1672 lines
38 KiB
C++
1672 lines
38 KiB
C++
/*************************************************************************/
|
|
/* visual_server.cpp */
|
|
/*************************************************************************/
|
|
/* This file is part of: */
|
|
/* GODOT ENGINE */
|
|
/* http://www.godotengine.org */
|
|
/*************************************************************************/
|
|
/* Copyright (c) 2007-2016 Juan Linietsky, Ariel Manzur. */
|
|
/* */
|
|
/* 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 THE AUTHORS OR COPYRIGHT HOLDERS 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. */
|
|
/*************************************************************************/
|
|
#include "visual_server.h"
|
|
#include "globals.h"
|
|
#include "method_bind_ext.inc"
|
|
|
|
VisualServer *VisualServer::singleton=NULL;
|
|
VisualServer* (*VisualServer::create_func)()=NULL;
|
|
|
|
VisualServer *VisualServer::get_singleton() {
|
|
|
|
return singleton;
|
|
}
|
|
|
|
|
|
|
|
DVector<String> VisualServer::_shader_get_param_list(RID p_shader) const {
|
|
|
|
//remove at some point
|
|
|
|
DVector<String> pl;
|
|
|
|
|
|
#if 0
|
|
List<StringName> params;
|
|
shader_get_param_list(p_shader,¶ms);
|
|
|
|
|
|
for(List<StringName>::Element *E=params.front();E;E=E->next()) {
|
|
|
|
pl.push_back(E->get());
|
|
}
|
|
#endif
|
|
return pl;
|
|
}
|
|
|
|
VisualServer *VisualServer::create() {
|
|
|
|
ERR_FAIL_COND_V(singleton,NULL);
|
|
|
|
if (create_func)
|
|
return create_func();
|
|
|
|
return NULL;
|
|
}
|
|
|
|
RID VisualServer::texture_create_from_image(const Image& p_image,uint32_t p_flags) {
|
|
|
|
RID texture = texture_create();
|
|
texture_allocate(texture,p_image.get_width(), p_image.get_height(), p_image.get_format(), p_flags); //if it has mipmaps, use, else generate
|
|
ERR_FAIL_COND_V(!texture.is_valid(),texture);
|
|
|
|
texture_set_data(texture, p_image );
|
|
|
|
return texture;
|
|
}
|
|
|
|
RID VisualServer::get_test_texture() {
|
|
|
|
if (test_texture.is_valid()) {
|
|
return test_texture;
|
|
};
|
|
|
|
#define TEST_TEXTURE_SIZE 256
|
|
|
|
|
|
DVector<uint8_t> test_data;
|
|
test_data.resize(TEST_TEXTURE_SIZE*TEST_TEXTURE_SIZE*3);
|
|
|
|
{
|
|
DVector<uint8_t>::Write w=test_data.write();
|
|
|
|
for (int x=0;x<TEST_TEXTURE_SIZE;x++) {
|
|
|
|
for (int y=0;y<TEST_TEXTURE_SIZE;y++) {
|
|
|
|
Color c;
|
|
int r=255-(x+y)/2;
|
|
|
|
if ((x%(TEST_TEXTURE_SIZE/8))<2 ||(y%(TEST_TEXTURE_SIZE/8))<2) {
|
|
|
|
c.r=y;
|
|
c.g=r;
|
|
c.b=x;
|
|
|
|
} else {
|
|
|
|
c.r=r;
|
|
c.g=x;
|
|
c.b=y;
|
|
}
|
|
|
|
w[(y*TEST_TEXTURE_SIZE+x)*3+0]=uint8_t(CLAMP(c.r*255,0,255));
|
|
w[(y*TEST_TEXTURE_SIZE+x)*3+1]=uint8_t(CLAMP(c.g*255,0,255));
|
|
w[(y*TEST_TEXTURE_SIZE+x)*3+2]=uint8_t(CLAMP(c.b*255,0,255));
|
|
}
|
|
}
|
|
}
|
|
|
|
Image data(TEST_TEXTURE_SIZE,TEST_TEXTURE_SIZE,false,Image::FORMAT_RGB8,test_data);
|
|
|
|
test_texture = texture_create_from_image(data);
|
|
|
|
return test_texture;
|
|
};
|
|
|
|
void VisualServer::_free_internal_rids() {
|
|
|
|
if (test_texture.is_valid())
|
|
free(test_texture);
|
|
if (white_texture.is_valid())
|
|
free(white_texture);
|
|
if (test_material.is_valid())
|
|
free(test_material);
|
|
|
|
for(int i=0;i<16;i++) {
|
|
if (material_2d[i].is_valid())
|
|
free(material_2d[i]);
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
RID VisualServer::_make_test_cube() {
|
|
|
|
DVector<Vector3> vertices;
|
|
DVector<Vector3> normals;
|
|
DVector<float> tangents;
|
|
DVector<Vector3> uvs;
|
|
|
|
int vtx_idx=0;
|
|
#define ADD_VTX(m_idx);\
|
|
vertices.push_back( face_points[m_idx] );\
|
|
normals.push_back( normal_points[m_idx] );\
|
|
tangents.push_back( normal_points[m_idx][1] );\
|
|
tangents.push_back( normal_points[m_idx][2] );\
|
|
tangents.push_back( normal_points[m_idx][0] );\
|
|
tangents.push_back( 1.0 );\
|
|
uvs.push_back( Vector3(uv_points[m_idx*2+0],uv_points[m_idx*2+1],0) );\
|
|
vtx_idx++;\
|
|
|
|
for (int i=0;i<6;i++) {
|
|
|
|
|
|
Vector3 face_points[4];
|
|
Vector3 normal_points[4];
|
|
float uv_points[8]={0,0,0,1,1,1,1,0};
|
|
|
|
for (int j=0;j<4;j++) {
|
|
|
|
float v[3];
|
|
v[0]=1.0;
|
|
v[1]=1-2*((j>>1)&1);
|
|
v[2]=v[1]*(1-2*(j&1));
|
|
|
|
for (int k=0;k<3;k++) {
|
|
|
|
if (i<3)
|
|
face_points[j][(i+k)%3]=v[k]*(i>=3?-1:1);
|
|
else
|
|
face_points[3-j][(i+k)%3]=v[k]*(i>=3?-1:1);
|
|
}
|
|
normal_points[j]=Vector3();
|
|
normal_points[j][i%3]=(i>=3?-1:1);
|
|
}
|
|
|
|
//tri 1
|
|
ADD_VTX(0);
|
|
ADD_VTX(1);
|
|
ADD_VTX(2);
|
|
//tri 2
|
|
ADD_VTX(2);
|
|
ADD_VTX(3);
|
|
ADD_VTX(0);
|
|
|
|
}
|
|
|
|
RID test_cube = mesh_create();
|
|
|
|
Array d;
|
|
d.resize(VS::ARRAY_MAX);
|
|
d[VisualServer::ARRAY_NORMAL]= normals ;
|
|
d[VisualServer::ARRAY_TANGENT]= tangents ;
|
|
d[VisualServer::ARRAY_TEX_UV]= uvs ;
|
|
d[VisualServer::ARRAY_VERTEX]= vertices ;
|
|
|
|
DVector<int> indices;
|
|
indices.resize(vertices.size());
|
|
for(int i=0;i<vertices.size();i++)
|
|
indices.set(i,i);
|
|
d[VisualServer::ARRAY_INDEX]=indices;
