r300: Fix R300_CMD_SCRATCH on big endian.
[mesa.git] / src / mesa / drivers / dri / radeon / radeon_common_context.h
1
2 #ifndef COMMON_CONTEXT_H
3 #define COMMON_CONTEXT_H
4
5 #include "main/mm.h"
6 #include "math/m_vector.h"
7 #include "texmem.h"
8 #include "tnl/t_context.h"
9 #include "main/colormac.h"
10
11 #include "radeon_screen.h"
12 #include "radeon_drm.h"
13 #include "dri_util.h"
14 #include "tnl/t_vertex.h"
15
16 /* This union is used to avoid warnings/miscompilation
17 with float to uint32_t casts due to strict-aliasing */
18 typedef union { GLfloat f; uint32_t ui32; } float_ui32_type;
19
20 struct radeon_context;
21 typedef struct radeon_context radeonContextRec;
22 typedef struct radeon_context *radeonContextPtr;
23
24
25 #define TEX_0 0x1
26 #define TEX_1 0x2
27 #define TEX_2 0x4
28 #define TEX_3 0x8
29 #define TEX_4 0x10
30 #define TEX_5 0x20
31
32 /* Rasterizing fallbacks */
33 /* See correponding strings in r200_swtcl.c */
34 #define RADEON_FALLBACK_TEXTURE 0x0001
35 #define RADEON_FALLBACK_DRAW_BUFFER 0x0002
36 #define RADEON_FALLBACK_STENCIL 0x0004
37 #define RADEON_FALLBACK_RENDER_MODE 0x0008
38 #define RADEON_FALLBACK_BLEND_EQ 0x0010
39 #define RADEON_FALLBACK_BLEND_FUNC 0x0020
40 #define RADEON_FALLBACK_DISABLE 0x0040
41 #define RADEON_FALLBACK_BORDER_MODE 0x0080
42
43 #define R200_FALLBACK_TEXTURE 0x01
44 #define R200_FALLBACK_DRAW_BUFFER 0x02
45 #define R200_FALLBACK_STENCIL 0x04
46 #define R200_FALLBACK_RENDER_MODE 0x08
47 #define R200_FALLBACK_DISABLE 0x10
48 #define R200_FALLBACK_BORDER_MODE 0x20
49
50 #define RADEON_TCL_FALLBACK_RASTER 0x1 /* rasterization */
51 #define RADEON_TCL_FALLBACK_UNFILLED 0x2 /* unfilled tris */
52 #define RADEON_TCL_FALLBACK_LIGHT_TWOSIDE 0x4 /* twoside tris */
53 #define RADEON_TCL_FALLBACK_MATERIAL 0x8 /* material in vb */
54 #define RADEON_TCL_FALLBACK_TEXGEN_0 0x10 /* texgen, unit 0 */
55 #define RADEON_TCL_FALLBACK_TEXGEN_1 0x20 /* texgen, unit 1 */
56 #define RADEON_TCL_FALLBACK_TEXGEN_2 0x40 /* texgen, unit 2 */
57 #define RADEON_TCL_FALLBACK_TCL_DISABLE 0x80 /* user disable */
58 #define RADEON_TCL_FALLBACK_FOGCOORDSPEC 0x100 /* fogcoord, sep. spec light */
59
60 /* The blit width for texture uploads
61 */
62 #define BLIT_WIDTH_BYTES 1024
63
64 /* Use the templated vertex format:
65 */
66 #define COLOR_IS_RGBA
67 #define TAG(x) radeon##x
68 #include "tnl_dd/t_dd_vertex.h"
69 #undef TAG
70
71 struct radeon_renderbuffer
72 {
73 struct gl_renderbuffer base;
74 struct radeon_bo *bo;
75 unsigned int cpp;
76 /* unsigned int offset; */
77 unsigned int pitch;
78 unsigned int width;
79 unsigned int height;
80
81 /* boo Xorg 6.8.2 compat */
82 int has_surface;
83
84 __DRIdrawablePrivate *dPriv;
85 };
86
87 struct radeon_colorbuffer_state {
88 GLuint clear;
89 int roundEnable;
90 struct radeon_renderbuffer *rrb;
91 };
92
93 struct radeon_depthbuffer_state {
94 GLuint clear;
95 GLfloat scale;
96 struct radeon_renderbuffer *rrb;
97 };
98
99 struct radeon_scissor_state {
100 drm_clip_rect_t rect;
101 GLboolean enabled;
102
103 GLuint numClipRects; /* Cliprects active */
104 GLuint numAllocedClipRects; /* Cliprects available */
105 drm_clip_rect_t *pClipRects;
106 };
107
108 struct radeon_stencilbuffer_state {
109 GLboolean hwBuffer;
110 GLuint clear; /* rb3d_stencilrefmask value */
111 };
112
113 struct radeon_stipple_state {
114 GLuint mask[32];
115 };
116
117 struct radeon_state_atom {
118 struct radeon_state_atom *next, *prev;
119 const char *name; /* for debug */
120 int cmd_size; /* size in bytes */
121 GLuint idx;
122 GLuint is_tcl;
123 GLuint *cmd; /* one or more cmd's */
124 GLuint *lastcmd; /* one or more cmd's */
125 GLboolean dirty; /* dirty-mark in emit_state_list */
126 int (*check) (GLcontext *, struct radeon_state_atom *atom); /* is this state active? */
