF_C 1
161 /* execution mask, each value is either 0 or ~0 */
162 #define TGSI_EXEC_MASK_I (TGSI_EXEC_NUM_TEMPS + 3)
163 #define TGSI_EXEC_MASK_C 2
165 /* 4 register buffer for various purposes */
166 #define TGSI_EXEC_TEMP_R0 (TGSI_EXEC_NUM_TEMPS + 4)
167 #define TGSI_EXEC_NUM_TEMP_R 4
169 #define TGSI_EXEC_TEMP_ADDR (TGSI_EXEC_NUM_TEMPS + 8)
170 #define TGSI_EXEC_NUM_ADDRS 1
171 #define TGSI_EXEC_NUM_TEMP_EXTRAS 9
175 #define TGSI_EXEC_MAX_COND_NESTING 20
176 #define TGSI_EXEC_MAX_LOOP_NESTING 20
177 #define TGSI_EXEC_MAX_CALL_NESTING 20
179 /* The maximum number of input attributes per vertex. For 2D
180 * input register files, this is the stride between two 1D
183 #define TGSI_EXEC_MAX_INPUT_ATTRIBS 17
185 /* The maximum number of constant vectors per constant buffer.
187 #define TGSI_EXEC_MAX_CONST_BUFFER 4096
190 /** function call/activation record */
191 struct tgsi_call_record
201 * Run-time virtual machine state for executing TGSI shader.
203 struct tgsi_exec_machine
205 /* Total = program temporaries + internal temporaries
207 struct tgsi_exec_vector Temps
[TGSI_EXEC_NUM_TEMPS
+
208 TGSI_EXEC_NUM_TEMP_EXTRAS
];
210 float Imms
[TGSI_EXEC_NUM_IMMEDIATES
][4];
212 struct tgsi_exec_vector Inputs
[PIPE_MAX_ATTRIBS
];
213 struct tgsi_exec_vector Outputs
[PIPE_MAX_ATTRIBS
];
215 struct tgsi_exec_vector
*Addrs
;
217 struct tgsi_sampler
**Samplers
;
220 const float (*Consts
)[4];
221 const struct tgsi_token
*Tokens
; /**< Declarations, instructions */
222 unsigned Processor
; /**< TGSI_PROCESSOR_x */
224 /* GEOMETRY processor only. */
225 unsigned *Primitives
;
227 /* FRAGMENT processor only. */
228 const struct tgsi_interp_coef
*InterpCoefs
;
229 struct tgsi_exec_vector QuadPos
;
231 /* Conditional execution masks */
232 uint CondMask
; /**< For IF/ELSE/ENDIF */
233 uint LoopMask
; /**< For BGNLOOP/ENDLOOP */
234 uint ContMask
; /**< For loop CONT statements */
235 uint FuncMask
; /**< For function calls */
236 uint ExecMask
; /**< = CondMask & LoopMask */
238 /** Condition mask stack (for nested conditionals) */
239 uint CondStack
[TGSI_EXEC_MAX_COND_NESTING
];
242 /** Loop mask stack (for nested loops) */
243 uint LoopStack
[TGSI_EXEC_MAX_LOOP_NESTING
];
246 /** Loop label stack */
247 uint LoopLabelStack
[TGSI_EXEC_MAX_LOOP_NESTING
];
248 int LoopLabelStackTop
;
250 /** Loop counter stack (x = count, y = current, z = step) */
251 struct tgsi_exec_vector LoopCounterStack
[TGSI_EXEC_MAX_LOOP_NESTING
];
252 int LoopCounterStackTop
;
254 /** Loop continue mask stack (see comments in tgsi_exec.c) */
255 uint ContStack
[TGSI_EXEC_MAX_LOOP_NESTING
];
258 /** Function execution mask stack (for executing subroutine code) */
259 uint FuncStack
[TGSI_EXEC_MAX_CALL_NESTING
];
262 /** Function call stack for saving/restoring the program counter */
263 struct tgsi_call_record CallStack
[TGSI_EXEC_MAX_CALL_NESTING
];
266 struct tgsi_full_instruction
*Instructions
;
267 uint NumInstructions
;
269 struct tgsi_full_declaration
*Declarations
;
270 uint NumDeclarations
;
272 struct tgsi_exec_labels Labels
;
275 struct tgsi_exec_machine
*
276 tgsi_exec_machine_create( void );
279 tgsi_exec_machine_destroy(struct tgsi_exec_machine
*mach
);
283 tgsi_exec_machine_bind_shader(
284 struct tgsi_exec_machine
*mach
,
285 const struct tgsi_token
*tokens
,
287 struct tgsi_sampler
**samplers
);
290 tgsi_exec_machine_run(
291 struct tgsi_exec_machine
*mach
);
295 tgsi_exec_machine_free_data(struct tgsi_exec_machine
*mach
);
299 tgsi_check_soa_dependencies(const struct tgsi_full_instruction
*inst
);
303 tgsi_set_kill_mask(struct tgsi_exec_machine
*mach
, unsigned mask
)
305 mach
->Temps
[TGSI_EXEC_TEMP_KILMASK_I
].xyzw
[TGSI_EXEC_TEMP_KILMASK_C
].u
[0] =
310 /** Set execution mask values prior to executing the shader */
312 tgsi_set_exec_mask(struct tgsi_exec_machine
*mach
,
313 boolean ch0
, boolean ch1
, boolean ch2
, boolean ch3
)
315 int *mask
= mach
->Temps
[TGSI_EXEC_MASK_I
].xyzw
[TGSI_EXEC_MASK_C
].i
;
316 mask
[0] = ch0
? ~0 : 0;
317 mask
[1] = ch1
? ~0 : 0;
318 mask
[2] = ch2
? ~0 : 0;
319 mask
[3] = ch3
? ~0 : 0;
323 #if defined __cplusplus
327 #endif /* TGSI_EXEC_H */