|
|
|
|
mesh_add_surface_from_arrays( test_cube, PRIMITIVE_TRIANGLES,d );
|
|
|
|
|
|
/*
|
|
test_material = fixed_material_create();
|
|
//material_set_flag(material, MATERIAL_FLAG_BILLBOARD_TOGGLE,true);
|
|
fixed_material_set_texture( test_material, FIXED_MATERIAL_PARAM_DIFFUSE, get_test_texture() );
|
|
fixed_material_set_param( test_material, FIXED_MATERIAL_PARAM_SPECULAR_EXP, 70 );
|
|
fixed_material_set_param( test_material, FIXED_MATERIAL_PARAM_EMISSION, Color(0.2,0.2,0.2) );
|
|
|
|
fixed_material_set_param( test_material, FIXED_MATERIAL_PARAM_DIFFUSE, Color(1, 1, 1) );
|
|
fixed_material_set_param( test_material, FIXED_MATERIAL_PARAM_SPECULAR, Color(1,1,1) );
|
|
*/
|
|
mesh_surface_set_material(test_cube, 0, test_material );
|
|
|
|
return test_cube;
|
|
}
|
|
|
|
|
|
RID VisualServer::make_sphere_mesh(int p_lats,int p_lons,float p_radius) {
|
|
|
|
DVector<Vector3> vertices;
|
|
DVector<Vector3> normals;
|
|
|
|
for(int i = 1; i <= p_lats; i++) {
|
|
double lat0 = Math_PI * (-0.5 + (double) (i - 1) / p_lats);
|
|
double z0 = Math::sin(lat0);
|
|
double zr0 = Math::cos(lat0);
|
|
|
|
double lat1 = Math_PI * (-0.5 + (double) i / p_lats);
|
|
double z1 = Math::sin(lat1);
|
|
double zr1 = Math::cos(lat1);
|
|
|
|
for(int j = p_lons; j >= 1; j--) {
|
|
|
|
double lng0 = 2 * Math_PI * (double) (j - 1) / p_lons;
|
|
double x0 = Math::cos(lng0);
|
|
double y0 = Math::sin(lng0);
|
|
|
|
double lng1 = 2 * Math_PI * (double) (j) / p_lons;
|
|
double x1 = Math::cos(lng1);
|
|
double y1 = Math::sin(lng1);
|
|
|
|
|
|
Vector3 v[4]={
|
|
Vector3(x1 * zr0, z0, y1 *zr0),
|
|
Vector3(x1 * zr1, z1, y1 *zr1),
|
|
Vector3(x0 * zr1, z1, y0 *zr1),
|
|
Vector3(x0 * zr0, z0, y0 *zr0)
|
|
};
|
|
|
|
#define ADD_POINT(m_idx)\
|
|
normals.push_back(v[m_idx]); \
|
|
vertices.push_back(v[m_idx]*p_radius);\
|
|
|
|
ADD_POINT(0);
|
|
ADD_POINT(1);
|
|
ADD_POINT(2);
|
|
|
|
ADD_POINT(2);
|
|
ADD_POINT(3);
|
|
ADD_POINT(0);
|
|
}
|
|
}
|
|
|
|
RID mesh = mesh_create();
|
|
Array d;
|
|
d.resize(VS::ARRAY_MAX);
|
|
|
|
d[ARRAY_VERTEX]=vertices;
|
|
d[ARRAY_NORMAL]=normals;
|
|
|
|
mesh_add_surface_from_arrays(mesh,PRIMITIVE_TRIANGLES,d);
|
|
|
|
return mesh;
|
|
}
|
|
|
|
|
|
RID VisualServer::material_2d_get(bool p_shaded, bool p_transparent, bool p_cut_alpha, bool p_opaque_prepass) {
|
|
|
|
int version=0;
|
|
if (p_shaded)
|
|
version=1;
|
|
if (p_transparent)
|
|
version|=2;
|
|
if (p_cut_alpha)
|
|
version|=4;
|
|
if (p_opaque_prepass)
|
|
version|=8;
|
|
if (material_2d[version].is_valid())
|
|
return material_2d[version];
|
|
|
|
//not valid, make
|
|
|
|
/* material_2d[version]=fixed_material_create();
|
|
fixed_material_set_flag(material_2d[version],FIXED_MATERIAL_FLAG_USE_ALPHA,p_transparent);
|
|
fixed_material_set_flag(material_2d[version],FIXED_MATERIAL_FLAG_USE_COLOR_ARRAY,true);
|
|
fixed_material_set_flag(material_2d[version],FIXED_MATERIAL_FLAG_DISCARD_ALPHA,p_cut_alpha);
|
|
material_set_flag(material_2d[version],MATERIAL_FLAG_UNSHADED,!p_shaded);
|
|
material_set_flag(material_2d[version],MATERIAL_FLAG_DOUBLE_SIDED,true);
|
|
material_set_depth_draw_mode(material_2d[version],p_opaque_prepass?MATERIAL_DEPTH_DRAW_OPAQUE_PRE_PASS_ALPHA:MATERIAL_DEPTH_DRAW_OPAQUE_ONLY);
|
|
fixed_material_set_texture(material_2d[version],FIXED_MATERIAL_PARAM_DIFFUSE,get_white_texture());
|
|
//material cut alpha?*/
|
|
|
|
return material_2d[version];
|
|
}
|
|
|
|
RID VisualServer::get_white_texture() {
|
|
|
|
if (white_texture.is_valid())
|
|
return white_texture;
|
|
|
|
DVector<uint8_t> wt;
|
|
wt.resize(16*3);
|
|
{
|
|
DVector<uint8_t>::Write w =wt.write();
|
|
for(int i=0;i<16*3;i++)
|
|
w[i]=255;
|
|
}
|
|
Image white(4,4,0,Image::FORMAT_RGB8,wt);
|
|
white_texture=texture_create();
|
|
texture_allocate(white_texture,4,4,Image::FORMAT_RGB8);
|
|
texture_set_data(white_texture,white);
|
|
return white_texture;
|
|
|
|
}
|
|
|
|
|
|
Error VisualServer::_surface_set_data(Array p_arrays,uint32_t p_format,uint32_t *p_offsets,uint32_t p_stride,DVector<uint8_t> &r_vertex_array,int p_vertex_array_len,DVector<uint8_t> &r_index_array,int p_index_array_len,AABB &r_aabb,Vector<AABB> r_bone_aabb) {
|
|
|
|
DVector<uint8_t>::Write vw = r_vertex_array.write();
|
|
|
|
DVector<uint8_t>::Write iw;
|
|
if (r_index_array.size()) {
|
|
iw=r_index_array.write();
|
|
}
|
|
|
|
int max_bone=0;
|
|
|
|
|
|
for(int ai=0;ai<VS::ARRAY_MAX;ai++) {
|
|
|
|
if (!(p_format&(1<<ai))) // no array
|
|
continue;
|
|
|
|
|
|
switch(ai) {
|
|
|
|
case VS::ARRAY_VERTEX: {
|
|
|
|
if (p_format& VS::ARRAY_FLAG_USE_2D_VERTICES) {
|
|
|
|
DVector<Vector2> array = p_arrays[ai];
|
|
ERR_FAIL_COND_V( array.size() != p_vertex_array_len, ERR_INVALID_PARAMETER );
|
|
|
|
|
|
DVector<Vector2>::Read read = array.read();
|
|
const Vector2* src=read.ptr();
|
|
|
|
// setting vertices means regenerating the AABB
|
|
Rect2 aabb;
|
|
|
|
|
|
if (p_format&ARRAY_COMPRESS_VERTEX) {
|
|
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
