xref: /linux/drivers/gpu/drm/radeon/r300_reg.h (revision b7e1e969c887c897947fdc3754fe9b0c24acb155)
1 /*
2  * Copyright 2005 Nicolai Haehnle et al.
3  * Copyright 2008 Advanced Micro Devices, Inc.
4  * Copyright 2009 Jerome Glisse.
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a
7  * copy of this software and associated documentation files (the "Software"),
8  * to deal in the Software without restriction, including without limitation
9  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10  * and/or sell copies of the Software, and to permit persons to whom the
11  * Software is furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the 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 NONINFRINGEMENT.  IN NO EVENT SHALL
19  * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
20  * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
21  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
22  * OTHER DEALINGS IN THE SOFTWARE.
23  *
24  * Authors: Nicolai Haehnle
25  *          Jerome Glisse
26  */
27 #ifndef _R300_REG_H_
28 #define _R300_REG_H_
29 
30 #define R300_SURF_TILE_MACRO (1<<16)
31 #define R300_SURF_TILE_MICRO (2<<16)
32 #define R300_SURF_TILE_BOTH (3<<16)
33 
34 
35 #define R300_MC_INIT_MISC_LAT_TIMER	0x180
36 #	define R300_MC_MISC__MC_CPR_INIT_LAT_SHIFT	0
37 #	define R300_MC_MISC__MC_VF_INIT_LAT_SHIFT	4
38 #	define R300_MC_MISC__MC_DISP0R_INIT_LAT_SHIFT	8
39 #	define R300_MC_MISC__MC_DISP1R_INIT_LAT_SHIFT	12
40 #	define R300_MC_MISC__MC_FIXED_INIT_LAT_SHIFT	16
41 #	define R300_MC_MISC__MC_E2R_INIT_LAT_SHIFT	20
42 #	define R300_MC_MISC__MC_SAME_PAGE_PRIO_SHIFT	24
43 #	define R300_MC_MISC__MC_GLOBW_INIT_LAT_SHIFT	28
44 
45 #define R300_MC_INIT_GFX_LAT_TIMER	0x154
46 #	define R300_MC_MISC__MC_G3D0R_INIT_LAT_SHIFT	0
47 #	define R300_MC_MISC__MC_G3D1R_INIT_LAT_SHIFT	4
48 #	define R300_MC_MISC__MC_G3D2R_INIT_LAT_SHIFT	8
49 #	define R300_MC_MISC__MC_G3D3R_INIT_LAT_SHIFT	12
50 #	define R300_MC_MISC__MC_TX0R_INIT_LAT_SHIFT	16
51 #	define R300_MC_MISC__MC_TX1R_INIT_LAT_SHIFT	20
52 #	define R300_MC_MISC__MC_GLOBR_INIT_LAT_SHIFT	24
53 #	define R300_MC_MISC__MC_GLOBW_FULL_LAT_SHIFT	28
54 
55 /*
56  * This file contains registers and constants for the R300. They have been
57  * found mostly by examining command buffers captured using glxtest, as well
58  * as by extrapolating some known registers and constants from the R200.
59  * I am fairly certain that they are correct unless stated otherwise
60  * in comments.
61  */
62 
63 #define R300_SE_VPORT_XSCALE                0x1D98
64 #define R300_SE_VPORT_XOFFSET               0x1D9C
65 #define R300_SE_VPORT_YSCALE                0x1DA0
66 #define R300_SE_VPORT_YOFFSET               0x1DA4
67 #define R300_SE_VPORT_ZSCALE                0x1DA8
68 #define R300_SE_VPORT_ZOFFSET               0x1DAC
69 
70 
71 /*
72  * Vertex Array Processing (VAP) Control
73  * Stolen from r200 code from Christoph Brill (It's a guess!)
74  */
75 #define R300_VAP_CNTL	0x2080
76 
77 /* This register is written directly and also starts data section
78  * in many 3d CP_PACKET3's
79  */
80 #define R300_VAP_VF_CNTL	0x2084
81 #	define	R300_VAP_VF_CNTL__PRIM_TYPE__SHIFT              0
82 #	define  R300_VAP_VF_CNTL__PRIM_NONE                     (0<<0)
83 #	define  R300_VAP_VF_CNTL__PRIM_POINTS                   (1<<0)
84 #	define  R300_VAP_VF_CNTL__PRIM_LINES                    (2<<0)
85 #	define  R300_VAP_VF_CNTL__PRIM_LINE_STRIP               (3<<0)
86 #	define  R300_VAP_VF_CNTL__PRIM_TRIANGLES                (4<<0)
87 #	define  R300_VAP_VF_CNTL__PRIM_TRIANGLE_FAN             (5<<0)
88 #	define  R300_VAP_VF_CNTL__PRIM_TRIANGLE_STRIP           (6<<0)
89 #	define  R300_VAP_VF_CNTL__PRIM_LINE_LOOP                (12<<0)
90 #	define  R300_VAP_VF_CNTL__PRIM_QUADS                    (13<<0)
91 #	define  R300_VAP_VF_CNTL__PRIM_QUAD_STRIP               (14<<0)
92 #	define  R300_VAP_VF_CNTL__PRIM_POLYGON                  (15<<0)
93 
94 #	define	R300_VAP_VF_CNTL__PRIM_WALK__SHIFT              4
95 	/* State based - direct writes to registers trigger vertex
96            generation */
97 #	define	R300_VAP_VF_CNTL__PRIM_WALK_STATE_BASED         (0<<4)
98 #	define	R300_VAP_VF_CNTL__PRIM_WALK_INDICES             (1<<4)
99 #	define	R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST         (2<<4)
100 #	define	R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_EMBEDDED     (3<<4)
101 
102 	/* I don't think I saw these three used.. */
103 #	define	R300_VAP_VF_CNTL__COLOR_ORDER__SHIFT            6
104 #	define	R300_VAP_VF_CNTL__TCL_OUTPUT_CTL_ENA__SHIFT     9
105 #	define	R300_VAP_VF_CNTL__PROG_STREAM_ENA__SHIFT        10
106 
107 	/* index size - when not set the indices are assumed to be 16 bit */
108 #	define	R300_VAP_VF_CNTL__INDEX_SIZE_32bit              (1<<11)
109 	/* number of vertices */
110 #	define	R300_VAP_VF_CNTL__NUM_VERTICES__SHIFT           16
111 
112 /* BEGIN: Wild guesses */
113 #define R300_VAP_OUTPUT_VTX_FMT_0           0x2090
114 #       define R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT     (1<<0)
115 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT   (1<<1)
116 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT (1<<2)  /* GUESS */
117 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT (1<<3)  /* GUESS */
118 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT (1<<4)  /* GUESS */
119 #       define R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT (1<<16) /* GUESS */
120 
121 #define R300_VAP_OUTPUT_VTX_FMT_1           0x2094
122 	/* each of the following is 3 bits wide, specifies number
123 	   of components */
124 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT 0
125 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT 3
126 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT 6
127 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT 9
128 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT 12
129 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT 15
130 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT 18
131 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT 21
132 /* END: Wild guesses */
133 
134 #define R300_SE_VTE_CNTL                  0x20b0
135 #	define     R300_VPORT_X_SCALE_ENA                0x00000001
136 #	define     R300_VPORT_X_OFFSET_ENA               0x00000002
137 #	define     R300_VPORT_Y_SCALE_ENA                0x00000004
138 #	define     R300_VPORT_Y_OFFSET_ENA               0x00000008
139 #	define     R300_VPORT_Z_SCALE_ENA                0x00000010
140 #	define     R300_VPORT_Z_OFFSET_ENA               0x00000020
141 #	define     R300_VTX_XY_FMT                       0x00000100
142 #	define     R300_VTX_Z_FMT                        0x00000200
143 #	define     R300_VTX_W0_FMT                       0x00000400
144 #	define     R300_VTX_W0_NORMALIZE                 0x00000800
145 #	define     R300_VTX_ST_DENORMALIZED              0x00001000
146 
147 /* BEGIN: Vertex data assembly - lots of uncertainties */
148 
149 /* gap */
150 
151 #define R300_VAP_CNTL_STATUS              0x2140
152 #	define R300_VC_NO_SWAP                  (0 << 0)
153 #	define R300_VC_16BIT_SWAP               (1 << 0)
154 #	define R300_VC_32BIT_SWAP               (2 << 0)
155 #	define R300_VAP_TCL_BYPASS		(1 << 8)
156 
157 /* gap */
158 
159 /* Where do we get our vertex data?
160  *
161  * Vertex data either comes either from immediate mode registers or from
162  * vertex arrays.
163  * There appears to be no mixed mode (though we can force the pitch of
164  * vertex arrays to 0, effectively reusing the same element over and over
165  * again).
166  *
167  * Immediate mode is controlled by the INPUT_CNTL registers. I am not sure
168  * if these registers influence vertex array processing.
169  *
170  * Vertex arrays are controlled via the 3D_LOAD_VBPNTR packet3.
171  *
172  * In both cases, vertex attributes are then passed through INPUT_ROUTE.
173  *
174  * Beginning with INPUT_ROUTE_0_0 is a list of WORDs that route vertex data
175  * into the vertex processor's input registers.
176  * The first word routes the first input, the second word the second, etc.
177  * The corresponding input is routed into the register with the given index.
178  * The list is ended by a word with INPUT_ROUTE_END set.
179  *
180  * Always set COMPONENTS_4 in immediate mode.
181  */
182 
183 #define R300_VAP_INPUT_ROUTE_0_0            0x2150
184 #       define R300_INPUT_ROUTE_COMPONENTS_1     (0 << 0)
185 #       define R300_INPUT_ROUTE_COMPONENTS_2     (1 << 0)
186 #       define R300_INPUT_ROUTE_COMPONENTS_3     (2 << 0)
187 #       define R300_INPUT_ROUTE_COMPONENTS_4     (3 << 0)
188 #       define R300_INPUT_ROUTE_COMPONENTS_RGBA  (4 << 0) /* GUESS */
189 #       define R300_VAP_INPUT_ROUTE_IDX_SHIFT    8
190 #       define R300_VAP_INPUT_ROUTE_IDX_MASK     (31 << 8) /* GUESS */
191 #       define R300_VAP_INPUT_ROUTE_END          (1 << 13)
192 #       define R300_INPUT_ROUTE_IMMEDIATE_MODE   (0 << 14) /* GUESS */
193 #       define R300_INPUT_ROUTE_FLOAT            (1 << 14) /* GUESS */
194 #       define R300_INPUT_ROUTE_UNSIGNED_BYTE    (2 << 14) /* GUESS */
195 #       define R300_INPUT_ROUTE_FLOAT_COLOR      (3 << 14) /* GUESS */
196 #define R300_VAP_INPUT_ROUTE_0_1            0x2154
197 #define R300_VAP_INPUT_ROUTE_0_2            0x2158
198 #define R300_VAP_INPUT_ROUTE_0_3            0x215C
199 #define R300_VAP_INPUT_ROUTE_0_4            0x2160
200 #define R300_VAP_INPUT_ROUTE_0_5            0x2164
201 #define R300_VAP_INPUT_ROUTE_0_6            0x2168
202 #define R300_VAP_INPUT_ROUTE_0_7            0x216C
203 
204 /* gap */
205 
206 /* Notes:
207  *  - always set up to produce at least two attributes:
208  *    if vertex program uses only position, fglrx will set normal, too
209  *  - INPUT_CNTL_0_COLOR and INPUT_CNTL_COLOR bits are always equal.
210  */
211 #define R300_VAP_INPUT_CNTL_0               0x2180
212 #       define R300_INPUT_CNTL_0_COLOR           0x00000001
213 #define R300_VAP_INPUT_CNTL_1               0x2184
214 #       define R300_INPUT_CNTL_POS               0x00000001
215 #       define R300_INPUT_CNTL_NORMAL            0x00000002
216 #       define R300_INPUT_CNTL_COLOR             0x00000004
217 #       define R300_INPUT_CNTL_TC0               0x00000400
218 #       define R300_INPUT_CNTL_TC1               0x00000800
219 #       define R300_INPUT_CNTL_TC2               0x00001000 /* GUESS */
220 #       define R300_INPUT_CNTL_TC3               0x00002000 /* GUESS */
221 #       define R300_INPUT_CNTL_TC4               0x00004000 /* GUESS */
222 #       define R300_INPUT_CNTL_TC5               0x00008000 /* GUESS */
223 #       define R300_INPUT_CNTL_TC6               0x00010000 /* GUESS */
224 #       define R300_INPUT_CNTL_TC7               0x00020000 /* GUESS */
225 
226 /* gap */
227 
228 /* Words parallel to INPUT_ROUTE_0; All words that are active in INPUT_ROUTE_0
229  * are set to a swizzling bit pattern, other words are 0.
230  *
231  * In immediate mode, the pattern is always set to xyzw. In vertex array
232  * mode, the swizzling pattern is e.g. used to set zw components in texture
233  * coordinates with only tweo components.
