Home | History | Annotate | Line # | Download | only in radeon
      1 /*	$NetBSD: r300_reg.h,v 1.3 2021/12/18 23:45:42 riastradh Exp $	*/
      2 
      3 /*
      4  * Copyright 2005 Nicolai Haehnle et al.
      5  * Copyright 2008 Advanced Micro Devices, Inc.
      6  * Copyright 2009 Jerome Glisse.
      7  *
      8  * Permission is hereby granted, free of charge, to any person obtaining a
      9  * copy of this software and associated documentation files (the "Software"),
     10  * to deal in the Software without restriction, including without limitation
     11  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
     12  * and/or sell copies of the Software, and to permit persons to whom the
     13  * Software is furnished to do so, subject to the following conditions:
     14  *
     15  * The above copyright notice and this permission notice shall be included in
     16  * all copies or substantial portions of the Software.
     17  *
     18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
     19  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
     20  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
     21  * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
     22  * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
     23  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
     24  * OTHER DEALINGS IN THE SOFTWARE.
     25  *
     26  * Authors: Nicolai Haehnle
     27  *          Jerome Glisse
     28  */
     29 #ifndef _R300_REG_H_
     30 #define _R300_REG_H_
     31 
     32 #define R300_SURF_TILE_MACRO (1<<16)
     33 #define R300_SURF_TILE_MICRO (2<<16)
     34 #define R300_SURF_TILE_BOTH (3<<16)
     35 
     36 
     37 #define R300_MC_INIT_MISC_LAT_TIMER	0x180
     38 #	define R300_MC_MISC__MC_CPR_INIT_LAT_SHIFT	0
     39 #	define R300_MC_MISC__MC_VF_INIT_LAT_SHIFT	4
     40 #	define R300_MC_MISC__MC_DISP0R_INIT_LAT_SHIFT	8
     41 #	define R300_MC_MISC__MC_DISP1R_INIT_LAT_SHIFT	12
     42 #	define R300_MC_MISC__MC_FIXED_INIT_LAT_SHIFT	16
     43 #	define R300_MC_MISC__MC_E2R_INIT_LAT_SHIFT	20
     44 #	define R300_MC_MISC__MC_SAME_PAGE_PRIO_SHIFT	24
     45 #	define R300_MC_MISC__MC_GLOBW_INIT_LAT_SHIFT	28
     46 
     47 #define R300_MC_INIT_GFX_LAT_TIMER	0x154
     48 #	define R300_MC_MISC__MC_G3D0R_INIT_LAT_SHIFT	0
     49 #	define R300_MC_MISC__MC_G3D1R_INIT_LAT_SHIFT	4
     50 #	define R300_MC_MISC__MC_G3D2R_INIT_LAT_SHIFT	8
     51 #	define R300_MC_MISC__MC_G3D3R_INIT_LAT_SHIFT	12
     52 #	define R300_MC_MISC__MC_TX0R_INIT_LAT_SHIFT	16
     53 #	define R300_MC_MISC__MC_TX1R_INIT_LAT_SHIFT	20
     54 #	define R300_MC_MISC__MC_GLOBR_INIT_LAT_SHIFT	24
     55 #	define R300_MC_MISC__MC_GLOBW_FULL_LAT_SHIFT	28
     56 
     57 /*
     58  * This file contains registers and constants for the R300. They have been
     59  * found mostly by examining command buffers captured using glxtest, as well
     60  * as by extrapolating some known registers and constants from the R200.
     61  * I am fairly certain that they are correct unless stated otherwise
     62  * in comments.
     63  */
     64 
     65 #define R300_SE_VPORT_XSCALE                0x1D98
     66 #define R300_SE_VPORT_XOFFSET               0x1D9C
     67 #define R300_SE_VPORT_YSCALE                0x1DA0
     68 #define R300_SE_VPORT_YOFFSET               0x1DA4
     69 #define R300_SE_VPORT_ZSCALE                0x1DA8
     70 #define R300_SE_VPORT_ZOFFSET               0x1DAC
     71 
     72 
     73 /*
     74  * Vertex Array Processing (VAP) Control
     75  * Stolen from r200 code from Christoph Brill (It's a guess!)
     76  */
     77 #define R300_VAP_CNTL	0x2080
     78 
     79 /* This register is written directly and also starts data section
     80  * in many 3d CP_PACKET3's
     81  */
     82 #define R300_VAP_VF_CNTL	0x2084
     83 #	define	R300_VAP_VF_CNTL__PRIM_TYPE__SHIFT              0
     84 #	define  R300_VAP_VF_CNTL__PRIM_NONE                     (0<<0)
     85 #	define  R300_VAP_VF_CNTL__PRIM_POINTS                   (1<<0)
     86 #	define  R300_VAP_VF_CNTL__PRIM_LINES                    (2<<0)
     87 #	define  R300_VAP_VF_CNTL__PRIM_LINE_STRIP               (3<<0)
     88 #	define  R300_VAP_VF_CNTL__PRIM_TRIANGLES                (4<<0)
     89 #	define  R300_VAP_VF_CNTL__PRIM_TRIANGLE_FAN             (5<<0)
     90 #	define  R300_VAP_VF_CNTL__PRIM_TRIANGLE_STRIP           (6<<0)
     91 #	define  R300_VAP_VF_CNTL__PRIM_LINE_LOOP                (12<<0)
     92 #	define  R300_VAP_VF_CNTL__PRIM_QUADS                    (13<<0)
     93 #	define  R300_VAP_VF_CNTL__PRIM_QUAD_STRIP               (14<<0)
     94 #	define  R300_VAP_VF_CNTL__PRIM_POLYGON                  (15<<0)
     95 
     96 #	define	R300_VAP_VF_CNTL__PRIM_WALK__SHIFT              4
     97 	/* State based - direct writes to registers trigger vertex
     98            generation */
     99 #	define	R300_VAP_VF_CNTL__PRIM_WALK_STATE_BASED         (0<<4)
    100 #	define	R300_VAP_VF_CNTL__PRIM_WALK_INDICES             (1<<4)
    101 #	define	R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST         (2<<4)
    102 #	define	R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_EMBEDDED     (3<<4)
    103 
    104 	/* I don't think I saw these three used.. */
    105 #	define	R300_VAP_VF_CNTL__COLOR_ORDER__SHIFT            6
    106 #	define	R300_VAP_VF_CNTL__TCL_OUTPUT_CTL_ENA__SHIFT     9
    107 #	define	R300_VAP_VF_CNTL__PROG_STREAM_ENA__SHIFT        10
    108 
    109 	/* index size - when not set the indices are assumed to be 16 bit */
    110 #	define	R300_VAP_VF_CNTL__INDEX_SIZE_32bit              (1<<11)
    111 	/* number of vertices */
    112 #	define	R300_VAP_VF_CNTL__NUM_VERTICES__SHIFT           16
    113 
    114 /* BEGIN: Wild guesses */
    115 #define R300_VAP_OUTPUT_VTX_FMT_0           0x2090
    116 #       define R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT     (1<<0)
    117 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT   (1<<1)
    118 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT (1<<2)  /* GUESS */
    119 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT (1<<3)  /* GUESS */
    120 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT (1<<4)  /* GUESS */
    121 #       define R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT (1<<16) /* GUESS */
    122 
    123 #define R300_VAP_OUTPUT_VTX_FMT_1           0x2094
    124 	/* each of the following is 3 bits wide, specifies number
    125 	   of components */
    126 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT 0
    127 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT 3
    128 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT 6
    129 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT 9
    130 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT 12
    131 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT 15
    132 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT 18
    133 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT 21
    134 /* END: Wild guesses */
    135 
    136 #define R300_SE_VTE_CNTL                  0x20b0
    137 #	define     R300_VPORT_X_SCALE_ENA                0x00000001
    138 #	define     R300_VPORT_X_OFFSET_ENA               0x00000002
    139 #	define     R300_VPORT_Y_SCALE_ENA                0x00000004
    140 #	define     R300_VPORT_Y_OFFSET_ENA               0x00000008
    141 #	define     R300_VPORT_Z_SCALE_ENA                0x00000010
    142 #	define     R300_VPORT_Z_OFFSET_ENA               0x00000020
    143 #	define     R300_VTX_XY_FMT                       0x00000100
    144 #	define     R300_VTX_Z_FMT                        0x00000200
    145 #	define     R300_VTX_W0_FMT                       0x00000400
    146 #	define     R300_VTX_W0_NORMALIZE                 0x00000800
    147 #	define     R300_VTX_ST_DENORMALIZED              0x00001000
    148 
    149 /* BEGIN: Vertex data assembly - lots of uncertainties */
    150 
    151 /* gap */
    152 
    153 #define R300_VAP_CNTL_STATUS              0x2140
    154 #	define R300_VC_NO_SWAP                  (0 << 0)
    155 #	define R300_VC_16BIT_SWAP               (1 << 0)
    156 #	define R300_VC_32BIT_SWAP               (2 << 0)
    157 #	define R300_VAP_TCL_BYPASS		(1 << 8)
    158 
    159 /* gap */
    160 
    161 /* Where do we get our vertex data?
    162  *
    163  * Vertex data either comes either from immediate mode registers or from
    164  * vertex arrays.
    165  * There appears to be no mixed mode (though we can force the pitch of
    166  * vertex arrays to 0, effectively reusing the same element over and over
    167  * again).
    168  *
    169  * Immediate mode is controlled by the INPUT_CNTL registers. I am not sure
    170  * if these registers influence vertex array processing.
    171  *
    172  * Vertex arrays are controlled via the 3D_LOAD_VBPNTR packet3.
    173  *
    174  * In both cases, vertex attributes are then passed through INPUT_ROUTE.
    175  *
    176  * Beginning with INPUT_ROUTE_0_0 is a list of WORDs that route vertex data
    177  * into the vertex processor's input registers.
    178  * The first word routes the first input, the second word the second, etc.
    179  * The corresponding input is routed into the register with the given index.
    180  * The list is ended by a word with INPUT_ROUTE_END set.
    181  *
    182  * Always set COMPONENTS_4 in immediate mode.
    183  */
    184 
    185 #define R300_VAP_INPUT_ROUTE_0_0            0x2150
    186 #       define R300_INPUT_ROUTE_COMPONENTS_1     (0 << 0)
    187 #       define R300_INPUT_ROUTE_COMPONENTS_2     (1 << 0)
    188 #       define R300_INPUT_ROUTE_COMPONENTS_3     (2 << 0)
    189 #       define R300_INPUT_ROUTE_COMPONENTS_4     (3 << 0)
    190 #       define R300_INPUT_ROUTE_COMPONENTS_RGBA  (4 << 0) /* GUESS */
    191 #       define R300_VAP_INPUT_ROUTE_IDX_SHIFT    8
    192 #       define R300_VAP_INPUT_ROUTE_IDX_MASK     (31 << 8) /* GUESS */
    193 #       define R300_VAP_INPUT_ROUTE_END          (1 << 13)
    194 #       define R300_INPUT_ROUTE_IMMEDIATE_MODE   (0 << 14) /* GUESS */
    195 #       define R300_INPUT_ROUTE_FLOAT            (1 << 14) /* GUESS */
    196 #       define R300_INPUT_ROUTE_UNSIGNED_BYTE    (2 << 14) /* GUESS */
    197 #       define R300_INPUT_ROUTE_FLOAT_COLOR      (3 << 14) /* GUESS */
    198 #define R300_VAP_INPUT_ROUTE_0_1            0x2154
    199 #define R300_VAP_INPUT_ROUTE_0_2            0x2158
    200 #define R300_VAP_INPUT_ROUTE_0_3            0x215C
    201 #define R300_VAP_INPUT_ROUTE_0_4            0x2160
    202 #define R300_VAP_INPUT_ROUTE_0_5            0x2164
    203 #define R300_VAP_INPUT_ROUTE_0_6            0x2168
    204 #define R300_VAP_INPUT_ROUTE_0_7            0x216C
    205 
    206 /* gap */
    207 
    208 /* Notes:
    209  *  - always set up to produce at least two attributes:
    210  *    if vertex program uses only position, fglrx will set normal, too
    211  *  - INPUT_CNTL_0_COLOR and INPUT_CNTL_COLOR bits are always equal.
    212  */
    213 #define R300_VAP_INPUT_CNTL_0               0x2180
    214 #       define R300_INPUT_CNTL_0_COLOR           0x00000001
    215 #define R300_VAP_INPUT_CNTL_1               0x2184
    216 #       define R300_INPUT_CNTL_POS               0x00000001
    217 #       define R300_INPUT_CNTL_NORMAL            0x00000002
    218 #       define R300_INPUT_CNTL_COLOR             0x00000004
    219 #       define R300_INPUT_CNTL_TC0               0x00000400
    220 #       define R300_INPUT_CNTL_TC1               0x00000800
    221 #       define R300_INPUT_CNTL_TC2               0x00001000 /* GUESS */
    222 #       define R300_INPUT_CNTL_TC3               0x00002000 /* GUESS */
    223 #       define R300_INPUT_CNTL_TC4               0x00004000 /* GUESS */
    224 #       define R300_INPUT_CNTL_TC5               0x00008000 /* GUESS */
    225 #       define R300_INPUT_CNTL_TC6               0x00010000 /* GUESS */
    226 #       define R300_INPUT_CNTL_TC7               0x00020000 /* GUESS */
    227 
    228 /* gap */
    229 
    230 /* Words parallel to INPUT_ROUTE_0; All words that are active in INPUT_ROUTE_0
    231  * are set to a swizzling bit pattern, other words are 0.
    232  *
    233  * In immediate mode, the pattern is always set to xyzw. In vertex array
    234  * mode, the swizzling pattern is e.g. used to set zw components in texture
    235  * coordinates with only tweo components.
