Rev 56 |
Blame |
Compare with Previous |
Last modification |
View Log
| RSS feed
/* $Id: t_imm_exec.c,v 1.2 2003-04-24 14:22:20 giacomo Exp $ */
/*
* Mesa 3-D graphics library
* Version: 4.1
*
* Copyright (C) 1999-2002 Brian Paul All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included
* in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
* AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
/**
* \file tnl/t_imm_exec.c
* \brief Setup to execute immediate-mode vertex data.
* \author Keith Whitwell
*/
#include "glheader.h"
#include "colormac.h"
#include "context.h"
#include "enums.h"
#include "dlist.h"
#include "macros.h"
#include "imports.h"
#include "mmath.h"
#include "light.h"
#include "state.h"
#include "mtypes.h"
#include "math/m_matrix.h"
#include "math/m_xform.h"
#include "t_context.h"
#include "t_import_array.h"
#include "t_imm_alloc.h"
#include "t_imm_api.h"
#include "t_imm_debug.h"
#include "t_imm_dlist.h"
#include "t_imm_eval.h"
#include "t_imm_elt.h"
#include "t_imm_exec.h"
#include "t_imm_fixup.h"
#include "t_pipeline.h"
static void reset_input( GLcontext *ctx,
GLuint start,
GLuint beginstate,
GLuint savedbeginstate )
{
struct immediate *IM = TNL_CURRENT_IM(ctx);
/* Clear the dirty part of the flag array.
*/
if (start < IM->Count+2)
MEMSET(IM->Flag + start, 0, sizeof(GLuint) * (IM->Count+2-start));
if (MESA_VERBOSE & VERBOSE_IMMEDIATE)
_mesa_debug(ctx, "reset_input: IM(%d) new %x\n", IM->id, beginstate);
IM->Start = start;
IM->Count = start;
IM->LastMaterial = start;
IM->BeginState = beginstate;
IM->SavedBeginState = savedbeginstate;
IM->TexSize = 0;
IM->MaterialOrMask = 0;
if (IM->MaterialMask)
IM->MaterialMask[IM->Start] = 0;
IM->ArrayEltFlags = ~ctx->Array._Enabled;
IM->ArrayEltIncr = ctx->Array.Vertex.Enabled ? 1 : 0;
IM->ArrayEltFlush = ctx->Array.LockCount ? FLUSH_ELT_LAZY : FLUSH_ELT_EAGER;
}
void _tnl_reset_exec_input( GLcontext *ctx,
GLuint start,
GLuint beginstate,
GLuint savedbeginstate )
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
struct immediate *IM = TNL_CURRENT_IM(ctx);
reset_input( ctx, start, beginstate, savedbeginstate );
IM->CopyStart = start - tnl->ExecCopyCount;
IM->Primitive[IM->CopyStart] = ctx->Driver.CurrentExecPrimitive;
if (tnl->ExecParity)
IM->Primitive[IM->CopyStart] |= PRIM_PARITY;
IM->LastPrimitive = IM->CopyStart;
}
void _tnl_reset_compile_input( GLcontext *ctx,
GLuint start,
GLuint beginstate,
GLuint savedbeginstate )
{
struct immediate *IM = TNL_CURRENT_IM(ctx);
reset_input( ctx, start, beginstate, savedbeginstate );
IM->CopyStart = start;
IM->LastPrimitive = IM->Start;
}
/**
* Copy the last specified normal, color, texcoord, edge flag, etc
* from the immediate struct into the ctx->Current attribute group.
*/
void _tnl_copy_to_current( GLcontext *ctx, struct immediate *IM,
GLuint flag, GLuint count )
{
if (MESA_VERBOSE&VERBOSE_IMMEDIATE)
_tnl_print_vert_flags("copy to current", flag);
/* XXX should be able to replace these conditions with a loop over
* the 16 vertex attributes.
