Home | History | Annotate | Download | only in r300
      1 /**************************************************************************
      2 
      3 Copyright (C) 2004-2005 Nicolai Haehnle et al.
      4 
      5 Permission is hereby granted, free of charge, to any person obtaining a
      6 copy of this software and associated documentation files (the "Software"),
      7 to deal in the Software without restriction, including without limitation
      8 on the rights to use, copy, modify, merge, publish, distribute, sub
      9 license, and/or sell copies of the Software, and to permit persons to whom
     10 the Software is furnished to do so, subject to the following conditions:
     11 
     12 The above copyright notice and this permission notice (including the next
     13 paragraph) shall be included in all copies or substantial portions of the
     14 Software.
     15 
     16 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
     17 IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
     18 FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
     19 THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
     20 DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
     21 OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
     22 USE OR OTHER DEALINGS IN THE SOFTWARE.
     23 
     24 **************************************************************************/
     25 
     26 /* *INDENT-OFF* */
     27 
     28 #ifndef _R300_REG_H
     29 #define _R300_REG_H
     30 
     31 #define R300_MC_INIT_MISC_LAT_TIMER	0x180
     32 #	define R300_MC_MISC__MC_CPR_INIT_LAT_SHIFT	0
     33 #	define R300_MC_MISC__MC_VF_INIT_LAT_SHIFT	4
     34 #	define R300_MC_MISC__MC_DISP0R_INIT_LAT_SHIFT	8
     35 #	define R300_MC_MISC__MC_DISP1R_INIT_LAT_SHIFT	12
     36 #	define R300_MC_MISC__MC_FIXED_INIT_LAT_SHIFT	16
     37 #	define R300_MC_MISC__MC_E2R_INIT_LAT_SHIFT	20
     38 #	define R300_MC_MISC__MC_SAME_PAGE_PRIO_SHIFT	24
     39 #	define R300_MC_MISC__MC_GLOBW_INIT_LAT_SHIFT	28
     40 
     41 
     42 #define R300_MC_INIT_GFX_LAT_TIMER	0x154
     43 #	define R300_MC_MISC__MC_G3D0R_INIT_LAT_SHIFT	0
     44 #	define R300_MC_MISC__MC_G3D1R_INIT_LAT_SHIFT	4
     45 #	define R300_MC_MISC__MC_G3D2R_INIT_LAT_SHIFT	8
     46 #	define R300_MC_MISC__MC_G3D3R_INIT_LAT_SHIFT	12
     47 #	define R300_MC_MISC__MC_TX0R_INIT_LAT_SHIFT	16
     48 #	define R300_MC_MISC__MC_TX1R_INIT_LAT_SHIFT	20
     49 #	define R300_MC_MISC__MC_GLOBR_INIT_LAT_SHIFT	24
     50 #	define R300_MC_MISC__MC_GLOBW_FULL_LAT_SHIFT	28
     51 
     52 /*
     53  * This file contains registers and constants for the R300. They have been
     54  * found mostly by examining command buffers captured using glxtest, as well
     55  * as by extrapolating some known registers and constants from the R200.
     56  * I am fairly certain that they are correct unless stated otherwise
     57  * in comments.
     58  */
     59 
     60 #define R300_SE_VPORT_XSCALE                0x1D98
     61 #define R300_SE_VPORT_XOFFSET               0x1D9C
     62 #define R300_SE_VPORT_YSCALE                0x1DA0
     63 #define R300_SE_VPORT_YOFFSET               0x1DA4
     64 #define R300_SE_VPORT_ZSCALE                0x1DA8
     65 #define R300_SE_VPORT_ZOFFSET               0x1DAC
     66 
     67 #define R300_VAP_PORT_IDX0		    0x2040
     68 /*
     69  * Vertex Array Processing (VAP) Control
     70  */
     71 #define R300_VAP_CNTL	0x2080
     72 #       define R300_PVS_NUM_SLOTS_SHIFT                 0
     73 #       define R300_PVS_NUM_CNTLRS_SHIFT                4
     74 #       define R300_PVS_NUM_FPUS_SHIFT                  8
     75 #       define R300_VF_MAX_VTX_NUM_SHIFT                18
     76 #       define R300_PVS_NUM_SLOTS(x)                    ((x) << 0)
     77 #       define R300_PVS_NUM_CNTLRS(x)                   ((x) << 4)
     78 #       define R300_PVS_NUM_FPUS(x)                     ((x) << 8)
     79 #       define R300_PVS_VF_MAX_VTX_NUM(x)               ((x) << 18)
     80 #       define R300_GL_CLIP_SPACE_DEF                   (0 << 22)
     81 #       define R300_DX_CLIP_SPACE_DEF                   (1 << 22)
     82 #       define R500_TCL_STATE_OPTIMIZATION              (1 << 23)
     83 
     84 /* This register is written directly and also starts data section
     85  * in many 3d CP_PACKET3's
     86  */
     87 #define R300_VAP_VF_CNTL	0x2084
     88 #	define	R300_VAP_VF_CNTL__PRIM_TYPE__SHIFT              0
     89 #	define  R300_VAP_VF_CNTL__PRIM_NONE                     (0<<0)
     90 #	define  R300_VAP_VF_CNTL__PRIM_POINTS                   (1<<0)
     91 #	define  R300_VAP_VF_CNTL__PRIM_LINES                    (2<<0)
     92 #	define  R300_VAP_VF_CNTL__PRIM_LINE_STRIP               (3<<0)
     93 #	define  R300_VAP_VF_CNTL__PRIM_TRIANGLES                (4<<0)
     94 #	define  R300_VAP_VF_CNTL__PRIM_TRIANGLE_FAN             (5<<0)
     95 #	define  R300_VAP_VF_CNTL__PRIM_TRIANGLE_STRIP           (6<<0)
     96 #	define  R300_VAP_VF_CNTL__PRIM_LINE_LOOP                (12<<0)
     97 #	define  R300_VAP_VF_CNTL__PRIM_QUADS                    (13<<0)
     98 #	define  R300_VAP_VF_CNTL__PRIM_QUAD_STRIP               (14<<0)
     99 #	define  R300_VAP_VF_CNTL__PRIM_POLYGON                  (15<<0)
    100 
    101 #	define	R300_VAP_VF_CNTL__PRIM_WALK__SHIFT              4
    102 	/* State based - direct writes to registers trigger vertex
    103            generation */
    104 #	define	R300_VAP_VF_CNTL__PRIM_WALK_STATE_BASED         (0<<4)
    105 #	define	R300_VAP_VF_CNTL__PRIM_WALK_INDICES             (1<<4)
    106 #	define	R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST         (2<<4)
    107 #	define	R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_EMBEDDED     (3<<4)
    108 
    109 	/* I don't think I saw these three used.. */
    110 #	define	R300_VAP_VF_CNTL__COLOR_ORDER__SHIFT            6
    111 #	define	R300_VAP_VF_CNTL__TCL_OUTPUT_CTL_ENA__SHIFT     9
    112 #	define	R300_VAP_VF_CNTL__PROG_STREAM_ENA__SHIFT        10
    113 
    114 	/* index size - when not set the indices are assumed to be 16 bit */
    115 #	define	R300_VAP_VF_CNTL__INDEX_SIZE_32bit              (1<<11)
    116 #       define R500_VAP_VF_CNTL__USE_ALT_NUM_VERTS          (1<<14)
    117 	/* number of vertices */
    118 #	define	R300_VAP_VF_CNTL__NUM_VERTICES__SHIFT           16
    119 
    120 #define R500_VAP_INDEX_OFFSET		    0x208c
    121 
    122 #define R500_VAP_ALT_NUM_VERTICES                           0x2088
    123 
    124 #define R300_VAP_OUTPUT_VTX_FMT_0           0x2090
    125 #       define R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT     (1<<0)
    126 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT (1<<1)
    127 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT (1<<2)
    128 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT (1<<3)
    129 #       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT (1<<4)
    130 #       define R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT (1<<16)
    131 
    132 #define R300_VAP_OUTPUT_VTX_FMT_1           0x2094
    133 	/* each of the following is 3 bits wide, specifies number
    134 	   of components */
    135 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT 0
    136 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT 3
    137 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT 6
    138 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT 9
    139 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT 12
    140 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT 15
    141 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT 18
    142 #       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT 21
    143 #	define R300_VAP_OUTPUT_VTX_FMT_1__NOT_PRESENT  0
    144 #	define R300_VAP_OUTPUT_VTX_FMT_1__1_COMPONENT  1
    145 #	define R300_VAP_OUTPUT_VTX_FMT_1__2_COMPONENTS 2
    146 #	define R300_VAP_OUTPUT_VTX_FMT_1__3_COMPONENTS 3
    147 #	define R300_VAP_OUTPUT_VTX_FMT_1__4_COMPONENTS 4
    148 
    149 #define R300_VAP_VPORT_XSCALE                     0x2098
    150 #define R300_VAP_VPORT_XOFFSET                    0x209c
    151 #define R300_VAP_VPORT_YSCALE                     0x20a0
    152 #define R300_VAP_VPORT_YOFFSET                    0x20a4
    153 #define R300_VAP_VPORT_ZSCALE                     0x20a8
    154 #define R300_VAP_VPORT_ZOFFSET                    0x20ac
    155 
    156 #define R300_VAP_VTE_CNTL                         0x20b0
    157 #define R300_SE_VTE_CNTL R300_VAP_VTE_CNTL
    158 #   define R300_VPORT_X_SCALE_ENA                           (1 << 0)
    159 #   define R300_VPORT_X_OFFSET_ENA                          (1 << 1)
    160 #   define R300_VPORT_Y_SCALE_ENA                           (1 << 2)
    161 #   define R300_VPORT_Y_OFFSET_ENA                          (1 << 3)
    162 #   define R300_VPORT_Z_SCALE_ENA                           (1 << 4)
    163 #   define R300_VPORT_Z_OFFSET_ENA                          (1 << 5)
    164 #   define R300_VTX_XY_FMT                                  (1 << 8)
    165 #   define R300_VTX_Z_FMT                                   (1 << 9)
    166 #   define R300_VTX_W0_FMT                                  (1 << 10)
    167 #   define R300_SERIAL_PROC_ENA                             (1 << 11)
    168 
    169 #define R300_VAP_VTX_SIZE               0x20b4
    170 
    171 /* BEGIN: Vertex data assembly - lots of uncertainties */
    172 
    173 /* gap */
    174 
    175 /* Maximum Vertex Indx Clamp */
    176 #define R300_VAP_VF_MAX_VTX_INDX         0x2134
    177 /* Minimum Vertex Indx Clamp */
    178 #define R300_VAP_VF_MIN_VTX_INDX         0x2138
    179 
    180 /** Vertex assembler/processor control status */
    181 #define R300_VAP_CNTL_STATUS              0x2140
    182 /* No swap at all (default) */
    183 #	define R300_VC_NO_SWAP                  (0 << 0)
    184 /* 16-bit swap: 0xAABBCCDD becomes 0xBBAADDCC */
    185 #	define R300_VC_16BIT_SWAP               (1 << 0)
    186 /* 32-bit swap: 0xAABBCCDD becomes 0xDDCCBBAA */
    187 #	define R300_VC_32BIT_SWAP               (2 << 0)
    188 /* Half-dword swap: 0xAABBCCDD becomes 0xCCDDAABB */
    189 #	define R300_VC_HALF_DWORD_SWAP          (3 << 0)
    190 /* The TCL engine will not be used (as it is logically or even physically removed) */
    191 #	define R300_VAP_TCL_BYPASS		(1 << 8)
    192 /* Read only flag if TCL engine is busy. */
    193 #	define R300_VAP_PVS_BUSY                (1 << 11)
    194 /* TODO: gap for MAX_MPS */
    195 /* Read only flag if the vertex store is busy. */
    196 #	define R300_VAP_VS_BUSY                 (1 << 24)
    197 /* Read only flag if the reciprocal engine is busy. */
    198 #	define R300_VAP_RCP_BUSY                (1 << 25)
    199 /* Read only flag if the viewport transform engine is busy. */
    200 #	define R300_VAP_VTE_BUSY                (1 << 26)
    201 /* Read only flag if the memory interface unit is busy. */
    202 #	define R300_VAP_MUI_BUSY                (1 << 27)
    203 /* Read only flag if the vertex cache is busy. */
    204 #	define R300_VAP_VC_BUSY                 (1 << 28)
    205 /* Read only flag if the vertex fetcher is busy. */
    206 #	define R300_VAP_VF_BUSY                 (1 << 29)
    207 /* Read only flag if the register pipeline is busy. */
    208 #	define R300_VAP_REGPIPE_BUSY            (1 << 30)
    209 /* Read only flag if the VAP engine is busy. */
    210 #	define R300_VAP_VAP_BUSY                (1 << 31)
    211 
    212 /* gap */
    213 
    214 /* Where do we get our vertex data?
    215  *
    216  * Vertex data either comes either from immediate mode registers or from
    217  * vertex arrays.
    218  * There appears to be no mixed mode (though we can force the pitch of
    219  * vertex arrays to 0, effectively reusing the same element over and over
    220  * again).
    221  *
    222  * Immediate mode is controlled by the INPUT_CNTL registers. I am not sure
    223  * if these registers influence vertex array processing.
    224  *
    225  * Vertex arrays are controlled via the 3D_LOAD_VBPNTR packet3.
    226  *
    227  * In both cases, vertex attributes are then passed through INPUT_ROUTE.
    228  *
    229  * Beginning with INPUT_ROUTE_0_0 is a list of WORDs that route vertex data
    230  * into the vertex processor's input registers.
    231  * The first word routes the first input, the second word the second, etc.
    232  * The corresponding input is routed into the register with the given index.
    233  * The list is ended by a word with INPUT_ROUTE_END set.
    234  *
    235  * Always set COMPONENTS_4 in immediate mode.
    236  */
    237 
    238 #define R300_VAP_PROG_STREAM_CNTL_0                     0x2150
    239 #       define R300_DATA_TYPE_0_SHIFT                   0
    240 #       define R300_DATA_TYPE_FLOAT_1                   0
    241 #       define R300_DATA_TYPE_FLOAT_2                   1
    242 #       define R300_DATA_TYPE_FLOAT_3                   2
    243 #       define R300_DATA_TYPE_FLOAT_4                   3
    244 #       define R300_DATA_TYPE_BYTE                      4
    245 #       define R300_DATA_TYPE_D3DCOLOR                  5
    246 #       define R300_DATA_TYPE_SHORT_2                   6
    247 #       define R300_DATA_TYPE_SHORT_4                   7
    248 #       define R300_DATA_TYPE_VECTOR_3_TTT              8
    249 #       define R300_DATA_TYPE_VECTOR_3_EET              9
    250 #       define R300_DATA_TYPE_FLOAT_8                   10
    251 #       define R300_DATA_TYPE_FLT16_2                   11
    252 #       define R300_DATA_TYPE_FLT16_4                   12
    253 #       define R300_SKIP_DWORDS_SHIFT                   4
    254 #       define R300_DST_VEC_LOC_SHIFT                   8
    255 #       define R300_LAST_VEC                            (1 << 13)
    256 #       define R300_SIGNED                              (1 << 14)
    257 #       define R300_NORMALIZE                           (1 << 15)
    258 #       define R300_DATA_TYPE_1_SHIFT                   16
    259 #define R300_VAP_PROG_STREAM_CNTL_1                     0x2154
    260 #define R300_VAP_PROG_STREAM_CNTL_2                     0x2158
    261 #define R300_VAP_PROG_STREAM_CNTL_3                     0x215C
    262 #define R300_VAP_PROG_STREAM_CNTL_4                     0x2160
    263 #define R300_VAP_PROG_STREAM_CNTL_5                     0x2164
    264 #define R300_VAP_PROG_STREAM_CNTL_6                     0x2168
    265 #define R300_VAP_PROG_STREAM_CNTL_7                     0x216C
    266 /* gap */
    267 
    268 /* Notes:
    269  *  - always set up to produce at least two attributes:
    270  *    if vertex program uses only position, fglrx will set normal, too
    271  *  - INPUT_CNTL_0_COLOR and INPUT_CNTL_COLOR bits are always equal.
    272  */
    273 #define R300_VAP_VTX_STATE_CNTL               0x2180
    274 #       define R300_COLOR_0_ASSEMBLY_SHIFT    0
    275 #       define R300_SEL_COLOR                 0
    276 #       define R300_SEL_USER_COLOR_0          1
    277 #       define R300_SEL_USER_COLOR_1          2
    278 #       define R300_COLOR_1_ASSEMBLY_SHIFT    2
    279 #       define R300_COLOR_2_ASSEMBLY_SHIFT    4
    280 #       define R300_COLOR_3_ASSEMBLY_SHIFT    6
    281 #       define R300_COLOR_4_ASSEMBLY_SHIFT    8
    282 #       define R300_COLOR_5_ASSEMBLY_SHIFT    10
    283 #       define R300_COLOR_6_ASSEMBLY_SHIFT    12
    284 #       define R300_COLOR_7_ASSEMBLY_SHIFT    14
    285 #       define R300_UPDATE_USER_COLOR_0_ENA   (1 << 16)
    286 
    287 /*
    288  * Each bit in this field applies to the corresponding vector in the VSM
    289  * memory (i.e. Bit 0 applies to VECTOR_0 (POSITION), etc.). If the bit
    290  * is set, then the corresponding 4-Dword Vector is output into the Vertex Stream.
    291  */
    292 #define R300_VAP_VSM_VTX_ASSM               0x2184
    293 #       define R300_INPUT_CNTL_POS               0x00000001
    294 #       define R300_INPUT_CNTL_NORMAL            0x00000002
    295 #       define R300_INPUT_CNTL_COLOR             0x00000004
    296 #       define R300_INPUT_CNTL_TC0               0x00000400
    297 #       define R300_INPUT_CNTL_TC1               0x00000800
    298 #       define R300_INPUT_CNTL_TC2               0x00001000 /* GUESS */
    299 #       define R300_INPUT_CNTL_TC3               0x00002000 /* GUESS */
    300 #       define R300_INPUT_CNTL_TC4               0x00004000 /* GUESS */
    301 #       define R300_INPUT_CNTL_TC5               0x00008000 /* GUESS */
    302 #       define R300_INPUT_CNTL_TC6               0x00010000 /* GUESS */
    303 #       define R300_INPUT_CNTL_TC7               0x00020000 /* GUESS */
    304 
    305 /* Programmable Stream Control Signed Normalize Control */
    306 #define R300_VAP_PSC_SGN_NORM_CNTL                0x21dc
    307 #   define SGN_NORM_ZERO                                    0
    308 #   define SGN_NORM_ZERO_CLAMP_MINUS_ONE                    1
    309 #   define SGN_NORM_NO_ZERO                                 2
    310 #   define R300_SGN_NORM_NO_ZERO (SGN_NORM_NO_ZERO | \
    311         (SGN_NORM_NO_ZERO << 2) | (SGN_NORM_NO_ZERO << 4) | \
    312         (SGN_NORM_NO_ZERO << 6) | (SGN_NORM_NO_ZERO << 8) | \
    313         (SGN_NORM_NO_ZERO << 10) | (SGN_NORM_NO_ZERO << 12) | \
    314         (SGN_NORM_NO_ZERO << 14) | (SGN_NORM_NO_ZERO << 16) | \
    315         (SGN_NORM_NO_ZERO << 18) | (SGN_NORM_NO_ZERO << 20) | \
    316         (SGN_NORM_NO_ZERO << 22) | (SGN_NORM_NO_ZERO << 24) | \
    317         (SGN_NORM_NO_ZERO << 26) | (SGN_NORM_NO_ZERO << 28) | \
    318         (SGN_NORM_NO_ZERO << 30))
    319 
    320 /* gap */
    321 
    322 /* Words parallel to INPUT_ROUTE_0; All words that are active in INPUT_ROUTE_0
    323  * are set to a swizzling bit pattern, other words are 0.
    324  *
    325  * In immediate mode, the pattern is always set to xyzw. In vertex array
    326  * mode, the swizzling pattern is e.g. used to set zw components in texture
    327  * coordinates with only tweo components.
    328  */
    329 #define R300_VAP_PROG_STREAM_CNTL_EXT_0                 0x21e0
    330 #       define R300_SWIZZLE0_SHIFT                      0
    331 #       define R300_SWIZZLE_SELECT_X_SHIFT              0
    332 #       define R300_SWIZZLE_SELECT_Y_SHIFT              3
    333 #       define R300_SWIZZLE_SELECT_Z_SHIFT              6
    334 #       define R300_SWIZZLE_SELECT_W_SHIFT              9
    335 
    336 #       define R300_SWIZZLE_SELECT_X                    0
    337 #       define R300_SWIZZLE_SELECT_Y                    1
    338 #       define R300_SWIZZLE_SELECT_Z                    2
    339 #       define R300_SWIZZLE_SELECT_W                    3
    340 #       define R300_SWIZZLE_SELECT_FP_ZERO              4
    341 #       define R300_SWIZZLE_SELECT_FP_ONE               5
    342 /* alternate forms for r300_emit.c */
    343 #       define R300_INPUT_ROUTE_SELECT_X    0
    344 #       define R300_INPUT_ROUTE_SELECT_Y    1
    345 #       define R300_INPUT_ROUTE_SELECT_Z    2
    346 #       define R300_INPUT_ROUTE_SELECT_W    3
    347 #       define R300_INPUT_ROUTE_SELECT_ZERO 4
    348 #       define R300_INPUT_ROUTE_SELECT_ONE  5
    349 
    350 #       define R300_WRITE_ENA_SHIFT                     12
    351 #       define R300_WRITE_ENA_X                         1
    352 #       define R300_WRITE_ENA_Y                         2
    353 #       define R300_WRITE_ENA_Z                         4
    354 #       define R300_WRITE_ENA_W                         8
    355 #       define R300_SWIZZLE1_SHIFT                      16
    356 
    357 #       define R300_VAP_SWIZZLE_X001 \
    358         ((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | \
    359          (R300_SWIZZLE_SELECT_FP_ZERO << R300_SWIZZLE_SELECT_Y_SHIFT) | \
    360          (R300_SWIZZLE_SELECT_FP_ZERO << R300_SWIZZLE_SELECT_Z_SHIFT) | \
    361          (R300_SWIZZLE_SELECT_FP_ONE << R300_SWIZZLE_SELECT_W_SHIFT) | \
    362          (0xf << R300_WRITE_ENA_SHIFT))
    363 
    364 #       define R300_VAP_SWIZZLE_XY01 \
    365         ((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | \
    366          (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | \
    367          (R300_SWIZZLE_SELECT_FP_ZERO << R300_SWIZZLE_SELECT_Z_SHIFT) | \
    368          (R300_SWIZZLE_SELECT_FP_ONE << R300_SWIZZLE_SELECT_W_SHIFT) | \
    369          (0xf << R300_WRITE_ENA_SHIFT))
    370 
    371 #       define R300_VAP_SWIZZLE_XYZ1 \
    372         ((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | \
    373          (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | \
    374          (R300_SWIZZLE_SELECT_Z << R300_SWIZZLE_SELECT_Z_SHIFT) | \
    375          (R300_SWIZZLE_SELECT_FP_ONE << R300_SWIZZLE_SELECT_W_SHIFT) | \
    376          (0xf << R300_WRITE_ENA_SHIFT))
    377 
    378 #       define R300_VAP_SWIZZLE_XYZW \
    379         ((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | \
    380          (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | \
    381          (R300_SWIZZLE_SELECT_Z << R300_SWIZZLE_SELECT_Z_SHIFT) | \
    382          (R300_SWIZZLE_SELECT_W << R300_SWIZZLE_SELECT_W_SHIFT) | \
    383          (0xf << R300_WRITE_ENA_SHIFT))
    384 
    385 #define R300_VAP_PROG_STREAM_CNTL_EXT_1                 0x21e4
    386 #define R300_VAP_PROG_STREAM_CNTL_EXT_2                 0x21e8
    387 #define R300_VAP_PROG_STREAM_CNTL_EXT_3                 0x21ec
    388 #define R300_VAP_PROG_STREAM_CNTL_EXT_4                 0x21f0
    389 #define R300_VAP_PROG_STREAM_CNTL_EXT_5                 0x21f4
    390 #define R300_VAP_PROG_STREAM_CNTL_EXT_6                 0x21f8
    391 #define R300_VAP_PROG_STREAM_CNTL_EXT_7                 0x21fc
    392 
    393 /* END: Vertex data assembly */
    394 
    395 /* gap */
    396 
    397 /* BEGIN: Upload vertex program and data */
    398 
    399 /*
    400  * The programmable vertex shader unit has a memory bank of unknown size
    401  * that can be written to in 16 byte units by writing the address into
    402  * UPLOAD_ADDRESS, followed by data in UPLOAD_DATA (multiples of 4 DWORDs).