    236  */
    237 #define R300_VAP_INPUT_ROUTE_1_0            0x21E0
    238 #       define R300_INPUT_ROUTE_SELECT_X    0
    239 #       define R300_INPUT_ROUTE_SELECT_Y    1
    240 #       define R300_INPUT_ROUTE_SELECT_Z    2
    241 #       define R300_INPUT_ROUTE_SELECT_W    3
    242 #       define R300_INPUT_ROUTE_SELECT_ZERO 4
    243 #       define R300_INPUT_ROUTE_SELECT_ONE  5
    244 #       define R300_INPUT_ROUTE_SELECT_MASK 7
    245 #       define R300_INPUT_ROUTE_X_SHIFT     0
    246 #       define R300_INPUT_ROUTE_Y_SHIFT     3
    247 #       define R300_INPUT_ROUTE_Z_SHIFT     6
    248 #       define R300_INPUT_ROUTE_W_SHIFT     9
    249 #       define R300_INPUT_ROUTE_ENABLE      (15 << 12)
    250 #define R300_VAP_INPUT_ROUTE_1_1            0x21E4
    251 #define R300_VAP_INPUT_ROUTE_1_2            0x21E8
    252 #define R300_VAP_INPUT_ROUTE_1_3            0x21EC
    253 #define R300_VAP_INPUT_ROUTE_1_4            0x21F0
    254 #define R300_VAP_INPUT_ROUTE_1_5            0x21F4
    255 #define R300_VAP_INPUT_ROUTE_1_6            0x21F8
    256 #define R300_VAP_INPUT_ROUTE_1_7            0x21FC
    257 
    258 /* END: Vertex data assembly */
    259 
    260 /* gap */
    261 
    262 /* BEGIN: Upload vertex program and data */
    263 
    264 /*
    265  * The programmable vertex shader unit has a memory bank of unknown size
    266  * that can be written to in 16 byte units by writing the address into
    267  * UPLOAD_ADDRESS, followed by data in UPLOAD_DATA (multiples of 4 DWORDs).
    268  *
    269  * Pointers into the memory bank are always in multiples of 16 bytes.
    270  *
    271  * The memory bank is divided into areas with fixed meaning.
    272  *
    273  * Starting at address UPLOAD_PROGRAM: Vertex program instructions.
    274  * Native limits reported by drivers from ATI suggest size 256 (i.e. 4KB),
    275  * whereas the difference between known addresses suggests size 512.
    276  *
    277  * Starting at address UPLOAD_PARAMETERS: Vertex program parameters.
    278  * Native reported limits and the VPI layout suggest size 256, whereas
    279  * difference between known addresses suggests size 512.
    280  *
    281  * At address UPLOAD_POINTSIZE is a vector (0, 0, ps, 0), where ps is the
    282  * floating point pointsize. The exact purpose of this state is uncertain,
    283  * as there is also the R300_RE_POINTSIZE register.
    284  *
    285  * Multiple vertex programs and parameter sets can be loaded at once,
    286  * which could explain the size discrepancy.
    287  */
    288 #define R300_VAP_PVS_UPLOAD_ADDRESS         0x2200
    289 #       define R300_PVS_UPLOAD_PROGRAM           0x00000000
    290 #       define R300_PVS_UPLOAD_PARAMETERS        0x00000200
    291 #       define R300_PVS_UPLOAD_POINTSIZE         0x00000406
    292 
    293 /* gap */
    294 
    295 #define R300_VAP_PVS_UPLOAD_DATA            0x2208
    296 
    297 /* END: Upload vertex program and data */
    298 
    299 /* gap */
    300 
    301 /* I do not know the purpose of this register. However, I do know that
    302  * it is set to 221C_CLEAR for clear operations and to 221C_NORMAL
    303  * for normal rendering.
    304  */
    305 #define R300_VAP_UNKNOWN_221C               0x221C
    306 #       define R300_221C_NORMAL                  0x00000000
    307 #       define R300_221C_CLEAR                   0x0001C000
    308 
    309 /* These seem to be per-pixel and per-vertex X and Y clipping planes. The first
    310  * plane is per-pixel and the second plane is per-vertex.
    311  *
    312  * This was determined by experimentation alone but I believe it is correct.
    313  *
    314  * These registers are called X_QUAD0_1_FL to X_QUAD0_4_FL by glxtest.
    315  */
    316 #define R300_VAP_CLIP_X_0                   0x2220
    317 #define R300_VAP_CLIP_X_1                   0x2224
    318 #define R300_VAP_CLIP_Y_0                   0x2228
    319 #define R300_VAP_CLIP_Y_1                   0x2230
    320 
    321 /* gap */
    322 
    323 /* Sometimes, END_OF_PKT and 0x2284=0 are the only commands sent between
    324  * rendering commands and overwriting vertex program parameters.
    325  * Therefore, I suspect writing zero to 0x2284 synchronizes the engine and
    326  * avoids bugs caused by still running shaders reading bad data from memory.
    327  */
    328 #define R300_VAP_PVS_STATE_FLUSH_REG        0x2284
    329 
    330 /* Absolutely no clue what this register is about. */
    331 #define R300_VAP_UNKNOWN_2288               0x2288
    332 #       define R300_2288_R300                    0x00750000 /* -- nh */
    333 #       define R300_2288_RV350                   0x0000FFFF /* -- Vladimir */
    334 
    335 /* gap */
    336 
    337 /* Addresses are relative to the vertex program instruction area of the
    338  * memory bank. PROGRAM_END points to the last instruction of the active
    339  * program
    340  *
    341  * The meaning of the two UNKNOWN fields is obviously not known. However,
    342  * experiments so far have shown that both *must* point to an instruction
    343  * inside the vertex program, otherwise the GPU locks up.
    344  *
    345  * fglrx usually sets CNTL_3_UNKNOWN to the end of the program and
    346  * R300_PVS_CNTL_1_POS_END_SHIFT points to instruction where last write to
    347  * position takes place.
    348  *
    349  * Most likely this is used to ignore rest of the program in cases
    350  * where group of verts arent visible. For some reason this "section"
    351  * is sometimes accepted other instruction that have no relationship with
    352  * position calculations.
    353  */
    354 #define R300_VAP_PVS_CNTL_1                 0x22D0
    355 #       define R300_PVS_CNTL_1_PROGRAM_START_SHIFT   0
    356 #       define R300_PVS_CNTL_1_POS_END_SHIFT         10
    357 #       define R300_PVS_CNTL_1_PROGRAM_END_SHIFT     20
    358 /* Addresses are relative the the vertex program parameters area. */
    359 #define R300_VAP_PVS_CNTL_2                 0x22D4
    360 #       define R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT 0
    361 #       define R300_PVS_CNTL_2_PARAM_COUNT_SHIFT  16
    362 #define R300_VAP_PVS_CNTL_3	           0x22D8
    363 #       define R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT 10
    364 #       define R300_PVS_CNTL_3_PROGRAM_UNKNOWN2_SHIFT 0
    365 
    366 /* The entire range from 0x2300 to 0x2AC inclusive seems to be used for
    367  * immediate vertices
    368  */
    369 #define R300_VAP_VTX_COLOR_R                0x2464
    370 #define R300_VAP_VTX_COLOR_G                0x2468
    371 #define R300_VAP_VTX_COLOR_B                0x246C
    372 #define R300_VAP_VTX_POS_0_X_1              0x2490 /* used for glVertex2*() */
    373 #define R300_VAP_VTX_POS_0_Y_1              0x2494
    374 #define R300_VAP_VTX_COLOR_PKD              0x249C /* RGBA */
    375 #define R300_VAP_VTX_POS_0_X_2              0x24A0 /* used for glVertex3*() */
    376 #define R300_VAP_VTX_POS_0_Y_2              0x24A4
    377 #define R300_VAP_VTX_POS_0_Z_2              0x24A8
    378 /* write 0 to indicate end of packet? */
    379 #define R300_VAP_VTX_END_OF_PKT             0x24AC
    380 
    381 /* gap */
    382 
    383 /* These are values from r300_reg/r300_reg.h - they are known to be correct
    384  * and are here so we can use one register file instead of several
    385  * - Vladimir
    386  */
    387 #define R300_GB_VAP_RASTER_VTX_FMT_0	0x4000
    388 #	define R300_GB_VAP_RASTER_VTX_FMT_0__POS_PRESENT	(1<<0)
    389 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_0_PRESENT	(1<<1)
    390 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_1_PRESENT	(1<<2)
    391 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_2_PRESENT	(1<<3)
    392 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_3_PRESENT	(1<<4)
    393 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_SPACE	(0xf<<5)
    394 #	define R300_GB_VAP_RASTER_VTX_FMT_0__PT_SIZE_PRESENT	(0x1<<16)
    395 
    396 #define R300_GB_VAP_RASTER_VTX_FMT_1	0x4004
    397 	/* each of the following is 3 bits wide, specifies number
    398 	   of components */
    399 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT	0
    400 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT	3
    401 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT	6
    402 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT	9
    403 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT	12
    404 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT	15
    405 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT	18
    406 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT	21
    407 
    408 /* UNK30 seems to enables point to quad transformation on textures
    409  * (or something closely related to that).
    410  * This bit is rather fatal at the time being due to lackings at pixel
    411  * shader side
    412  */
    413 #define R300_GB_ENABLE	0x4008
    414 #	define R300_GB_POINT_STUFF_ENABLE	(1<<0)
    415 #	define R300_GB_LINE_STUFF_ENABLE	(1<<1)
    416 #	define R300_GB_TRIANGLE_STUFF_ENABLE	(1<<2)
    417 #	define R300_GB_STENCIL_AUTO_ENABLE	(1<<4)
    418 #	define R300_GB_UNK31			(1<<31)
    419 	/* each of the following is 2 bits wide */
    420 #define R300_GB_TEX_REPLICATE	0
    421 #define R300_GB_TEX_ST		1
    422 #define R300_GB_TEX_STR		2
    423 #	define R300_GB_TEX0_SOURCE_SHIFT	16
    424 #	define R300_GB_TEX1_SOURCE_SHIFT	18
    425 #	define R300_GB_TEX2_SOURCE_SHIFT	20
    426 #	define R300_GB_TEX3_SOURCE_SHIFT	22
    427 #	define R300_GB_TEX4_SOURCE_SHIFT	24
    428 #	define R300_GB_TEX5_SOURCE_SHIFT	26
    429 #	define R300_GB_TEX6_SOURCE_SHIFT	28
    430 #	define R300_GB_TEX7_SOURCE_SHIFT	30
    431 
    432 /* MSPOS - positions for multisample antialiasing (?) */
    433 #define R300_GB_MSPOS0	0x4010
    434 	/* shifts - each of the fields is 4 bits */
    435 #	define R300_GB_MSPOS0__MS_X0_SHIFT	0
    436 #	define R300_GB_MSPOS0__MS_Y0_SHIFT	4
    437 #	define R300_GB_MSPOS0__MS_X1_SHIFT	8
    438 #	define R300_GB_MSPOS0__MS_Y1_SHIFT	12
    439 #	define R300_GB_MSPOS0__MS_X2_SHIFT	16
    440 #	define R300_GB_MSPOS0__MS_Y2_SHIFT	20
    441 #	define R300_GB_MSPOS0__MSBD0_Y		24
    442 #	define R300_GB_MSPOS0__MSBD0_X		28
    443 
    444 #define R300_GB_MSPOS1	0x4014
    445 #	define R300_GB_MSPOS1__MS_X3_SHIFT	0
    446 #	define R300_GB_MSPOS1__MS_Y3_SHIFT	4
    447 #	define R300_GB_MSPOS1__MS_X4_SHIFT	8
    448 #	define R300_GB_MSPOS1__MS_Y4_SHIFT	12
    449 #	define R300_GB_MSPOS1__MS_X5_SHIFT	16
    450 #	define R300_GB_MSPOS1__MS_Y5_SHIFT	20
    451 #	define R300_GB_MSPOS1__MSBD1		24
    452 
    453 
    454 #define R300_GB_TILE_CONFIG	0x4018
    455 #	define R300_GB_TILE_ENABLE	(1<<0)
    456 #	define R300_GB_TILE_PIPE_COUNT_RV300	0
    457 #	define R300_GB_TILE_PIPE_COUNT_R300	(3<<1)
    458 #	define R300_GB_TILE_PIPE_COUNT_R420	(7<<1)
    459 #	define R300_GB_TILE_PIPE_COUNT_RV410	(3<<1)
    460 #	define R300_GB_TILE_SIZE_8		0
    461 #	define R300_GB_TILE_SIZE_16		(1<<4)
    462 #	define R300_GB_TILE_SIZE_32		(2<<4)
    463 #	define R300_GB_SUPER_SIZE_1		(0<<6)
    464 #	define R300_GB_SUPER_SIZE_2		(1<<6)
    465 #	define R300_GB_SUPER_SIZE_4		(2<<6)
    466 #	define R300_GB_SUPER_SIZE_8		(3<<6)
    467 #	define R300_GB_SUPER_SIZE_16		(4<<6)
    468 #	define R300_GB_SUPER_SIZE_32		(5<<6)
    469 #	define R300_GB_SUPER_SIZE_64		(6<<6)
    470 #	define R300_GB_SUPER_SIZE_128		(7<<6)
    471 #	define R300_GB_SUPER_X_SHIFT		9	/* 3 bits wide */
    472 #	define R300_GB_SUPER_Y_SHIFT		12	/* 3 bits wide */
    473 #	define R300_GB_SUPER_TILE_A		0
    474 #	define R300_GB_SUPER_TILE_B		(1<<15)
    475 #	define R300_GB_SUBPIXEL_1_12		0
    476 #	define R300_GB_SUBPIXEL_1_16		(1<<16)
    477 
    478 #define R300_GB_FIFO_SIZE	0x4024
    479 	/* each of the following is 2 bits wide */
    480 #define R300_GB_FIFO_SIZE_32	0
    481 #define R300_GB_FIFO_SIZE_64	1
    482 #define R300_GB_FIFO_SIZE_128	2
    483 #define R300_GB_FIFO_SIZE_256	3
    484 #	define R300_SC_IFIFO_SIZE_SHIFT	0
    485 #	define R300_SC_TZFIFO_SIZE_SHIFT	2
    486 #	define R300_SC_BFIFO_SIZE_SHIFT	4
    487 
    488 #	define R300_US_OFIFO_SIZE_SHIFT	12
    489 #	define R300_US_WFIFO_SIZE_SHIFT	14
    490 	/* the following use the same constants as above, but meaning is
    491 	   is times 2 (i.e. instead of 32 words it means 64 */
    492 #	define R300_RS_TFIFO_SIZE_SHIFT	6
    493 #	define R300_RS_CFIFO_SIZE_SHIFT	8
    494 #	define R300_US_RAM_SIZE_SHIFT		10
    495 	/* watermarks, 3 bits wide */
    496 #	define R300_RS_HIGHWATER_COL_SHIFT	16
    497 #	define R300_RS_HIGHWATER_TEX_SHIFT	19
    498 #	define R300_OFIFO_HIGHWATER_SHIFT	22	/* two bits only */
    499 #	define R300_CUBE_FIFO_HIGHWATER_COL_SHIFT	24
    500 
    501 #define R300_GB_SELECT	0x401C
    502 #	define R300_GB_FOG_SELECT_C0A		0
    503 #	define R300_GB_FOG_SELECT_C1A		1
    504 #	define R300_GB_FOG_SELECT_C2A		2
    505 #	define R300_GB_FOG_SELECT_C3A		3
    506 #	define R300_GB_FOG_SELECT_1_1_W	4
    507 #	define R300_GB_FOG_SELECT_Z		5
    508 #	define R300_GB_DEPTH_SELECT_Z		0
    509 #	define R300_GB_DEPTH_SELECT_1_1_W	(1<<3)
    510 #	define R300_GB_W_SELECT_1_W		0
    511 #	define R300_GB_W_SELECT_1		(1<<4)
    512 
    513 #define R300_GB_AA_CONFIG		0x4020
    514 #	define R300_AA_DISABLE			0x00
    515 #	define R300_AA_ENABLE			0x01
    516 #	define R300_AA_SUBSAMPLES_2		0
    517 #	define R300_AA_SUBSAMPLES_3		(1<<1)
    518 #	define R300_AA_SUBSAMPLES_4		(2<<1)
    519 #	define R300_AA_SUBSAMPLES_6		(3<<1)
    520 
    521 /* gap */
    522 
    523 /* Zero to flush caches. */
    524 #define R300_TX_INVALTAGS                   0x4100
    525 #define R300_TX_FLUSH                       0x0
    526 
    527 /* The upper enable bits are guessed, based on fglrx reported limits. */
    528 #define R300_TX_ENABLE                      0x4104
    529 #       define R300_TX_ENABLE_0                  (1 << 0)
    530 #       define R300_TX_ENABLE_1                  (1 << 1)
    531 #       define R300_TX_ENABLE_2                  (1 << 2)
    532 #       define R300_TX_ENABLE_3                  (1 << 3)
    533 #       define R300_TX_ENABLE_4                  (1 << 4)
    534 #       define R300_TX_ENABLE_5                  (1 << 5)
    535 #       define R300_TX_ENABLE_6                  (1 << 6)
    536 #       define R300_TX_ENABLE_7                  (1 << 7)
    537 #       define R300_TX_ENABLE_8                  (1 << 8)
    538 #       define R300_TX_ENABLE_9                  (1 << 9)
    539 #       define R300_TX_ENABLE_10                 (1 << 10)
    540 #       define R300_TX_ENABLE_11                 (1 << 11)
    541 #       define R300_TX_ENABLE_12                 (1 << 12)
    542 #       define R300_TX_ENABLE_13                 (1 << 13)
    543 #       define R300_TX_ENABLE_14                 (1 << 14)
    544 #       define R300_TX_ENABLE_15                 (1 << 15)
    545 
    546 /* The pointsize is given in multiples of 6. The pointsize can be
    547  * enormous: Clear() renders a single point that fills the entire
    548  * framebuffer.
