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 "r600_tex.h"
  51
  52 #include "r700_vertprog.h"
  53 #include "r700_fragprog.h"
  54 #include "r700_state.h"
  55
  56 void r700WaitForIdle(context_t *context);
  57 void r700WaitForIdleClean(context_t *context);
  58 void r700Start3D(context_t *context);
  59 GLboolean r700SendTextureState(context_t *context);
  60 static unsigned int r700PrimitiveType(int prim);
  61 void r600UpdateTextureState(GLcontext * ctx);
  62 GLboolean r700SyncSurf(context_t *context,
  63                        struct radeon_bo *pbo,
  64                        uint32_t read_domain,
  65                        uint32_t write_domain,
  66                        uint32_t sync_type);
  67
  68 void r700WaitForIdle(context_t *context)
  69 {
  70     BATCH_LOCALS(&context->radeon);
  71     BEGIN_BATCH_NO_AUTOSTATE(3);
  72
  73     R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
  74     R600_OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX);
  75     R600_OUT_BATCH(WAIT_3D_IDLE_bit);
  76
  77     END_BATCH();
  78     COMMIT_BATCH();
  79 }
  80
  81 void r700WaitForIdleClean(context_t *context)
  82 {
  83     BATCH_LOCALS(&context->radeon);
  84     BEGIN_BATCH_NO_AUTOSTATE(5);
  85
  86     R600_OUT_BATCH(CP_PACKET3(R600_IT_EVENT_WRITE, 0));
  87     R600_OUT_BATCH(CACHE_FLUSH_AND_INV_EVENT);
  88
  89     R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
  90     R600_OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX);
  91     R600_OUT_BATCH(WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit);
  92
  93     END_BATCH();
  94     COMMIT_BATCH();
  95 }
  96
  97 void r700Start3D(context_t *context)
  98 {
  99     BATCH_LOCALS(&context->radeon);
 100     if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
 101     {
 102         BEGIN_BATCH_NO_AUTOSTATE(2);
 103         R600_OUT_BATCH(CP_PACKET3(R600_IT_START_3D_CMDBUF, 0));
 104         R600_OUT_BATCH(0);
 105         END_BATCH();
 106     }
 107
 108     BEGIN_BATCH_NO_AUTOSTATE(3);
 109     R600_OUT_BATCH(CP_PACKET3(R600_IT_CONTEXT_CONTROL, 1));
 110     R600_OUT_BATCH(0x80000000);
 111     R600_OUT_BATCH(0x80000000);
 112     END_BATCH();
 113
 114     COMMIT_BATCH();
 115
 116     r700WaitForIdleClean(context);
 117 }
 118
 119 static GLboolean r700SetupShaders(GLcontext * ctx)
 120 {
 121     context_t *context = R700_CONTEXT(ctx);
 122
 123     R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
 124
 125     GLuint exportCount;
 126
 127     r700->ps.SQ_PGM_RESOURCES_PS.u32All = 0;
 128     r700->vs.SQ_PGM_RESOURCES_VS.u32All = 0;
 129
 130     SETbit(r700->ps.SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
 131     SETbit(r700->vs.SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
 132
 133     r700SetupVertexProgram(ctx);
 134
 135     r700SetupFragmentProgram(ctx);
 136
 137     exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift);
 138     r700->CB_SHADER_CONTROL.u32All = (1 << exportCount) - 1;
 139
 140     r600UpdateTextureState(ctx);
 141
 142     return GL_TRUE;
 143 }
 144
 145 GLboolean r700SyncSurf(context_t *context,
 146                        struct radeon_bo *pbo,
 147                        uint32_t read_domain,
 148                        uint32_t write_domain,
 149                        uint32_t sync_type)
 150 {
 151     BATCH_LOCALS(&context->radeon);
 152     uint32_t cp_coher_size;
 153
 154     if (pbo->size == 0xffffffff)
 155             cp_coher_size = 0xffffffff;
 156     else
 157             cp_coher_size = ((pbo->size + 255) >> 8);
 158
 159     BEGIN_BATCH_NO_AUTOSTATE(5);
 160     R600_OUT_BATCH(CP_PACKET3(R600_IT_SURFACE_SYNC, 3));
 161     R600_OUT_BATCH(sync_type);
 162     R600_OUT_BATCH(cp_coher_size);
 163     R600_OUT_BATCH_RELOC(0,
 164                          pbo,
 165                          0,
 166                          read_domain, write_domain, 0); // ???