    403  *
    404  * Pointers into the memory bank are always in multiples of 16 bytes.
    405  *
    406  * The memory bank is divided into areas with fixed meaning.
    407  *
    408  * Starting at address UPLOAD_PROGRAM: Vertex program instructions.
    409  * Native limits reported by drivers from ATI suggest size 256 (i.e. 4KB),
    410  * whereas the difference between known addresses suggests size 512.
    411  *
    412  * Starting at address UPLOAD_PARAMETERS: Vertex program parameters.
    413  * Native reported limits and the VPI layout suggest size 256, whereas
    414  * difference between known addresses suggests size 512.
    415  *
    416  * At address UPLOAD_POINTSIZE is a vector (0, 0, ps, 0), where ps is the
    417  * floating point pointsize. The exact purpose of this state is uncertain,
    418  * as there is also the R300_RE_POINTSIZE register.
    419  *
    420  * Multiple vertex programs and parameter sets can be loaded at once,
    421  * which could explain the size discrepancy.
    422  */
    423 #define R300_VAP_PVS_VECTOR_INDX_REG         0x2200
    424 #       define R300_PVS_CODE_START           0
    425 #       define R300_MAX_PVS_CODE_LINES       256
    426 #       define R500_MAX_PVS_CODE_LINES       1024
    427 #       define R300_PVS_CONST_START          512
    428 #       define R500_PVS_CONST_START          1024
    429 #       define R300_MAX_PVS_CONST_VECS       256
    430 #       define R500_MAX_PVS_CONST_VECS       256
    431 #       define R300_PVS_UCP_START            1024
    432 #       define R500_PVS_UCP_START            1536
    433 #       define R300_POINT_VPORT_SCALE_OFFSET 1030
    434 #       define R500_POINT_VPORT_SCALE_OFFSET 1542
    435 #       define R300_POINT_GEN_TEX_OFFSET     1031
    436 #       define R500_POINT_GEN_TEX_OFFSET     1543
    437 
    438 /*
    439  * These are obsolete defines form r300_context.h, but they might give some
    440  * clues when investigating the addresses further...
    441  */
    442 #if 0
    443 #define VSF_DEST_PROGRAM        0x0
    444 #define VSF_DEST_MATRIX0        0x200
    445 #define VSF_DEST_MATRIX1        0x204
    446 #define VSF_DEST_MATRIX2        0x208
    447 #define VSF_DEST_VECTOR0        0x20c
    448 #define VSF_DEST_VECTOR1        0x20d
    449 #define VSF_DEST_UNKNOWN1       0x400
    450 #define VSF_DEST_UNKNOWN2       0x406
    451 #endif
    452 
    453 /* gap */
    454 
    455 #define R300_VAP_PVS_UPLOAD_DATA            0x2208
    456 
    457 /* END: Upload vertex program and data */
    458 
    459 /* gap */
    460 
    461 /* I do not know the purpose of this register. However, I do know that
    462  * it is set to 221C_CLEAR for clear operations and to 221C_NORMAL
    463  * for normal rendering.
    464  *
    465  * 2007-11-05: This register is the user clip plane control register, but there
    466  * also seems to be a rendering mode control; the NORMAL/CLEAR defines.
    467  *
    468  * See bug #9871. http://bugs.freedesktop.org/attachment.cgi?id=10672&action=view
    469  */
    470 #define R500_VAP_TEX_TO_COLOR_CNTL		0x2218
    471 
    472 #define R300_VAP_CLIP_CNTL                       0x221C
    473 #       define R300_VAP_UCP_ENABLE_0             (1 << 0)
    474 #       define R300_VAP_UCP_ENABLE_1             (1 << 1)
    475 #       define R300_VAP_UCP_ENABLE_2             (1 << 2)
    476 #       define R300_VAP_UCP_ENABLE_3             (1 << 3)
    477 #       define R300_VAP_UCP_ENABLE_4             (1 << 4)
    478 #       define R300_VAP_UCP_ENABLE_5             (1 << 5)
    479 #       define R300_PS_UCP_MODE_DIST_COP         (0 << 14)
    480 #       define R300_PS_UCP_MODE_RADIUS_COP       (1 << 14)
    481 #       define R300_PS_UCP_MODE_RADIUS_COP_CLIP  (2 << 14)
    482 #       define R300_PS_UCP_MODE_CLIP_AS_TRIFAN   (3 << 14)
    483 #       define R300_CLIP_DISABLE                 (1 << 16)
    484 #       define R300_UCP_CULL_ONLY_ENABLE         (1 << 17)
    485 #       define R300_BOUNDARY_EDGE_FLAG_ENABLE    (1 << 18)
    486 #       define R500_COLOR2_IS_TEXTURE            (1 << 20)
    487 #       define R500_COLOR3_IS_TEXTURE            (1 << 21)
    488 
    489 /* These seem to be per-pixel and per-vertex X and Y clipping planes. The first
    490  * plane is per-pixel and the second plane is per-vertex.
    491  *
    492  * This was determined by experimentation alone but I believe it is correct.
    493  *
    494  * These registers are called X_QUAD0_1_FL to X_QUAD0_4_FL by glxtest.
    495  */
    496 #define R300_VAP_GB_VERT_CLIP_ADJ                   0x2220
    497 #define R300_VAP_GB_VERT_DISC_ADJ                   0x2224
    498 #define R300_VAP_GB_HORZ_CLIP_ADJ                   0x2228
    499 #define R300_VAP_GB_HORZ_DISC_ADJ                   0x222c
    500 
    501 #define R300_VAP_PVS_FLOW_CNTL_ADDRS_0      0x2230
    502 #define R300_PVS_FC_ACT_ADRS(x)             ((x) << 0)
    503 #define R300_PVS_FC_LOOP_CNT_JMP_INST(x)    ((x) << 8)
    504 #define R300_PVS_FC_LAST_INST(x)            ((x) << 16)
    505 #define R300_PVS_FC_RTN_INST(x)             ((x) << 24)
    506 
    507 /* gap */
    508 
    509 /* Sometimes, END_OF_PKT and 0x2284=0 are the only commands sent between
    510  * rendering commands and overwriting vertex program parameters.
    511  * Therefore, I suspect writing zero to 0x2284 synchronizes the engine and
    512  * avoids bugs caused by still running shaders reading bad data from memory.
    513  */
    514 #define R300_VAP_PVS_STATE_FLUSH_REG        0x2284
    515 
    516 /* This register is used to define the number of core clocks to wait for a
    517  * vertex to be received by the VAP input controller (while the primitive
    518  * path is backed up) before forcing any accumulated vertices to be submitted
    519  * to the vertex processing path.
    520  */
    521 #define VAP_PVS_VTX_TIMEOUT_REG             0x2288
    522 #       define R300_2288_R300                    0x00750000 /* -- nh */
    523 #       define R300_2288_RV350                   0x0000FFFF /* -- Vladimir */
    524 
    525 #define R300_VAP_PVS_FLOW_CNTL_LOOP_INDEX_0 0x2290
    526 #define R300_PVS_FC_LOOP_INIT_VAL(x)        ((x) << 0)
    527 #define R300_PVS_FC_LOOP_STEP_VAL(x)        ((x) << 8)
    528 
    529 /* gap */
    530 
    531 /* Addresses are relative to the vertex program instruction area of the
    532  * memory bank. PROGRAM_END points to the last instruction of the active
    533  * program
    534  *
    535  * The meaning of the two UNKNOWN fields is obviously not known. However,
    536  * experiments so far have shown that both *must* point to an instruction
    537  * inside the vertex program, otherwise the GPU locks up.
    538  *
    539  * fglrx usually sets CNTL_3_UNKNOWN to the end of the program and
    540  * R300_PVS_CNTL_1_POS_END_SHIFT points to instruction where last write to
    541  * position takes place.
    542  *
    543  * Most likely this is used to ignore rest of the program in cases
    544  * where group of verts arent visible. For some reason this "section"
    545  * is sometimes accepted other instruction that have no relationship with
    546  * position calculations.
    547  */
    548 #define R300_VAP_PVS_CODE_CNTL_0            0x22D0
    549 #       define R300_PVS_FIRST_INST_SHIFT         0
    550 #       define R300_PVS_XYZW_VALID_INST_SHIFT    10
    551 #       define R300_PVS_LAST_INST_SHIFT          20
    552 #       define R300_PVS_FIRST_INST(x)            ((x) << 0)
    553 #       define R300_PVS_XYZW_VALID_INST(x)       ((x) << 10)
    554 #       define R300_PVS_LAST_INST(x)             ((x) << 20)
    555 /* Addresses are relative to the vertex program parameters area. */
    556 #define R300_VAP_PVS_CONST_CNTL             0x22D4
    557 #       define R300_PVS_CONST_BASE_OFFSET_SHIFT  0
    558 #       define R300_PVS_CONST_BASE_OFFSET(x)     (x)
    559 #       define R300_PVS_MAX_CONST_ADDR_SHIFT     16
    560 #       define R300_PVS_MAX_CONST_ADDR(x)        ((x) << 16)
    561 #define R300_VAP_PVS_CODE_CNTL_1	    0x22D8
    562 #       define R300_PVS_LAST_VTX_SRC_INST_SHIFT  0
    563 #define R300_VAP_PVS_FLOW_CNTL_OPC          0x22DC
    564 #define R300_VAP_PVS_FC_OPC_JUMP(x)         (1 << (2 * (x)))
    565 #define R300_VAP_PVS_FC_OPC_LOOP(x)         (2 << (2 * (x)))
    566 #define R300_VAP_PVS_FC_OPC_JSR(x)          (3 << (2 * (x)))
    567 
    568 /* The entire range from 0x2300 to 0x2AC inclusive seems to be used for
    569  * immediate vertices
    570  */
    571 #define R300_VAP_VTX_COLOR_R                0x2464
    572 #define R300_VAP_VTX_COLOR_G                0x2468
    573 #define R300_VAP_VTX_COLOR_B                0x246C
    574 #define R300_VAP_VTX_POS_0_X_1              0x2490 /* used for glVertex2*() */
    575 #define R300_VAP_VTX_POS_0_Y_1              0x2494
    576 #define R300_VAP_VTX_COLOR_PKD              0x249C /* RGBA */
    577 #define R300_VAP_VTX_POS_0_X_2              0x24A0 /* used for glVertex3*() */
    578 #define R300_VAP_VTX_POS_0_Y_2              0x24A4
    579 #define R300_VAP_VTX_POS_0_Z_2              0x24A8
    580 /* write 0 to indicate end of packet? */
    581 #define R300_VAP_VTX_END_OF_PKT             0x24AC
    582 
    583 #define R500_VAP_PVS_FLOW_CNTL_ADDRS_LW_0   0x2500
    584 #define R500_PVS_FC_ACT_ADRS(x)             ((x) << 0)
    585 #define R500_PVS_FC_LOOP_CNT_JMP_INST(x)    ((x) << 16)
    586 
    587 #define R500_VAP_PVS_FLOW_CNTL_ADDRS_UW_0   0x2504
    588 #define R500_PVS_FC_LAST_INST(x)            ((x) << 0)
    589 #define R500_PVS_FC_RTN_INST(x)             ((x) << 16)
    590 
    591 /* gap */
    592 
    593 /* These are values from r300_reg/r300_reg.h - they are known to be correct
    594  * and are here so we can use one register file instead of several
    595  * - Vladimir
    596  */
    597 #define R300_GB_VAP_RASTER_VTX_FMT_0	0x4000
    598 #	define R300_GB_VAP_RASTER_VTX_FMT_0__POS_PRESENT	(1<<0)
    599 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_0_PRESENT	(1<<1)
    600 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_1_PRESENT	(1<<2)
    601 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_2_PRESENT	(1<<3)
    602 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_3_PRESENT	(1<<4)
    603 #	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_SPACE	(0xf<<5)
    604 #	define R300_GB_VAP_RASTER_VTX_FMT_0__PT_SIZE_PRESENT	(0x1<<16)
    605 
    606 #define R300_GB_VAP_RASTER_VTX_FMT_1	0x4004
    607 	/* each of the following is 3 bits wide, specifies number
    608 	   of components */
    609 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT	0
    610 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT	3
    611 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT	6
    612 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT	9
    613 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT	12
    614 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT	15
    615 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT	18
    616 #	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT	21
    617 
    618 /* UNK30 seems to enables point to quad transformation on textures
    619  * (or something closely related to that).
    620  * This bit is rather fatal at the time being due to lackings at pixel
    621  * shader side
    622  * Specifies top of Raster pipe specific enable controls.
    623  */
    624 #define R300_GB_ENABLE	0x4008
    625 #	define R300_GB_POINT_STUFF_DISABLE     (0 << 0)
    626 #	define R300_GB_POINT_STUFF_ENABLE      (1 << 0) /* Specifies if points will have stuffed texture coordinates. */
    627 #	define R300_GB_LINE_STUFF_DISABLE      (0 << 1)
    628 #	define R300_GB_LINE_STUFF_ENABLE       (1 << 1) /* Specifies if lines will have stuffed texture coordinates. */
    629 #	define R300_GB_TRIANGLE_STUFF_DISABLE  (0 << 2)
    630 #	define R300_GB_TRIANGLE_STUFF_ENABLE   (1 << 2) /* Specifies if triangles will have stuffed texture coordinates. */
    631 #	define R300_GB_STENCIL_AUTO_DISABLE    (0 << 4)
    632 #	define R300_GB_STENCIL_AUTO_ENABLE     (1 << 4) /* Enable stencil auto inc/dec based on triangle cw/ccw, force into dzy low bit. */
    633 #	define R300_GB_STENCIL_AUTO_FORCE      (2 << 4) /* Force 0 into dzy low bit. */
    634 
    635 	/* each of the following is 2 bits wide */
    636 #define R300_GB_TEX_REPLICATE	0 /* Replicate VAP source texture coordinates (S,T,[R,Q]). */
    637 #define R300_GB_TEX_ST		1 /* Stuff with source texture coordinates (S,T). */
    638 #define R300_GB_TEX_STR		2 /* Stuff with source texture coordinates (S,T,R). */
    639 #	define R300_GB_TEX0_SOURCE_SHIFT	16
    640 #	define R300_GB_TEX1_SOURCE_SHIFT	18
    641 #	define R300_GB_TEX2_SOURCE_SHIFT	20
    642 #	define R300_GB_TEX3_SOURCE_SHIFT	22
    643 #	define R300_GB_TEX4_SOURCE_SHIFT	24
    644 #	define R300_GB_TEX5_SOURCE_SHIFT	26
    645 #	define R300_GB_TEX6_SOURCE_SHIFT	28
    646 #	define R300_GB_TEX7_SOURCE_SHIFT	30
    647 
    648 /* MSPOS - positions for multisample antialiasing (?) */
    649 #define R300_GB_MSPOS0                           0x4010
    650 	/* shifts - each of the fields is 4 bits */
    651 #	define R300_GB_MSPOS0__MS_X0_SHIFT	0
    652 #	define R300_GB_MSPOS0__MS_Y0_SHIFT	4
    653 #	define R300_GB_MSPOS0__MS_X1_SHIFT	8
    654 #	define R300_GB_MSPOS0__MS_Y1_SHIFT	12
    655 #	define R300_GB_MSPOS0__MS_X2_SHIFT	16
    656 #	define R300_GB_MSPOS0__MS_Y2_SHIFT	20
    657 #	define R300_GB_MSPOS0__MSBD0_Y		24
    658 #	define R300_GB_MSPOS0__MSBD0_X		28
    659 
    660 #define R300_GB_MSPOS1                           0x4014
    661 #	define R300_GB_MSPOS1__MS_X3_SHIFT	0
    662 #	define R300_GB_MSPOS1__MS_Y3_SHIFT	4
    663 #	define R300_GB_MSPOS1__MS_X4_SHIFT	8
    664 #	define R300_GB_MSPOS1__MS_Y4_SHIFT	12
    665 #	define R300_GB_MSPOS1__MS_X5_SHIFT	16
    666 #	define R300_GB_MSPOS1__MS_Y5_SHIFT	20
    667 #	define R300_GB_MSPOS1__MSBD1		24
    668 
    669 /* Specifies the graphics pipeline configuration for rasterization. */
    670 #define R300_GB_TILE_CONFIG                      0x4018
    671 #	define R300_GB_TILE_DISABLE             (0 << 0)
    672 #	define R300_GB_TILE_ENABLE              (1 << 0)
    673 #	define R300_GB_TILE_PIPE_COUNT_RV300	(0 << 1) /* RV350 (1 pipe, 1 ctx) */
    674 #	define R300_GB_TILE_PIPE_COUNT_R300	(3 << 1) /* R300 (2 pipes, 1 ctx) */
    675 #	define R300_GB_TILE_PIPE_COUNT_R420_3P  (6 << 1) /* R420-3P (3 pipes, 1 ctx) */
    676 #	define R300_GB_TILE_PIPE_COUNT_R420	(7 << 1) /* R420 (4 pipes, 1 ctx) */
    677 #	define R300_GB_TILE_SIZE_8		(0 << 4)
    678 #	define R300_GB_TILE_SIZE_16		(1 << 4)
    679 #	define R300_GB_TILE_SIZE_32		(2 << 4)
    680 #	define R300_GB_SUPER_SIZE_1		(0 << 6)
    681 #	define R300_GB_SUPER_SIZE_2		(1 << 6)
    682 #	define R300_GB_SUPER_SIZE_4		(2 << 6)
    683 #	define R300_GB_SUPER_SIZE_8		(3 << 6)
    684 #	define R300_GB_SUPER_SIZE_16		(4 << 6)
    685 #	define R300_GB_SUPER_SIZE_32		(5 << 6)
    686 #	define R300_GB_SUPER_SIZE_64		(6 << 6)
    687 #	define R300_GB_SUPER_SIZE_128		(7 << 6)
    688 #	define R300_GB_SUPER_X_SHIFT		9	/* 3 bits wide */
    689 #	define R300_GB_SUPER_Y_SHIFT		12	/* 3 bits wide */
    690 #	define R300_GB_SUPER_TILE_A		(0 << 15)
    691 #	define R300_GB_SUPER_TILE_B		(1 << 15)
    692 #	define R300_GB_SUBPIXEL_1_12		(0 << 16)
    693 #	define R300_GB_SUBPIXEL_1_16		(1 << 16)
    694 #	define R300_GB_TILE_CONFIG_QUADS_PER_RAS_4   (0 << 17)
    695 #	define R300_GB_TILE_CONFIG_QUADS_PER_RAS_8   (1 << 17)
    696 #	define R300_GB_TILE_CONFIG_QUADS_PER_RAS_16  (2 << 17)
    697 #	define R300_GB_TILE_CONFIG_QUADS_PER_RAS_32  (3 << 17)
    698 #	define R300_GB_TILE_CONFIG_BB_SCAN_INTERCEPT (0 << 19)
    699 #	define R300_GB_TILE_CONFIG_BB_SCAN_BOUND_BOX (1 << 19)
    700 #	define R300_GB_TILE_CONFIG_ALT_SCAN_EN_LR    (0 << 20)
    701 #	define R300_GB_TILE_CONFIG_ALT_SCAN_EN_LRL   (1 << 20)
    702 #	define R300_GB_TILE_CONFIG_ALT_OFFSET        (0 << 21)
    703 #	define R300_GB_TILE_CONFIG_SUBPRECISION      (0 << 22)
    704 #	define R300_GB_TILE_CONFIG_ALT_TILING_DEF    (0 << 23)
    705 #	define R300_GB_TILE_CONFIG_ALT_TILING_3_2    (1 << 23)
    706 #	define R300_GB_TILE_CONFIG_Z_EXTENDED_24_1   (0 << 24)
    707 #	define R300_GB_TILE_CONFIG_Z_EXTENDED_S25_1  (1 << 24)
    708 
    709 /* Specifies the sizes of the various FIFO`s in the sc/rs/us. This register must be the first one written */
    710 #define R300_GB_FIFO_SIZE	0x4024
    711 	/* each of the following is 2 bits wide */
    712 #define R300_GB_FIFO_SIZE_32	0
