src/mesa/drivers/dri/r600/r700_render.c

   1 /*
   2  * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice shall be included
  12  * in all copies or substantial portions of the Software.
  13  *
  14  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  15  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  16  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  17  * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
  18  * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  19  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  20  */
  21
  22 /*
  23  * Authors:
  24  *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
  25  *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
  26  */
  27
  28 #include "main/glheader.h"
  29 #include "main/state.h"
  30 #include "main/imports.h"
  31 #include "main/enums.h"
  32 #include "main/macros.h"
  33 #include "main/context.h"
  34 #include "main/dd.h"
  35 #include "main/simple_list.h"
  36 #include "main/api_arrayelt.h"
  37 #include "swrast/swrast.h"
  38 #include "swrast_setup/swrast_setup.h"
  39 #include "vbo/vbo.h"
  40
  41 #include "tnl/tnl.h"
  42 #include "tnl/t_vp_build.h"
  43 #include "tnl/t_context.h"
  44 #include "tnl/t_vertex.h"
  45 #include "tnl/t_pipeline.h"
  46
  47 #include "r600_context.h"
  48 #include "r600_cmdbuf.h"
  49
  50 #include "r700_chip.h"
  51 #include "r700_tex.h"
  52
  53 #include "r700_vertprog.h"
  54 #include "r700_fragprog.h"
  55 #include "r700_state.h"
  56
  57 void r700WaitForIdle(context_t *context)
  58 {
  59     BATCH_LOCALS(&context->radeon);
  60     BEGIN_BATCH_NO_AUTOSTATE(3);
  61
  62     R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
  63     R600_OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX);
  64     R600_OUT_BATCH(WAIT_3D_IDLE_bit);
  65
  66     END_BATCH();
  67     COMMIT_BATCH();
  68 }
  69
  70 void r700WaitForIdleClean(context_t *context)
  71 {
  72     BATCH_LOCALS(&context->radeon);
  73     BEGIN_BATCH_NO_AUTOSTATE(5);
  74
  75     R600_OUT_BATCH(CP_PACKET3(R600_IT_EVENT_WRITE, 0));
  76     R600_OUT_BATCH(CACHE_FLUSH_AND_INV_EVENT);
  77
  78     R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
  79     R600_OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX);
  80     R600_OUT_BATCH(WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit);
  81
  82     END_BATCH();
  83     COMMIT_BATCH();
  84 }
  85
  86 static void r700Start3D(context_t *context)
  87 {
  88     BATCH_LOCALS(&context->radeon);
  89     if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
  90     {
  91         BEGIN_BATCH_NO_AUTOSTATE(2);
  92         R600_OUT_BATCH(CP_PACKET3(R600_IT_START_3D_CMDBUF, 0));
  93         R600_OUT_BATCH(0);
  94         END_BATCH();
  95     }
  96
  97     BEGIN_BATCH_NO_AUTOSTATE(3);
  98     R600_OUT_BATCH(CP_PACKET3(R600_IT_CONTEXT_CONTROL, 1));
  99     R600_OUT_BATCH(0x80000000);
 100     R600_OUT_BATCH(0x80000000);
 101     END_BATCH();
 102
 103     COMMIT_BATCH();
 104
 105     r700WaitForIdleClean(context);
 106 }
 107
 108 static GLboolean r700SetupShaders(GLcontext * ctx)
 109 {
 110     context_t *context = R700_CONTEXT(ctx);
 111
 112     R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj);
 113
 114     GLuint exportCount;
 115
 116         r700->SQ_PGM_RESOURCES_PS.u32All = 0;
 117         r700->SQ_PGM_RESOURCES_VS.u32All = 0;
 118
 119         SETbit(r700->SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
