Add CMPH instruction to fragprog
[mesa.git] / src / mesa / drivers / dri / r300 / r300_reg.h
1 /**************************************************************************
2
3 Copyright (C) 2004-2005 Nicolai Haehnle et al.
4
5 Permission is hereby granted, free of charge, to any person obtaining a
6 copy of this software and associated documentation files (the "Software"),
7 to deal in the Software without restriction, including without limitation
8 on the rights to use, copy, modify, merge, publish, distribute, sub
9 license, and/or sell copies of the Software, and to permit persons to whom
10 the Software is furnished to do so, subject to the following conditions:
11
12 The above copyright notice and this permission notice (including the next
13 paragraph) shall be included in all copies or substantial portions of the
14 Software.
15
16 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
19 THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
20 DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
21 OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
22 USE OR OTHER DEALINGS IN THE SOFTWARE.
23
24 **************************************************************************/
25
26 #ifndef _R300_REG_H
27 #define _R300_REG_H
28
29 #define R300_MC_INIT_MISC_LAT_TIMER 0x180
30 # define R300_MC_MISC__MC_CPR_INIT_LAT_SHIFT 0
31 # define R300_MC_MISC__MC_VF_INIT_LAT_SHIFT 4
32 # define R300_MC_MISC__MC_DISP0R_INIT_LAT_SHIFT 8
33 # define R300_MC_MISC__MC_DISP1R_INIT_LAT_SHIFT 12
34 # define R300_MC_MISC__MC_FIXED_INIT_LAT_SHIFT 16
35 # define R300_MC_MISC__MC_E2R_INIT_LAT_SHIFT 20
36 # define R300_MC_MISC__MC_SAME_PAGE_PRIO_SHIFT 24
37 # define R300_MC_MISC__MC_GLOBW_INIT_LAT_SHIFT 28
38
39
40 #define R300_MC_INIT_GFX_LAT_TIMER 0x154
41 # define R300_MC_MISC__MC_G3D0R_INIT_LAT_SHIFT 0
42 # define R300_MC_MISC__MC_G3D1R_INIT_LAT_SHIFT 4
43 # define R300_MC_MISC__MC_G3D2R_INIT_LAT_SHIFT 8
44 # define R300_MC_MISC__MC_G3D3R_INIT_LAT_SHIFT 12
45 # define R300_MC_MISC__MC_TX0R_INIT_LAT_SHIFT 16
46 # define R300_MC_MISC__MC_TX1R_INIT_LAT_SHIFT 20
47 # define R300_MC_MISC__MC_GLOBR_INIT_LAT_SHIFT 24
48 # define R300_MC_MISC__MC_GLOBW_FULL_LAT_SHIFT 28
49
50 /*
51 * This file contains registers and constants for the R300. They have been
52 * found mostly by examining command buffers captured using glxtest, as well
53 * as by extrapolating some known registers and constants from the R200.
54 * I am fairly certain that they are correct unless stated otherwise
55 * in comments.
56 */
57
58 #define R300_SE_VPORT_XSCALE 0x1D98
59 #define R300_SE_VPORT_XOFFSET 0x1D9C
60 #define R300_SE_VPORT_YSCALE 0x1DA0
61 #define R300_SE_VPORT_YOFFSET 0x1DA4
62 #define R300_SE_VPORT_ZSCALE 0x1DA8
63 #define R300_SE_VPORT_ZOFFSET 0x1DAC
64
65
66 /* This register is written directly and also starts data section
67 * in many 3d CP_PACKET3's
68 */
69 #define R300_VAP_VF_CNTL 0x2084
70 # define R300_VAP_VF_CNTL__PRIM_TYPE__SHIFT 0
71 # define R300_VAP_VF_CNTL__PRIM_NONE (0<<0)
72 # define R300_VAP_VF_CNTL__PRIM_POINTS (1<<0)
73 # define R300_VAP_VF_CNTL__PRIM_LINES (2<<0)
74 # define R300_VAP_VF_CNTL__PRIM_LINE_STRIP (3<<0)
75 # define R300_VAP_VF_CNTL__PRIM_TRIANGLES (4<<0)
76 # define R300_VAP_VF_CNTL__PRIM_TRIANGLE_FAN (5<<0)
77 # define R300_VAP_VF_CNTL__PRIM_TRIANGLE_STRIP (6<<0)
78 # define R300_VAP_VF_CNTL__PRIM_LINE_LOOP (12<<0)
79 # define R300_VAP_VF_CNTL__PRIM_QUADS (13<<0)
80 # define R300_VAP_VF_CNTL__PRIM_QUAD_STRIP (14<<0)
81 # define R300_VAP_VF_CNTL__PRIM_POLYGON (15<<0)
82
83 # define R300_VAP_VF_CNTL__PRIM_WALK__SHIFT 4
84 /* State based - direct writes to registers trigger vertex
85 generation */
86 # define R300_VAP_VF_CNTL__PRIM_WALK_STATE_BASED (0<<4)
87 # define R300_VAP_VF_CNTL__PRIM_WALK_INDICES (1<<4)
88 # define R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST (2<<4)
89 # define R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_EMBEDDED (3<<4)
90
91 /* I don't think I saw these three used.. */
92 # define R300_VAP_VF_CNTL__COLOR_ORDER__SHIFT 6
93 # define R300_VAP_VF_CNTL__TCL_OUTPUT_CTL_ENA__SHIFT 9
94 # define R300_VAP_VF_CNTL__PROG_STREAM_ENA__SHIFT 10
95
96 /* index size - when not set the indices are assumed to be 16 bit */
97 # define R300_VAP_VF_CNTL__INDEX_SIZE_32bit (1<<11)
98 /* number of vertices */
99 # define R300_VAP_VF_CNTL__NUM_VERTICES__SHIFT 16
100
101 /* BEGIN: Wild guesses */
102 #define R300_VAP_OUTPUT_VTX_FMT_0 0x2090
103 # define R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT (1<<0)
104 # define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT (1<<1)
105 # define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT (1<<2) /* GUESS */
106 # define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT (1<<3) /* GUESS */
107 # define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT (1<<4) /* GUESS */
108 # define R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT (1<<16) /* GUESS */
109
110 #define R300_VAP_OUTPUT_VTX_FMT_1 0x2094
111 # define R300_VAP_OUTPUT_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT 0
112 # define R300_VAP_OUTPUT_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT 3
113 # define R300_VAP_OUTPUT_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT 6
114 # define R300_VAP_OUTPUT_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT 9
115 # define R300_VAP_OUTPUT_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT 12
116 # define R300_VAP_OUTPUT_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT 15
117 # define R300_VAP_OUTPUT_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT 18
118 # define R300_VAP_OUTPUT_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT 21
119 /* END: Wild guesses */
120
121 #define R300_SE_VTE_CNTL 0x20b0
122 # define R300_VPORT_X_SCALE_ENA 0x00000001
123 # define R300_VPORT_X_OFFSET_ENA 0x00000002
124 # define R300_VPORT_Y_SCALE_ENA 0x00000004
125 # define R300_VPORT_Y_OFFSET_ENA 0x00000008
126 # define R300_VPORT_Z_SCALE_ENA 0x00000010
127 # define R300_VPORT_Z_OFFSET_ENA 0x00000020
128 # define R300_VTX_XY_FMT 0x00000100
129 # define R300_VTX_Z_FMT 0x00000200
130 # define R300_VTX_W0_FMT 0x00000400
131 # define R300_VTX_W0_NORMALIZE 0x00000800
132 # define R300_VTX_ST_DENORMALIZED 0x00001000
133
134 /* BEGIN: Vertex data assembly - lots of uncertainties */
135
136 /* gap */
137
138 #define R300_VAP_CNTL 0x2140
139 # define R300_VC_NO_SWAP (0 << 0)
140 # define R300_VC_16BIT_SWAP (1 << 0)
141 # define R300_VC_32BIT_SWAP (2 << 0)
142
143 /* gap */
144
145 /* Where do we get our vertex data?
146 *
147 * Vertex data either comes either from immediate mode registers or from
148 * vertex arrays.
149 * There appears to be no mixed mode (though we can force the pitch of
150 * vertex arrays to 0, effectively reusing the same element over and over
151 * again).
152 *
153 * Immediate mode is controlled by the INPUT_CNTL registers. I am not sure
154 * if these registers influence vertex array processing.
155 *
156 * Vertex arrays are controlled via the 3D_LOAD_VBPNTR packet3.
157 *
158 * In both cases, vertex attributes are then passed through INPUT_ROUTE.
159 *
160 * Beginning with INPUT_ROUTE_0_0 is a list of WORDs that route vertex data
161 * into the vertex processor's input registers.
162 * The first word routes the first input, the second word the second, etc.
163 * The corresponding input is routed into the register with the given index.
164 * The list is ended by a word with INPUT_ROUTE_END set.
165 *
166 * Always set COMPONENTS_4 in immediate mode.
167 */
168
169 #define R300_VAP_INPUT_ROUTE_0_0 0x2150
170 # define R300_INPUT_ROUTE_COMPONENTS_1 (0 << 0)
171 # define R300_INPUT_ROUTE_COMPONENTS_2 (1 << 0)
172 # define R300_INPUT_ROUTE_COMPONENTS_3 (2 << 0)
173 # define R300_INPUT_ROUTE_COMPONENTS_4 (3 << 0)
174 # define R300_INPUT_ROUTE_COMPONENTS_RGBA (4 << 0) /* GUESS */
175 # define R300_VAP_INPUT_ROUTE_IDX_SHIFT 8
176 # define R300_VAP_INPUT_ROUTE_IDX_MASK (31 << 8) /* GUESS */
177 # define R300_VAP_INPUT_ROUTE_END (1 << 13)
178 # define R300_INPUT_ROUTE_IMMEDIATE_MODE (0 << 14) /* GUESS */
179 # define R300_INPUT_ROUTE_FLOAT (1 << 14) /* GUESS */
180 # define R300_INPUT_ROUTE_UNSIGNED_BYTE (2 << 14) /* GUESS */
181 # define R300_INPUT_ROUTE_FLOAT_COLOR (3 << 14) /* GUESS */
182 #define R300_VAP_INPUT_ROUTE_0_1 0x2154
183 #define R300_VAP_INPUT_ROUTE_0_2 0x2158
184 #define R300_VAP_INPUT_ROUTE_0_3 0x215C
185 #define R300_VAP_INPUT_ROUTE_0_4 0x2160
186 #define R300_VAP_INPUT_ROUTE_0_5 0x2164
187 #define R300_VAP_INPUT_ROUTE_0_6 0x2168
188 #define R300_VAP_INPUT_ROUTE_0_7 0x216C
189
190 /* gap */
191
192 /* Notes:
193 * - always set up to produce at least two attributes:
194 * if vertex program uses only position, fglrx will set normal, too
195 * - INPUT_CNTL_0_COLOR and INPUT_CNTL_COLOR bits are always equal.
196 */
197 #define R300_VAP_INPUT_CNTL_0 0x2180
198 # define R300_INPUT_CNTL_0_COLOR 0x00000001
199 #define R300_VAP_INPUT_CNTL_1 0x2184
200 # define R300_INPUT_CNTL_POS 0x00000001
201 # define R300_INPUT_CNTL_NORMAL 0x00000002
202 # define R300_INPUT_CNTL_COLOR 0x00000004
203 # define R300_INPUT_CNTL_TC0 0x00000400
204 # define R300_INPUT_CNTL_TC1 0x00000800
205 # define R300_INPUT_CNTL_TC2 0x00001000 /* GUESS */
206 # define R300_INPUT_CNTL_TC3 0x00002000 /* GUESS */
207 # define R300_INPUT_CNTL_TC4 0x00004000 /* GUESS */
208 # define R300_INPUT_CNTL_TC5 0x00008000 /* GUESS */
209 # define R300_INPUT_CNTL_TC6 0x00010000 /* GUESS */
210 # define R300_INPUT_CNTL_TC7 0x00020000 /* GUESS */
211
212 /* gap */
213
214 /* Words parallel to INPUT_ROUTE_0; All words that are active in INPUT_ROUTE_0
215 * are set to a swizzling bit pattern, other words are 0.