*/
if (flag & VERT_BIT_NORMAL)
COPY_4FV( ctx->Current.Attrib[VERT_ATTRIB_NORMAL],
IM->Attrib[VERT_ATTRIB_NORMAL][count]);
if (flag & VERT_BIT_INDEX)
ctx->Current.Index = IM->Index[count];
if (flag & VERT_BIT_EDGEFLAG)
ctx->Current.EdgeFlag = IM->EdgeFlag[count];
if (flag & VERT_BIT_COLOR0) {
COPY_4FV(ctx->Current.Attrib[VERT_ATTRIB_COLOR0],
IM->Attrib[VERT_ATTRIB_COLOR0][count]);
if (ctx->Light.ColorMaterialEnabled) {
_mesa_update_color_material( ctx,
ctx->Current.Attrib[VERT_ATTRIB_COLOR0] );
TNL_CONTEXT(ctx)->Driver.NotifyMaterialChange( ctx );
}
}
if (flag & VERT_BIT_COLOR1)
COPY_4FV(ctx->Current.Attrib[VERT_ATTRIB_COLOR1],
IM->Attrib[VERT_ATTRIB_COLOR1][count]);
if (flag & VERT_BIT_FOG)
ctx->Current.Attrib[VERT_ATTRIB_FOG][0] = IM->Attrib[VERT_ATTRIB_FOG][count][0];
if (flag & VERT_BITS_TEX_ANY) {
GLuint i;
for (i = 0 ; i < ctx->Const.MaxTextureUnits ; i++) {
if (flag & VERT_BIT_TEX(i)) {
COPY_4FV( ctx->Current.Attrib[VERT_ATTRIB_TEX0 + i],
IM->Attrib[VERT_ATTRIB_TEX0 + i][count]);
}
}
}
if (flag & VERT_BIT_MATERIAL) {
_mesa_update_material( ctx,
IM->Material[IM->LastMaterial],
IM->MaterialOrMask );
TNL_CONTEXT(ctx)->Driver.NotifyMaterialChange( ctx );
}
}
void _tnl_compute_orflag( struct immediate *IM, GLuint start )
{
GLuint count = IM->Count;
GLuint orflag = 0;
GLuint andflag = ~0U;
GLuint i;
IM->LastData = count-1;
/* Compute the flags for the whole buffer.
*/
for (i = start ; i < count ; i++) {
andflag &= IM->Flag[i];
orflag |= IM->Flag[i];
}
/* It is possible there will be data in the buffer arising from
* calls like 'glNormal', 'glMaterial' that occur after the final
* glVertex, glEval, etc. Additionally, a buffer can consist of
* eg. a single glMaterial call, in which case IM->Start ==
* IM->Count, but the buffer is definitely not empty.
*/
if (IM->Flag[i] & VERT_BITS_DATA) {
IM->LastData++;
orflag |= IM->Flag[i];
}
IM->Flag[IM->LastData+1] |= VERT_BIT_END_VB;
IM->CopyAndFlag = IM->AndFlag = andflag;
IM->OrFlag = orflag;
IM->CopyOrFlag = orflag;
IM->Evaluated = 0;
}
/**
* This is where the vertex data is transfered from the 'struct immediate
* into the 'struct vertex_buffer'.
*
* Note: The 'start' member of the GLvector structs is now redundant
* because we always re-transform copied vertices, and the vectors
* below are set up so that the first copied vertex (if any) appears
* at position zero.
*/
static void _tnl_vb_bind_immediate( GLcontext *ctx, struct immediate *IM )
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
struct vertex_buffer *VB = &tnl->vb;
struct vertex_arrays *tmp = &tnl->imm_inputs;
GLuint inputs = tnl->pipeline.inputs; /* for copy-to-current */
const GLuint start = IM->CopyStart;
const GLuint count = IM->Count - start;
/* TODO: optimize the case where nothing has changed. (Just bind
* tmp to vb).
*/
/* Setup constant data in the VB.
*/
VB->Count = count;
VB->FirstClipped = IMM_MAXDATA - IM->CopyStart;
VB->import_data = NULL;
VB->importable_data = 0;
/* Need an IM->FirstPrimitive?
*/
VB->Primitive = IM->Primitive + IM->CopyStart;
VB->PrimitiveLength = IM->PrimitiveLength + IM->CopyStart;
VB->FirstPrimitive = 0;
VB->Flag = IM->Flag + start;
/* TexCoordPtr's are zeroed in loop below.