127 void (*emit) (GLcontext *, struct radeon_state_atom *atom);
128 };
129
130 struct radeon_hw_state {
131 /* Head of the linked list of state atoms. */
132 struct radeon_state_atom atomlist;
133 int max_state_size; /* Number of bytes necessary for a full state emit. */
134 GLboolean is_dirty, all_dirty;
135 };
136
137
138 /* Texture related */
139 typedef struct _radeon_texture_image radeon_texture_image;
140
141 struct _radeon_texture_image {
142 struct gl_texture_image base;
143
144 /**
145 * If mt != 0, the image is stored in hardware format in the
146 * given mipmap tree. In this case, base.Data may point into the
147 * mapping of the buffer object that contains the mipmap tree.
148 *
149 * If mt == 0, the image is stored in normal memory pointed to
150 * by base.Data.
151 */
152 struct _radeon_mipmap_tree *mt;
153 struct radeon_bo *bo;
154
155 int mtlevel; /** if mt != 0, this is the image's level in the mipmap tree */
156 int mtface; /** if mt != 0, this is the image's face in the mipmap tree */
157 };
158
159
160 static INLINE radeon_texture_image *get_radeon_texture_image(struct gl_texture_image *image)
161 {
162 return (radeon_texture_image*)image;
163 }
164
165
166 typedef struct radeon_tex_obj radeonTexObj, *radeonTexObjPtr;
167
168 #define RADEON_TXO_MICRO_TILE (1 << 3)
169
170 /* Texture object in locally shared texture space.
171 */
172 struct radeon_tex_obj {
173 // driTextureObject base;
174 struct gl_texture_object base;
175 struct _radeon_mipmap_tree *mt;
176
177 /**
178 * This is true if we've verified that the mipmap tree above is complete
179 * and so on.
180 */
181 GLboolean validated;
182
183 GLuint override_offset;
184 GLboolean image_override; /* Image overridden by GLX_EXT_tfp */
185 GLuint tile_bits; /* hw texture tile bits used on this texture */
186 struct radeon_bo *bo;
187
188 GLuint bufAddr; /* Offset to start of locally
189 shared texture block */
190
191 GLuint dirty_state; /* Flags (1 per texunit) for
192 whether or not this texobj
193 has dirty hardware state
194 (pp_*) that needs to be
195 brought into the
196 texunit. */
197
198 drm_radeon_tex_image_t image[6][RADEON_MAX_TEXTURE_LEVELS];
199 /* Six, for the cube faces */
200
201
202
203 GLuint pp_txfilter; /* hardware register values */
204 GLuint pp_txformat;
205 GLuint pp_txformat_x;
206 GLuint pp_txoffset; /* Image location in texmem.
207 All cube faces follow. */
208 GLuint pp_txsize; /* npot only */
209 GLuint pp_txpitch; /* npot only */
210 GLuint pp_border_color;
211 GLuint pp_cubic_faces; /* cube face 1,2,3,4 log2 sizes */
212
213 GLuint pp_txfilter_1; /* r300 */
214
215 GLboolean border_fallback;
216
217
218 };
219
220 static INLINE radeonTexObj* radeon_tex_obj(struct gl_texture_object *texObj)
221 {
222 return (radeonTexObj*)texObj;
223 }
224
225 /* Need refcounting on dma buffers:
226 */
227 struct radeon_dma_buffer {
228 int refcount; /* the number of retained regions in buf */
229 drmBufPtr buf;
230 };
231
232 struct radeon_aos {
233 struct radeon_bo *bo; /** Buffer object where vertex data is stored */
234 int offset; /** Offset into buffer object, in bytes */
235 int components; /** Number of components per vertex */
236 int stride; /** Stride in dwords (may be 0 for repeating) */
237 int count; /** Number of vertices */
238 };
239
240 struct radeon_dma {
241 /* Active dma region. Allocations for vertices and retained
242 * regions come from here. Also used for emitting random vertices,
243 * these may be flushed by calling flush_current();
244 */
245 struct radeon_bo *current; /** Buffer that DMA memory is allocated from */
246 int current_used; /** Number of bytes allocated and forgotten about */
247 int current_vertexptr; /** End of active vertex region */
248
249 /**
250 * If current_vertexptr != current_used then flush must be non-zero.