234  */
235 #define R300_VAP_INPUT_ROUTE_1_0            0x21E0
236 #       define R300_INPUT_ROUTE_SELECT_X    0
237 #       define R300_INPUT_ROUTE_SELECT_Y    1
238 #       define R300_INPUT_ROUTE_SELECT_Z    2
239 #       define R300_INPUT_ROUTE_SELECT_W    3
240 #       define R300_INPUT_ROUTE_SELECT_ZERO 4
241 #       define R300_INPUT_ROUTE_SELECT_ONE  5
242 #       define R300_INPUT_ROUTE_SELECT_MASK 7
243 #       define R300_INPUT_ROUTE_X_SHIFT     0
244 #       define R300_INPUT_ROUTE_Y_SHIFT     3
245 #       define R300_INPUT_ROUTE_Z_SHIFT     6
246 #       define R300_INPUT_ROUTE_W_SHIFT     9
247 #       define R300_INPUT_ROUTE_ENABLE      (15 << 12)
248 #define R300_VAP_INPUT_ROUTE_1_1            0x21E4
249 #define R300_VAP_INPUT_ROUTE_1_2            0x21E8
250 #define R300_VAP_INPUT_ROUTE_1_3            0x21EC
251 #define R300_VAP_INPUT_ROUTE_1_4            0x21F0
252 #define R300_VAP_INPUT_ROUTE_1_5            0x21F4
253 #define R300_VAP_INPUT_ROUTE_1_6            0x21F8
254 #define R300_VAP_INPUT_ROUTE_1_7            0x21FC
255 
256 /* END: Vertex data assembly */
257 
258 /* gap */
259 
260 /* BEGIN: Upload vertex program and data */
261 
262 /*
263  * The programmable vertex shader unit has a memory bank of unknown size
264  * that can be written to in 16 byte units by writing the address into
265  * UPLOAD_ADDRESS, followed by data in UPLOAD_DATA (multiples of 4 DWORDs).
266  *
267  * Pointers into the memory bank are always in multiples of 16 bytes.
268  *
269  * The memory bank is divided into areas with fixed meaning.
270  *
271  * Starting at address UPLOAD_PROGRAM: Vertex program instructions.
272  * Native limits reported by drivers from ATI suggest size 256 (i.e. 4KB),
273  * whereas the difference between known addresses suggests size 512.
274  *
275  * Starting at address UPLOAD_PARAMETERS: Vertex program parameters.
276  * Native reported limits and the VPI layout suggest size 256, whereas
277  * difference between known addresses suggests size 512.
278  *
279  * At address UPLOAD_POINTSIZE is a vector (0, 0, ps, 0), where ps is the
280  * floating point pointsize. The exact purpose of this state is uncertain,
281  * as there is also the R300_RE_POINTSIZE register.
282  *
283  * Multiple vertex programs and parameter sets can be loaded at once,
284  * which could explain the size discrepancy.
285  */
286 #define R300_VAP_PVS_UPLOAD_ADDRESS         0x2200
287 #       define R300_PVS_UPLOAD_PROGRAM           0x00000000
288 #       define R300_PVS_UPLOAD_PARAMETERS        0x00000200
289 #       define R300_PVS_UPLOAD_POINTSIZE         0x00000406
290 
291 /* gap */
292 
293 #define R300_VAP_PVS_UPLOAD_DATA            0x2208
294 
295 /* END: Upload vertex program and data */
296 
297 /* gap */
298 
299 /* I do not know the purpose of this register. However, I do know that
300  * it is set to 221C_CLEAR for clear operations and to 221C_NORMAL
301  * for normal rendering.
302  */
303 #define R300_VAP_UNKNOWN_221C               0x221C
304 #       define R300_221C_NORMAL                  0x00000000
305 #       define R300_221C_CLEAR                   0x0001C000
306 
307 /* These seem to be per-pixel and per-vertex X and Y clipping planes. The first
308  * plane is per-pixel and the second plane is per-vertex.
309  *
310  * This was determined by experimentation alone but I believe it is correct.
311  *
312  * These registers are called X_QUAD0_1_FL to X_QUAD0_4_FL by glxtest.
313  */
314 #define R300_VAP_CLIP_X_0                   0x2220
315 #define R300_VAP_CLIP_X_1                   0x2224
316 #define R300_VAP_CLIP_Y_0                   0x2228
317 #define R300_VAP_CLIP_Y_1                   0x2230
318 
319 /* gap */
320 
321 /* Sometimes, END_OF_PKT and 0x2284=0 are the only commands sent between
322  * rendering commands and overwriting vertex program parameters.
323  * Therefore, I suspect writing zero to 0x2284 synchronizes the engine and
324  * avoids bugs caused by still running shaders reading bad data from memory.
325  */
326 #define R300_VAP_PVS_STATE_FLUSH_REG        0x2284
327 
328 /* Absolutely no clue what this register is about. */
329 #define R300_VAP_UNKNOWN_2288               0x2288
330 #       define R300_2288_R300                    0x00750000 /* -- nh */
331 #       define R300_2288_RV350                   0x0000FFFF /* -- Vladimir */
332 
333 /* gap */
334 
335 /* Addresses are relative to the vertex program instruction area of the
336  * memory bank. PROGRAM_END points to the last instruction of the active
337  * program
338  *
339  * The meaning of the two UNKNOWN fields is obviously not known. However,
340  * experiments so far have shown that both *must* point to an instruction
341  * inside the vertex program, otherwise the GPU locks up.
342  *
343  * fglrx usually sets CNTL_3_UNKNOWN to the end of the program and
344  * R300_PVS_CNTL_1_POS_END_SHIFT points to instruction where last write to
345  * position takes place.
346  *
347  * Most likely this is used to ignore rest of the program in cases
348  * where group of verts arent visible. For some reason this "section"
349  * is sometimes accepted other instruction that have no relationship with
350  * position calculations.
351  */
352 #define R300_VAP_PVS_CNTL_1                 0x22D0
353 #       define R300_PVS_CNTL_1_PROGRAM_START_SHIFT   0
354 #       define R300_PVS_CNTL_1_POS_END_SHIFT         10
355 #       define R300_PVS_CNTL_1_PROGRAM_END_SHIFT     20
356 /* Addresses are relative the vertex program parameters area. */
357 #define R300_VAP_PVS_CNTL_2                 0x22D4
358 #       define R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT 0
359 #       define R300_PVS_CNTL_2_PARAM_COUNT_SHIFT  16
360 #define R300_VAP_PVS_CNTL_3	           0x22D8
361 #       define R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT 10
362 #       define R300_PVS_CNTL_3_PROGRAM_UNKNOWN2_SHIFT 0
363 
364 /* The entire range from 0x2300 to 0x2AC inclusive seems to be used for
365  * immediate vertices
366  */
367 #define R300_VAP_VTX_COLOR_R                0x2464
368 #define R300_VAP_VTX_COLOR_G                0x2468
369 #define R300_VAP_VTX_COLOR_B                0x246C
370 #define R300_VAP_VTX_POS_0_X_1              0x2490 /* used for glVertex2*() */
371 #define R300_VAP_VTX_POS_0_Y_1              0x2494
372 #define R300_VAP_VTX_COLOR_PKD              0x249C /* RGBA */
373 #define R300_VAP_VTX_POS_0_X_2              0x24A0 /* used for glVertex3*() */
374 #define R300_VAP_VTX_POS_0_Y_2              0x24A4
375 #define R300_VAP_VTX_POS_0_Z_2              0x24A8
376 /* write 0 to indicate end of packet? */
377 #define R300_VAP_VTX_END_OF_PKT             0x24AC
378 
379 /* gap */
380 
381 /* These are values from r300_reg/r300_reg.h - they are known to be correct
382  * and are here so we can use one register file instead of several
383  * - Vladimir
384  */
385 #define R300_GB_VAP_RASTER_VTX_FMT_0	0x4000
386 #	define R300_GB_VAP_RASTER_VTX_FMT_0__POS_PRESENT	(1<<0)
387 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_0_PRESENT	(1<<1)
388 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_1_PRESENT	(1<<2)
389 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_2_PRESENT	(1<<3)
390 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_3_PRESENT	(1<<4)
391 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_SPACE	(0xf<<5)
392 #	define R300_GB_VAP_RASTER_VTX_FMT_0__PT_SIZE_PRESENT	(0x1<<16)
393 
394 #define R300_GB_VAP_RASTER_VTX_FMT_1	0x4004
395 	/* each of the following is 3 bits wide, specifies number
396 	   of components */
397 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT	0
398 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT	3
399 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT	6
400 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT	9
401 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT	12
402 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT	15
403 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT	18
404 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT	21
405 
406 /* UNK30 seems to enables point to quad transformation on textures
407  * (or something closely related to that).
408  * This bit is rather fatal at the time being due to lackings at pixel
409  * shader side
410  */
411 #define R300_GB_ENABLE	0x4008
412 #	define R300_GB_POINT_STUFF_ENABLE	(1<<0)
413 #	define R300_GB_LINE_STUFF_ENABLE	(1<<1)
414 #	define R300_GB_TRIANGLE_STUFF_ENABLE	(1<<2)
415 #	define R300_GB_STENCIL_AUTO_ENABLE	(1<<4)
416 #	define R300_GB_UNK31			(1<<31)
417 	/* each of the following is 2 bits wide */
418 #define R300_GB_TEX_REPLICATE	0
419 #define R300_GB_TEX_ST		1
420 #define R300_GB_TEX_STR		2
421 #	define R300_GB_TEX0_SOURCE_SHIFT	16
422 #	define R300_GB_TEX1_SOURCE_SHIFT	18
423 #	define R300_GB_TEX2_SOURCE_SHIFT	20
424 #	define R300_GB_TEX3_SOURCE_SHIFT	22
425 #	define R300_GB_TEX4_SOURCE_SHIFT	24
426 #	define R300_GB_TEX5_SOURCE_SHIFT	26
427 #	define R300_GB_TEX6_SOURCE_SHIFT	28
428 #	define R300_GB_TEX7_SOURCE_SHIFT	30
429 
430 /* MSPOS - positions for multisample antialiasing (?) */
431 #define R300_GB_MSPOS0	0x4010
432 	/* shifts - each of the fields is 4 bits */
433 #	define R300_GB_MSPOS0__MS_X0_SHIFT	0
434 #	define R300_GB_MSPOS0__MS_Y0_SHIFT	4
435 #	define R300_GB_MSPOS0__MS_X1_SHIFT	8
436 #	define R300_GB_MSPOS0__MS_Y1_SHIFT	12
437 #	define R300_GB_MSPOS0__MS_X2_SHIFT	16
438 #	define R300_GB_MSPOS0__MS_Y2_SHIFT	20
439 #	define R300_GB_MSPOS0__MSBD0_Y		24
440 #	define R300_GB_MSPOS0__MSBD0_X		28
441 
442 #define R300_GB_MSPOS1	0x4014
443 #	define R300_GB_MSPOS1__MS_X3_SHIFT	0
444 #	define R300_GB_MSPOS1__MS_Y3_SHIFT	4
445 #	define R300_GB_MSPOS1__MS_X4_SHIFT	8
446 #	define R300_GB_MSPOS1__MS_Y4_SHIFT	12
447 #	define R300_GB_MSPOS1__MS_X5_SHIFT	16
448 #	define R300_GB_MSPOS1__MS_Y5_SHIFT	20
449 #	define R300_GB_MSPOS1__MSBD1		24
450 
451 
452 #define R300_GB_TILE_CONFIG	0x4018
453 #	define R300_GB_TILE_ENABLE	(1<<0)
454 #	define R300_GB_TILE_PIPE_COUNT_RV300	0
455 #	define R300_GB_TILE_PIPE_COUNT_R300	(3<<1)
456 #	define R300_GB_TILE_PIPE_COUNT_R420	(7<<1)
457 #	define R300_GB_TILE_PIPE_COUNT_RV410	(3<<1)
458 #	define R300_GB_TILE_SIZE_8		0
459 #	define R300_GB_TILE_SIZE_16		(1<<4)
460 #	define R300_GB_TILE_SIZE_32		(2<<4)
461 #	define R300_GB_SUPER_SIZE_1		(0<<6)
462 #	define R300_GB_SUPER_SIZE_2		(1<<6)
463 #	define R300_GB_SUPER_SIZE_4		(2<<6)
464 #	define R300_GB_SUPER_SIZE_8		(3<<6)
465 #	define R300_GB_SUPER_SIZE_16		(4<<6)
466 #	define R300_GB_SUPER_SIZE_32		(5<<6)
467 #	define R300_GB_SUPER_SIZE_64		(6<<6)
468 #	define R300_GB_SUPER_SIZE_128		(7<<6)
469 #	define R300_GB_SUPER_X_SHIFT		9	/* 3 bits wide */
470 #	define R300_GB_SUPER_Y_SHIFT		12	/* 3 bits wide */
471 #	define R300_GB_SUPER_TILE_A		0
472 #	define R300_GB_SUPER_TILE_B		(1<<15)
473 #	define R300_GB_SUBPIXEL_1_12		0
474 #	define R300_GB_SUBPIXEL_1_16		(1<<16)
475 
476 #define R300_GB_FIFO_SIZE	0x4024
477 	/* each of the following is 2 bits wide */
478 #define R300_GB_FIFO_SIZE_32	0
479 #define R300_GB_FIFO_SIZE_64	1
480 #define R300_GB_FIFO_SIZE_128	2
481 #define R300_GB_FIFO_SIZE_256	3
482 #	define R300_SC_IFIFO_SIZE_SHIFT	0
483 #	define R300_SC_TZFIFO_SIZE_SHIFT	2
484 #	define R300_SC_BFIFO_SIZE_SHIFT	4
485 
486 #	define R300_US_OFIFO_SIZE_SHIFT	12
487 #	define R300_US_WFIFO_SIZE_SHIFT	14
488 	/* the following use the same constants as above, but meaning is
489 	   is times 2 (i.e. instead of 32 words it means 64 */
490 #	define R300_RS_TFIFO_SIZE_SHIFT	6
491 #	define R300_RS_CFIFO_SIZE_SHIFT	8
492 #	define R300_US_RAM_SIZE_SHIFT		10
493 	/* watermarks, 3 bits wide */
494 #	define R300_RS_HIGHWATER_COL_SHIFT	16
495 #	define R300_RS_HIGHWATER_TEX_SHIFT	19
496 #	define R300_OFIFO_HIGHWATER_SHIFT	22	/* two bits only */
497 #	define R300_CUBE_FIFO_HIGHWATER_COL_SHIFT	24
498 
499 #define R300_GB_SELECT	0x401C
500 #	define R300_GB_FOG_SELECT_C0A		0
501 #	define R300_GB_FOG_SELECT_C1A		1
502 #	define R300_GB_FOG_SELECT_C2A		2
503 #	define R300_GB_FOG_SELECT_C3A		3
504 #	define R300_GB_FOG_SELECT_1_1_W	4
505 #	define R300_GB_FOG_SELECT_Z		5
506 #	define R300_GB_DEPTH_SELECT_Z		0
507 #	define R300_GB_DEPTH_SELECT_1_1_W	(1<<3)
508 #	define R300_GB_W_SELECT_1_W		0
509 #	define R300_GB_W_SELECT_1		(1<<4)
510 
511 #define R300_GB_AA_CONFIG		0x4020
512 #	define R300_AA_DISABLE			0x00
513 #	define R300_AA_ENABLE			0x01
514 #	define R300_AA_SUBSAMPLES_2		0
515 #	define R300_AA_SUBSAMPLES_3		(1<<1)
516 #	define R300_AA_SUBSAMPLES_4		(2<<1)
517 #	define R300_AA_SUBSAMPLES_6		(3<<1)
518 
519 /* gap */
520 
521 /* Zero to flush caches. */
522 #define R300_TX_INVALTAGS                   0x4100
523 #define R300_TX_FLUSH                       0x0
524 
525 /* The upper enable bits are guessed, based on fglrx reported limits. */
526 #define R300_TX_ENABLE                      0x4104
527 #       define R300_TX_ENABLE_0                  (1 << 0)
528 #       define R300_TX_ENABLE_1                  (1 << 1)
529 #       define R300_TX_ENABLE_2                  (1 << 2)
530 #       define R300_TX_ENABLE_3                  (1 << 3)
531 #       define R300_TX_ENABLE_4                  (1 << 4)
532 #       define R300_TX_ENABLE_5                  (1 << 5)
533 #       define R300_TX_ENABLE_6                  (1 << 6)
534 #       define R300_TX_ENABLE_7                  (1 << 7)
535 #       define R300_TX_ENABLE_8                  (1 << 8)
536 #       define R300_TX_ENABLE_9                  (1 << 9)
537 #       define R300_TX_ENABLE_10                 (1 << 10)
538 #       define R300_TX_ENABLE_11                 (1 << 11)
539 #       define R300_TX_ENABLE_12                 (1 << 12)
540 #       define R300_TX_ENABLE_13                 (1 << 13)
541 #       define R300_TX_ENABLE_14                 (1 << 14)
542 #       define R300_TX_ENABLE_15                 (1 << 15)
543 
544 /* The pointsize is given in multiples of 6. The pointsize can be
545  * enormous: Clear() renders a single point that fills the entire
546  * framebuffer.