    549  */
    550 #define R300_RE_POINTSIZE                   0x421C
    551 #       define R300_POINTSIZE_Y_SHIFT            0
    552 #       define R300_POINTSIZE_Y_MASK             (0xFFFF << 0) /* GUESS */
    553 #       define R300_POINTSIZE_X_SHIFT            16
    554 #       define R300_POINTSIZE_X_MASK             (0xFFFF << 16) /* GUESS */
    555 #       define R300_POINTSIZE_MAX             (R300_POINTSIZE_Y_MASK / 6)
    556 
    557 /* The line width is given in multiples of 6.
    558  * In default mode lines are classified as vertical lines.
    559  * HO: horizontal
    560  * VE: vertical or horizontal
    561  * HO & VE: no classification
    562  */
    563 #define R300_RE_LINE_CNT                      0x4234
    564 #       define R300_LINESIZE_SHIFT            0
    565 #       define R300_LINESIZE_MASK             (0xFFFF << 0) /* GUESS */
    566 #       define R300_LINESIZE_MAX             (R300_LINESIZE_MASK / 6)
    567 #       define R300_LINE_CNT_HO               (1 << 16)
    568 #       define R300_LINE_CNT_VE               (1 << 17)
    569 
    570 /* Some sort of scale or clamp value for texcoordless textures. */
    571 #define R300_RE_UNK4238                       0x4238
    572 
    573 /* Something shade related */
    574 #define R300_RE_SHADE                         0x4274
    575 
    576 #define R300_RE_SHADE_MODEL                   0x4278
    577 #	define R300_RE_SHADE_MODEL_SMOOTH     0x3aaaa
    578 #	define R300_RE_SHADE_MODEL_FLAT       0x39595
    579 
    580 /* Dangerous */
    581 #define R300_RE_POLYGON_MODE                  0x4288
    582 #	define R300_PM_ENABLED                (1 << 0)
    583 #	define R300_PM_FRONT_POINT            (0 << 0)
    584 #	define R300_PM_BACK_POINT             (0 << 0)
    585 #	define R300_PM_FRONT_LINE             (1 << 4)
    586 #	define R300_PM_FRONT_FILL             (1 << 5)
    587 #	define R300_PM_BACK_LINE              (1 << 7)
    588 #	define R300_PM_BACK_FILL              (1 << 8)
    589 
    590 /* Fog parameters */
    591 #define R300_RE_FOG_SCALE                     0x4294
    592 #define R300_RE_FOG_START                     0x4298
    593 
    594 /* Not sure why there are duplicate of factor and constant values.
    595  * My best guess so far is that there are separate zbiases for test and write.
    596  * Ordering might be wrong.
    597  * Some of the tests indicate that fgl has a fallback implementation of zbias
    598  * via pixel shaders.
    599  */
    600 #define R300_RE_ZBIAS_CNTL                    0x42A0 /* GUESS */
    601 #define R300_RE_ZBIAS_T_FACTOR                0x42A4
    602 #define R300_RE_ZBIAS_T_CONSTANT              0x42A8
    603 #define R300_RE_ZBIAS_W_FACTOR                0x42AC
    604 #define R300_RE_ZBIAS_W_CONSTANT              0x42B0
    605 
    606 /* This register needs to be set to (1<<1) for RV350 to correctly
    607  * perform depth test (see --vb-triangles in r300_demo)
    608  * Don't know about other chips. - Vladimir
    609  * This is set to 3 when GL_POLYGON_OFFSET_FILL is on.
    610  * My guess is that there are two bits for each zbias primitive
    611  * (FILL, LINE, POINT).
    612  *  One to enable depth test and one for depth write.
    613  * Yet this doesn't explain why depth writes work ...
    614  */
    615 #define R300_RE_OCCLUSION_CNTL		    0x42B4
    616 #	define R300_OCCLUSION_ON		(1<<1)
    617 
    618 #define R300_RE_CULL_CNTL                   0x42B8
    619 #       define R300_CULL_FRONT                   (1 << 0)
    620 #       define R300_CULL_BACK                    (1 << 1)
    621 #       define R300_FRONT_FACE_CCW               (0 << 2)
    622 #       define R300_FRONT_FACE_CW                (1 << 2)
    623 
    624 
    625 /* BEGIN: Rasterization / Interpolators - many guesses */
    626 
    627 /* 0_UNKNOWN_18 has always been set except for clear operations.
    628  * TC_CNT is the number of incoming texture coordinate sets (i.e. it depends
    629  * on the vertex program, *not* the fragment program)
    630  */
    631 #define R300_RS_CNTL_0                      0x4300
    632 #       define R300_RS_CNTL_TC_CNT_SHIFT         2
    633 #       define R300_RS_CNTL_TC_CNT_MASK          (7 << 2)
    634 	/* number of color interpolators used */
    635 #	define R300_RS_CNTL_CI_CNT_SHIFT         7
    636 #       define R300_RS_CNTL_0_UNKNOWN_18         (1 << 18)
    637 	/* Guess: RS_CNTL_1 holds the index of the highest used RS_ROUTE_n
    638 	   register. */
    639 #define R300_RS_CNTL_1                      0x4304
    640 
    641 /* gap */
    642 
    643 /* Only used for texture coordinates.
    644  * Use the source field to route texture coordinate input from the
    645  * vertex program to the desired interpolator. Note that the source
    646  * field is relative to the outputs the vertex program *actually*
    647  * writes. If a vertex program only writes texcoord[1], this will
    648  * be source index 0.
    649  * Set INTERP_USED on all interpolators that produce data used by
    650  * the fragment program. INTERP_USED looks like a swizzling mask,
    651  * but I haven't seen it used that way.
    652  *
    653  * Note: The _UNKNOWN constants are always set in their respective
    654  * register. I don't know if this is necessary.
    655  */
    656 #define R300_RS_INTERP_0                    0x4310
    657 #define R300_RS_INTERP_1                    0x4314
    658 #       define R300_RS_INTERP_1_UNKNOWN          0x40
    659 #define R300_RS_INTERP_2                    0x4318
    660 #       define R300_RS_INTERP_2_UNKNOWN          0x80
    661 #define R300_RS_INTERP_3                    0x431C
    662 #       define R300_RS_INTERP_3_UNKNOWN          0xC0
    663 #define R300_RS_INTERP_4                    0x4320
    664 #define R300_RS_INTERP_5                    0x4324
    665 #define R300_RS_INTERP_6                    0x4328
    666 #define R300_RS_INTERP_7                    0x432C
    667 #       define R300_RS_INTERP_SRC_SHIFT          2
    668 #       define R300_RS_INTERP_SRC_MASK           (7 << 2)
    669 #       define R300_RS_INTERP_USED               0x00D10000
    670 
    671 /* These DWORDs control how vertex data is routed into fragment program
    672  * registers, after interpolators.
    673  */
    674 #define R300_RS_ROUTE_0                     0x4330
    675 #define R300_RS_ROUTE_1                     0x4334
    676 #define R300_RS_ROUTE_2                     0x4338
    677 #define R300_RS_ROUTE_3                     0x433C /* GUESS */
    678 #define R300_RS_ROUTE_4                     0x4340 /* GUESS */
    679 #define R300_RS_ROUTE_5                     0x4344 /* GUESS */
    680 #define R300_RS_ROUTE_6                     0x4348 /* GUESS */
    681 #define R300_RS_ROUTE_7                     0x434C /* GUESS */
    682 #       define R300_RS_ROUTE_SOURCE_INTERP_0     0
    683 #       define R300_RS_ROUTE_SOURCE_INTERP_1     1
    684 #       define R300_RS_ROUTE_SOURCE_INTERP_2     2
    685 #       define R300_RS_ROUTE_SOURCE_INTERP_3     3
    686 #       define R300_RS_ROUTE_SOURCE_INTERP_4     4
    687 #       define R300_RS_ROUTE_SOURCE_INTERP_5     5 /* GUESS */
    688 #       define R300_RS_ROUTE_SOURCE_INTERP_6     6 /* GUESS */
    689 #       define R300_RS_ROUTE_SOURCE_INTERP_7     7 /* GUESS */
    690 #       define R300_RS_ROUTE_ENABLE              (1 << 3) /* GUESS */
    691 #       define R300_RS_ROUTE_DEST_SHIFT          6
    692 #       define R300_RS_ROUTE_DEST_MASK           (31 << 6) /* GUESS */
    693 
    694 /* Special handling for color: When the fragment program uses color,
    695  * the ROUTE_0_COLOR bit is set and ROUTE_0_COLOR_DEST contains the
    696  * color register index.
    697  *
    698  * Apperently you may set the R300_RS_ROUTE_0_COLOR bit, but not provide any
    699  * R300_RS_ROUTE_0_COLOR_DEST value; this setup is used for clearing the state.
    700  * See r300_ioctl.c:r300EmitClearState. I'm not sure if this setup is strictly
    701  * correct or not. - Oliver.
    702  */
    703 #       define R300_RS_ROUTE_0_COLOR             (1 << 14)
    704 #       define R300_RS_ROUTE_0_COLOR_DEST_SHIFT  17
    705 #       define R300_RS_ROUTE_0_COLOR_DEST_MASK   (31 << 17) /* GUESS */
    706 /* As above, but for secondary color */
    707 #		define R300_RS_ROUTE_1_COLOR1            (1 << 14)
    708 #		define R300_RS_ROUTE_1_COLOR1_DEST_SHIFT 17
    709 #		define R300_RS_ROUTE_1_COLOR1_DEST_MASK  (31 << 17)
    710 #		define R300_RS_ROUTE_1_UNKNOWN11         (1 << 11)
    711 /* END: Rasterization / Interpolators - many guesses */
    712 
    713 /* Hierarchical Z Enable */
    714 #define R300_SC_HYPERZ                   0x43a4
    715 #	define R300_SC_HYPERZ_DISABLE     (0 << 0)
    716 #	define R300_SC_HYPERZ_ENABLE      (1 << 0)
    717 #	define R300_SC_HYPERZ_MIN         (0 << 1)
    718 #	define R300_SC_HYPERZ_MAX         (1 << 1)
    719 #	define R300_SC_HYPERZ_ADJ_256     (0 << 2)
    720 #	define R300_SC_HYPERZ_ADJ_128     (1 << 2)
    721 #	define R300_SC_HYPERZ_ADJ_64      (2 << 2)
    722 #	define R300_SC_HYPERZ_ADJ_32      (3 << 2)
    723 #	define R300_SC_HYPERZ_ADJ_16      (4 << 2)
    724 #	define R300_SC_HYPERZ_ADJ_8       (5 << 2)
    725 #	define R300_SC_HYPERZ_ADJ_4       (6 << 2)
    726 #	define R300_SC_HYPERZ_ADJ_2       (7 << 2)
    727 #	define R300_SC_HYPERZ_HZ_Z0MIN_NO (0 << 5)
    728 #	define R300_SC_HYPERZ_HZ_Z0MIN    (1 << 5)
    729 #	define R300_SC_HYPERZ_HZ_Z0MAX_NO (0 << 6)
    730 #	define R300_SC_HYPERZ_HZ_Z0MAX    (1 << 6)
    731 
    732 #define R300_SC_EDGERULE                 0x43a8
    733 
    734 /* BEGIN: Scissors and cliprects */
    735 
    736 /* There are four clipping rectangles. Their corner coordinates are inclusive.
