src/gallium/drivers/nv20/nv20_prim_vbuf.c

   1 /**************************************************************************
   2  *
   3  * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
   4  * All Rights Reserved.
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a
   7  * copy of this software and associated documentation files (the
   8  * "Software"), to deal in the Software without restriction, including
   9  * without limitation the rights to use, copy, modify, merge, publish,
  10  * distribute, sub license, and/or sell copies of the Software, and to
  11  * permit persons to whom the Software is furnished to do so, subject to
  12  * the following conditions:
  13  *
  14  * The above copyright notice and this permission notice (including the
  15  * next paragraph) shall be included in all copies or substantial portions
  16  * of the Software.
  17  *
  18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  19  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  20  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
  21  * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
  22  * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
  23  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
  24  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  25  *
  26  **************************************************************************/
  27
  28 /**
  29  * \file
  30  * Build post-transformation, post-clipping vertex buffers and element
  31  * lists by hooking into the end of the primitive pipeline and
  32  * manipulating the vertex_id field in the vertex headers.
  33  *
  34  * XXX: work in progress
  35  *
  36  * \author José Fonseca <jrfonseca@tungstengraphics.com>
  37  * \author Keith Whitwell <keith@tungstengraphics.com>
  38  */
  39
  40
  41 #include "pipe/p_debug.h"
  42 #include "pipe/p_inlines.h"
  43 #include "pipe/internal/p_winsys_screen.h"
  44
  45 #include "nv20_context.h"
  46 #include "nv20_state.h"
  47
  48 #include "draw/draw_vbuf.h"
  49
  50 /**
  51  * Primitive renderer for nv20.
  52  */
  53 struct nv20_vbuf_render {
  54         struct vbuf_render base;
  55
  56         struct nv20_context *nv20;
  57
  58         /** Vertex buffer in VRAM */
  59         struct pipe_buffer *pbuffer;
  60
  61         /** Vertex buffer in normal memory */
  62         void *mbuffer;
  63
  64         /** Vertex size in bytes */
  65         /*unsigned vertex_size;*/
  66
  67         /** Hardware primitive */
  68         unsigned hwprim;
  69 };
  70
  71 /**
  72  * Basically a cast wrapper.
  73  */
  74 static INLINE struct nv20_vbuf_render *
  75 nv20_vbuf_render(struct vbuf_render *render)
  76 {
  77         assert(render);
  78         return (struct nv20_vbuf_render *)render;
  79 }
  80
  81 void nv20_vtxbuf_bind( struct nv20_context* nv20 )
  82 {
  83 #if 0
  84         int i;
  85         for(i = 0; i < NV20TCL_VTXBUF_ADDRESS__SIZE; i++) {
  86                 BEGIN_RING(kelvin, NV20TCL_VTXBUF_ADDRESS(i), 1);
  87                 OUT_RING(0/*nv20->vtxbuf*/);
  88                 BEGIN_RING(kelvin, NV20TCL_VTXFMT(i) ,1);
  89                 OUT_RING(0/*XXX*/);
  90         }
  91 #endif
  92 }
  93
  94 static const struct vertex_info *
  95 nv20_vbuf_render_get_vertex_info( struct vbuf_render *render )
  96 {
  97         struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
  98         struct nv20_context *nv20 = nv20_render->nv20;
  99
 100         nv20_emit_hw_state(nv20);
 101
 102         return &nv20->vertex_info;
 103 }
 104
 105 static void *
 106 nv20__allocate_mbuffer(struct nv20_vbuf_render *nv20_render, size_t size)
 107 {
 108         nv20_render->mbuffer = MALLOC(size);
 109         return nv20_render->mbuffer;
 110 }
 111
 112 static void *
 113 nv20__allocate_pbuffer(struct nv20_vbuf_render *nv20_render, size_t size)
 114 {
 115         struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys;
 116         nv20_render->pbuffer = winsys->buffer_create(winsys, 64,
 117                                         PIPE_BUFFER_USAGE_VERTEX, size);
 118         return winsys->buffer_map(winsys,
 119                         nv20_render->pbuffer,
 120                         PIPE_BUFFER_USAGE_CPU_WRITE);
 121 }
 122
 123 static void *
 124 nv20_vbuf_render_allocate_vertices( struct vbuf_render *render,
 125                 ushort vertex_size,
 126                 ushort nr_vertices )
 127 {
 128         struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
 129         size_t size = (size_t)vertex_size * (size_t)nr_vertices;
 130         void *buf;
 131
 132         assert(!nv20_render->pbuffer);
 133         assert(!nv20_render->mbuffer);
 134
 135         /*
 136          * For small amount of vertices, don't bother with pipe vertex
 137          * buffer, the data will be passed directly via the fifo.
