r300_reg.h revision 261455
1251599Sdas/*
2141296Sdas * Copyright 2005 Nicolai Haehnle et al.
3141296Sdas * Copyright 2008 Advanced Micro Devices, Inc.
42116Sjkh * Copyright 2009 Jerome Glisse.
52116Sjkh *
62116Sjkh * Permission is hereby granted, free of charge, to any person obtaining a
72116Sjkh * copy of this software and associated documentation files (the "Software"),
8141296Sdas * to deal in the Software without restriction, including without limitation
92116Sjkh * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10141296Sdas * and/or sell copies of the Software, and to permit persons to whom the
112116Sjkh * Software is furnished to do so, subject to the following conditions:
122116Sjkh *
13141296Sdas * The above copyright notice and this permission notice shall be included in
142116Sjkh * all copies or substantial portions of the Software.
152116Sjkh *
16176451Sdas * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17176451Sdas * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
182116Sjkh * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
19251599Sdas * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
20251599Sdas * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
212116Sjkh * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
22251599Sdas * OTHER DEALINGS IN THE SOFTWARE.
23251599Sdas *
242116Sjkh * Authors: Nicolai Haehnle
252116Sjkh *          Jerome Glisse
26251599Sdas */
27251599Sdas#ifndef _R300_REG_H_
28251599Sdas#define _R300_REG_H_
29251599Sdas
30251599Sdas#include <sys/cdefs.h>
31251599Sdas__FBSDID("$FreeBSD: stable/10/sys/dev/drm2/radeon/r300_reg.h 261455 2014-02-04 03:36:42Z eadler $");
322116Sjkh
332116Sjkh#define R300_SURF_TILE_MACRO (1<<16)
342116Sjkh#define R300_SURF_TILE_MICRO (2<<16)
35251599Sdas#define R300_SURF_TILE_BOTH (3<<16)
36251599Sdas
37251599Sdas
38251599Sdas#define R300_MC_INIT_MISC_LAT_TIMER	0x180
39251599Sdas#	define R300_MC_MISC__MC_CPR_INIT_LAT_SHIFT	0
40251599Sdas#	define R300_MC_MISC__MC_VF_INIT_LAT_SHIFT	4
41251599Sdas#	define R300_MC_MISC__MC_DISP0R_INIT_LAT_SHIFT	8
42251599Sdas#	define R300_MC_MISC__MC_DISP1R_INIT_LAT_SHIFT	12
43251599Sdas#	define R300_MC_MISC__MC_FIXED_INIT_LAT_SHIFT	16
44251599Sdas#	define R300_MC_MISC__MC_E2R_INIT_LAT_SHIFT	20
45251599Sdas#	define R300_MC_MISC__MC_SAME_PAGE_PRIO_SHIFT	24
46251599Sdas#	define R300_MC_MISC__MC_GLOBW_INIT_LAT_SHIFT	28
47251599Sdas
48251599Sdas#define R300_MC_INIT_GFX_LAT_TIMER	0x154
49251599Sdas#	define R300_MC_MISC__MC_G3D0R_INIT_LAT_SHIFT	0
50251599Sdas#	define R300_MC_MISC__MC_G3D1R_INIT_LAT_SHIFT	4
518870Srgrimes#	define R300_MC_MISC__MC_G3D2R_INIT_LAT_SHIFT	8
52251599Sdas#	define R300_MC_MISC__MC_G3D3R_INIT_LAT_SHIFT	12
532116Sjkh#	define R300_MC_MISC__MC_TX0R_INIT_LAT_SHIFT	16
54251599Sdas#	define R300_MC_MISC__MC_TX1R_INIT_LAT_SHIFT	20
55251599Sdas#	define R300_MC_MISC__MC_GLOBR_INIT_LAT_SHIFT	24
56251599Sdas#	define R300_MC_MISC__MC_GLOBW_FULL_LAT_SHIFT	28
57251599Sdas
58251599Sdas/*
59251599Sdas * This file contains registers and constants for the R300. They have been
60251599Sdas * found mostly by examining command buffers captured using glxtest, as well
61251599Sdas * as by extrapolating some known registers and constants from the R200.
62251599Sdas * I am fairly certain that they are correct unless stated otherwise
63251599Sdas * in comments.
64251599Sdas */
65251599Sdas
66251599Sdas#define R300_SE_VPORT_XSCALE                0x1D98
678870Srgrimes#define R300_SE_VPORT_XOFFSET               0x1D9C
68251599Sdas#define R300_SE_VPORT_YSCALE                0x1DA0
69251599Sdas#define R300_SE_VPORT_YOFFSET               0x1DA4
70251599Sdas#define R300_SE_VPORT_ZSCALE                0x1DA8
71251599Sdas#define R300_SE_VPORT_ZOFFSET               0x1DAC
72251599Sdas
73251599Sdas
74251599Sdas/*
75251599Sdas * Vertex Array Processing (VAP) Control
76251599Sdas * Stolen from r200 code from Christoph Brill (It's a guess!)
77251599Sdas */
78141296Sdas#define R300_VAP_CNTL	0x2080
79251599Sdas
80251599Sdas/* This register is written directly and also starts data section
81251599Sdas * in many 3d CP_PACKET3's
82251599Sdas */
832116Sjkh#define R300_VAP_VF_CNTL	0x2084
84251599Sdas#	define	R300_VAP_VF_CNTL__PRIM_TYPE__SHIFT              0
852116Sjkh#	define  R300_VAP_VF_CNTL__PRIM_NONE                     (0<<0)
862116Sjkh#	define  R300_VAP_VF_CNTL__PRIM_POINTS                   (1<<0)
87251599Sdas#	define  R300_VAP_VF_CNTL__PRIM_LINES                    (2<<0)
882116Sjkh#	define  R300_VAP_VF_CNTL__PRIM_LINE_STRIP               (3<<0)
892116Sjkh#	define  R300_VAP_VF_CNTL__PRIM_TRIANGLES                (4<<0)
90#	define  R300_VAP_VF_CNTL__PRIM_TRIANGLE_FAN             (5<<0)
91#	define  R300_VAP_VF_CNTL__PRIM_TRIANGLE_STRIP           (6<<0)
92#	define  R300_VAP_VF_CNTL__PRIM_LINE_LOOP                (12<<0)
93#	define  R300_VAP_VF_CNTL__PRIM_QUADS                    (13<<0)
94#	define  R300_VAP_VF_CNTL__PRIM_QUAD_STRIP               (14<<0)
95#	define  R300_VAP_VF_CNTL__PRIM_POLYGON                  (15<<0)
96
97#	define	R300_VAP_VF_CNTL__PRIM_WALK__SHIFT              4
98	/* State based - direct writes to registers trigger vertex
99           generation */
100#	define	R300_VAP_VF_CNTL__PRIM_WALK_STATE_BASED         (0<<4)
101#	define	R300_VAP_VF_CNTL__PRIM_WALK_INDICES             (1<<4)
102#	define	R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST         (2<<4)
103#	define	R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_EMBEDDED     (3<<4)
104
105	/* I don't think I saw these three used.. */
106#	define	R300_VAP_VF_CNTL__COLOR_ORDER__SHIFT            6
107#	define	R300_VAP_VF_CNTL__TCL_OUTPUT_CTL_ENA__SHIFT     9
108#	define	R300_VAP_VF_CNTL__PROG_STREAM_ENA__SHIFT        10
109
110	/* index size - when not set the indices are assumed to be 16 bit */
111#	define	R300_VAP_VF_CNTL__INDEX_SIZE_32bit              (1<<11)
112	/* number of vertices */
113#	define	R300_VAP_VF_CNTL__NUM_VERTICES__SHIFT           16
114
115/* BEGIN: Wild guesses */
116#define R300_VAP_OUTPUT_VTX_FMT_0           0x2090
117#       define R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT     (1<<0)
118#       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT   (1<<1)
119#       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT (1<<2)  /* GUESS */
120#       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT (1<<3)  /* GUESS */
121#       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT (1<<4)  /* GUESS */
122#       define R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT (1<<16) /* GUESS */
123
124#define R300_VAP_OUTPUT_VTX_FMT_1           0x2094
125	/* each of the following is 3 bits wide, specifies number
126	   of components */
127#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT 0
128#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT 3
129#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT 6
130#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT 9
131#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT 12
132#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT 15
133#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT 18
134#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT 21
135/* END: Wild guesses */
136
137#define R300_SE_VTE_CNTL                  0x20b0
138#	define     R300_VPORT_X_SCALE_ENA                0x00000001
139#	define     R300_VPORT_X_OFFSET_ENA               0x00000002
140#	define     R300_VPORT_Y_SCALE_ENA                0x00000004
141#	define     R300_VPORT_Y_OFFSET_ENA               0x00000008
142#	define     R300_VPORT_Z_SCALE_ENA                0x00000010
143#	define     R300_VPORT_Z_OFFSET_ENA               0x00000020
144#	define     R300_VTX_XY_FMT                       0x00000100
145#	define     R300_VTX_Z_FMT                        0x00000200
146#	define     R300_VTX_W0_FMT                       0x00000400
147#	define     R300_VTX_W0_NORMALIZE                 0x00000800
148#	define     R300_VTX_ST_DENORMALIZED              0x00001000
149
150/* BEGIN: Vertex data assembly - lots of uncertainties */
151
152/* gap */
153
154#define R300_VAP_CNTL_STATUS              0x2140
155#	define R300_VC_NO_SWAP                  (0 << 0)
156#	define R300_VC_16BIT_SWAP               (1 << 0)
157#	define R300_VC_32BIT_SWAP               (2 << 0)
158#	define R300_VAP_TCL_BYPASS		(1 << 8)
159
160/* gap */
161
162/* Where do we get our vertex data?
163 *
164 * Vertex data either comes either from immediate mode registers or from
165 * vertex arrays.
166 * There appears to be no mixed mode (though we can force the pitch of
167 * vertex arrays to 0, effectively reusing the same element over and over
168 * again).
169 *
170 * Immediate mode is controlled by the INPUT_CNTL registers. I am not sure
171 * if these registers influence vertex array processing.
172 *
173 * Vertex arrays are controlled via the 3D_LOAD_VBPNTR packet3.
174 *
175 * In both cases, vertex attributes are then passed through INPUT_ROUTE.
176 *
177 * Beginning with INPUT_ROUTE_0_0 is a list of WORDs that route vertex data
178 * into the vertex processor's input registers.
179 * The first word routes the first input, the second word the second, etc.
180 * The corresponding input is routed into the register with the given index.
181 * The list is ended by a word with INPUT_ROUTE_END set.
182 *
183 * Always set COMPONENTS_4 in immediate mode.
184 */
185
186#define R300_VAP_INPUT_ROUTE_0_0            0x2150
187#       define R300_INPUT_ROUTE_COMPONENTS_1     (0 << 0)
188#       define R300_INPUT_ROUTE_COMPONENTS_2     (1 << 0)
189#       define R300_INPUT_ROUTE_COMPONENTS_3     (2 << 0)
190#       define R300_INPUT_ROUTE_COMPONENTS_4     (3 << 0)
191#       define R300_INPUT_ROUTE_COMPONENTS_RGBA  (4 << 0) /* GUESS */
192#       define R300_VAP_INPUT_ROUTE_IDX_SHIFT    8
193#       define R300_VAP_INPUT_ROUTE_IDX_MASK     (31 << 8) /* GUESS */
194#       define R300_VAP_INPUT_ROUTE_END          (1 << 13)
195#       define R300_INPUT_ROUTE_IMMEDIATE_MODE   (0 << 14) /* GUESS */
196#       define R300_INPUT_ROUTE_FLOAT            (1 << 14) /* GUESS */
197#       define R300_INPUT_ROUTE_UNSIGNED_BYTE    (2 << 14) /* GUESS */
198#       define R300_INPUT_ROUTE_FLOAT_COLOR      (3 << 14) /* GUESS */
199#define R300_VAP_INPUT_ROUTE_0_1            0x2154
200#define R300_VAP_INPUT_ROUTE_0_2            0x2158
201#define R300_VAP_INPUT_ROUTE_0_3            0x215C
202#define R300_VAP_INPUT_ROUTE_0_4            0x2160
203#define R300_VAP_INPUT_ROUTE_0_5            0x2164
204#define R300_VAP_INPUT_ROUTE_0_6            0x2168
205#define R300_VAP_INPUT_ROUTE_0_7            0x216C
206
207/* gap */
208
209/* Notes:
210 *  - always set up to produce at least two attributes:
211 *    if vertex program uses only position, fglrx will set normal, too
212 *  - INPUT_CNTL_0_COLOR and INPUT_CNTL_COLOR bits are always equal.