*/
VB->NormalPtr = NULL;
VB->NormalLengthPtr = NULL;
VB->EdgeFlag = NULL;
VB->IndexPtr[0] = NULL;
VB->IndexPtr[1] = NULL;
VB->ColorPtr[0] = NULL;
VB->ColorPtr[1] = NULL;
VB->SecondaryColorPtr[0] = NULL;
VB->SecondaryColorPtr[1] = NULL;
VB->Elts = NULL;
VB->MaterialMask = NULL;
VB->Material = NULL;
/* _tnl_print_vert_flags("copy-orflag", IM->CopyOrFlag); */
/* _tnl_print_vert_flags("orflag", IM->OrFlag); */
/* _tnl_print_vert_flags("inputs", inputs); */
/* Setup the initial values of array pointers in the vb.
*/
if (inputs & VERT_BIT_POS) {
tmp->Obj.data = IM->Attrib[VERT_ATTRIB_POS] + start;
tmp->Obj.start = (GLfloat *)(IM->Attrib[VERT_ATTRIB_POS] + start);
tmp->Obj.count = count;
VB->ObjPtr = &tmp->Obj;
if ((IM->CopyOrFlag & VERT_BITS_OBJ_234) == VERT_BITS_OBJ_234)
tmp->Obj.size = 4;
else if ((IM->CopyOrFlag & VERT_BITS_OBJ_234) == VERT_BITS_OBJ_23)
tmp->Obj.size = 3;
else
tmp->Obj.size = 2;
}
if (inputs & VERT_BIT_NORMAL) {
tmp->Normal.data = IM->Attrib[VERT_ATTRIB_NORMAL] + start;
tmp->Normal.start = (GLfloat *) (IM->Attrib[VERT_ATTRIB_NORMAL] + start);
tmp->Normal.count = count;
tmp->Normal.size = 3; /* just to be safe */
VB->NormalPtr = &tmp->Normal;
if (IM->NormalLengthPtr)
VB->NormalLengthPtr = IM->NormalLengthPtr + start;
}
if (inputs & VERT_BIT_INDEX) {
tmp->Index.count = count;
tmp->Index.data = IM->Index + start;
tmp->Index.start = IM->Index + start;
VB->IndexPtr[0] = &tmp->Index;
}
if (inputs & VERT_BIT_FOG) {
tmp->FogCoord.data = IM->Attrib[VERT_ATTRIB_FOG] + start;
tmp->FogCoord.start = (GLfloat *) (IM->Attrib[VERT_ATTRIB_FOG] + start);
tmp->FogCoord.count = count;
VB->FogCoordPtr = &tmp->FogCoord;
}
if (inputs & VERT_BIT_COLOR1) {
tmp->SecondaryColor.Ptr = IM->Attrib[VERT_ATTRIB_COLOR1] + start;
VB->SecondaryColorPtr[0] = &tmp->SecondaryColor;
}
if (inputs & VERT_BIT_EDGEFLAG) {
VB->EdgeFlag = IM->EdgeFlag + start;
}
if (inputs & VERT_BIT_COLOR0) {
if (IM->CopyOrFlag & VERT_BIT_COLOR0) {
tmp->Color.Ptr = IM->Attrib[VERT_ATTRIB_COLOR0] + start;
tmp->Color.StrideB = 4 * sizeof(GLfloat);
tmp->Color.Flags = 0;
}
else {
tmp->Color.Ptr = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
tmp->Color.StrideB = 0;
tmp->Color.Flags = CA_CLIENT_DATA; /* hack */
VB->import_source = IM;
VB->importable_data |= VERT_BIT_COLOR0;
VB->import_data = _tnl_upgrade_current_data;
}
VB->ColorPtr[0] = &tmp->Color;
}
if (inputs & VERT_BITS_TEX_ANY) {
GLuint i;
for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
VB->TexCoordPtr[i] = NULL;
if (inputs & VERT_BIT_TEX(i)) {
tmp->TexCoord[i].count = count;
tmp->TexCoord[i].data = IM->Attrib[VERT_ATTRIB_TEX0 + i] + start;
tmp->TexCoord[i].start = (GLfloat *)(IM->Attrib[VERT_ATTRIB_TEX0 + i] + start);
tmp->TexCoord[i].size = 2;
if (IM->TexSize & TEX_SIZE_3(i)) {
tmp->TexCoord[i].size = 3;
if (IM->TexSize & TEX_SIZE_4(i))
tmp->TexCoord[i].size = 4;
}
VB->TexCoordPtr[i] = &tmp->TexCoord[i];
}
}
}
if ((inputs & IM->OrFlag & VERT_BIT_MATERIAL) && IM->Material) {
VB->MaterialMask = IM->MaterialMask + start;
VB->Material = IM->Material + start;
}
/* GL_NV_vertex_program */
if (ctx->VertexProgram.Enabled) {
GLuint attr;
for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) {
tmp->Attribs[attr].count = count;
tmp->Attribs[attr].data = IM->Attrib[attr] + start;
tmp->Attribs[attr].start = (GLfloat *) (IM->Attrib[attr] + start);
tmp->Attribs[attr].size = 4;
VB->AttribPtr[attr] = &(tmp->Attribs[attr]);
}
}
}
/**
* Called by exec_vert_cassette, execute_compiled_cassette, but not
* exec_elt_cassette.