547  */
548 #define R300_RE_POINTSIZE                   0x421C
549 #       define R300_POINTSIZE_Y_SHIFT            0
550 #       define R300_POINTSIZE_Y_MASK             (0xFFFF << 0) /* GUESS */
551 #       define R300_POINTSIZE_X_SHIFT            16
552 #       define R300_POINTSIZE_X_MASK             (0xFFFF << 16) /* GUESS */
553 #       define R300_POINTSIZE_MAX             (R300_POINTSIZE_Y_MASK / 6)
554 
555 /* The line width is given in multiples of 6.
556  * In default mode lines are classified as vertical lines.
557  * HO: horizontal
558  * VE: vertical or horizontal
559  * HO & VE: no classification
560  */
561 #define R300_RE_LINE_CNT                      0x4234
562 #       define R300_LINESIZE_SHIFT            0
563 #       define R300_LINESIZE_MASK             (0xFFFF << 0) /* GUESS */
564 #       define R300_LINESIZE_MAX             (R300_LINESIZE_MASK / 6)
565 #       define R300_LINE_CNT_HO               (1 << 16)
566 #       define R300_LINE_CNT_VE               (1 << 17)
567 
568 /* Some sort of scale or clamp value for texcoordless textures. */
569 #define R300_RE_UNK4238                       0x4238
570 
571 /* Something shade related */
572 #define R300_RE_SHADE                         0x4274
573 
574 #define R300_RE_SHADE_MODEL                   0x4278
575 #	define R300_RE_SHADE_MODEL_SMOOTH     0x3aaaa
576 #	define R300_RE_SHADE_MODEL_FLAT       0x39595
577 
578 /* Dangerous */
579 #define R300_RE_POLYGON_MODE                  0x4288
580 #	define R300_PM_ENABLED                (1 << 0)
581 #	define R300_PM_FRONT_POINT            (0 << 0)
582 #	define R300_PM_BACK_POINT             (0 << 0)
583 #	define R300_PM_FRONT_LINE             (1 << 4)
584 #	define R300_PM_FRONT_FILL             (1 << 5)
585 #	define R300_PM_BACK_LINE              (1 << 7)
586 #	define R300_PM_BACK_FILL              (1 << 8)
587 
588 /* Fog parameters */
589 #define R300_RE_FOG_SCALE                     0x4294
590 #define R300_RE_FOG_START                     0x4298
591 
592 /* Not sure why there are duplicate of factor and constant values.
593  * My best guess so far is that there are separate zbiases for test and write.
594  * Ordering might be wrong.
595  * Some of the tests indicate that fgl has a fallback implementation of zbias
596  * via pixel shaders.
597  */
598 #define R300_RE_ZBIAS_CNTL                    0x42A0 /* GUESS */
599 #define R300_RE_ZBIAS_T_FACTOR                0x42A4
600 #define R300_RE_ZBIAS_T_CONSTANT              0x42A8
601 #define R300_RE_ZBIAS_W_FACTOR                0x42AC
602 #define R300_RE_ZBIAS_W_CONSTANT              0x42B0
603 
604 /* This register needs to be set to (1<<1) for RV350 to correctly
605  * perform depth test (see --vb-triangles in r300_demo)
606  * Don't know about other chips. - Vladimir
607  * This is set to 3 when GL_POLYGON_OFFSET_FILL is on.
608  * My guess is that there are two bits for each zbias primitive
609  * (FILL, LINE, POINT).
610  *  One to enable depth test and one for depth write.
611  * Yet this doesn't explain why depth writes work ...
612  */
613 #define R300_RE_OCCLUSION_CNTL		    0x42B4
614 #	define R300_OCCLUSION_ON		(1<<1)
615 
616 #define R300_RE_CULL_CNTL                   0x42B8
617 #       define R300_CULL_FRONT                   (1 << 0)
618 #       define R300_CULL_BACK                    (1 << 1)
619 #       define R300_FRONT_FACE_CCW               (0 << 2)
620 #       define R300_FRONT_FACE_CW                (1 << 2)
621 
622 
623 /* BEGIN: Rasterization / Interpolators - many guesses */
624 
625 /* 0_UNKNOWN_18 has always been set except for clear operations.
626  * TC_CNT is the number of incoming texture coordinate sets (i.e. it depends
627  * on the vertex program, *not* the fragment program)
628  */
629 #define R300_RS_CNTL_0                      0x4300
630 #       define R300_RS_CNTL_TC_CNT_SHIFT         2
631 #       define R300_RS_CNTL_TC_CNT_MASK          (7 << 2)
632 	/* number of color interpolators used */
633 #	define R300_RS_CNTL_CI_CNT_SHIFT         7
634 #       define R300_RS_CNTL_0_UNKNOWN_18         (1 << 18)
635 	/* Guess: RS_CNTL_1 holds the index of the highest used RS_ROUTE_n
636 	   register. */
637 #define R300_RS_CNTL_1                      0x4304
638 
639 /* gap */
640 
641 /* Only used for texture coordinates.
642  * Use the source field to route texture coordinate input from the
643  * vertex program to the desired interpolator. Note that the source
644  * field is relative to the outputs the vertex program *actually*
645  * writes. If a vertex program only writes texcoord[1], this will
646  * be source index 0.
647  * Set INTERP_USED on all interpolators that produce data used by
648  * the fragment program. INTERP_USED looks like a swizzling mask,
649  * but I haven't seen it used that way.
650  *
651  * Note: The _UNKNOWN constants are always set in their respective
652  * register. I don't know if this is necessary.
653  */
654 #define R300_RS_INTERP_0                    0x4310
655 #define R300_RS_INTERP_1                    0x4314
656 #       define R300_RS_INTERP_1_UNKNOWN          0x40
657 #define R300_RS_INTERP_2                    0x4318
658 #       define R300_RS_INTERP_2_UNKNOWN          0x80
659 #define R300_RS_INTERP_3                    0x431C
660 #       define R300_RS_INTERP_3_UNKNOWN          0xC0
661 #define R300_RS_INTERP_4                    0x4320
662 #define R300_RS_INTERP_5                    0x4324
663 #define R300_RS_INTERP_6                    0x4328
664 #define R300_RS_INTERP_7                    0x432C
665 #       define R300_RS_INTERP_SRC_SHIFT          2
666 #       define R300_RS_INTERP_SRC_MASK           (7 << 2)
667 #       define R300_RS_INTERP_USED               0x00D10000
668 
669 /* These DWORDs control how vertex data is routed into fragment program
670  * registers, after interpolators.
671  */
672 #define R300_RS_ROUTE_0                     0x4330
673 #define R300_RS_ROUTE_1                     0x4334
674 #define R300_RS_ROUTE_2                     0x4338
675 #define R300_RS_ROUTE_3                     0x433C /* GUESS */
676 #define R300_RS_ROUTE_4                     0x4340 /* GUESS */
677 #define R300_RS_ROUTE_5                     0x4344 /* GUESS */
678 #define R300_RS_ROUTE_6                     0x4348 /* GUESS */
679 #define R300_RS_ROUTE_7                     0x434C /* GUESS */
680 #       define R300_RS_ROUTE_SOURCE_INTERP_0     0
681 #       define R300_RS_ROUTE_SOURCE_INTERP_1     1
682 #       define R300_RS_ROUTE_SOURCE_INTERP_2     2
683 #       define R300_RS_ROUTE_SOURCE_INTERP_3     3
684 #       define R300_RS_ROUTE_SOURCE_INTERP_4     4
685 #       define R300_RS_ROUTE_SOURCE_INTERP_5     5 /* GUESS */
686 #       define R300_RS_ROUTE_SOURCE_INTERP_6     6 /* GUESS */
687 #       define R300_RS_ROUTE_SOURCE_INTERP_7     7 /* GUESS */
688 #       define R300_RS_ROUTE_ENABLE              (1 << 3) /* GUESS */
689 #       define R300_RS_ROUTE_DEST_SHIFT          6
690 #       define R300_RS_ROUTE_DEST_MASK           (31 << 6) /* GUESS */
691 
692 /* Special handling for color: When the fragment program uses color,
693  * the ROUTE_0_COLOR bit is set and ROUTE_0_COLOR_DEST contains the
694  * color register index.
695  *
696  * Apperently you may set the R300_RS_ROUTE_0_COLOR bit, but not provide any
697  * R300_RS_ROUTE_0_COLOR_DEST value; this setup is used for clearing the state.
698  * See r300_ioctl.c:r300EmitClearState. I'm not sure if this setup is strictly
699  * correct or not. - Oliver.
700  */
701 #       define R300_RS_ROUTE_0_COLOR             (1 << 14)
702 #       define R300_RS_ROUTE_0_COLOR_DEST_SHIFT  17
703 #       define R300_RS_ROUTE_0_COLOR_DEST_MASK   (31 << 17) /* GUESS */
704 /* As above, but for secondary color */
705 #		define R300_RS_ROUTE_1_COLOR1            (1 << 14)
706 #		define R300_RS_ROUTE_1_COLOR1_DEST_SHIFT 17
707 #		define R300_RS_ROUTE_1_COLOR1_DEST_MASK  (31 << 17)
708 #		define R300_RS_ROUTE_1_UNKNOWN11         (1 << 11)
709 /* END: Rasterization / Interpolators - many guesses */
710 
711 /* Hierarchical Z Enable */
712 #define R300_SC_HYPERZ                   0x43a4
713 #	define R300_SC_HYPERZ_DISABLE     (0 << 0)
714 #	define R300_SC_HYPERZ_ENABLE      (1 << 0)
715 #	define R300_SC_HYPERZ_MIN         (0 << 1)
716 #	define R300_SC_HYPERZ_MAX         (1 << 1)
717 #	define R300_SC_HYPERZ_ADJ_256     (0 << 2)
718 #	define R300_SC_HYPERZ_ADJ_128     (1 << 2)
719 #	define R300_SC_HYPERZ_ADJ_64      (2 << 2)
720 #	define R300_SC_HYPERZ_ADJ_32      (3 << 2)
721 #	define R300_SC_HYPERZ_ADJ_16      (4 << 2)
722 #	define R300_SC_HYPERZ_ADJ_8       (5 << 2)
723 #	define R300_SC_HYPERZ_ADJ_4       (6 << 2)
724 #	define R300_SC_HYPERZ_ADJ_2       (7 << 2)
725 #	define R300_SC_HYPERZ_HZ_Z0MIN_NO (0 << 5)
726 #	define R300_SC_HYPERZ_HZ_Z0MIN    (1 << 5)
727 #	define R300_SC_HYPERZ_HZ_Z0MAX_NO (0 << 6)
728 #	define R300_SC_HYPERZ_HZ_Z0MAX    (1 << 6)
729 
730 #define R300_SC_EDGERULE                 0x43a8
731 
732 /* BEGIN: Scissors and cliprects */
733 
734 /* There are four clipping rectangles. Their corner coordinates are inclusive.