|
|
uint16_t vector[2]={ Math::make_half_float(src[i].x), Math::make_half_float(src[i].y) };
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], vector, sizeof(uint16_t)*2);
|
|
|
|
if (i==0) {
|
|
|
|
aabb=Rect2(src[i],Vector2());
|
|
} else {
|
|
|
|
aabb.expand_to( src[i] );
|
|
}
|
|
}
|
|
|
|
|
|
} else {
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
|
|
float vector[2]={ src[i].x, src[i].y };
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], vector, sizeof(float)*2);
|
|
|
|
if (i==0) {
|
|
|
|
aabb=Rect2(src[i],Vector2());
|
|
} else {
|
|
|
|
aabb.expand_to( src[i] );
|
|
}
|
|
}
|
|
}
|
|
|
|
r_aabb=AABB(Vector3(aabb.pos.x,aabb.pos.y,0),Vector3(aabb.size.x,aabb.size.y,0));
|
|
|
|
|
|
} else {
|
|
DVector<Vector3> array = p_arrays[ai];
|
|
ERR_FAIL_COND_V( array.size() != p_vertex_array_len, ERR_INVALID_PARAMETER );
|
|
|
|
|
|
DVector<Vector3>::Read read = array.read();
|
|
const Vector3* src=read.ptr();
|
|
|
|
// setting vertices means regenerating the AABB
|
|
AABB aabb;
|
|
|
|
|
|
if (p_format&ARRAY_COMPRESS_VERTEX) {
|
|
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
|
|
uint16_t vector[3]={ Math::make_half_float(src[i].x), Math::make_half_float(src[i].y), Math::make_half_float(src[i].z) };
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], vector, sizeof(uint16_t)*3);
|
|
|
|
if (i==0) {
|
|
|
|
aabb=AABB(src[i],Vector3());
|
|
} else {
|
|
|
|
aabb.expand_to( src[i] );
|
|
}
|
|
}
|
|
|
|
|
|
} else {
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
|
|
float vector[3]={ src[i].x, src[i].y, src[i].z };
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], vector, sizeof(float)*3);
|
|
|
|
if (i==0) {
|
|
|
|
aabb=AABB(src[i],Vector3());
|
|
} else {
|
|
|
|
aabb.expand_to( src[i] );
|
|
}
|
|
}
|
|
}
|
|
|
|
r_aabb=aabb;
|
|
|
|
}
|
|
|
|
|
|
} break;
|
|
case VS::ARRAY_NORMAL: {
|
|
|
|
ERR_FAIL_COND_V( p_arrays[ai].get_type() != Variant::VECTOR3_ARRAY, ERR_INVALID_PARAMETER );
|
|
|
|
DVector<Vector3> array = p_arrays[ai];
|
|
ERR_FAIL_COND_V( array.size() != p_vertex_array_len, ERR_INVALID_PARAMETER );
|
|
|
|
|
|
DVector<Vector3>::Read read = array.read();
|
|
const Vector3* src=read.ptr();
|
|
|
|
// setting vertices means regenerating the AABB
|
|
|
|
if (p_format&ARRAY_COMPRESS_NORMAL) {
|
|
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
uint8_t vector[4]={
|
|
CLAMP(src[i].x*127,-128,127),
|
|
CLAMP(src[i].y*127,-128,127),
|
|
CLAMP(src[i].z*127,-128,127),
|
|
0,
|
|
};
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], vector, 4);
|
|
|
|
}
|
|
|
|
} else {
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
|
|
float vector[3]={ src[i].x, src[i].y, src[i].z };
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], vector, 3*4);
|
|
|
|
}
|
|
}
|
|
|
|
} break;
|
|
|
|
case VS::ARRAY_TANGENT: {
|
|
|
|
ERR_FAIL_COND_V( p_arrays[ai].get_type() != Variant::REAL_ARRAY, ERR_INVALID_PARAMETER );
|
|
|
|
DVector<real_t> array = p_arrays[ai];
|
|
|
|
ERR_FAIL_COND_V( array.size() != p_vertex_array_len*4, ERR_INVALID_PARAMETER );
|
|
|
|
|
|
DVector<real_t>::Read read = array.read();
|
|
const real_t* src = read.ptr();
|
|
|
|
if (p_format&ARRAY_COMPRESS_TANGENT) {
|
|
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
uint8_t xyzw[4]={
|
|
CLAMP(src[i*4+0]*127,-128,127),
|
|
CLAMP(src[i*4+1]*127,-128,127),
|
|
CLAMP(src[i*4+2]*127,-128,127),
|
|
CLAMP(src[i*4+3]*127,-128,127)
|
|
};
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], xyzw, 4);
|
|
|
|
}
|
|
|
|
|
|
} else {
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
float xyzw[4]={
|
|
src[i*4+0],
|
|
src[i*4+1],
|
|
src[i*4+2],
|
|
src[i*4+3]
|
|
};
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], xyzw, 4*4);
|
|
|
|
}
|
|
}
|
|
|
|
} break;
|
|
case VS::ARRAY_COLOR: {
|
|
|
|
ERR_FAIL_COND_V( p_arrays[ai].get_type() != Variant::COLOR_ARRAY, ERR_INVALID_PARAMETER );
|
|
|
|
|
|
DVector<Color> array = p_arrays[ai];
|
|
|
|
ERR_FAIL_COND_V( array.size() != p_vertex_array_len, ERR_INVALID_PARAMETER );
|
|
|
|
|
|
DVector<Color>::Read read = array.read();
|
|
const Color* src = read.ptr();
|
|
|
|
if (p_format&ARRAY_COMPRESS_COLOR) {
|
|
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
|
|
uint8_t colors[4];
|
|
|
|
for(int j=0;j<4;j++) {
|
|
|
|
colors[j]=CLAMP( int((src[i][j])*255.0), 0,255 );
|
|
}
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], colors, 4);
|
|
|
|
}
|
|
} else {
|
|
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], &src[i], 4*4);
|
|
}
|
|
|
|
}
|
|
|
|
|
|
} break;
|
|
case VS::ARRAY_TEX_UV: {
|
|
|
|
ERR_FAIL_COND_V( p_arrays[ai].get_type() != Variant::VECTOR3_ARRAY && p_arrays[ai].get_type() != Variant::VECTOR2_ARRAY, ERR_INVALID_PARAMETER );
|
|
|
|
DVector<Vector2> array = p_arrays[ai];
|
|
|
|
ERR_FAIL_COND_V( array.size() != p_vertex_array_len , ERR_INVALID_PARAMETER);
|
|
|
|
DVector<Vector2>::Read read = array.read();
|
|
|
|
const Vector2 * src=read.ptr();
|
|
|
|
|
|
|
|
