#include <xtl.h>
#include <xgraphics.h>

#include <sys/timer.h>
#include <sys/file/filesys.h>
#include <core/macros.h>

#include <stdio.h>
#include <stdlib.h>
#include "nx_init.h"
#include "texture.h"
#include "scene.h"
#include "mesh.h"
#include "anim.h"
#include "render.h"
#include "billboard.h"

namespace NxXbox
{

bool			s_meshScalingEnabled = false;
char*			s_pWeightIndices = NULL;
float*			s_pWeights = NULL;
Mth::Vector*	s_pBonePositions = NULL;
Mth::Vector*	s_pBoneScales = NULL;
int				s_currentVertIndex = 0;

/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void SetMeshScalingParameters( Nx::SMeshScalingParameters* pParams )
{
	Dbg_Assert( pParams );

	s_meshScalingEnabled	= true;
	s_pWeights				= pParams->pWeights;
	s_pWeightIndices		= pParams->pWeightIndices;
	s_pBoneScales			= pParams->pBoneScales;
	s_pBonePositions		= pParams->pBonePositions;
	s_currentVertIndex		= 0;
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void DisableMeshScaling( void )
{
	s_meshScalingEnabled	= false;
	s_pWeights				= NULL;
	s_pWeightIndices		= NULL;
	s_pBoneScales			= NULL;
	s_pBonePositions		= NULL;
	s_currentVertIndex		= 0;
}

	
	
/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
static inline Mth::Vector get_bone_scale( int bone_index )
{
	Mth::Vector returnVec( 1.0f, 1.0f, 1.0f, 1.0f );

	if( bone_index >= 29 && bone_index <= 33 )
	{
		// this only works with the thps5 skeleton, whose
		// head bones are between 29 and 33...
		// (eventually, we can remove the subtract 29
		// once the exporter is massaging the data correctly)
		returnVec = s_pBoneScales[ bone_index - 29 ];
		
		// Y & Z are reversed...  odd!
		Mth::Vector tempVec = returnVec;
		returnVec[Y] = tempVec[Z];
		returnVec[Z] = tempVec[Y];
	}
	else if( bone_index == -1 )
	{
		// implies that it's not weighted to a bone
		return returnVec;
	}
	else
	{
		// implies that it's weighted to the wrong bone
		return returnVec;
	}
	return returnVec;
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
static inline Mth::Vector get_bone_pos( int bone_index )
{
	Mth::Vector returnVec( 0.0f, 0.0f, 0.0f, 1.0f );
	
	if( bone_index >= 29 && bone_index <= 33 )
	{
		// this only works with the thps5 skeleton, whose
		// head bones are between 29 and 33...
		// (eventually, we can remove the subtract 29
		// once the exporter is massaging the data correctly)
		returnVec = s_pBonePositions[ bone_index - 29 ];
	}
	else if( bone_index == -1 )
	{
		// implies that it's not weighted to a bone
		return returnVec;
	}
	else
	{
		// implies that it's weighted to the wrong bone
		return returnVec;
	}
	returnVec[W] = 1.0f;

	return returnVec;
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void ApplyMeshScaling( float* p_vertices, int num_verts )
{
	if( s_meshScalingEnabled )
	{
		for( int v = 0; v < num_verts; ++v, p_vertices += 3 )
		{
			float x = p_vertices[0];
			float y = p_vertices[1];
			float z = p_vertices[2];

    		Mth::Vector origPos( x, y, z, 1.0f );

			Mth::Vector bonePos0 = get_bone_pos( s_pWeightIndices[v * 3] );
			Mth::Vector bonePos1 = get_bone_pos( s_pWeightIndices[v * 3 + 1] );
			Mth::Vector bonePos2 = get_bone_pos( s_pWeightIndices[v * 3 + 2] );

			// Need to scale each vert relative to its parent bone.
			Mth::Vector localPos0 = origPos - bonePos0;
			Mth::Vector localPos1 = origPos - bonePos1;
			Mth::Vector localPos2 = origPos - bonePos2;

			localPos0.Scale( get_bone_scale( s_pWeightIndices[v * 3] ) );
			localPos1.Scale( get_bone_scale( s_pWeightIndices[v * 3 + 1] ) );
			localPos2.Scale( get_bone_scale( s_pWeightIndices[v * 3 + 2] ) );

			localPos0 += bonePos0;
			localPos1 += bonePos1;
			localPos2 += bonePos2;
			
			Mth::Vector scaledPos = ( localPos0 * s_pWeights[v * 3] ) +
									( localPos1 * s_pWeights[v * 3 + 1] ) +
									( localPos2 * s_pWeights[v * 3 + 2] );

			p_vertices[0] = scaledPos[X];
			p_vertices[1] = scaledPos[Y];
			p_vertices[2] = scaledPos[Z];
		}
	}
}

	
	
/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
sMesh::sMesh( void )
{
	m_flags							= 0;
	m_num_vertex_buffers			= 1;
	m_current_write_vertex_buffer	= 0;
	for( int vb = 0; vb < MAX_VERTEX_BUFFERS; ++vb )
	{
		mp_vertex_buffer[vb]		= NULL;
	}
	for( int ib = 0; ib < MAX_INDEX_BUFFERS; ++ib )
	{
		mp_index_buffer[ib]			= NULL;
		m_num_indices[ib]			= 0;
	}
	mp_vc_wibble_data				= NULL;
	mp_index_lod_data				= NULL;
	mp_billboard_data				= NULL;
	m_bone_index					= -1;
	mp_transform					= NULL;
	m_diffuse_offset				= 0;
	m_uv0_offset					= 0;
	m_normal_offset					= 0;
	m_vertex_stride					= 0;
	m_vertex_shader[0]				= 0;
	m_pixel_shader					= 0;

