src/mesa/drivers/dri/r200/r200_span.c

   1 /*
   2 Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
   3
   4 The Weather Channel (TM) funded Tungsten Graphics to develop the
   5 initial release of the Radeon 8500 driver under the XFree86 license.
   6 This notice must be preserved.
   7
   8 Permission is hereby granted, free of charge, to any person obtaining
   9 a copy of this software and associated documentation files (the
  10 "Software"), to deal in the Software without restriction, including
  11 without limitation the rights to use, copy, modify, merge, publish,
  12 distribute, sublicense, and/or sell copies of the Software, and to
  13 permit persons to whom the Software is furnished to do so, subject to
  14 the following conditions:
  15
  16 The above copyright notice and this permission notice (including the
  17 next paragraph) shall be included in all copies or substantial
  18 portions of the Software.
  19
  20 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  21 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  22 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
  23 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
  24 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
  25 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
  26 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  27
  28 **************************************************************************/
  29
  30 /*
  31  * Authors:
  32  *   Keith Whitwell <keith@tungstengraphics.com>
  33  */
  34
  35 #include "glheader.h"
  36 #include "imports.h"
  37 #include "swrast/swrast.h"
  38 #include "colormac.h"
  39
  40 #include "r200_context.h"
  41 #include "r200_ioctl.h"
  42 #include "r200_state.h"
  43 #include "r200_span.h"
  44 #include "r200_tex.h"
  45
  46 #define DBG 0
  47
  48 /*
  49  * Note that all information needed to access pixels in a renderbuffer
  50  * should be obtained through the gl_renderbuffer parameter, not per-context
  51  * information.
  52  */
  53 #define LOCAL_VARS                                              \
  54    driRenderbuffer *drb = (driRenderbuffer *) rb;               \
  55    const __DRIdrawablePrivate *dPriv = drb->dPriv;              \
  56    const GLuint bottom = dPriv->h - 1;                          \
  57    GLubyte *buf = (GLubyte *) drb->flippedData                  \
  58       + (dPriv->y * drb->flippedPitch + dPriv->x) * drb->cpp;   \
  59    GLuint p;                                                    \
  60    (void) p;
  61
  62 #define LOCAL_DEPTH_VARS                                \
  63    driRenderbuffer *drb = (driRenderbuffer *) rb;       \
  64    const __DRIdrawablePrivate *dPriv = drb->dPriv;      \
  65    const GLuint bottom = dPriv->h - 1;                  \
  66    GLuint xo = dPriv->x;                                \
  67    GLuint yo = dPriv->y;                                \
  68    GLubyte *buf = (GLubyte *) drb->Base.Data;
  69
  70 #define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
  71
  72 #define Y_FLIP(Y) (bottom - (Y))
  73
  74 #define HW_LOCK()
  75
  76 #define HW_UNLOCK()
  77
  78
  79
  80 /* ================================================================
  81  * Color buffer
  82  */
  83
  84 /* 16 bit, RGB565 color spanline and pixel functions
  85  */
  86 #define SPANTMP_PIXEL_FMT GL_RGB
  87 #define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
  88
  89 #define TAG(x)    r200##x##_RGB565
  90 #define TAG2(x,y) r200##x##_RGB565##y
  91 #define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 2)
  92 #include "spantmp2.h"
  93
  94 /* 32 bit, ARGB8888 color spanline and pixel functions
  95  */
  96 #define SPANTMP_PIXEL_FMT GL_BGRA
  97 #define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
  98
  99 #define TAG(x)    r200##x##_ARGB8888
 100 #define TAG2(x,y) r200##x##_ARGB8888##y
 101 #define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 4)
 102 #include "spantmp2.h"
 103
 104
 105 /* ================================================================
 106  * Depth buffer
 107  */
 108
 109 /* The Radeon family has depth tiling on all the time, so we have to convert
 110  * the x,y coordinates into the memory bus address (mba) in the same
 111  * manner as the engine.  In each case, the linear block address (ba)
 112  * is calculated, and then wired with x and y to produce the final
 113  * memory address.
 114  * The chip will do address translation on its own if the surface registers
 115  * are set up correctly. It is not quite enough to get it working with hyperz too...