216 *
217 * In immediate mode, the pattern is always set to xyzw. In vertex array
218 * mode, the swizzling pattern is e.g. used to set zw components in texture
219 * coordinates with only tweo components.
220 */
221 #define R300_VAP_INPUT_ROUTE_1_0 0x21E0
222 # define R300_INPUT_ROUTE_SELECT_X 0
223 # define R300_INPUT_ROUTE_SELECT_Y 1
224 # define R300_INPUT_ROUTE_SELECT_Z 2
225 # define R300_INPUT_ROUTE_SELECT_W 3
226 # define R300_INPUT_ROUTE_SELECT_ZERO 4
227 # define R300_INPUT_ROUTE_SELECT_ONE 5
228 # define R300_INPUT_ROUTE_SELECT_MASK 7
229 # define R300_INPUT_ROUTE_X_SHIFT 0
230 # define R300_INPUT_ROUTE_Y_SHIFT 3
231 # define R300_INPUT_ROUTE_Z_SHIFT 6
232 # define R300_INPUT_ROUTE_W_SHIFT 9
233 # define R300_INPUT_ROUTE_ENABLE (15 << 12)
234 #define R300_VAP_INPUT_ROUTE_1_1 0x21E4
235 #define R300_VAP_INPUT_ROUTE_1_2 0x21E8
236 #define R300_VAP_INPUT_ROUTE_1_3 0x21EC
237 #define R300_VAP_INPUT_ROUTE_1_4 0x21F0
238 #define R300_VAP_INPUT_ROUTE_1_5 0x21F4
239 #define R300_VAP_INPUT_ROUTE_1_6 0x21F8
240 #define R300_VAP_INPUT_ROUTE_1_7 0x21FC
241
242 /* END: Vertex data assembly */
243
244 /* gap */
245
246 /* BEGIN: Upload vertex program and data */
247
248 /*
249 * The programmable vertex shader unit has a memory bank of unknown size
250 * that can be written to in 16 byte units by writing the address into
251 * UPLOAD_ADDRESS, followed by data in UPLOAD_DATA (multiples of 4 DWORDs).
252 *
253 * Pointers into the memory bank are always in multiples of 16 bytes.
254 *
255 * The memory bank is divided into areas with fixed meaning.
256 *
257 * Starting at address UPLOAD_PROGRAM: Vertex program instructions.
258 * Native limits reported by drivers from ATI suggest size 256 (i.e. 4KB),
259 * whereas the difference between known addresses suggests size 512.
260 *
261 * Starting at address UPLOAD_PARAMETERS: Vertex program parameters.
262 * Native reported limits and the VPI layout suggest size 256, whereas
263 * difference between known addresses suggests size 512.
264 *
265 * At address UPLOAD_POINTSIZE is a vector (0, 0, ps, 0), where ps is the
266 * floating point pointsize. The exact purpose of this state is uncertain,
267 * as there is also the R300_RE_POINTSIZE register.
268 *
269 * Multiple vertex programs and parameter sets can be loaded at once,
270 * which could explain the size discrepancy.
271 */
272 #define R300_VAP_PVS_UPLOAD_ADDRESS 0x2200
273 # define R300_PVS_UPLOAD_PROGRAM 0x00000000
274 # define R300_PVS_UPLOAD_PARAMETERS 0x00000200
275 # define R300_PVS_UPLOAD_POINTSIZE 0x00000406
276
277 /* gap */
278
279 #define R300_VAP_PVS_UPLOAD_DATA 0x2208
280
281 /* END: Upload vertex program and data */
282
283 /* gap */
284
285 /* I do not know the purpose of this register. However, I do know that
286 * it is set to 221C_CLEAR for clear operations and to 221C_NORMAL
287 * for normal rendering.
288 */
289 #define R300_VAP_UNKNOWN_221C 0x221C
290 # define R300_221C_NORMAL 0x00000000
291 # define R300_221C_CLEAR 0x0001C000
292
293 /* gap */
294
295 /* Sometimes, END_OF_PKT and 0x2284=0 are the only commands sent between
296 * rendering commands and overwriting vertex program parameters.
297 * Therefore, I suspect writing zero to 0x2284 synchronizes the engine and
298 * avoids bugs caused by still running shaders reading bad data from memory.
299 */
300 #define R300_VAP_PVS_WAITIDLE 0x2284 /* GUESS */
301
302 /* Absolutely no clue what this register is about. */
303 #define R300_VAP_UNKNOWN_2288 0x2288
304 # define R300_2288_R300 0x00750000 /* -- nh */
305 # define R300_2288_RV350 0x0000FFFF /* -- Vladimir */
306
307 /* gap */
308
309 /* Addresses are relative to the vertex program instruction area of the
310 * memory bank. PROGRAM_END points to the last instruction of the active
311 * program
312 *
313 * The meaning of the two UNKNOWN fields is obviously not known. However,
314 * experiments so far have shown that both *must* point to an instruction
315 * inside the vertex program, otherwise the GPU locks up.
316 * fglrx usually sets CNTL_3_UNKNOWN to the end of the program and
317 * CNTL_1_UNKNOWN points to instruction where last write to position takes
318 * place.
319 * Most likely this is used to ignore rest of the program in cases
320 * where group of verts arent visible. For some reason this "section"
321 * is sometimes accepted other instruction that have no relationship with
322 *position calculations.
323 */
324 #define R300_VAP_PVS_CNTL_1 0x22D0
325 # define R300_PVS_CNTL_1_PROGRAM_START_SHIFT 0
326 # define R300_PVS_CNTL_1_POS_END_SHIFT 10
327 # define R300_PVS_CNTL_1_PROGRAM_END_SHIFT 20
328 /* Addresses are relative the the vertex program parameters area. */
329 #define R300_VAP_PVS_CNTL_2 0x22D4
330 # define R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT 0
331 # define R300_PVS_CNTL_2_PARAM_COUNT_SHIFT 16
332 #define R300_VAP_PVS_CNTL_3 0x22D8
333 # define R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT 10
334 # define R300_PVS_CNTL_3_PROGRAM_UNKNOWN2_SHIFT 0
335
336 /* The entire range from 0x2300 to 0x2AC inclusive seems to be used for
337 * immediate vertices
338 */
339 #define R300_VAP_VTX_COLOR_R 0x2464
340 #define R300_VAP_VTX_COLOR_G 0x2468
341 #define R300_VAP_VTX_COLOR_B 0x246C
342 #define R300_VAP_VTX_POS_0_X_1 0x2490 /* used for glVertex2*() */
343 #define R300_VAP_VTX_POS_0_Y_1 0x2494
344 #define R300_VAP_VTX_COLOR_PKD 0x249C /* RGBA */
345 #define R300_VAP_VTX_POS_0_X_2 0x24A0 /* used for glVertex3*() */
346 #define R300_VAP_VTX_POS_0_Y_2 0x24A4
347 #define R300_VAP_VTX_POS_0_Z_2 0x24A8
348 /* write 0 to indicate end of packet? */
349 #define R300_VAP_VTX_END_OF_PKT 0x24AC
350
351 /* gap */
352
353 /* These are values from r300_reg/r300_reg.h - they are known to be correct
354 * and are here so we can use one register file instead of several
355 * - Vladimir
356 */
357 #define R300_GB_VAP_RASTER_VTX_FMT_0 0x4000
358 # define R300_GB_VAP_RASTER_VTX_FMT_0__POS_PRESENT (1<<0)
359 # define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_0_PRESENT (1<<1)
360 # define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_1_PRESENT (1<<2)
361 # define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_2_PRESENT (1<<3)
362 # define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_3_PRESENT (1<<4)
363 # define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_SPACE (0xf<<5)
364 # define R300_GB_VAP_RASTER_VTX_FMT_0__PT_SIZE_PRESENT (0x1<<16)
365
366 #define R300_GB_VAP_RASTER_VTX_FMT_1 0x4004
367 /* each of the following is 3 bits wide, specifies number
368 of components */
369 # define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT 0
370 # define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT 3
371 # define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT 6
372 # define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT 9
373 # define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT 12
374 # define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT 15
375 # define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT 18
376 # define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT 21
377
378 /* UNK30 seems to enables point to quad transformation on textures
379 * (or something closely related to that).