    713 #define R300_GB_FIFO_SIZE_64	1
    714 #define R300_GB_FIFO_SIZE_128	2
    715 #define R300_GB_FIFO_SIZE_256	3
    716 #	define R300_SC_IFIFO_SIZE_SHIFT	0
    717 #	define R300_SC_TZFIFO_SIZE_SHIFT	2
    718 #	define R300_SC_BFIFO_SIZE_SHIFT	4
    719 
    720 #	define R300_US_OFIFO_SIZE_SHIFT	12
    721 #	define R300_US_WFIFO_SIZE_SHIFT	14
    722 	/* the following use the same constants as above, but meaning is
    723 	   is times 2 (i.e. instead of 32 words it means 64 */
    724 #	define R300_RS_TFIFO_SIZE_SHIFT	6
    725 #	define R300_RS_CFIFO_SIZE_SHIFT	8
    726 #	define R300_US_RAM_SIZE_SHIFT		10
    727 	/* watermarks, 3 bits wide */
    728 #	define R300_RS_HIGHWATER_COL_SHIFT	16
    729 #	define R300_RS_HIGHWATER_TEX_SHIFT	19
    730 #	define R300_OFIFO_HIGHWATER_SHIFT	22	/* two bits only */
    731 #	define R300_CUBE_FIFO_HIGHWATER_COL_SHIFT	24
    732 
    733 #define R300_GB_Z_PEQ_CONFIG                          0x4028
    734 #	define R300_GB_Z_PEQ_CONFIG_Z_PEQ_SIZE_4_4    (0 << 0)
    735 #	define R300_GB_Z_PEQ_CONFIG_Z_PEQ_SIZE_8_8    (1 << 0)
    736 
    737 /* Specifies various polygon specific selects (fog, depth, perspective). */
    738 #define R300_GB_SELECT                           0x401c
    739 #	define R300_GB_FOG_SELECT_C0A		(0 << 0)
    740 #	define R300_GB_FOG_SELECT_C1A           (1 << 0)
    741 #	define R300_GB_FOG_SELECT_C2A           (2 << 0)
    742 #	define R300_GB_FOG_SELECT_C3A           (3 << 0)
    743 #	define R300_GB_FOG_SELECT_1_1_W         (4 << 0)
    744 #	define R300_GB_FOG_SELECT_Z		(5 << 0)
    745 #	define R300_GB_DEPTH_SELECT_Z		(0 << 3)
    746 #	define R300_GB_DEPTH_SELECT_1_1_W	(1 << 3)
    747 #	define R300_GB_W_SELECT_1_W		(0 << 4)
    748 #	define R300_GB_W_SELECT_1		(1 << 4)
    749 #	define R300_GB_FOG_STUFF_DISABLE        (0 << 5)
    750 #	define R300_GB_FOG_STUFF_ENABLE         (1 << 5)
    751 #	define R300_GB_FOG_STUFF_TEX_SHIFT      6
    752 #	define R300_GB_FOG_STUFF_TEX_MASK       0x000003c0
    753 #	define R300_GB_FOG_STUFF_COMP_SHIFT     10
    754 #	define R300_GB_FOG_STUFF_COMP_MASK      0x00000c00
    755 
    756 /* Specifies the graphics pipeline configuration for antialiasing. */
    757 #define R300_GB_AA_CONFIG                         0x4020
    758 #	define R300_GB_AA_CONFIG_AA_DISABLE           (0 << 0)
    759 #	define R300_GB_AA_CONFIG_AA_ENABLE            (1 << 0)
    760 #	define R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_2  (0 << 1)
    761 #	define R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_3  (1 << 1)
    762 #	define R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_4  (2 << 1)
    763 #	define R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_6  (3 << 1)
    764 
    765 /* Selects which of 4 pipes are active. */
    766 #define R300_GB_PIPE_SELECT                           0x402c
    767 #	define R300_GB_PIPE_SELECT_PIPE0_ID_SHIFT  0
    768 #	define R300_GB_PIPE_SELECT_PIPE1_ID_SHIFT  2
    769 #	define R300_GB_PIPE_SELECT_PIPE2_ID_SHIFT  4
    770 #	define R300_GB_PIPE_SELECT_PIPE3_ID_SHIFT  6
    771 #	define R300_GB_PIPE_SELECT_PIPE_MASK_SHIFT 8
    772 #	define R300_GB_PIPE_SELECT_MAX_PIPE        12
    773 #	define R300_GB_PIPE_SELECT_BAD_PIPES       14
    774 #	define R300_GB_PIPE_SELECT_CONFIG_PIPES    18
    775 
    776 
    777 /* Specifies the sizes of the various FIFO`s in the sc/rs. */
    778 #define R300_GB_FIFO_SIZE1                            0x4070
    779 /* High water mark for SC input fifo */
    780 #	define R300_GB_FIFO_SIZE1_SC_HIGHWATER_IFIFO_SHIFT 0
    781 #	define R300_GB_FIFO_SIZE1_SC_HIGHWATER_IFIFO_MASK  0x0000003f
    782 /* High water mark for SC input fifo (B) */
    783 #	define R300_GB_FIFO_SIZE1_SC_HIGHWATER_BFIFO_SHIFT 6
    784 #	define R300_GB_FIFO_SIZE1_SC_HIGHWATER_BFIFO_MASK  0x00000fc0
    785 /* High water mark for RS colors' fifo */
    786 #	define R300_GB_FIFO_SIZE1_SC_HIGHWATER_COL_SHIFT   12
    787 #	define R300_GB_FIFO_SIZE1_SC_HIGHWATER_COL_MASK    0x0003f000
    788 /* High water mark for RS textures' fifo */
    789 #	define R300_GB_FIFO_SIZE1_SC_HIGHWATER_TEX_SHIFT   18
    790 #	define R300_GB_FIFO_SIZE1_SC_HIGHWATER_TEX_MASK    0x00fc0000
    791 
    792 /* This table specifies the source location and format for up to 16 texture
    793  * addresses (i[0]:i[15]) and four colors (c[0]:c[3])
    794  */
    795 #define R500_RS_IP_0					0x4074
    796 #define R500_RS_IP_1					0x4078
    797 #define R500_RS_IP_2					0x407C
    798 #define R500_RS_IP_3					0x4080
    799 #define R500_RS_IP_4					0x4084
    800 #define R500_RS_IP_5					0x4088
    801 #define R500_RS_IP_6					0x408C
    802 #define R500_RS_IP_7					0x4090
    803 #define R500_RS_IP_8					0x4094
    804 #define R500_RS_IP_9					0x4098
    805 #define R500_RS_IP_10					0x409C
    806 #define R500_RS_IP_11					0x40A0
    807 #define R500_RS_IP_12					0x40A4
    808 #define R500_RS_IP_13					0x40A8
    809 #define R500_RS_IP_14					0x40AC
    810 #define R500_RS_IP_15					0x40B0
    811 #define R500_RS_IP_PTR_K0                               62
    812 #define R500_RS_IP_PTR_K1                               63
    813 #define R500_RS_IP_TEX_PTR_S_SHIFT 			0
    814 #define R500_RS_IP_TEX_PTR_T_SHIFT 			6
    815 #define R500_RS_IP_TEX_PTR_R_SHIFT 			12
    816 #define R500_RS_IP_TEX_PTR_Q_SHIFT 			18
    817 #define R500_RS_IP_COL_PTR_SHIFT 			24
    818 #define R500_RS_IP_COL_FMT_SHIFT 			27
    819 #       define R500_RS_SEL_S(x)                         ((x) << 0)
    820 #       define R500_RS_SEL_T(x)                         ((x) << 6)
    821 #       define R500_RS_SEL_R(x)                         ((x) << 12)
    822 #       define R500_RS_SEL_Q(x)                         ((x) << 18)
    823 #	define R500_RS_COL_PTR(x)		        ((x) << 24)
    824 #       define R500_RS_COL_FMT(x)                       ((x) << 27)
    825 /* gap */
    826 #define R500_RS_IP_OFFSET_DIS 				(0 << 31)
    827 #define R500_RS_IP_OFFSET_EN 				(1 << 31)
    828 
    829 /* gap */
    830 
    831 /* Zero to flush caches. */
    832 #define R300_TX_INVALTAGS                   0x4100
    833 #define R300_TX_FLUSH                       0x0
    834 
    835 /* The upper enable bits are guessed, based on fglrx reported limits. */
    836 #define R300_TX_ENABLE                      0x4104
    837 #       define R300_TX_ENABLE_0                  (1 << 0)
    838 #       define R300_TX_ENABLE_1                  (1 << 1)
    839 #       define R300_TX_ENABLE_2                  (1 << 2)
    840 #       define R300_TX_ENABLE_3                  (1 << 3)
    841 #       define R300_TX_ENABLE_4                  (1 << 4)
    842 #       define R300_TX_ENABLE_5                  (1 << 5)
    843 #       define R300_TX_ENABLE_6                  (1 << 6)
    844 #       define R300_TX_ENABLE_7                  (1 << 7)
    845 #       define R300_TX_ENABLE_8                  (1 << 8)
    846 #       define R300_TX_ENABLE_9                  (1 << 9)
    847 #       define R300_TX_ENABLE_10                 (1 << 10)
    848 #       define R300_TX_ENABLE_11                 (1 << 11)
    849 #       define R300_TX_ENABLE_12                 (1 << 12)
    850 #       define R300_TX_ENABLE_13                 (1 << 13)
    851 #       define R300_TX_ENABLE_14                 (1 << 14)
    852 #       define R300_TX_ENABLE_15                 (1 << 15)
    853 
    854 #define R500_TX_FILTER_4		    0x4110
    855 #	define R500_TX_WEIGHT_1_SHIFT            (0)
    856 #	define R500_TX_WEIGHT_0_SHIFT            (11)
    857 #	define R500_TX_WEIGHT_PAIR               (1<<22)
    858 #	define R500_TX_PHASE_SHIFT               (23)
    859 #	define R500_TX_DIRECTION_HORIZONTAL	 (0<<27)
    860 #	define R500_TX_DIRECTION_VERITCAL	 (1<<27)
    861 
    862 #define R500_SU_TEX_WRAP_PS3		    0x4114
    863 
    864 /* S Texture Coordinate of Vertex 0 for Point texture stuffing (LLC) */
    865 #define R300_GA_POINT_S0                              0x4200
    866 
    867 /* T Texture Coordinate of Vertex 0 for Point texture stuffing (LLC) */
    868 #define R300_GA_POINT_T0                              0x4204
    869 
    870 /* S Texture Coordinate of Vertex 2 for Point texture stuffing (URC) */
    871 #define R300_GA_POINT_S1                              0x4208
    872 
    873 /* T Texture Coordinate of Vertex 2 for Point texture stuffing (URC) */
    874 #define R300_GA_POINT_T1                              0x420c
    875 
    876 /* Specifies amount to shift integer position of vertex (screen space) before
    877  * converting to float for triangle stipple.
    878  */
    879 #define R300_GA_TRIANGLE_STIPPLE            0x4214
    880 #	define R300_GA_TRIANGLE_STIPPLE_X_SHIFT_SHIFT 0
    881 #	define R300_GA_TRIANGLE_STIPPLE_X_SHIFT_MASK  0x0000000f
    882 #	define R300_GA_TRIANGLE_STIPPLE_Y_SHIFT_SHIFT 16
    883 #	define R300_GA_TRIANGLE_STIPPLE_Y_SHIFT_MASK  0x000f0000
    884 
    885 /* The pointsize is given in multiples of 6. The pointsize can be enormous:
    886  * Clear() renders a single point that fills the entire framebuffer.
    887  * 1/2 Height of point; fixed (16.0), subpixel format (1/12 or 1/16, even if in
    888  * 8b precision).
    889  */
    890 #define R300_GA_POINT_SIZE                   0x421C
    891 #       define R300_POINTSIZE_Y_SHIFT         0
    892 #       define R300_POINTSIZE_Y_MASK          0x0000ffff
    893 #       define R300_POINTSIZE_X_SHIFT         16
    894 #       define R300_POINTSIZE_X_MASK          0xffff0000
    895 #       define R300_POINTSIZE_MAX             (R300_POINTSIZE_Y_MASK / 6)
    896 
    897 /* Red fill color */
    898 #define R500_GA_FILL_R                                0x4220
    899 
    900 /* Green fill color */
    901 #define R500_GA_FILL_G                                0x4224
    902 
    903 /* Blue fill color */
    904 #define R500_GA_FILL_B                                0x4228
    905 
    906 /* Alpha fill color */
    907 #define R500_GA_FILL_A                                0x422c
    908 
    909 
    910 /* Specifies maximum and minimum point & sprite sizes for per vertex size
    911  * specification. The lower part (15:0) is MIN and (31:16) is max.
    912  */
    913 #define R300_GA_POINT_MINMAX                0x4230
    914 #       define R300_GA_POINT_MINMAX_MIN_SHIFT          0
    915 #       define R300_GA_POINT_MINMAX_MIN_MASK           (0xFFFF << 0)
    916 #       define R300_GA_POINT_MINMAX_MAX_SHIFT          16
    917 #       define R300_GA_POINT_MINMAX_MAX_MASK           (0xFFFF << 16)
    918 
    919 /* 1/2 width of line, in subpixels (1/12 or 1/16 only, even in 8b
    920  * subprecision); (16.0) fixed format.
    921  *
    922  * The line width is given in multiples of 6.
    923  * In default mode lines are classified as vertical lines.
    924  * HO: horizontal
    925  * VE: vertical or horizontal
    926  * HO & VE: no classification
    927  */
    928 #define R300_GA_LINE_CNTL                             0x4234
    929 #       define R300_GA_LINE_CNTL_WIDTH_SHIFT       0
    930 #       define R300_GA_LINE_CNTL_WIDTH_MASK        0x0000ffff
    931 #	define R300_GA_LINE_CNTL_END_TYPE_HOR      (0 << 16)
    932 #	define R300_GA_LINE_CNTL_END_TYPE_VER      (1 << 16)
    933 #	define R300_GA_LINE_CNTL_END_TYPE_SQR      (2 << 16) /* horizontal or vertical depending upon slope */
    934 #	define R300_GA_LINE_CNTL_END_TYPE_COMP     (3 << 16) /* Computed (perpendicular to slope) */
    935 #	define R500_GA_LINE_CNTL_SORT_NO           (0 << 18)
    936 #	define R500_GA_LINE_CNTL_SORT_MINX_MINY    (1 << 18)
    937 /** TODO: looks wrong */
    938 #       define R300_LINESIZE_MAX              (R300_GA_LINE_CNTL_WIDTH_MASK / 6)
    939 /** TODO: looks wrong */
    940 #       define R300_LINE_CNT_HO               (1 << 16)
    941 /** TODO: looks wrong */
    942 #       define R300_LINE_CNT_VE               (1 << 17)
    943 
    944 /* Line Stipple configuration information. */
    945 #define R300_GA_LINE_STIPPLE_CONFIG                   0x4238
    946 #	define R300_GA_LINE_STIPPLE_CONFIG_LINE_RESET_NO     (0 << 0)
    947 #	define R300_GA_LINE_STIPPLE_CONFIG_LINE_RESET_LINE   (1 << 0)
    948 #	define R300_GA_LINE_STIPPLE_CONFIG_LINE_RESET_PACKET (2 << 0)
    949 #	define R300_GA_LINE_STIPPLE_CONFIG_STIPPLE_SCALE_SHIFT 2
    950 #	define R300_GA_LINE_STIPPLE_CONFIG_STIPPLE_SCALE_MASK  0xfffffffc
    951 
    952 /* Used to load US instructions and constants */
    953 #define R500_GA_US_VECTOR_INDEX               0x4250
    954 #	define R500_GA_US_VECTOR_INDEX_SHIFT       0
    955 #	define R500_GA_US_VECTOR_INDEX_MASK        0x000000ff
    956 #	define R500_GA_US_VECTOR_INDEX_TYPE_INSTR  (0 << 16)
    957 #	define R500_GA_US_VECTOR_INDEX_TYPE_CONST  (1 << 16)
    958 #	define R500_GA_US_VECTOR_INDEX_CLAMP_NO    (0 << 17)
    959 #	define R500_GA_US_VECTOR_INDEX_CLAMP_CONST (1 << 17)
    960 
    961 /* Data register for loading US instructions and constants */
    962 #define R500_GA_US_VECTOR_DATA                0x4254
    963 
    964 /* Specifies color properties and mappings of textures. */
    965 #define R500_GA_COLOR_CONTROL_PS3                     0x4258
    966 #	define R500_TEX0_SHADING_PS3_SOLID       (0 << 0)
    967 #	define R500_TEX0_SHADING_PS3_FLAT        (1 << 0)
    968 #	define R500_TEX0_SHADING_PS3_GOURAUD     (2 << 0)
    969 #	define R500_TEX1_SHADING_PS3_SOLID       (0 << 2)
    970 #	define R500_TEX1_SHADING_PS3_FLAT        (1 << 2)
    971 #	define R500_TEX1_SHADING_PS3_GOURAUD     (2 << 2)
    972 #	define R500_TEX2_SHADING_PS3_SOLID       (0 << 4)
    973 #	define R500_TEX2_SHADING_PS3_FLAT        (1 << 4)
    974 #	define R500_TEX2_SHADING_PS3_GOURAUD     (2 << 4)
    975 #	define R500_TEX3_SHADING_PS3_SOLID       (0 << 6)
    976 #	define R500_TEX3_SHADING_PS3_FLAT        (1 << 6)
    977 #	define R500_TEX3_SHADING_PS3_GOURAUD     (2 << 6)
    978 #	define R500_TEX4_SHADING_PS3_SOLID       (0 << 8)
    979 #	define R500_TEX4_SHADING_PS3_FLAT        (1 << 8)
    980 #	define R500_TEX4_SHADING_PS3_GOURAUD     (2 << 8)
    981 #	define R500_TEX5_SHADING_PS3_SOLID       (0 << 10)
    982 #	define R500_TEX5_SHADING_PS3_FLAT        (1 << 10)
    983 #	define R500_TEX5_SHADING_PS3_GOURAUD     (2 << 10)
    984 #	define R500_TEX6_SHADING_PS3_SOLID       (0 << 12)
    985 #	define R500_TEX6_SHADING_PS3_FLAT        (1 << 12)
    986 #	define R500_TEX6_SHADING_PS3_GOURAUD     (2 << 12)
    987 #	define R500_TEX7_SHADING_PS3_SOLID       (0 << 14)
    988 #	define R500_TEX7_SHADING_PS3_FLAT        (1 << 14)
    989 #	define R500_TEX7_SHADING_PS3_GOURAUD     (2 << 14)
    990 #	define R500_TEX8_SHADING_PS3_SOLID       (0 << 16)
    991 #	define R500_TEX8_SHADING_PS3_FLAT        (1 << 16)
    992 #	define R500_TEX8_SHADING_PS3_GOURAUD     (2 << 16)
    993 #	define R500_TEX9_SHADING_PS3_SOLID       (0 << 18)
    994 #	define R500_TEX9_SHADING_PS3_FLAT        (1 << 18)
    995 #	define R500_TEX9_SHADING_PS3_GOURAUD     (2 << 18)
    996 #	define R500_TEX10_SHADING_PS3_SOLID      (0 << 20)
    997 #	define R500_TEX10_SHADING_PS3_FLAT       (1 << 20)
    998 #	define R500_TEX10_SHADING_PS3_GOURAUD    (2 << 20)
    999 #	define R500_COLOR0_TEX_OVERRIDE_NO       (0 << 22)
   1000 #	define R500_COLOR0_TEX_OVERRIDE_TEX_0    (1 << 22)
   1001 #	define R500_COLOR0_TEX_OVERRIDE_TEX_1    (2 << 22)
   1002 #	define R500_COLOR0_TEX_OVERRIDE_TEX_2    (3 << 22)
   1003 #	define R500_COLOR0_TEX_OVERRIDE_TEX_3    (4 << 22)
   1004 #	define R500_COLOR0_TEX_OVERRIDE_TEX_4    (5 << 22)
   1005 #	define R500_COLOR0_TEX_OVERRIDE_TEX_5    (6 << 22)
   1006 #	define R500_COLOR0_TEX_OVERRIDE_TEX_6    (7 << 22)
   1007 #	define R500_COLOR0_TEX_OVERRIDE_TEX_7    (8 << 22)
   1008 #	define R500_COLOR0_TEX_OVERRIDE_TEX_8_C2 (9 << 22)
   1009 #	define R500_COLOR0_TEX_OVERRIDE_TEX_9_C3 (10 << 22)
   1010 #	define R500_COLOR1_TEX_OVERRIDE_NO       (0 << 26)
   1011 #	define R500_COLOR1_TEX_OVERRIDE_TEX_0    (1 << 26)
   1012 #	define R500_COLOR1_TEX_OVERRIDE_TEX_1    (2 << 26)
   1013 #	define R500_COLOR1_TEX_OVERRIDE_TEX_2    (3 << 26)
   1014 #	define R500_COLOR1_TEX_OVERRIDE_TEX_3    (4 << 26)
   1015 #	define R500_COLOR1_TEX_OVERRIDE_TEX_4    (5 << 26)
   1016 #	define R500_COLOR1_TEX_OVERRIDE_TEX_5    (6 << 26)
   1017 #	define R500_COLOR1_TEX_OVERRIDE_TEX_6    (7 << 26)
   1018 #	define R500_COLOR1_TEX_OVERRIDE_TEX_7    (8 << 26)
   1019 #	define R500_COLOR1_TEX_OVERRIDE_TEX_8_C2 (9 << 26)
   1020 #	define R500_COLOR1_TEX_OVERRIDE_TEX_9_C3 (10 << 26)
   1021 
   1022 /* Returns idle status of various G3D block, captured when GA_IDLE written or
   1023  * when hard or soft reset asserted.