if (p_format&ARRAY_COMPRESS_TEX_UV) {
|
|
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
uint16_t uv[2]={ Math::make_half_float(src[i].x) , Math::make_half_float(src[i].y) };
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], uv, 2*2);
|
|
}
|
|
|
|
} else {
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
float uv[2]={ src[i].x , src[i].y };
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], uv, 2*4);
|
|
|
|
}
|
|
}
|
|
|
|
|
|
} break;
|
|
|
|
case VS::ARRAY_TEX_UV2: {
|
|
|
|
|
|
ERR_FAIL_COND_V( p_arrays[ai].get_type() != Variant::VECTOR3_ARRAY && p_arrays[ai].get_type() != Variant::VECTOR2_ARRAY, ERR_INVALID_PARAMETER );
|
|
|
|
DVector<Vector2> array = p_arrays[ai];
|
|
|
|
ERR_FAIL_COND_V( array.size() != p_vertex_array_len , ERR_INVALID_PARAMETER);
|
|
|
|
DVector<Vector2>::Read read = array.read();
|
|
|
|
const Vector2 * src=read.ptr();
|
|
|
|
|
|
|
|
if (p_format&ARRAY_COMPRESS_TEX_UV2) {
|
|
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
uint16_t uv[2]={ Math::make_half_float(src[i].x) , Math::make_half_float(src[i].y) };
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], uv, 2*2);
|
|
}
|
|
|
|
} else {
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
float uv[2]={ src[i].x , src[i].y };
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], uv, 2*4);
|
|
|
|
}
|
|
}
|
|
} break;
|
|
case VS::ARRAY_WEIGHTS: {
|
|
|
|
ERR_FAIL_COND_V( p_arrays[ai].get_type() != Variant::REAL_ARRAY, ERR_INVALID_PARAMETER );
|
|
|
|
DVector<real_t> array = p_arrays[ai];
|
|
|
|
ERR_FAIL_COND_V( array.size() != p_vertex_array_len*VS::ARRAY_WEIGHTS_SIZE, ERR_INVALID_PARAMETER );
|
|
|
|
|
|
DVector<real_t>::Read read = array.read();
|
|
|
|
const real_t * src = read.ptr();
|
|
|
|
if (p_format&ARRAY_COMPRESS_WEIGHTS) {
|
|
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
uint16_t data[VS::ARRAY_WEIGHTS_SIZE];
|
|
for (int j=0;j<VS::ARRAY_WEIGHTS_SIZE;j++) {
|
|
data[j]=CLAMP(src[i*VS::ARRAY_WEIGHTS_SIZE+j]*65535,0,65535);
|
|
}
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], data, 2*4);
|
|
}
|
|
} else {
|
|
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
float data[VS::ARRAY_WEIGHTS_SIZE];
|
|
for (int j=0;j<VS::ARRAY_WEIGHTS_SIZE;j++) {
|
|
data[j]=src[i*VS::ARRAY_WEIGHTS_SIZE+j];
|
|
}
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], data, 4*4);
|
|
|
|
|
|
}
|
|
}
|
|
|
|
} break;
|
|
case VS::ARRAY_BONES: {
|
|
|
|
ERR_FAIL_COND_V( p_arrays[ai].get_type() != Variant::REAL_ARRAY, ERR_INVALID_PARAMETER );
|
|
|
|
DVector<int> array = p_arrays[ai];
|
|
|
|
ERR_FAIL_COND_V( array.size() != p_vertex_array_len*VS::ARRAY_WEIGHTS_SIZE, ERR_INVALID_PARAMETER );
|
|
|
|
|
|
DVector<int>::Read read = array.read();
|
|
|
|
const int * src = read.ptr();
|
|
|
|
|
|
if (!(p_format&ARRAY_FLAG_USE_16_BIT_BONES)) {
|
|
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
uint8_t data[VS::ARRAY_WEIGHTS_SIZE];
|
|
for (int j=0;j<VS::ARRAY_WEIGHTS_SIZE;j++) {
|
|
data[j]=CLAMP(src[i*VS::ARRAY_WEIGHTS_SIZE+j],0,255);
|
|
max_bone=MAX(data[j],max_bone);
|
|
|
|
}
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], data, 4);
|
|
|
|
|
|
}
|
|
|
|
} else {
|
|
for (int i=0;i<p_vertex_array_len;i++) {
|
|
|
|
uint16_t data[VS::ARRAY_WEIGHTS_SIZE];
|
|
for (int j=0;j<VS::ARRAY_WEIGHTS_SIZE;j++) {
|
|
data[j]=src[i*VS::ARRAY_WEIGHTS_SIZE+j];
|
|
max_bone=MAX(data[j],max_bone);
|
|
|
|
}
|
|
|
|
copymem(&vw[p_offsets[ai]+i*p_stride], data, 2*4);
|
|
|
|
|
|
}
|
|
}
|
|
|
|
|
|
} break;
|
|
case VS::ARRAY_INDEX: {
|
|
|
|
|
|
ERR_FAIL_COND_V( p_index_array_len<=0, ERR_INVALID_DATA );
|
|
ERR_FAIL_COND_V( p_arrays[ai].get_type() != Variant::INT_ARRAY, ERR_INVALID_PARAMETER );
|
|
|
|
DVector<int> indices = p_arrays[ai];
|
|
ERR_FAIL_COND_V( indices.size() == 0, ERR_INVALID_PARAMETER );
|
|
ERR_FAIL_COND_V( indices.size() != p_index_array_len, ERR_INVALID_PARAMETER );
|
|
|
|
/* determine wether using 16 or 32 bits indices */
|
|
|
|
DVector<int>::Read read = indices.read();
|
|
const int *src=read.ptr();
|
|
|
|
for (int i=0;i<p_index_array_len;i++) {
|
|
|
|
|
|
if (p_vertex_array_len<(1<<16)) {
|
|
uint16_t v=src[i];
|
|
|
|
copymem(&iw[i*2], &v, 2);
|
|
} else {
|
|
uint32_t v=src[i];
|
|
|
|
copymem(&iw[i*4], &v, 4);
|
|
}
|
|
}
|
|
} break;
|
|
default: {
|
|
ERR_FAIL_V( ERR_INVALID_DATA );
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
if (p_format&VS::ARRAY_FORMAT_BONES) {
|
|
//create AABBs for each detected bone
|
|
int total_bones = max_bone+1;
|
|
|
|
bool first = r_bone_aabb.size()==0;
|
|
|
|
r_bone_aabb.resize(total_bones);
|
|
|
|
if (first) {
|
|
for(int i=0;i<total_bones;i++) {
|
|
r_bone_aabb[i].size==Vector3(-1,-1,-1); //negative means unused
|
|
}
|
|
}
|
|
|
|
DVector<Vector3> vertices = p_arrays[VS::ARRAY_VERTEX];
|
|
DVector<int> bones = p_arrays[VS::ARRAY_BONES];
|
|
DVector<float> weights = p_arrays[VS::ARRAY_WEIGHTS];
|
|
|
|
bool any_valid=false;
|
|
|
|
if (vertices.size() && bones.size()==vertices.size()*4 && weights.size()==bones.size()) {