213 */
214#define R300_VAP_INPUT_CNTL_0               0x2180
215#       define R300_INPUT_CNTL_0_COLOR           0x00000001
216#define R300_VAP_INPUT_CNTL_1               0x2184
217#       define R300_INPUT_CNTL_POS               0x00000001
218#       define R300_INPUT_CNTL_NORMAL            0x00000002
219#       define R300_INPUT_CNTL_COLOR             0x00000004
220#       define R300_INPUT_CNTL_TC0               0x00000400
221#       define R300_INPUT_CNTL_TC1               0x00000800
222#       define R300_INPUT_CNTL_TC2               0x00001000 /* GUESS */
223#       define R300_INPUT_CNTL_TC3               0x00002000 /* GUESS */
224#       define R300_INPUT_CNTL_TC4               0x00004000 /* GUESS */
225#       define R300_INPUT_CNTL_TC5               0x00008000 /* GUESS */
226#       define R300_INPUT_CNTL_TC6               0x00010000 /* GUESS */
227#       define R300_INPUT_CNTL_TC7               0x00020000 /* GUESS */
228
229/* gap */
230
231/* Words parallel to INPUT_ROUTE_0; All words that are active in INPUT_ROUTE_0
232 * are set to a swizzling bit pattern, other words are 0.
233 *
234 * In immediate mode, the pattern is always set to xyzw. In vertex array
235 * mode, the swizzling pattern is e.g. used to set zw components in texture
236 * coordinates with only tweo components.
237 */
238#define R300_VAP_INPUT_ROUTE_1_0            0x21E0
239#       define R300_INPUT_ROUTE_SELECT_X    0
240#       define R300_INPUT_ROUTE_SELECT_Y    1
241#       define R300_INPUT_ROUTE_SELECT_Z    2
242#       define R300_INPUT_ROUTE_SELECT_W    3
243#       define R300_INPUT_ROUTE_SELECT_ZERO 4
244#       define R300_INPUT_ROUTE_SELECT_ONE  5
245#       define R300_INPUT_ROUTE_SELECT_MASK 7
246#       define R300_INPUT_ROUTE_X_SHIFT     0
247#       define R300_INPUT_ROUTE_Y_SHIFT     3
248#       define R300_INPUT_ROUTE_Z_SHIFT     6
249#       define R300_INPUT_ROUTE_W_SHIFT     9
250#       define R300_INPUT_ROUTE_ENABLE      (15 << 12)
251#define R300_VAP_INPUT_ROUTE_1_1            0x21E4
252#define R300_VAP_INPUT_ROUTE_1_2            0x21E8
253#define R300_VAP_INPUT_ROUTE_1_3            0x21EC
254#define R300_VAP_INPUT_ROUTE_1_4            0x21F0
255#define R300_VAP_INPUT_ROUTE_1_5            0x21F4
256#define R300_VAP_INPUT_ROUTE_1_6            0x21F8
257#define R300_VAP_INPUT_ROUTE_1_7            0x21FC
258
259/* END: Vertex data assembly */
260
261/* gap */
262
263/* BEGIN: Upload vertex program and data */
264
265/*
266 * The programmable vertex shader unit has a memory bank of unknown size
267 * that can be written to in 16 byte units by writing the address into
268 * UPLOAD_ADDRESS, followed by data in UPLOAD_DATA (multiples of 4 DWORDs).
269 *
270 * Pointers into the memory bank are always in multiples of 16 bytes.
271 *
272 * The memory bank is divided into areas with fixed meaning.
273 *
274 * Starting at address UPLOAD_PROGRAM: Vertex program instructions.
275 * Native limits reported by drivers from ATI suggest size 256 (i.e. 4KB),
276 * whereas the difference between known addresses suggests size 512.
277 *
278 * Starting at address UPLOAD_PARAMETERS: Vertex program parameters.
279 * Native reported limits and the VPI layout suggest size 256, whereas
280 * difference between known addresses suggests size 512.
281 *
282 * At address UPLOAD_POINTSIZE is a vector (0, 0, ps, 0), where ps is the
283 * floating point pointsize. The exact purpose of this state is uncertain,
284 * as there is also the R300_RE_POINTSIZE register.
285 *
286 * Multiple vertex programs and parameter sets can be loaded at once,
287 * which could explain the size discrepancy.
288 */
289#define R300_VAP_PVS_UPLOAD_ADDRESS         0x2200
290#       define R300_PVS_UPLOAD_PROGRAM           0x00000000
291#       define R300_PVS_UPLOAD_PARAMETERS        0x00000200
292#       define R300_PVS_UPLOAD_POINTSIZE         0x00000406
293
294/* gap */
295
296#define R300_VAP_PVS_UPLOAD_DATA            0x2208
297
298/* END: Upload vertex program and data */
299
300/* gap */
301
302/* I do not know the purpose of this register. However, I do know that
303 * it is set to 221C_CLEAR for clear operations and to 221C_NORMAL
304 * for normal rendering.
305 */
306#define R300_VAP_UNKNOWN_221C               0x221C
307#       define R300_221C_NORMAL                  0x00000000
308#       define R300_221C_CLEAR                   0x0001C000
309
310/* These seem to be per-pixel and per-vertex X and Y clipping planes. The first
311 * plane is per-pixel and the second plane is per-vertex.
312 *
313 * This was determined by experimentation alone but I believe it is correct.
314 *
315 * These registers are called X_QUAD0_1_FL to X_QUAD0_4_FL by glxtest.
316 */
317#define R300_VAP_CLIP_X_0                   0x2220
318#define R300_VAP_CLIP_X_1                   0x2224
319#define R300_VAP_CLIP_Y_0                   0x2228
320#define R300_VAP_CLIP_Y_1                   0x2230
321
322/* gap */
323
324/* Sometimes, END_OF_PKT and 0x2284=0 are the only commands sent between
325 * rendering commands and overwriting vertex program parameters.
326 * Therefore, I suspect writing zero to 0x2284 synchronizes the engine and
327 * avoids bugs caused by still running shaders reading bad data from memory.
328 */
329#define R300_VAP_PVS_STATE_FLUSH_REG        0x2284
330
331/* Absolutely no clue what this register is about. */
332#define R300_VAP_UNKNOWN_2288               0x2288
333#       define R300_2288_R300                    0x00750000 /* -- nh */
334#       define R300_2288_RV350                   0x0000FFFF /* -- Vladimir */
335
336/* gap */
337
338/* Addresses are relative to the vertex program instruction area of the
339 * memory bank. PROGRAM_END points to the last instruction of the active
340 * program
341 *
342 * The meaning of the two UNKNOWN fields is obviously not known. However,
343 * experiments so far have shown that both *must* point to an instruction
344 * inside the vertex program, otherwise the GPU locks up.
345 *
346 * fglrx usually sets CNTL_3_UNKNOWN to the end of the program and
347 * R300_PVS_CNTL_1_POS_END_SHIFT points to instruction where last write to
348 * position takes place.
349 *
350 * Most likely this is used to ignore rest of the program in cases
351 * where group of verts arent visible. For some reason this "section"
352 * is sometimes accepted other instruction that have no relationship with
353 * position calculations.
354 */
355#define R300_VAP_PVS_CNTL_1                 0x22D0
356#       define R300_PVS_CNTL_1_PROGRAM_START_SHIFT   0
357#       define R300_PVS_CNTL_1_POS_END_SHIFT         10
358#       define R300_PVS_CNTL_1_PROGRAM_END_SHIFT     20
359/* Addresses are relative the the vertex program parameters area. */
360#define R300_VAP_PVS_CNTL_2                 0x22D4
361#       define R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT 0
362#       define R300_PVS_CNTL_2_PARAM_COUNT_SHIFT  16
363#define R300_VAP_PVS_CNTL_3	           0x22D8
364#       define R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT 10
365#       define R300_PVS_CNTL_3_PROGRAM_UNKNOWN2_SHIFT 0
366
367/* The entire range from 0x2300 to 0x2AC inclusive seems to be used for
368 * immediate vertices
369 */
370#define R300_VAP_VTX_COLOR_R                0x2464
371#define R300_VAP_VTX_COLOR_G                0x2468
372#define R300_VAP_VTX_COLOR_B                0x246C
373#define R300_VAP_VTX_POS_0_X_1              0x2490 /* used for glVertex2*() */
374#define R300_VAP_VTX_POS_0_Y_1              0x2494
375#define R300_VAP_VTX_COLOR_PKD              0x249C /* RGBA */
376#define R300_VAP_VTX_POS_0_X_2              0x24A0 /* used for glVertex3*() */
377#define R300_VAP_VTX_POS_0_Y_2              0x24A4
378#define R300_VAP_VTX_POS_0_Z_2              0x24A8
379/* write 0 to indicate end of packet? */
380#define R300_VAP_VTX_END_OF_PKT             0x24AC
381
382/* gap */
383
384/* These are values from r300_reg/r300_reg.h - they are known to be correct
385 * and are here so we can use one register file instead of several
386 * - Vladimir
387 */
388#define R300_GB_VAP_RASTER_VTX_FMT_0	0x4000
389#	define R300_GB_VAP_RASTER_VTX_FMT_0__POS_PRESENT	(1<<0)
390#	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_0_PRESENT	(1<<1)
391#	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_1_PRESENT	(1<<2)
392#	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_2_PRESENT	(1<<3)
393#	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_3_PRESENT	(1<<4)
394#	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_SPACE	(0xf<<5)
395#	define R300_GB_VAP_RASTER_VTX_FMT_0__PT_SIZE_PRESENT	(0x1<<16)
396
397#define R300_GB_VAP_RASTER_VTX_FMT_1	0x4004
398	/* each of the following is 3 bits wide, specifies number
399	   of components */
400#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT	0
401#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT	3
402#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT	6
403#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT	9
404#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT	12
405#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT	15
406#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT	18
407#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT	21
408
409/* UNK30 seems to enables point to quad transformation on textures
410 * (or something closely related to that).
411 * This bit is rather fatal at the time being due to lackings at pixel
412 * shader side
413 */
414#define R300_GB_ENABLE	0x4008
415#	define R300_GB_POINT_STUFF_ENABLE	(1<<0)
416#	define R300_GB_LINE_STUFF_ENABLE	(1<<1)
417#	define R300_GB_TRIANGLE_STUFF_ENABLE	(1<<2)
418#	define R300_GB_STENCIL_AUTO_ENABLE	(1<<4)
419#	define R300_GB_UNK31			(1<<31)
420	/* each of the following is 2 bits wide */
421#define R300_GB_TEX_REPLICATE	0
422#define R300_GB_TEX_ST		1
423#define R300_GB_TEX_STR		2
424#	define R300_GB_TEX0_SOURCE_SHIFT	16
425#	define R300_GB_TEX1_SOURCE_SHIFT	18
426#	define R300_GB_TEX2_SOURCE_SHIFT	20
427#	define R300_GB_TEX3_SOURCE_SHIFT	22
428#	define R300_GB_TEX4_SOURCE_SHIFT	24
429#	define R300_GB_TEX5_SOURCE_SHIFT	26
430#	define R300_GB_TEX6_SOURCE_SHIFT	28
431#	define R300_GB_TEX7_SOURCE_SHIFT	30
432
433/* MSPOS - positions for multisample antialiasing (?) */
434#define R300_GB_MSPOS0	0x4010
435	/* shifts - each of the fields is 4 bits */
436#	define R300_GB_MSPOS0__MS_X0_SHIFT	0
437#	define R300_GB_MSPOS0__MS_Y0_SHIFT	4
438#	define R300_GB_MSPOS0__MS_X1_SHIFT	8
439#	define R300_GB_MSPOS0__MS_Y1_SHIFT	12
440#	define R300_GB_MSPOS0__MS_X2_SHIFT	16
441#	define R300_GB_MSPOS0__MS_Y2_SHIFT	20
442#	define R300_GB_MSPOS0__MSBD0_Y		24
443#	define R300_GB_MSPOS0__MSBD0_X		28
444
445#define R300_GB_MSPOS1	0x4014
446#	define R300_GB_MSPOS1__MS_X3_SHIFT	0
447#	define R300_GB_MSPOS1__MS_Y3_SHIFT	4
448#	define R300_GB_MSPOS1__MS_X4_SHIFT	8
449#	define R300_GB_MSPOS1__MS_Y4_SHIFT	12
450#	define R300_GB_MSPOS1__MS_X5_SHIFT	16
451#	define R300_GB_MSPOS1__MS_Y5_SHIFT	20
452#	define R300_GB_MSPOS1__MSBD1		24
453
454
455#define R300_GB_TILE_CONFIG	0x4018
456#	define R300_GB_TILE_ENABLE	(1<<0)
457#	define R300_GB_TILE_PIPE_COUNT_RV300	0
458#	define R300_GB_TILE_PIPE_COUNT_R300	(3<<1)