 167     R600_OUT_BATCH(10);
 168
 169     END_BATCH();
 170     COMMIT_BATCH();
 171
 172     return GL_TRUE;
 173 }
 174
 175 static unsigned int r700PrimitiveType(int prim)
 176 {
 177     switch (prim & PRIM_MODE_MASK)
 178     {
 179     case GL_POINTS:
 180         return DI_PT_POINTLIST;
 181         break;
 182     case GL_LINES:
 183         return DI_PT_LINELIST;
 184         break;
 185     case GL_LINE_STRIP:
 186         return DI_PT_LINESTRIP;
 187         break;
 188     case GL_LINE_LOOP:
 189         return DI_PT_LINELOOP;
 190         break;
 191     case GL_TRIANGLES:
 192         return DI_PT_TRILIST;
 193         break;
 194     case GL_TRIANGLE_STRIP:
 195         return DI_PT_TRISTRIP;
 196         break;
 197     case GL_TRIANGLE_FAN:
 198         return DI_PT_TRIFAN;
 199         break;
 200     case GL_QUADS:
 201         return DI_PT_QUADLIST;
 202         break;
 203     case GL_QUAD_STRIP:
 204         return DI_PT_QUADSTRIP;
 205         break;
 206     case GL_POLYGON:
 207         return DI_PT_POLYGON;
 208         break;
 209     default:
 210         assert(0);
 211         return -1;
 212         break;
 213     }
 214 }
 215
 216 static void r700RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
 217 {
 218         context_t *context = R700_CONTEXT(ctx);
 219         BATCH_LOCALS(&context->radeon);
 220         int type, i, total_emit;
 221         int num_indices = end - start;
 222         uint32_t vgt_draw_initiator = 0;
 223         uint32_t vgt_index_type     = 0;
 224         uint32_t vgt_primitive_type = 0;
 225         uint32_t vgt_num_indices    = 0;
 226
 227         type = r700PrimitiveType(prim);
 228
 229         if (type < 0 || num_indices <= 0)
 230                 return;
 231
 232         total_emit =   3 /* VGT_PRIMITIVE_TYPE */
 233                      + 2 /* VGT_INDEX_TYPE */
 234                      + 2 /* NUM_INSTANCES */
 235                      + num_indices + 3; /* DRAW_INDEX_IMMD */
 236
 237         BEGIN_BATCH_NO_AUTOSTATE(total_emit);
 238         // prim
 239         SETfield(vgt_primitive_type, type,
 240                  VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift, VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask);
 241         R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
 242         R600_OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX);
 243         R600_OUT_BATCH(vgt_primitive_type);
 244
 245         // index type
 246         SETfield(vgt_index_type, DI_INDEX_SIZE_32_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask);
 247         R600_OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0));
 248         R600_OUT_BATCH(vgt_index_type);
 249
 250         // num instances
 251         R600_OUT_BATCH(CP_PACKET3(R600_IT_NUM_INSTANCES, 0));
 252         R600_OUT_BATCH(1);
 253
 254         // draw packet
 255         vgt_num_indices = num_indices;
 256         SETfield(vgt_draw_initiator, DI_SRC_SEL_IMMEDIATE, SOURCE_SELECT_shift, SOURCE_SELECT_mask);
 257         SETfield(vgt_draw_initiator, DI_MAJOR_MODE_0, MAJOR_MODE_shift, MAJOR_MODE_mask);
 258
 259         R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (num_indices + 1)));
 260         R600_OUT_BATCH(vgt_num_indices);
 261         R600_OUT_BATCH(vgt_draw_initiator);
 262
 263         for (i = start; i < end; i++) {
 264             R600_OUT_BATCH(i);
 265         }
 266         END_BATCH();
 267         COMMIT_BATCH();
 268
 269 }
 270
 271 void r700EmitState(GLcontext * ctx)
 272 {
 273         context_t *context = R700_CONTEXT(ctx);
 274
 275         rcommonEnsureCmdBufSpace(&context->radeon,
 276                                  context->radeon.hw.max_state_size, __FUNCTION__);