*/
void _tnl_run_cassette( GLcontext *ctx, struct immediate *IM )
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
_tnl_vb_bind_immediate( ctx, IM );
if (IM->OrFlag & VERT_BITS_EVAL_ANY)
_tnl_eval_immediate( ctx, IM );
/* Invalidate all stored data before and after run:
*/
tnl->pipeline.run_input_changes |= tnl->pipeline.inputs;
tnl->Driver.RunPipeline( ctx );
tnl->pipeline.run_input_changes |= tnl->pipeline.inputs;
_tnl_copy_to_current( ctx, IM, IM->OrFlag, IM->LastData );
}
/**
* Called for regular vertex cassettes.
*/
static void exec_vert_cassette( GLcontext *ctx, struct immediate *IM )
{
if (IM->FlushElt) {
/* Orflag is computed twice, but only reach this code if app is
* using a mixture of glArrayElement() and glVertex() while
* arrays are locked (else would be in exec_elt_cassette now).
*/
ASSERT(ctx->Array.LockCount);
ASSERT(IM->FlushElt == FLUSH_ELT_LAZY);
_tnl_translate_array_elts( ctx, IM, IM->CopyStart, IM->Count );
_tnl_compute_orflag( IM, IM->CopyStart );
}
_tnl_fixup_input( ctx, IM );
/* _tnl_print_cassette( IM ); */
_tnl_run_cassette( ctx, IM );
}
/* Called for pure, locked VERT_BIT_ELT cassettes instead of
* _tnl_run_cassette.
*/
static void exec_elt_cassette( GLcontext *ctx, struct immediate *IM )
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
struct vertex_buffer *VB = &tnl->vb;
_tnl_vb_bind_arrays( ctx, ctx->Array.LockFirst, ctx->Array.LockCount );
/* Take only elements and primitive information from the immediate:
*/
VB->Elts = IM->Elt + IM->CopyStart;
VB->Primitive = IM->Primitive + IM->CopyStart;
VB->PrimitiveLength = IM->PrimitiveLength + IM->CopyStart;
VB->FirstPrimitive = 0;
/* Run the pipeline. No input changes as a result of this action.
*/
tnl->Driver.RunPipeline( ctx );
/* Still need to update current values:
*/
if (ctx->Driver.CurrentExecPrimitive == GL_POLYGON+1) {
_tnl_translate_array_elts( ctx, IM, IM->LastData, IM->LastData );
_tnl_copy_to_current( ctx, IM, ctx->Array._Enabled, IM->LastData );
}
}
static void
exec_empty_cassette( GLcontext *ctx, struct immediate *IM )
{
if (IM->FlushElt)
_tnl_translate_array_elts( ctx, IM, IM->CopyStart, IM->CopyStart );
_tnl_copy_to_current( ctx, IM, IM->OrFlag, IM->LastData );
}
/**
* Called for all cassettes when not compiling or playing a display
* list.