459#	define R300_GB_TILE_PIPE_COUNT_R420	(7<<1)
460#	define R300_GB_TILE_PIPE_COUNT_RV410	(3<<1)
461#	define R300_GB_TILE_SIZE_8		0
462#	define R300_GB_TILE_SIZE_16		(1<<4)
463#	define R300_GB_TILE_SIZE_32		(2<<4)
464#	define R300_GB_SUPER_SIZE_1		(0<<6)
465#	define R300_GB_SUPER_SIZE_2		(1<<6)
466#	define R300_GB_SUPER_SIZE_4		(2<<6)
467#	define R300_GB_SUPER_SIZE_8		(3<<6)
468#	define R300_GB_SUPER_SIZE_16		(4<<6)
469#	define R300_GB_SUPER_SIZE_32		(5<<6)
470#	define R300_GB_SUPER_SIZE_64		(6<<6)
471#	define R300_GB_SUPER_SIZE_128		(7<<6)
472#	define R300_GB_SUPER_X_SHIFT		9	/* 3 bits wide */
473#	define R300_GB_SUPER_Y_SHIFT		12	/* 3 bits wide */
474#	define R300_GB_SUPER_TILE_A		0
475#	define R300_GB_SUPER_TILE_B		(1<<15)
476#	define R300_GB_SUBPIXEL_1_12		0
477#	define R300_GB_SUBPIXEL_1_16		(1<<16)
478
479#define R300_GB_FIFO_SIZE	0x4024
480	/* each of the following is 2 bits wide */
481#define R300_GB_FIFO_SIZE_32	0
482#define R300_GB_FIFO_SIZE_64	1
483#define R300_GB_FIFO_SIZE_128	2
484#define R300_GB_FIFO_SIZE_256	3
485#	define R300_SC_IFIFO_SIZE_SHIFT	0
486#	define R300_SC_TZFIFO_SIZE_SHIFT	2
487#	define R300_SC_BFIFO_SIZE_SHIFT	4
488
489#	define R300_US_OFIFO_SIZE_SHIFT	12
490#	define R300_US_WFIFO_SIZE_SHIFT	14
491	/* the following use the same constants as above, but meaning is
492	   is times 2 (i.e. instead of 32 words it means 64 */
493#	define R300_RS_TFIFO_SIZE_SHIFT	6
494#	define R300_RS_CFIFO_SIZE_SHIFT	8
495#	define R300_US_RAM_SIZE_SHIFT		10
496	/* watermarks, 3 bits wide */
497#	define R300_RS_HIGHWATER_COL_SHIFT	16
498#	define R300_RS_HIGHWATER_TEX_SHIFT	19
499#	define R300_OFIFO_HIGHWATER_SHIFT	22	/* two bits only */
500#	define R300_CUBE_FIFO_HIGHWATER_COL_SHIFT	24
501
502#define R300_GB_SELECT	0x401C
503#	define R300_GB_FOG_SELECT_C0A		0
504#	define R300_GB_FOG_SELECT_C1A		1
505#	define R300_GB_FOG_SELECT_C2A		2
506#	define R300_GB_FOG_SELECT_C3A		3
507#	define R300_GB_FOG_SELECT_1_1_W	4
508#	define R300_GB_FOG_SELECT_Z		5
509#	define R300_GB_DEPTH_SELECT_Z		0
510#	define R300_GB_DEPTH_SELECT_1_1_W	(1<<3)
511#	define R300_GB_W_SELECT_1_W		0
512#	define R300_GB_W_SELECT_1		(1<<4)
513
514#define R300_GB_AA_CONFIG		0x4020
515#	define R300_AA_DISABLE			0x00
516#	define R300_AA_ENABLE			0x01
517#	define R300_AA_SUBSAMPLES_2		0
518#	define R300_AA_SUBSAMPLES_3		(1<<1)
519#	define R300_AA_SUBSAMPLES_4		(2<<1)
520#	define R300_AA_SUBSAMPLES_6		(3<<1)
521
522/* gap */
523
524/* Zero to flush caches. */
525#define R300_TX_INVALTAGS                   0x4100
526#define R300_TX_FLUSH                       0x0
527
528/* The upper enable bits are guessed, based on fglrx reported limits. */
529#define R300_TX_ENABLE                      0x4104
530#       define R300_TX_ENABLE_0                  (1 << 0)
531#       define R300_TX_ENABLE_1                  (1 << 1)
532#       define R300_TX_ENABLE_2                  (1 << 2)
533#       define R300_TX_ENABLE_3                  (1 << 3)
534#       define R300_TX_ENABLE_4                  (1 << 4)
535#       define R300_TX_ENABLE_5                  (1 << 5)
536#       define R300_TX_ENABLE_6                  (1 << 6)
537#       define R300_TX_ENABLE_7                  (1 << 7)
538#       define R300_TX_ENABLE_8                  (1 << 8)
539#       define R300_TX_ENABLE_9                  (1 << 9)
540#       define R300_TX_ENABLE_10                 (1 << 10)
541#       define R300_TX_ENABLE_11                 (1 << 11)
542#       define R300_TX_ENABLE_12                 (1 << 12)
543#       define R300_TX_ENABLE_13                 (1 << 13)
544#       define R300_TX_ENABLE_14                 (1 << 14)
545#       define R300_TX_ENABLE_15                 (1 << 15)
546
547/* The pointsize is given in multiples of 6. The pointsize can be
548 * enormous: Clear() renders a single point that fills the entire
549 * framebuffer.
550 */
551#define R300_RE_POINTSIZE                   0x421C
552#       define R300_POINTSIZE_Y_SHIFT            0
553#       define R300_POINTSIZE_Y_MASK             (0xFFFF << 0) /* GUESS */
554#       define R300_POINTSIZE_X_SHIFT            16
555#       define R300_POINTSIZE_X_MASK             (0xFFFF << 16) /* GUESS */
556#       define R300_POINTSIZE_MAX             (R300_POINTSIZE_Y_MASK / 6)
557
558/* The line width is given in multiples of 6.
559 * In default mode lines are classified as vertical lines.
560 * HO: horizontal
561 * VE: vertical or horizontal
562 * HO & VE: no classification
563 */
564#define R300_RE_LINE_CNT                      0x4234
565#       define R300_LINESIZE_SHIFT            0
566#       define R300_LINESIZE_MASK             (0xFFFF << 0) /* GUESS */
567#       define R300_LINESIZE_MAX             (R300_LINESIZE_MASK / 6)
568#       define R300_LINE_CNT_HO               (1 << 16)
569#       define R300_LINE_CNT_VE               (1 << 17)
570
571/* Some sort of scale or clamp value for texcoordless textures. */
572#define R300_RE_UNK4238                       0x4238
573
574/* Something shade related */
575#define R300_RE_SHADE                         0x4274
576
577#define R300_RE_SHADE_MODEL                   0x4278
578#	define R300_RE_SHADE_MODEL_SMOOTH     0x3aaaa
579#	define R300_RE_SHADE_MODEL_FLAT       0x39595
580
581/* Dangerous */
582#define R300_RE_POLYGON_MODE                  0x4288
583#	define R300_PM_ENABLED                (1 << 0)
584#	define R300_PM_FRONT_POINT            (0 << 0)
585#	define R300_PM_BACK_POINT             (0 << 0)
586#	define R300_PM_FRONT_LINE             (1 << 4)
587#	define R300_PM_FRONT_FILL             (1 << 5)
588#	define R300_PM_BACK_LINE              (1 << 7)
589#	define R300_PM_BACK_FILL              (1 << 8)
590
591/* Fog parameters */
592#define R300_RE_FOG_SCALE                     0x4294
593#define R300_RE_FOG_START                     0x4298
594
595/* Not sure why there are duplicate of factor and constant values.
596 * My best guess so far is that there are separate zbiases for test and write.
597 * Ordering might be wrong.
598 * Some of the tests indicate that fgl has a fallback implementation of zbias
599 * via pixel shaders.
600 */
601#define R300_RE_ZBIAS_CNTL                    0x42A0 /* GUESS */
602#define R300_RE_ZBIAS_T_FACTOR                0x42A4
603#define R300_RE_ZBIAS_T_CONSTANT              0x42A8
604#define R300_RE_ZBIAS_W_FACTOR                0x42AC
605#define R300_RE_ZBIAS_W_CONSTANT              0x42B0
606
607/* This register needs to be set to (1<<1) for RV350 to correctly
608 * perform depth test (see --vb-triangles in r300_demo)
609 * Don't know about other chips. - Vladimir
610 * This is set to 3 when GL_POLYGON_OFFSET_FILL is on.
611 * My guess is that there are two bits for each zbias primitive
612 * (FILL, LINE, POINT).
613 *  One to enable depth test and one for depth write.
614 * Yet this doesn't explain why depth writes work ...
615 */
616#define R300_RE_OCCLUSION_CNTL		    0x42B4
617#	define R300_OCCLUSION_ON		(1<<1)
618
619#define R300_RE_CULL_CNTL                   0x42B8
620#       define R300_CULL_FRONT                   (1 << 0)
621#       define R300_CULL_BACK                    (1 << 1)
622#       define R300_FRONT_FACE_CCW               (0 << 2)
623#       define R300_FRONT_FACE_CW                (1 << 2)
624
625
626/* BEGIN: Rasterization / Interpolators - many guesses */
627
628/* 0_UNKNOWN_18 has always been set except for clear operations.
629 * TC_CNT is the number of incoming texture coordinate sets (i.e. it depends
630 * on the vertex program, *not* the fragment program)
631 */
632#define R300_RS_CNTL_0                      0x4300
633#       define R300_RS_CNTL_TC_CNT_SHIFT         2
634#       define R300_RS_CNTL_TC_CNT_MASK          (7 << 2)
635	/* number of color interpolators used */
636#	define R300_RS_CNTL_CI_CNT_SHIFT         7
637#       define R300_RS_CNTL_0_UNKNOWN_18         (1 << 18)
638	/* Guess: RS_CNTL_1 holds the index of the highest used RS_ROUTE_n
639	   register. */
640#define R300_RS_CNTL_1                      0x4304
641
642/* gap */
643
644/* Only used for texture coordinates.
645 * Use the source field to route texture coordinate input from the
646 * vertex program to the desired interpolator. Note that the source
647 * field is relative to the outputs the vertex program *actually*
648 * writes. If a vertex program only writes texcoord[1], this will
649 * be source index 0.
650 * Set INTERP_USED on all interpolators that produce data used by
651 * the fragment program. INTERP_USED looks like a swizzling mask,
652 * but I haven't seen it used that way.
653 *
654 * Note: The _UNKNOWN constants are always set in their respective
655 * register. I don't know if this is necessary.
656 */
657#define R300_RS_INTERP_0                    0x4310
658#define R300_RS_INTERP_1                    0x4314
659#       define R300_RS_INTERP_1_UNKNOWN          0x40
660#define R300_RS_INTERP_2                    0x4318
661#       define R300_RS_INTERP_2_UNKNOWN          0x80
662#define R300_RS_INTERP_3                    0x431C
663#       define R300_RS_INTERP_3_UNKNOWN          0xC0
664#define R300_RS_INTERP_4                    0x4320
665#define R300_RS_INTERP_5                    0x4324
666#define R300_RS_INTERP_6                    0x4328
667#define R300_RS_INTERP_7                    0x432C
668#       define R300_RS_INTERP_SRC_SHIFT          2
669#       define R300_RS_INTERP_SRC_MASK           (7 << 2)
670#       define R300_RS_INTERP_USED               0x00D10000
671
672/* These DWORDs control how vertex data is routed into fragment program
673 * registers, after interpolators.
674 */
675#define R300_RS_ROUTE_0                     0x4330
676#define R300_RS_ROUTE_1                     0x4334
677#define R300_RS_ROUTE_2                     0x4338
678#define R300_RS_ROUTE_3                     0x433C /* GUESS */
679#define R300_RS_ROUTE_4                     0x4340 /* GUESS */
680#define R300_RS_ROUTE_5                     0x4344 /* GUESS */
681#define R300_RS_ROUTE_6                     0x4348 /* GUESS */
682#define R300_RS_ROUTE_7                     0x434C /* GUESS */
683#       define R300_RS_ROUTE_SOURCE_INTERP_0     0
684#       define R300_RS_ROUTE_SOURCE_INTERP_1     1
685#       define R300_RS_ROUTE_SOURCE_INTERP_2     2
686#       define R300_RS_ROUTE_SOURCE_INTERP_3     3
687#       define R300_RS_ROUTE_SOURCE_INTERP_4     4
688#       define R300_RS_ROUTE_SOURCE_INTERP_5     5 /* GUESS */
689#       define R300_RS_ROUTE_SOURCE_INTERP_6     6 /* GUESS */
690#       define R300_RS_ROUTE_SOURCE_INTERP_7     7 /* GUESS */
691#       define R300_RS_ROUTE_ENABLE              (1 << 3) /* GUESS */
692#       define R300_RS_ROUTE_DEST_SHIFT          6
693#       define R300_RS_ROUTE_DEST_MASK           (31 << 6) /* GUESS */
694
695/* Special handling for color: When the fragment program uses color,
696 * the ROUTE_0_COLOR bit is set and ROUTE_0_COLOR_DEST contains the
697 * color register index.
698 *
699 * Apperently you may set the R300_RS_ROUTE_0_COLOR bit, but not provide any
700 * R300_RS_ROUTE_0_COLOR_DEST value; this setup is used for clearing the state.
701 * See r300_ioctl.c:r300EmitClearState. I'm not sure if this setup is strictly
702 * correct or not. - Oliver.