 138          */
 139         /* XXX: Pipe vertex buffers don't work. */
 140         if (0 && size > 16 * 1024)
 141                 buf = nv20__allocate_pbuffer(nv20_render, size);
 142         else
 143                 buf = nv20__allocate_mbuffer(nv20_render, size);
 144
 145         if (buf)
 146                 nv20_render->nv20->dirty |= NV20_NEW_VTXARRAYS;
 147
 148         return buf;
 149 }
 150
 151 static boolean
 152 nv20_vbuf_render_set_primitive( struct vbuf_render *render,
 153                 unsigned prim )
 154 {
 155         struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
 156         unsigned hwp = nvgl_primitive(prim);
 157         if (hwp == 0)
 158                 return FALSE;
 159
 160         nv20_render->hwprim = hwp;
 161         return TRUE;
 162 }
 163
 164 static uint32_t
 165 nv20__vtxhwformat(unsigned stride, unsigned fields, unsigned type)
 166 {
 167         return (stride << NV20TCL_VTXFMT_STRIDE_SHIFT) |
 168                 (fields << NV20TCL_VTXFMT_SIZE_SHIFT) |
 169                 (type << NV20TCL_VTXFMT_TYPE_SHIFT);
 170 }
 171
 172 static unsigned
 173 nv20__emit_format(struct nv20_context *nv20, enum attrib_emit type, int hwattr)
 174 {
 175         uint32_t hwfmt = 0;
 176         unsigned fields;
 177
 178         switch (type) {
 179         case EMIT_OMIT:
 180                 hwfmt = nv20__vtxhwformat(0, 0, 2);
 181                 fields = 0;
 182                 break;
 183         case EMIT_1F:
 184                 hwfmt = nv20__vtxhwformat(4, 1, 2);
 185                 fields = 1;
 186                 break;
 187         case EMIT_2F:
 188                 hwfmt = nv20__vtxhwformat(8, 2, 2);
 189                 fields = 2;
 190                 break;
 191         case EMIT_3F:
 192                 hwfmt = nv20__vtxhwformat(12, 3, 2);
 193                 fields = 3;
 194                 break;
 195         case EMIT_4F:
 196                 hwfmt = nv20__vtxhwformat(16, 4, 2);
 197                 fields = 4;
 198                 break;
 199         default:
 200                 NOUVEAU_ERR("unhandled attrib_emit %d\n", type);
 201                 return 0;
 202         }
 203
 204         BEGIN_RING(kelvin, NV20TCL_VTXFMT(hwattr), 1);
 205         OUT_RING(hwfmt);
 206         return fields;
 207 }
 208
 209 static unsigned
 210 nv20__emit_vertex_array_format(struct nv20_context *nv20)
 211 {
 212         struct vertex_info *vinfo = &nv20->vertex_info;
 213         int hwattr = NV20TCL_VTXFMT__SIZE;
 214         int attr = 0;
 215         unsigned nr_fields = 0;
 216
 217         while (hwattr-- > 0) {
 218                 if (vinfo->hwfmt[0] & (1 << hwattr)) {
 219                         nr_fields += nv20__emit_format(nv20,
 220                                         vinfo->attrib[attr].emit, hwattr);
 221                         attr++;
 222                 } else
 223                         nv20__emit_format(nv20, EMIT_OMIT, hwattr);
 224         }
 225
 226         return nr_fields;
 227 }
 228
 229 static void
 230 nv20__draw_mbuffer(struct nv20_vbuf_render *nv20_render,
 231                 const ushort *indices,
 232                 uint nr_indices)
 233 {
 234         struct nv20_context *nv20 = nv20_render->nv20;
 235         struct vertex_info *vinfo = &nv20->vertex_info;
 236         unsigned nr_fields;
 237         int max_push;
 238         ubyte *data = nv20_render->mbuffer;
 239         int vsz = 4 * vinfo->size;
 240
 241         nr_fields = nv20__emit_vertex_array_format(nv20);
 242
 243         BEGIN_RING(kelvin, NV20TCL_VERTEX_BEGIN_END, 1);
 244         OUT_RING(nv20_render->hwprim);
 245
 246         max_push = 1200 / nr_fields;
 247         while (nr_indices) {
 248                 int i;
 249                 int push = MIN2(nr_indices, max_push);
 250
 251                 BEGIN_RING_NI(kelvin, NV20TCL_VERTEX_DATA, push * nr_fields);
 252                 for (i = 0; i < push; i++) {
 253                         /* XXX: fixme to handle other than floats? */
 254                         int f = nr_fields;
 255                         float *attrv = (float*)&data[indices[i] * vsz];
 256                         while (f-- > 0)
 257                                 OUT_RINGf(*attrv++);
 258                 }
 259
 260                 nr_indices -= push;
 261                 indices += push;
 262         }
 263
 264         BEGIN_RING(kelvin, NV20TCL_VERTEX_BEGIN_END, 1);
 265         OUT_RING(NV20TCL_VERTEX_BEGIN_END_STOP);
 266 }
 267
 268 static void
 269 nv20__draw_pbuffer(struct nv20_vbuf_render *nv20_render,
 270                 const ushort *indices,
