src/mesa/drivers/dri/r600/r700_vertprog.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  */
  26
  27
  28 #include <stdio.h>
  29 #include <stdarg.h>
  30 #include <stdlib.h>
  31 #include <string.h>
  32 #include <math.h>
  33
  34 #include "main/imports.h"
  35 #include "main/mtypes.h"
  36
  37 #include "tnl/t_context.h"
  38 #include "program/program.h"
  39 #include "program/prog_parameter.h"
  40 #include "program/prog_statevars.h"
  41
  42 #include "radeon_debug.h"
  43 #include "r600_context.h"
  44 #include "r600_cmdbuf.h"
  45 #include "r600_emit.h"
  46 #include "program/programopt.h"
  47
  48 #include "r700_debug.h"
  49 #include "r700_vertprog.h"
  50
  51 unsigned int Map_Vertex_Output(r700_AssemblerBase       *pAsm,
  52                                                    struct gl_vertex_program *mesa_vp,
  53                                                    unsigned int unStart)
  54 {
  55     unsigned int i;
  56         unsigned int unBit;
  57         unsigned int unTotal = unStart;
  58
  59     //!!!!!!! THE ORDER MATCH FS INPUT
  60
  61         unBit = 1 << VERT_RESULT_HPOS;
  62         if(mesa_vp->Base.OutputsWritten & unBit)
  63         {
  64                 pAsm->ucVP_OutputMap[VERT_RESULT_HPOS] = unTotal++;
  65         }
  66
  67         unBit = 1 << VERT_RESULT_COL0;
  68         if(mesa_vp->Base.OutputsWritten & unBit)
  69         {
  70                 pAsm->ucVP_OutputMap[VERT_RESULT_COL0] = unTotal++;
  71         }
  72
  73         unBit = 1 << VERT_RESULT_COL1;
  74         if(mesa_vp->Base.OutputsWritten & unBit)
  75         {
  76                 pAsm->ucVP_OutputMap[VERT_RESULT_COL1] = unTotal++;
  77         }
  78
  79         //TODO : dealing back face.
  80         unBit = 1 << VERT_RESULT_BFC0;
  81         if(mesa_vp->Base.OutputsWritten & unBit)
  82         {
  83                 pAsm->ucVP_OutputMap[VERT_RESULT_BFC0] = unTotal++;
  84         }
  85
  86         unBit = 1 << VERT_RESULT_BFC1;
  87         if(mesa_vp->Base.OutputsWritten & unBit)
  88         {
  89                 pAsm->ucVP_OutputMap[VERT_RESULT_BFC1] = unTotal++;
  90         }
  91
  92         //TODO : dealing fog.
  93         unBit = 1 << VERT_RESULT_FOGC;
  94         if(mesa_vp->Base.OutputsWritten & unBit)
  95         {
  96                 pAsm->ucVP_OutputMap[VERT_RESULT_FOGC] = unTotal++;
  97         }
  98
  99         //TODO : dealing point size.
 100         unBit = 1 << VERT_RESULT_PSIZ;
 101         if(mesa_vp->Base.OutputsWritten & unBit)
 102         {
 103                 pAsm->ucVP_OutputMap[VERT_RESULT_PSIZ] = unTotal++;
 104         }
 105
 106         for(i=0; i<8; i++)
 107         {
 108                 unBit = 1 << (VERT_RESULT_TEX0 + i);
 109                 if(mesa_vp->Base.OutputsWritten & unBit)
 110                 {
 111                         pAsm->ucVP_OutputMap[VERT_RESULT_TEX0 + i] = unTotal++;
 112                 }
 113         }
 114
 115     for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++)
 116         {
 117                 unBit = 1 << i;
 118                 if(mesa_vp->Base.OutputsWritten & unBit)
 119                 {
 120                         pAsm->ucVP_OutputMap[i] = unTotal++;
 121                 }
 122         }
 123
 124         return (unTotal - unStart);
 125 }
 126
 127 unsigned int Map_Vertex_Input(r700_AssemblerBase       *pAsm,
 128                                           struct gl_vertex_program *mesa_vp,
 129                                           unsigned int unStart)
 130 {
 131         int i;
 132         unsigned int unBit;
 133         unsigned int unTotal = unStart;
 134         for(i=0; i<VERT_ATTRIB_MAX; i++)
 135         {
 136                 unBit = 1 << i;
 137                 if(mesa_vp->Base.InputsRead & unBit)
 138                 {
 139                         pAsm->ucVP_AttributeMap[i] = unTotal++;