    737  * Every pixel is assigned a number from 0 and 15 by setting bits 0-3 depending
    738  * on whether the pixel is inside cliprects 0-3, respectively. For example,
    739  * if a pixel is inside cliprects 0 and 1, but outside 2 and 3, it is assigned
    740  * the number 3 (binary 0011).
    741  * Iff the bit corresponding to the pixel's number in RE_CLIPRECT_CNTL is set,
    742  * the pixel is rasterized.
    743  *
    744  * In addition to this, there is a scissors rectangle. Only pixels inside the
    745  * scissors rectangle are drawn. (coordinates are inclusive)
    746  *
    747  * For some reason, the top-left corner of the framebuffer is at (1440, 1440)
    748  * for the purpose of clipping and scissors.
    749  */
    750 #define R300_RE_CLIPRECT_TL_0               0x43B0
    751 #define R300_RE_CLIPRECT_BR_0               0x43B4
    752 #define R300_RE_CLIPRECT_TL_1               0x43B8
    753 #define R300_RE_CLIPRECT_BR_1               0x43BC
    754 #define R300_RE_CLIPRECT_TL_2               0x43C0
    755 #define R300_RE_CLIPRECT_BR_2               0x43C4
    756 #define R300_RE_CLIPRECT_TL_3               0x43C8
    757 #define R300_RE_CLIPRECT_BR_3               0x43CC
    758 #       define R300_CLIPRECT_OFFSET              1440
    759 #       define R300_CLIPRECT_MASK                0x1FFF
    760 #       define R300_CLIPRECT_X_SHIFT             0
    761 #       define R300_CLIPRECT_X_MASK              (0x1FFF << 0)
    762 #       define R300_CLIPRECT_Y_SHIFT             13
    763 #       define R300_CLIPRECT_Y_MASK              (0x1FFF << 13)
    764 #define R300_RE_CLIPRECT_CNTL               0x43D0
    765 #       define R300_CLIP_OUT                     (1 << 0)
    766 #       define R300_CLIP_0                       (1 << 1)
    767 #       define R300_CLIP_1                       (1 << 2)
    768 #       define R300_CLIP_10                      (1 << 3)
    769 #       define R300_CLIP_2                       (1 << 4)
    770 #       define R300_CLIP_20                      (1 << 5)
    771 #       define R300_CLIP_21                      (1 << 6)
    772 #       define R300_CLIP_210                     (1 << 7)
    773 #       define R300_CLIP_3                       (1 << 8)
    774 #       define R300_CLIP_30                      (1 << 9)
    775 #       define R300_CLIP_31                      (1 << 10)
    776 #       define R300_CLIP_310                     (1 << 11)
    777 #       define R300_CLIP_32                      (1 << 12)
    778 #       define R300_CLIP_320                     (1 << 13)
    779 #       define R300_CLIP_321                     (1 << 14)
    780 #       define R300_CLIP_3210                    (1 << 15)
    781 
    782 /* gap */
    783 
    784 #define R300_RE_SCISSORS_TL                 0x43E0
    785 #define R300_RE_SCISSORS_BR                 0x43E4
    786 #       define R300_SCISSORS_OFFSET              1440
    787 #       define R300_SCISSORS_X_SHIFT             0
    788 #       define R300_SCISSORS_X_MASK              (0x1FFF << 0)
    789 #       define R300_SCISSORS_Y_SHIFT             13
    790 #       define R300_SCISSORS_Y_MASK              (0x1FFF << 13)
    791 /* END: Scissors and cliprects */
    792 
    793 /* BEGIN: Texture specification */
    794 
    795 /*
    796  * The texture specification dwords are grouped by meaning and not by texture
    797  * unit. This means that e.g. the offset for texture image unit N is found in
    798  * register TX_OFFSET_0 + (4*N)
    799  */
    800 #define R300_TX_FILTER_0                    0x4400
    801 #       define R300_TX_REPEAT                    0
    802 #       define R300_TX_MIRRORED                  1
    803 #       define R300_TX_CLAMP                     4
    804 #       define R300_TX_CLAMP_TO_EDGE             2
    805 #       define R300_TX_CLAMP_TO_BORDER           6
    806 #       define R300_TX_WRAP_S_SHIFT              0
    807 #       define R300_TX_WRAP_S_MASK               (7 << 0)
    808 #       define R300_TX_WRAP_T_SHIFT              3
    809 #       define R300_TX_WRAP_T_MASK               (7 << 3)
    810 #       define R300_TX_WRAP_Q_SHIFT              6
    811 #       define R300_TX_WRAP_Q_MASK               (7 << 6)
    812 #       define R300_TX_MAG_FILTER_NEAREST        (1 << 9)
    813 #       define R300_TX_MAG_FILTER_LINEAR         (2 << 9)
    814 #       define R300_TX_MAG_FILTER_MASK           (3 << 9)
    815 #       define R300_TX_MIN_FILTER_NEAREST        (1 << 11)
    816 #       define R300_TX_MIN_FILTER_LINEAR         (2 << 11)
    817 #	define R300_TX_MIN_FILTER_NEAREST_MIP_NEAREST       (5  <<  11)
    818 #	define R300_TX_MIN_FILTER_NEAREST_MIP_LINEAR        (9  <<  11)
    819 #	define R300_TX_MIN_FILTER_LINEAR_MIP_NEAREST        (6  <<  11)
    820 #	define R300_TX_MIN_FILTER_LINEAR_MIP_LINEAR         (10 <<  11)
    821 
    822 /* NOTE: NEAREST doesn't seem to exist.
    823  * Im not seting MAG_FILTER_MASK and (3 << 11) on for all
    824  * anisotropy modes because that would void selected mag filter
    825  */
    826 #	define R300_TX_MIN_FILTER_ANISO_NEAREST             (0 << 13)
    827 #	define R300_TX_MIN_FILTER_ANISO_LINEAR              (0 << 13)
    828 #	define R300_TX_MIN_FILTER_ANISO_NEAREST_MIP_NEAREST (1 << 13)
    829 #	define R300_TX_MIN_FILTER_ANISO_NEAREST_MIP_LINEAR  (2 << 13)
    830 #       define R300_TX_MIN_FILTER_MASK   ( (15 << 11) | (3 << 13) )
    831 #	define R300_TX_MAX_ANISO_1_TO_1  (0 << 21)
    832 #	define R300_TX_MAX_ANISO_2_TO_1  (2 << 21)
    833 #	define R300_TX_MAX_ANISO_4_TO_1  (4 << 21)
    834 #	define R300_TX_MAX_ANISO_8_TO_1  (6 << 21)
    835 #	define R300_TX_MAX_ANISO_16_TO_1 (8 << 21)
    836 #	define R300_TX_MAX_ANISO_MASK    (14 << 21)
    837 
    838 #define R300_TX_FILTER1_0                      0x4440
    839 #	define R300_CHROMA_KEY_MODE_DISABLE    0
    840 #	define R300_CHROMA_KEY_FORCE	       1
    841 #	define R300_CHROMA_KEY_BLEND           2
    842 #	define R300_MC_ROUND_NORMAL            (0<<2)
    843 #	define R300_MC_ROUND_MPEG4             (1<<2)
    844 #	define R300_LOD_BIAS_MASK	    0x1fff
    845 #	define R300_EDGE_ANISO_EDGE_DIAG       (0<<13)
    846 #	define R300_EDGE_ANISO_EDGE_ONLY       (1<<13)
    847 #	define R300_MC_COORD_TRUNCATE_DISABLE  (0<<14)
    848 #	define R300_MC_COORD_TRUNCATE_MPEG     (1<<14)
    849 #	define R300_TX_TRI_PERF_0_8            (0<<15)
    850 #	define R300_TX_TRI_PERF_1_8            (1<<15)
    851 #	define R300_TX_TRI_PERF_1_4            (2<<15)
    852 #	define R300_TX_TRI_PERF_3_8            (3<<15)
    853 #	define R300_ANISO_THRESHOLD_MASK       (7<<17)
    854 
    855 #define R300_TX_SIZE_0                      0x4480
    856 #       define R300_TX_WIDTHMASK_SHIFT           0
    857 #       define R300_TX_WIDTHMASK_MASK            (2047 << 0)
    858 #       define R300_TX_HEIGHTMASK_SHIFT          11
    859 #       define R300_TX_HEIGHTMASK_MASK           (2047 << 11)
    860 #       define R300_TX_UNK23                     (1 << 23)
    861 #       define R300_TX_MAX_MIP_LEVEL_SHIFT       26
    862 #       define R300_TX_MAX_MIP_LEVEL_MASK        (0xf << 26)
    863 #       define R300_TX_SIZE_PROJECTED            (1<<30)
    864 #       define R300_TX_SIZE_TXPITCH_EN           (1<<31)
    865 #define R300_TX_FORMAT_0                    0x44C0
    866 	/* The interpretation of the format word by Wladimir van der Laan */
    867 	/* The X, Y, Z and W refer to the layout of the components.
    868 	   They are given meanings as R, G, B and Alpha by the swizzle
    869 	   specification */
    870 #	define R300_TX_FORMAT_X8		    0x0
    871 #	define R300_TX_FORMAT_X16		    0x1
    872 #	define R300_TX_FORMAT_Y4X4		    0x2
    873 #	define R300_TX_FORMAT_Y8X8		    0x3
    874 #	define R300_TX_FORMAT_Y16X16		    0x4
    875 #	define R300_TX_FORMAT_Z3Y3X2		    0x5
    876 #	define R300_TX_FORMAT_Z5Y6X5		    0x6
    877 #	define R300_TX_FORMAT_Z6Y5X5		    0x7
    878 #	define R300_TX_FORMAT_Z11Y11X10		    0x8
    879 #	define R300_TX_FORMAT_Z10Y11X11		    0x9
    880 #	define R300_TX_FORMAT_W4Z4Y4X4		    0xA
    881 #	define R300_TX_FORMAT_W1Z5Y5X5		    0xB
    882 #	define R300_TX_FORMAT_W8Z8Y8X8		    0xC
    883 #	define R300_TX_FORMAT_W2Z10Y10X10	    0xD
    884 #	define R300_TX_FORMAT_W16Z16Y16X16	    0xE
    885 #	define R300_TX_FORMAT_DXT1		    0xF
    886 #	define R300_TX_FORMAT_DXT3		    0x10
    887 #	define R300_TX_FORMAT_DXT5		    0x11
    888 #	define R300_TX_FORMAT_D3DMFT_CxV8U8	    0x12     /* no swizzle */
    889 #	define R300_TX_FORMAT_A8R8G8B8		    0x13     /* no swizzle */
    890 #	define R300_TX_FORMAT_B8G8_B8G8		    0x14     /* no swizzle */
    891 #	define R300_TX_FORMAT_G8R8_G8B8		    0x15     /* no swizzle */
    892 	/* 0x16 - some 16 bit green format.. ?? */
    893 #	define R300_TX_FORMAT_UNK25		   (1 << 25) /* no swizzle */
    894 #	define R300_TX_FORMAT_CUBIC_MAP		   (1 << 26)
    895 
    896 	/* gap */
    897 	/* Floating point formats */
    898 	/* Note - hardware supports both 16 and 32 bit floating point */
    899 #	define R300_TX_FORMAT_FL_I16		    0x18
    900 #	define R300_TX_FORMAT_FL_I16A16		    0x19
    901 #	define R300_TX_FORMAT_FL_R16G16B16A16	    0x1A
    902 #	define R300_TX_FORMAT_FL_I32		    0x1B
    903 #	define R300_TX_FORMAT_FL_I32A32		    0x1C
    904 #	define R300_TX_FORMAT_FL_R32G32B32A32	    0x1D
    905 #	define R300_TX_FORMAT_ATI2N		    0x1F
    906 	/* alpha modes, convenience mostly */
    907 	/* if you have alpha, pick constant appropriate to the
    908 	   number of channels (1 for I8, 2 for I8A8, 4 for R8G8B8A8, etc */
    909 #	define R300_TX_FORMAT_ALPHA_1CH		    0x000
    910 #	define R300_TX_FORMAT_ALPHA_2CH		    0x200
    911 #	define R300_TX_FORMAT_ALPHA_4CH		    0x600
    912 #	define R300_TX_FORMAT_ALPHA_NONE	    0xA00
    913 	/* Swizzling */
    914 	/* constants */
    915 #	define R300_TX_FORMAT_X		0
    916 #	define R300_TX_FORMAT_Y		1
    917 #	define R300_TX_FORMAT_Z		2
    918 #	define R300_TX_FORMAT_W		3
    919 #	define R300_TX_FORMAT_ZERO	4
    920 #	define R300_TX_FORMAT_ONE	5
    921 	/* 2.0*Z, everything above 1.0 is set to 0.0 */
    922 #	define R300_TX_FORMAT_CUT_Z	6
    923 	/* 2.0*W, everything above 1.0 is set to 0.0 */
    924 #	define R300_TX_FORMAT_CUT_W	7
    925 
    926 #	define R300_TX_FORMAT_B_SHIFT	18
    927 #	define R300_TX_FORMAT_G_SHIFT	15
    928 #	define R300_TX_FORMAT_R_SHIFT	12
    929 #	define R300_TX_FORMAT_A_SHIFT	9
    930 	/* Convenience macro to take care of layout and swizzling */
    931 #	define R300_EASY_TX_FORMAT(B, G, R, A, FMT)	(		\
    932 		((R300_TX_FORMAT_##B)<<R300_TX_FORMAT_B_SHIFT)		\
    933 		| ((R300_TX_FORMAT_##G)<<R300_TX_FORMAT_G_SHIFT)	\
    934 		| ((R300_TX_FORMAT_##R)<<R300_TX_FORMAT_R_SHIFT)	\
    935 		| ((R300_TX_FORMAT_##A)<<R300_TX_FORMAT_A_SHIFT)	\
    936 		| (R300_TX_FORMAT_##FMT)				\
    937 		)
    938 	/* These can be ORed with result of R300_EASY_TX_FORMAT()
    939 	   We don't really know what they do. Take values from a
    940            constant color ? */
    941 #	define R300_TX_FORMAT_CONST_X		(1<<5)
    942 #	define R300_TX_FORMAT_CONST_Y		(2<<5)
    943 #	define R300_TX_FORMAT_CONST_Z		(4<<5)
    944 #	define R300_TX_FORMAT_CONST_W		(8<<5)
    945 
    946 #	define R300_TX_FORMAT_YUV_MODE		0x00800000
    947 
    948 #define R300_TX_PITCH_0			    0x4500 /* obvious missing in gap */
    949 #define R300_TX_OFFSET_0                    0x4540
    950 	/* BEGIN: Guess from R200 */
    951 #       define R300_TXO_ENDIAN_NO_SWAP           (0 << 0)
    952 #       define R300_TXO_ENDIAN_BYTE_SWAP         (1 << 0)
    953 #       define R300_TXO_ENDIAN_WORD_SWAP         (2 << 0)
    954 #       define R300_TXO_ENDIAN_HALFDW_SWAP       (3 << 0)
    955 #       define R300_TXO_MACRO_TILE               (1 << 2)
    956 #       define R300_TXO_MICRO_TILE               (1 << 3)
    957 #       define R300_TXO_MICRO_TILE_SQUARE        (2 << 3)
    958 #       define R300_TXO_OFFSET_MASK              0xffffffe0
    959 #       define R300_TXO_OFFSET_SHIFT             5
    960 	/* END: Guess from R200 */
    961 
    962 /* 32 bit chroma key */
    963 #define R300_TX_CHROMA_KEY_0                      0x4580
    964 /* ff00ff00 == { 0, 1.0, 0, 1.0 } */
    965 #define R300_TX_BORDER_COLOR_0              0x45C0
    966 
    967 /* END: Texture specification */
    968 
    969 /* BEGIN: Fragment program instruction set */
    970 
    971 /* Fragment programs are written directly into register space.