|
|
|
|
int vs = vertices.size();
|
|
DVector<Vector3>::Read rv =vertices.read();
|
|
DVector<int>::Read rb=bones.read();
|
|
DVector<float>::Read rw=weights.read();
|
|
|
|
AABB *bptr = r_bone_aabb.ptr();
|
|
|
|
for(int i=0;i<vs;i++) {
|
|
|
|
Vector3 v = rv[i];
|
|
for(int j=0;j<4;j++) {
|
|
|
|
int idx = rb[i*4+j];
|
|
float w = rw[i*4+j];
|
|
if (w==0)
|
|
continue;//break;
|
|
ERR_FAIL_INDEX_V(idx,total_bones,ERR_INVALID_DATA);
|
|
|
|
if (bptr->size.x<0) {
|
|
//first
|
|
bptr[idx]=AABB();
|
|
bptr[idx].pos=v;
|
|
any_valid=true;
|
|
} else {
|
|
bptr[idx].expand_to(v);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
if (!any_valid && first) {
|
|
|
|
r_bone_aabb.clear();
|
|
}
|
|
}
|
|
return OK;
|
|
}
|
|
|
|
|
|
void VisualServer::mesh_add_surface_from_arrays(RID p_mesh,PrimitiveType p_primitive,const Array& p_arrays,const Array& p_blend_shapes,uint32_t p_compress_format) {
|
|
|
|
ERR_FAIL_INDEX( p_primitive, VS::PRIMITIVE_MAX );
|
|
ERR_FAIL_COND(p_arrays.size()!=VS::ARRAY_MAX);
|
|
|
|
uint32_t format=0;
|
|
|
|
// validation
|
|
int index_array_len=0;
|
|
int array_len=0;
|
|
|
|
for(int i=0;i<p_arrays.size();i++) {
|
|
|
|
if (p_arrays[i].get_type()==Variant::NIL)
|
|
continue;
|
|
|
|
format|=(1<<i);
|
|
|
|
if (i==VS::ARRAY_VERTEX) {
|
|
|
|
Variant var = p_arrays[i];
|
|
switch(var.get_type()) {
|
|
case Variant::VECTOR2_ARRAY: {
|
|
DVector<Vector2> v2 = var;
|
|
array_len=v2.size();
|
|
} break;
|
|
case Variant::VECTOR3_ARRAY: {
|
|
DVector<Vector3> v3 = var;
|
|
array_len=v3.size();
|
|
} break;
|
|
default: {
|
|
Array v = var;
|
|
array_len=v.size();
|
|
} break;
|
|
}
|
|
|
|
array_len=Vector3Array(p_arrays[i]).size();
|
|
ERR_FAIL_COND(array_len==0);
|
|
} else if (i==VS::ARRAY_INDEX) {
|
|
|
|
index_array_len=IntArray(p_arrays[i]).size();
|
|
}
|
|
}
|
|
|
|
ERR_FAIL_COND((format&VS::ARRAY_FORMAT_VERTEX)==0); // mandatory
|
|
|
|
|
|
if (p_blend_shapes.size()) {
|
|
//validate format for morphs
|
|
for(int i=0;i<p_blend_shapes.size();i++) {
|
|
|
|
uint32_t bsformat=0;
|
|
Array arr = p_blend_shapes[i];
|
|
for(int j=0;j<arr.size();j++) {
|
|
|
|
|
|
if (arr[j].get_type()!=Variant::NIL)
|
|
bsformat|=(1<<j);
|
|
}
|
|
|
|
ERR_FAIL_COND( (bsformat)!=(format&(VS::ARRAY_FORMAT_BONES-1)));
|
|
}
|
|
}
|
|
|
|
uint32_t offsets[VS::ARRAY_MAX];
|
|
|
|
int total_elem_size=0;
|
|
|
|
for (int i=0;i<VS::ARRAY_MAX;i++) {
|
|
|
|
|
|
offsets[i]=0; //reset
|
|
|
|
if (!(format&(1<<i))) // no array
|
|
continue;
|
|
|
|
|
|
int elem_size=0;
|
|
|
|
switch(i) {
|
|
|
|
case VS::ARRAY_VERTEX: {
|
|
|
|
Variant arr = p_arrays[0];
|
|
if (arr.get_type()==Variant::VECTOR2_ARRAY) {
|
|
elem_size=2;
|
|
p_compress_format|=ARRAY_FLAG_USE_2D_VERTICES;
|
|
} else if (arr.get_type()==Variant::VECTOR3_ARRAY) {
|
|
p_compress_format&=~ARRAY_FLAG_USE_2D_VERTICES;
|
|
elem_size=3;
|
|
} else {
|
|
elem_size=(p_compress_format&ARRAY_FLAG_USE_2D_VERTICES)?2:3;
|
|
}
|
|
|
|
if (p_compress_format&ARRAY_COMPRESS_VERTEX) {
|
|
elem_size*=sizeof(int16_t);
|
|
} else {
|
|
elem_size*=sizeof(float);
|
|
}
|
|
|
|
} break;
|
|
case VS::ARRAY_NORMAL: {
|
|
|
|
if (p_compress_format&ARRAY_COMPRESS_NORMAL) {
|
|
elem_size=sizeof(uint32_t);
|
|
} else {
|
|
elem_size=sizeof(float)*3;
|
|
}
|
|
|
|
} break;
|
|
|
|
case VS::ARRAY_TANGENT: {
|
|
if (p_compress_format&ARRAY_COMPRESS_TANGENT) {
|
|
elem_size=sizeof(uint32_t);
|
|
} else {
|
|
elem_size=sizeof(float)*4;
|
|
}
|
|
|
|
} break;
|
|
case VS::ARRAY_COLOR: {
|
|
|
|
if (p_compress_format&ARRAY_COMPRESS_COLOR) {
|
|
elem_size=sizeof(uint32_t);
|
|
} else {
|
|
elem_size=sizeof(float)*4;
|
|
}
|
|
} break;
|
|
case VS::ARRAY_TEX_UV: {
|
|
if (p_compress_format&ARRAY_COMPRESS_TEX_UV) {
|
|
elem_size=sizeof(uint32_t);
|
|
} else {
|
|
elem_size=sizeof(float)*2;
|
|
}
|
|
|
|
} break;
|
|
|
|
case VS::ARRAY_TEX_UV2: {
|
|
if (p_compress_format&ARRAY_COMPRESS_TEX_UV2) {
|
|
elem_size=sizeof(uint32_t);
|
|
} else {
|
|
elem_size=sizeof(float)*2;
|
|
}
|
|
|
|
} break;
|
|
case VS::ARRAY_WEIGHTS: {
|
|
|
|
if (p_compress_format&ARRAY_COMPRESS_WEIGHTS) {
|
|
elem_size=sizeof(uint16_t)*4;
|
|
} else {
|
|
elem_size=sizeof(float)*4;
|
|
}
|
|
|
|
} break;
|
|
case VS::ARRAY_BONES: {
|
|
|
|
DVector<int> bones = p_arrays[VS::ARRAY_BONES];
|
|
int max_bone=0;
|
|
|
|
{
|
|
int bc = bones.size();
|
|
DVector<int>::Read r=bones.read();
|
|
for(int j=0;j<bc;j++) {
|
|
max_bone=MAX(r[j],max_bone);
|
|
}
|
|
}
|
|
|
|
if (max_bone > 255) {
|
|
p_compress_format|=ARRAY_FLAG_USE_16_BIT_BONES;
|
|
elem_size=sizeof(uint16_t)*4;
|
|
} else {
|
|
p_compress_format&=~ARRAY_FLAG_USE_16_BIT_BONES;
|
|
elem_size=sizeof(uint32_t);
|
|
}
|
|
|
|
|
|
} break;
|
|
case VS::ARRAY_INDEX: {
|
|
|
|
if (index_array_len<=0) {
|
|
ERR_PRINT("index_array_len==NO_INDEX_ARRAY");
|
|
break;
|
|
}
|
|
/* determine wether using 16 or 32 bits indices */
|
|
if (array_len>=(1<<16)) {
|
|
|
|
elem_size=4;
|
|
|
|
} else {
|
|
elem_size=2;
|
|
}
|
|