 140                 }
 141         }
 142         return (unTotal - unStart);
 143 }
 144
 145 GLboolean Process_Vertex_Program_Vfetch_Instructions(
 146                                                 struct r700_vertex_program *vp,
 147                                                 struct gl_vertex_program   *mesa_vp)
 148 {
 149         int i;
 150     unsigned int unBit;
 151         VTX_FETCH_METHOD vtxFetchMethod;
 152         vtxFetchMethod.bEnableMini          = GL_FALSE;
 153         vtxFetchMethod.mega_fetch_remainder = 0;
 154
 155         for(i=0; i<VERT_ATTRIB_MAX; i++)
 156         {
 157                 unBit = 1 << i;
 158                 if(mesa_vp->Base.InputsRead & unBit)
 159                 {
 160                         assemble_vfetch_instruction(&vp->r700AsmCode,
 161                                                     i,
 162                                                     vp->r700AsmCode.ucVP_AttributeMap[i],
 163                                                     vp->aos_desc[i].size,
 164                                                     vp->aos_desc[i].type,
 165                                                     &vtxFetchMethod);
 166                 }
 167         }
 168
 169         return GL_TRUE;
 170 }
 171
 172 GLboolean Process_Vertex_Program_Vfetch_Instructions2(
 173     struct gl_context *ctx,
 174         struct r700_vertex_program *vp,
 175         struct gl_vertex_program   *mesa_vp)
 176 {
 177     int i;
 178     context_t *context = R700_CONTEXT(ctx);
 179
 180     VTX_FETCH_METHOD vtxFetchMethod;
 181         vtxFetchMethod.bEnableMini          = GL_FALSE;
 182         vtxFetchMethod.mega_fetch_remainder = 0;
 183
 184     for(i=0; i<context->nNumActiveAos; i++)
 185     {
 186         assemble_vfetch_instruction2(&vp->r700AsmCode,
 187                                       vp->r700AsmCode.ucVP_AttributeMap[context->stream_desc[i].element],
 188                                       context->stream_desc[i].type,
 189                                       context->stream_desc[i].size,
 190                                       context->stream_desc[i].element,
 191                                       context->stream_desc[i]._signed,
 192                                       context->stream_desc[i].normalize,
 193                                       context->stream_desc[i].format,
 194                                      &vtxFetchMethod);
 195     }
 196
 197     return GL_TRUE;
 198 }
 199
 200 void Map_Vertex_Program(struct gl_context *ctx,
 201                         struct r700_vertex_program *vp,
 202                                                 struct gl_vertex_program   *mesa_vp)
 203 {
 204     r700_AssemblerBase *pAsm = &(vp->r700AsmCode);
 205         unsigned int num_inputs;
 206
 207         // R0 will always be used for index into vertex buffer
 208         pAsm->number_used_registers = 1;
 209         pAsm->starting_vfetch_register_number = pAsm->number_used_registers;
 210
 211     // Map Inputs: Add 1 to mapping since R0 is used for index
 212         num_inputs = Map_Vertex_Input(pAsm, mesa_vp, pAsm->number_used_registers);
 213         pAsm->number_used_registers += num_inputs;
 214
 215         // Create VFETCH instructions for inputs
 216         if (GL_TRUE != Process_Vertex_Program_Vfetch_Instructions2(ctx, vp, mesa_vp) )
 217         {
 218                 radeon_error("Calling Process_Vertex_Program_Vfetch_Instructions2 return error. \n");
 219                 return;
 220         }
 221
 222         // Map Outputs
 223         pAsm->number_of_exports = Map_Vertex_Output(pAsm, mesa_vp, pAsm->number_used_registers);
 224
 225         pAsm->starting_export_register_number = pAsm->number_used_registers;
 226
 227         pAsm->number_used_registers += pAsm->number_of_exports;
 228
 229     /* Map temporary registers (GPRs) */