380 * This bit is rather fatal at the time being due to lackings at pixel
381 * shader side
382 */
383 #define R300_GB_ENABLE 0x4008
384 # define R300_GB_POINT_STUFF_ENABLE (1<<0)
385 # define R300_GB_LINE_STUFF_ENABLE (1<<1)
386 # define R300_GB_TRIANGLE_STUFF_ENABLE (1<<2)
387 # define R300_GB_STENCIL_AUTO_ENABLE (1<<4)
388 # define R300_GB_UNK31 (1<<31)
389 /* each of the following is 2 bits wide */
390 #define R300_GB_TEX_REPLICATE 0
391 #define R300_GB_TEX_ST 1
392 #define R300_GB_TEX_STR 2
393 # define R300_GB_TEX0_SOURCE_SHIFT 16
394 # define R300_GB_TEX1_SOURCE_SHIFT 18
395 # define R300_GB_TEX2_SOURCE_SHIFT 20
396 # define R300_GB_TEX3_SOURCE_SHIFT 22
397 # define R300_GB_TEX4_SOURCE_SHIFT 24
398 # define R300_GB_TEX5_SOURCE_SHIFT 26
399 # define R300_GB_TEX6_SOURCE_SHIFT 28
400 # define R300_GB_TEX7_SOURCE_SHIFT 30
401
402 /* MSPOS - positions for multisample antialiasing (?) */
403 #define R300_GB_MSPOS0 0x4010
404 /* shifts - each of the fields is 4 bits */
405 # define R300_GB_MSPOS0__MS_X0_SHIFT 0
406 # define R300_GB_MSPOS0__MS_Y0_SHIFT 4
407 # define R300_GB_MSPOS0__MS_X1_SHIFT 8
408 # define R300_GB_MSPOS0__MS_Y1_SHIFT 12
409 # define R300_GB_MSPOS0__MS_X2_SHIFT 16
410 # define R300_GB_MSPOS0__MS_Y2_SHIFT 20
411 # define R300_GB_MSPOS0__MSBD0_Y 24
412 # define R300_GB_MSPOS0__MSBD0_X 28
413
414 #define R300_GB_MSPOS1 0x4014
415 # define R300_GB_MSPOS1__MS_X3_SHIFT 0
416 # define R300_GB_MSPOS1__MS_Y3_SHIFT 4
417 # define R300_GB_MSPOS1__MS_X4_SHIFT 8
418 # define R300_GB_MSPOS1__MS_Y4_SHIFT 12
419 # define R300_GB_MSPOS1__MS_X5_SHIFT 16
420 # define R300_GB_MSPOS1__MS_Y5_SHIFT 20
421 # define R300_GB_MSPOS1__MSBD1 24
422
423
424 #define R300_GB_TILE_CONFIG 0x4018
425 # define R300_GB_TILE_ENABLE (1<<0)
426 # define R300_GB_TILE_PIPE_COUNT_RV300 0
427 # define R300_GB_TILE_PIPE_COUNT_R300 (3<<1)
428 # define R300_GB_TILE_PIPE_COUNT_R420 (7<<1)
429 # define R300_GB_TILE_PIPE_COUNT_RV410 (3<<1)
430 # define R300_GB_TILE_SIZE_8 0
431 # define R300_GB_TILE_SIZE_16 (1<<4)
432 # define R300_GB_TILE_SIZE_32 (2<<4)
433 # define R300_GB_SUPER_SIZE_1 (0<<6)
434 # define R300_GB_SUPER_SIZE_2 (1<<6)
435 # define R300_GB_SUPER_SIZE_4 (2<<6)
436 # define R300_GB_SUPER_SIZE_8 (3<<6)
437 # define R300_GB_SUPER_SIZE_16 (4<<6)
438 # define R300_GB_SUPER_SIZE_32 (5<<6)
439 # define R300_GB_SUPER_SIZE_64 (6<<6)
440 # define R300_GB_SUPER_SIZE_128 (7<<6)
441 # define R300_GB_SUPER_X_SHIFT 9 /* 3 bits wide */
442 # define R300_GB_SUPER_Y_SHIFT 12 /* 3 bits wide */
443 # define R300_GB_SUPER_TILE_A 0
444 # define R300_GB_SUPER_TILE_B (1<<15)
445 # define R300_GB_SUBPIXEL_1_12 0
446 # define R300_GB_SUBPIXEL_1_16 (1<<16)
447
448 #define R300_GB_FIFO_SIZE 0x4024
449 /* each of the following is 2 bits wide */
450 #define R300_GB_FIFO_SIZE_32 0
451 #define R300_GB_FIFO_SIZE_64 1
452 #define R300_GB_FIFO_SIZE_128 2
453 #define R300_GB_FIFO_SIZE_256 3
454 # define R300_SC_IFIFO_SIZE_SHIFT 0
455 # define R300_SC_TZFIFO_SIZE_SHIFT 2
456 # define R300_SC_BFIFO_SIZE_SHIFT 4
457
458 # define R300_US_OFIFO_SIZE_SHIFT 12
459 # define R300_US_WFIFO_SIZE_SHIFT 14
460 /* the following use the same constants as above, but meaning is
461 is times 2 (i.e. instead of 32 words it means 64 */
462 # define R300_RS_TFIFO_SIZE_SHIFT 6
463 # define R300_RS_CFIFO_SIZE_SHIFT 8
464 # define R300_US_RAM_SIZE_SHIFT 10
465 /* watermarks, 3 bits wide */
466 # define R300_RS_HIGHWATER_COL_SHIFT 16
467 # define R300_RS_HIGHWATER_TEX_SHIFT 19
468 # define R300_OFIFO_HIGHWATER_SHIFT 22 /* two bits only */
469 # define R300_CUBE_FIFO_HIGHWATER_COL_SHIFT 24
470
471 #define R300_GB_SELECT 0x401C
472 # define R300_GB_FOG_SELECT_C0A 0
473 # define R300_GB_FOG_SELECT_C1A 1
474 # define R300_GB_FOG_SELECT_C2A 2
475 # define R300_GB_FOG_SELECT_C3A 3
476 # define R300_GB_FOG_SELECT_1_1_W 4
477 # define R300_GB_FOG_SELECT_Z 5
478 # define R300_GB_DEPTH_SELECT_Z 0
479 # define R300_GB_DEPTH_SELECT_1_1_W (1<<3)
480 # define R300_GB_W_SELECT_1_W 0
481 # define R300_GB_W_SELECT_1 (1<<4)
482
483 #define R300_GB_AA_CONFIG 0x4020
484 # define R300_AA_ENABLE 0x01
485 # define R300_AA_SUBSAMPLES_2 0
486 # define R300_AA_SUBSAMPLES_3 (1<<1)
487 # define R300_AA_SUBSAMPLES_4 (2<<1)
488 # define R300_AA_SUBSAMPLES_6 (3<<1)
489
490 /* gap */
491
492 /* Zero to flush caches. */
493 #define R300_TX_CNTL 0x4100
494
495 /* The upper enable bits are guessed, based on fglrx reported limits. */
496 #define R300_TX_ENABLE 0x4104
497 # define R300_TX_ENABLE_0 (1 << 0)
498 # define R300_TX_ENABLE_1 (1 << 1)
499 # define R300_TX_ENABLE_2 (1 << 2)
500 # define R300_TX_ENABLE_3 (1 << 3)
501 # define R300_TX_ENABLE_4 (1 << 4)
502 # define R300_TX_ENABLE_5 (1 << 5)
503 # define R300_TX_ENABLE_6 (1 << 6)
504 # define R300_TX_ENABLE_7 (1 << 7)
505 # define R300_TX_ENABLE_8 (1 << 8)
506 # define R300_TX_ENABLE_9 (1 << 9)
507 # define R300_TX_ENABLE_10 (1 << 10)
508 # define R300_TX_ENABLE_11 (1 << 11)
509 # define R300_TX_ENABLE_12 (1 << 12)
510 # define R300_TX_ENABLE_13 (1 << 13)
511 # define R300_TX_ENABLE_14 (1 << 14)
512 # define R300_TX_ENABLE_15 (1 << 15)
513
514 /* The pointsize is given in multiples of 6. The pointsize can be
515 * enormous: Clear() renders a single point that fills the entire
516 * framebuffer.
517 */
518 #define R300_RE_POINTSIZE 0x421C
519 # define R300_POINTSIZE_Y_SHIFT 0
520 # define R300_POINTSIZE_Y_MASK (0xFFFF << 0) /* GUESS */
521 # define R300_POINTSIZE_X_SHIFT 16
522 # define R300_POINTSIZE_X_MASK (0xFFFF << 16) /* GUESS */
523 # define R300_POINTSIZE_MAX (R300_POINTSIZE_Y_MASK / 6)
524
525 /* The line width is given in multiples of 6.
526 * In default mode lines are classified as vertical lines.
527 * HO: horizontal
528 * VE: vertical or horizontal
529 * HO & VE: no classification
530 */
531 #define R300_RE_LINE_CNT 0x4234
532 # define R300_LINESIZE_SHIFT 0
533 # define R300_LINESIZE_MASK (0xFFFF << 0) /* GUESS */
534 # define R300_LINESIZE_MAX (R300_LINESIZE_MASK / 6)
535 # define R300_LINE_CNT_HO (1 << 16)
536 # define R300_LINE_CNT_VE (1 << 17)
537
538 /* Some sort of scale or clamp value for texcoordless textures. */
539 #define R300_RE_UNK4238 0x4238
540
541 #define R300_RE_SHADE_MODEL 0x4278
542 # define R300_RE_SHADE_MODEL_SMOOTH 0x3aaaa
543 # define R300_RE_SHADE_MODEL_FLAT 0x39595
544
545 /* Dangerous */
546 #define R300_RE_POLYGON_MODE 0x4288
547 # define R300_PM_ENABLED (1 << 0)
548 # define R300_PM_FRONT_POINT (0 << 0)
549 # define R300_PM_BACK_POINT (0 << 0)
550 # define R300_PM_FRONT_LINE (1 << 4)
551 # define R300_PM_FRONT_FILL (1 << 5)
552 # define R300_PM_BACK_LINE (1 << 7)
553 # define R300_PM_BACK_FILL (1 << 8)
554
555 /* Fog parameters */
556 #define R300_RE_FOG_SCALE 0x4294
557 #define R300_RE_FOG_START 0x4298
558
559 /* Not sure why there are duplicate of factor and constant values.
560 * My best guess so far is that there are seperate zbiases for test and write.
561 * Ordering might be wrong.
562 * Some of the tests indicate that fgl has a fallback implementation of zbias
563 * via pixel shaders.
564 */
565 #define R300_RE_ZBIAS_T_FACTOR 0x42A4
566 #define R300_RE_ZBIAS_T_CONSTANT 0x42A8
567 #define R300_RE_ZBIAS_W_FACTOR 0x42AC
568 #define R300_RE_ZBIAS_W_CONSTANT 0x42B0
569
570 /* This register needs to be set to (1<<1) for RV350 to correctly
571 * perform depth test (see --vb-triangles in r300_demo)
572 * Don't know about other chips. - Vladimir
573 * This is set to 3 when GL_POLYGON_OFFSET_FILL is on.
574 * My guess is that there are two bits for each zbias primitive
575 * (FILL, LINE, POINT).
576 * One to enable depth test and one for depth write.
577 * Yet this doesnt explain why depth writes work ...
578 */
579 #define R300_RE_OCCLUSION_CNTL 0x42B4
580 # define R300_OCCLUSION_ON (1<<1)
581
582 #define R300_RE_CULL_CNTL 0x42B8
583 # define R300_CULL_FRONT (1 << 0)
584 # define R300_CULL_BACK (1 << 1)
585 # define R300_FRONT_FACE_CCW (0 << 2)
586 # define R300_FRONT_FACE_CW (1 << 2)
587
588
589 /* BEGIN: Rasterization / Interpolators - many guesses */
590
591 /* 0_UNKNOWN_18 has always been set except for clear operations.
592 * TC_CNT is the number of incoming texture coordinate sets (i.e. it depends
593 * on the vertex program, *not* the fragment program)
594 */
595 #define R300_RS_CNTL_0 0x4300
596 # define R300_RS_CNTL_TC_CNT_SHIFT 2
597 # define R300_RS_CNTL_TC_CNT_MASK (7 << 2)
598 /* number of color interpolators used */
599 # define R300_RS_CNTL_CI_CNT_SHIFT 7
600 # define R300_RS_CNTL_0_UNKNOWN_18 (1 << 18)
601 /* Guess: RS_CNTL_1 holds the index of the highest used RS_ROUTE_n
602 register. */
603 #define R300_RS_CNTL_1 0x4304
604
605 /* gap */
606
607 /* Only used for texture coordinates.
608 * Use the source field to route texture coordinate input from the
609 * vertex program to the desired interpolator. Note that the source
610 * field is relative to the outputs the vertex program *actually*
611 * writes. If a vertex program only writes texcoord[1], this will
612 * be source index 0.
613 * Set INTERP_USED on all interpolators that produce data used by
614 * the fragment program. INTERP_USED looks like a swizzling mask,
615 * but I haven't seen it used that way.
616 *
617 * Note: The _UNKNOWN constants are always set in their respective
618 * register. I don't know if this is necessary.
619 */
620 #define R300_RS_INTERP_0 0x4310
621 #define R300_RS_INTERP_1 0x4314
622 # define R300_RS_INTERP_1_UNKNOWN 0x40
623 #define R300_RS_INTERP_2 0x4318
624 # define R300_RS_INTERP_2_UNKNOWN 0x80
625 #define R300_RS_INTERP_3 0x431C
626 # define R300_RS_INTERP_3_UNKNOWN 0xC0
627 #define R300_RS_INTERP_4 0x4320
628 #define R300_RS_INTERP_5 0x4324
629 #define R300_RS_INTERP_6 0x4328
630 #define R300_RS_INTERP_7 0x432C
631 # define R300_RS_INTERP_SRC_SHIFT 2
632 # define R300_RS_INTERP_SRC_MASK (7 << 2)
633 # define R300_RS_INTERP_USED 0x00D10000
634
635 /* These DWORDs control how vertex data is routed into fragment program
636 * registers, after interpolators.