offsets[i]=elem_size;
|
|
continue;
|
|
} break;
|
|
default: {
|
|
ERR_FAIL( );
|
|
}
|
|
}
|
|
|
|
offsets[i]=total_elem_size;
|
|
total_elem_size+=elem_size;
|
|
|
|
|
|
}
|
|
|
|
uint32_t mask = (1<<ARRAY_MAX)-1;
|
|
format|=(~mask)&p_compress_format; //make the full format
|
|
|
|
|
|
int array_size = total_elem_size * array_len;
|
|
|
|
DVector<uint8_t> vertex_array;
|
|
vertex_array.resize(array_size);
|
|
|
|
int index_array_size = offsets[VS::ARRAY_INDEX]*index_array_len;
|
|
|
|
DVector<uint8_t> index_array;
|
|
index_array.resize(index_array_size);
|
|
|
|
AABB aabb;
|
|
Vector<AABB> bone_aabb;
|
|
|
|
Error err = _surface_set_data(p_arrays,format,offsets,total_elem_size,vertex_array,array_len,index_array,index_array_len,aabb,bone_aabb);
|
|
|
|
if (err) {
|
|
ERR_EXPLAIN("Invalid array format for surface");
|
|
ERR_FAIL_COND(err!=OK);
|
|
}
|
|
|
|
Vector<DVector<uint8_t> > blend_shape_data;
|
|
|
|
for(int i=0;i<p_blend_shapes.size();i++) {
|
|
|
|
DVector<uint8_t> vertex_array_shape;
|
|
vertex_array_shape.resize(array_size);
|
|
DVector<uint8_t> noindex;
|
|
|
|
AABB laabb;
|
|
Error err = _surface_set_data(p_blend_shapes[i],format&~ARRAY_FORMAT_INDEX,offsets,total_elem_size,vertex_array,array_len,noindex,0,laabb,bone_aabb);
|
|
aabb.merge_with(laabb);
|
|
if (err) {
|
|
ERR_EXPLAIN("Invalid blend shape array format for surface");
|
|
ERR_FAIL_COND(err!=OK);
|
|
}
|
|
|
|
blend_shape_data.push_back(vertex_array_shape);
|
|
}
|
|
|
|
mesh_add_surface(p_mesh,format,p_primitive,vertex_array,array_len,index_array,index_array_len,aabb,blend_shape_data,bone_aabb);
|
|
|
|
}
|
|
|
|
Array VisualServer::_get_array_from_surface(uint32_t p_format,DVector<uint8_t> p_vertex_data,int p_vertex_len,DVector<uint8_t> p_index_data,int p_index_len) const {
|
|
|
|
|
|
uint32_t offsets[ARRAY_MAX];
|
|
|
|
int total_elem_size=0;
|
|
|
|
for (int i=0;i<VS::ARRAY_MAX;i++) {
|
|
|
|
|
|
offsets[i]=0; //reset
|
|
|
|
if (!(p_format&(1<<i))) // no array
|
|
continue;
|
|
|
|
|
|
int elem_size=0;
|
|
|
|
switch(i) {
|
|
|
|
case VS::ARRAY_VERTEX: {
|
|
|
|
|
|
if (p_format&ARRAY_FLAG_USE_2D_VERTICES) {
|
|
elem_size=2;
|
|
} else {
|
|
elem_size=3;
|
|
}
|
|
|
|
if (p_format&ARRAY_COMPRESS_VERTEX) {
|
|
elem_size*=sizeof(int16_t);
|
|
} else {
|
|
elem_size*=sizeof(float);
|
|
}
|
|
|
|
} break;
|
|
case VS::ARRAY_NORMAL: {
|
|
|
|
if (p_format&ARRAY_COMPRESS_NORMAL) {
|
|
elem_size=sizeof(uint32_t);
|
|
} else {
|
|
elem_size=sizeof(float)*3;
|
|
}
|
|
|
|
} break;
|
|
|
|
case VS::ARRAY_TANGENT: {
|
|
if (p_format&ARRAY_COMPRESS_TANGENT) {
|
|
elem_size=sizeof(uint32_t);
|
|
} else {
|
|
elem_size=sizeof(float)*4;
|
|
}
|
|
|
|
} break;
|
|
case VS::ARRAY_COLOR: {
|
|
|
|
if (p_format&ARRAY_COMPRESS_COLOR) {
|
|
elem_size=sizeof(uint32_t);
|
|
} else {
|
|
elem_size=sizeof(float)*4;
|
|
}
|
|
} break;
|
|
case VS::ARRAY_TEX_UV: {
|
|
if (p_format&ARRAY_COMPRESS_TEX_UV) {
|
|
elem_size=sizeof(uint32_t);
|
|
} else {
|
|
elem_size=sizeof(float)*2;
|
|
}
|
|
|
|
} break;
|
|
|
|
case VS::ARRAY_TEX_UV2: {
|
|
if (p_format&ARRAY_COMPRESS_TEX_UV2) {
|
|
elem_size=sizeof(uint32_t);
|
|
} else {
|
|
elem_size=sizeof(float)*2;
|
|
}
|
|
|
|
} break;
|
|
case VS::ARRAY_WEIGHTS: {
|
|
|
|
if (p_format&ARRAY_COMPRESS_WEIGHTS) {
|
|
elem_size=sizeof(uint16_t)*4;
|
|
} else {
|
|
elem_size=sizeof(float)*4;
|
|
}
|
|
|
|
} break;
|
|
case VS::ARRAY_BONES: {
|
|
|
|
if (p_format&ARRAY_FLAG_USE_16_BIT_BONES) {
|
|
elem_size=sizeof(uint32_t);
|
|
} else {
|
|
elem_size=sizeof(uint16_t)*4;
|
|
}
|
|
|
|
} break;
|
|
case VS::ARRAY_INDEX: {
|
|
|
|
if (p_index_len<=0) {
|
|
ERR_PRINT("index_array_len==NO_INDEX_ARRAY");
|
|
break;
|
|
}
|
|
/* determine wether using 16 or 32 bits indices */
|
|
if (p_index_len>=(1<<16)) {
|
|
|
|
elem_size=4;
|
|
|
|
} else {
|
|
elem_size=2;
|
|
}
|
|
offsets[i]=elem_size;
|
|
continue;
|
|
} break;
|
|
default: {
|
|
ERR_FAIL_V( Array() );
|
|
}
|
|
}
|
|
|
|
offsets[i]=total_elem_size;
|
|
total_elem_size+=elem_size;
|
|
|
|
|
|
}
|
|
|
|
Array ret;
|
|
ret.resize(VS::ARRAY_MAX);
|
|
|
|
DVector<uint8_t>::Read r = p_vertex_data.read();
|
|
|
|
for(int i=0;i<VS::ARRAY_MAX;i++) {
|
|
|
|
if (!(p_format&(1<<i)))
|
|
continue;
|
|
|
|
|
|
switch(i) {
|
|
|
|
case VS::ARRAY_VERTEX: {
|
|
|
|
|
|
if (p_format&ARRAY_FLAG_USE_2D_VERTICES) {
|
|
|
|
DVector<Vector2> arr_2d;
|
|
arr_2d.resize(p_vertex_len);
|
|
|
|
if (p_format&ARRAY_COMPRESS_VERTEX) {
|
|
|
|
DVector<Vector2>::Write w = arr_2d.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const uint16_t *v = (const uint16_t*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Vector2(Math::halfptr_to_float(&v[0]),Math::halfptr_to_float(&v[1]));
|
|
}
|
|
} else {
|
|
|
|
DVector<Vector2>::Write w = arr_2d.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const float *v = (const float*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Vector2(v[0],v[1]);
|
|
}
|
|
}
|
|
|
|
ret[i]=arr_2d;
|
|
} else {
|
|
|
|
DVector<Vector3> arr_3d;