703 */
704#       define R300_RS_ROUTE_0_COLOR             (1 << 14)
705#       define R300_RS_ROUTE_0_COLOR_DEST_SHIFT  17
706#       define R300_RS_ROUTE_0_COLOR_DEST_MASK   (31 << 17) /* GUESS */
707/* As above, but for secondary color */
708#		define R300_RS_ROUTE_1_COLOR1            (1 << 14)
709#		define R300_RS_ROUTE_1_COLOR1_DEST_SHIFT 17
710#		define R300_RS_ROUTE_1_COLOR1_DEST_MASK  (31 << 17)
711#		define R300_RS_ROUTE_1_UNKNOWN11         (1 << 11)
712/* END: Rasterization / Interpolators - many guesses */
713
714/* Hierarchical Z Enable */
715#define R300_SC_HYPERZ                   0x43a4
716#	define R300_SC_HYPERZ_DISABLE     (0 << 0)
717#	define R300_SC_HYPERZ_ENABLE      (1 << 0)
718#	define R300_SC_HYPERZ_MIN         (0 << 1)
719#	define R300_SC_HYPERZ_MAX         (1 << 1)
720#	define R300_SC_HYPERZ_ADJ_256     (0 << 2)
721#	define R300_SC_HYPERZ_ADJ_128     (1 << 2)
722#	define R300_SC_HYPERZ_ADJ_64      (2 << 2)
723#	define R300_SC_HYPERZ_ADJ_32      (3 << 2)
724#	define R300_SC_HYPERZ_ADJ_16      (4 << 2)
725#	define R300_SC_HYPERZ_ADJ_8       (5 << 2)
726#	define R300_SC_HYPERZ_ADJ_4       (6 << 2)
727#	define R300_SC_HYPERZ_ADJ_2       (7 << 2)
728#	define R300_SC_HYPERZ_HZ_Z0MIN_NO (0 << 5)
729#	define R300_SC_HYPERZ_HZ_Z0MIN    (1 << 5)
730#	define R300_SC_HYPERZ_HZ_Z0MAX_NO (0 << 6)
731#	define R300_SC_HYPERZ_HZ_Z0MAX    (1 << 6)
732
733#define R300_SC_EDGERULE                 0x43a8
734
735/* BEGIN: Scissors and cliprects */
736
737/* There are four clipping rectangles. Their corner coordinates are inclusive.
738 * Every pixel is assigned a number from 0 and 15 by setting bits 0-3 depending
739 * on whether the pixel is inside cliprects 0-3, respectively. For example,
740 * if a pixel is inside cliprects 0 and 1, but outside 2 and 3, it is assigned
741 * the number 3 (binary 0011).
742 * Iff the bit corresponding to the pixel's number in RE_CLIPRECT_CNTL is set,
743 * the pixel is rasterized.
744 *
745 * In addition to this, there is a scissors rectangle. Only pixels inside the
746 * scissors rectangle are drawn. (coordinates are inclusive)
747 *
748 * For some reason, the top-left corner of the framebuffer is at (1440, 1440)
749 * for the purpose of clipping and scissors.
750 */
751#define R300_RE_CLIPRECT_TL_0               0x43B0
752#define R300_RE_CLIPRECT_BR_0               0x43B4
753#define R300_RE_CLIPRECT_TL_1               0x43B8
754#define R300_RE_CLIPRECT_BR_1               0x43BC
755#define R300_RE_CLIPRECT_TL_2               0x43C0
756#define R300_RE_CLIPRECT_BR_2               0x43C4
757#define R300_RE_CLIPRECT_TL_3               0x43C8
758#define R300_RE_CLIPRECT_BR_3               0x43CC
759#       define R300_CLIPRECT_OFFSET              1440
760#       define R300_CLIPRECT_MASK                0x1FFF
761#       define R300_CLIPRECT_X_SHIFT             0
762#       define R300_CLIPRECT_X_MASK              (0x1FFF << 0)
763#       define R300_CLIPRECT_Y_SHIFT             13
764#       define R300_CLIPRECT_Y_MASK              (0x1FFF << 13)
765#define R300_RE_CLIPRECT_CNTL               0x43D0
766#       define R300_CLIP_OUT                     (1 << 0)
767#       define R300_CLIP_0                       (1 << 1)
768#       define R300_CLIP_1                       (1 << 2)
769#       define R300_CLIP_10                      (1 << 3)
770#       define R300_CLIP_2                       (1 << 4)
771#       define R300_CLIP_20                      (1 << 5)
772#       define R300_CLIP_21                      (1 << 6)
773#       define R300_CLIP_210                     (1 << 7)
774#       define R300_CLIP_3                       (1 << 8)
775#       define R300_CLIP_30                      (1 << 9)
776#       define R300_CLIP_31                      (1 << 10)
777#       define R300_CLIP_310                     (1 << 11)
778#       define R300_CLIP_32                      (1 << 12)
779#       define R300_CLIP_320                     (1 << 13)
780#       define R300_CLIP_321                     (1 << 14)
781#       define R300_CLIP_3210                    (1 << 15)
782
783/* gap */
784
785#define R300_RE_SCISSORS_TL                 0x43E0
786#define R300_RE_SCISSORS_BR                 0x43E4
787#       define R300_SCISSORS_OFFSET              1440
788#       define R300_SCISSORS_X_SHIFT             0
789#       define R300_SCISSORS_X_MASK              (0x1FFF << 0)
790#       define R300_SCISSORS_Y_SHIFT             13
791#       define R300_SCISSORS_Y_MASK              (0x1FFF << 13)
792/* END: Scissors and cliprects */
793
794/* BEGIN: Texture specification */
795
796/*
797 * The texture specification dwords are grouped by meaning and not by texture
798 * unit. This means that e.g. the offset for texture image unit N is found in
799 * register TX_OFFSET_0 + (4*N)
800 */
801#define R300_TX_FILTER_0                    0x4400
802#       define R300_TX_REPEAT                    0
803#       define R300_TX_MIRRORED                  1
804#       define R300_TX_CLAMP                     4
805#       define R300_TX_CLAMP_TO_EDGE             2
806#       define R300_TX_CLAMP_TO_BORDER           6
807#       define R300_TX_WRAP_S_SHIFT              0
808#       define R300_TX_WRAP_S_MASK               (7 << 0)
809#       define R300_TX_WRAP_T_SHIFT              3
810#       define R300_TX_WRAP_T_MASK               (7 << 3)
811#       define R300_TX_WRAP_Q_SHIFT              6
812#       define R300_TX_WRAP_Q_MASK               (7 << 6)
813#       define R300_TX_MAG_FILTER_NEAREST        (1 << 9)
814#       define R300_TX_MAG_FILTER_LINEAR         (2 << 9)
815#       define R300_TX_MAG_FILTER_MASK           (3 << 9)
816#       define R300_TX_MIN_FILTER_NEAREST        (1 << 11)
817#       define R300_TX_MIN_FILTER_LINEAR         (2 << 11)
818#	define R300_TX_MIN_FILTER_NEAREST_MIP_NEAREST       (5  <<  11)
819#	define R300_TX_MIN_FILTER_NEAREST_MIP_LINEAR        (9  <<  11)
820#	define R300_TX_MIN_FILTER_LINEAR_MIP_NEAREST        (6  <<  11)
821#	define R300_TX_MIN_FILTER_LINEAR_MIP_LINEAR         (10 <<  11)
822
823/* NOTE: NEAREST doesn't seem to exist.
824 * Im not seting MAG_FILTER_MASK and (3 << 11) on for all
825 * anisotropy modes because that would void selected mag filter
826 */
827#	define R300_TX_MIN_FILTER_ANISO_NEAREST             (0 << 13)
828#	define R300_TX_MIN_FILTER_ANISO_LINEAR              (0 << 13)
829#	define R300_TX_MIN_FILTER_ANISO_NEAREST_MIP_NEAREST (1 << 13)
830#	define R300_TX_MIN_FILTER_ANISO_NEAREST_MIP_LINEAR  (2 << 13)
831#       define R300_TX_MIN_FILTER_MASK   ( (15 << 11) | (3 << 13) )
832#	define R300_TX_MAX_ANISO_1_TO_1  (0 << 21)
833#	define R300_TX_MAX_ANISO_2_TO_1  (2 << 21)
834#	define R300_TX_MAX_ANISO_4_TO_1  (4 << 21)
835#	define R300_TX_MAX_ANISO_8_TO_1  (6 << 21)
836#	define R300_TX_MAX_ANISO_16_TO_1 (8 << 21)
837#	define R300_TX_MAX_ANISO_MASK    (14 << 21)
838
839#define R300_TX_FILTER1_0                      0x4440
840#	define R300_CHROMA_KEY_MODE_DISABLE    0
841#	define R300_CHROMA_KEY_FORCE	       1
842#	define R300_CHROMA_KEY_BLEND           2
843#	define R300_MC_ROUND_NORMAL            (0<<2)
844#	define R300_MC_ROUND_MPEG4             (1<<2)
845#	define R300_LOD_BIAS_MASK	    0x1fff
846#	define R300_EDGE_ANISO_EDGE_DIAG       (0<<13)
847#	define R300_EDGE_ANISO_EDGE_ONLY       (1<<13)
848#	define R300_MC_COORD_TRUNCATE_DISABLE  (0<<14)
849#	define R300_MC_COORD_TRUNCATE_MPEG     (1<<14)
850#	define R300_TX_TRI_PERF_0_8            (0<<15)
851#	define R300_TX_TRI_PERF_1_8            (1<<15)
852#	define R300_TX_TRI_PERF_1_4            (2<<15)
853#	define R300_TX_TRI_PERF_3_8            (3<<15)
854#	define R300_ANISO_THRESHOLD_MASK       (7<<17)
855
856#define R300_TX_SIZE_0                      0x4480
857#       define R300_TX_WIDTHMASK_SHIFT           0
858#       define R300_TX_WIDTHMASK_MASK            (2047 << 0)
859#       define R300_TX_HEIGHTMASK_SHIFT          11
860#       define R300_TX_HEIGHTMASK_MASK           (2047 << 11)
861#       define R300_TX_UNK23                     (1 << 23)
862#       define R300_TX_MAX_MIP_LEVEL_SHIFT       26
863#       define R300_TX_MAX_MIP_LEVEL_MASK        (0xf << 26)
864#       define R300_TX_SIZE_PROJECTED            (1<<30)
865#       define R300_TX_SIZE_TXPITCH_EN           (1<<31)
866#define R300_TX_FORMAT_0                    0x44C0
867	/* The interpretation of the format word by Wladimir van der Laan */
868	/* The X, Y, Z and W refer to the layout of the components.
869	   They are given meanings as R, G, B and Alpha by the swizzle
870	   specification */
871#	define R300_TX_FORMAT_X8		    0x0
872#	define R300_TX_FORMAT_X16		    0x1
873#	define R300_TX_FORMAT_Y4X4		    0x2
874#	define R300_TX_FORMAT_Y8X8		    0x3
875#	define R300_TX_FORMAT_Y16X16		    0x4
876#	define R300_TX_FORMAT_Z3Y3X2		    0x5
877#	define R300_TX_FORMAT_Z5Y6X5		    0x6
878#	define R300_TX_FORMAT_Z6Y5X5		    0x7
879#	define R300_TX_FORMAT_Z11Y11X10		    0x8
880#	define R300_TX_FORMAT_Z10Y11X11		    0x9
881#	define R300_TX_FORMAT_W4Z4Y4X4		    0xA
882#	define R300_TX_FORMAT_W1Z5Y5X5		    0xB
883#	define R300_TX_FORMAT_W8Z8Y8X8		    0xC
884#	define R300_TX_FORMAT_W2Z10Y10X10	    0xD
885#	define R300_TX_FORMAT_W16Z16Y16X16	    0xE
886#	define R300_TX_FORMAT_DXT1		    0xF
887#	define R300_TX_FORMAT_DXT3		    0x10
888#	define R300_TX_FORMAT_DXT5		    0x11
889#	define R300_TX_FORMAT_D3DMFT_CxV8U8	    0x12     /* no swizzle */
890#	define R300_TX_FORMAT_A8R8G8B8		    0x13     /* no swizzle */
891#	define R300_TX_FORMAT_B8G8_B8G8		    0x14     /* no swizzle */
892#	define R300_TX_FORMAT_G8R8_G8B8		    0x15     /* no swizzle */
893	/* 0x16 - some 16 bit green format.. ?? */
894#	define R300_TX_FORMAT_UNK25		   (1 << 25) /* no swizzle */
895#	define R300_TX_FORMAT_CUBIC_MAP		   (1 << 26)
896
897	/* gap */
898	/* Floating point formats */
899	/* Note - hardware supports both 16 and 32 bit floating point */
900#	define R300_TX_FORMAT_FL_I16		    0x18
901#	define R300_TX_FORMAT_FL_I16A16		    0x19
902#	define R300_TX_FORMAT_FL_R16G16B16A16	    0x1A
903#	define R300_TX_FORMAT_FL_I32		    0x1B
904#	define R300_TX_FORMAT_FL_I32A32		    0x1C
905#	define R300_TX_FORMAT_FL_R32G32B32A32	    0x1D
906#	define R300_TX_FORMAT_ATI2N		    0x1F
907	/* alpha modes, convenience mostly */
908	/* if you have alpha, pick constant appropriate to the
909	   number of channels (1 for I8, 2 for I8A8, 4 for R8G8B8A8, etc */
910#	define R300_TX_FORMAT_ALPHA_1CH		    0x000
911#	define R300_TX_FORMAT_ALPHA_2CH		    0x200
912#	define R300_TX_FORMAT_ALPHA_4CH		    0x600
913#	define R300_TX_FORMAT_ALPHA_NONE	    0xA00
914	/* Swizzling */
915	/* constants */
916#	define R300_TX_FORMAT_X		0
917#	define R300_TX_FORMAT_Y		1
918#	define R300_TX_FORMAT_Z		2
919#	define R300_TX_FORMAT_W		3
920#	define R300_TX_FORMAT_ZERO	4
921#	define R300_TX_FORMAT_ONE	5
922	/* 2.0*Z, everything above 1.0 is set to 0.0 */
923#	define R300_TX_FORMAT_CUT_Z	6
924	/* 2.0*W, everything above 1.0 is set to 0.0 */
925#	define R300_TX_FORMAT_CUT_W	7
926
927#	define R300_TX_FORMAT_B_SHIFT	18
928#	define R300_TX_FORMAT_G_SHIFT	15
929#	define R300_TX_FORMAT_R_SHIFT	12
930#	define R300_TX_FORMAT_A_SHIFT	9
931	/* Convenience macro to take care of layout and swizzling */
932#	define R300_EASY_TX_FORMAT(B, G, R, A, FMT)	(		\
933		((R300_TX_FORMAT_##B)<<R300_TX_FORMAT_B_SHIFT)		\
934		| ((R300_TX_FORMAT_##G)<<R300_TX_FORMAT_G_SHIFT)	\
935		| ((R300_TX_FORMAT_##R)<<R300_TX_FORMAT_R_SHIFT)	\
936		| ((R300_TX_FORMAT_##A)<<R300_TX_FORMAT_A_SHIFT)	\
937		| (R300_TX_FORMAT_##FMT)				\
938		)
939	/* These can be ORed with result of R300_EASY_TX_FORMAT()
940	   We don't really know what they do. Take values from a
941           constant color ? */
942#	define R300_TX_FORMAT_CONST_X		(1<<5)
943#	define R300_TX_FORMAT_CONST_Y		(2<<5)
944#	define R300_TX_FORMAT_CONST_Z		(4<<5)
945#	define R300_TX_FORMAT_CONST_W		(8<<5)
946
947#	define R300_TX_FORMAT_YUV_MODE		0x00800000
948
949#define R300_TX_PITCH_0			    0x4500 /* obvious missing in gap */
950#define R300_TX_OFFSET_0                    0x4540
951	/* BEGIN: Guess from R200 */
952#       define R300_TXO_ENDIAN_NO_SWAP           (0 << 0)
953#       define R300_TXO_ENDIAN_BYTE_SWAP         (1 << 0)
954#       define R300_TXO_ENDIAN_WORD_SWAP         (2 << 0)
955#       define R300_TXO_ENDIAN_HALFDW_SWAP       (3 << 0)
956#       define R300_TXO_MACRO_TILE               (1 << 2)
957#       define R300_TXO_MICRO_TILE               (1 << 3)
958#       define R300_TXO_MICRO_TILE_SQUARE        (2 << 3)
959#       define R300_TXO_OFFSET_MASK              0xffffffe0
960#       define R300_TXO_OFFSET_SHIFT             5
961	/* END: Guess from R200 */
962
963/* 32 bit chroma key */
964#define R300_TX_CHROMA_KEY_0                      0x4580
965/* ff00ff00 == { 0, 1.0, 0, 1.0 } */
966#define R300_TX_BORDER_COLOR_0              0x45C0
967
968/* END: Texture specification */
969
970/* BEGIN: Fragment program instruction set */
971
972/* Fragment programs are written directly into register space.