735  * Every pixel is assigned a number from 0 and 15 by setting bits 0-3 depending
736  * on whether the pixel is inside cliprects 0-3, respectively. For example,
737  * if a pixel is inside cliprects 0 and 1, but outside 2 and 3, it is assigned
738  * the number 3 (binary 0011).
739  * Iff the bit corresponding to the pixel's number in RE_CLIPRECT_CNTL is set,
740  * the pixel is rasterized.
741  *
742  * In addition to this, there is a scissors rectangle. Only pixels inside the
743  * scissors rectangle are drawn. (coordinates are inclusive)
744  *
745  * For some reason, the top-left corner of the framebuffer is at (1440, 1440)
746  * for the purpose of clipping and scissors.
747  */
748 #define R300_RE_CLIPRECT_TL_0               0x43B0
749 #define R300_RE_CLIPRECT_BR_0               0x43B4
750 #define R300_RE_CLIPRECT_TL_1               0x43B8
751 #define R300_RE_CLIPRECT_BR_1               0x43BC
752 #define R300_RE_CLIPRECT_TL_2               0x43C0
753 #define R300_RE_CLIPRECT_BR_2               0x43C4
754 #define R300_RE_CLIPRECT_TL_3               0x43C8
755 #define R300_RE_CLIPRECT_BR_3               0x43CC
756 #       define R300_CLIPRECT_OFFSET              1440
757 #       define R300_CLIPRECT_MASK                0x1FFF
758 #       define R300_CLIPRECT_X_SHIFT             0
759 #       define R300_CLIPRECT_X_MASK              (0x1FFF << 0)
760 #       define R300_CLIPRECT_Y_SHIFT             13
761 #       define R300_CLIPRECT_Y_MASK              (0x1FFF << 13)
762 #define R300_RE_CLIPRECT_CNTL               0x43D0
763 #       define R300_CLIP_OUT                     (1 << 0)
764 #       define R300_CLIP_0                       (1 << 1)
765 #       define R300_CLIP_1                       (1 << 2)
766 #       define R300_CLIP_10                      (1 << 3)
767 #       define R300_CLIP_2                       (1 << 4)
768 #       define R300_CLIP_20                      (1 << 5)
769 #       define R300_CLIP_21                      (1 << 6)
770 #       define R300_CLIP_210                     (1 << 7)
771 #       define R300_CLIP_3                       (1 << 8)
772 #       define R300_CLIP_30                      (1 << 9)
773 #       define R300_CLIP_31                      (1 << 10)
774 #       define R300_CLIP_310                     (1 << 11)
775 #       define R300_CLIP_32                      (1 << 12)
776 #       define R300_CLIP_320                     (1 << 13)
777 #       define R300_CLIP_321                     (1 << 14)
778 #       define R300_CLIP_3210                    (1 << 15)
779 
780 /* gap */
781 
782 #define R300_RE_SCISSORS_TL                 0x43E0
783 #define R300_RE_SCISSORS_BR                 0x43E4
784 #       define R300_SCISSORS_OFFSET              1440
785 #       define R300_SCISSORS_X_SHIFT             0
786 #       define R300_SCISSORS_X_MASK              (0x1FFF << 0)
787 #       define R300_SCISSORS_Y_SHIFT             13
788 #       define R300_SCISSORS_Y_MASK              (0x1FFF << 13)
789 /* END: Scissors and cliprects */
790 
791 /* BEGIN: Texture specification */
792 
793 /*
794  * The texture specification dwords are grouped by meaning and not by texture
795  * unit. This means that e.g. the offset for texture image unit N is found in
796  * register TX_OFFSET_0 + (4*N)
797  */
798 #define R300_TX_FILTER_0                    0x4400
799 #       define R300_TX_REPEAT                    0
800 #       define R300_TX_MIRRORED                  1
801 #       define R300_TX_CLAMP                     4
802 #       define R300_TX_CLAMP_TO_EDGE             2
803 #       define R300_TX_CLAMP_TO_BORDER           6
804 #       define R300_TX_WRAP_S_SHIFT              0
805 #       define R300_TX_WRAP_S_MASK               (7 << 0)
806 #       define R300_TX_WRAP_T_SHIFT              3
807 #       define R300_TX_WRAP_T_MASK               (7 << 3)
808 #       define R300_TX_WRAP_Q_SHIFT              6
809 #       define R300_TX_WRAP_Q_MASK               (7 << 6)
810 #       define R300_TX_MAG_FILTER_NEAREST        (1 << 9)
811 #       define R300_TX_MAG_FILTER_LINEAR         (2 << 9)
812 #       define R300_TX_MAG_FILTER_MASK           (3 << 9)
813 #       define R300_TX_MIN_FILTER_NEAREST        (1 << 11)
814 #       define R300_TX_MIN_FILTER_LINEAR         (2 << 11)
815 #	define R300_TX_MIN_FILTER_NEAREST_MIP_NEAREST       (5  <<  11)
816 #	define R300_TX_MIN_FILTER_NEAREST_MIP_LINEAR        (9  <<  11)
817 #	define R300_TX_MIN_FILTER_LINEAR_MIP_NEAREST        (6  <<  11)
818 #	define R300_TX_MIN_FILTER_LINEAR_MIP_LINEAR         (10 <<  11)
819 
820 /* NOTE: NEAREST doesn't seem to exist.
821  * Im not seting MAG_FILTER_MASK and (3 << 11) on for all
822  * anisotropy modes because that would void selected mag filter
823  */
824 #	define R300_TX_MIN_FILTER_ANISO_NEAREST             (0 << 13)
825 #	define R300_TX_MIN_FILTER_ANISO_LINEAR              (0 << 13)
826 #	define R300_TX_MIN_FILTER_ANISO_NEAREST_MIP_NEAREST (1 << 13)
827 #	define R300_TX_MIN_FILTER_ANISO_NEAREST_MIP_LINEAR  (2 << 13)
828 #       define R300_TX_MIN_FILTER_MASK   ((15 << 11) | (3 << 13))
829 #	define R300_TX_MAX_ANISO_1_TO_1  (0 << 21)
830 #	define R300_TX_MAX_ANISO_2_TO_1  (2 << 21)
831 #	define R300_TX_MAX_ANISO_4_TO_1  (4 << 21)
832 #	define R300_TX_MAX_ANISO_8_TO_1  (6 << 21)
833 #	define R300_TX_MAX_ANISO_16_TO_1 (8 << 21)
834 #	define R300_TX_MAX_ANISO_MASK    (14 << 21)
835 
836 #define R300_TX_FILTER1_0                      0x4440
837 #	define R300_CHROMA_KEY_MODE_DISABLE    0
838 #	define R300_CHROMA_KEY_FORCE	       1
839 #	define R300_CHROMA_KEY_BLEND           2
840 #	define R300_MC_ROUND_NORMAL            (0<<2)
841 #	define R300_MC_ROUND_MPEG4             (1<<2)
842 #	define R300_LOD_BIAS_MASK	    0x1fff
843 #	define R300_EDGE_ANISO_EDGE_DIAG       (0<<13)
844 #	define R300_EDGE_ANISO_EDGE_ONLY       (1<<13)
845 #	define R300_MC_COORD_TRUNCATE_DISABLE  (0<<14)
846 #	define R300_MC_COORD_TRUNCATE_MPEG     (1<<14)
847 #	define R300_TX_TRI_PERF_0_8            (0<<15)
848 #	define R300_TX_TRI_PERF_1_8            (1<<15)
849 #	define R300_TX_TRI_PERF_1_4            (2<<15)
850 #	define R300_TX_TRI_PERF_3_8            (3<<15)
851 #	define R300_ANISO_THRESHOLD_MASK       (7<<17)
852 
853 #define R300_TX_SIZE_0                      0x4480
854 #       define R300_TX_WIDTHMASK_SHIFT           0
855 #       define R300_TX_WIDTHMASK_MASK            (2047 << 0)
856 #       define R300_TX_HEIGHTMASK_SHIFT          11
857 #       define R300_TX_HEIGHTMASK_MASK           (2047 << 11)
858 #       define R300_TX_UNK23                     (1 << 23)
859 #       define R300_TX_MAX_MIP_LEVEL_SHIFT       26
860 #       define R300_TX_MAX_MIP_LEVEL_MASK        (0xf << 26)
861 #       define R300_TX_SIZE_PROJECTED            (1<<30)
862 #       define R300_TX_SIZE_TXPITCH_EN           (1<<31)
863 #define R300_TX_FORMAT_0                    0x44C0
864 	/* The interpretation of the format word by Wladimir van der Laan */
865 	/* The X, Y, Z and W refer to the layout of the components.
866 	   They are given meanings as R, G, B and Alpha by the swizzle
867 	   specification */
868 #	define R300_TX_FORMAT_X8		    0x0
869 #	define R300_TX_FORMAT_X16		    0x1
870 #	define R300_TX_FORMAT_Y4X4		    0x2
871 #	define R300_TX_FORMAT_Y8X8		    0x3
872 #	define R300_TX_FORMAT_Y16X16		    0x4
873 #	define R300_TX_FORMAT_Z3Y3X2		    0x5
874 #	define R300_TX_FORMAT_Z5Y6X5		    0x6
875 #	define R300_TX_FORMAT_Z6Y5X5		    0x7
876 #	define R300_TX_FORMAT_Z11Y11X10		    0x8
877 #	define R300_TX_FORMAT_Z10Y11X11		    0x9
878 #	define R300_TX_FORMAT_W4Z4Y4X4		    0xA
879 #	define R300_TX_FORMAT_W1Z5Y5X5		    0xB
880 #	define R300_TX_FORMAT_W8Z8Y8X8		    0xC
881 #	define R300_TX_FORMAT_W2Z10Y10X10	    0xD
882 #	define R300_TX_FORMAT_W16Z16Y16X16	    0xE
883 #	define R300_TX_FORMAT_DXT1		    0xF
884 #	define R300_TX_FORMAT_DXT3		    0x10
885 #	define R300_TX_FORMAT_DXT5		    0x11
886 #	define R300_TX_FORMAT_D3DMFT_CxV8U8	    0x12     /* no swizzle */
887 #	define R300_TX_FORMAT_A8R8G8B8		    0x13     /* no swizzle */
888 #	define R300_TX_FORMAT_B8G8_B8G8		    0x14     /* no swizzle */
889 #	define R300_TX_FORMAT_G8R8_G8B8		    0x15     /* no swizzle */
890 	/* 0x16 - some 16 bit green format.. ?? */
891 #	define R300_TX_FORMAT_UNK25		   (1 << 25) /* no swizzle */
892 #	define R300_TX_FORMAT_CUBIC_MAP		   (1 << 26)
893 
894 	/* gap */
895 	/* Floating point formats */
896 	/* Note - hardware supports both 16 and 32 bit floating point */
897 #	define R300_TX_FORMAT_FL_I16		    0x18
898 #	define R300_TX_FORMAT_FL_I16A16		    0x19
899 #	define R300_TX_FORMAT_FL_R16G16B16A16	    0x1A
900 #	define R300_TX_FORMAT_FL_I32		    0x1B
901 #	define R300_TX_FORMAT_FL_I32A32		    0x1C
902 #	define R300_TX_FORMAT_FL_R32G32B32A32	    0x1D
903 #	define R300_TX_FORMAT_ATI2N		    0x1F
904 	/* alpha modes, convenience mostly */
905 	/* if you have alpha, pick constant appropriate to the
906 	   number of channels (1 for I8, 2 for I8A8, 4 for R8G8B8A8, etc */
907 #	define R300_TX_FORMAT_ALPHA_1CH		    0x000
908 #	define R300_TX_FORMAT_ALPHA_2CH		    0x200
909 #	define R300_TX_FORMAT_ALPHA_4CH		    0x600
910 #	define R300_TX_FORMAT_ALPHA_NONE	    0xA00
911 	/* Swizzling */
912 	/* constants */
913 #	define R300_TX_FORMAT_X		0
914 #	define R300_TX_FORMAT_Y		1
915 #	define R300_TX_FORMAT_Z		2
916 #	define R300_TX_FORMAT_W		3
917 #	define R300_TX_FORMAT_ZERO	4
918 #	define R300_TX_FORMAT_ONE	5
919 	/* 2.0*Z, everything above 1.0 is set to 0.0 */
920 #	define R300_TX_FORMAT_CUT_Z	6
921 	/* 2.0*W, everything above 1.0 is set to 0.0 */
922 #	define R300_TX_FORMAT_CUT_W	7
923 
924 #	define R300_TX_FORMAT_B_SHIFT	18
925 #	define R300_TX_FORMAT_G_SHIFT	15
926 #	define R300_TX_FORMAT_R_SHIFT	12
927 #	define R300_TX_FORMAT_A_SHIFT	9
928 	/* Convenience macro to take care of layout and swizzling */
929 #	define R300_EASY_TX_FORMAT(B, G, R, A, FMT)	(		\
930 		((R300_TX_FORMAT_##B)<<R300_TX_FORMAT_B_SHIFT)		\
931 		| ((R300_TX_FORMAT_##G)<<R300_TX_FORMAT_G_SHIFT)	\
932 		| ((R300_TX_FORMAT_##R)<<R300_TX_FORMAT_R_SHIFT)	\
933 		| ((R300_TX_FORMAT_##A)<<R300_TX_FORMAT_A_SHIFT)	\
934 		| (R300_TX_FORMAT_##FMT)				\
935 		)
936 	/* These can be ORed with result of R300_EASY_TX_FORMAT()
937 	   We don't really know what they do. Take values from a
938            constant color ? */
939 #	define R300_TX_FORMAT_CONST_X		(1<<5)
940 #	define R300_TX_FORMAT_CONST_Y		(2<<5)
941 #	define R300_TX_FORMAT_CONST_Z		(4<<5)
942 #	define R300_TX_FORMAT_CONST_W		(8<<5)
943 
944 #	define R300_TX_FORMAT_YUV_MODE		0x00800000
945 
946 #define R300_TX_PITCH_0			    0x4500 /* obvious missing in gap */
947 #define R300_TX_OFFSET_0                    0x4540
948 	/* BEGIN: Guess from R200 */
949 #       define R300_TXO_ENDIAN_NO_SWAP           (0 << 0)
950 #       define R300_TXO_ENDIAN_BYTE_SWAP         (1 << 0)
951 #       define R300_TXO_ENDIAN_WORD_SWAP         (2 << 0)
952 #       define R300_TXO_ENDIAN_HALFDW_SWAP       (3 << 0)
953 #       define R300_TXO_MACRO_TILE               (1 << 2)
954 #       define R300_TXO_MICRO_TILE               (1 << 3)
955 #       define R300_TXO_MICRO_TILE_SQUARE        (2 << 3)
956 #       define R300_TXO_OFFSET_MASK              0xffffffe0
957 #       define R300_TXO_OFFSET_SHIFT             5
958 	/* END: Guess from R200 */
959 
960 /* 32 bit chroma key */
961 #define R300_TX_CHROMA_KEY_0                      0x4580
962 /* ff00ff00 == { 0, 1.0, 0, 1.0 } */
963 #define R300_TX_BORDER_COLOR_0              0x45C0
964 
965 /* END: Texture specification */
966 
967 /* BEGIN: Fragment program instruction set */
968 
969 /* Fragment programs are written directly into register space.