*/
void _tnl_execute_cassette( GLcontext *ctx, struct immediate *IM )
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
_tnl_compute_orflag( IM, IM->Start );
_tnl_copy_immediate_vertices( ctx, IM );
_tnl_get_exec_copy_verts( ctx, IM );
if (tnl->pipeline.build_state_changes)
_tnl_validate_pipeline( ctx );
if (IM->CopyStart == IM->Count) {
exec_empty_cassette( ctx, IM );
}
else if ((IM->CopyOrFlag & VERT_BITS_DATA) == VERT_BIT_ELT &&
ctx->Array.LockCount &&
ctx->Array.Vertex.Enabled) {
exec_elt_cassette( ctx, IM );
}
else {
exec_vert_cassette( ctx, IM );
}
/* Only reuse the immediate if there are no copied vertices living
* inside it:
*/
{
GLuint begin_state = IM->BeginState & (VERT_BEGIN_0|VERT_BEGIN_1);
GLuint saved_begin_state = IM->SavedBeginState;
if (--IM->ref_count != 0) {
IM = _tnl_alloc_immediate( ctx );
SET_IMMEDIATE( ctx, IM );
}
IM->ref_count++;
_tnl_reset_exec_input( ctx, IMM_MAX_COPIED_VERTS,
begin_state, saved_begin_state );
}
if (ctx->Driver.CurrentExecPrimitive == GL_POLYGON+1)
ctx->Driver.NeedFlush &= ~FLUSH_STORED_VERTICES;
}
/**
* Setup vector pointers that will be used to bind immediates to VB's.
*/
void _tnl_imm_init( GLcontext *ctx )
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
struct vertex_arrays *tmp = &tnl->imm_inputs;
GLuint i;
static int firsttime = 1;
if (firsttime) {
firsttime = 0;
_tnl_imm_elt_init();
}
ctx->swtnl_im = _tnl_alloc_immediate( ctx );
TNL_CURRENT_IM(ctx)->ref_count++;
tnl->ExecCopyTexSize = 0;
tnl->ExecCopyCount = 0;
tnl->ExecCopySource = 0;
TNL_CURRENT_IM(ctx)->CopyStart = IMM_MAX_COPIED_VERTS;
_mesa_vector4f_init( &tmp->Obj, 0, 0 );
_mesa_vector4f_init( &tmp->Normal, 0, 0 );
tmp->Color.Ptr = NULL;
tmp->Color.Type = GL_FLOAT;
tmp->Color.Size = 4;
tmp->Color.Stride = 0;
tmp->Color.StrideB = 4 * sizeof(GLfloat);
tmp->Color.Flags = 0;
tmp->SecondaryColor.Ptr = NULL;
tmp->SecondaryColor.Type = GL_FLOAT;
tmp->SecondaryColor.Size = 4;
tmp->SecondaryColor.Stride = 0;
tmp->SecondaryColor.StrideB = 4 * sizeof(GLfloat);
tmp->SecondaryColor.Flags = 0;
_mesa_vector4f_init( &tmp->FogCoord, 0, 0 );
_mesa_vector1ui_init( &tmp->Index, 0, 0 );
_mesa_vector1ub_init( &tmp->EdgeFlag, 0, 0 );
for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
_mesa_vector4f_init( &tmp->TexCoord[i], 0, 0);
/* Install the first immediate. Intially outside begin/end.
*/
_tnl_reset_exec_input( ctx, IMM_MAX_COPIED_VERTS, 0, 0 );
tnl->ReplayHardBeginEnd = 0;
_tnl_imm_vtxfmt_init( ctx );
}
/**
* Deallocate the immediate-mode buffer for the given context, if
* its reference count goes to zero.
*/
void _tnl_imm_destroy( GLcontext *ctx )
{
if (TNL_CURRENT_IM(ctx)) {
TNL_CURRENT_IM(ctx)->ref_count--;
if (TNL_CURRENT_IM(ctx)->ref_count == 0)
_tnl_free_immediate( ctx, TNL_CURRENT_IM(ctx) );
/*
* Don't use SET_IMMEDIATE here, or else we'll whack the
* _tnl_CurrentInput pointer - not good when another
* context has already been made current.
* So we just set the context's own tnl immediate pointer
* to 0.
*/
ctx->swtnl_im = NULL;
}
}