973 * There are separate instruction streams for texture instructions and ALU
974 * instructions.
975 * In order to synchronize these streams, the program is divided into up
976 * to 4 nodes. Each node begins with a number of TEX operations, followed
977 * by a number of ALU operations.
978 * The first node can have zero TEX ops, all subsequent nodes must have at
979 * least
980 * one TEX ops.
981 * All nodes must have at least one ALU op.
982 *
983 * The index of the last node is stored in PFS_CNTL_0: A value of 0 means
984 * 1 node, a value of 3 means 4 nodes.
985 * The total amount of instructions is defined in PFS_CNTL_2. The offsets are
986 * offsets into the respective instruction streams, while *_END points to the
987 * last instruction relative to this offset.
988 */
989#define R300_PFS_CNTL_0                     0x4600
990#       define R300_PFS_CNTL_LAST_NODES_SHIFT    0
991#       define R300_PFS_CNTL_LAST_NODES_MASK     (3 << 0)
992#       define R300_PFS_CNTL_FIRST_NODE_HAS_TEX  (1 << 3)
993#define R300_PFS_CNTL_1                     0x4604
994/* There is an unshifted value here which has so far always been equal to the
995 * index of the highest used temporary register.
996 */
997#define R300_PFS_CNTL_2                     0x4608
998#       define R300_PFS_CNTL_ALU_OFFSET_SHIFT    0
999#       define R300_PFS_CNTL_ALU_OFFSET_MASK     (63 << 0)
1000#       define R300_PFS_CNTL_ALU_END_SHIFT       6
1001#       define R300_PFS_CNTL_ALU_END_MASK        (63 << 6)
1002#       define R300_PFS_CNTL_TEX_OFFSET_SHIFT    12
1003#       define R300_PFS_CNTL_TEX_OFFSET_MASK     (31 << 12) /* GUESS */
1004#       define R300_PFS_CNTL_TEX_END_SHIFT       18
1005#       define R300_PFS_CNTL_TEX_END_MASK        (31 << 18) /* GUESS */
1006
1007/* gap */
1008
1009/* Nodes are stored backwards. The last active node is always stored in
1010 * PFS_NODE_3.
1011 * Example: In a 2-node program, NODE_0 and NODE_1 are set to 0. The
1012 * first node is stored in NODE_2, the second node is stored in NODE_3.
1013 *
1014 * Offsets are relative to the master offset from PFS_CNTL_2.
1015 */
1016#define R300_PFS_NODE_0                     0x4610
1017#define R300_PFS_NODE_1                     0x4614
1018#define R300_PFS_NODE_2                     0x4618
1019#define R300_PFS_NODE_3                     0x461C
1020#       define R300_PFS_NODE_ALU_OFFSET_SHIFT    0
1021#       define R300_PFS_NODE_ALU_OFFSET_MASK     (63 << 0)
1022#       define R300_PFS_NODE_ALU_END_SHIFT       6
1023#       define R300_PFS_NODE_ALU_END_MASK        (63 << 6)
1024#       define R300_PFS_NODE_TEX_OFFSET_SHIFT    12
1025#       define R300_PFS_NODE_TEX_OFFSET_MASK     (31 << 12)
1026#       define R300_PFS_NODE_TEX_END_SHIFT       17
1027#       define R300_PFS_NODE_TEX_END_MASK        (31 << 17)
1028#		define R300_PFS_NODE_OUTPUT_COLOR        (1 << 22)
1029#		define R300_PFS_NODE_OUTPUT_DEPTH        (1 << 23)
1030
1031/* TEX
1032 * As far as I can tell, texture instructions cannot write into output
1033 * registers directly. A subsequent ALU instruction is always necessary,
1034 * even if it's just MAD o0, r0, 1, 0
1035 */
1036#define R300_PFS_TEXI_0                     0x4620
1037#	define R300_FPITX_SRC_SHIFT              0
1038#	define R300_FPITX_SRC_MASK               (31 << 0)
1039	/* GUESS */
1040#	define R300_FPITX_SRC_CONST              (1 << 5)
1041#	define R300_FPITX_DST_SHIFT              6
1042#	define R300_FPITX_DST_MASK               (31 << 6)
1043#	define R300_FPITX_IMAGE_SHIFT            11
1044	/* GUESS based on layout and native limits */
1045#       define R300_FPITX_IMAGE_MASK             (15 << 11)
1046/* Unsure if these are opcodes, or some kind of bitfield, but this is how
1047 * they were set when I checked
1048 */
1049#	define R300_FPITX_OPCODE_SHIFT		15
1050#		define R300_FPITX_OP_TEX	1
1051#		define R300_FPITX_OP_KIL	2
1052#		define R300_FPITX_OP_TXP	3
1053#		define R300_FPITX_OP_TXB	4
1054#	define R300_FPITX_OPCODE_MASK           (7 << 15)
1055
1056/* ALU
1057 * The ALU instructions register blocks are enumerated according to the order
1058 * in which fglrx. I assume there is space for 64 instructions, since
1059 * each block has space for a maximum of 64 DWORDs, and this matches reported
1060 * native limits.
1061 *
1062 * The basic functional block seems to be one MAD for each color and alpha,
1063 * and an adder that adds all components after the MUL.
1064 *  - ADD, MUL, MAD etc.: use MAD with appropriate neutral operands
1065 *  - DP4: Use OUTC_DP4, OUTA_DP4
1066 *  - DP3: Use OUTC_DP3, OUTA_DP4, appropriate alpha operands
1067 *  - DPH: Use OUTC_DP4, OUTA_DP4, appropriate alpha operands
1068 *  - CMPH: If ARG2 > 0.5, return ARG0, else return ARG1
1069 *  - CMP: If ARG2 < 0, return ARG1, else return ARG0
1070 *  - FLR: use FRC+MAD
1071 *  - XPD: use MAD+MAD
1072 *  - SGE, SLT: use MAD+CMP
1073 *  - RSQ: use ABS modifier for argument
1074 *  - Use OUTC_REPL_ALPHA to write results of an alpha-only operation
1075 *    (e.g. RCP) into color register
1076 *  - apparently, there's no quick DST operation
1077 *  - fglrx set FPI2_UNKNOWN_31 on a "MAD fragment.color, tmp0, tmp1, tmp2"
1078 *  - fglrx set FPI2_UNKNOWN_31 on a "MAX r2, r1, c0"
1079 *  - fglrx once set FPI0_UNKNOWN_31 on a "FRC r1, r1"
1080 *
1081 * Operand selection
1082 * First stage selects three sources from the available registers and
1083 * constant parameters. This is defined in INSTR1 (color) and INSTR3 (alpha).
1084 * fglrx sorts the three source fields: Registers before constants,
1085 * lower indices before higher indices; I do not know whether this is
1086 * necessary.
1087 *
1088 * fglrx fills unused sources with "read constant 0"
1089 * According to specs, you cannot select more than two different constants.
1090 *
1091 * Second stage selects the operands from the sources. This is defined in
1092 * INSTR0 (color) and INSTR2 (alpha). You can also select the special constants
1093 * zero and one.
1094 * Swizzling and negation happens in this stage, as well.
1095 *
1096 * Important: Color and alpha seem to be mostly separate, i.e. their sources
1097 * selection appears to be fully independent (the register storage is probably
1098 * physically split into a color and an alpha section).
1099 * However (because of the apparent physical split), there is some interaction
1100 * WRT swizzling. If, for example, you want to load an R component into an
1101 * Alpha operand, this R component is taken from a *color* source, not from
1102 * an alpha source. The corresponding register doesn't even have to appear in
1103 * the alpha sources list. (I hope this all makes sense to you)
1104 *
1105 * Destination selection
1106 * The destination register index is in FPI1 (color) and FPI3 (alpha)
1107 * together with enable bits.
1108 * There are separate enable bits for writing into temporary registers
1109 * (DSTC_REG_* /DSTA_REG) and and program output registers (DSTC_OUTPUT_*
1110 * /DSTA_OUTPUT). You can write to both at once, or not write at all (the
1111 * same index must be used for both).
1112 *
1113 * Note: There is a special form for LRP
1114 *  - Argument order is the same as in ARB_fragment_program.