 277
 278         r700Start3D(context);
 279         r700SendSQConfig(context);
 280         r700SendFSState(context); // FIXME just a place holder for now
 281         r700SendPSState(context);
 282         r700SendVSState(context);
 283         r700SendVSConstants(ctx);
 284         r700SendPSConstants(ctx);
 285
 286         r700SendTextureState(context);
 287         r700SetupStreams(ctx);
 288
 289         r700SendUCPState(context);
 290         r700SendContextStates(context);
 291         r700SendViewportState(context, 0);
 292         r700SendRenderTargetState(context, 0);
 293         r700SendDepthTargetState(context);
 294
 295 }
 296
 297 static GLboolean r700RunRender(GLcontext * ctx,
 298                                            struct tnl_pipeline_stage *stage)
 299 {
 300     context_t *context = R700_CONTEXT(ctx);
 301     unsigned int i;
 302     TNLcontext *tnl = TNL_CONTEXT(ctx);
 303     struct vertex_buffer *vb = &tnl->vb;
 304
 305     r700UpdateShaders(ctx);
 306     r700SetScissor(context);
 307     r700SetupShaders(ctx);
 308
 309     r700SetRenderTarget(context, 0);
 310     r700SetDepthTarget(context);
 311
 312     r700EmitState(ctx);
 313
 314     /* richard test code */
 315     for (i = 0; i < vb->PrimitiveCount; i++) {
 316         GLuint prim = _tnl_translate_prim(&vb->Primitive[i]);
 317         GLuint start = vb->Primitive[i].start;
 318         GLuint end = vb->Primitive[i].start + vb->Primitive[i].count;
 319         r700RunRenderPrimitive(ctx, start, end, prim);
 320     }
 321
 322     /* Flush render op cached for last several quads. */
 323     r700WaitForIdleClean(context);
 324
 325     radeonReleaseArrays(ctx, ~0);
 326
 327     rcommonFlushCmdBuf( &context->radeon, __FUNCTION__ );
 328
 329     return GL_FALSE;
 330 }
 331
 332 static GLboolean r700RunNonTCLRender(GLcontext * ctx,
 333                                      struct tnl_pipeline_stage *stage) /* -------------------- */
 334 {
 335         GLboolean bRet = GL_TRUE;
 336
 337         return bRet;
 338 }
 339
 340 static GLboolean r700RunTCLRender(GLcontext * ctx,  /*----------------------*/
 341                                   struct tnl_pipeline_stage *stage)
 342 {
 343         GLboolean bRet = GL_FALSE;
 344
 345     /* TODO : sw fallback */
 346
 347     /**
 348     * Ensure all enabled and complete textures are uploaded along with any buffers being used.
 349     */
 350     if(!r600ValidateBuffers(ctx))
 351     {
 352         return GL_TRUE;
 353     }
 354
 355     bRet = r700RunRender(ctx, stage);
 356
 357     return bRet;
 358         //GL_FALSE will stop to do other pipe stage in _tnl_run_pipeline
 359     //The render here DOES finish the whole pipe, so GL_FALSE should be returned for success.
 360 }
 361
 362 const struct tnl_pipeline_stage _r700_render_stage = {
 363         "r700 Hardware Rasterization",
 364         NULL,
 365         NULL,
 366         NULL,
 367         NULL,
 368         r700RunNonTCLRender
 369 };
 370
 371 const struct tnl_pipeline_stage _r700_tcl_stage = {
 372         "r700 Hardware Transform, Clipping and Lighting",
 373         NULL,
 374         NULL,
 375         NULL,
 376         NULL,
 377         r700RunTCLRender
 378 };
 379
 380 const struct tnl_pipeline_stage *r700_pipeline[] =
 381 {
 382     &_r700_tcl_stage,
 383     &_tnl_vertex_transform_stage,
 384         &_tnl_normal_transform_stage,
 385         &_tnl_lighting_stage,
 386         &_tnl_fog_coordinate_stage,
 387         &_tnl_texgen_stage,
 388         &_tnl_texture_transform_stage,
 389         &_tnl_vertex_program_stage,
 390
 391     &_r700_render_stage,
 392     &_tnl_render_stage,
 393     0,
 394 };
 395
 396