970  * There are separate instruction streams for texture instructions and ALU
971  * instructions.
972  * In order to synchronize these streams, the program is divided into up
973  * to 4 nodes. Each node begins with a number of TEX operations, followed
974  * by a number of ALU operations.
975  * The first node can have zero TEX ops, all subsequent nodes must have at
976  * least
977  * one TEX ops.
978  * All nodes must have at least one ALU op.
979  *
980  * The index of the last node is stored in PFS_CNTL_0: A value of 0 means
981  * 1 node, a value of 3 means 4 nodes.
982  * The total amount of instructions is defined in PFS_CNTL_2. The offsets are
983  * offsets into the respective instruction streams, while *_END points to the
984  * last instruction relative to this offset.
985  */
986 #define R300_PFS_CNTL_0                     0x4600
987 #       define R300_PFS_CNTL_LAST_NODES_SHIFT    0
988 #       define R300_PFS_CNTL_LAST_NODES_MASK     (3 << 0)
989 #       define R300_PFS_CNTL_FIRST_NODE_HAS_TEX  (1 << 3)
990 #define R300_PFS_CNTL_1                     0x4604
991 /* There is an unshifted value here which has so far always been equal to the
992  * index of the highest used temporary register.
993  */
994 #define R300_PFS_CNTL_2                     0x4608
995 #       define R300_PFS_CNTL_ALU_OFFSET_SHIFT    0
996 #       define R300_PFS_CNTL_ALU_OFFSET_MASK     (63 << 0)
997 #       define R300_PFS_CNTL_ALU_END_SHIFT       6
998 #       define R300_PFS_CNTL_ALU_END_MASK        (63 << 6)
999 #       define R300_PFS_CNTL_TEX_OFFSET_SHIFT    12
1000 #       define R300_PFS_CNTL_TEX_OFFSET_MASK     (31 << 12) /* GUESS */
1001 #       define R300_PFS_CNTL_TEX_END_SHIFT       18
1002 #       define R300_PFS_CNTL_TEX_END_MASK        (31 << 18) /* GUESS */
1003 
1004 /* gap */
1005 
1006 /* Nodes are stored backwards. The last active node is always stored in
1007  * PFS_NODE_3.
1008  * Example: In a 2-node program, NODE_0 and NODE_1 are set to 0. The
1009  * first node is stored in NODE_2, the second node is stored in NODE_3.
1010  *
1011  * Offsets are relative to the master offset from PFS_CNTL_2.
1012  */
1013 #define R300_PFS_NODE_0                     0x4610
1014 #define R300_PFS_NODE_1                     0x4614
1015 #define R300_PFS_NODE_2                     0x4618
1016 #define R300_PFS_NODE_3                     0x461C
1017 #       define R300_PFS_NODE_ALU_OFFSET_SHIFT    0
1018 #       define R300_PFS_NODE_ALU_OFFSET_MASK     (63 << 0)
1019 #       define R300_PFS_NODE_ALU_END_SHIFT       6
1020 #       define R300_PFS_NODE_ALU_END_MASK        (63 << 6)
1021 #       define R300_PFS_NODE_TEX_OFFSET_SHIFT    12
1022 #       define R300_PFS_NODE_TEX_OFFSET_MASK     (31 << 12)
1023 #       define R300_PFS_NODE_TEX_END_SHIFT       17
1024 #       define R300_PFS_NODE_TEX_END_MASK        (31 << 17)
1025 #		define R300_PFS_NODE_OUTPUT_COLOR        (1 << 22)
1026 #		define R300_PFS_NODE_OUTPUT_DEPTH        (1 << 23)
1027 
1028 /* TEX
1029  * As far as I can tell, texture instructions cannot write into output
1030  * registers directly. A subsequent ALU instruction is always necessary,
1031  * even if it's just MAD o0, r0, 1, 0
1032  */
1033 #define R300_PFS_TEXI_0                     0x4620
1034 #	define R300_FPITX_SRC_SHIFT              0
1035 #	define R300_FPITX_SRC_MASK               (31 << 0)
1036 	/* GUESS */
1037 #	define R300_FPITX_SRC_CONST              (1 << 5)
1038 #	define R300_FPITX_DST_SHIFT              6
1039 #	define R300_FPITX_DST_MASK               (31 << 6)
1040 #	define R300_FPITX_IMAGE_SHIFT            11
1041 	/* GUESS based on layout and native limits */
1042 #       define R300_FPITX_IMAGE_MASK             (15 << 11)
1043 /* Unsure if these are opcodes, or some kind of bitfield, but this is how
1044  * they were set when I checked
1045  */
1046 #	define R300_FPITX_OPCODE_SHIFT		15
1047 #		define R300_FPITX_OP_TEX	1
1048 #		define R300_FPITX_OP_KIL	2
1049 #		define R300_FPITX_OP_TXP	3
1050 #		define R300_FPITX_OP_TXB	4
1051 #	define R300_FPITX_OPCODE_MASK           (7 << 15)
1052 
1053 /* ALU
1054  * The ALU instructions register blocks are enumerated according to the order
1055  * in which fglrx. I assume there is space for 64 instructions, since
1056  * each block has space for a maximum of 64 DWORDs, and this matches reported
1057  * native limits.
1058  *
1059  * The basic functional block seems to be one MAD for each color and alpha,
1060  * and an adder that adds all components after the MUL.
1061  *  - ADD, MUL, MAD etc.: use MAD with appropriate neutral operands
1062  *  - DP4: Use OUTC_DP4, OUTA_DP4
1063  *  - DP3: Use OUTC_DP3, OUTA_DP4, appropriate alpha operands
1064  *  - DPH: Use OUTC_DP4, OUTA_DP4, appropriate alpha operands
1065  *  - CMPH: If ARG2 > 0.5, return ARG0, else return ARG1
1066  *  - CMP: If ARG2 < 0, return ARG1, else return ARG0
1067  *  - FLR: use FRC+MAD
1068  *  - XPD: use MAD+MAD
1069  *  - SGE, SLT: use MAD+CMP
1070  *  - RSQ: use ABS modifier for argument
1071  *  - Use OUTC_REPL_ALPHA to write results of an alpha-only operation
1072  *    (e.g. RCP) into color register
1073  *  - apparently, there's no quick DST operation
1074  *  - fglrx set FPI2_UNKNOWN_31 on a "MAD fragment.color, tmp0, tmp1, tmp2"
1075  *  - fglrx set FPI2_UNKNOWN_31 on a "MAX r2, r1, c0"
1076  *  - fglrx once set FPI0_UNKNOWN_31 on a "FRC r1, r1"
1077  *
1078  * Operand selection
1079  * First stage selects three sources from the available registers and
1080  * constant parameters. This is defined in INSTR1 (color) and INSTR3 (alpha).
1081  * fglrx sorts the three source fields: Registers before constants,
1082  * lower indices before higher indices; I do not know whether this is
1083  * necessary.
1084  *
1085  * fglrx fills unused sources with "read constant 0"
1086  * According to specs, you cannot select more than two different constants.
1087  *
1088  * Second stage selects the operands from the sources. This is defined in
1089  * INSTR0 (color) and INSTR2 (alpha). You can also select the special constants
1090  * zero and one.
1091  * Swizzling and negation happens in this stage, as well.
1092  *
1093  * Important: Color and alpha seem to be mostly separate, i.e. their sources
1094  * selection appears to be fully independent (the register storage is probably
1095  * physically split into a color and an alpha section).
1096  * However (because of the apparent physical split), there is some interaction
1097  * WRT swizzling. If, for example, you want to load an R component into an
1098  * Alpha operand, this R component is taken from a *color* source, not from
1099  * an alpha source. The corresponding register doesn't even have to appear in
1100  * the alpha sources list. (I hope this all makes sense to you)
1101  *
1102  * Destination selection
1103  * The destination register index is in FPI1 (color) and FPI3 (alpha)
1104  * together with enable bits.
1105  * There are separate enable bits for writing into temporary registers
1106  * (DSTC_REG_* /DSTA_REG) and program output registers (DSTC_OUTPUT_*
1107  * /DSTA_OUTPUT). You can write to both at once, or not write at all (the
1108  * same index must be used for both).
1109  *
1110  * Note: There is a special form for LRP
1111  *  - Argument order is the same as in ARB_fragment_program.