1115 *  - Operation is MAD
1116 *  - ARG1 is set to ARGC_SRC1C_LRP/ARGC_SRC1A_LRP
1117 *  - Set FPI0/FPI2_SPECIAL_LRP
1118 * Arbitrary LRP (including support for swizzling) requires vanilla MAD+MAD
1119 */
1120#define R300_PFS_INSTR1_0                   0x46C0
1121#       define R300_FPI1_SRC0C_SHIFT             0
1122#       define R300_FPI1_SRC0C_MASK              (31 << 0)
1123#       define R300_FPI1_SRC0C_CONST             (1 << 5)
1124#       define R300_FPI1_SRC1C_SHIFT             6
1125#       define R300_FPI1_SRC1C_MASK              (31 << 6)
1126#       define R300_FPI1_SRC1C_CONST             (1 << 11)
1127#       define R300_FPI1_SRC2C_SHIFT             12
1128#       define R300_FPI1_SRC2C_MASK              (31 << 12)
1129#       define R300_FPI1_SRC2C_CONST             (1 << 17)
1130#       define R300_FPI1_SRC_MASK                0x0003ffff
1131#       define R300_FPI1_DSTC_SHIFT              18
1132#       define R300_FPI1_DSTC_MASK               (31 << 18)
1133#		define R300_FPI1_DSTC_REG_MASK_SHIFT     23
1134#       define R300_FPI1_DSTC_REG_X              (1 << 23)
1135#       define R300_FPI1_DSTC_REG_Y              (1 << 24)
1136#       define R300_FPI1_DSTC_REG_Z              (1 << 25)
1137#		define R300_FPI1_DSTC_OUTPUT_MASK_SHIFT  26
1138#       define R300_FPI1_DSTC_OUTPUT_X           (1 << 26)
1139#       define R300_FPI1_DSTC_OUTPUT_Y           (1 << 27)
1140#       define R300_FPI1_DSTC_OUTPUT_Z           (1 << 28)
1141
1142#define R300_PFS_INSTR3_0                   0x47C0
1143#       define R300_FPI3_SRC0A_SHIFT             0
1144#       define R300_FPI3_SRC0A_MASK              (31 << 0)
1145#       define R300_FPI3_SRC0A_CONST             (1 << 5)
1146#       define R300_FPI3_SRC1A_SHIFT             6
1147#       define R300_FPI3_SRC1A_MASK              (31 << 6)
1148#       define R300_FPI3_SRC1A_CONST             (1 << 11)
1149#       define R300_FPI3_SRC2A_SHIFT             12
1150#       define R300_FPI3_SRC2A_MASK              (31 << 12)
1151#       define R300_FPI3_SRC2A_CONST             (1 << 17)
1152#       define R300_FPI3_SRC_MASK                0x0003ffff
1153#       define R300_FPI3_DSTA_SHIFT              18
1154#       define R300_FPI3_DSTA_MASK               (31 << 18)
1155#       define R300_FPI3_DSTA_REG                (1 << 23)
1156#       define R300_FPI3_DSTA_OUTPUT             (1 << 24)
1157#		define R300_FPI3_DSTA_DEPTH              (1 << 27)
1158
1159#define R300_PFS_INSTR0_0                   0x48C0
1160#       define R300_FPI0_ARGC_SRC0C_XYZ          0
1161#       define R300_FPI0_ARGC_SRC0C_XXX          1
1162#       define R300_FPI0_ARGC_SRC0C_YYY          2
1163#       define R300_FPI0_ARGC_SRC0C_ZZZ          3
1164#       define R300_FPI0_ARGC_SRC1C_XYZ          4
1165#       define R300_FPI0_ARGC_SRC1C_XXX          5
1166#       define R300_FPI0_ARGC_SRC1C_YYY          6
1167#       define R300_FPI0_ARGC_SRC1C_ZZZ          7
1168#       define R300_FPI0_ARGC_SRC2C_XYZ          8
1169#       define R300_FPI0_ARGC_SRC2C_XXX          9
1170#       define R300_FPI0_ARGC_SRC2C_YYY          10
1171#       define R300_FPI0_ARGC_SRC2C_ZZZ          11
1172#       define R300_FPI0_ARGC_SRC0A              12
1173#       define R300_FPI0_ARGC_SRC1A              13
1174#       define R300_FPI0_ARGC_SRC2A              14
1175#       define R300_FPI0_ARGC_SRC1C_LRP          15
1176#       define R300_FPI0_ARGC_ZERO               20
1177#       define R300_FPI0_ARGC_ONE                21
1178	/* GUESS */
1179#       define R300_FPI0_ARGC_HALF               22
1180#       define R300_FPI0_ARGC_SRC0C_YZX          23
1181#       define R300_FPI0_ARGC_SRC1C_YZX          24
1182#       define R300_FPI0_ARGC_SRC2C_YZX          25
1183#       define R300_FPI0_ARGC_SRC0C_ZXY          26
1184#       define R300_FPI0_ARGC_SRC1C_ZXY          27
1185#       define R300_FPI0_ARGC_SRC2C_ZXY          28
1186#       define R300_FPI0_ARGC_SRC0CA_WZY         29
1187#       define R300_FPI0_ARGC_SRC1CA_WZY         30
1188#       define R300_FPI0_ARGC_SRC2CA_WZY         31
1189
1190#       define R300_FPI0_ARG0C_SHIFT             0
1191#       define R300_FPI0_ARG0C_MASK              (31 << 0)
1192#       define R300_FPI0_ARG0C_NEG               (1 << 5)
1193#       define R300_FPI0_ARG0C_ABS               (1 << 6)
1194#       define R300_FPI0_ARG1C_SHIFT             7
1195#       define R300_FPI0_ARG1C_MASK              (31 << 7)
1196#       define R300_FPI0_ARG1C_NEG               (1 << 12)
1197#       define R300_FPI0_ARG1C_ABS               (1 << 13)
1198#       define R300_FPI0_ARG2C_SHIFT             14
1199#       define R300_FPI0_ARG2C_MASK              (31 << 14)
1200#       define R300_FPI0_ARG2C_NEG               (1 << 19)
1201#       define R300_FPI0_ARG2C_ABS               (1 << 20)
1202#       define R300_FPI0_SPECIAL_LRP             (1 << 21)
1203#       define R300_FPI0_OUTC_MAD                (0 << 23)
1204#       define R300_FPI0_OUTC_DP3                (1 << 23)
1205#       define R300_FPI0_OUTC_DP4                (2 << 23)
1206#       define R300_FPI0_OUTC_MIN                (4 << 23)
1207#       define R300_FPI0_OUTC_MAX                (5 << 23)
1208#       define R300_FPI0_OUTC_CMPH               (7 << 23)
1209#       define R300_FPI0_OUTC_CMP                (8 << 23)
1210#       define R300_FPI0_OUTC_FRC                (9 << 23)
1211#       define R300_FPI0_OUTC_REPL_ALPHA         (10 << 23)
1212#       define R300_FPI0_OUTC_SAT                (1 << 30)
1213#       define R300_FPI0_INSERT_NOP              (1U << 31)
1214
1215#define R300_PFS_INSTR2_0                   0x49C0
1216#       define R300_FPI2_ARGA_SRC0C_X            0
1217#       define R300_FPI2_ARGA_SRC0C_Y            1
1218#       define R300_FPI2_ARGA_SRC0C_Z            2
1219#       define R300_FPI2_ARGA_SRC1C_X            3
1220#       define R300_FPI2_ARGA_SRC1C_Y            4
1221#       define R300_FPI2_ARGA_SRC1C_Z            5
1222#       define R300_FPI2_ARGA_SRC2C_X            6
1223#       define R300_FPI2_ARGA_SRC2C_Y            7
1224#       define R300_FPI2_ARGA_SRC2C_Z            8
1225#       define R300_FPI2_ARGA_SRC0A              9
1226#       define R300_FPI2_ARGA_SRC1A              10
1227#       define R300_FPI2_ARGA_SRC2A              11
1228#       define R300_FPI2_ARGA_SRC1A_LRP          15
1229#       define R300_FPI2_ARGA_ZERO               16
1230#       define R300_FPI2_ARGA_ONE                17
1231	/* GUESS */
1232#       define R300_FPI2_ARGA_HALF               18
1233#       define R300_FPI2_ARG0A_SHIFT             0
1234#       define R300_FPI2_ARG0A_MASK              (31 << 0)
1235#       define R300_FPI2_ARG0A_NEG               (1 << 5)
1236	/* GUESS */
1237#	define R300_FPI2_ARG0A_ABS		 (1 << 6)
1238#       define R300_FPI2_ARG1A_SHIFT             7
1239#       define R300_FPI2_ARG1A_MASK              (31 << 7)
1240#       define R300_FPI2_ARG1A_NEG               (1 << 12)
1241	/* GUESS */
1242#	define R300_FPI2_ARG1A_ABS		 (1 << 13)
1243#       define R300_FPI2_ARG2A_SHIFT             14
1244#       define R300_FPI2_ARG2A_MASK              (31 << 14)
1245#       define R300_FPI2_ARG2A_NEG               (1 << 19)
1246	/* GUESS */
1247#	define R300_FPI2_ARG2A_ABS		 (1 << 20)
1248#       define R300_FPI2_SPECIAL_LRP             (1 << 21)
1249#       define R300_FPI2_OUTA_MAD                (0 << 23)
1250#       define R300_FPI2_OUTA_DP4                (1 << 23)
1251#       define R300_FPI2_OUTA_MIN                (2 << 23)
1252#       define R300_FPI2_OUTA_MAX                (3 << 23)
1253#       define R300_FPI2_OUTA_CMP                (6 << 23)
1254#       define R300_FPI2_OUTA_FRC                (7 << 23)
1255#       define R300_FPI2_OUTA_EX2                (8 << 23)
1256#       define R300_FPI2_OUTA_LG2                (9 << 23)
1257#       define R300_FPI2_OUTA_RCP                (10 << 23)
1258#       define R300_FPI2_OUTA_RSQ                (11 << 23)
1259#       define R300_FPI2_OUTA_SAT                (1 << 30)
1260#       define R300_FPI2_UNKNOWN_31              (1U << 31)
1261/* END: Fragment program instruction set */
1262
1263/* Fog state and color */
1264#define R300_RE_FOG_STATE                   0x4BC0
1265#       define R300_FOG_ENABLE                   (1 << 0)
1266#	define R300_FOG_MODE_LINEAR              (0 << 1)
1267#	define R300_FOG_MODE_EXP                 (1 << 1)
1268#	define R300_FOG_MODE_EXP2                (2 << 1)
1269#	define R300_FOG_MODE_MASK                (3 << 1)
1270#define R300_FOG_COLOR_R                    0x4BC8
1271#define R300_FOG_COLOR_G                    0x4BCC
1272#define R300_FOG_COLOR_B                    0x4BD0
1273
1274#define R300_PP_ALPHA_TEST                  0x4BD4
1275#       define R300_REF_ALPHA_MASK               0x000000ff
1276#       define R300_ALPHA_TEST_FAIL              (0 << 8)
1277#       define R300_ALPHA_TEST_LESS              (1 << 8)
1278#       define R300_ALPHA_TEST_LEQUAL            (3 << 8)
1279#       define R300_ALPHA_TEST_EQUAL             (2 << 8)
1280#       define R300_ALPHA_TEST_GEQUAL            (6 << 8)
1281#       define R300_ALPHA_TEST_GREATER           (4 << 8)
1282#       define R300_ALPHA_TEST_NEQUAL            (5 << 8)
1283#       define R300_ALPHA_TEST_PASS              (7 << 8)
1284#       define R300_ALPHA_TEST_OP_MASK           (7 << 8)
1285#       define R300_ALPHA_TEST_ENABLE            (1 << 11)
1286
1287/* gap */
1288
1289/* Fragment program parameters in 7.16 floating point */
1290#define R300_PFS_PARAM_0_X                  0x4C00
1291#define R300_PFS_PARAM_0_Y                  0x4C04
1292#define R300_PFS_PARAM_0_Z                  0x4C08
1293#define R300_PFS_PARAM_0_W                  0x4C0C
1294/* GUESS: PARAM_31 is last, based on native limits reported by fglrx */
1295#define R300_PFS_PARAM_31_X                 0x4DF0
1296#define R300_PFS_PARAM_31_Y                 0x4DF4
1297#define R300_PFS_PARAM_31_Z                 0x4DF8
1298#define R300_PFS_PARAM_31_W                 0x4DFC
1299
1300/* Notes:
1301 * - AFAIK fglrx always sets BLEND_UNKNOWN when blending is used in
1302 *   the application
1303 * - AFAIK fglrx always sets BLEND_NO_SEPARATE when CBLEND and ABLEND
1304 *    are set to the same
1305 *   function (both registers are always set up completely in any case)
1306 * - Most blend flags are simply copied from R200 and not tested yet
1307 */
1308#define R300_RB3D_CBLEND                    0x4E04
1309#define R300_RB3D_ABLEND                    0x4E08
1310/* the following only appear in CBLEND */
1311#       define R300_BLEND_ENABLE                     (1 << 0)
1312#       define R300_BLEND_UNKNOWN                    (3 << 1)
1313#       define R300_BLEND_NO_SEPARATE                (1 << 3)
1314/* the following are shared between CBLEND and ABLEND */
1315#       define R300_FCN_MASK                         (3  << 12)
1316#       define R300_COMB_FCN_ADD_CLAMP               (0  << 12)
1317#       define R300_COMB_FCN_ADD_NOCLAMP             (1  << 12)
1318#       define R300_COMB_FCN_SUB_CLAMP               (2  << 12)
1319#       define R300_COMB_FCN_SUB_NOCLAMP             (3  << 12)
1320#       define R300_COMB_FCN_MIN                     (4  << 12)
1321#       define R300_COMB_FCN_MAX                     (5  << 12)
1322#       define R300_COMB_FCN_RSUB_CLAMP              (6  << 12)
1323#       define R300_COMB_FCN_RSUB_NOCLAMP            (7  << 12)
1324#       define R300_BLEND_GL_ZERO                    (32)
1325#       define R300_BLEND_GL_ONE                     (33)
1326#       define R300_BLEND_GL_SRC_COLOR               (34)
1327#       define R300_BLEND_GL_ONE_MINUS_SRC_COLOR     (35)
1328#       define R300_BLEND_GL_DST_COLOR               (36)
1329#       define R300_BLEND_GL_ONE_MINUS_DST_COLOR     (37)
1330#       define R300_BLEND_GL_SRC_ALPHA               (38)
1331#       define R300_BLEND_GL_ONE_MINUS_SRC_ALPHA     (39)
1332#       define R300_BLEND_GL_DST_ALPHA               (40)
1333#       define R300_BLEND_GL_ONE_MINUS_DST_ALPHA     (41)
1334#       define R300_BLEND_GL_SRC_ALPHA_SATURATE      (42)
1335#       define R300_BLEND_GL_CONST_COLOR             (43)
1336#       define R300_BLEND_GL_ONE_MINUS_CONST_COLOR   (44)
1337#       define R300_BLEND_GL_CONST_ALPHA             (45)
1338#       define R300_BLEND_GL_ONE_MINUS_CONST_ALPHA   (46)
1339#       define R300_BLEND_MASK                       (63)
1340#       define R300_SRC_BLEND_SHIFT                  (16)
1341#       define R300_DST_BLEND_SHIFT                  (24)
1342#define R300_RB3D_BLEND_COLOR               0x4E10
1343#define R300_RB3D_COLORMASK                 0x4E0C
1344#       define R300_COLORMASK0_B                 (1<<0)
1345#       define R300_COLORMASK0_G                 (1<<1)
1346#       define R300_COLORMASK0_R                 (1<<2)
1347#       define R300_COLORMASK0_A                 (1<<3)
1348
1349/* gap */
1350
1351#define R300_RB3D_COLOROFFSET0              0x4E28
1352#       define R300_COLOROFFSET_MASK             0xFFFFFFF0 /* GUESS */
1353#define R300_RB3D_COLOROFFSET1              0x4E2C /* GUESS */
1354#define R300_RB3D_COLOROFFSET2              0x4E30 /* GUESS */
1355#define R300_RB3D_COLOROFFSET3              0x4E34 /* GUESS */
1356
1357/* gap */
1358
1359/* Bit 16: Larger tiles
1360 * Bit 17: 4x2 tiles
1361 * Bit 18: Extremely weird tile like, but some pixels duplicated?