637 */
638 #define R300_RS_ROUTE_0 0x4330
639 #define R300_RS_ROUTE_1 0x4334
640 #define R300_RS_ROUTE_2 0x4338
641 #define R300_RS_ROUTE_3 0x433C /* GUESS */
642 #define R300_RS_ROUTE_4 0x4340 /* GUESS */
643 #define R300_RS_ROUTE_5 0x4344 /* GUESS */
644 #define R300_RS_ROUTE_6 0x4348 /* GUESS */
645 #define R300_RS_ROUTE_7 0x434C /* GUESS */
646 # define R300_RS_ROUTE_SOURCE_INTERP_0 0
647 # define R300_RS_ROUTE_SOURCE_INTERP_1 1
648 # define R300_RS_ROUTE_SOURCE_INTERP_2 2
649 # define R300_RS_ROUTE_SOURCE_INTERP_3 3
650 # define R300_RS_ROUTE_SOURCE_INTERP_4 4
651 # define R300_RS_ROUTE_SOURCE_INTERP_5 5 /* GUESS */
652 # define R300_RS_ROUTE_SOURCE_INTERP_6 6 /* GUESS */
653 # define R300_RS_ROUTE_SOURCE_INTERP_7 7 /* GUESS */
654 # define R300_RS_ROUTE_ENABLE (1 << 3) /* GUESS */
655 # define R300_RS_ROUTE_DEST_SHIFT 6
656 # define R300_RS_ROUTE_DEST_MASK (31 << 6) /* GUESS */
657
658 /* Special handling for color: When the fragment program uses color,
659 * the ROUTE_0_COLOR bit is set and ROUTE_0_COLOR_DEST contains the
660 * color register index.
661 */
662 # define R300_RS_ROUTE_0_COLOR (1 << 14)
663 # define R300_RS_ROUTE_0_COLOR_DEST_SHIFT 17
664 # define R300_RS_ROUTE_0_COLOR_DEST_MASK (31 << 17) /* GUESS */
665 /* As above, but for secondary color */
666 # define R300_RS_ROUTE_1_COLOR1 (1 << 14)
667 # define R300_RS_ROUTE_1_COLOR1_DEST_SHIFT 17
668 # define R300_RS_ROUTE_1_COLOR1_DEST_MASK (31 << 17)
669 # define R300_RS_ROUTE_1_UNKNOWN11 (1 << 11)
670 /* END: Rasterization / Interpolators - many guesses */
671
672 /* BEGIN: Scissors and cliprects */
673
674 /* There are four clipping rectangles. Their corner coordinates are inclusive.
675 * Every pixel is assigned a number from 0 and 15 by setting bits 0-3 depending
676 * on whether the pixel is inside cliprects 0-3, respectively. For example,
677 * if a pixel is inside cliprects 0 and 1, but outside 2 and 3, it is assigned
678 * the number 3 (binary 0011).
679 * Iff the bit corresponding to the pixel's number in RE_CLIPRECT_CNTL is set,
680 * the pixel is rasterized.
681 *
682 * In addition to this, there is a scissors rectangle. Only pixels inside the
683 * scissors rectangle are drawn. (coordinates are inclusive)
684 *
685 * For some reason, the top-left corner of the framebuffer is at (1440, 1440)
686 * for the purpose of clipping and scissors.
687 */
688 #define R300_RE_CLIPRECT_TL_0 0x43B0
689 #define R300_RE_CLIPRECT_BR_0 0x43B4
690 #define R300_RE_CLIPRECT_TL_1 0x43B8
691 #define R300_RE_CLIPRECT_BR_1 0x43BC
692 #define R300_RE_CLIPRECT_TL_2 0x43C0
693 #define R300_RE_CLIPRECT_BR_2 0x43C4
694 #define R300_RE_CLIPRECT_TL_3 0x43C8
695 #define R300_RE_CLIPRECT_BR_3 0x43CC
696 # define R300_CLIPRECT_OFFSET 1440
697 # define R300_CLIPRECT_MASK 0x1FFF
698 # define R300_CLIPRECT_X_SHIFT 0
699 # define R300_CLIPRECT_X_MASK (0x1FFF << 0)
700 # define R300_CLIPRECT_Y_SHIFT 13
701 # define R300_CLIPRECT_Y_MASK (0x1FFF << 13)
702 #define R300_RE_CLIPRECT_CNTL 0x43D0
703 # define R300_CLIP_OUT (1 << 0)
704 # define R300_CLIP_0 (1 << 1)
705 # define R300_CLIP_1 (1 << 2)
706 # define R300_CLIP_10 (1 << 3)
707 # define R300_CLIP_2 (1 << 4)
708 # define R300_CLIP_20 (1 << 5)
709 # define R300_CLIP_21 (1 << 6)
710 # define R300_CLIP_210 (1 << 7)
711 # define R300_CLIP_3 (1 << 8)
712 # define R300_CLIP_30 (1 << 9)
713 # define R300_CLIP_31 (1 << 10)
714 # define R300_CLIP_310 (1 << 11)
715 # define R300_CLIP_32 (1 << 12)
716 # define R300_CLIP_320 (1 << 13)
717 # define R300_CLIP_321 (1 << 14)
718 # define R300_CLIP_3210 (1 << 15)
719
720 /* gap */
721
722 #define R300_RE_SCISSORS_TL 0x43E0
723 #define R300_RE_SCISSORS_BR 0x43E4
724 # define R300_SCISSORS_OFFSET 1440
725 # define R300_SCISSORS_X_SHIFT 0
726 # define R300_SCISSORS_X_MASK (0x1FFF << 0)
727 # define R300_SCISSORS_Y_SHIFT 13
728 # define R300_SCISSORS_Y_MASK (0x1FFF << 13)
729 /* END: Scissors and cliprects */
730
731 /* BEGIN: Texture specification */
732
733 /*
734 * The texture specification dwords are grouped by meaning and not by texture
735 * unit. This means that e.g. the offset for texture image unit N is found in
736 * register TX_OFFSET_0 + (4*N)
737 */
738 #define R300_TX_FILTER_0 0x4400
739 # define R300_TX_REPEAT 0
740 # define R300_TX_MIRRORED 1
741 # define R300_TX_CLAMP 4
742 # define R300_TX_CLAMP_TO_EDGE 2
743 # define R300_TX_CLAMP_TO_BORDER 6
744 # define R300_TX_WRAP_S_SHIFT 0
745 # define R300_TX_WRAP_S_MASK (7 << 0)
746 # define R300_TX_WRAP_T_SHIFT 3
747 # define R300_TX_WRAP_T_MASK (7 << 3)
748 # define R300_TX_WRAP_Q_SHIFT 6
749 # define R300_TX_WRAP_Q_MASK (7 << 6)
750 # define R300_TX_MAG_FILTER_NEAREST (1 << 9)
751 # define R300_TX_MAG_FILTER_LINEAR (2 << 9)
752 # define R300_TX_MAG_FILTER_MASK (3 << 9)
753 # define R300_TX_MIN_FILTER_NEAREST (1 << 11)
754 # define R300_TX_MIN_FILTER_LINEAR (2 << 11)
755 # define R300_TX_MIN_FILTER_NEAREST_MIP_NEAREST (5 << 11)
756 # define R300_TX_MIN_FILTER_NEAREST_MIP_LINEAR (9 << 11)
757 # define R300_TX_MIN_FILTER_LINEAR_MIP_NEAREST (6 << 11)
758 # define R300_TX_MIN_FILTER_LINEAR_MIP_LINEAR (10 << 11)
759
760 /* NOTE: NEAREST doesnt seem to exist.
761 * Im not seting MAG_FILTER_MASK and (3 << 11) on for all
762 * anisotropy modes because that would void selected mag filter
763 */
764 # define R300_TX_MIN_FILTER_ANISO_NEAREST (0 << 13)
765 # define R300_TX_MIN_FILTER_ANISO_LINEAR (0 << 13)
766 # define R300_TX_MIN_FILTER_ANISO_NEAREST_MIP_NEAREST (1 << 13)
767 # define R300_TX_MIN_FILTER_ANISO_NEAREST_MIP_LINEAR (2 << 13)
768 # define R300_TX_MIN_FILTER_MASK ( (15 << 11) | (3 << 13) )
769 # define R300_TX_MAX_ANISO_1_TO_1 (0 << 21)
770 # define R300_TX_MAX_ANISO_2_TO_1 (2 << 21)
771 # define R300_TX_MAX_ANISO_4_TO_1 (4 << 21)
772 # define R300_TX_MAX_ANISO_8_TO_1 (6 << 21)
773 # define R300_TX_MAX_ANISO_16_TO_1 (8 << 21)
774 # define R300_TX_MAX_ANISO_MASK (14 << 21)
775
776 #define R300_TX_FILTER1_0 0x4440
777 # define R300_CHROMA_KEY_MODE_DISABLE 0
778 # define R300_CHROMA_KEY_FORCE 1
779 # define R300_CHROMA_KEY_BLEND 2
780 # define R300_MC_ROUND_NORMAL (0<<2)
781 # define R300_MC_ROUND_MPEG4 (1<<2)
782 # define R300_LOD_BIAS_MASK 0x1fff
783 # define R300_EDGE_ANISO_EDGE_DIAG (0<<13)
784 # define R300_EDGE_ANISO_EDGE_ONLY (1<<13)
785 # define R300_MC_COORD_TRUNCATE_DISABLE (0<<14)
786 # define R300_MC_COORD_TRUNCATE_MPEG (1<<14)
787 # define R300_TX_TRI_PERF_0_8 (0<<15)
788 # define R300_TX_TRI_PERF_1_8 (1<<15)
789 # define R300_TX_TRI_PERF_1_4 (2<<15)
790 # define R300_TX_TRI_PERF_3_8 (3<<15)
791 # define R300_ANISO_THRESHOLD_MASK (7<<17)
792
793 #define R300_TX_SIZE_0 0x4480
794 # define R300_TX_WIDTHMASK_SHIFT 0
795 # define R300_TX_WIDTHMASK_MASK (2047 << 0)
796 # define R300_TX_HEIGHTMASK_SHIFT 11
797 # define R300_TX_HEIGHTMASK_MASK (2047 << 11)
798 # define R300_TX_UNK23 (1 << 23)
799 # define R300_TX_MAX_MIP_LEVEL_SHIFT 26
800 # define R300_TX_MAX_MIP_LEVEL_MASK (0xf << 26)
801 # define R300_TX_SIZE_PROJECTED (1<<30)
802 # define R300_TX_SIZE_TXPITCH_EN (1<<31)
803 #define R300_TX_FORMAT_0 0x44C0
804 /* The interpretation of the format word by Wladimir van der Laan */
805 /* The X, Y, Z and W refer to the layout of the components.