1112  *  - Operation is MAD
1113  *  - ARG1 is set to ARGC_SRC1C_LRP/ARGC_SRC1A_LRP
1114  *  - Set FPI0/FPI2_SPECIAL_LRP
1115  * Arbitrary LRP (including support for swizzling) requires vanilla MAD+MAD
1116  */
1117 #define R300_PFS_INSTR1_0                   0x46C0
1118 #       define R300_FPI1_SRC0C_SHIFT             0
1119 #       define R300_FPI1_SRC0C_MASK              (31 << 0)
1120 #       define R300_FPI1_SRC0C_CONST             (1 << 5)
1121 #       define R300_FPI1_SRC1C_SHIFT             6
1122 #       define R300_FPI1_SRC1C_MASK              (31 << 6)
1123 #       define R300_FPI1_SRC1C_CONST             (1 << 11)
1124 #       define R300_FPI1_SRC2C_SHIFT             12
1125 #       define R300_FPI1_SRC2C_MASK              (31 << 12)
1126 #       define R300_FPI1_SRC2C_CONST             (1 << 17)
1127 #       define R300_FPI1_SRC_MASK                0x0003ffff
1128 #       define R300_FPI1_DSTC_SHIFT              18
1129 #       define R300_FPI1_DSTC_MASK               (31 << 18)
1130 #		define R300_FPI1_DSTC_REG_MASK_SHIFT     23
1131 #       define R300_FPI1_DSTC_REG_X              (1 << 23)
1132 #       define R300_FPI1_DSTC_REG_Y              (1 << 24)
1133 #       define R300_FPI1_DSTC_REG_Z              (1 << 25)
1134 #		define R300_FPI1_DSTC_OUTPUT_MASK_SHIFT  26
1135 #       define R300_FPI1_DSTC_OUTPUT_X           (1 << 26)
1136 #       define R300_FPI1_DSTC_OUTPUT_Y           (1 << 27)
1137 #       define R300_FPI1_DSTC_OUTPUT_Z           (1 << 28)
1138 
1139 #define R300_PFS_INSTR3_0                   0x47C0
1140 #       define R300_FPI3_SRC0A_SHIFT             0
1141 #       define R300_FPI3_SRC0A_MASK              (31 << 0)
1142 #       define R300_FPI3_SRC0A_CONST             (1 << 5)
1143 #       define R300_FPI3_SRC1A_SHIFT             6
1144 #       define R300_FPI3_SRC1A_MASK              (31 << 6)
1145 #       define R300_FPI3_SRC1A_CONST             (1 << 11)
1146 #       define R300_FPI3_SRC2A_SHIFT             12
1147 #       define R300_FPI3_SRC2A_MASK              (31 << 12)
1148 #       define R300_FPI3_SRC2A_CONST             (1 << 17)
1149 #       define R300_FPI3_SRC_MASK                0x0003ffff
1150 #       define R300_FPI3_DSTA_SHIFT              18
1151 #       define R300_FPI3_DSTA_MASK               (31 << 18)
1152 #       define R300_FPI3_DSTA_REG                (1 << 23)
1153 #       define R300_FPI3_DSTA_OUTPUT             (1 << 24)
1154 #		define R300_FPI3_DSTA_DEPTH              (1 << 27)
1155 
1156 #define R300_PFS_INSTR0_0                   0x48C0
1157 #       define R300_FPI0_ARGC_SRC0C_XYZ          0
1158 #       define R300_FPI0_ARGC_SRC0C_XXX          1
1159 #       define R300_FPI0_ARGC_SRC0C_YYY          2
1160 #       define R300_FPI0_ARGC_SRC0C_ZZZ          3
1161 #       define R300_FPI0_ARGC_SRC1C_XYZ          4
1162 #       define R300_FPI0_ARGC_SRC1C_XXX          5
1163 #       define R300_FPI0_ARGC_SRC1C_YYY          6
1164 #       define R300_FPI0_ARGC_SRC1C_ZZZ          7
1165 #       define R300_FPI0_ARGC_SRC2C_XYZ          8
1166 #       define R300_FPI0_ARGC_SRC2C_XXX          9
1167 #       define R300_FPI0_ARGC_SRC2C_YYY          10
1168 #       define R300_FPI0_ARGC_SRC2C_ZZZ          11
1169 #       define R300_FPI0_ARGC_SRC0A              12
1170 #       define R300_FPI0_ARGC_SRC1A              13
1171 #       define R300_FPI0_ARGC_SRC2A              14
1172 #       define R300_FPI0_ARGC_SRC1C_LRP          15
1173 #       define R300_FPI0_ARGC_ZERO               20
1174 #       define R300_FPI0_ARGC_ONE                21
1175 	/* GUESS */
1176 #       define R300_FPI0_ARGC_HALF               22
1177 #       define R300_FPI0_ARGC_SRC0C_YZX          23
1178 #       define R300_FPI0_ARGC_SRC1C_YZX          24
1179 #       define R300_FPI0_ARGC_SRC2C_YZX          25
1180 #       define R300_FPI0_ARGC_SRC0C_ZXY          26
1181 #       define R300_FPI0_ARGC_SRC1C_ZXY          27
1182 #       define R300_FPI0_ARGC_SRC2C_ZXY          28
1183 #       define R300_FPI0_ARGC_SRC0CA_WZY         29
1184 #       define R300_FPI0_ARGC_SRC1CA_WZY         30
1185 #       define R300_FPI0_ARGC_SRC2CA_WZY         31
1186 
1187 #       define R300_FPI0_ARG0C_SHIFT             0
1188 #       define R300_FPI0_ARG0C_MASK              (31 << 0)
1189 #       define R300_FPI0_ARG0C_NEG               (1 << 5)
1190 #       define R300_FPI0_ARG0C_ABS               (1 << 6)
1191 #       define R300_FPI0_ARG1C_SHIFT             7
1192 #       define R300_FPI0_ARG1C_MASK              (31 << 7)
1193 #       define R300_FPI0_ARG1C_NEG               (1 << 12)
1194 #       define R300_FPI0_ARG1C_ABS               (1 << 13)
1195 #       define R300_FPI0_ARG2C_SHIFT             14
1196 #       define R300_FPI0_ARG2C_MASK              (31 << 14)
1197 #       define R300_FPI0_ARG2C_NEG               (1 << 19)
1198 #       define R300_FPI0_ARG2C_ABS               (1 << 20)
1199 #       define R300_FPI0_SPECIAL_LRP             (1 << 21)
1200 #       define R300_FPI0_OUTC_MAD                (0 << 23)
1201 #       define R300_FPI0_OUTC_DP3                (1 << 23)
1202 #       define R300_FPI0_OUTC_DP4                (2 << 23)
1203 #       define R300_FPI0_OUTC_MIN                (4 << 23)
1204 #       define R300_FPI0_OUTC_MAX                (5 << 23)
1205 #       define R300_FPI0_OUTC_CMPH               (7 << 23)
1206 #       define R300_FPI0_OUTC_CMP                (8 << 23)
1207 #       define R300_FPI0_OUTC_FRC                (9 << 23)
1208 #       define R300_FPI0_OUTC_REPL_ALPHA         (10 << 23)
1209 #       define R300_FPI0_OUTC_SAT                (1 << 30)
1210 #       define R300_FPI0_INSERT_NOP              (1 << 31)
1211 
1212 #define R300_PFS_INSTR2_0                   0x49C0
1213 #       define R300_FPI2_ARGA_SRC0C_X            0
1214 #       define R300_FPI2_ARGA_SRC0C_Y            1
1215 #       define R300_FPI2_ARGA_SRC0C_Z            2
1216 #       define R300_FPI2_ARGA_SRC1C_X            3
1217 #       define R300_FPI2_ARGA_SRC1C_Y            4
1218 #       define R300_FPI2_ARGA_SRC1C_Z            5
1219 #       define R300_FPI2_ARGA_SRC2C_X            6
1220 #       define R300_FPI2_ARGA_SRC2C_Y            7
1221 #       define R300_FPI2_ARGA_SRC2C_Z            8
1222 #       define R300_FPI2_ARGA_SRC0A              9
1223 #       define R300_FPI2_ARGA_SRC1A              10
1224 #       define R300_FPI2_ARGA_SRC2A              11
1225 #       define R300_FPI2_ARGA_SRC1A_LRP          15
1226 #       define R300_FPI2_ARGA_ZERO               16
1227 #       define R300_FPI2_ARGA_ONE                17
1228 	/* GUESS */
1229 #       define R300_FPI2_ARGA_HALF               18
1230 #       define R300_FPI2_ARG0A_SHIFT             0
1231 #       define R300_FPI2_ARG0A_MASK              (31 << 0)
1232 #       define R300_FPI2_ARG0A_NEG               (1 << 5)
1233 	/* GUESS */
1234 #	define R300_FPI2_ARG0A_ABS		 (1 << 6)
1235 #       define R300_FPI2_ARG1A_SHIFT             7
1236 #       define R300_FPI2_ARG1A_MASK              (31 << 7)
1237 #       define R300_FPI2_ARG1A_NEG               (1 << 12)
1238 	/* GUESS */
1239 #	define R300_FPI2_ARG1A_ABS		 (1 << 13)
1240 #       define R300_FPI2_ARG2A_SHIFT             14
1241 #       define R300_FPI2_ARG2A_MASK              (31 << 14)
1242 #       define R300_FPI2_ARG2A_NEG               (1 << 19)
1243 	/* GUESS */
1244 #	define R300_FPI2_ARG2A_ABS		 (1 << 20)
1245 #       define R300_FPI2_SPECIAL_LRP             (1 << 21)
1246 #       define R300_FPI2_OUTA_MAD                (0 << 23)
1247 #       define R300_FPI2_OUTA_DP4                (1 << 23)
1248 #       define R300_FPI2_OUTA_MIN                (2 << 23)
1249 #       define R300_FPI2_OUTA_MAX                (3 << 23)
1250 #       define R300_FPI2_OUTA_CMP                (6 << 23)
1251 #       define R300_FPI2_OUTA_FRC                (7 << 23)
1252 #       define R300_FPI2_OUTA_EX2                (8 << 23)
1253 #       define R300_FPI2_OUTA_LG2                (9 << 23)
1254 #       define R300_FPI2_OUTA_RCP                (10 << 23)
1255 #       define R300_FPI2_OUTA_RSQ                (11 << 23)
1256 #       define R300_FPI2_OUTA_SAT                (1 << 30)
1257 #       define R300_FPI2_UNKNOWN_31              (1 << 31)
1258 /* END: Fragment program instruction set */
1259 
1260 /* Fog state and color */
1261 #define R300_RE_FOG_STATE                   0x4BC0
1262 #       define R300_FOG_ENABLE                   (1 << 0)
1263 #	define R300_FOG_MODE_LINEAR              (0 << 1)
1264 #	define R300_FOG_MODE_EXP                 (1 << 1)
1265 #	define R300_FOG_MODE_EXP2                (2 << 1)
1266 #	define R300_FOG_MODE_MASK                (3 << 1)
1267 #define R300_FOG_COLOR_R                    0x4BC8
1268 #define R300_FOG_COLOR_G                    0x4BCC
1269 #define R300_FOG_COLOR_B                    0x4BD0
1270 
1271 #define R300_PP_ALPHA_TEST                  0x4BD4
1272 #       define R300_REF_ALPHA_MASK               0x000000ff
1273 #       define R300_ALPHA_TEST_FAIL              (0 << 8)
1274 #       define R300_ALPHA_TEST_LESS              (1 << 8)
1275 #       define R300_ALPHA_TEST_LEQUAL            (3 << 8)
1276 #       define R300_ALPHA_TEST_EQUAL             (2 << 8)
1277 #       define R300_ALPHA_TEST_GEQUAL            (6 << 8)
1278 #       define R300_ALPHA_TEST_GREATER           (4 << 8)
1279 #       define R300_ALPHA_TEST_NEQUAL            (5 << 8)
1280 #       define R300_ALPHA_TEST_PASS              (7 << 8)
1281 #       define R300_ALPHA_TEST_OP_MASK           (7 << 8)
1282 #       define R300_ALPHA_TEST_ENABLE            (1 << 11)
1283 
1284 /* gap */
1285 
1286 /* Fragment program parameters in 7.16 floating point */
1287 #define R300_PFS_PARAM_0_X                  0x4C00
1288 #define R300_PFS_PARAM_0_Y                  0x4C04
1289 #define R300_PFS_PARAM_0_Z                  0x4C08
1290 #define R300_PFS_PARAM_0_W                  0x4C0C
1291 /* GUESS: PARAM_31 is last, based on native limits reported by fglrx */
1292 #define R300_PFS_PARAM_31_X                 0x4DF0
1293 #define R300_PFS_PARAM_31_Y                 0x4DF4
1294 #define R300_PFS_PARAM_31_Z                 0x4DF8
1295 #define R300_PFS_PARAM_31_W                 0x4DFC
1296 
1297 /* Notes:
1298  * - AFAIK fglrx always sets BLEND_UNKNOWN when blending is used in
1299  *   the application
1300  * - AFAIK fglrx always sets BLEND_NO_SEPARATE when CBLEND and ABLEND
1301  *    are set to the same
1302  *   function (both registers are always set up completely in any case)
1303  * - Most blend flags are simply copied from R200 and not tested yet
1304  */
1305 #define R300_RB3D_CBLEND                    0x4E04
1306 #define R300_RB3D_ABLEND                    0x4E08
1307 /* the following only appear in CBLEND */
1308 #       define R300_BLEND_ENABLE                     (1 << 0)
1309 #       define R300_BLEND_UNKNOWN                    (3 << 1)
1310 #       define R300_BLEND_NO_SEPARATE                (1 << 3)
1311 /* the following are shared between CBLEND and ABLEND */
1312 #       define R300_FCN_MASK                         (3  << 12)
1313 #       define R300_COMB_FCN_ADD_CLAMP               (0  << 12)
1314 #       define R300_COMB_FCN_ADD_NOCLAMP             (1  << 12)
1315 #       define R300_COMB_FCN_SUB_CLAMP               (2  << 12)
1316 #       define R300_COMB_FCN_SUB_NOCLAMP             (3  << 12)
1317 #       define R300_COMB_FCN_MIN                     (4  << 12)
1318 #       define R300_COMB_FCN_MAX                     (5  << 12)
1319 #       define R300_COMB_FCN_RSUB_CLAMP              (6  << 12)
1320 #       define R300_COMB_FCN_RSUB_NOCLAMP            (7  << 12)
1321 #       define R300_BLEND_GL_ZERO                    (32)
1322 #       define R300_BLEND_GL_ONE                     (33)
1323 #       define R300_BLEND_GL_SRC_COLOR               (34)
1324 #       define R300_BLEND_GL_ONE_MINUS_SRC_COLOR     (35)
1325 #       define R300_BLEND_GL_DST_COLOR               (36)
1326 #       define R300_BLEND_GL_ONE_MINUS_DST_COLOR     (37)
1327 #       define R300_BLEND_GL_SRC_ALPHA               (38)
1328 #       define R300_BLEND_GL_ONE_MINUS_SRC_ALPHA     (39)
1329 #       define R300_BLEND_GL_DST_ALPHA               (40)
1330 #       define R300_BLEND_GL_ONE_MINUS_DST_ALPHA     (41)
1331 #       define R300_BLEND_GL_SRC_ALPHA_SATURATE      (42)
1332 #       define R300_BLEND_GL_CONST_COLOR             (43)
1333 #       define R300_BLEND_GL_ONE_MINUS_CONST_COLOR   (44)
1334 #       define R300_BLEND_GL_CONST_ALPHA             (45)
1335 #       define R300_BLEND_GL_ONE_MINUS_CONST_ALPHA   (46)
1336 #       define R300_BLEND_MASK                       (63)
1337 #       define R300_SRC_BLEND_SHIFT                  (16)
1338 #       define R300_DST_BLEND_SHIFT                  (24)
1339 #define R300_RB3D_BLEND_COLOR               0x4E10
1340 #define R300_RB3D_COLORMASK                 0x4E0C
1341 #       define R300_COLORMASK0_B                 (1<<0)
1342 #       define R300_COLORMASK0_G                 (1<<1)
1343 #       define R300_COLORMASK0_R                 (1<<2)
1344 #       define R300_COLORMASK0_A                 (1<<3)
1345 
1346 /* gap */
1347 
1348 #define R300_RB3D_COLOROFFSET0              0x4E28
1349 #       define R300_COLOROFFSET_MASK             0xFFFFFFF0 /* GUESS */
1350 #define R300_RB3D_COLOROFFSET1              0x4E2C /* GUESS */
1351 #define R300_RB3D_COLOROFFSET2              0x4E30 /* GUESS */
1352 #define R300_RB3D_COLOROFFSET3              0x4E34 /* GUESS */
1353 
1354 /* gap */
1355 
1356 /* Bit 16: Larger tiles
1357  * Bit 17: 4x2 tiles
1358  * Bit 18: Extremely weird tile like, but some pixels duplicated?