1362 */
1363#define R300_RB3D_COLORPITCH0               0x4E38
1364#       define R300_COLORPITCH_MASK              0x00001FF8 /* GUESS */
1365#       define R300_COLOR_TILE_ENABLE            (1 << 16) /* GUESS */
1366#       define R300_COLOR_MICROTILE_ENABLE       (1 << 17) /* GUESS */
1367#       define R300_COLOR_MICROTILE_SQUARE_ENABLE (2 << 17)
1368#       define R300_COLOR_ENDIAN_NO_SWAP         (0 << 18) /* GUESS */
1369#       define R300_COLOR_ENDIAN_WORD_SWAP       (1 << 18) /* GUESS */
1370#       define R300_COLOR_ENDIAN_DWORD_SWAP      (2 << 18) /* GUESS */
1371#       define R300_COLOR_FORMAT_RGB565          (2 << 22)
1372#       define R300_COLOR_FORMAT_ARGB8888        (3 << 22)
1373#define R300_RB3D_COLORPITCH1               0x4E3C /* GUESS */
1374#define R300_RB3D_COLORPITCH2               0x4E40 /* GUESS */
1375#define R300_RB3D_COLORPITCH3               0x4E44 /* GUESS */
1376
1377#define R300_RB3D_AARESOLVE_OFFSET          0x4E80
1378#define R300_RB3D_AARESOLVE_PITCH           0x4E84
1379#define R300_RB3D_AARESOLVE_CTL             0x4E88
1380/* gap */
1381
1382/* Guess by Vladimir.
1383 * Set to 0A before 3D operations, set to 02 afterwards.
1384 */
1385/*#define R300_RB3D_DSTCACHE_CTLSTAT          0x4E4C*/
1386#       define R300_RB3D_DSTCACHE_UNKNOWN_02             0x00000002
1387#       define R300_RB3D_DSTCACHE_UNKNOWN_0A             0x0000000A
1388
1389/* gap */
1390/* There seems to be no "write only" setting, so use Z-test = ALWAYS
1391 * for this.
1392 * Bit (1<<8) is the "test" bit. so plain write is 6  - vd
1393 */
1394#define R300_ZB_CNTL                             0x4F00
1395#	define R300_STENCIL_ENABLE		 (1 << 0)
1396#	define R300_Z_ENABLE		         (1 << 1)
1397#	define R300_Z_WRITE_ENABLE		 (1 << 2)
1398#	define R300_Z_SIGNED_COMPARE		 (1 << 3)
1399#	define R300_STENCIL_FRONT_BACK		 (1 << 4)
1400
1401#define R300_ZB_ZSTENCILCNTL                   0x4f04
1402	/* functions */
1403#	define R300_ZS_NEVER			0
1404#	define R300_ZS_LESS			1
1405#	define R300_ZS_LEQUAL			2
1406#	define R300_ZS_EQUAL			3
1407#	define R300_ZS_GEQUAL			4
1408#	define R300_ZS_GREATER			5
1409#	define R300_ZS_NOTEQUAL			6
1410#	define R300_ZS_ALWAYS			7
1411#       define R300_ZS_MASK                     7
1412	/* operations */
1413#	define R300_ZS_KEEP			0
1414#	define R300_ZS_ZERO			1
1415#	define R300_ZS_REPLACE			2
1416#	define R300_ZS_INCR			3
1417#	define R300_ZS_DECR			4
1418#	define R300_ZS_INVERT			5
1419#	define R300_ZS_INCR_WRAP		6
1420#	define R300_ZS_DECR_WRAP		7
1421#	define R300_Z_FUNC_SHIFT		0
1422	/* front and back refer to operations done for front
1423	   and back faces, i.e. separate stencil function support */
1424#	define R300_S_FRONT_FUNC_SHIFT	        3
1425#	define R300_S_FRONT_SFAIL_OP_SHIFT	6
1426#	define R300_S_FRONT_ZPASS_OP_SHIFT	9
1427#	define R300_S_FRONT_ZFAIL_OP_SHIFT      12
1428#	define R300_S_BACK_FUNC_SHIFT           15
1429#	define R300_S_BACK_SFAIL_OP_SHIFT       18
1430#	define R300_S_BACK_ZPASS_OP_SHIFT       21
1431#	define R300_S_BACK_ZFAIL_OP_SHIFT       24
1432
1433#define R300_ZB_STENCILREFMASK                        0x4f08
1434#	define R300_STENCILREF_SHIFT       0
1435#	define R300_STENCILREF_MASK        0x000000ff
1436#	define R300_STENCILMASK_SHIFT      8
1437#	define R300_STENCILMASK_MASK       0x0000ff00
1438#	define R300_STENCILWRITEMASK_SHIFT 16
1439#	define R300_STENCILWRITEMASK_MASK  0x00ff0000
1440
1441/* gap */
1442
1443#define R300_ZB_FORMAT                             0x4f10
1444#	define R300_DEPTHFORMAT_16BIT_INT_Z   (0 << 0)
1445#	define R300_DEPTHFORMAT_16BIT_13E3    (1 << 0)
1446#	define R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL   (2 << 0)
1447/* reserved up to (15 << 0) */
1448#	define R300_INVERT_13E3_LEADING_ONES  (0 << 4)
1449#	define R300_INVERT_13E3_LEADING_ZEROS (1 << 4)
1450
1451#define R300_ZB_ZTOP                             0x4F14
1452#	define R300_ZTOP_DISABLE                 (0 << 0)
1453#	define R300_ZTOP_ENABLE                  (1 << 0)
1454
1455/* gap */
1456
1457#define R300_ZB_ZCACHE_CTLSTAT            0x4f18
1458#       define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_NO_EFFECT      (0 << 0)
1459#       define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE (1 << 0)
1460#       define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_NO_EFFECT       (0 << 1)
1461#       define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE            (1 << 1)
1462#       define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_IDLE            (0 << 31)
1463#       define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_BUSY            (1U << 31)
1464
1465#define R300_ZB_BW_CNTL                     0x4f1c
1466#	define R300_HIZ_DISABLE                              (0 << 0)
1467#	define R300_HIZ_ENABLE                               (1 << 0)
1468#	define R300_HIZ_MIN                                  (0 << 1)
1469#	define R300_HIZ_MAX                                  (1 << 1)
1470#	define R300_FAST_FILL_DISABLE                        (0 << 2)
1471#	define R300_FAST_FILL_ENABLE                         (1 << 2)
1472#	define R300_RD_COMP_DISABLE                          (0 << 3)
1473#	define R300_RD_COMP_ENABLE                           (1 << 3)
1474#	define R300_WR_COMP_DISABLE                          (0 << 4)
1475#	define R300_WR_COMP_ENABLE                           (1 << 4)
1476#	define R300_ZB_CB_CLEAR_RMW                          (0 << 5)
1477#	define R300_ZB_CB_CLEAR_CACHE_LINEAR                 (1 << 5)
1478#	define R300_FORCE_COMPRESSED_STENCIL_VALUE_DISABLE   (0 << 6)
1479#	define R300_FORCE_COMPRESSED_STENCIL_VALUE_ENABLE    (1 << 6)
1480
1481#	define R500_ZEQUAL_OPTIMIZE_ENABLE                   (0 << 7)
1482#	define R500_ZEQUAL_OPTIMIZE_DISABLE                  (1 << 7)
1483#	define R500_SEQUAL_OPTIMIZE_ENABLE                   (0 << 8)
1484#	define R500_SEQUAL_OPTIMIZE_DISABLE                  (1 << 8)
1485
1486#	define R500_BMASK_ENABLE                             (0 << 10)
1487#	define R500_BMASK_DISABLE                            (1 << 10)
1488#	define R500_HIZ_EQUAL_REJECT_DISABLE                 (0 << 11)
1489#	define R500_HIZ_EQUAL_REJECT_ENABLE                  (1 << 11)
1490#	define R500_HIZ_FP_EXP_BITS_DISABLE                  (0 << 12)
1491#	define R500_HIZ_FP_EXP_BITS_1                        (1 << 12)
1492#	define R500_HIZ_FP_EXP_BITS_2                        (2 << 12)
1493#	define R500_HIZ_FP_EXP_BITS_3                        (3 << 12)
1494#	define R500_HIZ_FP_EXP_BITS_4                        (4 << 12)
1495#	define R500_HIZ_FP_EXP_BITS_5                        (5 << 12)
1496#	define R500_HIZ_FP_INVERT_LEADING_ONES               (0 << 15)
1497#	define R500_HIZ_FP_INVERT_LEADING_ZEROS              (1 << 15)
1498#	define R500_TILE_OVERWRITE_RECOMPRESSION_ENABLE      (0 << 16)
1499#	define R500_TILE_OVERWRITE_RECOMPRESSION_DISABLE     (1 << 16)
1500#	define R500_CONTIGUOUS_6XAA_SAMPLES_ENABLE           (0 << 17)
1501#	define R500_CONTIGUOUS_6XAA_SAMPLES_DISABLE          (1 << 17)
1502#	define R500_PEQ_PACKING_DISABLE                      (0 << 18)
1503#	define R500_PEQ_PACKING_ENABLE                       (1 << 18)
1504#	define R500_COVERED_PTR_MASKING_DISABLE              (0 << 18)
1505#	define R500_COVERED_PTR_MASKING_ENABLE               (1 << 18)
1506
1507
1508/* gap */
1509
1510/* Z Buffer Address Offset.
1511 * Bits 31 to 5 are used for aligned Z buffer address offset for macro tiles.