 120     SETbit(r700->SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
 121
 122     r700SetupVertexProgram(ctx);
 123
 124     r700SetupFragmentProgram(ctx);
 125
 126         exportCount = (r700->SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift);
 127     r700->CB_SHADER_CONTROL.u32All = (1 << exportCount) - 1;
 128
 129     return GL_TRUE;
 130 }
 131
 132 GLboolean r700SendTextureState(context_t *context)
 133 {
 134     unsigned int i;
 135
 136     R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj);
 137 #if 0 /* to be enabled */
 138     for(i=0; i<R700_TEXTURE_NUMBERUNITS; i++)
 139     {
 140         if(r700->texture_states.textures[i] != 0)
 141         {
 142             R700_CMDBUF_CHECK_SPACE(9);
 143             R700EP3 (context, IT_SET_RESOURCE, 7);
 144             R700E32 (context, i * 7);
 145             R700E32 (context, r700->texture_states.textures[i]->SQ_TEX_RESOURCE0.u32All);
 146             R700E32 (context, r700->texture_states.textures[i]->SQ_TEX_RESOURCE1.u32All);
 147             R700E32 (context, r700->texture_states.textures[i]->SQ_TEX_RESOURCE2.u32All);
 148             R700E32 (context, r700->texture_states.textures[i]->SQ_TEX_RESOURCE3.u32All);
 149             R700E32 (context, r700->texture_states.textures[i]->SQ_TEX_RESOURCE4.u32All);
 150             R700E32 (context, r700->texture_states.textures[i]->SQ_TEX_RESOURCE5.u32All);
 151             R700E32 (context, r700->texture_states.textures[i]->SQ_TEX_RESOURCE6.u32All);
 152         }
 153
 154         if(r700->texture_states.samplers[i] != 0)
 155         {
 156             R700_CMDBUF_CHECK_SPACE(5);
 157             R700EP3 (context, IT_SET_SAMPLER, 3);
 158             R700E32 (context, i * 3);   // Base at 0x7000
 159             R700E32 (context, r700->texture_states.samplers[i]->SQ_TEX_SAMPLER0.u32All);
 160             R700E32 (context, r700->texture_states.samplers[i]->SQ_TEX_SAMPLER1.u32All);
 161             R700E32 (context, r700->texture_states.samplers[i]->SQ_TEX_SAMPLER2.u32All);
 162         }
 163     }
 164 #endif
 165     return GL_TRUE;
 166 }
 167
 168 GLboolean r700SyncSurf(context_t *context)
 169 {
 170     BATCH_LOCALS(&context->radeon);
 171
 172     /* TODO : too heavy? */
 173     unsigned int CP_COHER_CNTL   = 0;
 174
 175     CP_COHER_CNTL |= (TC_ACTION_ENA_bit
 176                       | VC_ACTION_ENA_bit
 177                       | CB_ACTION_ENA_bit
 178                       | DB_ACTION_ENA_bit
 179                       | SH_ACTION_ENA_bit
 180                       | SMX_ACTION_ENA_bit);
 181
 182
 183     BEGIN_BATCH_NO_AUTOSTATE(5);
 184     R600_OUT_BATCH(CP_PACKET3(R600_IT_SURFACE_SYNC, 3));
 185     R600_OUT_BATCH(CP_COHER_CNTL);
 186     R600_OUT_BATCH(0xFFFFFFFF);
 187     R600_OUT_BATCH(0x00000000);
 188     R600_OUT_BATCH(10);
 189
 190     END_BATCH();
 191     COMMIT_BATCH();
 192
 193     return GL_TRUE;
 194 }
 195
 196 unsigned int r700PrimitiveType(int prim)
 197 {
 198     switch (prim & PRIM_MODE_MASK)
 199     {
 200     case GL_POINTS:
 201         return DI_PT_POINTLIST;
 202         break;
 203     case GL_LINES:
 204         return DI_PT_LINELIST;
 205         break;
 206     case GL_LINE_STRIP:
 207         return DI_PT_LINESTRIP;
 208         break;
 209     case GL_LINE_LOOP:
 210         return DI_PT_LINELOOP;
 211         break;
 212     case GL_TRIANGLES:
 213         return DI_PT_TRILIST;
 214         break;
 215     case GL_TRIANGLE_STRIP:
 216         return DI_PT_TRISTRIP;
 217         break;
 218     case GL_TRIANGLE_FAN:
 219         return DI_PT_TRIFAN;
 220         break;
 221     case GL_QUADS:
 222         return DI_PT_QUADLIST;
 223         break;
 224     case GL_QUAD_STRIP:
 225         return DI_PT_QUADSTRIP;
 226         break;
 227     case GL_POLYGON:
 228         return DI_PT_POLYGON;
 229         break;
 230     default:
 231         assert(0);
 232         return -1;
 233         break;
 234     }
 235 }
 236
 237 static GLboolean r700RunRender(GLcontext * ctx,
 238                                            struct tnl_pipeline_stage *stage)
 239 {
 240     context_t *context = R700_CONTEXT(ctx);
 241     R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj);
 242 #if 1
 243     BATCH_LOCALS(&context->radeon);
 244
 245     unsigned int i, j;
 246     TNLcontext *tnl = TNL_CONTEXT(ctx);
 247     struct vertex_buffer *vb = &tnl->vb;
 248
 249     struct r700_fragment_program *fp = (struct r700_fragment_program *)
 250                                            (ctx->FragmentProgram._Current);
 251     if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
 252     {
 253         fp->r700AsmCode.bR6xx = 1;
 254     }
 255
 256     r700Start3D(context); /* TODO : this is too much. */
 257
 258     r700SyncSurf(context); /* TODO : make it light. */
 259
 260     r700UpdateShaders(ctx);
 261
 262     r700SetRenderTarget(context);
 263
 264     if(r700SetupStreams(ctx))
 265     {
 266         return GL_TRUE;
 267     }
 268
 269     /* flush TX */
 270     //r700SyncSurf(context); /*  */
 271
 272     r700UpdateTextureState(context);
 273     r700SendTextureState(context);
 274
 275     if(GL_FALSE == fp->translated)
 276     {
 277         if( GL_FALSE == r700TranslateFragmentShader(fp, &(fp->mesa_program)) )
 278         {
 279             return GL_TRUE;
 280         }
 281     }
 282
 283     /* flush SQ */
 284     //r700SyncSurf(context); /*  */
 285     //r700SyncSurf(context); /*  */
 286
 287     r700SetupShaders(ctx);
 288
 289     /* set a valid base address to make the command checker happy */
 290     r700->SQ_PGM_START_FS.u32All     = r700->SQ_PGM_START_PS.u32All;
 291     r700->SQ_PGM_START_ES.u32All     = r700->SQ_PGM_START_PS.u32All;
 292     r700->SQ_PGM_START_GS.u32All     = r700->SQ_PGM_START_PS.u32All;
 293
 294     /* flush vtx */
 295     //r700SyncSurf(context); /*  */
 296
 297     r700SendContextStates(context, GL_FALSE);
 298
 299     /* richard test code */
 300     for (i = 0; i < vb->PrimitiveCount; i++)
 301     {
 302         GLuint prim = _tnl_translate_prim(&vb->Primitive[i]);
 303         GLuint start = vb->Primitive[i].start;
 304         GLuint end = vb->Primitive[i].start + vb->Primitive[i].count;
 305         GLuint numIndices = vb->Primitive[i].count;
 306         GLuint numEntires;
 307                 //r300RunRenderPrimitive(rmesa, ctx, start, end, prim);
 308
 309         unsigned int VGT_DRAW_INITIATOR = 0;
 310         unsigned int VGT_INDEX_TYPE     = 0;
 311         unsigned int VGT_PRIMITIVE_TYPE = 0;
 312         unsigned int VGT_NUM_INDICES    = 0;
 313
 314         numEntires = 2 /* VGT_INDEX_TYPE */
 315                      + 3 /* VGT_PRIMITIVE_TYPE */
 316                      + numIndices + 3; /* DRAW_INDEX_IMMD */
 317
 318         BEGIN_BATCH_NO_AUTOSTATE(numEntires);
 319
 320         VGT_INDEX_TYPE |= DI_INDEX_SIZE_32_BIT << INDEX_TYPE_shift;
 321
 322         R600_OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0));
 323         R600_OUT_BATCH(VGT_INDEX_TYPE);
 324