1359  */
1360 #define R300_RB3D_COLORPITCH0               0x4E38
1361 #       define R300_COLORPITCH_MASK              0x00001FF8 /* GUESS */
1362 #       define R300_COLOR_TILE_ENABLE            (1 << 16) /* GUESS */
1363 #       define R300_COLOR_MICROTILE_ENABLE       (1 << 17) /* GUESS */
1364 #       define R300_COLOR_MICROTILE_SQUARE_ENABLE (2 << 17)
1365 #       define R300_COLOR_ENDIAN_NO_SWAP         (0 << 18) /* GUESS */
1366 #       define R300_COLOR_ENDIAN_WORD_SWAP       (1 << 18) /* GUESS */
1367 #       define R300_COLOR_ENDIAN_DWORD_SWAP      (2 << 18) /* GUESS */
1368 #       define R300_COLOR_FORMAT_RGB565          (2 << 22)
1369 #       define R300_COLOR_FORMAT_ARGB8888        (3 << 22)
1370 #define R300_RB3D_COLORPITCH1               0x4E3C /* GUESS */
1371 #define R300_RB3D_COLORPITCH2               0x4E40 /* GUESS */
1372 #define R300_RB3D_COLORPITCH3               0x4E44 /* GUESS */
1373 
1374 #define R300_RB3D_AARESOLVE_OFFSET          0x4E80
1375 #define R300_RB3D_AARESOLVE_PITCH           0x4E84
1376 #define R300_RB3D_AARESOLVE_CTL             0x4E88
1377 /* gap */
1378 
1379 /* Guess by Vladimir.
1380  * Set to 0A before 3D operations, set to 02 afterwards.
1381  */
1382 /*#define R300_RB3D_DSTCACHE_CTLSTAT          0x4E4C*/
1383 #       define R300_RB3D_DSTCACHE_UNKNOWN_02             0x00000002
1384 #       define R300_RB3D_DSTCACHE_UNKNOWN_0A             0x0000000A
1385 
1386 /* gap */
1387 /* There seems to be no "write only" setting, so use Z-test = ALWAYS
1388  * for this.
1389  * Bit (1<<8) is the "test" bit. so plain write is 6  - vd
1390  */
1391 #define R300_ZB_CNTL                             0x4F00
1392 #	define R300_STENCIL_ENABLE		 (1 << 0)
1393 #	define R300_Z_ENABLE		         (1 << 1)
1394 #	define R300_Z_WRITE_ENABLE		 (1 << 2)
1395 #	define R300_Z_SIGNED_COMPARE		 (1 << 3)
1396 #	define R300_STENCIL_FRONT_BACK		 (1 << 4)
1397 
1398 #define R300_ZB_ZSTENCILCNTL                   0x4f04
1399 	/* functions */
1400 #	define R300_ZS_NEVER			0
1401 #	define R300_ZS_LESS			1
1402 #	define R300_ZS_LEQUAL			2
1403 #	define R300_ZS_EQUAL			3
1404 #	define R300_ZS_GEQUAL			4
1405 #	define R300_ZS_GREATER			5
1406 #	define R300_ZS_NOTEQUAL			6
1407 #	define R300_ZS_ALWAYS			7
1408 #       define R300_ZS_MASK                     7
1409 	/* operations */
1410 #	define R300_ZS_KEEP			0
1411 #	define R300_ZS_ZERO			1
1412 #	define R300_ZS_REPLACE			2
1413 #	define R300_ZS_INCR			3
1414 #	define R300_ZS_DECR			4
1415 #	define R300_ZS_INVERT			5
1416 #	define R300_ZS_INCR_WRAP		6
1417 #	define R300_ZS_DECR_WRAP		7
1418 #	define R300_Z_FUNC_SHIFT		0
1419 	/* front and back refer to operations done for front
1420 	   and back faces, i.e. separate stencil function support */
1421 #	define R300_S_FRONT_FUNC_SHIFT	        3
1422 #	define R300_S_FRONT_SFAIL_OP_SHIFT	6
1423 #	define R300_S_FRONT_ZPASS_OP_SHIFT	9
1424 #	define R300_S_FRONT_ZFAIL_OP_SHIFT      12
1425 #	define R300_S_BACK_FUNC_SHIFT           15
1426 #	define R300_S_BACK_SFAIL_OP_SHIFT       18
1427 #	define R300_S_BACK_ZPASS_OP_SHIFT       21
1428 #	define R300_S_BACK_ZFAIL_OP_SHIFT       24
1429 
1430 #define R300_ZB_STENCILREFMASK                        0x4f08
1431 #	define R300_STENCILREF_SHIFT       0
1432 #	define R300_STENCILREF_MASK        0x000000ff
1433 #	define R300_STENCILMASK_SHIFT      8
1434 #	define R300_STENCILMASK_MASK       0x0000ff00
1435 #	define R300_STENCILWRITEMASK_SHIFT 16
1436 #	define R300_STENCILWRITEMASK_MASK  0x00ff0000
1437 
1438 /* gap */
1439 
1440 #define R300_ZB_FORMAT                             0x4f10
1441 #	define R300_DEPTHFORMAT_16BIT_INT_Z   (0 << 0)
1442 #	define R300_DEPTHFORMAT_16BIT_13E3    (1 << 0)
1443 #	define R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL   (2 << 0)
1444 /* reserved up to (15 << 0) */
1445 #	define R300_INVERT_13E3_LEADING_ONES  (0 << 4)
1446 #	define R300_INVERT_13E3_LEADING_ZEROS (1 << 4)
1447 
1448 #define R300_ZB_ZTOP                             0x4F14
1449 #	define R300_ZTOP_DISABLE                 (0 << 0)
1450 #	define R300_ZTOP_ENABLE                  (1 << 0)
1451 
1452 /* gap */
1453 
1454 #define R300_ZB_ZCACHE_CTLSTAT            0x4f18
1455 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_NO_EFFECT      (0 << 0)
1456 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE (1 << 0)
1457 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_NO_EFFECT       (0 << 1)
1458 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE            (1 << 1)
1459 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_IDLE            (0 << 31)
1460 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_BUSY            (1 << 31)
1461 
1462 #define R300_ZB_BW_CNTL                     0x4f1c
1463 #	define R300_HIZ_DISABLE                              (0 << 0)
1464 #	define R300_HIZ_ENABLE                               (1 << 0)
1465 #	define R300_HIZ_MIN                                  (0 << 1)
1466 #	define R300_HIZ_MAX                                  (1 << 1)
1467 #	define R300_FAST_FILL_DISABLE                        (0 << 2)
1468 #	define R300_FAST_FILL_ENABLE                         (1 << 2)
1469 #	define R300_RD_COMP_DISABLE                          (0 << 3)
1470 #	define R300_RD_COMP_ENABLE                           (1 << 3)
1471 #	define R300_WR_COMP_DISABLE                          (0 << 4)
1472 #	define R300_WR_COMP_ENABLE                           (1 << 4)
1473 #	define R300_ZB_CB_CLEAR_RMW                          (0 << 5)
1474 #	define R300_ZB_CB_CLEAR_CACHE_LINEAR                 (1 << 5)
1475 #	define R300_FORCE_COMPRESSED_STENCIL_VALUE_DISABLE   (0 << 6)
1476 #	define R300_FORCE_COMPRESSED_STENCIL_VALUE_ENABLE    (1 << 6)
1477 
1478 #	define R500_ZEQUAL_OPTIMIZE_ENABLE                   (0 << 7)
1479 #	define R500_ZEQUAL_OPTIMIZE_DISABLE                  (1 << 7)
1480 #	define R500_SEQUAL_OPTIMIZE_ENABLE                   (0 << 8)
1481 #	define R500_SEQUAL_OPTIMIZE_DISABLE                  (1 << 8)
1482 
1483 #	define R500_BMASK_ENABLE                             (0 << 10)
1484 #	define R500_BMASK_DISABLE                            (1 << 10)
1485 #	define R500_HIZ_EQUAL_REJECT_DISABLE                 (0 << 11)
1486 #	define R500_HIZ_EQUAL_REJECT_ENABLE                  (1 << 11)
1487 #	define R500_HIZ_FP_EXP_BITS_DISABLE                  (0 << 12)
1488 #	define R500_HIZ_FP_EXP_BITS_1                        (1 << 12)
1489 #	define R500_HIZ_FP_EXP_BITS_2                        (2 << 12)
1490 #	define R500_HIZ_FP_EXP_BITS_3                        (3 << 12)
1491 #	define R500_HIZ_FP_EXP_BITS_4                        (4 << 12)
1492 #	define R500_HIZ_FP_EXP_BITS_5                        (5 << 12)
1493 #	define R500_HIZ_FP_INVERT_LEADING_ONES               (0 << 15)
1494 #	define R500_HIZ_FP_INVERT_LEADING_ZEROS              (1 << 15)
1495 #	define R500_TILE_OVERWRITE_RECOMPRESSION_ENABLE      (0 << 16)
1496 #	define R500_TILE_OVERWRITE_RECOMPRESSION_DISABLE     (1 << 16)
1497 #	define R500_CONTIGUOUS_6XAA_SAMPLES_ENABLE           (0 << 17)
1498 #	define R500_CONTIGUOUS_6XAA_SAMPLES_DISABLE          (1 << 17)
1499 #	define R500_PEQ_PACKING_DISABLE                      (0 << 18)
1500 #	define R500_PEQ_PACKING_ENABLE                       (1 << 18)
1501 #	define R500_COVERED_PTR_MASKING_DISABLE              (0 << 18)
1502 #	define R500_COVERED_PTR_MASKING_ENABLE               (1 << 18)
1503 
1504 
1505 /* gap */
1506 
1507 /* Z Buffer Address Offset.
1508  * Bits 31 to 5 are used for aligned Z buffer address offset for macro tiles.