 230     pAsm->starting_temp_register_number = pAsm->number_used_registers;
 231
 232     if(mesa_vp->Base.NumNativeTemporaries >= mesa_vp->Base.NumTemporaries)
 233     {   /* arb uses NumNativeTemporaries */
 234         pAsm->number_used_registers += mesa_vp->Base.NumNativeTemporaries;
 235     }
 236     else
 237     {   /* fix func t_vp uses NumTemporaries */
 238         pAsm->number_used_registers += mesa_vp->Base.NumTemporaries;
 239     }
 240
 241     pAsm->flag_reg_index = pAsm->number_used_registers++;
 242
 243     pAsm->uFirstHelpReg = pAsm->number_used_registers;
 244 }
 245
 246 GLboolean Find_Instruction_Dependencies_vp(struct r700_vertex_program *vp,
 247                                                                 struct gl_vertex_program   *mesa_vp)
 248 {
 249     GLuint i, j;
 250     GLint * puiTEMPwrites;
 251     struct prog_instruction *pILInst;
 252     InstDeps         *pInstDeps;
 253
 254     puiTEMPwrites = (GLint*) MALLOC(sizeof(GLuint)*mesa_vp->Base.NumTemporaries);
 255     for(i=0; i<mesa_vp->Base.NumTemporaries; i++)
 256     {
 257         puiTEMPwrites[i] = -1;
 258     }
 259
 260     pInstDeps = (InstDeps*)MALLOC(sizeof(InstDeps)*mesa_vp->Base.NumInstructions);
 261
 262     for(i=0; i<mesa_vp->Base.NumInstructions; i++)
 263     {
 264         pInstDeps[i].nDstDep = -1;
 265         pILInst = &(mesa_vp->Base.Instructions[i]);
 266
 267         //Dst
 268         if(pILInst->DstReg.File == PROGRAM_TEMPORARY)
 269         {
 270             //Set lastwrite for the temp
 271             puiTEMPwrites[pILInst->DstReg.Index] = i;
 272         }
 273
 274         //Src
 275         for(j=0; j<3; j++)
 276         {
 277             if(pILInst->SrcReg[j].File == PROGRAM_TEMPORARY)
 278             {
 279                 //Set dep.
 280                 pInstDeps[i].nSrcDeps[j] = puiTEMPwrites[pILInst->SrcReg[j].Index];
 281             }
 282             else
 283             {
 284                 pInstDeps[i].nSrcDeps[j] = -1;
 285             }
 286         }
 287     }
 288
 289     vp->r700AsmCode.pInstDeps = pInstDeps;
 290
 291     FREE(puiTEMPwrites);
 292
 293     return GL_TRUE;
 294 }
 295
 296 struct r700_vertex_program* r700TranslateVertexShader(struct gl_context *ctx,
 297                                                       struct gl_vertex_program *mesa_vp)
 298 {
 299         context_t *context = R700_CONTEXT(ctx);
 300
 301     R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
 302
 303         struct r700_vertex_program *vp;
 304         unsigned int i;
 305
 306         vp = calloc(1, sizeof(*vp));
 307         vp->mesa_program = _mesa_clone_vertex_program(ctx, mesa_vp);
 308
 309     vp->constbo0 = NULL;
 310
 311         if (mesa_vp->IsPositionInvariant)
 312         {
 313                 _mesa_insert_mvp_code(ctx, vp->mesa_program);
 314         }
 315
 316         for(i=0; i<context->nNumActiveAos; i++)
 317         {
 318                 vp->aos_desc[i].size   = context->stream_desc[i].size;
 319                 vp->aos_desc[i].stride = context->stream_desc[i].stride;
 320                 vp->aos_desc[i].type   = context->stream_desc[i].type;
 321                 vp->aos_desc[i].format = context->stream_desc[i].format;
 322         }
 323
 324         if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
 325         {
 326                 vp->r700AsmCode.bR6xx = 1;
 327         }
 328
 329         //Init_Program
 330         Init_r700_AssemblerBase(SPT_VP, &(vp->r700AsmCode), &(vp->r700Shader) );
 331
 332     if(GL_TRUE == r700->bShaderUseMemConstant)
 333     {
 334         vp->r700AsmCode.bUseMemConstant = GL_TRUE;
 335     }
 336     else
 337     {
 338         vp->r700AsmCode.bUseMemConstant = GL_FALSE;
 339     }
 340
 341     vp->r700AsmCode.unAsic = 7;
 342
 343         Map_Vertex_Program(ctx, vp, vp->mesa_program );
 344