806 They are given meanings as R, G, B and Alpha by the swizzle
807 specification */
808 # define R300_TX_FORMAT_X8 0x0
809 # define R300_TX_FORMAT_X16 0x1
810 # define R300_TX_FORMAT_Y4X4 0x2
811 # define R300_TX_FORMAT_Y8X8 0x3
812 # define R300_TX_FORMAT_Y16X16 0x4
813 # define R300_TX_FORMAT_Z3Y3X2 0x5
814 # define R300_TX_FORMAT_Z5Y6X5 0x6
815 # define R300_TX_FORMAT_Z6Y5X5 0x7
816 # define R300_TX_FORMAT_Z11Y11X10 0x8
817 # define R300_TX_FORMAT_Z10Y11X11 0x9
818 # define R300_TX_FORMAT_W4Z4Y4X4 0xA
819 # define R300_TX_FORMAT_W1Z5Y5X5 0xB
820 # define R300_TX_FORMAT_W8Z8Y8X8 0xC
821 # define R300_TX_FORMAT_W2Z10Y10X10 0xD
822 # define R300_TX_FORMAT_W16Z16Y16X16 0xE
823 # define R300_TX_FORMAT_DXT1 0xF
824 # define R300_TX_FORMAT_DXT3 0x10
825 # define R300_TX_FORMAT_DXT5 0x11
826 # define R300_TX_FORMAT_D3DMFT_CxV8U8 0x12 /* no swizzle */
827 # define R300_TX_FORMAT_A8R8G8B8 0x13 /* no swizzle */
828 # define R300_TX_FORMAT_B8G8_B8G8 0x14 /* no swizzle */
829 # define R300_TX_FORMAT_G8R8_G8B8 0x15 /* no swizzle */
830 /* 0x16 - some 16 bit green format.. ?? */
831 # define R300_TX_FORMAT_UNK25 (1 << 25) /* no swizzle */
832 # define R300_TX_FORMAT_CUBIC_MAP (1 << 26)
833
834 /* gap */
835 /* Floating point formats */
836 /* Note - hardware supports both 16 and 32 bit floating point */
837 # define R300_TX_FORMAT_FL_I16 0x18
838 # define R300_TX_FORMAT_FL_I16A16 0x19
839 # define R300_TX_FORMAT_FL_R16G16B16A16 0x1A
840 # define R300_TX_FORMAT_FL_I32 0x1B
841 # define R300_TX_FORMAT_FL_I32A32 0x1C
842 # define R300_TX_FORMAT_FL_R32G32B32A32 0x1D
843 /* alpha modes, convenience mostly */
844 /* if you have alpha, pick constant appropriate to the
845 number of channels (1 for I8, 2 for I8A8, 4 for R8G8B8A8, etc */
846 # define R300_TX_FORMAT_ALPHA_1CH 0x000
847 # define R300_TX_FORMAT_ALPHA_2CH 0x200
848 # define R300_TX_FORMAT_ALPHA_4CH 0x600
849 # define R300_TX_FORMAT_ALPHA_NONE 0xA00
850 /* Swizzling */
851 /* constants */
852 # define R300_TX_FORMAT_X 0
853 # define R300_TX_FORMAT_Y 1
854 # define R300_TX_FORMAT_Z 2
855 # define R300_TX_FORMAT_W 3
856 # define R300_TX_FORMAT_ZERO 4
857 # define R300_TX_FORMAT_ONE 5
858 /* 2.0*Z, everything above 1.0 is set to 0.0 */
859 # define R300_TX_FORMAT_CUT_Z 6
860 /* 2.0*W, everything above 1.0 is set to 0.0 */
861 # define R300_TX_FORMAT_CUT_W 7
862
863 # define R300_TX_FORMAT_B_SHIFT 18
864 # define R300_TX_FORMAT_G_SHIFT 15
865 # define R300_TX_FORMAT_R_SHIFT 12
866 # define R300_TX_FORMAT_A_SHIFT 9
867 /* Convenience macro to take care of layout and swizzling */
868 # define R300_EASY_TX_FORMAT(B, G, R, A, FMT) ( \
869 ((R300_TX_FORMAT_##B)<<R300_TX_FORMAT_B_SHIFT) \
870 | ((R300_TX_FORMAT_##G)<<R300_TX_FORMAT_G_SHIFT) \
871 | ((R300_TX_FORMAT_##R)<<R300_TX_FORMAT_R_SHIFT) \
872 | ((R300_TX_FORMAT_##A)<<R300_TX_FORMAT_A_SHIFT) \
873 | (R300_TX_FORMAT_##FMT) \
874 )
875 /* These can be ORed with result of R300_EASY_TX_FORMAT()
876 We don't really know what they do. Take values from a
877 constant color ? */
878 # define R300_TX_FORMAT_CONST_X (1<<5)
879 # define R300_TX_FORMAT_CONST_Y (2<<5)
880 # define R300_TX_FORMAT_CONST_Z (4<<5)
881 # define R300_TX_FORMAT_CONST_W (8<<5)
882
883 # define R300_TX_FORMAT_YUV_MODE 0x00800000
884
885 #define R300_TX_PITCH_0 0x4500 /* obvious missing in gap */
886 #define R300_TX_OFFSET_0 0x4540
887 /* BEGIN: Guess from R200 */
888 # define R300_TXO_ENDIAN_NO_SWAP (0 << 0)
889 # define R300_TXO_ENDIAN_BYTE_SWAP (1 << 0)
890 # define R300_TXO_ENDIAN_WORD_SWAP (2 << 0)
891 # define R300_TXO_ENDIAN_HALFDW_SWAP (3 << 0)
892 # define R300_TXO_MACRO_TILE (1 << 2)
893 # define R300_TXO_MICRO_TILE (1 << 3)
894 # define R300_TXO_OFFSET_MASK 0xffffffe0
895 # define R300_TXO_OFFSET_SHIFT 5
896 /* END: Guess from R200 */
897
898 /* 32 bit chroma key */
899 #define R300_TX_CHROMA_KEY_0 0x4580
900 /* ff00ff00 == { 0, 1.0, 0, 1.0 } */
901 #define R300_TX_BORDER_COLOR_0 0x45C0
902
903 /* END: Texture specification */
904
905 /* BEGIN: Fragment program instruction set */
906
907 /* Fragment programs are written directly into register space.
908 * There are separate instruction streams for texture instructions and ALU
909 * instructions.
910 * In order to synchronize these streams, the program is divided into up
911 * to 4 nodes. Each node begins with a number of TEX operations, followed
912 * by a number of ALU operations.
913 * The first node can have zero TEX ops, all subsequent nodes must have at
914 * least
915 * one TEX ops.
916 * All nodes must have at least one ALU op.
917 *
918 * The index of the last node is stored in PFS_CNTL_0: A value of 0 means
919 * 1 node, a value of 3 means 4 nodes.
920 * The total amount of instructions is defined in PFS_CNTL_2. The offsets are
921 * offsets into the respective instruction streams, while *_END points to the
922 * last instruction relative to this offset.
923 */
924 #define R300_PFS_CNTL_0 0x4600
925 # define R300_PFS_CNTL_LAST_NODES_SHIFT 0
926 # define R300_PFS_CNTL_LAST_NODES_MASK (3 << 0)
927 # define R300_PFS_CNTL_FIRST_NODE_HAS_TEX (1 << 3)
928 #define R300_PFS_CNTL_1 0x4604
929 /* There is an unshifted value here which has so far always been equal to the
930 * index of the highest used temporary register.
931 */
932 #define R300_PFS_CNTL_2 0x4608
933 # define R300_PFS_CNTL_ALU_OFFSET_SHIFT 0
934 # define R300_PFS_CNTL_ALU_OFFSET_MASK (63 << 0)
935 # define R300_PFS_CNTL_ALU_END_SHIFT 6
936 # define R300_PFS_CNTL_ALU_END_MASK (63 << 6)
937 # define R300_PFS_CNTL_TEX_OFFSET_SHIFT 12
938 # define R300_PFS_CNTL_TEX_OFFSET_MASK (31 << 12) /* GUESS */
939 # define R300_PFS_CNTL_TEX_END_SHIFT 18
940 # define R300_PFS_CNTL_TEX_END_MASK (31 << 18) /* GUESS */
941
942 /* gap */
943
944 /* Nodes are stored backwards. The last active node is always stored in
945 * PFS_NODE_3.
946 * Example: In a 2-node program, NODE_0 and NODE_1 are set to 0. The
947 * first node is stored in NODE_2, the second node is stored in NODE_3.
948 *
949 * Offsets are relative to the master offset from PFS_CNTL_2.
950 * LAST_NODE is set for the last node, and only for the last node.
951 */
952 #define R300_PFS_NODE_0 0x4610
953 #define R300_PFS_NODE_1 0x4614
954 #define R300_PFS_NODE_2 0x4618
955 #define R300_PFS_NODE_3 0x461C
956 # define R300_PFS_NODE_ALU_OFFSET_SHIFT 0
957 # define R300_PFS_NODE_ALU_OFFSET_MASK (63 << 0)
958 # define R300_PFS_NODE_ALU_END_SHIFT 6
959 # define R300_PFS_NODE_ALU_END_MASK (63 << 6)
960 # define R300_PFS_NODE_TEX_OFFSET_SHIFT 12
961 # define R300_PFS_NODE_TEX_OFFSET_MASK (31 << 12)
962 # define R300_PFS_NODE_TEX_END_SHIFT 17
963 # define R300_PFS_NODE_TEX_END_MASK (31 << 17)
964 /*# define R300_PFS_NODE_LAST_NODE (1 << 22) */
965 # define R300_PFS_NODE_OUTPUT_COLOR (1 << 22)
966 # define R300_PFS_NODE_OUTPUT_DEPTH (1 << 23)
967
968 /* TEX
969 * As far as I can tell, texture instructions cannot write into output
970 * registers directly. A subsequent ALU instruction is always necessary,
971 * even if it's just MAD o0, r0, 1, 0
972 */
973 #define R300_PFS_TEXI_0 0x4620
974 # define R300_FPITX_SRC_SHIFT 0
975 # define R300_FPITX_SRC_MASK (31 << 0)
976 /* GUESS */
977 # define R300_FPITX_SRC_CONST (1 << 5)
978 # define R300_FPITX_DST_SHIFT 6
979 # define R300_FPITX_DST_MASK (31 << 6)
980 # define R300_FPITX_IMAGE_SHIFT 11
981 /* GUESS based on layout and native limits */
982 # define R300_FPITX_IMAGE_MASK (15 << 11)
983 /* Unsure if these are opcodes, or some kind of bitfield, but this is how
984 * they were set when I checked
985 */
986 # define R300_FPITX_OPCODE_SHIFT 15
987 # define R300_FPITX_OP_TEX 1
988 # define R300_FPITX_OP_KIL 2
989 # define R300_FPITX_OP_TXP 3
990 # define R300_FPITX_OP_TXB 4
991
992 /* ALU
993 * The ALU instructions register blocks are enumerated according to the order
994 * in which fglrx. I assume there is space for 64 instructions, since
995 * each block has space for a maximum of 64 DWORDs, and this matches reported
996 * native limits.
997 *
998 * The basic functional block seems to be one MAD for each color and alpha,
999 * and an adder that adds all components after the MUL.
1000 * - ADD, MUL, MAD etc.: use MAD with appropriate neutral operands
1001 * - DP4: Use OUTC_DP4, OUTA_DP4
1002 * - DP3: Use OUTC_DP3, OUTA_DP4, appropriate alpha operands
1003 * - DPH: Use OUTC_DP4, OUTA_DP4, appropriate alpha operands
1004 * - CMPH: If ARG2 > 0.5, return ARG0, else return ARG1
1005 * - CMP: If ARG2 < 0, return ARG1, else return ARG0
1006 * - FLR: use FRC+MAD
1007 * - XPD: use MAD+MAD
1008 * - SGE, SLT: use MAD+CMP
1009 * - RSQ: use ABS modifier for argument
1010 * - Use OUTC_REPL_ALPHA to write results of an alpha-only operation
1011 * (e.g. RCP) into color register
1012 * - apparently, there's no quick DST operation
1013 * - fglrx set FPI2_UNKNOWN_31 on a "MAD fragment.color, tmp0, tmp1, tmp2"
1014 * - fglrx set FPI2_UNKNOWN_31 on a "MAX r2, r1, c0"
1015 * - fglrx once set FPI0_UNKNOWN_31 on a "FRC r1, r1"
1016 *
1017 * Operand selection
1018 * First stage selects three sources from the available registers and
1019 * constant parameters. This is defined in INSTR1 (color) and INSTR3 (alpha).