|
|
arr_3d.resize(p_vertex_len);
|
|
|
|
if (p_format&ARRAY_COMPRESS_VERTEX) {
|
|
|
|
DVector<Vector3>::Write w = arr_3d.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const uint16_t *v = (const uint16_t*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Vector3(Math::halfptr_to_float(&v[0]),Math::halfptr_to_float(&v[1]),Math::halfptr_to_float(&v[2]));
|
|
}
|
|
} else {
|
|
|
|
DVector<Vector3>::Write w = arr_3d.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const float *v = (const float*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Vector3(v[0],v[1],v[2]);
|
|
}
|
|
}
|
|
|
|
ret[i]=arr_3d;
|
|
}
|
|
|
|
|
|
} break;
|
|
case VS::ARRAY_NORMAL: {
|
|
DVector<Vector3> arr;
|
|
arr.resize(p_vertex_len);
|
|
|
|
if (p_format&ARRAY_COMPRESS_NORMAL) {
|
|
|
|
DVector<Vector3>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const uint8_t *v = (const uint8_t*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Vector3( float(v[0]/255.0)*2.0-1.0, float(v[1]/255.0)*2.0-1.0, float(v[2]/255.0)*2.0-1.0 );
|
|
}
|
|
} else {
|
|
DVector<Vector3>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const float *v = (const float*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Vector3(v[0],v[1],v[2]);
|
|
}
|
|
}
|
|
|
|
ret[i]=arr;
|
|
|
|
} break;
|
|
|
|
case VS::ARRAY_TANGENT: {
|
|
DVector<float> arr;
|
|
arr.resize(p_vertex_len*4);
|
|
if (p_format&ARRAY_COMPRESS_TANGENT) {
|
|
DVector<float>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const uint8_t *v = (const uint8_t*)&r[j*total_elem_size+offsets[i]];
|
|
for(int k=0;k<4;k++) {
|
|
w[j*4+k]=float(v[k]/255.0)*2.0-1.0;
|
|
}
|
|
}
|
|
} else {
|
|
|
|
DVector<float>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
const float *v = (const float*)&r[j*total_elem_size+offsets[i]];
|
|
for(int k=0;k<4;k++) {
|
|
w[j*4+k]=v[k];
|
|
}
|
|
}
|
|
|
|
}
|
|
|
|
ret[i]=arr;
|
|
|
|
} break;
|
|
case VS::ARRAY_COLOR: {
|
|
|
|
DVector<Color> arr;
|
|
arr.resize(p_vertex_len);
|
|
|
|
if (p_format&ARRAY_COMPRESS_COLOR) {
|
|
|
|
DVector<Color>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const uint8_t *v = (const uint8_t*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Color( float(v[0]/255.0)*2.0-1.0, float(v[1]/255.0)*2.0-1.0, float(v[2]/255.0)*2.0-1.0, float(v[3]/255.0)*2.0-1.0 );
|
|
}
|
|
} else {
|
|
DVector<Color>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const float *v = (const float*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Color(v[0],v[1],v[2],v[3]);
|
|
}
|
|
}
|
|
|
|
ret[i]=arr;
|
|
} break;
|
|
case VS::ARRAY_TEX_UV: {
|
|
|
|
DVector<Vector2> arr;
|
|
arr.resize(p_vertex_len);
|
|
|
|
if (p_format&ARRAY_COMPRESS_TEX_UV) {
|
|
|
|
DVector<Vector2>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const uint16_t *v = (const uint16_t*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Vector2(Math::halfptr_to_float(&v[0]),Math::halfptr_to_float(&v[1]));
|
|
}
|
|
} else {
|
|
|
|
DVector<Vector2>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const float *v = (const float*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Vector2(v[0],v[1]);
|
|
}
|
|
}
|
|
|
|
ret[i]=arr;
|
|
} break;
|
|
|
|
case VS::ARRAY_TEX_UV2: {
|
|
DVector<Vector2> arr;
|
|
arr.resize(p_vertex_len);
|
|
|
|
if (p_format&ARRAY_COMPRESS_TEX_UV2) {
|
|
|
|
DVector<Vector2>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const uint16_t *v = (const uint16_t*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Vector2(Math::halfptr_to_float(&v[0]),Math::halfptr_to_float(&v[1]));
|
|
}
|
|
} else {
|
|
|
|
DVector<Vector2>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const float *v = (const float*)&r[j*total_elem_size+offsets[i]];
|
|
w[j]=Vector2(v[0],v[1]);
|
|
}
|
|
}
|
|
|
|
ret[i]=arr;
|
|
|
|
} break;
|
|
case VS::ARRAY_WEIGHTS: {
|
|
|
|
DVector<float> arr;
|
|
arr.resize(p_vertex_len*4);
|
|
if (p_format&ARRAY_COMPRESS_WEIGHTS) {
|
|
DVector<float>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const uint16_t *v = (const uint16_t*)&r[j*total_elem_size+offsets[i]];
|
|
for(int k=0;k<4;k++) {
|
|
w[j*4+k]=float(v[k]/65535.0)*2.0-1.0;
|
|
}
|
|
}
|
|
} else {
|
|
|
|
DVector<float>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
const float *v = (const float*)&r[j*total_elem_size+offsets[i]];
|
|
for(int k=0;k<4;k++) {
|
|
w[j*4+k]=v[k];
|
|
}
|
|
}
|
|
|
|
}
|
|
|
|
ret[i]=arr;
|
|
|
|
} break;
|
|
case VS::ARRAY_BONES: {
|
|
|
|
DVector<int> arr;
|
|
arr.resize(p_vertex_len*4);
|
|
if (p_format&ARRAY_FLAG_USE_16_BIT_BONES) {
|
|
|
|
DVector<int>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
|
|
const uint16_t *v = (const uint16_t*)&r[j*total_elem_size+offsets[i]];
|
|
for(int k=0;k<4;k++) {
|
|
w[j*4+k]=v[k];
|
|
}
|
|
}
|
|
} else {
|
|
|
|
DVector<int>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_vertex_len;j++) {
|
|
const int *v = (const int*)&r[j*total_elem_size+offsets[i]];
|
|
for(int k=0;k<4;k++) {
|
|
w[j*4+k]=v[k];
|
|
}
|
|
}
|
|
|
|
}