   1024  */
   1025 #define R500_GA_IDLE                                  0x425c
   1026 #	define R500_GA_IDLE_PIPE3_Z_IDLE  (0 << 0)
   1027 #	define R500_GA_IDLE_PIPE2_Z_IDLE  (0 << 1)
   1028 #	define R500_GA_IDLE_PIPE3_CD_IDLE (0 << 2)
   1029 #	define R500_GA_IDLE_PIPE2_CD_IDLE (0 << 3)
   1030 #	define R500_GA_IDLE_PIPE3_FG_IDLE (0 << 4)
   1031 #	define R500_GA_IDLE_PIPE2_FG_IDLE (0 << 5)
   1032 #	define R500_GA_IDLE_PIPE3_US_IDLE (0 << 6)
   1033 #	define R500_GA_IDLE_PIPE2_US_IDLE (0 << 7)
   1034 #	define R500_GA_IDLE_PIPE3_SC_IDLE (0 << 8)
   1035 #	define R500_GA_IDLE_PIPE2_SC_IDLE (0 << 9)
   1036 #	define R500_GA_IDLE_PIPE3_RS_IDLE (0 << 10)
   1037 #	define R500_GA_IDLE_PIPE2_RS_IDLE (0 << 11)
   1038 #	define R500_GA_IDLE_PIPE1_Z_IDLE  (0 << 12)
   1039 #	define R500_GA_IDLE_PIPE0_Z_IDLE  (0 << 13)
   1040 #	define R500_GA_IDLE_PIPE1_CD_IDLE (0 << 14)
   1041 #	define R500_GA_IDLE_PIPE0_CD_IDLE (0 << 15)
   1042 #	define R500_GA_IDLE_PIPE1_FG_IDLE (0 << 16)
   1043 #	define R500_GA_IDLE_PIPE0_FG_IDLE (0 << 17)
   1044 #	define R500_GA_IDLE_PIPE1_US_IDLE (0 << 18)
   1045 #	define R500_GA_IDLE_PIPE0_US_IDLE (0 << 19)
   1046 #	define R500_GA_IDLE_PIPE1_SC_IDLE (0 << 20)
   1047 #	define R500_GA_IDLE_PIPE0_SC_IDLE (0 << 21)
   1048 #	define R500_GA_IDLE_PIPE1_RS_IDLE (0 << 22)
   1049 #	define R500_GA_IDLE_PIPE0_RS_IDLE (0 << 23)
   1050 #	define R500_GA_IDLE_SU_IDLE       (0 << 24)
   1051 #	define R500_GA_IDLE_GA_IDLE       (0 << 25)
   1052 #	define R500_GA_IDLE_GA_UNIT2_IDLE (0 << 26)
   1053 
   1054 /* Current value of stipple accumulator. */
   1055 #define R300_GA_LINE_STIPPLE_VALUE            0x4260
   1056 
   1057 /* S Texture Coordinate Value for Vertex 0 of Line (stuff textures -- i.e. AA) */
   1058 #define R300_GA_LINE_S0                               0x4264
   1059 /* S Texture Coordinate Value for Vertex 1 of Lines (V2 of parallelogram -- stuff textures -- i.e. AA) */
   1060 #define R300_GA_LINE_S1                               0x4268
   1061 
   1062 /* GA Input fifo high water marks */
   1063 #define R500_GA_FIFO_CNTL                             0x4270
   1064 #	define R500_GA_FIFO_CNTL_VERTEX_FIFO_MASK   0x00000007
   1065 #	define R500_GA_FIFO_CNTL_VERTEX_FIFO_SHIFT  0
   1066 #	define R500_GA_FIFO_CNTL_VERTEX_INDEX_MASK  0x00000038
   1067 #	define R500_GA_FIFO_CNTL_VERTEX_INDEX_SHIFT 3
   1068 #	define R500_GA_FIFO_CNTL_VERTEX_REG_MASK    0x00003fc0
   1069 #	define R500_GA_FIFO_CNTL_VERTEX_REG_SHIFT   6
   1070 
   1071 /* GA enhance/tweaks */
   1072 #define R300_GA_ENHANCE                               0x4274
   1073 #	define R300_GA_ENHANCE_DEADLOCK_CNTL_NO_EFFECT   (0 << 0)
   1074 #	define R300_GA_ENHANCE_DEADLOCK_CNTL_PREVENT_TCL (1 << 0) /* Prevents TCL interface from deadlocking on GA side. */
   1075 #	define R300_GA_ENHANCE_FASTSYNC_CNTL_NO_EFFECT   (0 << 1)
   1076 #	define R300_GA_ENHANCE_FASTSYNC_CNTL_ENABLE      (1 << 1) /* Enables high-performance register/primitive switching. */
   1077 #	define R500_GA_ENHANCE_REG_READWRITE_NO_EFFECT   (0 << 2) /* R520+ only */
   1078 #	define R500_GA_ENHANCE_REG_READWRITE_ENABLE      (1 << 2) /* R520+ only, Enables GA support of simultaneous register reads and writes. */
   1079 #	define R500_GA_ENHANCE_REG_NOSTALL_NO_EFFECT     (0 << 3)
   1080 #	define R500_GA_ENHANCE_REG_NOSTALL_ENABLE        (1 << 3) /* Enables GA support of no-stall reads for register read back. */
   1081 
   1082 #define R300_GA_COLOR_CONTROL                   0x4278
   1083 #	define R300_GA_COLOR_CONTROL_RGB0_SHADING_SOLID      (0 << 0)
   1084 #	define R300_GA_COLOR_CONTROL_RGB0_SHADING_FLAT       (1 << 0)
   1085 #	define R300_GA_COLOR_CONTROL_RGB0_SHADING_GOURAUD    (2 << 0)
   1086 #	define R300_GA_COLOR_CONTROL_ALPHA0_SHADING_SOLID    (0 << 2)
   1087 #	define R300_GA_COLOR_CONTROL_ALPHA0_SHADING_FLAT     (1 << 2)
   1088 #	define R300_GA_COLOR_CONTROL_ALPHA0_SHADING_GOURAUD  (2 << 2)
   1089 #	define R300_GA_COLOR_CONTROL_RGB1_SHADING_SOLID      (0 << 4)
   1090 #	define R300_GA_COLOR_CONTROL_RGB1_SHADING_FLAT       (1 << 4)
   1091 #	define R300_GA_COLOR_CONTROL_RGB1_SHADING_GOURAUD    (2 << 4)
   1092 #	define R300_GA_COLOR_CONTROL_ALPHA1_SHADING_SOLID    (0 << 6)
   1093 #	define R300_GA_COLOR_CONTROL_ALPHA1_SHADING_FLAT     (1 << 6)
   1094 #	define R300_GA_COLOR_CONTROL_ALPHA1_SHADING_GOURAUD  (2 << 6)
   1095 #	define R300_GA_COLOR_CONTROL_RGB2_SHADING_SOLID      (0 << 8)
   1096 #	define R300_GA_COLOR_CONTROL_RGB2_SHADING_FLAT       (1 << 8)
   1097 #	define R300_GA_COLOR_CONTROL_RGB2_SHADING_GOURAUD    (2 << 8)
   1098 #	define R300_GA_COLOR_CONTROL_ALPHA2_SHADING_SOLID    (0 << 10)
   1099 #	define R300_GA_COLOR_CONTROL_ALPHA2_SHADING_FLAT     (1 << 10)
   1100 #	define R300_GA_COLOR_CONTROL_ALPHA2_SHADING_GOURAUD  (2 << 10)
   1101 #	define R300_GA_COLOR_CONTROL_RGB3_SHADING_SOLID      (0 << 12)
   1102 #	define R300_GA_COLOR_CONTROL_RGB3_SHADING_FLAT       (1 << 12)
   1103 #	define R300_GA_COLOR_CONTROL_RGB3_SHADING_GOURAUD    (2 << 12)
   1104 #	define R300_GA_COLOR_CONTROL_ALPHA3_SHADING_SOLID    (0 << 14)
   1105 #	define R300_GA_COLOR_CONTROL_ALPHA3_SHADING_FLAT     (1 << 14)
   1106 #	define R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD  (2 << 14)
   1107 #	define R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_FIRST  (0 << 16)
   1108 #	define R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_SECOND (1 << 16)
   1109 #	define R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_THIRD  (2 << 16)
   1110 #	define R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST   (3 << 16)
   1111 
   1112 #       define R300_SHADE_MODEL_FLAT ( \
   1113         R300_GA_COLOR_CONTROL_RGB0_SHADING_FLAT | \
   1114         R300_GA_COLOR_CONTROL_ALPHA0_SHADING_FLAT | \
   1115         R300_GA_COLOR_CONTROL_RGB1_SHADING_FLAT | \
   1116         R300_GA_COLOR_CONTROL_ALPHA1_SHADING_FLAT | \
   1117         R300_GA_COLOR_CONTROL_RGB2_SHADING_FLAT | \
   1118         R300_GA_COLOR_CONTROL_ALPHA2_SHADING_FLAT | \
   1119         R300_GA_COLOR_CONTROL_RGB3_SHADING_FLAT | \
   1120         R300_GA_COLOR_CONTROL_ALPHA3_SHADING_FLAT )
   1121 
   1122 #       define R300_SHADE_MODEL_SMOOTH ( \
   1123         R300_GA_COLOR_CONTROL_RGB0_SHADING_GOURAUD | \
   1124         R300_GA_COLOR_CONTROL_ALPHA0_SHADING_GOURAUD | \
   1125         R300_GA_COLOR_CONTROL_RGB1_SHADING_GOURAUD | \
   1126         R300_GA_COLOR_CONTROL_ALPHA1_SHADING_GOURAUD | \
   1127         R300_GA_COLOR_CONTROL_RGB2_SHADING_GOURAUD | \
   1128         R300_GA_COLOR_CONTROL_ALPHA2_SHADING_GOURAUD | \
   1129         R300_GA_COLOR_CONTROL_RGB3_SHADING_GOURAUD | \
   1130         R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD )
   1131 
   1132 /* Specifies red & green components of fill color -- S312 format -- Backwards comp. */
   1133 #define R300_GA_SOLID_RG                         0x427c
   1134 #	define GA_SOLID_RG_COLOR_GREEN_SHIFT 0
   1135 #	define GA_SOLID_RG_COLOR_GREEN_MASK  0x0000ffff
   1136 #	define GA_SOLID_RG_COLOR_RED_SHIFT   16
   1137 #	define GA_SOLID_RG_COLOR_RED_MASK    0xffff0000
   1138 /* Specifies blue & alpha components of fill color -- S312 format -- Backwards comp. */
   1139 #define R300_GA_SOLID_BA                         0x4280
   1140 #	define GA_SOLID_BA_COLOR_ALPHA_SHIFT 0
   1141 #	define GA_SOLID_BA_COLOR_ALPHA_MASK  0x0000ffff
   1142 #	define GA_SOLID_BA_COLOR_BLUE_SHIFT  16
   1143 #	define GA_SOLID_BA_COLOR_BLUE_MASK   0xffff0000
   1144 
   1145 /* Polygon Mode
   1146  * Dangerous
   1147  */
   1148 #define R300_GA_POLY_MODE                             0x4288
   1149 #	define R300_GA_POLY_MODE_DISABLE           (0 << 0)
   1150 #	define R300_GA_POLY_MODE_DUAL              (1 << 0) /* send 2 sets of 3 polys with specified poly type */
   1151 /* reserved */
   1152 #	define R300_GA_POLY_MODE_FRONT_PTYPE_POINT (0 << 4)
   1153 #	define R300_GA_POLY_MODE_FRONT_PTYPE_LINE  (1 << 4)
   1154 #	define R300_GA_POLY_MODE_FRONT_PTYPE_TRI   (2 << 4)
   1155 /* reserved */
   1156 #	define R300_GA_POLY_MODE_BACK_PTYPE_POINT  (0 << 7)
   1157 #	define R300_GA_POLY_MODE_BACK_PTYPE_LINE   (1 << 7)
   1158 #	define R300_GA_POLY_MODE_BACK_PTYPE_TRI    (2 << 7)
   1159 /* reserved */
   1160 
   1161 /* Specifies the rouding mode for geometry & color SPFP to FP conversions. */
   1162 #define R300_GA_ROUND_MODE                            0x428c
   1163 #	define R300_GA_ROUND_MODE_GEOMETRY_ROUND_TRUNC   (0 << 0)
   1164 #	define R300_GA_ROUND_MODE_GEOMETRY_ROUND_NEAREST (1 << 0)
   1165 #	define R300_GA_ROUND_MODE_COLOR_ROUND_TRUNC      (0 << 2)
   1166 #	define R300_GA_ROUND_MODE_COLOR_ROUND_NEAREST    (1 << 2)
   1167 #	define R300_GA_ROUND_MODE_RGB_CLAMP_RGB          (0 << 4)
   1168 #	define R300_GA_ROUND_MODE_RGB_CLAMP_FP20         (1 << 4)
   1169 #	define R300_GA_ROUND_MODE_ALPHA_CLAMP_RGB        (0 << 5)
   1170 #	define R300_GA_ROUND_MODE_ALPHA_CLAMP_FP20       (1 << 5)
   1171 #	define R500_GA_ROUND_MODE_GEOMETRY_MASK_SHIFT    6
   1172 #	define R500_GA_ROUND_MODE_GEOMETRY_MASK_MASK     0x000003c0
   1173 
   1174 /* Specifies x & y offsets for vertex data after conversion to FP.
   1175  * Offsets are in S15 format (subpixels -- 1/12 or 1/16, even in 8b
   1176  * subprecision).
   1177  */
   1178 #define R300_GA_OFFSET                                0x4290
   1179 #	define R300_GA_OFFSET_X_OFFSET_SHIFT 0
   1180 #	define R300_GA_OFFSET_X_OFFSET_MASK  0x0000ffff
   1181 #	define R300_GA_OFFSET_Y_OFFSET_SHIFT 16
   1182 #	define R300_GA_OFFSET_Y_OFFSET_MASK  0xffff0000
   1183 
   1184 /* Specifies the scale to apply to fog. */
   1185 #define R300_GA_FOG_SCALE                     0x4294
   1186 /* Specifies the offset to apply to fog. */
   1187 #define R300_GA_FOG_OFFSET                    0x4298
   1188 /* Specifies number of cycles to assert reset, and also causes RB3D soft reset to assert. */
   1189 #define R300_GA_SOFT_RESET                    0x429c
   1190 
   1191 /* Not sure why there are duplicate of factor and constant values.
   1192  * My best guess so far is that there are seperate zbiases for test and write.
   1193  * Ordering might be wrong.
   1194  * Some of the tests indicate that fgl has a fallback implementation of zbias
   1195  * via pixel shaders.
   1196  */
   1197 #define R300_SU_TEX_WRAP                      0x42A0
   1198 #define R300_SU_POLY_OFFSET_FRONT_SCALE       0x42A4
   1199 #define R300_SU_POLY_OFFSET_FRONT_OFFSET      0x42A8
   1200 #define R300_SU_POLY_OFFSET_BACK_SCALE        0x42AC
   1201 #define R300_SU_POLY_OFFSET_BACK_OFFSET       0x42B0
   1202 
   1203 /* This register needs to be set to (1<<1) for RV350 to correctly
   1204  * perform depth test (see --vb-triangles in r300_demo)
   1205  * Don't know about other chips. - Vladimir
   1206  * This is set to 3 when GL_POLYGON_OFFSET_FILL is on.
   1207  * My guess is that there are two bits for each zbias primitive
   1208  * (FILL, LINE, POINT).
   1209  *  One to enable depth test and one for depth write.
   1210  * Yet this doesnt explain why depth writes work ...
   1211  */
   1212 #define R300_SU_POLY_OFFSET_ENABLE	       0x42B4
   1213 #	define R300_FRONT_ENABLE	       (1 << 0)
   1214 #	define R300_BACK_ENABLE 	       (1 << 1)
   1215 #	define R300_PARA_ENABLE 	       (1 << 2)
   1216 
   1217 #define R300_SU_CULL_MODE                      0x42B8
   1218 #       define R300_CULL_FRONT                   (1 << 0)
   1219 #       define R300_CULL_BACK                    (1 << 1)
   1220 #       define R300_FRONT_FACE_CCW               (0 << 2)
   1221 #       define R300_FRONT_FACE_CW                (1 << 2)
   1222 
   1223 /* SU Depth Scale value */
   1224 #define R300_SU_DEPTH_SCALE                 0x42c0
   1225 /* SU Depth Offset value */
   1226 #define R300_SU_DEPTH_OFFSET                0x42c4
   1227 
   1228 #define R300_SU_REG_DEST		    0x42c8
   1229 #	define R300_RASTER_PIPE_SELECT_0	(1 << 0)
   1230 #	define R300_RASTER_PIPE_SELECT_1	(1 << 1)
   1231 #	define R300_RASTER_PIPE_SELECT_2	(1 << 2)
   1232 #	define R300_RASTER_PIPE_SELECT_3	(1 << 3)
   1233 #	define R300_RASTER_PIPE_SELECT_ALL	0xf
   1234 
   1235 
   1236 /* BEGIN: Rasterization / Interpolators - many guesses */
   1237 
   1238 /*
   1239  * TC_CNT is the number of incoming texture coordinate sets (i.e. it depends
   1240  * on the vertex program, *not* the fragment program)
   1241  */
   1242 #define R300_RS_COUNT                      0x4300
   1243 #       define R300_IT_COUNT_SHIFT               0
   1244 #       define R300_IT_COUNT_MASK                0x0000007f
   1245 #       define R300_IC_COUNT_SHIFT               7
   1246 #       define R300_IC_COUNT_MASK                0x00000780
   1247 #       define R300_W_ADDR_SHIFT                 12
   1248 #       define R300_W_ADDR_MASK                  0x0003f000
   1249 #       define R300_HIRES_DIS                    (0 << 18)
   1250 #       define R300_HIRES_EN                     (1 << 18)
   1251 #       define R300_IT_COUNT(x)                  ((x) << 0)
   1252 #       define R300_IC_COUNT(x)                  ((x) << 7)
   1253 #       define R300_W_COUNT(x)                   ((x) << 12)
   1254 
   1255 #define R300_RS_INST_COUNT                       0x4304
   1256 #       define R300_RS_INST_COUNT_SHIFT          0
   1257 #       define R300_RS_INST_COUNT_MASK           0x0000000f
   1258 #       define R300_RS_TX_OFFSET_SHIFT           5
   1259 #	define R300_RS_TX_OFFSET_MASK            0x000000e0
   1260 #       define R300_RS_TX_OFFSET(x)              ((x) << 5)
   1261 
   1262 /* gap */
   1263 
   1264 /* Only used for texture coordinates.
   1265  * Use the source field to route texture coordinate input from the
   1266  * vertex program to the desired interpolator. Note that the source
   1267  * field is relative to the outputs the vertex program *actually*
   1268  * writes. If a vertex program only writes texcoord[1], this will
   1269  * be source index 0.
   1270  * Set INTERP_USED on all interpolators that produce data used by
   1271  * the fragment program. INTERP_USED looks like a swizzling mask,
   1272  * but I haven't seen it used that way.
   1273  *
   1274  * Note: The _UNKNOWN constants are always set in their respective
   1275  * register. I don't know if this is necessary.
   1276  */
   1277 #define R300_RS_IP_0				        0x4310
   1278 #define R300_RS_IP_1				        0x4314
   1279 #define R300_RS_IP_2				        0x4318
   1280 #define R300_RS_IP_3				        0x431C
   1281 #       define R300_RS_INTERP_SRC_SHIFT          2 /* TODO: check for removal */
   1282 #       define R300_RS_INTERP_SRC_MASK           (7 << 2) /* TODO: check for removal */
   1283 #	define R300_RS_TEX_PTR(x)		        (x << 0)
   1284 #	define R300_RS_COL_PTR(x)		        ((x) << 6)
   1285 #	define R300_RS_COL_FMT(x)		        ((x) << 9)
   1286 #	define R300_RS_COL_FMT_RGBA		        0
   1287 #	define R300_RS_COL_FMT_RGB0		        1
   1288 #	define R300_RS_COL_FMT_RGB1		        2
   1289 #	define R300_RS_COL_FMT_000A		        4
   1290 #	define R300_RS_COL_FMT_0000		        5
   1291 #	define R300_RS_COL_FMT_0001		        6
   1292 #	define R300_RS_COL_FMT_111A		        8
   1293 #	define R300_RS_COL_FMT_1110		        9
   1294 #	define R300_RS_COL_FMT_1111		        10
   1295 #	define R300_RS_SEL_S(x)		                ((x) << 13)
   1296 #	define R300_RS_SEL_T(x)		                ((x) << 16)
   1297 #	define R300_RS_SEL_R(x)		                ((x) << 19)
   1298 #	define R300_RS_SEL_Q(x)		                ((x) << 22)
   1299 #	define R300_RS_SEL_C0		                0
   1300 #	define R300_RS_SEL_C1		                1
   1301 #	define R300_RS_SEL_C2		                2
   1302 #	define R300_RS_SEL_C3		                3
   1303 #	define R300_RS_SEL_K0		                4
   1304 #	define R300_RS_SEL_K1		                5
   1305 
   1306 
   1307 /*  */
   1308 #define R500_RS_INST_0					0x4320
   1309 #define R500_RS_INST_1					0x4324
   1310 #define R500_RS_INST_2					0x4328
   1311 #define R500_RS_INST_3					0x432c
   1312 #define R500_RS_INST_4					0x4330
   1313 #define R500_RS_INST_5					0x4334
   1314 #define R500_RS_INST_6					0x4338
   1315 #define R500_RS_INST_7					0x433c
   1316 #define R500_RS_INST_8					0x4340
   1317 #define R500_RS_INST_9					0x4344
   1318 #define R500_RS_INST_10					0x4348
   1319 #define R500_RS_INST_11					0x434c
   1320 #define R500_RS_INST_12					0x4350
   1321 #define R500_RS_INST_13					0x4354
   1322 #define R500_RS_INST_14					0x4358
   1323 #define R500_RS_INST_15					0x435c
   1324 #define R500_RS_INST_TEX_ID_SHIFT			0
   1325 #        define R500_RS_INST_TEX_ID(x)                  ((x) << 0)
   1326 #define R500_RS_INST_TEX_CN_WRITE			(1 << 4)
   1327 #define R500_RS_INST_TEX_ADDR_SHIFT			5
   1328 #        define R500_RS_INST_TEX_ADDR(x)                ((x) << 5)
   1329 #define R500_RS_INST_COL_ID_SHIFT			12
   1330 #        define R500_RS_INST_COL_ID(x)                  ((x) << 12)
   1331 #define R500_RS_INST_COL_CN_NO_WRITE			(0 << 16)
   1332 #define R500_RS_INST_COL_CN_WRITE			(1 << 16)
   1333 #define R500_RS_INST_COL_CN_WRITE_FBUFFER		(2 << 16)
   1334 #define R500_RS_INST_COL_CN_WRITE_BACKFACE		(3 << 16)
   1335 #define R500_RS_INST_COL_ADDR_SHIFT			18
   1336 #        define R500_RS_INST_COL_ADDR(x)                ((x) << 18)
   1337 #define R500_RS_INST_TEX_ADJ				(1 << 25)
   1338 #define R500_RS_INST_W_CN				(1 << 26)
   1339 
   1340 /* These DWORDs control how vertex data is routed into fragment program
   1341  * registers, after interpolators.