1020 * fglrx sorts the three source fields: Registers before constants,
1021 * lower indices before higher indices; I do not know whether this is
1022 * necessary.
1023 *
1024 * fglrx fills unused sources with "read constant 0"
1025 * According to specs, you cannot select more than two different constants.
1026 *
1027 * Second stage selects the operands from the sources. This is defined in
1028 * INSTR0 (color) and INSTR2 (alpha). You can also select the special constants
1029 * zero and one.
1030 * Swizzling and negation happens in this stage, as well.
1031 *
1032 * Important: Color and alpha seem to be mostly separate, i.e. their sources
1033 * selection appears to be fully independent (the register storage is probably
1034 * physically split into a color and an alpha section).
1035 * However (because of the apparent physical split), there is some interaction
1036 * WRT swizzling. If, for example, you want to load an R component into an
1037 * Alpha operand, this R component is taken from a *color* source, not from
1038 * an alpha source. The corresponding register doesn't even have to appear in
1039 * the alpha sources list. (I hope this alll makes sense to you)
1040 *
1041 * Destination selection
1042 * The destination register index is in FPI1 (color) and FPI3 (alpha)
1043 * together with enable bits.
1044 * There are separate enable bits for writing into temporary registers
1045 * (DSTC_REG_* /DSTA_REG) and and program output registers (DSTC_OUTPUT_*
1046 * /DSTA_OUTPUT). You can write to both at once, or not write at all (the
1047 * same index must be used for both).
1048 *
1049 * Note: There is a special form for LRP
1050 * - Argument order is the same as in ARB_fragment_program.
1051 * - Operation is MAD
1052 * - ARG1 is set to ARGC_SRC1C_LRP/ARGC_SRC1A_LRP
1053 * - Set FPI0/FPI2_SPECIAL_LRP
1054 * Arbitrary LRP (including support for swizzling) requires vanilla MAD+MAD
1055 */
1056 #define R300_PFS_INSTR1_0 0x46C0
1057 # define R300_FPI1_SRC0C_SHIFT 0
1058 # define R300_FPI1_SRC0C_MASK (31 << 0)
1059 # define R300_FPI1_SRC0C_CONST (1 << 5)
1060 # define R300_FPI1_SRC1C_SHIFT 6
1061 # define R300_FPI1_SRC1C_MASK (31 << 6)
1062 # define R300_FPI1_SRC1C_CONST (1 << 11)
1063 # define R300_FPI1_SRC2C_SHIFT 12
1064 # define R300_FPI1_SRC2C_MASK (31 << 12)
1065 # define R300_FPI1_SRC2C_CONST (1 << 17)
1066 # define R300_FPI1_DSTC_SHIFT 18
1067 # define R300_FPI1_DSTC_MASK (31 << 18)
1068 # define R300_FPI1_DSTC_REG_MASK_SHIFT 23
1069 # define R300_FPI1_DSTC_REG_X (1 << 23)
1070 # define R300_FPI1_DSTC_REG_Y (1 << 24)
1071 # define R300_FPI1_DSTC_REG_Z (1 << 25)
1072 # define R300_FPI1_DSTC_OUTPUT_MASK_SHIFT 26
1073 # define R300_FPI1_DSTC_OUTPUT_X (1 << 26)
1074 # define R300_FPI1_DSTC_OUTPUT_Y (1 << 27)
1075 # define R300_FPI1_DSTC_OUTPUT_Z (1 << 28)
1076
1077 #define R300_PFS_INSTR3_0 0x47C0
1078 # define R300_FPI3_SRC0A_SHIFT 0
1079 # define R300_FPI3_SRC0A_MASK (31 << 0)
1080 # define R300_FPI3_SRC0A_CONST (1 << 5)
1081 # define R300_FPI3_SRC1A_SHIFT 6
1082 # define R300_FPI3_SRC1A_MASK (31 << 6)
1083 # define R300_FPI3_SRC1A_CONST (1 << 11)
1084 # define R300_FPI3_SRC2A_SHIFT 12
1085 # define R300_FPI3_SRC2A_MASK (31 << 12)
1086 # define R300_FPI3_SRC2A_CONST (1 << 17)
1087 # define R300_FPI3_DSTA_SHIFT 18
1088 # define R300_FPI3_DSTA_MASK (31 << 18)
1089 # define R300_FPI3_DSTA_REG (1 << 23)
1090 # define R300_FPI3_DSTA_OUTPUT (1 << 24)
1091 # define R300_FPI3_DSTA_DEPTH (1 << 27)
1092
1093 #define R300_PFS_INSTR0_0 0x48C0
1094 # define R300_FPI0_ARGC_SRC0C_XYZ 0
1095 # define R300_FPI0_ARGC_SRC0C_XXX 1
1096 # define R300_FPI0_ARGC_SRC0C_YYY 2
1097 # define R300_FPI0_ARGC_SRC0C_ZZZ 3
1098 # define R300_FPI0_ARGC_SRC1C_XYZ 4
1099 # define R300_FPI0_ARGC_SRC1C_XXX 5
1100 # define R300_FPI0_ARGC_SRC1C_YYY 6
1101 # define R300_FPI0_ARGC_SRC1C_ZZZ 7
1102 # define R300_FPI0_ARGC_SRC2C_XYZ 8
1103 # define R300_FPI0_ARGC_SRC2C_XXX 9
1104 # define R300_FPI0_ARGC_SRC2C_YYY 10
1105 # define R300_FPI0_ARGC_SRC2C_ZZZ 11
1106 # define R300_FPI0_ARGC_SRC0A 12
1107 # define R300_FPI0_ARGC_SRC1A 13
1108 # define R300_FPI0_ARGC_SRC2A 14
1109 # define R300_FPI0_ARGC_SRC1C_LRP 15
1110 # define R300_FPI0_ARGC_ZERO 20
1111 # define R300_FPI0_ARGC_ONE 21
1112 /* GUESS */
1113 # define R300_FPI0_ARGC_HALF 22
1114 # define R300_FPI0_ARGC_SRC0C_YZX 23
1115 # define R300_FPI0_ARGC_SRC1C_YZX 24
1116 # define R300_FPI0_ARGC_SRC2C_YZX 25
1117 # define R300_FPI0_ARGC_SRC0C_ZXY 26
1118 # define R300_FPI0_ARGC_SRC1C_ZXY 27
1119 # define R300_FPI0_ARGC_SRC2C_ZXY 28
1120 # define R300_FPI0_ARGC_SRC0CA_WZY 29
1121 # define R300_FPI0_ARGC_SRC1CA_WZY 30
1122 # define R300_FPI0_ARGC_SRC2CA_WZY 31
1123
1124 # define R300_FPI0_ARG0C_SHIFT 0
1125 # define R300_FPI0_ARG0C_MASK (31 << 0)
1126 # define R300_FPI0_ARG0C_NEG (1 << 5)
1127 # define R300_FPI0_ARG0C_ABS (1 << 6)
1128 # define R300_FPI0_ARG1C_SHIFT 7
1129 # define R300_FPI0_ARG1C_MASK (31 << 7)
1130 # define R300_FPI0_ARG1C_NEG (1 << 12)
1131 # define R300_FPI0_ARG1C_ABS (1 << 13)
1132 # define R300_FPI0_ARG2C_SHIFT 14
1133 # define R300_FPI0_ARG2C_MASK (31 << 14)
1134 # define R300_FPI0_ARG2C_NEG (1 << 19)
1135 # define R300_FPI0_ARG2C_ABS (1 << 20)
1136 # define R300_FPI0_SPECIAL_LRP (1 << 21)
1137 # define R300_FPI0_OUTC_MAD (0 << 23)
1138 # define R300_FPI0_OUTC_DP3 (1 << 23)
1139 # define R300_FPI0_OUTC_DP4 (2 << 23)
1140 # define R300_FPI0_OUTC_MIN (4 << 23)
1141 # define R300_FPI0_OUTC_MAX (5 << 23)
1142 # define R300_FPI0_OUTC_CMPH (7 << 23)
1143 # define R300_FPI0_OUTC_CMP (8 << 23)
1144 # define R300_FPI0_OUTC_FRC (9 << 23)
1145 # define R300_FPI0_OUTC_REPL_ALPHA (10 << 23)
1146 # define R300_FPI0_OUTC_SAT (1 << 30)
1147 # define R300_FPI0_INSERT_NOP (1 << 31)
1148
1149 #define R300_PFS_INSTR2_0 0x49C0
1150 # define R300_FPI2_ARGA_SRC0C_X 0
1151 # define R300_FPI2_ARGA_SRC0C_Y 1
1152 # define R300_FPI2_ARGA_SRC0C_Z 2
1153 # define R300_FPI2_ARGA_SRC1C_X 3
1154 # define R300_FPI2_ARGA_SRC1C_Y 4
1155 # define R300_FPI2_ARGA_SRC1C_Z 5
1156 # define R300_FPI2_ARGA_SRC2C_X 6
1157 # define R300_FPI2_ARGA_SRC2C_Y 7
1158 # define R300_FPI2_ARGA_SRC2C_Z 8
1159 # define R300_FPI2_ARGA_SRC0A 9
1160 # define R300_FPI2_ARGA_SRC1A 10
1161 # define R300_FPI2_ARGA_SRC2A 11
1162 # define R300_FPI2_ARGA_SRC1A_LRP 15
1163 # define R300_FPI2_ARGA_ZERO 16
1164 # define R300_FPI2_ARGA_ONE 17
1165 /* GUESS */
1166 # define R300_FPI2_ARGA_HALF 18
1167 # define R300_FPI2_ARG0A_SHIFT 0
1168 # define R300_FPI2_ARG0A_MASK (31 << 0)
1169 # define R300_FPI2_ARG0A_NEG (1 << 5)
1170 /* GUESS */
1171 # define R300_FPI2_ARG0A_ABS (1 << 6)
1172 # define R300_FPI2_ARG1A_SHIFT 7
1173 # define R300_FPI2_ARG1A_MASK (31 << 7)
1174 # define R300_FPI2_ARG1A_NEG (1 << 12)
1175 /* GUESS */
1176 # define R300_FPI2_ARG1A_ABS (1 << 13)
1177 # define R300_FPI2_ARG2A_SHIFT 14
1178 # define R300_FPI2_ARG2A_MASK (31 << 14)
1179 # define R300_FPI2_ARG2A_NEG (1 << 19)
1180 /* GUESS */
1181 # define R300_FPI2_ARG2A_ABS (1 << 20)
1182 # define R300_FPI2_SPECIAL_LRP (1 << 21)
1183 # define R300_FPI2_OUTA_MAD (0 << 23)
1184 # define R300_FPI2_OUTA_DP4 (1 << 23)
1185 # define R300_FPI2_OUTA_MIN (2 << 23)
1186 # define R300_FPI2_OUTA_MAX (3 << 23)
1187 # define R300_FPI2_OUTA_CMP (6 << 23)
1188 # define R300_FPI2_OUTA_FRC (7 << 23)
1189 # define R300_FPI2_OUTA_EX2 (8 << 23)
1190 # define R300_FPI2_OUTA_LG2 (9 << 23)
1191 # define R300_FPI2_OUTA_RCP (10 << 23)
1192 # define R300_FPI2_OUTA_RSQ (11 << 23)
1193 # define R300_FPI2_OUTA_SAT (1 << 30)
1194 # define R300_FPI2_UNKNOWN_31 (1 << 31)
1195 /* END: Fragment program instruction set */
1196
1197 /* Fog state and color */
1198 #define R300_RE_FOG_STATE 0x4BC0
1199 # define R300_FOG_ENABLE (1 << 0)
1200 # define R300_FOG_MODE_LINEAR (0 << 1)