251 * flush must be called before non-active vertex allocations can be
252 * performed.
253 */
254 void (*flush) (GLcontext *);
255
256 /* Number of "in-flight" DMA buffers, i.e. the number of buffers
257 * for which a DISCARD command is currently queued in the command buffer
258 .
259 */
260 GLuint nr_released_bufs;
261 };
262
263 /* radeon_swtcl.c
264 */
265 struct radeon_swtcl_info {
266
267 GLuint RenderIndex;
268 GLuint vertex_size;
269 GLubyte *verts;
270
271 /* Fallback rasterization functions
272 */
273 GLuint hw_primitive;
274 GLenum render_primitive;
275 GLuint numverts;
276
277 struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
278 GLuint vertex_attr_count;
279
280 };
281
282 struct radeon_ioctl {
283 GLuint vertex_offset;
284 struct radeon_bo *bo;
285 GLuint vertex_size;
286 };
287
288 #define RADEON_MAX_PRIMS 64
289
290 struct radeon_prim {
291 GLuint start;
292 GLuint end;
293 GLuint prim;
294 };
295
296 static INLINE GLuint radeonPackColor(GLuint cpp,
297 GLubyte r, GLubyte g,
298 GLubyte b, GLubyte a)
299 {
300 switch (cpp) {
301 case 2:
302 return PACK_COLOR_565(r, g, b);
303 case 4:
304 return PACK_COLOR_8888(a, r, g, b);
305 default:
306 return 0;
307 }
308 }
309
310 #define MAX_CMD_BUF_SZ (16*1024)
311
312 #define MAX_DMA_BUF_SZ (64*1024)
313
314 struct radeon_store {
315 GLuint statenr;
316 GLuint primnr;
317 char cmd_buf[MAX_CMD_BUF_SZ];
318 int cmd_used;
319 int elts_start;
320 };
321
322 struct radeon_dri_mirror {
323 __DRIcontextPrivate *context; /* DRI context */
324 __DRIscreenPrivate *screen; /* DRI screen */
325
326 /**
327 * DRI drawable bound to this context for drawing.
328 */
329 __DRIdrawablePrivate *drawable;
330
331 /**
332 * DRI drawable bound to this context for reading.
333 */
334 __DRIdrawablePrivate *readable;
335
336 drm_context_t hwContext;
337 drm_hw_lock_t *hwLock;
338 int fd;
339 int drmMinor;
340 };
341
342 #define DEBUG_TEXTURE 0x001
343 #define DEBUG_STATE 0x002
344 #define DEBUG_IOCTL 0x004
345 #define DEBUG_PRIMS 0x008
346 #define DEBUG_VERTS 0x010
347 #define DEBUG_FALLBACKS 0x020
348 #define DEBUG_VFMT 0x040
349 #define DEBUG_CODEGEN 0x080
350 #define DEBUG_VERBOSE 0x100
351 #define DEBUG_DRI 0x200
352 #define DEBUG_DMA 0x400
353 #define DEBUG_SANITY 0x800
354 #define DEBUG_SYNC 0x1000
355 #define DEBUG_PIXEL 0x2000
356 #define DEBUG_MEMORY 0x4000
357
358
359
360 typedef void (*radeon_tri_func) (radeonContextPtr,
361 radeonVertex *,
362 radeonVertex *, radeonVertex *);
363
364 typedef void (*radeon_line_func) (radeonContextPtr,
365 radeonVertex *, radeonVertex *);
366
367 typedef void (*radeon_point_func) (radeonContextPtr, radeonVertex *);
368
369 struct radeon_state {
370 struct radeon_colorbuffer_state color;
371 struct radeon_depthbuffer_state depth;
372 struct radeon_scissor_state scissor;
373 struct radeon_stencilbuffer_state stencil;
374 };
375
376 /**
377 * This structure holds the command buffer while it is being constructed.
378 *
379 * The first batch of commands in the buffer is always the state that needs
380 * to be re-emitted when the context is lost. This batch can be skipped
381 * otherwise.