    972  * There are separate instruction streams for texture instructions and ALU
    973  * instructions.
    974  * In order to synchronize these streams, the program is divided into up
    975  * to 4 nodes. Each node begins with a number of TEX operations, followed
    976  * by a number of ALU operations.
    977  * The first node can have zero TEX ops, all subsequent nodes must have at
    978  * least
    979  * one TEX ops.
    980  * All nodes must have at least one ALU op.
    981  *
    982  * The index of the last node is stored in PFS_CNTL_0: A value of 0 means
    983  * 1 node, a value of 3 means 4 nodes.
    984  * The total amount of instructions is defined in PFS_CNTL_2. The offsets are
    985  * offsets into the respective instruction streams, while *_END points to the
    986  * last instruction relative to this offset.
    987  */
    988 #define R300_PFS_CNTL_0                     0x4600
    989 #       define R300_PFS_CNTL_LAST_NODES_SHIFT    0
    990 #       define R300_PFS_CNTL_LAST_NODES_MASK     (3 << 0)
    991 #       define R300_PFS_CNTL_FIRST_NODE_HAS_TEX  (1 << 3)
    992 #define R300_PFS_CNTL_1                     0x4604
    993 /* There is an unshifted value here which has so far always been equal to the
    994  * index of the highest used temporary register.
    995  */
    996 #define R300_PFS_CNTL_2                     0x4608
    997 #       define R300_PFS_CNTL_ALU_OFFSET_SHIFT    0
    998 #       define R300_PFS_CNTL_ALU_OFFSET_MASK     (63 << 0)
    999 #       define R300_PFS_CNTL_ALU_END_SHIFT       6
   1000 #       define R300_PFS_CNTL_ALU_END_MASK        (63 << 6)
   1001 #       define R300_PFS_CNTL_TEX_OFFSET_SHIFT    12
   1002 #       define R300_PFS_CNTL_TEX_OFFSET_MASK     (31 << 12) /* GUESS */
   1003 #       define R300_PFS_CNTL_TEX_END_SHIFT       18
   1004 #       define R300_PFS_CNTL_TEX_END_MASK        (31 << 18) /* GUESS */
   1005 
   1006 /* gap */
   1007 
   1008 /* Nodes are stored backwards. The last active node is always stored in
   1009  * PFS_NODE_3.
   1010  * Example: In a 2-node program, NODE_0 and NODE_1 are set to 0. The
   1011  * first node is stored in NODE_2, the second node is stored in NODE_3.
   1012  *
   1013  * Offsets are relative to the master offset from PFS_CNTL_2.
   1014  */
   1015 #define R300_PFS_NODE_0                     0x4610
   1016 #define R300_PFS_NODE_1                     0x4614
   1017 #define R300_PFS_NODE_2                     0x4618
   1018 #define R300_PFS_NODE_3                     0x461C
   1019 #       define R300_PFS_NODE_ALU_OFFSET_SHIFT    0
   1020 #       define R300_PFS_NODE_ALU_OFFSET_MASK     (63 << 0)
   1021 #       define R300_PFS_NODE_ALU_END_SHIFT       6
   1022 #       define R300_PFS_NODE_ALU_END_MASK        (63 << 6)
   1023 #       define R300_PFS_NODE_TEX_OFFSET_SHIFT    12
   1024 #       define R300_PFS_NODE_TEX_OFFSET_MASK     (31 << 12)
   1025 #       define R300_PFS_NODE_TEX_END_SHIFT       17
   1026 #       define R300_PFS_NODE_TEX_END_MASK        (31 << 17)
   1027 #		define R300_PFS_NODE_OUTPUT_COLOR        (1 << 22)
   1028 #		define R300_PFS_NODE_OUTPUT_DEPTH        (1 << 23)
   1029 
   1030 /* TEX
   1031  * As far as I can tell, texture instructions cannot write into output
   1032  * registers directly. A subsequent ALU instruction is always necessary,
   1033  * even if it's just MAD o0, r0, 1, 0
   1034  */
   1035 #define R300_PFS_TEXI_0                     0x4620
   1036 #	define R300_FPITX_SRC_SHIFT              0
   1037 #	define R300_FPITX_SRC_MASK               (31 << 0)
   1038 	/* GUESS */
   1039 #	define R300_FPITX_SRC_CONST              (1 << 5)
   1040 #	define R300_FPITX_DST_SHIFT              6
   1041 #	define R300_FPITX_DST_MASK               (31 << 6)
   1042 #	define R300_FPITX_IMAGE_SHIFT            11
   1043 	/* GUESS based on layout and native limits */
   1044 #       define R300_FPITX_IMAGE_MASK             (15 << 11)
   1045 /* Unsure if these are opcodes, or some kind of bitfield, but this is how
   1046  * they were set when I checked
   1047  */
   1048 #	define R300_FPITX_OPCODE_SHIFT		15
   1049 #		define R300_FPITX_OP_TEX	1
   1050 #		define R300_FPITX_OP_KIL	2
   1051 #		define R300_FPITX_OP_TXP	3
   1052 #		define R300_FPITX_OP_TXB	4
   1053 #	define R300_FPITX_OPCODE_MASK           (7 << 15)
   1054 
   1055 /* ALU
   1056  * The ALU instructions register blocks are enumerated according to the order
   1057  * in which fglrx. I assume there is space for 64 instructions, since
   1058  * each block has space for a maximum of 64 DWORDs, and this matches reported
   1059  * native limits.
   1060  *
   1061  * The basic functional block seems to be one MAD for each color and alpha,
   1062  * and an adder that adds all components after the MUL.
   1063  *  - ADD, MUL, MAD etc.: use MAD with appropriate neutral operands
   1064  *  - DP4: Use OUTC_DP4, OUTA_DP4
   1065  *  - DP3: Use OUTC_DP3, OUTA_DP4, appropriate alpha operands
   1066  *  - DPH: Use OUTC_DP4, OUTA_DP4, appropriate alpha operands
   1067  *  - CMPH: If ARG2 > 0.5, return ARG0, else return ARG1
   1068  *  - CMP: If ARG2 < 0, return ARG1, else return ARG0
   1069  *  - FLR: use FRC+MAD
   1070  *  - XPD: use MAD+MAD
   1071  *  - SGE, SLT: use MAD+CMP
   1072  *  - RSQ: use ABS modifier for argument
   1073  *  - Use OUTC_REPL_ALPHA to write results of an alpha-only operation
   1074  *    (e.g. RCP) into color register
   1075  *  - apparently, there's no quick DST operation
   1076  *  - fglrx set FPI2_UNKNOWN_31 on a "MAD fragment.color, tmp0, tmp1, tmp2"
   1077  *  - fglrx set FPI2_UNKNOWN_31 on a "MAX r2, r1, c0"
   1078  *  - fglrx once set FPI0_UNKNOWN_31 on a "FRC r1, r1"
   1079  *
   1080  * Operand selection
   1081  * First stage selects three sources from the available registers and
   1082  * constant parameters. This is defined in INSTR1 (color) and INSTR3 (alpha).
   1083  * fglrx sorts the three source fields: Registers before constants,
   1084  * lower indices before higher indices; I do not know whether this is
   1085  * necessary.
   1086  *
   1087  * fglrx fills unused sources with "read constant 0"
   1088  * According to specs, you cannot select more than two different constants.
   1089  *
   1090  * Second stage selects the operands from the sources. This is defined in
   1091  * INSTR0 (color) and INSTR2 (alpha). You can also select the special constants
   1092  * zero and one.
   1093  * Swizzling and negation happens in this stage, as well.
   1094  *
   1095  * Important: Color and alpha seem to be mostly separate, i.e. their sources
   1096  * selection appears to be fully independent (the register storage is probably
   1097  * physically split into a color and an alpha section).
   1098  * However (because of the apparent physical split), there is some interaction
   1099  * WRT swizzling. If, for example, you want to load an R component into an
   1100  * Alpha operand, this R component is taken from a *color* source, not from
   1101  * an alpha source. The corresponding register doesn't even have to appear in
   1102  * the alpha sources list. (I hope this all makes sense to you)
   1103  *
   1104  * Destination selection
   1105  * The destination register index is in FPI1 (color) and FPI3 (alpha)
   1106  * together with enable bits.
   1107  * There are separate enable bits for writing into temporary registers
   1108  * (DSTC_REG_* /DSTA_REG) and and program output registers (DSTC_OUTPUT_*
   1109  * /DSTA_OUTPUT). You can write to both at once, or not write at all (the
   1110  * same index must be used for both).
   1111  *
   1112  * Note: There is a special form for LRP
   1113  *  - Argument order is the same as in ARB_fragment_program.