1512 */
1513#define R300_ZB_DEPTHOFFSET               0x4f20
1514
1515/* Z Buffer Pitch and Endian Control */
1516#define R300_ZB_DEPTHPITCH                0x4f24
1517#       define R300_DEPTHPITCH_MASK              0x00003FFC
1518#       define R300_DEPTHMACROTILE_DISABLE      (0 << 16)
1519#       define R300_DEPTHMACROTILE_ENABLE       (1 << 16)
1520#       define R300_DEPTHMICROTILE_LINEAR       (0 << 17)
1521#       define R300_DEPTHMICROTILE_TILED        (1 << 17)
1522#       define R300_DEPTHMICROTILE_TILED_SQUARE (2 << 17)
1523#       define R300_DEPTHENDIAN_NO_SWAP         (0 << 18)
1524#       define R300_DEPTHENDIAN_WORD_SWAP       (1 << 18)
1525#       define R300_DEPTHENDIAN_DWORD_SWAP      (2 << 18)
1526#       define R300_DEPTHENDIAN_HALF_DWORD_SWAP (3 << 18)
1527
1528/* Z Buffer Clear Value */
1529#define R300_ZB_DEPTHCLEARVALUE                  0x4f28
1530
1531#define R300_ZB_ZMASK_OFFSET			 0x4f30
1532#define R300_ZB_ZMASK_PITCH			 0x4f34
1533#define R300_ZB_ZMASK_WRINDEX			 0x4f38
1534#define R300_ZB_ZMASK_DWORD			 0x4f3c
1535#define R300_ZB_ZMASK_RDINDEX			 0x4f40
1536
1537/* Hierarchical Z Memory Offset */
1538#define R300_ZB_HIZ_OFFSET                       0x4f44
1539
1540/* Hierarchical Z Write Index */
1541#define R300_ZB_HIZ_WRINDEX                      0x4f48
1542
1543/* Hierarchical Z Data */
1544#define R300_ZB_HIZ_DWORD                        0x4f4c
1545
1546/* Hierarchical Z Read Index */
1547#define R300_ZB_HIZ_RDINDEX                      0x4f50
1548
1549/* Hierarchical Z Pitch */
1550#define R300_ZB_HIZ_PITCH                        0x4f54
1551
1552/* Z Buffer Z Pass Counter Data */
1553#define R300_ZB_ZPASS_DATA                       0x4f58
1554
1555/* Z Buffer Z Pass Counter Address */
1556#define R300_ZB_ZPASS_ADDR                       0x4f5c
1557
1558/* Depth buffer X and Y coordinate offset */
1559#define R300_ZB_DEPTHXY_OFFSET                   0x4f60
1560#	define R300_DEPTHX_OFFSET_SHIFT  1
1561#	define R300_DEPTHX_OFFSET_MASK   0x000007FE
1562#	define R300_DEPTHY_OFFSET_SHIFT  17
1563#	define R300_DEPTHY_OFFSET_MASK   0x07FE0000
1564
1565/* Sets the fifo sizes */
1566#define R500_ZB_FIFO_SIZE                        0x4fd0
1567#	define R500_OP_FIFO_SIZE_FULL   (0 << 0)
1568#	define R500_OP_FIFO_SIZE_HALF   (1 << 0)
1569#	define R500_OP_FIFO_SIZE_QUATER (2 << 0)
1570#	define R500_OP_FIFO_SIZE_EIGTHS (4 << 0)
1571
1572/* Stencil Reference Value and Mask for backfacing quads */
1573/* R300_ZB_STENCILREFMASK handles front face */
1574#define R500_ZB_STENCILREFMASK_BF                0x4fd4
1575#	define R500_STENCILREF_SHIFT       0
1576#	define R500_STENCILREF_MASK        0x000000ff
1577#	define R500_STENCILMASK_SHIFT      8
1578#	define R500_STENCILMASK_MASK       0x0000ff00
1579#	define R500_STENCILWRITEMASK_SHIFT 16
1580#	define R500_STENCILWRITEMASK_MASK  0x00ff0000
1581
1582/* BEGIN: Vertex program instruction set */
1583
1584/* Every instruction is four dwords long:
1585 *  DWORD 0: output and opcode
1586 *  DWORD 1: first argument
1587 *  DWORD 2: second argument
1588 *  DWORD 3: third argument
1589 *
1590 * Notes:
1591 *  - ABS r, a is implemented as MAX r, a, -a
1592 *  - MOV is implemented as ADD to zero
1593 *  - XPD is implemented as MUL + MAD
1594 *  - FLR is implemented as FRC + ADD
1595 *  - apparently, fglrx tries to schedule instructions so that there is at
1596 *    least one instruction between the write to a temporary and the first
1597 *    read from said temporary; however, violations of this scheduling are
1598 *    allowed
1599 *  - register indices seem to be unrelated with OpenGL aliasing to
1600 *    conventional state
1601 *  - only one attribute and one parameter can be loaded at a time; however,
1602 *    the same attribute/parameter can be used for more than one argument
1603 *  - the second software argument for POW is the third hardware argument
1604 *    (no idea why)
1605 *  - MAD with only temporaries as input seems to use VPI_OUT_SELECT_MAD_2
1606 *
1607 * There is some magic surrounding LIT:
1608 *   The single argument is replicated across all three inputs, but swizzled:
1609 *     First argument: xyzy
1610 *     Second argument: xyzx
1611 *     Third argument: xyzw
1612 *   Whenever the result is used later in the fragment program, fglrx forces
1613 *   x and w to be 1.0 in the input selection; I don't know whether this is
1614 *   strictly necessary
1615 */
1616#define R300_VPI_OUT_OP_DOT                     (1 << 0)
1617#define R300_VPI_OUT_OP_MUL                     (2 << 0)
1618#define R300_VPI_OUT_OP_ADD                     (3 << 0)
1619#define R300_VPI_OUT_OP_MAD                     (4 << 0)
1620#define R300_VPI_OUT_OP_DST                     (5 << 0)
1621#define R300_VPI_OUT_OP_FRC                     (6 << 0)
1622#define R300_VPI_OUT_OP_MAX                     (7 << 0)
1623#define R300_VPI_OUT_OP_MIN                     (8 << 0)
1624#define R300_VPI_OUT_OP_SGE                     (9 << 0)
1625#define R300_VPI_OUT_OP_SLT                     (10 << 0)
1626	/* Used in GL_POINT_DISTANCE_ATTENUATION_ARB, vector(scalar, vector) */
1627#define R300_VPI_OUT_OP_UNK12                   (12 << 0)
1628#define R300_VPI_OUT_OP_ARL                     (13 << 0)
1629#define R300_VPI_OUT_OP_EXP                     (65 << 0)
1630#define R300_VPI_OUT_OP_LOG                     (66 << 0)
1631	/* Used in fog computations, scalar(scalar) */
1632#define R300_VPI_OUT_OP_UNK67                   (67 << 0)
1633#define R300_VPI_OUT_OP_LIT                     (68 << 0)
1634#define R300_VPI_OUT_OP_POW                     (69 << 0)
1635#define R300_VPI_OUT_OP_RCP                     (70 << 0)
1636#define R300_VPI_OUT_OP_RSQ                     (72 << 0)
1637	/* Used in GL_POINT_DISTANCE_ATTENUATION_ARB, scalar(scalar) */
1638#define R300_VPI_OUT_OP_UNK73                   (73 << 0)
1639#define R300_VPI_OUT_OP_EX2                     (75 << 0)
1640#define R300_VPI_OUT_OP_LG2                     (76 << 0)
1641#define R300_VPI_OUT_OP_MAD_2                   (128 << 0)
1642	/* all temps, vector(scalar, vector, vector) */
1643#define R300_VPI_OUT_OP_UNK129                  (129 << 0)
1644
1645#define R300_VPI_OUT_REG_CLASS_TEMPORARY        (0 << 8)
1646#define R300_VPI_OUT_REG_CLASS_ADDR             (1 << 8)
1647#define R300_VPI_OUT_REG_CLASS_RESULT           (2 << 8)
1648#define R300_VPI_OUT_REG_CLASS_MASK             (31 << 8)
1649
1650#define R300_VPI_OUT_REG_INDEX_SHIFT            13
1651	/* GUESS based on fglrx native limits */
1652#define R300_VPI_OUT_REG_INDEX_MASK             (31 << 13)
1653
1654#define R300_VPI_OUT_WRITE_X                    (1 << 20)
1655#define R300_VPI_OUT_WRITE_Y                    (1 << 21)
1656#define R300_VPI_OUT_WRITE_Z                    (1 << 22)
1657#define R300_VPI_OUT_WRITE_W                    (1 << 23)
1658
1659#define R300_VPI_IN_REG_CLASS_TEMPORARY         (0 << 0)
1660#define R300_VPI_IN_REG_CLASS_ATTRIBUTE         (1 << 0)
1661#define R300_VPI_IN_REG_CLASS_PARAMETER         (2 << 0)
1662#define R300_VPI_IN_REG_CLASS_NONE              (9 << 0)
1663#define R300_VPI_IN_REG_CLASS_MASK              (31 << 0)
1664
1665#define R300_VPI_IN_REG_INDEX_SHIFT             5
1666	/* GUESS based on fglrx native limits */
1667#define R300_VPI_IN_REG_INDEX_MASK              (255 << 5)
1668
1669/* The R300 can select components from the input register arbitrarily.
1670 * Use the following constants, shifted by the component shift you
1671 * want to select
1672 */
1673#define R300_VPI_IN_SELECT_X    0
1674#define R300_VPI_IN_SELECT_Y    1
1675#define R300_VPI_IN_SELECT_Z    2
1676#define R300_VPI_IN_SELECT_W    3
1677#define R300_VPI_IN_SELECT_ZERO 4
1678#define R300_VPI_IN_SELECT_ONE  5
1679#define R300_VPI_IN_SELECT_MASK 7
1680
1681#define R300_VPI_IN_X_SHIFT                     13
1682#define R300_VPI_IN_Y_SHIFT                     16
1683#define R300_VPI_IN_Z_SHIFT                     19
1684#define R300_VPI_IN_W_SHIFT                     22
1685
1686#define R300_VPI_IN_NEG_X                       (1 << 25)
1687#define R300_VPI_IN_NEG_Y                       (1 << 26)
1688#define R300_VPI_IN_NEG_Z                       (1 << 27)
1689#define R300_VPI_IN_NEG_W                       (1 << 28)
1690/* END: Vertex program instruction set */
1691
1692/* BEGIN: Packet 3 commands */
1693
1694/* A primitive emission dword. */
1695#define R300_PRIM_TYPE_NONE                     (0 << 0)
1696#define R300_PRIM_TYPE_POINT                    (1 << 0)
1697#define R300_PRIM_TYPE_LINE                     (2 << 0)
1698#define R300_PRIM_TYPE_LINE_STRIP               (3 << 0)
1699#define R300_PRIM_TYPE_TRI_LIST                 (4 << 0)
1700#define R300_PRIM_TYPE_TRI_FAN                  (5 << 0)
1701#define R300_PRIM_TYPE_TRI_STRIP                (6 << 0)
1702#define R300_PRIM_TYPE_TRI_TYPE2                (7 << 0)
1703#define R300_PRIM_TYPE_RECT_LIST                (8 << 0)
1704#define R300_PRIM_TYPE_3VRT_POINT_LIST          (9 << 0)
1705#define R300_PRIM_TYPE_3VRT_LINE_LIST           (10 << 0)
1706	/* GUESS (based on r200) */
1707#define R300_PRIM_TYPE_POINT_SPRITES            (11 << 0)
1708#define R300_PRIM_TYPE_LINE_LOOP                (12 << 0)
1709#define R300_PRIM_TYPE_QUADS                    (13 << 0)
1710#define R300_PRIM_TYPE_QUAD_STRIP               (14 << 0)
1711#define R300_PRIM_TYPE_POLYGON                  (15 << 0)
1712#define R300_PRIM_TYPE_MASK                     0xF
1713#define R300_PRIM_WALK_IND                      (1 << 4)
1714#define R300_PRIM_WALK_LIST                     (2 << 4)
1715#define R300_PRIM_WALK_RING                     (3 << 4)
1716#define R300_PRIM_WALK_MASK                     (3 << 4)
1717	/* GUESS (based on r200) */
1718#define R300_PRIM_COLOR_ORDER_BGRA              (0 << 6)
1719#define R300_PRIM_COLOR_ORDER_RGBA              (1 << 6)
1720#define R300_PRIM_NUM_VERTICES_SHIFT            16
1721#define R300_PRIM_NUM_VERTICES_MASK             0xffff
1722
1723/* Draw a primitive from vertex data in arrays loaded via 3D_LOAD_VBPNTR.
1724 * Two parameter dwords:
1725 * 0. The first parameter appears to be always 0
1726 * 1. The second parameter is a standard primitive emission dword.
1727 */
1728#define R300_PACKET3_3D_DRAW_VBUF           0x00002800
1729
1730/* Specify the full set of vertex arrays as (address, stride).
1731 * The first parameter is the number of vertex arrays specified.
1732 * The rest of the command is a variable length list of blocks, where
1733 * each block is three dwords long and specifies two arrays.
1734 * The first dword of a block is split into two words, the lower significant
1735 * word refers to the first array, the more significant word to the second
1736 * array in the block.
1737 * The low byte of each word contains the size of an array entry in dwords,
1738 * the high byte contains the stride of the array.
1739 * The second dword of a block contains the pointer to the first array,
1740 * the third dword of a block contains the pointer to the second array.
1741 * Note that if the total number of arrays is odd, the third dword of
1742 * the last block is omitted.
1743 */
1744#define R300_PACKET3_3D_LOAD_VBPNTR         0x00002F00
1745
1746#define R300_PACKET3_INDX_BUFFER            0x00003300
1747#    define R300_EB_UNK1_SHIFT                      24
1748#    define R300_EB_UNK1                    (0x80<<24)
1749#    define R300_EB_UNK2                        0x0810
1750#define R300_PACKET3_3D_DRAW_VBUF_2         0x00003400
1751#define R300_PACKET3_3D_DRAW_INDX_2         0x00003600
1752
1753/* END: Packet 3 commands */
1754
1755
1756/* Color formats for 2d packets
1757 */
1758#define R300_CP_COLOR_FORMAT_CI8	2
1759#define R300_CP_COLOR_FORMAT_ARGB1555	3
1760#define R300_CP_COLOR_FORMAT_RGB565	4
1761#define R300_CP_COLOR_FORMAT_ARGB8888	6
1762#define R300_CP_COLOR_FORMAT_RGB332	7
1763#define R300_CP_COLOR_FORMAT_RGB8	9
1764#define R300_CP_COLOR_FORMAT_ARGB4444	15
1765
1766/*
1767 * CP type-3 packets
1768 */
1769#define R300_CP_CMD_BITBLT_MULTI	0xC0009B00
1770
1771#define R500_VAP_INDEX_OFFSET		0x208c
1772
1773#define R500_GA_US_VECTOR_INDEX         0x4250
1774#define R500_GA_US_VECTOR_DATA          0x4254
1775
1776#define R500_RS_IP_0                    0x4074
1777#define R500_RS_INST_0                  0x4320
1778
1779#define R500_US_CONFIG                  0x4600
1780
1781#define R500_US_FC_CTRL			0x4624
1782#define R500_US_CODE_ADDR		0x4630
1783
1784#define R500_RB3D_COLOR_CLEAR_VALUE_AR  0x46c0
1785#define R500_RB3D_CONSTANT_COLOR_AR     0x4ef8
1786
1787#define R300_SU_REG_DEST                0x42c8
1788#define RV530_FG_ZBREG_DEST             0x4be8
1789#define R300_ZB_ZPASS_DATA              0x4f58
1790#define R300_ZB_ZPASS_ADDR              0x4f5c
1791
1792#endif /* _R300_REG_H */
1793