   1342  */
   1343 #define R300_RS_INST_0                     0x4330
   1344 #define R300_RS_INST_1                     0x4334
   1345 #define R300_RS_INST_2                     0x4338
   1346 #define R300_RS_INST_3                     0x433C
   1347 #define R300_RS_INST_4                     0x4340
   1348 #define R300_RS_INST_5                     0x4344
   1349 #define R300_RS_INST_6                     0x4348
   1350 #define R300_RS_INST_7                     0x434C
   1351 #	define R300_RS_INST_TEX_ID(x)  		((x) << 0)
   1352 #	define R300_RS_INST_TEX_CN_WRITE 	(1 << 3)
   1353 #	define R300_RS_INST_TEX_ADDR(x)		((x) << 6)
   1354 #	define R300_RS_INST_TEX_ADDR_SHIFT 	6
   1355 #	define R300_RS_INST_COL_ID(x)		((x) << 11)
   1356 #	define R300_RS_INST_COL_CN_WRITE	(1 << 14)
   1357 #	define R300_RS_INST_COL_ADDR(x)		((x) << 17)
   1358 #	define R300_RS_INST_COL_ADDR_SHIFT	17
   1359 #	define R300_RS_INST_TEX_ADJ		(1 << 22)
   1360 #	define R300_RS_COL_BIAS_UNUSED_SHIFT    23
   1361 
   1362 /* END: Rasterization / Interpolators - many guesses */
   1363 
   1364 /* Hierarchical Z Enable */
   1365 #define R300_SC_HYPERZ                   0x43a4
   1366 #	define R300_SC_HYPERZ_DISABLE     (0 << 0)
   1367 #	define R300_SC_HYPERZ_ENABLE      (1 << 0)
   1368 #	define R300_SC_HYPERZ_MIN         (0 << 1)
   1369 #	define R300_SC_HYPERZ_MAX         (1 << 1)
   1370 #	define R300_SC_HYPERZ_ADJ_256     (0 << 2)
   1371 #	define R300_SC_HYPERZ_ADJ_128     (1 << 2)
   1372 #	define R300_SC_HYPERZ_ADJ_64      (2 << 2)
   1373 #	define R300_SC_HYPERZ_ADJ_32      (3 << 2)
   1374 #	define R300_SC_HYPERZ_ADJ_16      (4 << 2)
   1375 #	define R300_SC_HYPERZ_ADJ_8       (5 << 2)
   1376 #	define R300_SC_HYPERZ_ADJ_4       (6 << 2)
   1377 #	define R300_SC_HYPERZ_ADJ_2       (7 << 2)
   1378 #	define R300_SC_HYPERZ_HZ_Z0MIN_NO (0 << 5)
   1379 #	define R300_SC_HYPERZ_HZ_Z0MIN    (1 << 5)
   1380 #	define R300_SC_HYPERZ_HZ_Z0MAX_NO (0 << 6)
   1381 #	define R300_SC_HYPERZ_HZ_Z0MAX    (1 << 6)
   1382 
   1383 #define R300_SC_EDGERULE                 0x43a8
   1384 
   1385 /* BEGIN: Scissors and cliprects */
   1386 
   1387 /* There are four clipping rectangles. Their corner coordinates are inclusive.
   1388  * Every pixel is assigned a number from 0 and 15 by setting bits 0-3 depending
   1389  * on whether the pixel is inside cliprects 0-3, respectively. For example,
   1390  * if a pixel is inside cliprects 0 and 1, but outside 2 and 3, it is assigned
   1391  * the number 3 (binary 0011).
   1392  * Iff the bit corresponding to the pixel's number in RE_CLIPRECT_CNTL is set,
   1393  * the pixel is rasterized.
   1394  *
   1395  * In addition to this, there is a scissors rectangle. Only pixels inside the
   1396  * scissors rectangle are drawn. (coordinates are inclusive)
   1397  *
   1398  * For some reason, the top-left corner of the framebuffer is at (1440, 1440)
   1399  * for the purpose of clipping and scissors.
   1400  */
   1401 #define R300_SC_CLIPRECT_TL_0               0x43B0
   1402 #define R300_SC_CLIPRECT_BR_0               0x43B4
   1403 #define R300_SC_CLIPRECT_TL_1               0x43B8
   1404 #define R300_SC_CLIPRECT_BR_1               0x43BC
   1405 #define R300_SC_CLIPRECT_TL_2               0x43C0
   1406 #define R300_SC_CLIPRECT_BR_2               0x43C4
   1407 #define R300_SC_CLIPRECT_TL_3               0x43C8
   1408 #define R300_SC_CLIPRECT_BR_3               0x43CC
   1409 #       define R300_CLIPRECT_OFFSET              1440
   1410 #       define R300_CLIPRECT_MASK                0x1FFF
   1411 #       define R300_CLIPRECT_X_SHIFT             0
   1412 #       define R300_CLIPRECT_X_MASK              (0x1FFF << 0)
   1413 #       define R300_CLIPRECT_Y_SHIFT             13
   1414 #       define R300_CLIPRECT_Y_MASK              (0x1FFF << 13)
   1415 #define R300_SC_CLIP_RULE                   0x43D0
   1416 #       define R300_CLIP_OUT                     (1 << 0)
   1417 #       define R300_CLIP_0                       (1 << 1)
   1418 #       define R300_CLIP_1                       (1 << 2)
   1419 #       define R300_CLIP_10                      (1 << 3)
   1420 #       define R300_CLIP_2                       (1 << 4)
   1421 #       define R300_CLIP_20                      (1 << 5)
   1422 #       define R300_CLIP_21                      (1 << 6)
   1423 #       define R300_CLIP_210                     (1 << 7)
   1424 #       define R300_CLIP_3                       (1 << 8)
   1425 #       define R300_CLIP_30                      (1 << 9)
   1426 #       define R300_CLIP_31                      (1 << 10)
   1427 #       define R300_CLIP_310                     (1 << 11)
   1428 #       define R300_CLIP_32                      (1 << 12)
   1429 #       define R300_CLIP_320                     (1 << 13)
   1430 #       define R300_CLIP_321                     (1 << 14)
   1431 #       define R300_CLIP_3210                    (1 << 15)
   1432 
   1433 /* gap */
   1434 
   1435 #define R300_SC_SCISSORS_TL                 0x43E0
   1436 #define R300_SC_SCISSORS_BR                 0x43E4
   1437 #       define R300_SCISSORS_OFFSET              1440
   1438 #       define R300_SCISSORS_X_SHIFT             0
   1439 #       define R300_SCISSORS_X_MASK              (0x1FFF << 0)
   1440 #       define R300_SCISSORS_Y_SHIFT             13
   1441 #       define R300_SCISSORS_Y_MASK              (0x1FFF << 13)
   1442 
   1443 /* Screen door sample mask */
   1444 #define R300_SC_SCREENDOOR                 0x43e8
   1445 
   1446 /* END: Scissors and cliprects */
   1447 
   1448 /* BEGIN: Texture specification */
   1449 
   1450 /*
   1451  * The texture specification dwords are grouped by meaning and not by texture
   1452  * unit. This means that e.g. the offset for texture image unit N is found in
   1453  * register TX_OFFSET_0 + (4*N)
   1454  */
   1455 #define R300_TX_FILTER0_0                        0x4400
   1456 #define R300_TX_FILTER0_1                        0x4404
   1457 #define R300_TX_FILTER0_2                        0x4408
   1458 #define R300_TX_FILTER0_3                        0x440c
   1459 #define R300_TX_FILTER0_4                        0x4410
   1460 #define R300_TX_FILTER0_5                        0x4414
   1461 #define R300_TX_FILTER0_6                        0x4418
   1462 #define R300_TX_FILTER0_7                        0x441c
   1463 #define R300_TX_FILTER0_8                        0x4420
   1464 #define R300_TX_FILTER0_9                        0x4424
   1465 #define R300_TX_FILTER0_10                       0x4428
   1466 #define R300_TX_FILTER0_11                       0x442c
   1467 #define R300_TX_FILTER0_12                       0x4430
   1468 #define R300_TX_FILTER0_13                       0x4434
   1469 #define R300_TX_FILTER0_14                       0x4438
   1470 #define R300_TX_FILTER0_15                       0x443c
   1471 #       define R300_TX_REPEAT                    0
   1472 #       define R300_TX_MIRRORED                  1
   1473 #       define R300_TX_CLAMP_TO_EDGE             2
   1474 #	define R300_TX_MIRROR_ONCE_TO_EDGE       3
   1475 #       define R300_TX_CLAMP                     4
   1476 #	define R300_TX_MIRROR_ONCE               5
   1477 #       define R300_TX_CLAMP_TO_BORDER           6
   1478 #	define R300_TX_MIRROR_ONCE_TO_BORDER     7
   1479 #       define R300_TX_WRAP_S_SHIFT              0
   1480 #       define R300_TX_WRAP_S_MASK               (7 << 0)
   1481 #       define R300_TX_WRAP_T_SHIFT              3
   1482 #       define R300_TX_WRAP_T_MASK               (7 << 3)
   1483 #       define R300_TX_WRAP_R_SHIFT              6
   1484 #       define R300_TX_WRAP_R_MASK               (7 << 6)
   1485 #	define R300_TX_MAG_FILTER_4              (0 << 9)
   1486 #       define R300_TX_MAG_FILTER_NEAREST        (1 << 9)
   1487 #       define R300_TX_MAG_FILTER_LINEAR         (2 << 9)
   1488 #       define R300_TX_MAG_FILTER_ANISO          (3 << 9)
   1489 #       define R300_TX_MAG_FILTER_MASK           (3 << 9)
   1490 #       define R300_TX_MIN_FILTER_NEAREST        (1 << 11)
   1491 #       define R300_TX_MIN_FILTER_LINEAR         (2 << 11)
   1492 #	define R300_TX_MIN_FILTER_ANISO          (3 << 11)
   1493 #	define R300_TX_MIN_FILTER_MASK           (3 << 11)
   1494 #	define R300_TX_MIN_FILTER_MIP_NONE       (0 << 13)
   1495 #	define R300_TX_MIN_FILTER_MIP_NEAREST    (1 << 13)
   1496 #	define R300_TX_MIN_FILTER_MIP_LINEAR     (2 << 13)
   1497 #	define R300_TX_MIN_FILTER_MIP_MASK       (3 << 13)
   1498 #       define R300_TX_MAX_MIP_LEVEL_SHIFT       17
   1499 #       define R300_TX_MAX_MIP_LEVEL_MASK        (0xf << 17)
   1500 #	define R300_TX_MAX_ANISO_1_TO_1          (0 << 21)
   1501 #	define R300_TX_MAX_ANISO_2_TO_1          (1 << 21)
   1502 #	define R300_TX_MAX_ANISO_4_TO_1          (2 << 21)
   1503 #	define R300_TX_MAX_ANISO_8_TO_1          (3 << 21)
   1504 #	define R300_TX_MAX_ANISO_16_TO_1         (4 << 21)
   1505 #	define R300_TX_MAX_ANISO_MASK            (7 << 21)
   1506 #       define R300_TX_WRAP_S(x)                 ((x) << 0)
   1507 #       define R300_TX_WRAP_T(x)                 ((x) << 3)
   1508 #       define R300_TX_MAX_MIP_LEVEL(x)          ((x) << 17)
   1509 
   1510 #define R300_TX_FILTER1_0                      0x4440
   1511 #	define R300_CHROMA_KEY_MODE_DISABLE    0
   1512 #	define R300_CHROMA_KEY_FORCE	       1
   1513 #	define R300_CHROMA_KEY_BLEND           2
   1514 #	define R300_MC_ROUND_NORMAL            (0<<2)
   1515 #	define R300_MC_ROUND_MPEG4             (1<<2)
   1516 #	define R300_LOD_BIAS_SHIFT             3
   1517 #	define R300_LOD_BIAS_MASK	       0x1ff8
   1518 #	define R300_EDGE_ANISO_EDGE_DIAG       (0<<13)
   1519 #	define R300_EDGE_ANISO_EDGE_ONLY       (1<<13)
   1520 #	define R300_MC_COORD_TRUNCATE_DISABLE  (0<<14)
   1521 #	define R300_MC_COORD_TRUNCATE_MPEG     (1<<14)
   1522 #	define R300_TX_TRI_PERF_0_8            (0<<15)
   1523 #	define R300_TX_TRI_PERF_1_8            (1<<15)
   1524 #	define R300_TX_TRI_PERF_1_4            (2<<15)
   1525 #	define R300_TX_TRI_PERF_3_8            (3<<15)
   1526 #	define R300_ANISO_THRESHOLD_MASK       (7<<17)
   1527 
   1528 #       define R400_DXTC_SWIZZLE_ENABLE        (1<<21)
   1529 #	define R500_MACRO_SWITCH               (1<<22)
   1530 #       define R500_TX_MAX_ANISO(x)            ((x) << 23)
   1531 #       define R500_TX_MAX_ANISO_MASK          (63 << 23)
   1532 #       define R500_TX_ANISO_HIGH_QUALITY      (1 << 30)
   1533 #	define R500_BORDER_FIX                 (1<<31)
   1534 
   1535 #define R300_TX_FORMAT0_0                   0x4480
   1536 #       define R300_TX_WIDTHMASK_SHIFT           0
   1537 #       define R300_TX_WIDTHMASK_MASK            (2047 << 0)
   1538 #       define R300_TX_HEIGHTMASK_SHIFT          11
   1539 #       define R300_TX_HEIGHTMASK_MASK           (2047 << 11)
   1540 #	define R300_TX_DEPTHMASK_SHIFT           22
   1541 #	define R300_TX_DEPTHMASK_MASK            (0xf << 22)
   1542 #       define R300_TX_SIZE_PROJECTED            (1 << 30)
   1543 #       define R300_TX_PITCH_EN                  (1 << 31)
   1544 #       define R300_TX_WIDTH(x)                  ((x) << 0)
   1545 #       define R300_TX_HEIGHT(x)                 ((x) << 11)
   1546 #       define R300_TX_DEPTH(x)                  ((x) << 22)
   1547 #       define R300_TX_NUM_LEVELS(x)             ((x) << 26)
   1548 
   1549 #define R300_TX_FORMAT1_0                   0x44C0
   1550 	/* The interpretation of the format word by Wladimir van der Laan */
   1551 	/* The X, Y, Z and W refer to the layout of the components.
   1552 	   They are given meanings as R, G, B and Alpha by the swizzle
   1553 	   specification */
   1554 #	define R300_TX_FORMAT_X8		    0x0
   1555 #	define R300_TX_FORMAT_X16		    0x1
   1556 #	define R300_TX_FORMAT_Y4X4		    0x2
   1557 #	define R300_TX_FORMAT_Y8X8		    0x3
   1558 #	define R300_TX_FORMAT_Y16X16		    0x4
   1559 #	define R300_TX_FORMAT_Z3Y3X2		    0x5
   1560 #	define R300_TX_FORMAT_Z5Y6X5		    0x6
   1561 #	define R300_TX_FORMAT_Z6Y5X5		    0x7
   1562 #	define R300_TX_FORMAT_Z11Y11X10		    0x8
   1563 #	define R300_TX_FORMAT_Z10Y11X11		    0x9
   1564 #	define R300_TX_FORMAT_W4Z4Y4X4		    0xA
   1565 #	define R300_TX_FORMAT_W1Z5Y5X5		    0xB
   1566 #	define R300_TX_FORMAT_W8Z8Y8X8		    0xC
   1567 #	define R300_TX_FORMAT_W2Z10Y10X10	    0xD
   1568 #	define R300_TX_FORMAT_W16Z16Y16X16	    0xE
   1569 #	define R300_TX_FORMAT_DXT1	    	    0xF
   1570 #	define R300_TX_FORMAT_DXT3	    	    0x10
   1571 #	define R300_TX_FORMAT_DXT5	    	    0x11
   1572 #	define R300_TX_FORMAT_CxV8U8           	    0x12
   1573 #	define R300_TX_FORMAT_AVYU444 	    	    0x13
   1574 #	define R300_TX_FORMAT_VYUY422  	    	    0x14
   1575 #	define R300_TX_FORMAT_YVYU422  	    	    0x15
   1576 #	define R300_TX_FORMAT_16_MPEG  	    	    0x16
   1577 #	define R300_TX_FORMAT_16_16_MPEG    	    0x17
   1578 #	define R300_TX_FORMAT_16F     	    	    0x18
   1579 #	define R300_TX_FORMAT_16F_16F 	    	    0x19
   1580 #	define R300_TX_FORMAT_16F_16F_16F_16F  	    0x1A
   1581 #	define R300_TX_FORMAT_32F     	    	    0x1B
   1582 #	define R300_TX_FORMAT_32F_32F 	    	    0x1C
   1583 #	define R300_TX_FORMAT_32F_32F_32F_32F  	    0x1D
   1584 #       define R300_TX_FORMAT_W24_FP                0x1E
   1585 #       define R400_TX_FORMAT_ATI2N                 0x1F
   1586 
   1587 /* These need TX_FORMAT2_[0-15].TXFORMAT_MSB set.
   1588 
   1589    My guess is the 10-bit formats are the 8-bit ones but with filtering being
   1590    performed with the precision of 10 bits per channel. This makes sense
   1591    with sRGB textures since the conversion to linear space reduces the precision
   1592    significantly so the shader gets approximately the 8-bit precision
   1593    in the end. It might also improve the quality of HDR rendering where
   1594    high-precision filtering is desirable.