 345         if(GL_FALSE == Find_Instruction_Dependencies_vp(vp, vp->mesa_program))
 346         {
 347                 return NULL;
 348         }
 349
 350     InitShaderProgram(&(vp->r700AsmCode));
 351
 352     for(i=0; i < MAX_SAMPLERS; i++)
 353     {
 354         vp->r700AsmCode.SamplerUnits[i] = vp->mesa_program->Base.SamplerUnits[i];
 355     }
 356
 357     vp->r700AsmCode.unCurNumILInsts = vp->mesa_program->Base.NumInstructions;
 358
 359         if(GL_FALSE == AssembleInstr(0,
 360                                  0,
 361                                  vp->mesa_program->Base.NumInstructions,
 362                                  &(vp->mesa_program->Base.Instructions[0]),
 363                                  &(vp->r700AsmCode)) )
 364         {
 365                 return NULL;
 366         }
 367
 368     if(GL_FALSE == Process_Vertex_Exports(&(vp->r700AsmCode), vp->mesa_program->Base.OutputsWritten) )
 369     {
 370         return NULL;
 371     }
 372
 373     if( GL_FALSE == RelocProgram(&(vp->r700AsmCode), &(vp->mesa_program->Base)) )
 374     {
 375         return GL_FALSE;
 376     }
 377
 378     vp->r700Shader.nRegs = (vp->r700AsmCode.number_used_registers == 0) ? 0
 379                          : (vp->r700AsmCode.number_used_registers - 1);
 380
 381         vp->r700Shader.nParamExports = vp->r700AsmCode.number_of_exports;
 382
 383     vp->translated = GL_TRUE;
 384
 385         return vp;
 386 }
 387
 388 void r700SelectVertexShader(struct gl_context *ctx)
 389 {
 390     context_t *context = R700_CONTEXT(ctx);
 391     struct r700_vertex_program_cont *vpc;
 392     struct r700_vertex_program *vp;
 393     unsigned int i;
 394     GLboolean match;
 395     GLbitfield InputsRead;
 396
 397     vpc = (struct r700_vertex_program_cont *)ctx->VertexProgram._Current;
 398
 399     InputsRead = vpc->mesa_program.Base.InputsRead;
 400     if (vpc->mesa_program.IsPositionInvariant)
 401     {
 402         InputsRead |= VERT_BIT_POS;
 403     }
 404
 405     for (vp = vpc->progs; vp; vp = vp->next)
 406     {
 407         match = GL_TRUE;
 408         for(i=0; i<context->nNumActiveAos; i++)
 409         {
 410                 if (vp->aos_desc[i].size != context->stream_desc[i].size ||
 411                     vp->aos_desc[i].format != context->stream_desc[i].format)
 412                 {
 413                         match = GL_FALSE;
 414                         break;
 415                 }
 416         }
 417         if (match)
 418         {
 419                 context->selected_vp = vp;
 420                 return;
 421         }
 422     }
 423
 424     vp = r700TranslateVertexShader(ctx, &(vpc->mesa_program));
 425     if(!vp)
 426     {
 427         radeon_error("Failed to translate vertex shader. \n");
 428         return;
 429     }
 430     vp->next = vpc->progs;
 431     vpc->progs = vp;
 432     context->selected_vp = vp;
 433     return;
 434 }
 435
 436 int getTypeSize(GLenum type)
 437 {
 438     switch (type)
 439     {
 440     case GL_DOUBLE:
 441         return sizeof(GLdouble);
 442     case GL_FLOAT:
 443         return sizeof(GLfloat);
 444     case GL_INT:
 445         return sizeof(GLint);
 446     case GL_UNSIGNED_INT:
 447         return sizeof(GLuint);
 448     case GL_SHORT:
 449         return sizeof(GLshort);
 450     case GL_UNSIGNED_SHORT:
 451         return sizeof(GLushort);
 452     case GL_BYTE:
 453         return sizeof(GLbyte);
 454     case GL_UNSIGNED_BYTE:
 455         return sizeof(GLubyte);
 456     default:
 457         assert(0);
 458         return 0;
 459     }
 460 }
 461
 462 static void r700TranslateAttrib(struct gl_context *ctx, GLuint unLoc, int count, const struct gl_client_array *input)
 463 {
 464     context_t *context = R700_CONTEXT(ctx);
 465
 466     StreamDesc * pStreamDesc = &(context->stream_desc[context->nNumActiveAos]);
 467
 468         GLuint stride;
 469