 116  */
 117
 118 /* extract bit 'b' of x, result is zero or one */
 119 #define BIT(x,b) ((x & (1<<b))>>b)
 120
 121 static GLuint
 122 r200_mba_z32( driRenderbuffer *drb, GLint x, GLint y )
 123 {
 124    GLuint pitch = drb->pitch;
 125    if (drb->depthHasSurface) {
 126       return 4 * (x + y * pitch);
 127    }
 128    else {
 129       GLuint b = ((y & 0x7FF) >> 4) * ((pitch & 0xFFF) >> 5) + ((x & 0x7FF) >> 5);
 130       GLuint a =
 131          (BIT(x,0) << 2) |
 132          (BIT(y,0) << 3) |
 133          (BIT(x,1) << 4) |
 134          (BIT(y,1) << 5) |
 135          (BIT(x,3) << 6) |
 136          (BIT(x,4) << 7) |
 137          (BIT(x,2) << 8) |
 138          (BIT(y,2) << 9) |
 139          (BIT(y,3) << 10) |
 140          (((pitch & 0x20) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) |
 141          ((b >> 1) << 12);
 142       return a;
 143    }
 144 }
 145
 146 static GLuint
 147 r200_mba_z16( driRenderbuffer *drb, GLint x, GLint y )
 148 {
 149    GLuint pitch = drb->pitch;
 150    if (drb->depthHasSurface) {
 151       return 2 * (x + y * pitch);
 152    }
 153    else {
 154       GLuint b = ((y & 0x7FF) >> 4) * ((pitch & 0xFFF) >> 6) + ((x & 0x7FF) >> 6);
 155       GLuint a =
 156          (BIT(x,0) << 1) |
 157          (BIT(y,0) << 2) |
 158          (BIT(x,1) << 3) |
 159          (BIT(y,1) << 4) |
 160          (BIT(x,2) << 5) |
 161          (BIT(x,4) << 6) |
 162          (BIT(x,5) << 7) |
 163          (BIT(x,3) << 8) |
 164          (BIT(y,2) << 9) |
 165          (BIT(y,3) << 10) |
 166          (((pitch & 0x40) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) |
 167          ((b >> 1) << 12);
 168       return a;
 169    }
 170 }
 171
 172
 173 /* 16-bit depth buffer functions
 174  */
 175
 176 #define WRITE_DEPTH( _x, _y, d )                                        \
 177    *(GLushort *)(buf + r200_mba_z16( drb, _x + xo, _y + yo )) = d;
 178
 179 #define READ_DEPTH( d, _x, _y )                                         \
 180    d = *(GLushort *)(buf + r200_mba_z16( drb, _x + xo, _y + yo ));
 181
 182 #define TAG(x) r200##x##_z16
 183 #include "depthtmp.h"
 184
 185
 186 /* 24 bit depth, 8 bit stencil depthbuffer functions
 187  */
 188
 189 #define WRITE_DEPTH( _x, _y, d )                                        \
 190 do {                                                                    \
 191    GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo );               \
 192    GLuint tmp = *(GLuint *)(buf + offset);                              \
 193    tmp &= 0xff000000;                                                   \
 194    tmp |= ((d) & 0x00ffffff);                                           \
 195    *(GLuint *)(buf + offset) = tmp;                                     \
 196 } while (0)
 197
 198 #define READ_DEPTH( d, _x, _y )                                         \
 199    d = *(GLuint *)(buf + r200_mba_z32( drb, _x + xo,                    \
 200                                          _y + yo )) & 0x00ffffff;
 201
 202 #define TAG(x) r200##x##_z24_s8
 203 #include "depthtmp.h"
 204
 205
 206 /* ================================================================
 207  * Stencil buffer
 208  */
 209
 210 /* 24 bit depth, 8 bit stencil depthbuffer functions
 211  */
 212 #define WRITE_STENCIL( _x, _y, d )                                      \
 213 do {                                                                    \
 214    GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo );               \
 215    GLuint tmp = *(GLuint *)(buf + offset);                              \
 216    tmp &= 0x00ffffff;                                                   \
 217    tmp |= (((d) & 0xff) << 24);                                         \
 218    *(GLuint *)(buf + offset) = tmp;                                     \
 219 } while (0)
 220
 221 #define READ_STENCIL( d, _x, _y )                                       \
 222 do {                                                                    \
 223    GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo );               \
 224    GLuint tmp = *(GLuint *)(buf + offset);                              \
 225    tmp &= 0xff000000;                                                   \
 226    d = tmp >> 24;                                                       \
 227 } while (0)
 228
 229 #define TAG(x) r200##x##_z24_s8
 230 #include "stenciltmp.h"
 231
 232
 233 /* Move locking out to get reasonable span performance (10x better
 234  * than doing this in HW_LOCK above).  WaitForIdle() is the main
 235  * culprit.
 236  */
 237
 238 static void r200SpanRenderStart( GLcontext *ctx )
 239 {
 240    r200ContextPtr rmesa = R200_CONTEXT( ctx );
 241
 242    R200_FIREVERTICES( rmesa );
 243    LOCK_HARDWARE( rmesa );
 244    r200WaitForIdleLocked( rmesa );
 245
 246    /* Read & rewrite the first pixel in the frame buffer.  This should
 247     * be a noop, right?  In fact without this conform fails as reading
 248     * from the framebuffer sometimes produces old results -- the
 249     * on-card read cache gets mixed up and doesn't notice that the
 250     * framebuffer has been updated.
 251     *
 252     * In the worst case this is buggy too as p might get the wrong
 253     * value first time, so really need a hidden pixel somewhere for this.
 254     */
 255    {
 256       int p;
 257       driRenderbuffer *drb =
 258          (driRenderbuffer *) ctx->WinSysDrawBuffer->_ColorDrawBuffers[0][0];
 259       volatile int *buf =
 260          (volatile int *)(rmesa->dri.screen->pFB + drb->offset);
 261       p = *buf;
 262       *buf = p;
 263    }
 264 }
 265
 266 static void r200SpanRenderFinish( GLcontext *ctx )
 267 {
 268    r200ContextPtr rmesa = R200_CONTEXT( ctx );
 269    _swrast_flush( ctx );
 270    UNLOCK_HARDWARE( rmesa );
 271 }
 272
 273 void r200InitSpanFuncs( GLcontext *ctx )
 274 {
 275    struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
 276    swdd->SpanRenderStart          = r200SpanRenderStart;
 277    swdd->SpanRenderFinish         = r200SpanRenderFinish;
 278 }
 279
 280
 281
 282 /**
 283  * Plug in the Get/Put routines for the given driRenderbuffer.
 284  */
 285 void
 286 radeonSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis)
 287 {
 288    if (drb->Base.InternalFormat == GL_RGBA) {
 289       if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) {
 290          r200InitPointers_RGB565(&drb->Base);
 291       }
 292       else {
 293          r200InitPointers_ARGB8888(&drb->Base);
 294       }
 295    }
 296    else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) {
 297       r200InitDepthPointers_z16(&drb->Base);
 298    }
 299    else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) {
 300       r200InitDepthPointers_z24_s8(&drb->Base);
 301    }
 302    else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
 303       r200InitStencilPointers_z24_s8(&drb->Base);
 304    }
 305 }