1201 # define R300_FOG_MODE_EXP (1 << 1)
1202 # define R300_FOG_MODE_EXP2 (2 << 1)
1203 # define R300_FOG_MODE_MASK (3 << 1)
1204 #define R300_FOG_COLOR_R 0x4BC8
1205 #define R300_FOG_COLOR_G 0x4BCC
1206 #define R300_FOG_COLOR_B 0x4BD0
1207
1208 #define R300_PP_ALPHA_TEST 0x4BD4
1209 # define R300_REF_ALPHA_MASK 0x000000ff
1210 # define R300_ALPHA_TEST_FAIL (0 << 8)
1211 # define R300_ALPHA_TEST_LESS (1 << 8)
1212 # define R300_ALPHA_TEST_LEQUAL (3 << 8)
1213 # define R300_ALPHA_TEST_EQUAL (2 << 8)
1214 # define R300_ALPHA_TEST_GEQUAL (6 << 8)
1215 # define R300_ALPHA_TEST_GREATER (4 << 8)
1216 # define R300_ALPHA_TEST_NEQUAL (5 << 8)
1217 # define R300_ALPHA_TEST_PASS (7 << 8)
1218 # define R300_ALPHA_TEST_OP_MASK (7 << 8)
1219 # define R300_ALPHA_TEST_ENABLE (1 << 11)
1220
1221 /* gap */
1222
1223 /* Fragment program parameters in 7.16 floating point */
1224 #define R300_PFS_PARAM_0_X 0x4C00
1225 #define R300_PFS_PARAM_0_Y 0x4C04
1226 #define R300_PFS_PARAM_0_Z 0x4C08
1227 #define R300_PFS_PARAM_0_W 0x4C0C
1228 /* GUESS: PARAM_31 is last, based on native limits reported by fglrx */
1229 #define R300_PFS_PARAM_31_X 0x4DF0
1230 #define R300_PFS_PARAM_31_Y 0x4DF4
1231 #define R300_PFS_PARAM_31_Z 0x4DF8
1232 #define R300_PFS_PARAM_31_W 0x4DFC
1233
1234 /* Notes:
1235 * - AFAIK fglrx always sets BLEND_UNKNOWN when blending is used in
1236 * the application
1237 * - AFAIK fglrx always sets BLEND_NO_SEPARATE when CBLEND and ABLEND
1238 * are set to the same
1239 * function (both registers are always set up completely in any case)
1240 * - Most blend flags are simply copied from R200 and not tested yet
1241 */
1242 #define R300_RB3D_CBLEND 0x4E04
1243 #define R300_RB3D_ABLEND 0x4E08
1244 /* the following only appear in CBLEND */
1245 # define R300_BLEND_ENABLE (1 << 0)
1246 # define R300_BLEND_UNKNOWN (3 << 1)
1247 # define R300_BLEND_NO_SEPARATE (1 << 3)
1248 /* the following are shared between CBLEND and ABLEND */
1249 # define R300_FCN_MASK (3 << 12)
1250 # define R300_COMB_FCN_ADD_CLAMP (0 << 12)
1251 # define R300_COMB_FCN_ADD_NOCLAMP (1 << 12)
1252 # define R300_COMB_FCN_SUB_CLAMP (2 << 12)
1253 # define R300_COMB_FCN_SUB_NOCLAMP (3 << 12)
1254 # define R300_COMB_FCN_MIN (4 << 12)
1255 # define R300_COMB_FCN_MAX (5 << 12)
1256 # define R300_COMB_FCN_RSUB_CLAMP (6 << 12)
1257 # define R300_COMB_FCN_RSUB_NOCLAMP (7 << 12)
1258 # define R300_BLEND_GL_ZERO (32)
1259 # define R300_BLEND_GL_ONE (33)
1260 # define R300_BLEND_GL_SRC_COLOR (34)
1261 # define R300_BLEND_GL_ONE_MINUS_SRC_COLOR (35)
1262 # define R300_BLEND_GL_DST_COLOR (36)
1263 # define R300_BLEND_GL_ONE_MINUS_DST_COLOR (37)
1264 # define R300_BLEND_GL_SRC_ALPHA (38)
1265 # define R300_BLEND_GL_ONE_MINUS_SRC_ALPHA (39)
1266 # define R300_BLEND_GL_DST_ALPHA (40)
1267 # define R300_BLEND_GL_ONE_MINUS_DST_ALPHA (41)
1268 # define R300_BLEND_GL_SRC_ALPHA_SATURATE (42)
1269 # define R300_BLEND_GL_CONST_COLOR (43)
1270 # define R300_BLEND_GL_ONE_MINUS_CONST_COLOR (44)
1271 # define R300_BLEND_GL_CONST_ALPHA (45)
1272 # define R300_BLEND_GL_ONE_MINUS_CONST_ALPHA (46)
1273 # define R300_BLEND_MASK (63)
1274 # define R300_SRC_BLEND_SHIFT (16)
1275 # define R300_DST_BLEND_SHIFT (24)
1276 #define R300_RB3D_COLORMASK 0x4E0C
1277 # define R300_COLORMASK0_B (1<<0)
1278 # define R300_COLORMASK0_G (1<<1)
1279 # define R300_COLORMASK0_R (1<<2)
1280 # define R300_COLORMASK0_A (1<<3)
1281
1282 /* gap */
1283
1284 #define R300_RB3D_COLOROFFSET0 0x4E28
1285 # define R300_COLOROFFSET_MASK 0xFFFFFFF0 /* GUESS */
1286 #define R300_RB3D_COLOROFFSET1 0x4E2C /* GUESS */
1287 #define R300_RB3D_COLOROFFSET2 0x4E30 /* GUESS */
1288 #define R300_RB3D_COLOROFFSET3 0x4E34 /* GUESS */
1289
1290 /* gap */
1291
1292 /* Bit 16: Larger tiles
1293 * Bit 17: 4x2 tiles
1294 * Bit 18: Extremely weird tile like, but some pixels duplicated?
1295 */
1296 #define R300_RB3D_COLORPITCH0 0x4E38
1297 # define R300_COLORPITCH_MASK 0x00001FF8 /* GUESS */
1298 # define R300_COLOR_TILE_ENABLE (1 << 16) /* GUESS */
1299 # define R300_COLOR_MICROTILE_ENABLE (1 << 17) /* GUESS */
1300 # define R300_COLOR_ENDIAN_NO_SWAP (0 << 18) /* GUESS */
1301 # define R300_COLOR_ENDIAN_WORD_SWAP (1 << 18) /* GUESS */
1302 # define R300_COLOR_ENDIAN_DWORD_SWAP (2 << 18) /* GUESS */
1303 # define R300_COLOR_FORMAT_RGB565 (2 << 22)
1304 # define R300_COLOR_FORMAT_ARGB8888 (3 << 22)
1305 #define R300_RB3D_COLORPITCH1 0x4E3C /* GUESS */
1306 #define R300_RB3D_COLORPITCH2 0x4E40 /* GUESS */
1307 #define R300_RB3D_COLORPITCH3 0x4E44 /* GUESS */
1308
1309 /* gap */
1310
1311 /* Guess by Vladimir.
1312 * Set to 0A before 3D operations, set to 02 afterwards.
1313 */
1314 #define R300_RB3D_DSTCACHE_CTLSTAT 0x4E4C
1315 # define R300_RB3D_DSTCACHE_02 0x00000002
1316 # define R300_RB3D_DSTCACHE_0A 0x0000000A
1317
1318 /* gap */
1319 /* There seems to be no "write only" setting, so use Z-test = ALWAYS
1320 * for this.
1321 * Bit (1<<8) is the "test" bit. so plain write is 6 - vd
1322 */
1323 #define R300_RB3D_ZSTENCIL_CNTL_0 0x4F00
1324 # define R300_RB3D_Z_DISABLED_1 0x00000010
1325 # define R300_RB3D_Z_DISABLED_2 0x00000014
1326 # define R300_RB3D_Z_TEST 0x00000012
1327 # define R300_RB3D_Z_TEST_AND_WRITE 0x00000016
1328 # define R300_RB3D_Z_WRITE_ONLY 0x00000006
1329
1330 # define R300_RB3D_Z_TEST 0x00000012
1331 # define R300_RB3D_Z_TEST_AND_WRITE 0x00000016
1332 # define R300_RB3D_Z_WRITE_ONLY 0x00000006
1333 # define R300_RB3D_STENCIL_ENABLE 0x00000001
1334
1335 #define R300_RB3D_ZSTENCIL_CNTL_1 0x4F04
1336 /* functions */
1337 # define R300_ZS_NEVER 0
1338 # define R300_ZS_LESS 1
1339 # define R300_ZS_LEQUAL 2
1340 # define R300_ZS_EQUAL 3
1341 # define R300_ZS_GEQUAL 4
1342 # define R300_ZS_GREATER 5
1343 # define R300_ZS_NOTEQUAL 6
1344 # define R300_ZS_ALWAYS 7
1345 # define R300_ZS_MASK 7
1346 /* operations */
1347 # define R300_ZS_KEEP 0
1348 # define R300_ZS_ZERO 1
1349 # define R300_ZS_REPLACE 2
1350 # define R300_ZS_INCR 3
1351 # define R300_ZS_DECR 4
1352 # define R300_ZS_INVERT 5
1353 # define R300_ZS_INCR_WRAP 6
1354 # define R300_ZS_DECR_WRAP 7
1355 /* front and back refer to operations done for front
1356 and back faces, i.e. separate stencil function support */
1357 # define R300_RB3D_ZS1_DEPTH_FUNC_SHIFT 0
1358 # define R300_RB3D_ZS1_FRONT_FUNC_SHIFT 3
1359 # define R300_RB3D_ZS1_FRONT_FAIL_OP_SHIFT 6
1360 # define R300_RB3D_ZS1_FRONT_ZPASS_OP_SHIFT 9
1361 # define R300_RB3D_ZS1_FRONT_ZFAIL_OP_SHIFT 12
1362 # define R300_RB3D_ZS1_BACK_FUNC_SHIFT 15
1363 # define R300_RB3D_ZS1_BACK_FAIL_OP_SHIFT 18
1364 # define R300_RB3D_ZS1_BACK_ZPASS_OP_SHIFT 21
1365 # define R300_RB3D_ZS1_BACK_ZFAIL_OP_SHIFT 24
1366
1367 #define R300_RB3D_ZSTENCIL_CNTL_2 0x4F08
1368 # define R300_RB3D_ZS2_STENCIL_REF_SHIFT 0
1369 # define R300_RB3D_ZS2_STENCIL_MASK 0xFF
1370 # define R300_RB3D_ZS2_STENCIL_MASK_SHIFT 8
1371 # define R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT 16
1372
1373 /* gap */
1374
1375 #define R300_RB3D_ZSTENCIL_FORMAT 0x4F10
1376 # define R300_DEPTH_FORMAT_16BIT_INT_Z (0 << 0)
1377 # define R300_DEPTH_FORMAT_24BIT_INT_Z (2 << 0)
1378 /* 16 bit format or some aditional bit ? */
1379 # define R300_DEPTH_FORMAT_UNK32 (32 << 0)
1380
1381 /* gap */
1382
1383 #define R300_RB3D_DEPTHOFFSET 0x4F20
1384 #define R300_RB3D_DEPTHPITCH 0x4F24
1385 # define R300_DEPTHPITCH_MASK 0x00001FF8 /* GUESS */
1386 # define R300_DEPTH_TILE_ENABLE (1 << 16) /* GUESS */
1387 # define R300_DEPTH_MICROTILE_ENABLE (1 << 17) /* GUESS */
1388 # define R300_DEPTH_ENDIAN_NO_SWAP (0 << 18) /* GUESS */
1389 # define R300_DEPTH_ENDIAN_WORD_SWAP (1 << 18) /* GUESS */
1390 # define R300_DEPTH_ENDIAN_DWORD_SWAP (2 << 18) /* GUESS */
1391
1392 /* BEGIN: Vertex program instruction set */
1393