 325         VGT_NUM_INDICES = numIndices;
 326
 327         VGT_PRIMITIVE_TYPE |= r700PrimitiveType(prim) << VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift;
 328         R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
 329         R600_OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX);
 330         R600_OUT_BATCH(VGT_PRIMITIVE_TYPE);
 331
 332         VGT_DRAW_INITIATOR |= DI_SRC_SEL_IMMEDIATE << SOURCE_SELECT_shift;
 333         VGT_DRAW_INITIATOR |= DI_MAJOR_MODE_0 << MAJOR_MODE_shift;
 334
 335         R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (numIndices + 1)));
 336         R600_OUT_BATCH(VGT_NUM_INDICES);
 337         R600_OUT_BATCH(VGT_DRAW_INITIATOR);
 338
 339         for (j=0; j<numIndices; j++)
 340         {
 341             R600_OUT_BATCH(j);
 342         }
 343         END_BATCH();
 344         COMMIT_BATCH();
 345     }
 346
 347     /* Flush render op cached for last several quads. */
 348     r700WaitForIdleClean(context);
 349
 350     /* flush dst */
 351     //r700SyncSurf(context); /*  */
 352
 353     (context->chipobj.FlushCmdBuffer)(context);
 354
 355     (context->chipobj.ReleaseArrays)(ctx);
 356
 357     //richard test
 358     /* test stamp, write a number to mmSCRATCH4 */
 359     BEGIN_BATCH_NO_AUTOSTATE(3);
 360     R600_OUT_BATCH_REGVAL((0x2144 << 2), 0x56785678);
 361     END_BATCH();
 362     COMMIT_BATCH();
 363
 364
 365 #endif //0
 366     rcommonFlushCmdBuf( &context->radeon, __FUNCTION__ );
 367
 368     return GL_FALSE;
 369 }
 370
 371 static GLboolean r700RunNonTCLRender(GLcontext * ctx,
 372                                      struct tnl_pipeline_stage *stage) /* -------------------- */
 373 {
 374         GLboolean bRet = GL_TRUE;
 375
 376         return bRet;
 377 }
 378
 379 static GLboolean r700RunTCLRender(GLcontext * ctx,  /*----------------------*/
 380                                   struct tnl_pipeline_stage *stage)
 381 {
 382         GLboolean bRet = GL_FALSE;
 383
 384     /* TODO : sw fallback */
 385
 386     /**
 387     * Ensure all enabled and complete textures are uploaded along with any buffers being used.
 388     */
 389     if(!r700ValidateBuffers(ctx))
 390     {
 391         return GL_TRUE;
 392     }
 393
 394     context_t *context = R700_CONTEXT(ctx);
 395
 396     r700UpdateShaders(ctx);
 397
 398     bRet = r700RunRender(ctx, stage);
 399
 400     return bRet;
 401         //GL_FALSE will stop to do other pipe stage in _tnl_run_pipeline
 402     //The render here DOES finish the whole pipe, so GL_FALSE should be returned for success.
 403 }
 404
 405 const struct tnl_pipeline_stage _r700_render_stage = {
 406         "r700 Hardware Rasterization",
 407         NULL,
 408         NULL,
 409         NULL,
 410         NULL,
 411         r700RunNonTCLRender
 412 };
 413
 414 const struct tnl_pipeline_stage _r700_tcl_stage = {
 415         "r700 Hardware Transform, Clipping and Lighting",
 416         NULL,
 417         NULL,
 418         NULL,
 419         NULL,
 420         r700RunTCLRender
 421 };
 422
 423 const struct tnl_pipeline_stage *r700_pipeline[] =
 424 {
 425     &_r700_tcl_stage,
 426     &_tnl_vertex_transform_stage,
 427         &_tnl_normal_transform_stage,
 428         &_tnl_lighting_stage,
 429         &_tnl_fog_coordinate_stage,
 430         &_tnl_texgen_stage,
 431         &_tnl_texture_transform_stage,
 432         &_tnl_vertex_program_stage,
 433
 434     &_r700_render_stage,
 435     &_tnl_render_stage,
 436     0,
 437 };
 438
 439