   1114  *  - Operation is MAD
   1115  *  - ARG1 is set to ARGC_SRC1C_LRP/ARGC_SRC1A_LRP
   1116  *  - Set FPI0/FPI2_SPECIAL_LRP
   1117  * Arbitrary LRP (including support for swizzling) requires vanilla MAD+MAD
   1118  */
   1119 #define R300_PFS_INSTR1_0                   0x46C0
   1120 #       define R300_FPI1_SRC0C_SHIFT             0
   1121 #       define R300_FPI1_SRC0C_MASK              (31 << 0)
   1122 #       define R300_FPI1_SRC0C_CONST             (1 << 5)
   1123 #       define R300_FPI1_SRC1C_SHIFT             6
   1124 #       define R300_FPI1_SRC1C_MASK              (31 << 6)
   1125 #       define R300_FPI1_SRC1C_CONST             (1 << 11)
   1126 #       define R300_FPI1_SRC2C_SHIFT             12
   1127 #       define R300_FPI1_SRC2C_MASK              (31 << 12)
   1128 #       define R300_FPI1_SRC2C_CONST             (1 << 17)
   1129 #       define R300_FPI1_SRC_MASK                0x0003ffff
   1130 #       define R300_FPI1_DSTC_SHIFT              18
   1131 #       define R300_FPI1_DSTC_MASK               (31 << 18)
   1132 #		define R300_FPI1_DSTC_REG_MASK_SHIFT     23
   1133 #       define R300_FPI1_DSTC_REG_X              (1 << 23)
   1134 #       define R300_FPI1_DSTC_REG_Y              (1 << 24)
   1135 #       define R300_FPI1_DSTC_REG_Z              (1 << 25)
   1136 #		define R300_FPI1_DSTC_OUTPUT_MASK_SHIFT  26
   1137 #       define R300_FPI1_DSTC_OUTPUT_X           (1 << 26)
   1138 #       define R300_FPI1_DSTC_OUTPUT_Y           (1 << 27)
   1139 #       define R300_FPI1_DSTC_OUTPUT_Z           (1 << 28)
   1140 
   1141 #define R300_PFS_INSTR3_0                   0x47C0
   1142 #       define R300_FPI3_SRC0A_SHIFT             0
   1143 #       define R300_FPI3_SRC0A_MASK              (31 << 0)
   1144 #       define R300_FPI3_SRC0A_CONST             (1 << 5)
   1145 #       define R300_FPI3_SRC1A_SHIFT             6
   1146 #       define R300_FPI3_SRC1A_MASK              (31 << 6)
   1147 #       define R300_FPI3_SRC1A_CONST             (1 << 11)
   1148 #       define R300_FPI3_SRC2A_SHIFT             12
   1149 #       define R300_FPI3_SRC2A_MASK              (31 << 12)
   1150 #       define R300_FPI3_SRC2A_CONST             (1 << 17)
   1151 #       define R300_FPI3_SRC_MASK                0x0003ffff
   1152 #       define R300_FPI3_DSTA_SHIFT              18
   1153 #       define R300_FPI3_DSTA_MASK               (31 << 18)
   1154 #       define R300_FPI3_DSTA_REG                (1 << 23)
   1155 #       define R300_FPI3_DSTA_OUTPUT             (1 << 24)
   1156 #		define R300_FPI3_DSTA_DEPTH              (1 << 27)
   1157 
   1158 #define R300_PFS_INSTR0_0                   0x48C0
   1159 #       define R300_FPI0_ARGC_SRC0C_XYZ          0
   1160 #       define R300_FPI0_ARGC_SRC0C_XXX          1
   1161 #       define R300_FPI0_ARGC_SRC0C_YYY          2
   1162 #       define R300_FPI0_ARGC_SRC0C_ZZZ          3
   1163 #       define R300_FPI0_ARGC_SRC1C_XYZ          4
   1164 #       define R300_FPI0_ARGC_SRC1C_XXX          5
   1165 #       define R300_FPI0_ARGC_SRC1C_YYY          6
   1166 #       define R300_FPI0_ARGC_SRC1C_ZZZ          7
   1167 #       define R300_FPI0_ARGC_SRC2C_XYZ          8
   1168 #       define R300_FPI0_ARGC_SRC2C_XXX          9
   1169 #       define R300_FPI0_ARGC_SRC2C_YYY          10
   1170 #       define R300_FPI0_ARGC_SRC2C_ZZZ          11
   1171 #       define R300_FPI0_ARGC_SRC0A              12
   1172 #       define R300_FPI0_ARGC_SRC1A              13
   1173 #       define R300_FPI0_ARGC_SRC2A              14
   1174 #       define R300_FPI0_ARGC_SRC1C_LRP          15
   1175 #       define R300_FPI0_ARGC_ZERO               20
   1176 #       define R300_FPI0_ARGC_ONE                21
   1177 	/* GUESS */
   1178 #       define R300_FPI0_ARGC_HALF               22
   1179 #       define R300_FPI0_ARGC_SRC0C_YZX          23
   1180 #       define R300_FPI0_ARGC_SRC1C_YZX          24
   1181 #       define R300_FPI0_ARGC_SRC2C_YZX          25
   1182 #       define R300_FPI0_ARGC_SRC0C_ZXY          26
   1183 #       define R300_FPI0_ARGC_SRC1C_ZXY          27
   1184 #       define R300_FPI0_ARGC_SRC2C_ZXY          28
   1185 #       define R300_FPI0_ARGC_SRC0CA_WZY         29
   1186 #       define R300_FPI0_ARGC_SRC1CA_WZY         30
   1187 #       define R300_FPI0_ARGC_SRC2CA_WZY         31
   1188 
   1189 #       define R300_FPI0_ARG0C_SHIFT             0
   1190 #       define R300_FPI0_ARG0C_MASK              (31 << 0)
   1191 #       define R300_FPI0_ARG0C_NEG               (1 << 5)
   1192 #       define R300_FPI0_ARG0C_ABS               (1 << 6)
   1193 #       define R300_FPI0_ARG1C_SHIFT             7
   1194 #       define R300_FPI0_ARG1C_MASK              (31 << 7)
   1195 #       define R300_FPI0_ARG1C_NEG               (1 << 12)
   1196 #       define R300_FPI0_ARG1C_ABS               (1 << 13)
   1197 #       define R300_FPI0_ARG2C_SHIFT             14
   1198 #       define R300_FPI0_ARG2C_MASK              (31 << 14)
   1199 #       define R300_FPI0_ARG2C_NEG               (1 << 19)
   1200 #       define R300_FPI0_ARG2C_ABS               (1 << 20)
   1201 #       define R300_FPI0_SPECIAL_LRP             (1 << 21)
   1202 #       define R300_FPI0_OUTC_MAD                (0 << 23)
   1203 #       define R300_FPI0_OUTC_DP3                (1 << 23)
   1204 #       define R300_FPI0_OUTC_DP4                (2 << 23)
   1205 #       define R300_FPI0_OUTC_MIN                (4 << 23)
   1206 #       define R300_FPI0_OUTC_MAX                (5 << 23)
   1207 #       define R300_FPI0_OUTC_CMPH               (7 << 23)
   1208 #       define R300_FPI0_OUTC_CMP                (8 << 23)
   1209 #       define R300_FPI0_OUTC_FRC                (9 << 23)
   1210 #       define R300_FPI0_OUTC_REPL_ALPHA         (10 << 23)
   1211 #       define R300_FPI0_OUTC_SAT                (1 << 30)
   1212 #       define R300_FPI0_INSERT_NOP              (1 << 31)
   1213 
   1214 #define R300_PFS_INSTR2_0                   0x49C0
   1215 #       define R300_FPI2_ARGA_SRC0C_X            0
   1216 #       define R300_FPI2_ARGA_SRC0C_Y            1
   1217 #       define R300_FPI2_ARGA_SRC0C_Z            2
   1218 #       define R300_FPI2_ARGA_SRC1C_X            3
   1219 #       define R300_FPI2_ARGA_SRC1C_Y            4
   1220 #       define R300_FPI2_ARGA_SRC1C_Z            5
   1221 #       define R300_FPI2_ARGA_SRC2C_X            6
   1222 #       define R300_FPI2_ARGA_SRC2C_Y            7
   1223 #       define R300_FPI2_ARGA_SRC2C_Z            8
   1224 #       define R300_FPI2_ARGA_SRC0A              9
   1225 #       define R300_FPI2_ARGA_SRC1A              10
   1226 #       define R300_FPI2_ARGA_SRC2A              11
   1227 #       define R300_FPI2_ARGA_SRC1A_LRP          15
   1228 #       define R300_FPI2_ARGA_ZERO               16
   1229 #       define R300_FPI2_ARGA_ONE                17
   1230 	/* GUESS */
   1231 #       define R300_FPI2_ARGA_HALF               18
   1232 #       define R300_FPI2_ARG0A_SHIFT             0
   1233 #       define R300_FPI2_ARG0A_MASK              (31 << 0)
   1234 #       define R300_FPI2_ARG0A_NEG               (1 << 5)
   1235 	/* GUESS */
   1236 #	define R300_FPI2_ARG0A_ABS		 (1 << 6)
   1237 #       define R300_FPI2_ARG1A_SHIFT             7
   1238 #       define R300_FPI2_ARG1A_MASK              (31 << 7)
   1239 #       define R300_FPI2_ARG1A_NEG               (1 << 12)
   1240 	/* GUESS */
   1241 #	define R300_FPI2_ARG1A_ABS		 (1 << 13)
   1242 #       define R300_FPI2_ARG2A_SHIFT             14
   1243 #       define R300_FPI2_ARG2A_MASK              (31 << 14)
   1244 #       define R300_FPI2_ARG2A_NEG               (1 << 19)
   1245 	/* GUESS */
   1246 #	define R300_FPI2_ARG2A_ABS		 (1 << 20)
   1247 #       define R300_FPI2_SPECIAL_LRP             (1 << 21)
   1248 #       define R300_FPI2_OUTA_MAD                (0 << 23)
   1249 #       define R300_FPI2_OUTA_DP4                (1 << 23)
   1250 #       define R300_FPI2_OUTA_MIN                (2 << 23)
   1251 #       define R300_FPI2_OUTA_MAX                (3 << 23)
   1252 #       define R300_FPI2_OUTA_CMP                (6 << 23)
   1253 #       define R300_FPI2_OUTA_FRC                (7 << 23)
   1254 #       define R300_FPI2_OUTA_EX2                (8 << 23)
   1255 #       define R300_FPI2_OUTA_LG2                (9 << 23)
   1256 #       define R300_FPI2_OUTA_RCP                (10 << 23)
   1257 #       define R300_FPI2_OUTA_RSQ                (11 << 23)
   1258 #       define R300_FPI2_OUTA_SAT                (1 << 30)
   1259 #       define R300_FPI2_UNKNOWN_31              (1 << 31)
   1260 /* END: Fragment program instruction set */
   1261 
   1262 /* Fog state and color */
   1263 #define R300_RE_FOG_STATE                   0x4BC0
   1264 #       define R300_FOG_ENABLE                   (1 << 0)
   1265 #	define R300_FOG_MODE_LINEAR              (0 << 1)
   1266 #	define R300_FOG_MODE_EXP                 (1 << 1)
   1267 #	define R300_FOG_MODE_EXP2                (2 << 1)
   1268 #	define R300_FOG_MODE_MASK                (3 << 1)
   1269 #define R300_FOG_COLOR_R                    0x4BC8
   1270 #define R300_FOG_COLOR_G                    0x4BCC
   1271 #define R300_FOG_COLOR_B                    0x4BD0
   1272 
   1273 #define R300_PP_ALPHA_TEST                  0x4BD4
   1274 #       define R300_REF_ALPHA_MASK               0x000000ff
   1275 #       define R300_ALPHA_TEST_FAIL              (0 << 8)
   1276 #       define R300_ALPHA_TEST_LESS              (1 << 8)
   1277 #       define R300_ALPHA_TEST_LEQUAL            (3 << 8)
   1278 #       define R300_ALPHA_TEST_EQUAL             (2 << 8)
   1279 #       define R300_ALPHA_TEST_GEQUAL            (6 << 8)
   1280 #       define R300_ALPHA_TEST_GREATER           (4 << 8)
   1281 #       define R300_ALPHA_TEST_NEQUAL            (5 << 8)
   1282 #       define R300_ALPHA_TEST_PASS              (7 << 8)
   1283 #       define R300_ALPHA_TEST_OP_MASK           (7 << 8)
   1284 #       define R300_ALPHA_TEST_ENABLE            (1 << 11)
   1285 
   1286 /* gap */
   1287 
   1288 /* Fragment program parameters in 7.16 floating point */
   1289 #define R300_PFS_PARAM_0_X                  0x4C00
   1290 #define R300_PFS_PARAM_0_Y                  0x4C04
   1291 #define R300_PFS_PARAM_0_Z                  0x4C08
   1292 #define R300_PFS_PARAM_0_W                  0x4C0C
   1293 /* GUESS: PARAM_31 is last, based on native limits reported by fglrx */
   1294 #define R300_PFS_PARAM_31_X                 0x4DF0
   1295 #define R300_PFS_PARAM_31_Y                 0x4DF4
   1296 #define R300_PFS_PARAM_31_Z                 0x4DF8
   1297 #define R300_PFS_PARAM_31_W                 0x4DFC
   1298 
   1299 /* Notes:
   1300  * - AFAIK fglrx always sets BLEND_UNKNOWN when blending is used in
   1301  *   the application
   1302  * - AFAIK fglrx always sets BLEND_NO_SEPARATE when CBLEND and ABLEND
   1303  *    are set to the same
   1304  *   function (both registers are always set up completely in any case)
   1305  * - Most blend flags are simply copied from R200 and not tested yet
   1306  */
   1307 #define R300_RB3D_CBLEND                    0x4E04
   1308 #define R300_RB3D_ABLEND                    0x4E08
   1309 /* the following only appear in CBLEND */
   1310 #       define R300_BLEND_ENABLE                     (1 << 0)
   1311 #       define R300_BLEND_UNKNOWN                    (3 << 1)
   1312 #       define R300_BLEND_NO_SEPARATE                (1 << 3)
   1313 /* the following are shared between CBLEND and ABLEND */
   1314 #       define R300_FCN_MASK                         (3  << 12)
   1315 #       define R300_COMB_FCN_ADD_CLAMP               (0  << 12)
   1316 #       define R300_COMB_FCN_ADD_NOCLAMP             (1  << 12)
   1317 #       define R300_COMB_FCN_SUB_CLAMP               (2  << 12)
   1318 #       define R300_COMB_FCN_SUB_NOCLAMP             (3  << 12)
   1319 #       define R300_COMB_FCN_MIN                     (4  << 12)
   1320 #       define R300_COMB_FCN_MAX                     (5  << 12)
   1321 #       define R300_COMB_FCN_RSUB_CLAMP              (6  << 12)
   1322 #       define R300_COMB_FCN_RSUB_NOCLAMP            (7  << 12)
   1323 #       define R300_BLEND_GL_ZERO                    (32)
   1324 #       define R300_BLEND_GL_ONE                     (33)
   1325 #       define R300_BLEND_GL_SRC_COLOR               (34)
   1326 #       define R300_BLEND_GL_ONE_MINUS_SRC_COLOR     (35)
   1327 #       define R300_BLEND_GL_DST_COLOR               (36)
   1328 #       define R300_BLEND_GL_ONE_MINUS_DST_COLOR     (37)
   1329 #       define R300_BLEND_GL_SRC_ALPHA               (38)
   1330 #       define R300_BLEND_GL_ONE_MINUS_SRC_ALPHA     (39)
   1331 #       define R300_BLEND_GL_DST_ALPHA               (40)
   1332 #       define R300_BLEND_GL_ONE_MINUS_DST_ALPHA     (41)
   1333 #       define R300_BLEND_GL_SRC_ALPHA_SATURATE      (42)
   1334 #       define R300_BLEND_GL_CONST_COLOR             (43)
   1335 #       define R300_BLEND_GL_ONE_MINUS_CONST_COLOR   (44)
   1336 #       define R300_BLEND_GL_CONST_ALPHA             (45)
   1337 #       define R300_BLEND_GL_ONE_MINUS_CONST_ALPHA   (46)
   1338 #       define R300_BLEND_MASK                       (63)
   1339 #       define R300_SRC_BLEND_SHIFT                  (16)
   1340 #       define R300_DST_BLEND_SHIFT                  (24)
   1341 #define R300_RB3D_BLEND_COLOR               0x4E10
   1342 #define R300_RB3D_COLORMASK                 0x4E0C
   1343 #       define R300_COLORMASK0_B                 (1<<0)
   1344 #       define R300_COLORMASK0_G                 (1<<1)
   1345 #       define R300_COLORMASK0_R                 (1<<2)
   1346 #       define R300_COLORMASK0_A                 (1<<3)
   1347 
   1348 /* gap */
   1349 
   1350 #define R300_RB3D_COLOROFFSET0              0x4E28
   1351 #       define R300_COLOROFFSET_MASK             0xFFFFFFF0 /* GUESS */
   1352 #define R300_RB3D_COLOROFFSET1              0x4E2C /* GUESS */
   1353 #define R300_RB3D_COLOROFFSET2              0x4E30 /* GUESS */
   1354 #define R300_RB3D_COLOROFFSET3              0x4E34 /* GUESS */
   1355 
   1356 /* gap */
   1357 
   1358 /* Bit 16: Larger tiles
   1359  * Bit 17: 4x2 tiles
   1360  * Bit 18: Extremely weird tile like, but some pixels duplicated?