 470         stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size
 471                                    : input->StrideB;
 472
 473     if (input->Type == GL_DOUBLE || input->Type == GL_UNSIGNED_INT || input->Type == GL_INT
 474 #if MESA_BIG_ENDIAN
 475         || getTypeSize(input->Type) != 4
 476 #endif
 477        )
 478     {
 479         pStreamDesc->type = GL_FLOAT;
 480
 481         if (input->StrideB == 0)
 482         {
 483                 pStreamDesc->stride = 0;
 484         }
 485         else
 486         {
 487                 pStreamDesc->stride = sizeof(GLfloat) * input->Size;
 488         }
 489         pStreamDesc->dwords = input->Size;
 490         pStreamDesc->is_named_bo = GL_FALSE;
 491     }
 492     else
 493     {
 494         pStreamDesc->type = input->Type;
 495         pStreamDesc->dwords = (getTypeSize(input->Type) * input->Size + 3)/ 4;
 496         if (!input->BufferObj->Name)
 497         {
 498             if (input->StrideB == 0)
 499             {
 500                 pStreamDesc->stride = 0;
 501             }
 502             else
 503             {
 504                 pStreamDesc->stride = (getTypeSize(pStreamDesc->type) * input->Size + 3) & ~3;
 505             }
 506
 507             pStreamDesc->is_named_bo = GL_FALSE;
 508         }
 509     }
 510
 511         pStreamDesc->size = input->Size;
 512         pStreamDesc->dst_loc = context->nNumActiveAos;
 513         pStreamDesc->element = unLoc;
 514         pStreamDesc->format = input->Format;
 515
 516         switch (pStreamDesc->type)
 517         { //GetSurfaceFormat
 518         case GL_FLOAT:
 519                 pStreamDesc->_signed = 0;
 520                 pStreamDesc->normalize = GL_FALSE;
 521                 break;
 522         case GL_SHORT:
 523                 pStreamDesc->_signed = 1;
 524                 pStreamDesc->normalize = input->Normalized;
 525                 break;
 526         case GL_BYTE:
 527                 pStreamDesc->_signed = 1;
 528                 pStreamDesc->normalize = input->Normalized;
 529                 break;
 530         case GL_UNSIGNED_SHORT:
 531                 pStreamDesc->_signed = 0;
 532                 pStreamDesc->normalize = input->Normalized;
 533                 break;
 534         case GL_UNSIGNED_BYTE:
 535                 pStreamDesc->_signed = 0;
 536                 pStreamDesc->normalize = input->Normalized;
 537                 break;
 538         default:
 539         case GL_INT:
 540         case GL_UNSIGNED_INT:
 541         case GL_DOUBLE:
 542                 assert(0);
 543                 break;
 544         }
 545         context->nNumActiveAos++;
 546 }
 547
 548 void r700SetVertexFormat(struct gl_context *ctx, const struct gl_client_array *arrays[], int count)
 549 {
 550     context_t *context = R700_CONTEXT(ctx);
 551     struct r700_vertex_program *vpc
 552            = (struct r700_vertex_program *)ctx->VertexProgram._Current;
 553
 554     struct gl_vertex_program * mesa_vp = (struct gl_vertex_program *)&(vpc->mesa_program);
 555     unsigned int unLoc = 0;
 556     unsigned int unBit = mesa_vp->Base.InputsRead;
 557     context->nNumActiveAos = 0;
 558
 559     if (mesa_vp->IsPositionInvariant)
 560     {
 561         unBit |= VERT_BIT_POS;
 562     }
 563
 564     while(unBit)
 565     {
 566         if(unBit & 1)
 567         {
 568             r700TranslateAttrib(ctx, unLoc, count, arrays[unLoc]);
 569         }
 570
 571         unBit >>= 1;
 572         ++unLoc;
 573     }
 574     context->radeon.tcl.aos_count = context->nNumActiveAos;
 575 }
 576
 577 void * r700GetActiveVpShaderBo(struct gl_context * ctx)
 578 {
 579     context_t *context = R700_CONTEXT(ctx);
 580     struct r700_vertex_program *vp = context->selected_vp;;
 581
 582     if (vp)
 583         return vp->shaderbo;
 584     else
 585         return NULL;
 586 }
 587