   1595 
   1596    Again, this is guessed, the formats might mean something entirely else.
   1597    The others should be fine. */
   1598 #       define R500_TX_FORMAT_X1                    0x0
   1599 #       define R500_TX_FORMAT_X1_REV                0x1
   1600 #       define R500_TX_FORMAT_X10                   0x2
   1601 #       define R500_TX_FORMAT_Y10X10                0x3
   1602 #       define R500_TX_FORMAT_W10Z10Y10X10          0x4
   1603 #       define R500_TX_FORMAT_ATI1N                 0x5
   1604 #       define R500_TX_FORMAT_Y8X24                 0x6
   1605 
   1606 
   1607 #       define R300_TX_FORMAT_SIGNED_W             (1 << 5)
   1608 #       define R300_TX_FORMAT_SIGNED_Z             (1 << 6)
   1609 #       define R300_TX_FORMAT_SIGNED_Y             (1 << 7)
   1610 #       define R300_TX_FORMAT_SIGNED_X             (1 << 8)
   1611 #       define R300_TX_FORMAT_SIGNED               (0xf << 5)
   1612 
   1613 #	define R300_TX_FORMAT_3D		   (1 << 25)
   1614 #	define R300_TX_FORMAT_CUBIC_MAP		   (2 << 25)
   1615 #	define R300_TX_FORMAT_TEX_COORD_TYPE_MASK  (0x3 << 25)
   1616 
   1617 	/* alpha modes, convenience mostly */
   1618 	/* if you have alpha, pick constant appropriate to the
   1619 	   number of channels (1 for I8, 2 for I8A8, 4 for R8G8B8A8, etc */
   1620 # 	define R300_TX_FORMAT_ALPHA_1CH		    0x000
   1621 # 	define R300_TX_FORMAT_ALPHA_2CH		    0x200
   1622 # 	define R300_TX_FORMAT_ALPHA_4CH		    0x600
   1623 # 	define R300_TX_FORMAT_ALPHA_NONE	    0xA00
   1624 	/* Swizzling */
   1625 	/* constants */
   1626 #	define R300_TX_FORMAT_X		0
   1627 #	define R300_TX_FORMAT_Y		1
   1628 #	define R300_TX_FORMAT_Z		2
   1629 #	define R300_TX_FORMAT_W		3
   1630 #	define R300_TX_FORMAT_ZERO	4
   1631 #	define R300_TX_FORMAT_ONE	5
   1632 	/* 2.0*Z, everything above 1.0 is set to 0.0 */
   1633 #	define R300_TX_FORMAT_CUT_Z	6
   1634 	/* 2.0*W, everything above 1.0 is set to 0.0 */
   1635 #	define R300_TX_FORMAT_CUT_W	7
   1636 
   1637 #	define R300_TX_FORMAT_B_SHIFT	18
   1638 #	define R300_TX_FORMAT_G_SHIFT	15
   1639 #	define R300_TX_FORMAT_R_SHIFT	12
   1640 #	define R300_TX_FORMAT_A_SHIFT	9
   1641 	/* Convenience macro to take care of layout and swizzling */
   1642 #	define R300_EASY_TX_FORMAT(B, G, R, A, FMT)	(		\
   1643 		((R300_TX_FORMAT_##B)<<R300_TX_FORMAT_B_SHIFT)		\
   1644 		| ((R300_TX_FORMAT_##G)<<R300_TX_FORMAT_G_SHIFT)	\
   1645 		| ((R300_TX_FORMAT_##R)<<R300_TX_FORMAT_R_SHIFT)	\
   1646 		| ((R300_TX_FORMAT_##A)<<R300_TX_FORMAT_A_SHIFT)	\
   1647 		| (R300_TX_FORMAT_##FMT)				\
   1648 		)
   1649 	/* These can be ORed with result of R300_EASY_TX_FORMAT()
   1650 	   We don't really know what they do. Take values from a
   1651            constant color ? */
   1652 #	define R300_TX_FORMAT_CONST_X		(1<<5)
   1653 #	define R300_TX_FORMAT_CONST_Y		(2<<5)
   1654 #	define R300_TX_FORMAT_CONST_Z		(4<<5)
   1655 #	define R300_TX_FORMAT_CONST_W		(8<<5)
   1656 
   1657 #       define R300_TX_FORMAT_GAMMA               (1 << 21)
   1658 #       define R300_TX_FORMAT_YUV_TO_RGB          (1 << 22)
   1659 
   1660 #       define R300_TX_CACHE(x)                 ((x) << 27)
   1661 #       define R300_TX_CACHE_WHOLE              0
   1662 /* reserved */
   1663 #       define R300_TX_CACHE_HALF_0             2
   1664 #       define R300_TX_CACHE_HALF_1             3
   1665 #       define R300_TX_CACHE_FOURTH_0           4
   1666 #       define R300_TX_CACHE_FOURTH_1           5
   1667 #       define R300_TX_CACHE_FOURTH_2           6
   1668 #       define R300_TX_CACHE_FOURTH_3           7
   1669 #       define R300_TX_CACHE_EIGHTH_0           8
   1670 #       define R300_TX_CACHE_EIGHTH_1           9
   1671 #       define R300_TX_CACHE_EIGHTH_2           10
   1672 #       define R300_TX_CACHE_EIGHTH_3           11
   1673 #       define R300_TX_CACHE_EIGHTH_4           12
   1674 #       define R300_TX_CACHE_EIGHTH_5           13
   1675 #       define R300_TX_CACHE_EIGHTH_6           14
   1676 #       define R300_TX_CACHE_EIGHTH_7           15
   1677 #       define R300_TX_CACHE_SIXTEENTH_0        16
   1678 #       define R300_TX_CACHE_SIXTEENTH_1        17
   1679 #       define R300_TX_CACHE_SIXTEENTH_2        18
   1680 #       define R300_TX_CACHE_SIXTEENTH_3        19
   1681 #       define R300_TX_CACHE_SIXTEENTH_4        20
   1682 #       define R300_TX_CACHE_SIXTEENTH_5        21
   1683 #       define R300_TX_CACHE_SIXTEENTH_6        22
   1684 #       define R300_TX_CACHE_SIXTEENTH_7        23
   1685 #       define R300_TX_CACHE_SIXTEENTH_8        24
   1686 #       define R300_TX_CACHE_SIXTEENTH_9        25
   1687 #       define R300_TX_CACHE_SIXTEENTH_10       26
   1688 #       define R300_TX_CACHE_SIXTEENTH_11       27
   1689 #       define R300_TX_CACHE_SIXTEENTH_12       28
   1690 #       define R300_TX_CACHE_SIXTEENTH_13       29
   1691 #       define R300_TX_CACHE_SIXTEENTH_14       30
   1692 #       define R300_TX_CACHE_SIXTEENTH_15       31
   1693 
   1694 #define R300_TX_FORMAT2_0		    0x4500 /* obvious missing in gap */
   1695 #       define R300_TX_PITCHMASK_SHIFT           0
   1696 #       define R300_TX_PITCHMASK_MASK            (2047 << 0)
   1697 #	define R500_TXFORMAT_MSB		 (1 << 14)
   1698 #	define R500_TXWIDTH_BIT11	         (1 << 15)
   1699 #	define R500_TXHEIGHT_BIT11	         (1 << 16)
   1700 #	define R500_POW2FIX2FLT			 (1 << 17)
   1701 #	define R500_SEL_FILTER4_TC0		 (0 << 18)
   1702 #	define R500_SEL_FILTER4_TC1		 (1 << 18)
   1703 #	define R500_SEL_FILTER4_TC2		 (2 << 18)
   1704 #	define R500_SEL_FILTER4_TC3		 (3 << 18)
   1705 
   1706 #define R300_TX_OFFSET_0                    0x4540
   1707 #define R300_TX_OFFSET_1                    0x4544
   1708 #define R300_TX_OFFSET_2                    0x4548
   1709 #define R300_TX_OFFSET_3                    0x454C
   1710 #define R300_TX_OFFSET_4                    0x4550
   1711 #define R300_TX_OFFSET_5                    0x4554
   1712 #define R300_TX_OFFSET_6                    0x4558
   1713 #define R300_TX_OFFSET_7                    0x455C
   1714 
   1715 #       define R300_TXO_ENDIAN_NO_SWAP           (0 << 0)
   1716 #       define R300_TXO_ENDIAN_BYTE_SWAP         (1 << 0)
   1717 #       define R300_TXO_ENDIAN_WORD_SWAP         (2 << 0)
   1718 #       define R300_TXO_ENDIAN_HALFDW_SWAP       (3 << 0)
   1719 #       define R300_TXO_MACRO_TILE_LINEAR        (0 << 2)
   1720 #       define R300_TXO_MACRO_TILE_TILED         (1 << 2)
   1721 #       define R300_TXO_MACRO_TILE(x)            ((x) << 2)
   1722 #       define R300_TXO_MICRO_TILE_LINEAR        (0 << 3)
   1723 #       define R300_TXO_MICRO_TILE_TILED         (1 << 3)
   1724 #       define R300_TXO_MICRO_TILE_TILED_SQUARE  (2 << 3)
   1725 #       define R300_TXO_MICRO_TILE(x)            ((x) << 3)
   1726 #       define R300_TXO_OFFSET_MASK              0xffffffe0
   1727 #       define R300_TXO_OFFSET_SHIFT             5
   1728 
   1729 /* 32 bit chroma key */
   1730 #define R300_TX_CHROMA_KEY_0                      0x4580
   1731 #define R300_TX_CHROMA_KEY_1                      0x4584
   1732 #define R300_TX_CHROMA_KEY_2                      0x4588
   1733 #define R300_TX_CHROMA_KEY_3                      0x458c
   1734 #define R300_TX_CHROMA_KEY_4                      0x4590
   1735 #define R300_TX_CHROMA_KEY_5                      0x4594
   1736 #define R300_TX_CHROMA_KEY_6                      0x4598
   1737 #define R300_TX_CHROMA_KEY_7                      0x459c
   1738 #define R300_TX_CHROMA_KEY_8                      0x45a0
   1739 #define R300_TX_CHROMA_KEY_9                      0x45a4
   1740 #define R300_TX_CHROMA_KEY_10                     0x45a8
   1741 #define R300_TX_CHROMA_KEY_11                     0x45ac
   1742 #define R300_TX_CHROMA_KEY_12                     0x45b0
   1743 #define R300_TX_CHROMA_KEY_13                     0x45b4
   1744 #define R300_TX_CHROMA_KEY_14                     0x45b8
   1745 #define R300_TX_CHROMA_KEY_15                     0x45bc
   1746 /* ff00ff00 == { 0, 1.0, 0, 1.0 } */
   1747 
   1748 /* Border Color */
   1749 #define R300_TX_BORDER_COLOR_0              0x45c0
   1750 #define R300_TX_BORDER_COLOR_1              0x45c4
   1751 #define R300_TX_BORDER_COLOR_2              0x45c8
   1752 #define R300_TX_BORDER_COLOR_3              0x45cc
   1753 #define R300_TX_BORDER_COLOR_4              0x45d0
   1754 #define R300_TX_BORDER_COLOR_5              0x45d4
   1755 #define R300_TX_BORDER_COLOR_6              0x45d8
   1756 #define R300_TX_BORDER_COLOR_7              0x45dc
   1757 #define R300_TX_BORDER_COLOR_8              0x45e0
   1758 #define R300_TX_BORDER_COLOR_9              0x45e4
   1759 #define R300_TX_BORDER_COLOR_10             0x45e8
   1760 #define R300_TX_BORDER_COLOR_11             0x45ec
   1761 #define R300_TX_BORDER_COLOR_12             0x45f0
   1762 #define R300_TX_BORDER_COLOR_13             0x45f4
   1763 #define R300_TX_BORDER_COLOR_14             0x45f8
   1764 #define R300_TX_BORDER_COLOR_15             0x45fc
   1765 
   1766 
   1767 /* END: Texture specification */
   1768 
   1769 /* BEGIN: Fragment program instruction set */
   1770 
   1771 /* Fragment programs are written directly into register space.
   1772  * There are separate instruction streams for texture instructions and ALU
   1773  * instructions.
   1774  * In order to synchronize these streams, the program is divided into up
   1775  * to 4 nodes. Each node begins with a number of TEX operations, followed
   1776  * by a number of ALU operations.
   1777  * The first node can have zero TEX ops, all subsequent nodes must have at
   1778  * least
   1779  * one TEX ops.
   1780  * All nodes must have at least one ALU op.
   1781  *
   1782  * The index of the last node is stored in PFS_CNTL_0: A value of 0 means
   1783  * 1 node, a value of 3 means 4 nodes.
   1784  * The total amount of instructions is defined in PFS_CNTL_2. The offsets are
   1785  * offsets into the respective instruction streams, while *_END points to the
   1786  * last instruction relative to this offset.
   1787  */
   1788 #define R300_US_CONFIG                      0x4600
   1789 #       define R300_PFS_CNTL_LAST_NODES_SHIFT    0
   1790 #       define R300_PFS_CNTL_LAST_NODES_MASK     (3 << 0)
   1791 #       define R300_PFS_CNTL_FIRST_NODE_HAS_TEX  (1 << 3)
   1792 #define R300_US_PIXSIZE                     0x4604
   1793 /* There is an unshifted value here which has so far always been equal to the
   1794  * index of the highest used temporary register.
   1795  */
   1796 #define R300_US_CODE_OFFSET                 0x4608
   1797 #       define R300_PFS_CNTL_ALU_OFFSET_SHIFT    0
   1798 #       define R300_PFS_CNTL_ALU_OFFSET_MASK     (63 << 0)
   1799 #       define R300_PFS_CNTL_ALU_END_SHIFT       6
   1800 #       define R300_PFS_CNTL_ALU_END_MASK        (63 << 6)
   1801 #       define R300_PFS_CNTL_TEX_OFFSET_SHIFT    13
   1802 #       define R300_PFS_CNTL_TEX_OFFSET_MASK     (31 << 13)
   1803 #       define R300_PFS_CNTL_TEX_END_SHIFT       18
   1804 #       define R300_PFS_CNTL_TEX_END_MASK        (31 << 18)
   1805 #       define R400_PFS_CNTL_TEX_OFFSET_MSB_SHIFT 24
   1806 #       define R400_PFS_CNTL_TEX_OFFSET_MSB_MASK (0xf << 24)
   1807 #       define R400_PFS_CNTL_TEX_END_MSB_SHIFT   28
   1808 #       define R400_PFS_CNTL_TEX_END_MSB_MASK    (0xf << 28)
   1809 
   1810 /* gap */
   1811 
   1812 /* Nodes are stored backwards. The last active node is always stored in
   1813  * PFS_NODE_3.
   1814  * Example: In a 2-node program, NODE_0 and NODE_1 are set to 0. The
   1815  * first node is stored in NODE_2, the second node is stored in NODE_3.
   1816  *
   1817  * Offsets are relative to the master offset from PFS_CNTL_2.
   1818  */
   1819 #define R300_US_CODE_ADDR_0                 0x4610
   1820 #define R300_US_CODE_ADDR_1                 0x4614
   1821 #define R300_US_CODE_ADDR_2                 0x4618
   1822 #define R300_US_CODE_ADDR_3                 0x461C
   1823 #       define R300_ALU_START_SHIFT         0
   1824 #       define R300_ALU_START_MASK          (63 << 0)
   1825 #       define R300_ALU_SIZE_SHIFT          6
   1826 #       define R300_ALU_SIZE_MASK           (63 << 6)
   1827 #       define R300_TEX_START_SHIFT         12
   1828 #       define R300_TEX_START_MASK          (31 << 12)
   1829 #       define R300_TEX_SIZE_SHIFT          17
   1830 #       define R300_TEX_SIZE_MASK           (31 << 17)
   1831 #	define R300_RGBA_OUT                (1 << 22)
   1832 #	define R300_W_OUT                   (1 << 23)
   1833 #       define R400_TEX_START_MSB_SHIFT     24
   1834 #       define R400_TEX_START_MSG_MASK      (0xf << 24)
   1835 #       define R400_TEX_SIZE_MSB_SHIFT      28
   1836 #       define R400_TEX_SIZE_MSG_MASK       (0xf << 28)
   1837 
   1838 /* TEX
   1839  * As far as I can tell, texture instructions cannot write into output
   1840  * registers directly. A subsequent ALU instruction is always necessary,
   1841  * even if it's just MAD o0, r0, 1, 0
   1842  */
   1843 #define R300_US_TEX_INST_0                  0x4620
   1844 #	define R300_SRC_ADDR_SHIFT          0
   1845 #	define R300_SRC_ADDR_MASK           (31 << 0)
   1846 #	define R300_DST_ADDR_SHIFT          6
   1847 #	define R300_DST_ADDR_MASK           (31 << 6)
   1848 #	define R300_TEX_ID_SHIFT            11
   1849 #       define R300_TEX_ID_MASK             (15 << 11)
   1850 #	define R300_TEX_INST_SHIFT		15
   1851 #		define R300_TEX_OP_NOP	        0
   1852 #		define R300_TEX_OP_LD	        1
   1853 #		define R300_TEX_OP_KIL	        2
   1854 #		define R300_TEX_OP_TXP	        3
   1855 #		define R300_TEX_OP_TXB	        4
   1856 #	define R300_TEX_INST_MASK               (7 << 15)
   1857 #      define R400_SRC_ADDR_EXT_BIT         (1 << 19)
   1858 #      define R400_DST_ADDR_EXT_BIT         (1 << 20)
   1859 
   1860 /* Output format from the unfied shader */
   1861 #define R300_US_OUT_FMT_0                   0x46A4
   1862 #	define R300_US_OUT_FMT_C4_8         (0 << 0)
   1863 #	define R300_US_OUT_FMT_C4_10        (1 << 0)
   1864 #	define R300_US_OUT_FMT_C4_10_GAMMA  (2 << 0)
   1865 #	define R300_US_OUT_FMT_C_16         (3 << 0)
   1866 #	define R300_US_OUT_FMT_C2_16        (4 << 0)
   1867 #	define R300_US_OUT_FMT_C4_16        (5 << 0)
   1868 #	define R300_US_OUT_FMT_C_16_MPEG    (6 << 0)
   1869 #	define R300_US_OUT_FMT_C2_16_MPEG   (7 << 0)
   1870 #	define R300_US_OUT_FMT_C2_4         (8 << 0)
   1871 #	define R300_US_OUT_FMT_C_3_3_2      (9 << 0)
   1872 #	define R300_US_OUT_FMT_C_6_5_6      (10 << 0)
   1873 #	define R300_US_OUT_FMT_C_11_11_10   (11 << 0)
   1874 #	define R300_US_OUT_FMT_C_10_11_11   (12 << 0)
   1875 #	define R300_US_OUT_FMT_C_2_10_10_10 (13 << 0)
   1876 /* reserved */
   1877 #	define R300_US_OUT_FMT_UNUSED       (15 << 0)
   1878 #	define R300_US_OUT_FMT_C_16_FP      (16 << 0)
   1879 #	define R300_US_OUT_FMT_C2_16_FP     (17 << 0)
   1880 #	define R300_US_OUT_FMT_C4_16_FP     (18 << 0)
   1881 #	define R300_US_OUT_FMT_C_32_FP      (19 << 0)
   1882 #	define R300_US_OUT_FMT_C2_32_FP     (20 << 0)
   1883 #	define R300_US_OUT_FMT_C4_32_FP     (21 << 0)
   1884 #   define R300_C0_SEL_A				(0 << 8)
   1885 #   define R300_C0_SEL_R				(1 << 8)
   1886 #   define R300_C0_SEL_G				(2 << 8)
   1887 #   define R300_C0_SEL_B				(3 << 8)
   1888 #   define R300_C1_SEL_A				(0 << 10)
   1889 #   define R300_C1_SEL_R				(1 << 10)
   1890 #   define R300_C1_SEL_G				(2 << 10)
   1891 #   define R300_C1_SEL_B				(3 << 10)
   1892 #   define R300_C2_SEL_A				(0 << 12)
   1893 #   define R300_C2_SEL_R				(1 << 12)
   1894 #   define R300_C2_SEL_G				(2 << 12)
   1895 #   define R300_C2_SEL_B				(3 << 12)
   1896 #   define R300_C3_SEL_A				(0 << 14)
   1897 #   define R300_C3_SEL_R				(1 << 14)
   1898 #   define R300_C3_SEL_G				(2 << 14)
   1899 #   define R300_C3_SEL_B				(3 << 14)
   1900 #   define R300_OUT_SIGN(x)				((x) << 16)
   1901 #   define R500_ROUND_ADJ				(1 << 20)
   1902 
   1903 /* ALU
   1904  * The ALU instructions register blocks are enumerated according to the order
   1905  * in which fglrx. I assume there is space for 64 instructions, since
   1906  * each block has space for a maximum of 64 DWORDs, and this matches reported
   1907  * native limits.
   1908  *
   1909  * The basic functional block seems to be one MAD for each color and alpha,
   1910  * and an adder that adds all components after the MUL.
   1911  *  - ADD, MUL, MAD etc.: use MAD with appropriate neutral operands
   1912  *  - DP4: Use OUTC_DP4, OUTA_DP4
   1913  *  - DP3: Use OUTC_DP3, OUTA_DP4, appropriate alpha operands
   1914  *  - DPH: Use OUTC_DP4, OUTA_DP4, appropriate alpha operands
   1915  *  - CMPH: If ARG2 > 0.5, return ARG0, else return ARG1
   1916  *  - CMP: If ARG2 < 0, return ARG1, else return ARG0
   1917  *  - FLR: use FRC+MAD
   1918  *  - XPD: use MAD+MAD
   1919  *  - SGE, SLT: use MAD+CMP
   1920  *  - RSQ: use ABS modifier for argument
   1921  *  - Use OUTC_REPL_ALPHA to write results of an alpha-only operation
   1922  *    (e.g. RCP) into color register
   1923  *  - apparently, there's no quick DST operation
   1924  *  - fglrx set FPI2_UNKNOWN_31 on a "MAD fragment.color, tmp0, tmp1, tmp2"
   1925  *  - fglrx set FPI2_UNKNOWN_31 on a "MAX r2, r1, c0"
   1926  *  - fglrx once set FPI0_UNKNOWN_31 on a "FRC r1, r1"
   1927  *
   1928  * Operand selection
   1929  * First stage selects three sources from the available registers and
   1930  * constant parameters. This is defined in INSTR1 (color) and INSTR3 (alpha).
   1931  * fglrx sorts the three source fields: Registers before constants,
   1932  * lower indices before higher indices; I do not know whether this is
   1933  * necessary.
   1934  *
   1935  * fglrx fills unused sources with "read constant 0"
   1936  * According to specs, you cannot select more than two different constants.
   1937  *
   1938  * Second stage selects the operands from the sources. This is defined in
   1939  * INSTR0 (color) and INSTR2 (alpha). You can also select the special constants
   1940  * zero and one.
   1941  * Swizzling and negation happens in this stage, as well.
   1942  *
   1943  * Important: Color and alpha seem to be mostly separate, i.e. their sources
   1944  * selection appears to be fully independent (the register storage is probably
   1945  * physically split into a color and an alpha section).
   1946  * However (because of the apparent physical split), there is some interaction
   1947  * WRT swizzling. If, for example, you want to load an R component into an
   1948  * Alpha operand, this R component is taken from a *color* source, not from
   1949  * an alpha source. The corresponding register doesn't even have to appear in
   1950  * the alpha sources list. (I hope this all makes sense to you)
   1951  *
   1952  * Destination selection
   1953  * The destination register index is in FPI1 (color) and FPI3 (alpha)
   1954  * together with enable bits.
   1955  * There are separate enable bits for writing into temporary registers
   1956  * (DSTC_REG_* /DSTA_REG) and program output registers (DSTC_OUTPUT_*
   1957  * /DSTA_OUTPUT). You can write to both at once, or not write at all (the
   1958  * same index must be used for both).
   1959  *
   1960  * Note: There is a special form for LRP
   1961  *  - Argument order is the same as in ARB_fragment_program.