   1361  */
   1362 #define R300_RB3D_COLORPITCH0               0x4E38
   1363 #       define R300_COLORPITCH_MASK              0x00001FF8 /* GUESS */
   1364 #       define R300_COLOR_TILE_ENABLE            (1 << 16) /* GUESS */
   1365 #       define R300_COLOR_MICROTILE_ENABLE       (1 << 17) /* GUESS */
   1366 #       define R300_COLOR_MICROTILE_SQUARE_ENABLE (2 << 17)
   1367 #       define R300_COLOR_ENDIAN_NO_SWAP         (0 << 18) /* GUESS */
   1368 #       define R300_COLOR_ENDIAN_WORD_SWAP       (1 << 18) /* GUESS */
   1369 #       define R300_COLOR_ENDIAN_DWORD_SWAP      (2 << 18) /* GUESS */
   1370 #       define R300_COLOR_FORMAT_RGB565          (2 << 22)
   1371 #       define R300_COLOR_FORMAT_ARGB8888        (3 << 22)
   1372 #define R300_RB3D_COLORPITCH1               0x4E3C /* GUESS */
   1373 #define R300_RB3D_COLORPITCH2               0x4E40 /* GUESS */
   1374 #define R300_RB3D_COLORPITCH3               0x4E44 /* GUESS */
   1375 
   1376 #define R300_RB3D_AARESOLVE_OFFSET          0x4E80
   1377 #define R300_RB3D_AARESOLVE_PITCH           0x4E84
   1378 #define R300_RB3D_AARESOLVE_CTL             0x4E88
   1379 /* gap */
   1380 
   1381 /* Guess by Vladimir.
   1382  * Set to 0A before 3D operations, set to 02 afterwards.
   1383  */
   1384 /*#define R300_RB3D_DSTCACHE_CTLSTAT          0x4E4C*/
   1385 #       define R300_RB3D_DSTCACHE_UNKNOWN_02             0x00000002
   1386 #       define R300_RB3D_DSTCACHE_UNKNOWN_0A             0x0000000A
   1387 
   1388 /* gap */
   1389 /* There seems to be no "write only" setting, so use Z-test = ALWAYS
   1390  * for this.
   1391  * Bit (1<<8) is the "test" bit. so plain write is 6  - vd
   1392  */
   1393 #define R300_ZB_CNTL                             0x4F00
   1394 #	define R300_STENCIL_ENABLE		 (1 << 0)
   1395 #	define R300_Z_ENABLE		         (1 << 1)
   1396 #	define R300_Z_WRITE_ENABLE		 (1 << 2)
   1397 #	define R300_Z_SIGNED_COMPARE		 (1 << 3)
   1398 #	define R300_STENCIL_FRONT_BACK		 (1 << 4)
   1399 
   1400 #define R300_ZB_ZSTENCILCNTL                   0x4f04
   1401 	/* functions */
   1402 #	define R300_ZS_NEVER			0
   1403 #	define R300_ZS_LESS			1
   1404 #	define R300_ZS_LEQUAL			2
   1405 #	define R300_ZS_EQUAL			3
   1406 #	define R300_ZS_GEQUAL			4
   1407 #	define R300_ZS_GREATER			5
   1408 #	define R300_ZS_NOTEQUAL			6
   1409 #	define R300_ZS_ALWAYS			7
   1410 #       define R300_ZS_MASK                     7
   1411 	/* operations */
   1412 #	define R300_ZS_KEEP			0
   1413 #	define R300_ZS_ZERO			1
   1414 #	define R300_ZS_REPLACE			2
   1415 #	define R300_ZS_INCR			3
   1416 #	define R300_ZS_DECR			4
   1417 #	define R300_ZS_INVERT			5
   1418 #	define R300_ZS_INCR_WRAP		6
   1419 #	define R300_ZS_DECR_WRAP		7
   1420 #	define R300_Z_FUNC_SHIFT		0
   1421 	/* front and back refer to operations done for front
   1422 	   and back faces, i.e. separate stencil function support */
   1423 #	define R300_S_FRONT_FUNC_SHIFT	        3
   1424 #	define R300_S_FRONT_SFAIL_OP_SHIFT	6
   1425 #	define R300_S_FRONT_ZPASS_OP_SHIFT	9
   1426 #	define R300_S_FRONT_ZFAIL_OP_SHIFT      12
   1427 #	define R300_S_BACK_FUNC_SHIFT           15
   1428 #	define R300_S_BACK_SFAIL_OP_SHIFT       18
   1429 #	define R300_S_BACK_ZPASS_OP_SHIFT       21
   1430 #	define R300_S_BACK_ZFAIL_OP_SHIFT       24
   1431 
   1432 #define R300_ZB_STENCILREFMASK                        0x4f08
   1433 #	define R300_STENCILREF_SHIFT       0
   1434 #	define R300_STENCILREF_MASK        0x000000ff
   1435 #	define R300_STENCILMASK_SHIFT      8
   1436 #	define R300_STENCILMASK_MASK       0x0000ff00
   1437 #	define R300_STENCILWRITEMASK_SHIFT 16
   1438 #	define R300_STENCILWRITEMASK_MASK  0x00ff0000
   1439 
   1440 /* gap */
   1441 
   1442 #define R300_ZB_FORMAT                             0x4f10
   1443 #	define R300_DEPTHFORMAT_16BIT_INT_Z   (0 << 0)
   1444 #	define R300_DEPTHFORMAT_16BIT_13E3    (1 << 0)
   1445 #	define R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL   (2 << 0)
   1446 /* reserved up to (15 << 0) */
   1447 #	define R300_INVERT_13E3_LEADING_ONES  (0 << 4)
   1448 #	define R300_INVERT_13E3_LEADING_ZEROS (1 << 4)
   1449 
   1450 #define R300_ZB_ZTOP                             0x4F14
   1451 #	define R300_ZTOP_DISABLE                 (0 << 0)
   1452 #	define R300_ZTOP_ENABLE                  (1 << 0)
   1453 
   1454 /* gap */
   1455 
   1456 #define R300_ZB_ZCACHE_CTLSTAT            0x4f18
   1457 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_NO_EFFECT      (0 << 0)
   1458 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE (1 << 0)
   1459 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_NO_EFFECT       (0 << 1)
   1460 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE            (1 << 1)
   1461 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_IDLE            (0 << 31)
   1462 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_BUSY            (1 << 31)
   1463 
   1464 #define R300_ZB_BW_CNTL                     0x4f1c
   1465 #	define R300_HIZ_DISABLE                              (0 << 0)
   1466 #	define R300_HIZ_ENABLE                               (1 << 0)
   1467 #	define R300_HIZ_MIN                                  (0 << 1)
   1468 #	define R300_HIZ_MAX                                  (1 << 1)
   1469 #	define R300_FAST_FILL_DISABLE                        (0 << 2)
   1470 #	define R300_FAST_FILL_ENABLE                         (1 << 2)
   1471 #	define R300_RD_COMP_DISABLE                          (0 << 3)
   1472 #	define R300_RD_COMP_ENABLE                           (1 << 3)
   1473 #	define R300_WR_COMP_DISABLE                          (0 << 4)
   1474 #	define R300_WR_COMP_ENABLE                           (1 << 4)
   1475 #	define R300_ZB_CB_CLEAR_RMW                          (0 << 5)
   1476 #	define R300_ZB_CB_CLEAR_CACHE_LINEAR                 (1 << 5)
   1477 #	define R300_FORCE_COMPRESSED_STENCIL_VALUE_DISABLE   (0 << 6)
   1478 #	define R300_FORCE_COMPRESSED_STENCIL_VALUE_ENABLE    (1 << 6)
   1479 
   1480 #	define R500_ZEQUAL_OPTIMIZE_ENABLE                   (0 << 7)
   1481 #	define R500_ZEQUAL_OPTIMIZE_DISABLE                  (1 << 7)
   1482 #	define R500_SEQUAL_OPTIMIZE_ENABLE                   (0 << 8)
   1483 #	define R500_SEQUAL_OPTIMIZE_DISABLE                  (1 << 8)
   1484 
   1485 #	define R500_BMASK_ENABLE                             (0 << 10)
   1486 #	define R500_BMASK_DISABLE                            (1 << 10)
   1487 #	define R500_HIZ_EQUAL_REJECT_DISABLE                 (0 << 11)
   1488 #	define R500_HIZ_EQUAL_REJECT_ENABLE                  (1 << 11)
   1489 #	define R500_HIZ_FP_EXP_BITS_DISABLE                  (0 << 12)
   1490 #	define R500_HIZ_FP_EXP_BITS_1                        (1 << 12)
   1491 #	define R500_HIZ_FP_EXP_BITS_2                        (2 << 12)
   1492 #	define R500_HIZ_FP_EXP_BITS_3                        (3 << 12)
   1493 #	define R500_HIZ_FP_EXP_BITS_4                        (4 << 12)
   1494 #	define R500_HIZ_FP_EXP_BITS_5                        (5 << 12)
   1495 #	define R500_HIZ_FP_INVERT_LEADING_ONES               (0 << 15)
   1496 #	define R500_HIZ_FP_INVERT_LEADING_ZEROS              (1 << 15)
   1497 #	define R500_TILE_OVERWRITE_RECOMPRESSION_ENABLE      (0 << 16)
   1498 #	define R500_TILE_OVERWRITE_RECOMPRESSION_DISABLE     (1 << 16)
   1499 #	define R500_CONTIGUOUS_6XAA_SAMPLES_ENABLE           (0 << 17)
   1500 #	define R500_CONTIGUOUS_6XAA_SAMPLES_DISABLE          (1 << 17)
   1501 #	define R500_PEQ_PACKING_DISABLE                      (0 << 18)
   1502 #	define R500_PEQ_PACKING_ENABLE                       (1 << 18)
   1503 #	define R500_COVERED_PTR_MASKING_DISABLE              (0 << 18)
   1504 #	define R500_COVERED_PTR_MASKING_ENABLE               (1 << 18)
   1505 
   1506 
   1507 /* gap */
   1508 
   1509 /* Z Buffer Address Offset.
   1510  * Bits 31 to 5 are used for aligned Z buffer address offset for macro tiles.