 588 void * r700GetActiveVpShaderConstBo(struct gl_context * ctx)
 589 {
 590     context_t *context = R700_CONTEXT(ctx);
 591     struct r700_vertex_program *vp = context->selected_vp;;
 592
 593     if (vp)
 594         return vp->constbo0;
 595     else
 596         return NULL;
 597 }
 598
 599 GLboolean r700SetupVertexProgram(struct gl_context * ctx)
 600 {
 601     context_t *context = R700_CONTEXT(ctx);
 602     R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
 603     struct r700_vertex_program *vp = context->selected_vp;
 604
 605     struct gl_program_parameter_list *paramList;
 606     unsigned int unNumParamData;
 607     unsigned int ui;
 608     unsigned int num_sq_vs_gprs;
 609
 610     if(GL_FALSE == vp->loaded)
 611     {
 612             if(vp->r700Shader.bNeedsAssembly == GL_TRUE)
 613             {
 614                     Assemble( &(vp->r700Shader) );
 615             }
 616
 617         /* Load vp to gpu */
 618         r600EmitShader(ctx,
 619                        &(vp->shaderbo),
 620                        (GLvoid *)(vp->r700Shader.pProgram),
 621                        vp->r700Shader.uShaderBinaryDWORDSize,
 622                        "VS");
 623
 624         if(GL_TRUE == r700->bShaderUseMemConstant)
 625         {
 626             paramList = vp->mesa_program->Base.Parameters;
 627             if(NULL != paramList)
 628             {
 629                 unNumParamData = paramList->NumParameters;
 630                 r600AllocShaderConsts(ctx,
 631                                &(vp->constbo0),
 632                                unNumParamData *4*4,
 633                                "VSCON");
 634             }
 635         }
 636
 637         vp->loaded = GL_TRUE;
 638     }
 639
 640     DumpHwBinary(DUMP_VERTEX_SHADER, (GLvoid *)(vp->r700Shader.pProgram),
 641                  vp->r700Shader.uShaderBinaryDWORDSize);
 642
 643     /* TODO : enable this after MemUse fixed *=
 644     (context->chipobj.MemUse)(context, vp->shadercode.buf->id);
 645     */
 646
 647     R600_STATECHANGE(context, vs);
 648     R600_STATECHANGE(context, fs); /* hack */
 649
 650     r700->vs.SQ_PGM_RESOURCES_VS.u32All = 0;
 651     SETbit(r700->vs.SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
 652
 653     r700->vs.SQ_ALU_CONST_CACHE_VS_0.u32All = 0; /* set from buffer object. */
 654
 655     r700->vs.SQ_PGM_START_VS.u32All = 0;
 656
 657     SETfield(r700->vs.SQ_PGM_RESOURCES_VS.u32All, vp->r700Shader.nRegs + 1,
 658              NUM_GPRS_shift, NUM_GPRS_mask);
 659
 660     num_sq_vs_gprs = ((r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All & NUM_VS_GPRS_mask) >> NUM_VS_GPRS_shift);
 661
 662     if((vp->r700Shader.nRegs + 1) > num_sq_vs_gprs)
 663     {
 664         /* care! thich changes sq - needs idle state */
 665         R600_STATECHANGE(context, sq);
 666         SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All, vp->r700Shader.nRegs + 1,
 667                  NUM_VS_GPRS_shift, NUM_VS_GPRS_mask);
 668     }
 669
 670     if(vp->r700Shader.uStackSize) /* we don't use branch for now, it should be zero. */
 671         {
 672         SETfield(r700->vs.SQ_PGM_RESOURCES_VS.u32All, vp->r700Shader.uStackSize,
 673                  STACK_SIZE_shift, STACK_SIZE_mask);
 674     }
 675
 676     R600_STATECHANGE(context, spi);
 677
 678     if(vp->mesa_program->Base.OutputsWritten & (1 << VERT_RESULT_PSIZ)) {
 679         R600_STATECHANGE(context, cl);
 680         SETbit(r700->PA_CL_VS_OUT_CNTL.u32All, USE_VTX_POINT_SIZE_bit);
 681         SETbit(r700->PA_CL_VS_OUT_CNTL.u32All, VS_OUT_MISC_VEC_ENA_bit);
 682     } else if (r700->PA_CL_VS_OUT_CNTL.u32All != 0) {
 683         R600_STATECHANGE(context, cl);
 684         CLEARbit(r700->PA_CL_VS_OUT_CNTL.u32All, USE_VTX_POINT_SIZE_bit);
 685         CLEARbit(r700->PA_CL_VS_OUT_CNTL.u32All, VS_OUT_MISC_VEC_ENA_bit);
 686     }
 687