|
|
|
|
ret[i]=arr;
|
|
|
|
} break;
|
|
case VS::ARRAY_INDEX: {
|
|
/* determine wether using 16 or 32 bits indices */
|
|
|
|
DVector<uint8_t>::Read ir = p_index_data.read();
|
|
|
|
DVector<int> arr;
|
|
arr.resize(p_index_len);
|
|
if (p_index_len<(1<<16)) {
|
|
|
|
DVector<int>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_index_len;j++) {
|
|
|
|
const uint16_t *v = (const uint16_t*)&ir[j*2];
|
|
w[j]=*v;
|
|
}
|
|
} else {
|
|
|
|
DVector<int>::Write w = arr.write();
|
|
|
|
for(int j=0;j<p_index_len;j++) {
|
|
const int *v = (const int*)&ir[j*4];
|
|
w[j]=*v;
|
|
}
|
|
|
|
}
|
|
ret[i]=arr;
|
|
} break;
|
|
default: {
|
|
ERR_FAIL_V( ret );
|
|
}
|
|
}
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
|
|
Array VisualServer::mesh_surface_get_arrays(RID p_mesh,int p_surface) const {
|
|
|
|
DVector<uint8_t> vertex_data = mesh_surface_get_array(p_mesh,p_surface);
|
|
ERR_FAIL_COND_V(vertex_data.size()==0,Array());
|
|
int vertex_len = mesh_surface_get_array_len(p_mesh,p_surface);
|
|
|
|
DVector<uint8_t> index_data = mesh_surface_get_index_array(p_mesh,p_surface);
|
|
int index_len = mesh_surface_get_array_index_len(p_mesh,p_surface);
|
|
|
|
uint32_t format = mesh_surface_get_format(p_mesh,p_surface);
|
|
|
|
|
|
return _get_array_from_surface(format,vertex_data,vertex_len,index_data,index_len);
|
|
|
|
}
|
|
|
|
void VisualServer::_bind_methods() {
|
|
|
|
|
|
ObjectTypeDB::bind_method(_MD("texture_create"),&VisualServer::texture_create);
|
|
ObjectTypeDB::bind_method(_MD("texture_create_from_image"),&VisualServer::texture_create_from_image,DEFVAL( TEXTURE_FLAGS_DEFAULT ) );
|
|
//ObjectTypeDB::bind_method(_MD("texture_allocate"),&VisualServer::texture_allocate,DEFVAL( TEXTURE_FLAGS_DEFAULT ) );
|
|
//ObjectTypeDB::bind_method(_MD("texture_set_data"),&VisualServer::texture_blit_rect,DEFVAL( CUBEMAP_LEFT ) );
|
|
//ObjectTypeDB::bind_method(_MD("texture_get_rect"),&VisualServer::texture_get_rect );
|
|
ObjectTypeDB::bind_method(_MD("texture_set_flags"),&VisualServer::texture_set_flags );
|
|
ObjectTypeDB::bind_method(_MD("texture_get_flags"),&VisualServer::texture_get_flags );
|
|
ObjectTypeDB::bind_method(_MD("texture_get_width"),&VisualServer::texture_get_width );
|
|
ObjectTypeDB::bind_method(_MD("texture_get_height"),&VisualServer::texture_get_height );
|
|
|
|
ObjectTypeDB::bind_method(_MD("texture_set_shrink_all_x2_on_set_data","shrink"),&VisualServer::texture_set_shrink_all_x2_on_set_data );
|
|
|
|
|
|
|
|
|
|
}
|
|
|
|
void VisualServer::_canvas_item_add_style_box(RID p_item, const Rect2& p_rect, const Rect2& p_source, RID p_texture,const Vector<float>& p_margins, const Color& p_modulate) {
|
|
|
|
ERR_FAIL_COND(p_margins.size()!=4);
|
|
//canvas_item_add_style_box(p_item,p_rect,p_source,p_texture,Vector2(p_margins[0],p_margins[1]),Vector2(p_margins[2],p_margins[3]),true,p_modulate);
|
|
}
|
|
|
|
void VisualServer::_camera_set_orthogonal(RID p_camera,float p_size,float p_z_near,float p_z_far) {
|
|
|
|
camera_set_orthogonal(p_camera,p_size,p_z_near,p_z_far);
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
void VisualServer::mesh_add_surface_from_mesh_data( RID p_mesh, const Geometry::MeshData& p_mesh_data) {
|
|
|
|
#if 1
|
|
DVector<Vector3> vertices;
|
|
DVector<Vector3> normals;
|
|
|
|
for (int i=0;i<p_mesh_data.faces.size();i++) {
|
|
|
|
const Geometry::MeshData::Face& f = p_mesh_data.faces[i];
|
|
|
|
for (int j=2;j<f.indices.size();j++) {
|
|
|
|
#define _ADD_VERTEX(m_idx)\
|
|
vertices.push_back( p_mesh_data.vertices[ f.indices[m_idx] ] );\
|
|
normals.push_back( f.plane.normal );
|
|
|
|
_ADD_VERTEX( 0 );
|
|
_ADD_VERTEX( j-1 );
|
|
_ADD_VERTEX( j );
|
|
}
|
|
}
|
|
|
|
Array d;
|
|
d.resize(VS::ARRAY_MAX);
|
|
d[ARRAY_VERTEX]=vertices;
|
|
d[ARRAY_NORMAL]=normals;
|
|
mesh_add_surface_from_arrays(p_mesh,PRIMITIVE_TRIANGLES, d);
|
|
|
|
#else
|
|
|
|
|
|
DVector<Vector3> vertices;
|
|
|
|
|
|
|
|
for (int i=0;i<p_mesh_data.edges.size();i++) {
|
|
|
|
const Geometry::MeshData::Edge& f = p_mesh_data.edges[i];
|
|
vertices.push_back(p_mesh_data.vertices[ f.a]);
|
|
vertices.push_back(p_mesh_data.vertices[ f.b]);
|
|
}
|
|
|
|
Array d;
|
|
d.resize(VS::ARRAY_MAX);
|
|
d[ARRAY_VERTEX]=vertices;
|
|
mesh_add_surface(p_mesh,PRIMITIVE_LINES, d);
|
|
|
|
|
|
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
void VisualServer::mesh_add_surface_from_planes( RID p_mesh, const DVector<Plane>& p_planes) {
|
|
|
|
|
|
Geometry::MeshData mdata = Geometry::build_convex_mesh(p_planes);
|
|
mesh_add_surface_from_mesh_data(p_mesh,mdata);
|
|
|
|
}
|
|
|
|
RID VisualServer::instance_create2(RID p_base, RID p_scenario) {
|
|
|
|
RID instance = instance_create();
|
|
instance_set_base(instance,p_base);
|
|
instance_set_scenario(instance,p_scenario);
|
|
return instance;
|
|
}
|
|
|
|
|
|
VisualServer::VisualServer() {
|
|
|
|
// ERR_FAIL_COND(singleton);
|
|
singleton=this;
|
|
|
|
}
|
|
|
|
|
|
VisualServer::~VisualServer() {
|
|
|
|
singleton=NULL;
|
|
}
|