   1511  */
   1512 #define R300_ZB_DEPTHOFFSET               0x4f20
   1513 
   1514 /* Z Buffer Pitch and Endian Control */
   1515 #define R300_ZB_DEPTHPITCH                0x4f24
   1516 #       define R300_DEPTHPITCH_MASK              0x00003FFC
   1517 #       define R300_DEPTHMACROTILE_DISABLE      (0 << 16)
   1518 #       define R300_DEPTHMACROTILE_ENABLE       (1 << 16)
   1519 #       define R300_DEPTHMICROTILE_LINEAR       (0 << 17)
   1520 #       define R300_DEPTHMICROTILE_TILED        (1 << 17)
   1521 #       define R300_DEPTHMICROTILE_TILED_SQUARE (2 << 17)
   1522 #       define R300_DEPTHENDIAN_NO_SWAP         (0 << 18)
   1523 #       define R300_DEPTHENDIAN_WORD_SWAP       (1 << 18)
   1524 #       define R300_DEPTHENDIAN_DWORD_SWAP      (2 << 18)
   1525 #       define R300_DEPTHENDIAN_HALF_DWORD_SWAP (3 << 18)
   1526 
   1527 /* Z Buffer Clear Value */
   1528 #define R300_ZB_DEPTHCLEARVALUE                  0x4f28
   1529 
   1530 #define R300_ZB_ZMASK_OFFSET			 0x4f30
   1531 #define R300_ZB_ZMASK_PITCH			 0x4f34
   1532 #define R300_ZB_ZMASK_WRINDEX			 0x4f38
   1533 #define R300_ZB_ZMASK_DWORD			 0x4f3c
   1534 #define R300_ZB_ZMASK_RDINDEX			 0x4f40
   1535 
   1536 /* Hierarchical Z Memory Offset */
   1537 #define R300_ZB_HIZ_OFFSET                       0x4f44
   1538 
   1539 /* Hierarchical Z Write Index */
   1540 #define R300_ZB_HIZ_WRINDEX                      0x4f48
   1541 
   1542 /* Hierarchical Z Data */
   1543 #define R300_ZB_HIZ_DWORD                        0x4f4c
   1544 
   1545 /* Hierarchical Z Read Index */
   1546 #define R300_ZB_HIZ_RDINDEX                      0x4f50
   1547 
   1548 /* Hierarchical Z Pitch */
   1549 #define R300_ZB_HIZ_PITCH                        0x4f54
   1550 
   1551 /* Z Buffer Z Pass Counter Data */
   1552 #define R300_ZB_ZPASS_DATA                       0x4f58
   1553 
   1554 /* Z Buffer Z Pass Counter Address */
   1555 #define R300_ZB_ZPASS_ADDR                       0x4f5c
   1556 
   1557 /* Depth buffer X and Y coordinate offset */
   1558 #define R300_ZB_DEPTHXY_OFFSET                   0x4f60
   1559 #	define R300_DEPTHX_OFFSET_SHIFT  1
   1560 #	define R300_DEPTHX_OFFSET_MASK   0x000007FE
   1561 #	define R300_DEPTHY_OFFSET_SHIFT  17
   1562 #	define R300_DEPTHY_OFFSET_MASK   0x07FE0000
   1563 
   1564 /* Sets the fifo sizes */
   1565 #define R500_ZB_FIFO_SIZE                        0x4fd0
   1566 #	define R500_OP_FIFO_SIZE_FULL   (0 << 0)
   1567 #	define R500_OP_FIFO_SIZE_HALF   (1 << 0)
   1568 #	define R500_OP_FIFO_SIZE_QUATER (2 << 0)
   1569 #	define R500_OP_FIFO_SIZE_EIGTHS (4 << 0)
   1570 
   1571 /* Stencil Reference Value and Mask for backfacing quads */
   1572 /* R300_ZB_STENCILREFMASK handles front face */
   1573 #define R500_ZB_STENCILREFMASK_BF                0x4fd4
   1574 #	define R500_STENCILREF_SHIFT       0
   1575 #	define R500_STENCILREF_MASK        0x000000ff
   1576 #	define R500_STENCILMASK_SHIFT      8
   1577 #	define R500_STENCILMASK_MASK       0x0000ff00
   1578 #	define R500_STENCILWRITEMASK_SHIFT 16
   1579 #	define R500_STENCILWRITEMASK_MASK  0x00ff0000
   1580 
   1581 /* BEGIN: Vertex program instruction set */
   1582 
   1583 /* Every instruction is four dwords long:
   1584  *  DWORD 0: output and opcode
   1585  *  DWORD 1: first argument
   1586  *  DWORD 2: second argument
   1587  *  DWORD 3: third argument
   1588  *
   1589  * Notes:
   1590  *  - ABS r, a is implemented as MAX r, a, -a
   1591  *  - MOV is implemented as ADD to zero
   1592  *  - XPD is implemented as MUL + MAD
   1593  *  - FLR is implemented as FRC + ADD
   1594  *  - apparently, fglrx tries to schedule instructions so that there is at
   1595  *    least one instruction between the write to a temporary and the first
   1596  *    read from said temporary; however, violations of this scheduling are
   1597  *    allowed
   1598  *  - register indices seem to be unrelated with OpenGL aliasing to
   1599  *    conventional state
   1600  *  - only one attribute and one parameter can be loaded at a time; however,
   1601  *    the same attribute/parameter can be used for more than one argument
   1602  *  - the second software argument for POW is the third hardware argument
   1603  *    (no idea why)
   1604  *  - MAD with only temporaries as input seems to use VPI_OUT_SELECT_MAD_2
   1605  *
   1606  * There is some magic surrounding LIT:
   1607  *   The single argument is replicated across all three inputs, but swizzled:
   1608  *     First argument: xyzy
   1609  *     Second argument: xyzx
   1610  *     Third argument: xyzw
   1611  *   Whenever the result is used later in the fragment program, fglrx forces
   1612  *   x and w to be 1.0 in the input selection; I don't know whether this is
   1613  *   strictly necessary
   1614  */
   1615 #define R300_VPI_OUT_OP_DOT                     (1 << 0)
   1616 #define R300_VPI_OUT_OP_MUL                     (2 << 0)
   1617 #define R300_VPI_OUT_OP_ADD                     (3 << 0)
   1618 #define R300_VPI_OUT_OP_MAD                     (4 << 0)
   1619 #define R300_VPI_OUT_OP_DST                     (5 << 0)
   1620 #define R300_VPI_OUT_OP_FRC                     (6 << 0)
   1621 #define R300_VPI_OUT_OP_MAX                     (7 << 0)
   1622 #define R300_VPI_OUT_OP_MIN                     (8 << 0)
   1623 #define R300_VPI_OUT_OP_SGE                     (9 << 0)
   1624 #define R300_VPI_OUT_OP_SLT                     (10 << 0)
   1625 	/* Used in GL_POINT_DISTANCE_ATTENUATION_ARB, vector(scalar, vector) */
   1626 #define R300_VPI_OUT_OP_UNK12                   (12 << 0)
   1627 #define R300_VPI_OUT_OP_ARL                     (13 << 0)
   1628 #define R300_VPI_OUT_OP_EXP                     (65 << 0)
   1629 #define R300_VPI_OUT_OP_LOG                     (66 << 0)
   1630 	/* Used in fog computations, scalar(scalar) */
   1631 #define R300_VPI_OUT_OP_UNK67                   (67 << 0)
   1632 #define R300_VPI_OUT_OP_LIT                     (68 << 0)
   1633 #define R300_VPI_OUT_OP_POW                     (69 << 0)
   1634 #define R300_VPI_OUT_OP_RCP                     (70 << 0)
   1635 #define R300_VPI_OUT_OP_RSQ                     (72 << 0)
   1636 	/* Used in GL_POINT_DISTANCE_ATTENUATION_ARB, scalar(scalar) */
   1637 #define R300_VPI_OUT_OP_UNK73                   (73 << 0)
   1638 #define R300_VPI_OUT_OP_EX2                     (75 << 0)
   1639 #define R300_VPI_OUT_OP_LG2                     (76 << 0)
   1640 #define R300_VPI_OUT_OP_MAD_2                   (128 << 0)
   1641 	/* all temps, vector(scalar, vector, vector) */
   1642 #define R300_VPI_OUT_OP_UNK129                  (129 << 0)
   1643 
   1644 #define R300_VPI_OUT_REG_CLASS_TEMPORARY        (0 << 8)
   1645 #define R300_VPI_OUT_REG_CLASS_ADDR             (1 << 8)
   1646 #define R300_VPI_OUT_REG_CLASS_RESULT           (2 << 8)
   1647 #define R300_VPI_OUT_REG_CLASS_MASK             (31 << 8)
   1648 
   1649 #define R300_VPI_OUT_REG_INDEX_SHIFT            13
   1650 	/* GUESS based on fglrx native limits */
   1651 #define R300_VPI_OUT_REG_INDEX_MASK             (31 << 13)
   1652 
   1653 #define R300_VPI_OUT_WRITE_X                    (1 << 20)
   1654 #define R300_VPI_OUT_WRITE_Y                    (1 << 21)
   1655 #define R300_VPI_OUT_WRITE_Z                    (1 << 22)
   1656 #define R300_VPI_OUT_WRITE_W                    (1 << 23)
   1657 
   1658 #define R300_VPI_IN_REG_CLASS_TEMPORARY         (0 << 0)
   1659 #define R300_VPI_IN_REG_CLASS_ATTRIBUTE         (1 << 0)
   1660 #define R300_VPI_IN_REG_CLASS_PARAMETER         (2 << 0)
   1661 #define R300_VPI_IN_REG_CLASS_NONE              (9 << 0)
   1662 #define R300_VPI_IN_REG_CLASS_MASK              (31 << 0)
   1663 
   1664 #define R300_VPI_IN_REG_INDEX_SHIFT             5
   1665 	/* GUESS based on fglrx native limits */
   1666 #define R300_VPI_IN_REG_INDEX_MASK              (255 << 5)
   1667 
   1668 /* The R300 can select components from the input register arbitrarily.
   1669  * Use the following constants, shifted by the component shift you
   1670  * want to select
   1671  */
   1672 #define R300_VPI_IN_SELECT_X    0
   1673 #define R300_VPI_IN_SELECT_Y    1
   1674 #define R300_VPI_IN_SELECT_Z    2
   1675 #define R300_VPI_IN_SELECT_W    3
   1676 #define R300_VPI_IN_SELECT_ZERO 4
   1677 #define R300_VPI_IN_SELECT_ONE  5
   1678 #define R300_VPI_IN_SELECT_MASK 7
   1679 
   1680 #define R300_VPI_IN_X_SHIFT                     13
   1681 #define R300_VPI_IN_Y_SHIFT                     16
   1682 #define R300_VPI_IN_Z_SHIFT                     19
   1683 #define R300_VPI_IN_W_SHIFT                     22
   1684 
   1685 #define R300_VPI_IN_NEG_X                       (1 << 25)
   1686 #define R300_VPI_IN_NEG_Y                       (1 << 26)
   1687 #define R300_VPI_IN_NEG_Z                       (1 << 27)
   1688 #define R300_VPI_IN_NEG_W                       (1 << 28)
   1689 /* END: Vertex program instruction set */
   1690 
   1691 /* BEGIN: Packet 3 commands */
   1692 
   1693 /* A primitive emission dword. */
   1694 #define R300_PRIM_TYPE_NONE                     (0 << 0)
   1695 #define R300_PRIM_TYPE_POINT                    (1 << 0)
   1696 #define R300_PRIM_TYPE_LINE                     (2 << 0)
   1697 #define R300_PRIM_TYPE_LINE_STRIP               (3 << 0)
   1698 #define R300_PRIM_TYPE_TRI_LIST                 (4 << 0)
   1699 #define R300_PRIM_TYPE_TRI_FAN                  (5 << 0)
   1700 #define R300_PRIM_TYPE_TRI_STRIP                (6 << 0)
   1701 #define R300_PRIM_TYPE_TRI_TYPE2                (7 << 0)
   1702 #define R300_PRIM_TYPE_RECT_LIST                (8 << 0)
   1703 #define R300_PRIM_TYPE_3VRT_POINT_LIST          (9 << 0)
   1704 #define R300_PRIM_TYPE_3VRT_LINE_LIST           (10 << 0)
   1705 	/* GUESS (based on r200) */
   1706 #define R300_PRIM_TYPE_POINT_SPRITES            (11 << 0)
   1707 #define R300_PRIM_TYPE_LINE_LOOP                (12 << 0)
   1708 #define R300_PRIM_TYPE_QUADS                    (13 << 0)
   1709 #define R300_PRIM_TYPE_QUAD_STRIP               (14 << 0)
   1710 #define R300_PRIM_TYPE_POLYGON                  (15 << 0)
   1711 #define R300_PRIM_TYPE_MASK                     0xF
   1712 #define R300_PRIM_WALK_IND                      (1 << 4)
   1713 #define R300_PRIM_WALK_LIST                     (2 << 4)
   1714 #define R300_PRIM_WALK_RING                     (3 << 4)
   1715 #define R300_PRIM_WALK_MASK                     (3 << 4)
   1716 	/* GUESS (based on r200) */
   1717 #define R300_PRIM_COLOR_ORDER_BGRA              (0 << 6)
   1718 #define R300_PRIM_COLOR_ORDER_RGBA              (1 << 6)
   1719 #define R300_PRIM_NUM_VERTICES_SHIFT            16
   1720 #define R300_PRIM_NUM_VERTICES_MASK             0xffff
   1721 
   1722 /* Draw a primitive from vertex data in arrays loaded via 3D_LOAD_VBPNTR.
   1723  * Two parameter dwords:
   1724  * 0. The first parameter appears to be always 0
   1725  * 1. The second parameter is a standard primitive emission dword.
   1726  */
   1727 #define R300_PACKET3_3D_DRAW_VBUF           0x00002800
   1728 
   1729 /* Specify the full set of vertex arrays as (address, stride).
   1730  * The first parameter is the number of vertex arrays specified.
   1731  * The rest of the command is a variable length list of blocks, where
   1732  * each block is three dwords long and specifies two arrays.
   1733  * The first dword of a block is split into two words, the lower significant
   1734  * word refers to the first array, the more significant word to the second
   1735  * array in the block.
   1736  * The low byte of each word contains the size of an array entry in dwords,
   1737  * the high byte contains the stride of the array.
   1738  * The second dword of a block contains the pointer to the first array,
   1739  * the third dword of a block contains the pointer to the second array.
   1740  * Note that if the total number of arrays is odd, the third dword of
   1741  * the last block is omitted.
   1742  */
   1743 #define R300_PACKET3_3D_LOAD_VBPNTR         0x00002F00
   1744 
   1745 #define R300_PACKET3_INDX_BUFFER            0x00003300
   1746 #    define R300_EB_UNK1_SHIFT                      24
   1747 #    define R300_EB_UNK1                    (0x80<<24)
   1748 #    define R300_EB_UNK2                        0x0810
   1749 #define R300_PACKET3_3D_DRAW_VBUF_2         0x00003400
   1750 #define R300_PACKET3_3D_DRAW_INDX_2         0x00003600
   1751 
   1752 /* END: Packet 3 commands */
   1753 
   1754 
   1755 /* Color formats for 2d packets
   1756  */
   1757 #define R300_CP_COLOR_FORMAT_CI8	2
   1758 #define R300_CP_COLOR_FORMAT_ARGB1555	3
   1759 #define R300_CP_COLOR_FORMAT_RGB565	4
   1760 #define R300_CP_COLOR_FORMAT_ARGB8888	6
   1761 #define R300_CP_COLOR_FORMAT_RGB332	7
   1762 #define R300_CP_COLOR_FORMAT_RGB8	9
   1763 #define R300_CP_COLOR_FORMAT_ARGB4444	15
   1764 
   1765 /*
   1766  * CP type-3 packets
   1767  */
   1768 #define R300_CP_CMD_BITBLT_MULTI	0xC0009B00
   1769 
   1770 #define R500_VAP_INDEX_OFFSET		0x208c
   1771 
   1772 #define R500_GA_US_VECTOR_INDEX         0x4250
   1773 #define R500_GA_US_VECTOR_DATA          0x4254
   1774 
   1775 #define R500_RS_IP_0                    0x4074
   1776 #define R500_RS_INST_0                  0x4320
   1777 
   1778 #define R500_US_CONFIG                  0x4600
   1779 
   1780 #define R500_US_FC_CTRL			0x4624
   1781 #define R500_US_CODE_ADDR		0x4630
   1782 
   1783 #define R500_RB3D_COLOR_CLEAR_VALUE_AR  0x46c0
   1784 #define R500_RB3D_CONSTANT_COLOR_AR     0x4ef8
   1785 
   1786 #define R300_SU_REG_DEST                0x42c8
   1787 #define RV530_FG_ZBREG_DEST             0x4be8
   1788 #define R300_ZB_ZPASS_DATA              0x4f58
   1789 #define R300_ZB_ZPASS_ADDR              0x4f5c
   1790 
   1791 #endif /* _R300_REG_H */
   1792