 688     SETfield(r700->SPI_VS_OUT_CONFIG.u32All,
 689              vp->r700Shader.nParamExports ? (vp->r700Shader.nParamExports - 1) : 0,
 690              VS_EXPORT_COUNT_shift, VS_EXPORT_COUNT_mask);
 691     SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, vp->r700Shader.nParamExports,
 692              NUM_INTERP_shift, NUM_INTERP_mask);
 693
 694     /*
 695     SETbit(r700->SPI_PS_IN_CONTROL_0.u32All, PERSP_GRADIENT_ENA_bit);
 696     CLEARbit(r700->SPI_PS_IN_CONTROL_0.u32All, LINEAR_GRADIENT_ENA_bit);
 697     */
 698
 699     /* sent out shader constants. */
 700     paramList = vp->mesa_program->Base.Parameters;
 701
 702     if(NULL != paramList) {
 703         /* vp->mesa_program was cloned, not updated by glsl shader api. */
 704         /* _mesa_reference_program has already checked glsl shProg is ok and set ctx->VertexProgem._Current */
 705         /* so, use ctx->VertexProgem._Current */
 706         struct gl_program_parameter_list *paramListOrginal =
 707                          ctx->VertexProgram._Current->Base.Parameters;
 708
 709             _mesa_load_state_parameters(ctx, paramList);
 710
 711             if (paramList->NumParameters > R700_MAX_DX9_CONSTS)
 712                     return GL_FALSE;
 713
 714             R600_STATECHANGE(context, vs_consts);
 715
 716             r700->vs.num_consts = paramList->NumParameters;
 717
 718             unNumParamData = paramList->NumParameters;
 719
 720             for(ui=0; ui<unNumParamData; ui++) {
 721             if(paramList->Parameters[ui].Type == PROGRAM_UNIFORM)
 722             {
 723               r700->vs.consts[ui][0].f32All = paramListOrginal->ParameterValues[ui][0].f;
 724                         r700->vs.consts[ui][1].f32All = paramListOrginal->ParameterValues[ui][1].f;
 725                         r700->vs.consts[ui][2].f32All = paramListOrginal->ParameterValues[ui][2].f;
 726                         r700->vs.consts[ui][3].f32All = paramListOrginal->ParameterValues[ui][3].f;
 727             }
 728             else
 729             {
 730                         r700->vs.consts[ui][0].f32All = paramList->ParameterValues[ui][0].f;
 731                         r700->vs.consts[ui][1].f32All = paramList->ParameterValues[ui][1].f;
 732                         r700->vs.consts[ui][2].f32All = paramList->ParameterValues[ui][2].f;
 733                         r700->vs.consts[ui][3].f32All = paramList->ParameterValues[ui][3].f;
 734             }
 735             }
 736
 737         /* Load vp constants to gpu */
 738         if(GL_TRUE == r700->bShaderUseMemConstant)
 739         {
 740             r600EmitShaderConsts(ctx,
 741                            vp->constbo0,
 742                            0,
 743                            (GLvoid *)&(r700->vs.consts[0][0]),
 744                            unNumParamData * 4 * 4);
 745         }
 746     } else
 747             r700->vs.num_consts = 0;
 748
 749     COMPILED_SUB * pCompiledSub;
 750     GLuint uj;
 751     GLuint unConstOffset = r700->vs.num_consts;
 752     for(ui=0; ui<vp->r700AsmCode.unNumPresub; ui++)
 753     {
 754         pCompiledSub = vp->r700AsmCode.presubs[ui].pCompiledSub;
 755
 756         r700->vs.num_consts += pCompiledSub->NumParameters;
 757
 758         for(uj=0; uj<pCompiledSub->NumParameters; uj++)
 759         {
 760             r700->vs.consts[uj + unConstOffset][0].f32All = pCompiledSub->ParameterValues[uj][0];
 761                     r700->vs.consts[uj + unConstOffset][1].f32All = pCompiledSub->ParameterValues[uj][1];
 762                     r700->vs.consts[uj + unConstOffset][2].f32All = pCompiledSub->ParameterValues[uj][2];
 763                     r700->vs.consts[uj + unConstOffset][3].f32All = pCompiledSub->ParameterValues[uj][3];
 764         }
 765         unConstOffset += pCompiledSub->NumParameters;
 766     }
 767
 768     return GL_TRUE;
 769 }