   1962  *  - Operation is MAD
   1963  *  - ARG1 is set to ARGC_SRC1C_LRP/ARGC_SRC1A_LRP
   1964  *  - Set FPI0/FPI2_SPECIAL_LRP
   1965  * Arbitrary LRP (including support for swizzling) requires vanilla MAD+MAD
   1966  */
   1967 #define R300_US_ALU_RGB_ADDR_0                   0x46C0
   1968 #       define R300_ALU_SRC0C_SHIFT             0
   1969 #       define R300_ALU_SRC0C_MASK              (31 << 0)
   1970 #       define R300_ALU_SRC0C_CONST             (1 << 5)
   1971 #       define R300_ALU_SRC1C_SHIFT             6
   1972 #       define R300_ALU_SRC1C_MASK              (31 << 6)
   1973 #       define R300_ALU_SRC1C_CONST             (1 << 11)
   1974 #       define R300_ALU_SRC2C_SHIFT             12
   1975 #       define R300_ALU_SRC2C_MASK              (31 << 12)
   1976 #       define R300_ALU_SRC2C_CONST             (1 << 17)
   1977 #       define R300_ALU_SRC_MASK                0x0003ffff
   1978 #       define R300_ALU_DSTC_SHIFT              18
   1979 #       define R300_ALU_DSTC_MASK               (31 << 18)
   1980 #		define R300_ALU_DSTC_REG_MASK_SHIFT     23
   1981 #       define R300_ALU_DSTC_REG_X              (1 << 23)
   1982 #       define R300_ALU_DSTC_REG_Y              (1 << 24)
   1983 #       define R300_ALU_DSTC_REG_Z              (1 << 25)
   1984 #		define R300_ALU_DSTC_OUTPUT_MASK_SHIFT  26
   1985 #       define R300_ALU_DSTC_OUTPUT_X           (1 << 26)
   1986 #       define R300_ALU_DSTC_OUTPUT_Y           (1 << 27)
   1987 #       define R300_ALU_DSTC_OUTPUT_Z           (1 << 28)
   1988 #       define R300_ALU_DSTC_OUTPUT_XYZ         (7 << 26)
   1989 #       define R300_RGB_ADDR0(x)                ((x) << 0)
   1990 #       define R300_RGB_ADDR1(x)                ((x) << 6)
   1991 #       define R300_RGB_ADDR2(x)                ((x) << 12)
   1992 #       define R300_RGB_TARGET(x)               ((x) << 29)
   1993 
   1994 #define R300_US_ALU_ALPHA_ADDR_0                 0x47C0
   1995 #       define R300_ALU_SRC0A_SHIFT             0
   1996 #       define R300_ALU_SRC0A_MASK              (31 << 0)
   1997 #       define R300_ALU_SRC0A_CONST             (1 << 5)
   1998 #       define R300_ALU_SRC1A_SHIFT             6
   1999 #       define R300_ALU_SRC1A_MASK              (31 << 6)
   2000 #       define R300_ALU_SRC1A_CONST             (1 << 11)
   2001 #       define R300_ALU_SRC2A_SHIFT             12
   2002 #       define R300_ALU_SRC2A_MASK              (31 << 12)
   2003 #       define R300_ALU_SRC2A_CONST             (1 << 17)
   2004 #       define R300_ALU_SRC_MASK                0x0003ffff
   2005 #       define R300_ALU_DSTA_SHIFT              18
   2006 #       define R300_ALU_DSTA_MASK               (31 << 18)
   2007 #       define R300_ALU_DSTA_REG                (1 << 23)
   2008 #       define R300_ALU_DSTA_OUTPUT             (1 << 24)
   2009 #		define R300_ALU_DSTA_DEPTH              (1 << 27)
   2010 #       define R300_ALPHA_ADDR0(x)              ((x) << 0)
   2011 #       define R300_ALPHA_ADDR1(x)              ((x) << 6)
   2012 #       define R300_ALPHA_ADDR2(x)              ((x) << 12)
   2013 #       define R300_ALPHA_TARGET(x)             ((x) << 25)
   2014 
   2015 #define R300_US_ALU_RGB_INST_0                   0x48C0
   2016 #       define R300_ALU_ARGC_SRC0C_XYZ          0
   2017 #       define R300_ALU_ARGC_SRC0C_XXX          1
   2018 #       define R300_ALU_ARGC_SRC0C_YYY          2
   2019 #       define R300_ALU_ARGC_SRC0C_ZZZ          3
   2020 #       define R300_ALU_ARGC_SRC1C_XYZ          4
   2021 #       define R300_ALU_ARGC_SRC1C_XXX          5
   2022 #       define R300_ALU_ARGC_SRC1C_YYY          6
   2023 #       define R300_ALU_ARGC_SRC1C_ZZZ          7
   2024 #       define R300_ALU_ARGC_SRC2C_XYZ          8
   2025 #       define R300_ALU_ARGC_SRC2C_XXX          9
   2026 #       define R300_ALU_ARGC_SRC2C_YYY          10
   2027 #       define R300_ALU_ARGC_SRC2C_ZZZ          11
   2028 #       define R300_ALU_ARGC_SRC0A              12
   2029 #       define R300_ALU_ARGC_SRC1A              13
   2030 #       define R300_ALU_ARGC_SRC2A              14
   2031 #       define R300_ALU_ARGC_SRCP_XYZ           15
   2032 #       define R300_ALU_ARGC_SRCP_XXX           16
   2033 #       define R300_ALU_ARGC_SRCP_YYY           17
   2034 #       define R300_ALU_ARGC_SRCP_ZZZ           18
   2035 #       define R300_ALU_ARGC_SRCP_WWW           19
   2036 #       define R300_ALU_ARGC_ZERO               20
   2037 #       define R300_ALU_ARGC_ONE                21
   2038 #       define R300_ALU_ARGC_HALF               22
   2039 #       define R300_ALU_ARGC_SRC0C_YZX          23
   2040 #       define R300_ALU_ARGC_SRC1C_YZX          24
   2041 #       define R300_ALU_ARGC_SRC2C_YZX          25
   2042 #       define R300_ALU_ARGC_SRC0C_ZXY          26
   2043 #       define R300_ALU_ARGC_SRC1C_ZXY          27
   2044 #       define R300_ALU_ARGC_SRC2C_ZXY          28
   2045 #       define R300_ALU_ARGC_SRC0CA_WZY         29
   2046 #       define R300_ALU_ARGC_SRC1CA_WZY         30
   2047 #       define R300_ALU_ARGC_SRC2CA_WZY         31
   2048 #       define R300_RGB_SWIZA(x)                ((x) << 0)
   2049 #       define R300_RGB_SWIZB(x)                ((x) << 7)
   2050 #       define R300_RGB_SWIZC(x)                ((x) << 14)
   2051 
   2052 #       define R300_ALU_ARG0C_SHIFT             0
   2053 #       define R300_ALU_ARG0C_MASK              (31 << 0)
   2054 #       define R300_ALU_ARG0C_NOP               (0 << 5)
   2055 #       define R300_ALU_ARG0C_NEG               (1 << 5)
   2056 #       define R300_ALU_ARG0C_ABS               (2 << 5)
   2057 #       define R300_ALU_ARG0C_NAB               (3 << 5)
   2058 #       define R300_ALU_ARG1C_SHIFT             7
   2059 #       define R300_ALU_ARG1C_MASK              (31 << 7)
   2060 #       define R300_ALU_ARG1C_NOP               (0 << 12)
   2061 #       define R300_ALU_ARG1C_NEG               (1 << 12)
   2062 #       define R300_ALU_ARG1C_ABS               (2 << 12)
   2063 #       define R300_ALU_ARG1C_NAB               (3 << 12)
   2064 #       define R300_ALU_ARG2C_SHIFT             14
   2065 #       define R300_ALU_ARG2C_MASK              (31 << 14)
   2066 #       define R300_ALU_ARG2C_NOP               (0 << 19)
   2067 #       define R300_ALU_ARG2C_NEG               (1 << 19)
   2068 #       define R300_ALU_ARG2C_ABS               (2 << 19)
   2069 #       define R300_ALU_ARG2C_NAB               (3 << 19)
   2070 #       define R300_ALU_SRCP_1_MINUS_2_SRC0     (0 << 21)
   2071 #       define R300_ALU_SRCP_SRC1_MINUS_SRC0    (1 << 21)
   2072 #       define R300_ALU_SRCP_SRC1_PLUS_SRC0     (2 << 21)
   2073 #       define R300_ALU_SRCP_1_MINUS_SRC0       (3 << 21)
   2074 
   2075 #       define R300_ALU_OUTC_MAD                (0 << 23)
   2076 #       define R300_ALU_OUTC_DP3                (1 << 23)
   2077 #       define R300_ALU_OUTC_DP4                (2 << 23)
   2078 #       define R300_ALU_OUTC_D2A                (3 << 23)
   2079 #       define R300_ALU_OUTC_MIN                (4 << 23)
   2080 #       define R300_ALU_OUTC_MAX                (5 << 23)
   2081 #       define R300_ALU_OUTC_CND                (7 << 23)
   2082 #       define R300_ALU_OUTC_CMP                (8 << 23)
   2083 #       define R300_ALU_OUTC_FRC                (9 << 23)
   2084 #       define R300_ALU_OUTC_REPL_ALPHA         (10 << 23)
   2085 
   2086 #	define R300_ALU_OUTC_MOD_SHIFT		27
   2087 #       define R300_ALU_OUTC_MOD_NOP            (0 << R300_ALU_OUTC_MOD_SHIFT)
   2088 #       define R300_ALU_OUTC_MOD_MUL2           (1 << R300_ALU_OUTC_MOD_SHIFT)
   2089 #       define R300_ALU_OUTC_MOD_MUL4           (2 << R300_ALU_OUTC_MOD_SHIFT)
   2090 #       define R300_ALU_OUTC_MOD_MUL8           (3 << R300_ALU_OUTC_MOD_SHIFT)
   2091 #       define R300_ALU_OUTC_MOD_DIV2           (4 << R300_ALU_OUTC_MOD_SHIFT)
   2092 #       define R300_ALU_OUTC_MOD_DIV4           (5 << R300_ALU_OUTC_MOD_SHIFT)
   2093 #       define R300_ALU_OUTC_MOD_DIV8           (6 << R300_ALU_OUTC_MOD_SHIFT)
   2094 
   2095 #       define R300_ALU_OUTC_CLAMP              (1 << 30)
   2096 #       define R300_ALU_INSERT_NOP              (1 << 31)
   2097 
   2098 #define R300_US_ALU_ALPHA_INST_0                 0x49C0
   2099 #       define R300_ALU_ARGA_SRC0C_X            0
   2100 #       define R300_ALU_ARGA_SRC0C_Y            1
   2101 #       define R300_ALU_ARGA_SRC0C_Z            2
   2102 #       define R300_ALU_ARGA_SRC1C_X            3
   2103 #       define R300_ALU_ARGA_SRC1C_Y            4
   2104 #       define R300_ALU_ARGA_SRC1C_Z            5
   2105 #       define R300_ALU_ARGA_SRC2C_X            6
   2106 #       define R300_ALU_ARGA_SRC2C_Y            7
   2107 #       define R300_ALU_ARGA_SRC2C_Z            8
   2108 #       define R300_ALU_ARGA_SRC0A              9
   2109 #       define R300_ALU_ARGA_SRC1A              10
   2110 #       define R300_ALU_ARGA_SRC2A              11
   2111 #       define R300_ALU_ARGA_SRCP_X             12
   2112 #       define R300_ALU_ARGA_SRCP_Y             13
   2113 #       define R300_ALU_ARGA_SRCP_Z             14
   2114 #       define R300_ALU_ARGA_SRCP_W             15
   2115 #       define R300_ALU_ARGA_ZERO               16
   2116 #       define R300_ALU_ARGA_ONE                17
   2117 #       define R300_ALU_ARGA_HALF               18
   2118 #       define R300_ALPHA_SWIZA(x)              ((x) << 0)
   2119 #       define R300_ALPHA_SWIZB(x)              ((x) << 7)
   2120 #       define R300_ALPHA_SWIZC(x)              ((x) << 14)
   2121 
   2122 #       define R300_ALU_ARG0A_SHIFT             0
   2123 #       define R300_ALU_ARG0A_MASK              (31 << 0)
   2124 #       define R300_ALU_ARG0A_NOP               (0 << 5)
   2125 #       define R300_ALU_ARG0A_NEG               (1 << 5)
   2126 #	define R300_ALU_ARG0A_ABS		 (2 << 5)
   2127 #	define R300_ALU_ARG0A_NAB		 (3 << 5)
   2128 #       define R300_ALU_ARG1A_SHIFT             7
   2129 #       define R300_ALU_ARG1A_MASK              (31 << 7)
   2130 #       define R300_ALU_ARG1A_NOP               (0 << 12)
   2131 #       define R300_ALU_ARG1A_NEG               (1 << 12)
   2132 #	define R300_ALU_ARG1A_ABS		 (2 << 12)
   2133 #	define R300_ALU_ARG1A_NAB		 (3 << 12)
   2134 #       define R300_ALU_ARG2A_SHIFT             14
   2135 #       define R300_ALU_ARG2A_MASK              (31 << 14)
   2136 #       define R300_ALU_ARG2A_NOP               (0 << 19)
   2137 #       define R300_ALU_ARG2A_NEG               (1 << 19)
   2138 #	define R300_ALU_ARG2A_ABS		 (2 << 19)
   2139 #	define R300_ALU_ARG2A_NAB		 (3 << 19)
   2140 #       define R300_ALU_SRCP_1_MINUS_2_SRC0     (0 << 21)
   2141 #       define R300_ALU_SRCP_SRC1_MINUS_SRC0    (1 << 21)
   2142 #       define R300_ALU_SRCP_SRC1_PLUS_SRC0     (2 << 21)
   2143 #       define R300_ALU_SRCP_1_MINUS_SRC0       (3 << 21)
   2144 
   2145 #       define R300_ALU_OUTA_MAD                (0 << 23)
   2146 #       define R300_ALU_OUTA_DP4                (1 << 23)
   2147 #       define R300_ALU_OUTA_MIN                (2 << 23)
   2148 #       define R300_ALU_OUTA_MAX                (3 << 23)
   2149 #       define R300_ALU_OUTA_CND                (5 << 23)
   2150 #       define R300_ALU_OUTA_CMP                (6 << 23)
   2151 #       define R300_ALU_OUTA_FRC                (7 << 23)
   2152 #       define R300_ALU_OUTA_EX2                (8 << 23)
   2153 #       define R300_ALU_OUTA_LG2                (9 << 23)
   2154 #       define R300_ALU_OUTA_RCP                (10 << 23)
   2155 #       define R300_ALU_OUTA_RSQ                (11 << 23)
   2156 
   2157 #       define R300_ALU_OUTA_MOD_NOP            (0 << 27)
   2158 #       define R300_ALU_OUTA_MOD_MUL2           (1 << 27)
   2159 #       define R300_ALU_OUTA_MOD_MUL4           (2 << 27)
   2160 #       define R300_ALU_OUTA_MOD_MUL8           (3 << 27)
   2161 #       define R300_ALU_OUTA_MOD_DIV2           (4 << 27)
   2162 #       define R300_ALU_OUTA_MOD_DIV4           (5 << 27)
   2163 #       define R300_ALU_OUTA_MOD_DIV8           (6 << 27)
   2164 
   2165 #       define R300_ALU_OUTA_CLAMP              (1 << 30)
   2166 /* END: Fragment program instruction set */
   2167 
   2168 /* R4xx extended fragment shader registers. */
   2169 #define R400_US_ALU_EXT_ADDR_0              0x4ac0 /* up to 63 (0x4bbc) */
   2170 #   define R400_ADDR_EXT_RGB_MSB_BIT(x)     (1 << (x))
   2171 #   define R400_ADDRD_EXT_RGB_MSB_BIT       0x08
   2172 #   define R400_ADDR_EXT_A_MSB_BIT(x)       (1 << ((x) + 4))
   2173 #   define R400_ADDRD_EXT_A_MSB_BIT         0x80
   2174 
   2175 #define R400_US_CODE_BANK                   0x46b8
   2176 #   define R400_BANK_SHIFT                  0
   2177 #   define R400_BANK_MASK                   0xf
   2178 #   define R400_R390_MODE_ENABLE            (1 << 4)
   2179 #define R400_US_CODE_EXT                    0x46bc
   2180 #   define R400_ALU_OFFSET_MSB_SHIFT        0
   2181 #   define R400_ALU_OFFSET_MSB_MASK         (0x7 << 0)
   2182 #   define R400_ALU_SIZE_MSB_SHIFT          3
   2183 #   define R400_ALU_SIZE_MSB_MASK           (0x7 << 3)
   2184 #   define R400_ALU_START0_MSB_SHIFT        6
   2185 #   define R400_ALU_START0_MSB_MASK         (0x7 << 6)
   2186 #   define R400_ALU_SIZE0_MSB_SHIFT         9
   2187 #   define R400_ALU_SIZE0_MSB_MASK          (0x7 << 9)
   2188 #   define R400_ALU_START1_MSB_SHIFT        12
   2189 #   define R400_ALU_START1_MSB_MASK         (0x7 << 12)
   2190 #   define R400_ALU_SIZE1_MSB_SHIFT         15
   2191 #   define R400_ALU_SIZE1_MSB_MASK          (0x7 << 15)
   2192 #   define R400_ALU_START2_MSB_SHIFT        18
   2193 #   define R400_ALU_START2_MSB_MASK         (0x7 << 18)
   2194 #   define R400_ALU_SIZE2_MSB_SHIFT         21
   2195 #   define R400_ALU_SIZE2_MSB_MASK          (0x7 << 21)
   2196 #   define R400_ALU_START3_MSB_SHIFT        24
   2197 #   define R400_ALU_START3_MSB_MASK         (0x7 << 24)
   2198 #   define R400_ALU_SIZE3_MSB_SHIFT         27
   2199 #   define R400_ALU_SIZE3_MSB_MASK          (0x7 << 27)
   2200 /* END: R4xx extended fragment shader registers. */
   2201 
   2202 /* Fog: Fog Blending Enable */
   2203 #define R300_FG_FOG_BLEND                             0x4bc0
   2204 #       define R300_FG_FOG_BLEND_DISABLE              (0 << 0)
   2205 #       define R300_FG_FOG_BLEND_ENABLE               (1 << 0)
   2206 #	define R300_FG_FOG_BLEND_FN_LINEAR            (0 << 1)
   2207 #	define R300_FG_FOG_BLEND_FN_EXP               (1 << 1)
   2208 #	define R300_FG_FOG_BLEND_FN_EXP2              (2 << 1)
   2209 #	define R300_FG_FOG_BLEND_FN_CONSTANT          (3 << 1)
   2210 #	define R300_FG_FOG_BLEND_FN_MASK              (3 << 1)
   2211 
   2212 /* Fog: Red Component of Fog Color */
   2213 #define R300_FG_FOG_COLOR_R                           0x4bc8
   2214 /* Fog: Green Component of Fog Color */
   2215 #define R300_FG_FOG_COLOR_G                           0x4bcc
   2216 /* Fog: Blue Component of Fog Color */
   2217 #define R300_FG_FOG_COLOR_B                           0x4bd0
   2218 #	define R300_FG_FOG_COLOR_MASK 0x000003ff
   2219 
   2220 /* Fog: Constant Factor for Fog Blending */
   2221 #define R300_FG_FOG_FACTOR                            0x4bc4
   2222 #	define FG_FOG_FACTOR_MASK 0x000003ff
   2223 
   2224 /* Fog: Alpha function */
   2225 #define R300_FG_ALPHA_FUNC                            0x4bd4
   2226 #       define R300_FG_ALPHA_FUNC_VAL_MASK               0x000000ff
   2227 #       define R300_FG_ALPHA_FUNC_NEVER                     (0 << 8)
   2228 #       define R300_FG_ALPHA_FUNC_LESS                      (1 << 8)
   2229 #       define R300_FG_ALPHA_FUNC_EQUAL                     (2 << 8)
   2230 #       define R300_FG_ALPHA_FUNC_LE                        (3 << 8)
   2231 #       define R300_FG_ALPHA_FUNC_GREATER                   (4 << 8)
   2232 #       define R300_FG_ALPHA_FUNC_NOTEQUAL                  (5 << 8)
   2233 #       define R300_FG_ALPHA_FUNC_GE                        (6 << 8)
   2234 #       define R300_FG_ALPHA_FUNC_ALWAYS                    (7 << 8)
   2235 #       define R300_ALPHA_TEST_OP_MASK                      (7 << 8)
   2236 #       define R300_FG_ALPHA_FUNC_DISABLE                   (0 << 11)
   2237 #       define R300_FG_ALPHA_FUNC_ENABLE                    (1 << 11)
   2238 
   2239 #       define R500_FG_ALPHA_FUNC_10BIT                     (0 << 12)
   2240 #       define R500_FG_ALPHA_FUNC_8BIT                      (1 << 12)
   2241 
   2242 #       define R300_FG_ALPHA_FUNC_MASK_DISABLE              (0 << 16)
   2243 #       define R300_FG_ALPHA_FUNC_MASK_ENABLE               (1 << 16)
   2244 #       define R300_FG_ALPHA_FUNC_CFG_2_OF_4                (0 << 17)
   2245 #       define R300_FG_ALPHA_FUNC_CFG_3_OF_6                (1 << 17)
   2246 
   2247 #       define R300_FG_ALPHA_FUNC_DITH_DISABLE              (0 << 20)
   2248 #       define R300_FG_ALPHA_FUNC_DITH_ENABLE               (1 << 20)
   2249 
   2250 #       define R500_FG_ALPHA_FUNC_OFFSET_DISABLE            (0 << 24)
   2251 #       define R500_FG_ALPHA_FUNC_OFFSET_ENABLE             (1 << 24) /* Not supported in R520 */
   2252 #       define R500_FG_ALPHA_FUNC_DISC_ZERO_MASK_DISABLE    (0 << 25)
   2253 #       define R500_FG_ALPHA_FUNC_DISC_ZERO_MASK_ENABLE     (1 << 25)
   2254 
   2255 #       define R500_FG_ALPHA_FUNC_FP16_DISABLE              (0 << 28)
   2256 #       define R500_FG_ALPHA_FUNC_FP16_ENABLE               (1 << 28)
   2257 
   2258 
   2259 /* Fog: Where does the depth come from? */
   2260 #define R300_FG_DEPTH_SRC                  0x4bd8
   2261 #	define R300_FG_DEPTH_SRC_SCAN   (0 << 0)
   2262 #	define R300_FG_DEPTH_SRC_SHADER (1 << 0)
   2263 
   2264 /* Fog: Alpha Compare Value */
   2265 #define R500_FG_ALPHA_VALUE                0x4be0
   2266 #	define R500_FG_ALPHA_VALUE_MASK 0x0000ffff
   2267 
   2268 #define RV530_FG_ZBREG_DEST                 0x4be8
   2269 #	define RV530_FG_ZBREG_DEST_PIPE_SELECT_0             (1 << 0)
   2270 #	define RV530_FG_ZBREG_DEST_PIPE_SELECT_1             (1 << 1)
   2271 #	define RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL           (3 << 0)
   2272 /* gap */
   2273 
   2274 /* Fragment program parameters in 7.16 floating point */
   2275 #define R300_PFS_PARAM_0_X                  0x4C00
   2276 #define R300_PFS_PARAM_0_Y                  0x4C04
   2277 #define R300_PFS_PARAM_0_Z                  0x4C08
   2278 #define R300_PFS_PARAM_0_W                  0x4C0C
   2279 /* last consts */
   2280 #define R300_PFS_PARAM_31_X                 0x4DF0
   2281 #define R300_PFS_PARAM_31_Y                 0x4DF4
   2282 #define R300_PFS_PARAM_31_Z                 0x4DF8
   2283 #define R300_PFS_PARAM_31_W                 0x4DFC
   2284 
   2285 /* Unpipelined. */
   2286 #define R300_RB3D_CCTL                      0x4e00
   2287 #	define R300_RB3D_CCTL_NUM_MULTIWRITES(x)       (MAX2(((x)-1), 0) << 5)
   2288 #	define R300_RB3D_CCTL_NUM_MULTIWRITES_1_BUFFER                (0 << 5)
   2289 #	define R300_RB3D_CCTL_NUM_MULTIWRITES_2_BUFFERS               (1 << 5)
   2290 #	define R300_RB3D_CCTL_NUM_MULTIWRITES_3_BUFFERS               (2 << 5)
   2291 #	define R300_RB3D_CCTL_NUM_MULTIWRITES_4_BUFFERS               (3 << 5)
   2292 #	define R300_RB3D_CCTL_CLRCMP_FLIPE_DISABLE                    (0 << 7)
   2293 #	define R300_RB3D_CCTL_CLRCMP_FLIPE_ENABLE                     (1 << 7)
   2294 #	define R300_RB3D_CCTL_AA_COMPRESSION_DISABLE                  (0 << 9)
   2295 #	define R300_RB3D_CCTL_AA_COMPRESSION_ENABLE                   (1 << 9)
   2296 #	define R300_RB3D_CCTL_CMASK_DISABLE                           (0 << 10)
   2297 #	define R300_RB3D_CCTL_CMASK_ENABLE                            (1 << 10)
   2298 /* reserved */
   2299 #	define R300_RB3D_CCTL_INDEPENDENT_COLOR_CHANNEL_MASK_DISABLE  (0 << 12)
   2300 #	define R300_RB3D_CCTL_INDEPENDENT_COLOR_CHANNEL_MASK_ENABLE   (1 << 12)
   2301 #	define R300_RB3D_CCTL_WRITE_COMPRESSION_ENABLE                (0 << 13)
   2302 #	define R300_RB3D_CCTL_WRITE_COMPRESSION_DISABLE               (1 << 13)
   2303 #	define R300_RB3D_CCTL_INDEPENDENT_COLORFORMAT_ENABLE_DISABLE  (0 << 14)
   2304 #	define R300_RB3D_CCTL_INDEPENDENT_COLORFORMAT_ENABLE_ENABLE   (1 << 14)
   2305 
   2306 
   2307 /* Notes:
   2308  * - AFAIK fglrx always sets BLEND_UNKNOWN when blending is used in
   2309  *   the application
   2310  * - AFAIK fglrx always sets BLEND_NO_SEPARATE when CBLEND and ABLEND
   2311  *    are set to the same
   2312  *   function (both registers are always set up completely in any case)
   2313  * - Most blend flags are simply copied from R200 and not tested yet
   2314  */
   2315 #define R300_RB3D_CBLEND                    0x4E04
   2316 #define R300_RB3D_ABLEND                    0x4E08
   2317 /* the following only appear in CBLEND */
   2318 #       define R300_ALPHA_BLEND_ENABLE         (1 << 0)
   2319 #       define R300_SEPARATE_ALPHA_ENABLE      (1 << 1)
   2320 #       define R300_READ_ENABLE                (1 << 2)
   2321 #       define R300_DISCARD_SRC_PIXELS_DIS     (0 << 3)
   2322 #       define R300_DISCARD_SRC_PIXELS_SRC_ALPHA_0     (1 << 3)
   2323 #       define R300_DISCARD_SRC_PIXELS_SRC_COLOR_0     (2 << 3)
   2324 #       define R300_DISCARD_SRC_PIXELS_SRC_ALPHA_COLOR_0     (3 << 3)
   2325 #       define R300_DISCARD_SRC_PIXELS_SRC_ALPHA_1     (4 << 3)
   2326 #       define R300_DISCARD_SRC_PIXELS_SRC_COLOR_1     (5 << 3)
   2327 #       define R300_DISCARD_SRC_PIXELS_SRC_ALPHA_COLOR_1     (6 << 3)
   2328 #       define R500_SRC_ALPHA_0_NO_READ                (1 << 30)
   2329 #       define R500_SRC_ALPHA_1_NO_READ                (1 << 31)
   2330 
   2331 /* the following are shared between CBLEND and ABLEND */
   2332 #       define R300_FCN_MASK                         (3  << 12)
   2333 #       define R300_COMB_FCN_ADD_CLAMP               (0  << 12)
   2334 #       define R300_COMB_FCN_ADD_NOCLAMP             (1  << 12)
   2335 #       define R300_COMB_FCN_SUB_CLAMP               (2  << 12)
   2336 #       define R300_COMB_FCN_SUB_NOCLAMP             (3  << 12)
   2337 #       define R300_COMB_FCN_MIN                     (4  << 12)
   2338 #       define R300_COMB_FCN_MAX                     (5  << 12)
   2339 #       define R300_COMB_FCN_RSUB_CLAMP              (6  << 12)
   2340 #       define R300_COMB_FCN_RSUB_NOCLAMP            (7  << 12)
   2341 #       define R300_BLEND_GL_ZERO                    (32)
   2342 #       define R300_BLEND_GL_ONE                     (33)
   2343 #       define R300_BLEND_GL_SRC_COLOR               (34)
   2344 #       define R300_BLEND_GL_ONE_MINUS_SRC_COLOR     (35)
   2345 #       define R300_BLEND_GL_DST_COLOR               (36)
   2346 #       define R300_BLEND_GL_ONE_MINUS_DST_COLOR     (37)
   2347 #       define R300_BLEND_GL_SRC_ALPHA               (38)
   2348 #       define R300_BLEND_GL_ONE_MINUS_SRC_ALPHA     (39)
   2349 #       define R300_BLEND_GL_DST_ALPHA               (40)
   2350 #       define R300_BLEND_GL_ONE_MINUS_DST_ALPHA     (41)
   2351 #       define R300_BLEND_GL_SRC_ALPHA_SATURATE      (42)
   2352 #       define R300_BLEND_GL_CONST_COLOR             (43)
   2353 #       define R300_BLEND_GL_ONE_MINUS_CONST_COLOR   (44)
   2354 #       define R300_BLEND_GL_CONST_ALPHA             (45)
   2355 #       define R300_BLEND_GL_ONE_MINUS_CONST_ALPHA   (46)
   2356 #       define R300_BLEND_MASK                       (63)
   2357 #       define R300_SRC_BLEND_SHIFT                  (16)
   2358 #       define R300_DST_BLEND_SHIFT                  (24)
   2359 
   2360 /* Constant color used by the blender. Pipelined through the blender.