1394 /* Every instruction is four dwords long:
1395 * DWORD 0: output and opcode
1396 * DWORD 1: first argument
1397 * DWORD 2: second argument
1398 * DWORD 3: third argument
1399 *
1400 * Notes:
1401 * - ABS r, a is implemented as MAX r, a, -a
1402 * - MOV is implemented as ADD to zero
1403 * - XPD is implemented as MUL + MAD
1404 * - FLR is implemented as FRC + ADD
1405 * - apparently, fglrx tries to schedule instructions so that there is at
1406 * least one instruction between the write to a temporary and the first
1407 * read from said temporary; however, violations of this scheduling are
1408 * allowed
1409 * - register indices seem to be unrelated with OpenGL aliasing to
1410 * conventional state
1411 * - only one attribute and one parameter can be loaded at a time; however,
1412 * the same attribute/parameter can be used for more than one argument
1413 * - the second software argument for POW is the third hardware argument
1414 * (no idea why)
1415 * - MAD with only temporaries as input seems to use VPI_OUT_SELECT_MAD_2
1416 *
1417 * There is some magic surrounding LIT:
1418 * The single argument is replicated across all three inputs, but swizzled:
1419 * First argument: xyzy
1420 * Second argument: xyzx
1421 * Third argument: xyzw
1422 * Whenever the result is used later in the fragment program, fglrx forces
1423 * x and w to be 1.0 in the input selection; I don't know whether this is
1424 * strictly necessary
1425 */
1426 #define R300_VPI_OUT_OP_DOT (1 << 0)
1427 #define R300_VPI_OUT_OP_MUL (2 << 0)
1428 #define R300_VPI_OUT_OP_ADD (3 << 0)
1429 #define R300_VPI_OUT_OP_MAD (4 << 0)
1430 #define R300_VPI_OUT_OP_DST (5 << 0)
1431 #define R300_VPI_OUT_OP_FRC (6 << 0)
1432 #define R300_VPI_OUT_OP_MAX (7 << 0)
1433 #define R300_VPI_OUT_OP_MIN (8 << 0)
1434 #define R300_VPI_OUT_OP_SGE (9 << 0)
1435 #define R300_VPI_OUT_OP_SLT (10 << 0)
1436 /* Used in GL_POINT_DISTANCE_ATTENUATION_ARB, vector(scalar, vector) */
1437 #define R300_VPI_OUT_OP_UNK12 (12 << 0)
1438 #define R300_VPI_OUT_OP_ARL (13 << 0)
1439 #define R300_VPI_OUT_OP_EXP (65 << 0)
1440 #define R300_VPI_OUT_OP_LOG (66 << 0)
1441 /* Used in fog computations, scalar(scalar) */
1442 #define R300_VPI_OUT_OP_UNK67 (67 << 0)
1443 #define R300_VPI_OUT_OP_LIT (68 << 0)
1444 #define R300_VPI_OUT_OP_POW (69 << 0)
1445 #define R300_VPI_OUT_OP_RCP (70 << 0)
1446 #define R300_VPI_OUT_OP_RSQ (72 << 0)
1447 /* Used in GL_POINT_DISTANCE_ATTENUATION_ARB, scalar(scalar) */
1448 #define R300_VPI_OUT_OP_UNK73 (73 << 0)
1449 #define R300_VPI_OUT_OP_EX2 (75 << 0)
1450 #define R300_VPI_OUT_OP_LG2 (76 << 0)
1451 #define R300_VPI_OUT_OP_MAD_2 (128 << 0)
1452 /* all temps, vector(scalar, vector, vector) */
1453 #define R300_VPI_OUT_OP_UNK129 (129 << 0)
1454
1455 #define R300_VPI_OUT_REG_CLASS_TEMPORARY (0 << 8)
1456 #define R300_VPI_OUT_REG_CLASS_ADDR (1 << 8)
1457 #define R300_VPI_OUT_REG_CLASS_RESULT (2 << 8)
1458 #define R300_VPI_OUT_REG_CLASS_MASK (31 << 8)
1459
1460 #define R300_VPI_OUT_REG_INDEX_SHIFT 13
1461 /* GUESS based on fglrx native limits */
1462 #define R300_VPI_OUT_REG_INDEX_MASK (31 << 13)
1463
1464 #define R300_VPI_OUT_WRITE_X (1 << 20)
1465 #define R300_VPI_OUT_WRITE_Y (1 << 21)
1466 #define R300_VPI_OUT_WRITE_Z (1 << 22)
1467 #define R300_VPI_OUT_WRITE_W (1 << 23)
1468
1469 #define R300_VPI_IN_REG_CLASS_TEMPORARY (0 << 0)
1470 #define R300_VPI_IN_REG_CLASS_ATTRIBUTE (1 << 0)
1471 #define R300_VPI_IN_REG_CLASS_PARAMETER (2 << 0)
1472 #define R300_VPI_IN_REG_CLASS_NONE (9 << 0)
1473 #define R300_VPI_IN_REG_CLASS_MASK (31 << 0)
1474
1475 #define R300_VPI_IN_REG_INDEX_SHIFT 5
1476 /* GUESS based on fglrx native limits */
1477 #define R300_VPI_IN_REG_INDEX_MASK (255 << 5)
1478
1479 /* The R300 can select components from the input register arbitrarily.
1480 * Use the following constants, shifted by the component shift you
1481 * want to select
1482 */
1483 #define R300_VPI_IN_SELECT_X 0
1484 #define R300_VPI_IN_SELECT_Y 1
1485 #define R300_VPI_IN_SELECT_Z 2
1486 #define R300_VPI_IN_SELECT_W 3
1487 #define R300_VPI_IN_SELECT_ZERO 4
1488 #define R300_VPI_IN_SELECT_ONE 5
1489 #define R300_VPI_IN_SELECT_MASK 7
1490
1491 #define R300_VPI_IN_X_SHIFT 13
1492 #define R300_VPI_IN_Y_SHIFT 16
1493 #define R300_VPI_IN_Z_SHIFT 19
1494 #define R300_VPI_IN_W_SHIFT 22
1495
1496 #define R300_VPI_IN_NEG_X (1 << 25)
1497 #define R300_VPI_IN_NEG_Y (1 << 26)
1498 #define R300_VPI_IN_NEG_Z (1 << 27)
1499 #define R300_VPI_IN_NEG_W (1 << 28)
1500 /* END: Vertex program instruction set */
1501
1502 /* BEGIN: Packet 3 commands */
1503
1504 /* A primitive emission dword. */
1505 #define R300_PRIM_TYPE_NONE (0 << 0)
1506 #define R300_PRIM_TYPE_POINT (1 << 0)
1507 #define R300_PRIM_TYPE_LINE (2 << 0)
1508 #define R300_PRIM_TYPE_LINE_STRIP (3 << 0)
1509 #define R300_PRIM_TYPE_TRI_LIST (4 << 0)
1510 #define R300_PRIM_TYPE_TRI_FAN (5 << 0)
1511 #define R300_PRIM_TYPE_TRI_STRIP (6 << 0)
1512 #define R300_PRIM_TYPE_TRI_TYPE2 (7 << 0)
1513 #define R300_PRIM_TYPE_RECT_LIST (8 << 0)
1514 #define R300_PRIM_TYPE_3VRT_POINT_LIST (9 << 0)
1515 #define R300_PRIM_TYPE_3VRT_LINE_LIST (10 << 0)
1516 /* GUESS (based on r200) */
1517 #define R300_PRIM_TYPE_POINT_SPRITES (11 << 0)
1518 #define R300_PRIM_TYPE_LINE_LOOP (12 << 0)
1519 #define R300_PRIM_TYPE_QUADS (13 << 0)
1520 #define R300_PRIM_TYPE_QUAD_STRIP (14 << 0)
1521 #define R300_PRIM_TYPE_POLYGON (15 << 0)
1522 #define R300_PRIM_TYPE_MASK 0xF
1523 #define R300_PRIM_WALK_IND (1 << 4)
1524 #define R300_PRIM_WALK_LIST (2 << 4)
1525 #define R300_PRIM_WALK_RING (3 << 4)
1526 #define R300_PRIM_WALK_MASK (3 << 4)
1527 /* GUESS (based on r200) */
1528 #define R300_PRIM_COLOR_ORDER_BGRA (0 << 6)
1529 #define R300_PRIM_COLOR_ORDER_RGBA (1 << 6)
1530 #define R300_PRIM_NUM_VERTICES_SHIFT 16
1531
1532 /* Draw a primitive from vertex data in arrays loaded via 3D_LOAD_VBPNTR.
1533 * Two parameter dwords:
1534 * 0. The first parameter appears to be always 0
1535 * 1. The second parameter is a standard primitive emission dword.
1536 */
1537 #define R300_PACKET3_3D_DRAW_VBUF 0x00002800
1538
1539 /* Specify the full set of vertex arrays as (address, stride).
1540 * The first parameter is the number of vertex arrays specified.
1541 * The rest of the command is a variable length list of blocks, where
1542 * each block is three dwords long and specifies two arrays.
1543 * The first dword of a block is split into two words, the lower significant
1544 * word refers to the first array, the more significant word to the second
1545 * array in the block.
1546 * The low byte of each word contains the size of an array entry in dwords,
1547 * the high byte contains the stride of the array.
1548 * The second dword of a block contains the pointer to the first array,
1549 * the third dword of a block contains the pointer to the second array.
1550 * Note that if the total number of arrays is odd, the third dword of
1551 * the last block is omitted.
1552 */
1553 #define R300_PACKET3_3D_LOAD_VBPNTR 0x00002F00
1554
1555 #define R300_PACKET3_INDX_BUFFER 0x00003300
1556 # define R300_EB_UNK1_SHIFT 24
1557 # define R300_EB_UNK1 (0x80<<24)
1558 # define R300_EB_UNK2 0x0810
1559 #define R300_PACKET3_3D_DRAW_INDX_2 0x00003600
1560
1561 /* END: Packet 3 commands */
1562
1563
1564 /* Color formats for 2d packets
1565 */
1566 #define R300_CP_COLOR_FORMAT_CI8 2
1567 #define R300_CP_COLOR_FORMAT_ARGB1555 3
1568 #define R300_CP_COLOR_FORMAT_RGB565 4
1569 #define R300_CP_COLOR_FORMAT_ARGB8888 6
1570 #define R300_CP_COLOR_FORMAT_RGB332 7
1571 #define R300_CP_COLOR_FORMAT_RGB8 9
1572 #define R300_CP_COLOR_FORMAT_ARGB4444 15
1573
1574 /*
1575 * CP type-3 packets
1576 */
1577 #define R300_CP_CMD_BITBLT_MULTI 0xC0009B00
1578
1579
1580 #endif /* _R300_REG_H */