 271                 uint nr_indices)
 272 {
 273         struct nv20_context *nv20 = nv20_render->nv20;
 274         int push, i;
 275
 276         NOUVEAU_ERR("nv20__draw_pbuffer: this path is broken.\n");
 277
 278         BEGIN_RING(kelvin, NV10TCL_VERTEX_ARRAY_OFFSET_POS, 1);
 279         OUT_RELOCl(nv20_render->pbuffer, 0,
 280                         NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD);
 281
 282         BEGIN_RING(kelvin, NV10TCL_VERTEX_BUFFER_BEGIN_END, 1);
 283         OUT_RING(nv20_render->hwprim);
 284
 285         if (nr_indices & 1) {
 286                 BEGIN_RING(kelvin, NV10TCL_VB_ELEMENT_U32, 1);
 287                 OUT_RING  (indices[0]);
 288                 indices++; nr_indices--;
 289         }
 290
 291         while (nr_indices) {
 292                 // XXX too big/small ? check the size
 293                 push = MIN2(nr_indices, 1200 * 2);
 294
 295                 BEGIN_RING_NI(kelvin, NV10TCL_VB_ELEMENT_U16, push >> 1);
 296                 for (i = 0; i < push; i+=2)
 297                         OUT_RING((indices[i+1] << 16) | indices[i]);
 298
 299                 nr_indices -= push;
 300                 indices  += push;
 301         }
 302
 303         BEGIN_RING(kelvin, NV10TCL_VERTEX_BUFFER_BEGIN_END, 1);
 304         OUT_RING  (0);
 305 }
 306
 307 static void
 308 nv20_vbuf_render_draw( struct vbuf_render *render,
 309                 const ushort *indices,
 310                 uint nr_indices)
 311 {
 312         struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
 313
 314         nv20_emit_hw_state(nv20_render->nv20);
 315
 316         if (nv20_render->pbuffer)
 317                 nv20__draw_pbuffer(nv20_render, indices, nr_indices);
 318         else if (nv20_render->mbuffer)
 319                 nv20__draw_mbuffer(nv20_render, indices, nr_indices);
 320         else
 321                 assert(0);
 322 }
 323
 324
 325 static void
 326 nv20_vbuf_render_release_vertices( struct vbuf_render *render,
 327                 void *vertices,
 328                 unsigned vertex_size,
 329                 unsigned vertices_used )
 330 {
 331         struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
 332         struct nv20_context *nv20 = nv20_render->nv20;
 333         struct pipe_winsys *winsys = nv20->pipe.winsys;
 334         struct pipe_screen *pscreen = &nv20->screen->pipe;
 335
 336         if (nv20_render->pbuffer) {
 337                 winsys->buffer_unmap(winsys, nv20_render->pbuffer);
 338                 pipe_buffer_reference(pscreen, &nv20_render->pbuffer, NULL);
 339         } else if (nv20_render->mbuffer) {
 340                 FREE(nv20_render->mbuffer);
 341                 nv20_render->mbuffer = NULL;
 342         } else
 343                 assert(0);
 344 }
 345
 346
 347 static void
 348 nv20_vbuf_render_destroy( struct vbuf_render *render )
 349 {
 350         struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
 351
 352         assert(!nv20_render->pbuffer);
 353         assert(!nv20_render->mbuffer);
 354
 355         FREE(nv20_render);
 356 }
 357
 358
 359 /**
 360  * Create a new primitive render.
 361  */
 362 static struct vbuf_render *
 363 nv20_vbuf_render_create( struct nv20_context *nv20 )
 364 {
 365         struct nv20_vbuf_render *nv20_render = CALLOC_STRUCT(nv20_vbuf_render);
 366
 367         nv20_render->nv20 = nv20;
 368
 369         nv20_render->base.max_vertex_buffer_bytes = 16*1024;
 370         nv20_render->base.max_indices = 1024;
 371         nv20_render->base.get_vertex_info = nv20_vbuf_render_get_vertex_info;
 372         nv20_render->base.allocate_vertices =
 373                                         nv20_vbuf_render_allocate_vertices;
 374         nv20_render->base.set_primitive = nv20_vbuf_render_set_primitive;
 375         nv20_render->base.draw = nv20_vbuf_render_draw;
 376         nv20_render->base.release_vertices = nv20_vbuf_render_release_vertices;
 377         nv20_render->base.destroy = nv20_vbuf_render_destroy;
 378
 379         return &nv20_render->base;
 380 }
 381
 382
 383 /**
 384  * Create a new primitive vbuf/render stage.
 385  */
 386 struct draw_stage *nv20_draw_vbuf_stage( struct nv20_context *nv20 )
 387 {
 388         struct vbuf_render *render;
 389         struct draw_stage *stage;
 390
 391         render = nv20_vbuf_render_create(nv20);
 392         if(!render)
 393                 return NULL;
 394
 395         stage = draw_vbuf_stage( nv20->draw, render );
 396         if(!stage) {
 397                 render->destroy(render);
 398                 return NULL;
 399         }
 400
 401         return stage;
 402 }