   2361  * Note: For R520, this field is ignored, use RB3D_CONSTANT_COLOR_GB__BLUE,
   2362  * RB3D_CONSTANT_COLOR_GB__GREEN, etc. instead.
   2363  */
   2364 #define R300_RB3D_BLEND_COLOR               0x4E10
   2365 
   2366 
   2367 /* 3D Color Channel Mask. If all the channels used in the current color format
   2368  * are disabled, then the cb will discard all the incoming quads. Pipelined
   2369  * through the blender.
   2370  */
   2371 #define RB3D_COLOR_CHANNEL_MASK                  0x4E0C
   2372 #	define RB3D_COLOR_CHANNEL_MASK_BLUE_MASK0  (1 << 0)
   2373 #	define RB3D_COLOR_CHANNEL_MASK_GREEN_MASK0 (1 << 1)
   2374 #	define RB3D_COLOR_CHANNEL_MASK_RED_MASK0   (1 << 2)
   2375 #	define RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK0 (1 << 3)
   2376 #	define RB3D_COLOR_CHANNEL_MASK_BLUE_MASK1  (1 << 4)
   2377 #	define RB3D_COLOR_CHANNEL_MASK_GREEN_MASK1 (1 << 5)
   2378 #	define RB3D_COLOR_CHANNEL_MASK_RED_MASK1   (1 << 6)
   2379 #	define RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK1 (1 << 7)
   2380 #	define RB3D_COLOR_CHANNEL_MASK_BLUE_MASK2  (1 << 8)
   2381 #	define RB3D_COLOR_CHANNEL_MASK_GREEN_MASK2 (1 << 9)
   2382 #	define RB3D_COLOR_CHANNEL_MASK_RED_MASK2   (1 << 10)
   2383 #	define RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK2 (1 << 11)
   2384 #	define RB3D_COLOR_CHANNEL_MASK_BLUE_MASK3  (1 << 12)
   2385 #	define RB3D_COLOR_CHANNEL_MASK_GREEN_MASK3 (1 << 13)
   2386 #	define RB3D_COLOR_CHANNEL_MASK_RED_MASK3   (1 << 14)
   2387 #	define RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK3 (1 << 15)
   2388 
   2389 /* Clear color that is used when the color mask is set to 00. Unpipelined.
   2390  * Program this register with a 32-bit value in ARGB8888 or ARGB2101010
   2391  * formats, ignoring the fields.
   2392  */
   2393 #define RB3D_COLOR_CLEAR_VALUE                   0x4e14
   2394 
   2395 /* gap */
   2396 
   2397 /* Color Compare Color. Stalls the 2d/3d datapath until it is idle. */
   2398 #define RB3D_CLRCMP_CLR                     0x4e20
   2399 
   2400 /* Color Compare Mask. Stalls the 2d/3d datapath until it is idle. */
   2401 #define RB3D_CLRCMP_MSK                     0x4e24
   2402 
   2403 /* Color Buffer Address Offset of multibuffer 0. Unpipelined. */
   2404 #define R300_RB3D_COLOROFFSET0              0x4E28
   2405 #       define R300_COLOROFFSET_MASK             0xFFFFFFE0
   2406 /* Color Buffer Address Offset of multibuffer 1. Unpipelined. */
   2407 #define R300_RB3D_COLOROFFSET1              0x4E2C
   2408 /* Color Buffer Address Offset of multibuffer 2. Unpipelined. */
   2409 #define R300_RB3D_COLOROFFSET2              0x4E30
   2410 /* Color Buffer Address Offset of multibuffer 3. Unpipelined. */
   2411 #define R300_RB3D_COLOROFFSET3              0x4E34
   2412 
   2413 /* Color buffer format and tiling control for all the multibuffers and the
   2414  * pitch of multibuffer 0 to 3. Unpipelined. The cache must be empty before any
   2415  * of the registers are changed.
   2416  *
   2417  * Bit 16: Larger tiles
   2418  * Bit 17: 4x2 tiles
   2419  * Bit 18: Extremely weird tile like, but some pixels duplicated?
   2420  */
   2421 #define R300_RB3D_COLORPITCH0               0x4E38
   2422 #       define R300_COLORPITCH_MASK              0x00003FFE
   2423 #       define R300_COLOR_TILE_DISABLE            (0 << 16)
   2424 #       define R300_COLOR_TILE_ENABLE             (1 << 16)
   2425 #       define R300_COLOR_TILE(x)                 ((x) << 16)
   2426 #       define R300_COLOR_MICROTILE_DISABLE       (0 << 17)
   2427 #       define R300_COLOR_MICROTILE_ENABLE        (1 << 17)
   2428 #       define R300_COLOR_MICROTILE_ENABLE_SQUARE (2 << 17) /* Only available in 16-bit */
   2429 #       define R300_COLOR_MICROTILE(x)            ((x) << 17)
   2430 #       define R300_COLOR_ENDIAN_NO_SWAP          (0 << 19)
   2431 #       define R300_COLOR_ENDIAN_WORD_SWAP        (1 << 19)
   2432 #       define R300_COLOR_ENDIAN_DWORD_SWAP       (2 << 19)
   2433 #       define R300_COLOR_ENDIAN_HALF_DWORD_SWAP  (3 << 19)
   2434 #	define R500_COLOR_FORMAT_ARGB10101010     (0 << 21)
   2435 #	define R500_COLOR_FORMAT_UV1010           (1 << 21)
   2436 #	define R500_COLOR_FORMAT_CI8              (2 << 21) /* 2D only */
   2437 #	define R300_COLOR_FORMAT_ARGB1555         (3 << 21)
   2438 #       define R300_COLOR_FORMAT_RGB565           (4 << 21)
   2439 #       define R500_COLOR_FORMAT_ARGB2101010      (5 << 21)
   2440 #       define R300_COLOR_FORMAT_ARGB8888         (6 << 21)
   2441 #       define R300_COLOR_FORMAT_ARGB32323232     (7 << 21)
   2442 /* reserved */
   2443 #       define R300_COLOR_FORMAT_I8               (9 << 21)
   2444 #       define R300_COLOR_FORMAT_ARGB16161616     (10 << 21)
   2445 #       define R300_COLOR_FORMAT_VYUY             (11 << 21)
   2446 #       define R300_COLOR_FORMAT_YVYU             (12 << 21)
   2447 #       define R300_COLOR_FORMAT_UV88             (13 << 21)
   2448 #       define R500_COLOR_FORMAT_I10              (14 << 21)
   2449 #       define R300_COLOR_FORMAT_ARGB4444         (15 << 21)
   2450 #define R300_RB3D_COLORPITCH1               0x4E3C
   2451 #define R300_RB3D_COLORPITCH2               0x4E40
   2452 #define R300_RB3D_COLORPITCH3               0x4E44
   2453 
   2454 /* gap */
   2455 
   2456 /* Destination Color Buffer Cache Control/Status. If the cb is in e2 mode, then
   2457  * a flush or free will not occur upon a write to this register, but a sync
   2458  * will be immediately sent if one is requested. If both DC_FLUSH and DC_FREE
   2459  * are zero but DC_FINISH is one, then a sync will be sent immediately -- the
   2460  * cb will not wait for all the previous operations to complete before sending
   2461  * the sync. Unpipelined except when DC_FINISH and DC_FREE are both set to
   2462  * zero.
   2463  *
   2464  * Set to 0A before 3D operations, set to 02 afterwards.
   2465  */
   2466 #define R300_RB3D_DSTCACHE_CTLSTAT               0x4e4c
   2467 #	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_NO_EFFECT         (0 << 0)
   2468 #	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_NO_EFFECT_1       (1 << 0)
   2469 #	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D    (2 << 0)
   2470 #	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D_1  (3 << 0)
   2471 #	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_NO_EFFECT          (0 << 2)
   2472 #	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_NO_EFFECT_1        (1 << 2)
   2473 #	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS       (2 << 2)
   2474 #	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS_1     (3 << 2)
   2475 #	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FINISH_NO_SIGNAL        (0 << 4)
   2476 #	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FINISH_SIGNAL           (1 << 4)
   2477 
   2478 #define R300_RB3D_DITHER_CTL 0x4E50
   2479 #	define R300_RB3D_DITHER_CTL_DITHER_MODE_TRUNCATE         (0 << 0)
   2480 #	define R300_RB3D_DITHER_CTL_DITHER_MODE_ROUND            (1 << 0)
   2481 #	define R300_RB3D_DITHER_CTL_DITHER_MODE_LUT              (2 << 0)
   2482 /* reserved */
   2483 #	define R300_RB3D_DITHER_CTL_ALPHA_DITHER_MODE_TRUNCATE   (0 << 2)
   2484 #	define R300_RB3D_DITHER_CTL_ALPHA_DITHER_MODE_ROUND      (1 << 2)
   2485 #	define R300_RB3D_DITHER_CTL_ALPHA_DITHER_MODE_LUT        (2 << 2)
   2486 /* reserved */
   2487 
   2488 /* Resolve buffer destination address. The cache must be empty before changing
   2489  * this register if the cb is in resolve mode. Unpipelined
   2490  */
   2491 #define R300_RB3D_AARESOLVE_OFFSET        0x4e80
   2492 #	define R300_RB3D_AARESOLVE_OFFSET_SHIFT 5
   2493 #	define R300_RB3D_AARESOLVE_OFFSET_MASK 0xffffffe0 /* At least according to the calculations of Christoph Brill */
   2494 
   2495 /* Resolve Buffer Pitch and Tiling Control. The cache must be empty before
   2496  * changing this register if the cb is in resolve mode. Unpipelined
   2497  */
   2498 #define R300_RB3D_AARESOLVE_PITCH         0x4e84
   2499 #	define R300_RB3D_AARESOLVE_PITCH_SHIFT 1
   2500 #	define R300_RB3D_AARESOLVE_PITCH_MASK  0x00003ffe /* At least according to the calculations of Christoph Brill */
   2501 
   2502 /* Resolve Buffer Control. Unpipelined */
   2503 #define R300_RB3D_AARESOLVE_CTL           0x4e88
   2504 #	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_MODE_NORMAL   (0 << 0)
   2505 #	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_MODE_RESOLVE  (1 << 0)
   2506 #	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_GAMMA_10      (0 << 1)
   2507 #	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_GAMMA_22      (1 << 1)
   2508 #	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_ALPHA_SAMPLE0 (0 << 2)
   2509 #	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_ALPHA_AVERAGE (1 << 2)
   2510 
   2511 
   2512 /* Discard src pixels less than or equal to threshold. */
   2513 #define R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD 0x4ea0
   2514 /* Discard src pixels greater than or equal to threshold. */
   2515 #define R500_RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD 0x4ea4
   2516 #	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_BLUE_SHIFT 0
   2517 #	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_BLUE_MASK 0x000000ff
   2518 #	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_GREEN_SHIFT 8
   2519 #	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_GREEN_MASK 0x0000ff00
   2520 #	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_RED_SHIFT 16
   2521 #	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_RED_MASK 0x00ff0000
   2522 #	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_ALPHA_SHIFT 24
   2523 #	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_ALPHA_MASK 0xff000000
   2524 
   2525 /* 3D ROP Control. Stalls the 2d/3d datapath until it is idle. */
   2526 #define R300_RB3D_ROPCNTL                             0x4e18
   2527 #	define R300_RB3D_ROPCNTL_ROP_ENABLE            0x00000004
   2528 #	define R300_RB3D_ROPCNTL_ROP_MASK              (15 << 8)
   2529 #	define R300_RB3D_ROPCNTL_ROP_SHIFT             8
   2530 
   2531 /* Color Compare Flip. Stalls the 2d/3d datapath until it is idle. */
   2532 #define R300_RB3D_CLRCMP_FLIPE                        0x4e1c
   2533 
   2534 /* Sets the fifo sizes */
   2535 #define R500_RB3D_FIFO_SIZE                           0x4ef4
   2536 #	define R500_RB3D_FIFO_SIZE_OP_FIFO_SIZE_FULL   (0 << 0)
   2537 #	define R500_RB3D_FIFO_SIZE_OP_FIFO_SIZE_HALF   (1 << 0)
   2538 #	define R500_RB3D_FIFO_SIZE_OP_FIFO_SIZE_QUATER (2 << 0)
   2539 #	define R500_RB3D_FIFO_SIZE_OP_FIFO_SIZE_EIGTHS (3 << 0)
   2540 
   2541 /* Constant color used by the blender. Pipelined through the blender. */
   2542 #define R500_RB3D_CONSTANT_COLOR_AR                   0x4ef8
   2543 #	define R500_RB3D_CONSTANT_COLOR_AR_RED_MASK    0x0000ffff
   2544 #	define R500_RB3D_CONSTANT_COLOR_AR_RED_SHIFT   0
   2545 #	define R500_RB3D_CONSTANT_COLOR_AR_ALPHA_MASK  0xffff0000
   2546 #	define R500_RB3D_CONSTANT_COLOR_AR_ALPHA_SHIFT 16
   2547 
   2548 /* Constant color used by the blender. Pipelined through the blender. */
   2549 #define R500_RB3D_CONSTANT_COLOR_GB                   0x4efc
   2550 #	define R500_RB3D_CONSTANT_COLOR_AR_BLUE_MASK   0x0000ffff
   2551 #	define R500_RB3D_CONSTANT_COLOR_AR_BLUE_SHIFT  0
   2552 #	define R500_RB3D_CONSTANT_COLOR_AR_GREEN_MASK  0xffff0000
   2553 #	define R500_RB3D_CONSTANT_COLOR_AR_GREEN_SHIFT 16
   2554 
   2555 /* gap */
   2556 /* There seems to be no "write only" setting, so use Z-test = ALWAYS
   2557  * for this.
   2558  * Bit (1<<8) is the "test" bit. so plain write is 6  - vd
   2559  */
   2560 #define R300_ZB_CNTL                             0x4F00
   2561 #	define R300_STENCIL_ENABLE		 (1 << 0)
   2562 #	define R300_Z_ENABLE		         (1 << 1)
   2563 #	define R300_Z_WRITE_ENABLE		 (1 << 2)
   2564 #	define R300_Z_SIGNED_COMPARE		 (1 << 3)
   2565 #	define R300_STENCIL_FRONT_BACK		 (1 << 4)
   2566 #   define R500_STENCIL_ZSIGNED_MAGNITUDE (1 << 5)
   2567 #   define R500_STENCIL_REFMASK_FRONT_BACK (1 << 6)
   2568 
   2569 #define R300_ZB_ZSTENCILCNTL                   0x4f04
   2570 	/* functions */
   2571 #	define R300_ZS_NEVER			0
   2572 #	define R300_ZS_LESS			1
   2573 #	define R300_ZS_LEQUAL			2
   2574 #	define R300_ZS_EQUAL			3
   2575 #	define R300_ZS_GEQUAL			4
   2576 #	define R300_ZS_GREATER			5
   2577 #	define R300_ZS_NOTEQUAL			6
   2578 #	define R300_ZS_ALWAYS			7
   2579 #       define R300_ZS_MASK                     7
   2580 	/* operations */
   2581 #	define R300_ZS_KEEP			0
   2582 #	define R300_ZS_ZERO			1
   2583 #	define R300_ZS_REPLACE			2
   2584 #	define R300_ZS_INCR			3
   2585 #	define R300_ZS_DECR			4
   2586 #	define R300_ZS_INVERT			5
   2587 #	define R300_ZS_INCR_WRAP		6
   2588 #	define R300_ZS_DECR_WRAP		7
   2589 #	define R300_Z_FUNC_SHIFT		0
   2590 	/* front and back refer to operations done for front
   2591 	   and back faces, i.e. separate stencil function support */
   2592 #	define R300_S_FRONT_FUNC_SHIFT	        3
   2593 #	define R300_S_FRONT_SFAIL_OP_SHIFT	6
   2594 #	define R300_S_FRONT_ZPASS_OP_SHIFT	9
   2595 #	define R300_S_FRONT_ZFAIL_OP_SHIFT      12
   2596 #	define R300_S_BACK_FUNC_SHIFT           15
   2597 #	define R300_S_BACK_SFAIL_OP_SHIFT       18
   2598 #	define R300_S_BACK_ZPASS_OP_SHIFT       21
   2599 #	define R300_S_BACK_ZFAIL_OP_SHIFT       24
   2600 
   2601 #define R300_ZB_STENCILREFMASK                        0x4f08
   2602 #	define R300_STENCILREF_SHIFT       0
   2603 #	define R300_STENCILREF_MASK        0x000000ff
   2604 #	define R300_STENCILMASK_SHIFT      8
   2605 #	define R300_STENCILMASK_MASK       0x0000ff00
   2606 #	define R300_STENCILWRITEMASK_SHIFT 16
   2607 #	define R300_STENCILWRITEMASK_MASK  0x00ff0000
   2608 
   2609 /* gap */
   2610 
   2611 #define R300_ZB_FORMAT                             0x4f10
   2612 #	define R300_DEPTHFORMAT_16BIT_INT_Z   (0 << 0)
   2613 #	define R300_DEPTHFORMAT_16BIT_13E3    (1 << 0)
   2614 #	define R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL   (2 << 0)
   2615 /* reserved up to (15 << 0) */
   2616 #	define R300_INVERT_13E3_LEADING_ONES  (0 << 4)
   2617 #	define R300_INVERT_13E3_LEADING_ZEROS (1 << 4)
   2618 
   2619 #define R300_ZB_ZTOP                             0x4F14
   2620 #	define R300_ZTOP_DISABLE                 (0 << 0)
   2621 #	define R300_ZTOP_ENABLE                  (1 << 0)
   2622 
   2623 /* gap */
   2624 
   2625 #define R300_ZB_ZCACHE_CTLSTAT            0x4f18
   2626 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_NO_EFFECT      (0 << 0)
   2627 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE (1 << 0)
   2628 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_NO_EFFECT       (0 << 1)
   2629 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE            (1 << 1)
   2630 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_IDLE            (0 << 31)
   2631 #       define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_BUSY            (1 << 31)
   2632 
   2633 #define R300_ZB_BW_CNTL                     0x4f1c
   2634 #	define R300_HIZ_DISABLE                              (0 << 0)
   2635 #	define R300_HIZ_ENABLE                               (1 << 0)
   2636 #	define R300_HIZ_MAX                                  (0 << 1)
   2637 #	define R300_HIZ_MIN                                  (1 << 1)
   2638 #	define R300_FAST_FILL_DISABLE                        (0 << 2)
   2639 #	define R300_FAST_FILL_ENABLE                         (1 << 2)
   2640 #	define R300_RD_COMP_DISABLE                          (0 << 3)
   2641 #	define R300_RD_COMP_ENABLE                           (1 << 3)
   2642 #	define