382 */
383 struct radeon_cmdbuf {
384 struct radeon_cs_manager *csm;
385 struct radeon_cs *cs;
386 int size; /** # of dwords total */
387 unsigned int flushing:1; /** whether we're currently in FlushCmdBufLocked */
388 };
389
390 struct radeon_context {
391 GLcontext *glCtx;
392 radeonScreenPtr radeonScreen; /* Screen private DRI data */
393
394 /* Texture object bookkeeping
395 */
396 unsigned nr_heaps;
397 driTexHeap * texture_heaps[ RADEON_NR_TEX_HEAPS ];
398 driTextureObject swapped;
399 int texture_depth;
400 float initialMaxAnisotropy;
401
402 struct radeon_dma dma;
403 struct radeon_hw_state hw;
404 /* Rasterization and vertex state:
405 */
406 GLuint TclFallback;
407 GLuint Fallback;
408 GLuint NewGLState;
409 DECLARE_RENDERINPUTS(tnl_index_bitset); /* index of bits for last tnl_install_attrs */
410
411 /* Page flipping */
412 GLuint doPageFlip;
413
414 /* Drawable, cliprect and scissor information */
415 GLuint numClipRects; /* Cliprects for the draw buffer */
416 drm_clip_rect_t *pClipRects;
417 unsigned int lastStamp;
418 GLboolean lost_context;
419 drm_radeon_sarea_t *sarea; /* Private SAREA data */
420
421 /* Mirrors of some DRI state */
422 struct radeon_dri_mirror dri;
423
424 /* Busy waiting */
425 GLuint do_usleeps;
426 GLuint do_irqs;
427 GLuint irqsEmitted;
428 drm_radeon_irq_wait_t iw;
429
430 /* buffer swap */
431 int64_t swap_ust;
432 int64_t swap_missed_ust;
433
434 GLuint swap_count;
435 GLuint swap_missed_count;
436
437 /* Derived state - for r300 only */
438 struct radeon_state state;
439
440 struct radeon_swtcl_info swtcl;
441 /* Configuration cache
442 */
443 driOptionCache optionCache;
444
445 struct radeon_cmdbuf cmdbuf;
446
447 struct {
448 void (*get_lock)(radeonContextPtr radeon);
449 void (*update_viewport_offset)(GLcontext *ctx);
450 void (*update_draw_buffer)(GLcontext *ctx);
451 void (*emit_cs_header)(struct radeon_cs *cs, radeonContextPtr rmesa);
452 void (*swtcl_flush)(GLcontext *ctx, uint32_t offset);
453 void (*pre_emit_atoms)(radeonContextPtr rmesa);
454 void (*pre_emit_state)(radeonContextPtr rmesa);
455 } vtbl;
456 };
457
458 #define RADEON_CONTEXT(glctx) ((radeonContextPtr)(ctx->DriverCtx))
459
460 /**
461 * This function takes a float and packs it into a uint32_t
462 */
463 static INLINE uint32_t radeonPackFloat32(float fl)
464 {
465 union {
466 float fl;
467 uint32_t u;
468 } u;
469
470 u.fl = fl;
471 return u.u;
472 }
473
474 /* This is probably wrong for some values, I need to test this
475 * some more. Range checking would be a good idea also..
476 *
477 * But it works for most things. I'll fix it later if someone
478 * else with a better clue doesn't
479 */
480 static INLINE uint32_t radeonPackFloat24(float f)
481 {
482 float mantissa;
483 int exponent;
484 uint32_t float24 = 0;
485
486 if (f == 0.0)
487 return 0;
488
489 mantissa = frexpf(f, &exponent);
490
491 /* Handle -ve */
492 if (mantissa < 0) {
493 float24 |= (1 << 23);
494 mantissa = mantissa * -1.0;
495 }
496 /* Handle exponent, bias of 63 */
497 exponent += 62;
498 float24 |= (exponent << 16);
499 /* Kill 7 LSB of mantissa */
500 float24 |= (radeonPackFloat32(mantissa) & 0x7FFFFF) >> 7;
501
502 return float24;
503 }
504
505 GLboolean radeonInitContext(radeonContextPtr radeon,
506 struct dd_function_table* functions,
507 const __GLcontextModes * glVisual,
508 __DRIcontextPrivate * driContextPriv,
509 void *sharedContextPrivate);
510
511 void radeonCleanupContext(radeonContextPtr radeon);
512 GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv);
513 void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable);
514 GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv,
515 __DRIdrawablePrivate * driDrawPriv,
516 __DRIdrawablePrivate * driReadPriv);
517
518 /* ================================================================
519 * Debugging:
520 */
521 #define DO_DEBUG 1
522
523 #if DO_DEBUG
524 extern int RADEON_DEBUG;
525 #else
526 #define RADEON_DEBUG 0
527 #endif
528
529 #endif