1509  */
1510 #define R300_ZB_DEPTHOFFSET               0x4f20
1511 
1512 /* Z Buffer Pitch and Endian Control */
1513 #define R300_ZB_DEPTHPITCH                0x4f24
1514 #       define R300_DEPTHPITCH_MASK              0x00003FFC
1515 #       define R300_DEPTHMACROTILE_DISABLE      (0 << 16)
1516 #       define R300_DEPTHMACROTILE_ENABLE       (1 << 16)
1517 #       define R300_DEPTHMICROTILE_LINEAR       (0 << 17)
1518 #       define R300_DEPTHMICROTILE_TILED        (1 << 17)
1519 #       define R300_DEPTHMICROTILE_TILED_SQUARE (2 << 17)
1520 #       define R300_DEPTHENDIAN_NO_SWAP         (0 << 18)
1521 #       define R300_DEPTHENDIAN_WORD_SWAP       (1 << 18)
1522 #       define R300_DEPTHENDIAN_DWORD_SWAP      (2 << 18)
1523 #       define R300_DEPTHENDIAN_HALF_DWORD_SWAP (3 << 18)
1524 
1525 /* Z Buffer Clear Value */
1526 #define R300_ZB_DEPTHCLEARVALUE                  0x4f28
1527 
1528 #define R300_ZB_ZMASK_OFFSET			 0x4f30
1529 #define R300_ZB_ZMASK_PITCH			 0x4f34
1530 #define R300_ZB_ZMASK_WRINDEX			 0x4f38
1531 #define R300_ZB_ZMASK_DWORD			 0x4f3c
1532 #define R300_ZB_ZMASK_RDINDEX			 0x4f40
1533 
1534 /* Hierarchical Z Memory Offset */
1535 #define R300_ZB_HIZ_OFFSET                       0x4f44
1536 
1537 /* Hierarchical Z Write Index */
1538 #define R300_ZB_HIZ_WRINDEX                      0x4f48
1539 
1540 /* Hierarchical Z Data */
1541 #define R300_ZB_HIZ_DWORD                        0x4f4c
1542 
1543 /* Hierarchical Z Read Index */
1544 #define R300_ZB_HIZ_RDINDEX                      0x4f50
1545 
1546 /* Hierarchical Z Pitch */
1547 #define R300_ZB_HIZ_PITCH                        0x4f54
1548 
1549 /* Z Buffer Z Pass Counter Data */
1550 #define R300_ZB_ZPASS_DATA                       0x4f58
1551 
1552 /* Z Buffer Z Pass Counter Address */
1553 #define R300_ZB_ZPASS_ADDR                       0x4f5c
1554 
1555 /* Depth buffer X and Y coordinate offset */
1556 #define R300_ZB_DEPTHXY_OFFSET                   0x4f60
1557 #	define R300_DEPTHX_OFFSET_SHIFT  1
1558 #	define R300_DEPTHX_OFFSET_MASK   0x000007FE
1559 #	define R300_DEPTHY_OFFSET_SHIFT  17
1560 #	define R300_DEPTHY_OFFSET_MASK   0x07FE0000
1561 
1562 /* Sets the fifo sizes */
1563 #define R500_ZB_FIFO_SIZE                        0x4fd0
1564 #	define R500_OP_FIFO_SIZE_FULL   (0 << 0)
1565 #	define R500_OP_FIFO_SIZE_HALF   (1 << 0)
1566 #	define R500_OP_FIFO_SIZE_QUATER (2 << 0)
1567 #	define R500_OP_FIFO_SIZE_EIGTHS (4 << 0)
1568 
1569 /* Stencil Reference Value and Mask for backfacing quads */
1570 /* R300_ZB_STENCILREFMASK handles front face */
1571 #define R500_ZB_STENCILREFMASK_BF                0x4fd4
1572 #	define R500_STENCILREF_SHIFT       0
1573 #	define R500_STENCILREF_MASK        0x000000ff
1574 #	define R500_STENCILMASK_SHIFT      8
1575 #	define R500_STENCILMASK_MASK       0x0000ff00
1576 #	define R500_STENCILWRITEMASK_SHIFT 16
1577 #	define R500_STENCILWRITEMASK_MASK  0x00ff0000
1578 
1579 /* BEGIN: Vertex program instruction set */
1580 
1581 /* Every instruction is four dwords long:
1582  *  DWORD 0: output and opcode
1583  *  DWORD 1: first argument
1584  *  DWORD 2: second argument
1585  *  DWORD 3: third argument
1586  *
1587  * Notes:
1588  *  - ABS r, a is implemented as MAX r, a, -a
1589  *  - MOV is implemented as ADD to zero
1590  *  - XPD is implemented as MUL + MAD
1591  *  - FLR is implemented as FRC + ADD
1592  *  - apparently, fglrx tries to schedule instructions so that there is at
1593  *    least one instruction between the write to a temporary and the first
1594  *    read from said temporary; however, violations of this scheduling are
1595  *    allowed
1596  *  - register indices seem to be unrelated with OpenGL aliasing to
1597  *    conventional state
1598  *  - only one attribute and one parameter can be loaded at a time; however,
1599  *    the same attribute/parameter can be used for more than one argument
1600  *  - the second software argument for POW is the third hardware argument
1601  *    (no idea why)
1602  *  - MAD with only temporaries as input seems to use VPI_OUT_SELECT_MAD_2
1603  *
1604  * There is some magic surrounding LIT:
1605  *   The single argument is replicated across all three inputs, but swizzled:
1606  *     First argument: xyzy
1607  *     Second argument: xyzx
1608  *     Third argument: xyzw
1609  *   Whenever the result is used later in the fragment program, fglrx forces
1610  *   x and w to be 1.0 in the input selection; I don't know whether this is
1611  *   strictly necessary
1612  */
1613 #define R300_VPI_OUT_OP_DOT                     (1 << 0)
1614 #define R300_VPI_OUT_OP_MUL                     (2 << 0)
1615 #define R300_VPI_OUT_OP_ADD                     (3 << 0)
1616 #define R300_VPI_OUT_OP_MAD                     (4 << 0)
1617 #define R300_VPI_OUT_OP_DST                     (5 << 0)
1618 #define R300_VPI_OUT_OP_FRC                     (6 << 0)
1619 #define R300_VPI_OUT_OP_MAX                     (7 << 0)
1620 #define R300_VPI_OUT_OP_MIN                     (8 << 0)
1621 #define R300_VPI_OUT_OP_SGE                     (9 << 0)
1622 #define R300_VPI_OUT_OP_SLT                     (10 << 0)
1623 	/* Used in GL_POINT_DISTANCE_ATTENUATION_ARB, vector(scalar, vector) */
1624 #define R300_VPI_OUT_OP_UNK12                   (12 << 0)
1625 #define R300_VPI_OUT_OP_ARL                     (13 << 0)
1626 #define R300_VPI_OUT_OP_EXP                     (65 << 0)
1627 #define R300_VPI_OUT_OP_LOG                     (66 << 0)
1628 	/* Used in fog computations, scalar(scalar) */
1629 #define R300_VPI_OUT_OP_UNK67                   (67 << 0)
1630 #define R300_VPI_OUT_OP_LIT                     (68 << 0)
1631 #define R300_VPI_OUT_OP_POW                     (69 << 0)
1632 #define R300_VPI_OUT_OP_RCP                     (70 << 0)
1633 #define R300_VPI_OUT_OP_RSQ                     (72 << 0)
1634 	/* Used in GL_POINT_DISTANCE_ATTENUATION_ARB, scalar(scalar) */
1635 #define R300_VPI_OUT_OP_UNK73                   (73 << 0)
1636 #define R300_VPI_OUT_OP_EX2                     (75 << 0)
1637 #define R300_VPI_OUT_OP_LG2                     (76 << 0)
1638 #define R300_VPI_OUT_OP_MAD_2                   (128 << 0)
1639 	/* all temps, vector(scalar, vector, vector) */
1640 #define R300_VPI_OUT_OP_UNK129                  (129 << 0)
1641 
1642 #define R300_VPI_OUT_REG_CLASS_TEMPORARY        (0 << 8)
1643 #define R300_VPI_OUT_REG_CLASS_ADDR             (1 << 8)
1644 #define R300_VPI_OUT_REG_CLASS_RESULT           (2 << 8)
1645 #define R300_VPI_OUT_REG_CLASS_MASK             (31 << 8)
1646 
1647 #define R300_VPI_OUT_REG_INDEX_SHIFT            13
1648 	/* GUESS based on fglrx native limits */
1649 #define R300_VPI_OUT_REG_INDEX_MASK             (31 << 13)
1650 
1651 #define R300_VPI_OUT_WRITE_X                    (1 << 20)
1652 #define R300_VPI_OUT_WRITE_Y                    (1 << 21)
1653 #define R300_VPI_OUT_WRITE_Z                    (1 << 22)
1654 #define R300_VPI_OUT_WRITE_W                    (1 << 23)
1655 
1656 #define R300_VPI_IN_REG_CLASS_TEMPORARY         (0 << 0)
1657 #define R300_VPI_IN_REG_CLASS_ATTRIBUTE         (1 << 0)
1658 #define R300_VPI_IN_REG_CLASS_PARAMETER         (2 << 0)
1659 #define R300_VPI_IN_REG_CLASS_NONE              (9 << 0)
1660 #define R300_VPI_IN_REG_CLASS_MASK              (31 << 0)
1661 
1662 #define R300_VPI_IN_REG_INDEX_SHIFT             5
1663 	/* GUESS based on fglrx native limits */
1664 #define R300_VPI_IN_REG_INDEX_MASK              (255 << 5)
1665 
1666 /* The R300 can select components from the input register arbitrarily.
1667  * Use the following constants, shifted by the component shift you
1668  * want to select
1669  */
1670 #define R300_VPI_IN_SELECT_X    0
1671 #define R300_VPI_IN_SELECT_Y    1
1672 #define R300_VPI_IN_SELECT_Z    2
1673 #define R300_VPI_IN_SELECT_W    3
1674 #define R300_VPI_IN_SELECT_ZERO 4
1675 #define R300_VPI_IN_SELECT_ONE  5
1676 #define R300_VPI_IN_SELECT_MASK 7
1677 
1678 #define R300_VPI_IN_X_SHIFT                     13
1679 #define R300_VPI_IN_Y_SHIFT                     16
1680 #define R300_VPI_IN_Z_SHIFT                     19
1681 #define R300_VPI_IN_W_SHIFT                     22
1682 
1683 #define R300_VPI_IN_NEG_X                       (1 << 25)
1684 #define R300_VPI_IN_NEG_Y                       (1 << 26)
1685 #define R300_VPI_IN_NEG_Z                       (1 << 27)
1686 #define R300_VPI_IN_NEG_W                       (1 << 28)
1687 /* END: Vertex program instruction set */
1688 
1689 /* BEGIN: Packet 3 commands */
1690 
1691 /* A primitive emission dword. */
1692 #define R300_PRIM_TYPE_NONE                     (0 << 0)
1693 #define R300_PRIM_TYPE_POINT                    (1 << 0)
1694 #define R300_PRIM_TYPE_LINE                     (2 << 0)
1695 #define R300_PRIM_TYPE_LINE_STRIP               (3 << 0)
1696 #define R300_PRIM_TYPE_TRI_LIST                 (4 << 0)
1697 #define R300_PRIM_TYPE_TRI_FAN                  (5 << 0)
1698 #define R300_PRIM_TYPE_TRI_STRIP                (6 << 0)
1699 #define R300_PRIM_TYPE_TRI_TYPE2                (7 << 0)
1700 #define R300_PRIM_TYPE_RECT_LIST                (8 << 0)
1701 #define R300_PRIM_TYPE_3VRT_POINT_LIST          (9 << 0)
1702 #define R300_PRIM_TYPE_3VRT_LINE_LIST           (10 << 0)
1703 	/* GUESS (based on r200) */
1704 #define R300_PRIM_TYPE_POINT_SPRITES            (11 << 0)
1705 #define R300_PRIM_TYPE_LINE_LOOP                (12 << 0)
1706 #define R300_PRIM_TYPE_QUADS                    (13 << 0)
1707 #define R300_PRIM_TYPE_QUAD_STRIP               (14 << 0)
1708 #define R300_PRIM_TYPE_POLYGON                  (15 << 0)
1709 #define R300_PRIM_TYPE_MASK                     0xF
1710 #define R300_PRIM_WALK_IND                      (1 << 4)
1711 #define R300_PRIM_WALK_LIST                     (2 << 4)
1712 #define R300_PRIM_WALK_RING                     (3 << 4)
1713 #define R300_PRIM_WALK_MASK                     (3 << 4)
1714 	/* GUESS (based on r200) */
1715 #define R300_PRIM_COLOR_ORDER_BGRA              (0 << 6)
1716 #define R300_PRIM_COLOR_ORDER_RGBA              (1 << 6)
1717 #define R300_PRIM_NUM_VERTICES_SHIFT            16
1718 #define R300_PRIM_NUM_VERTICES_MASK             0xffff
1719 
1720 /* Draw a primitive from vertex data in arrays loaded via 3D_LOAD_VBPNTR.
1721  * Two parameter dwords:
1722  * 0. The first parameter appears to be always 0
1723  * 1. The second parameter is a standard primitive emission dword.
1724  */
1725 #define R300_PACKET3_3D_DRAW_VBUF           0x00002800
1726 
1727 /* Specify the full set of vertex arrays as (address, stride).
1728  * The first parameter is the number of vertex arrays specified.
1729  * The rest of the command is a variable length list of blocks, where
1730  * each block is three dwords long and specifies two arrays.
1731  * The first dword of a block is split into two words, the lower significant
1732  * word refers to the first array, the more significant word to the second
1733  * array in the block.
1734  * The low byte of each word contains the size of an array entry in dwords,
1735  * the high byte contains the stride of the array.
1736  * The second dword of a block contains the pointer to the first array,
1737  * the third dword of a block contains the pointer to the second array.
1738  * Note that if the total number of arrays is odd, the third dword of
1739  * the last block is omitted.
1740  */
1741 #define R300_PACKET3_3D_LOAD_VBPNTR         0x00002F00
1742 
1743 #define R300_PACKET3_INDX_BUFFER            0x00003300
1744 #    define R300_EB_UNK1_SHIFT                      24
1745 #    define R300_EB_UNK1                    (0x80<<24)
1746 #    define R300_EB_UNK2                        0x0810
1747 #define R300_PACKET3_3D_DRAW_VBUF_2         0x00003400
1748 #define R300_PACKET3_3D_DRAW_INDX_2         0x00003600
1749 
1750 /* END: Packet 3 commands */
1751 
1752 
1753 /* Color formats for 2d packets
1754  */
1755 #define R300_CP_COLOR_FORMAT_CI8	2
1756 #define R300_CP_COLOR_FORMAT_ARGB1555	3
1757 #define R300_CP_COLOR_FORMAT_RGB565	4
1758 #define R300_CP_COLOR_FORMAT_ARGB8888	6
1759 #define R300_CP_COLOR_FORMAT_RGB332	7
1760 #define R300_CP_COLOR_FORMAT_RGB8	9
1761 #define R300_CP_COLOR_FORMAT_ARGB4444	15
1762 
1763 /*
1764  * CP type-3 packets
1765  */
1766 #define R300_CP_CMD_BITBLT_MULTI	0xC0009B00
1767 
1768 #define R500_VAP_INDEX_OFFSET		0x208c
1769 
1770 #define R500_GA_US_VECTOR_INDEX         0x4250
1771 #define R500_GA_US_VECTOR_DATA          0x4254
1772 
1773 #define R500_RS_IP_0                    0x4074
1774 #define R500_RS_INST_0                  0x4320
1775 
1776 #define R500_US_CONFIG                  0x4600
1777 
1778 #define R500_US_FC_CTRL			0x4624
1779 #define R500_US_CODE_ADDR		0x4630
1780 
1781 #define R500_RB3D_COLOR_CLEAR_VALUE_AR  0x46c0
1782 #define R500_RB3D_CONSTANT_COLOR_AR     0x4ef8
1783 
1784 #define R300_SU_REG_DEST                0x42c8
1785 #define RV530_FG_ZBREG_DEST             0x4be8
1786 #define R300_ZB_ZPASS_DATA              0x4f58
1787 #define R300_ZB_ZPASS_ADDR              0x4f5c
1788 
1789 #endif /* _R300_REG_H */
1790