	SetActive( true );
	SetVisibility( 0xFF );

	// Set default primitive type.
	m_primitive_type				= D3DPT_TRIANGLESTRIP;
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
sMesh::~sMesh( void )
{
	// Remove this mesh from the billboard manager if appropriate.
	if( m_flags & sMesh::MESH_FLAG_BILLBOARD )
	{
		BillboardManager.RemoveEntry( this );
	}

	EngineGlobals.p_Device->BlockUntilIdle();

	if( mp_transform )
	{
		delete mp_transform;
	}
	
	if( !( m_flags & MESH_FLAG_IS_INSTANCE ))
	{
		for( int ib = 0; ib < MAX_INDEX_BUFFERS; ++ib )
		{
			delete [] mp_index_buffer[ib];
			mp_index_buffer[ib] = NULL;
		}

		if( mp_vc_wibble_data )
		{
			delete mp_vc_wibble_data;
			mp_vc_wibble_data = NULL;
		}

		if( mp_index_lod_data )
		{
			delete mp_index_lod_data;
			mp_index_lod_data = NULL;
		}

		if( mp_billboard_data )
		{
			delete mp_billboard_data;
			mp_billboard_data = NULL;
		}

		UINT					stride;
		IDirect3DVertexBuffer8	*p_vb;
		D3DDevice_GetStreamSource( 0, &p_vb, &stride );
		if( p_vb )
		{
			// GetStreamSource() increments the reference count, so call Release() here.
			p_vb->Release();
		}

		for( uint32 i = 0; i < m_num_vertex_buffers; ++i )
		{
			if( mp_vertex_buffer[i] )
			{
				if( p_vb == mp_vertex_buffer[i] )
				{
					// We are deleting a vertex buffer that is set as the current stream source. This can result in
					// problems with the internal D3D reference counter, so clear this up first.
					D3DDevice_SetStreamSource( 0, NULL, 0 );
				}
			
				uint8 *p_del = (uint8*)mp_vertex_buffer[i];
				delete p_del;
				mp_vertex_buffer[i]	= NULL;
			}
		}
	}
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::PushVertexShader( uint32 shader_id )
{
	for( uint32 i = sMesh::VERTEX_SHADER_STACK_SIZE - 1; i > 0; --i )
	{
		m_vertex_shader[i] = m_vertex_shader[i - 1];
	}
	m_vertex_shader[0] = shader_id;
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::PopVertexShader( void )
{
	for( uint32 i = 0; i < sMesh::VERTEX_SHADER_STACK_SIZE - 1; ++i )
	{
		m_vertex_shader[i] = m_vertex_shader[i + 1];
	}
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::wibble_normals( void )
{
	if( m_flags & 0 )
	{
		// Angle in the range [-PI/16, PI/16], period is 1 second.
		float time = (float)Tmr::GetTime() * 0.0005f;

		BYTE		*p_byte;
		float		*p_normal;
		float		*p_pos;
		mp_vertex_buffer[m_current_write_vertex_buffer]->Lock( 0, 0, &p_byte, 0 );
		p_pos		= (float*)( p_byte + 0 );
		p_normal	= (float*)( p_byte + m_normal_offset );

		for( uint32 i = 0; i < m_num_vertices; ++i )
		{
			float x				= p_pos[0] - m_sphere_center.x;
			float z				= p_pos[2] - m_sphere_center.z;
			
			float time_offset_x	= time + (( x / m_sphere_radius ) * 0.5f );
			float time_offset_z	= time + (( z / m_sphere_radius ) * 0.5f );

			float angle_x		= ( Mth::PI * ( 1.0f / 64.0f ) * (float)fabs( sinf( time_offset_x * Mth::PI ))) - ( Mth::PI * ( 1.0f / 128.0f ));
			float angle_z		= ( Mth::PI * ( 1.0f / 64.0f ) * (float)fabs( sinf( time_offset_z * Mth::PI ))) - ( Mth::PI * ( 1.0f / 129.0f ));
			
			Mth::Vector	n( sinf( angle_x ), cosf(( angle_x + angle_z ) * 0.5f ), sinf( angle_z ));
			n.Normalize();
			
			p_normal[0]			= n[X];
			p_normal[1]			= n[Y];
			p_normal[2]			= n[Z];
			
			p_pos				= (float*)((BYTE*)p_pos + m_vertex_stride );
			p_normal			= (float*)((BYTE*)p_normal + m_vertex_stride );
		}
	}
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::wibble_vc( void )
{
	if( mp_vc_wibble_data )
	{
		// Grab byte pointer to current 'write' vertex buffer.
		BYTE		*p_byte;
		D3DCOLOR	*p_color;
		mp_vertex_buffer[m_current_write_vertex_buffer]->Lock( 0, 0, &p_byte, 0 );
		p_color = (D3DCOLOR*)( p_byte + m_diffuse_offset );

		D3DCOLOR *p_color_array	= mp_material->mp_wibble_vc_colors;

		// Scan through each vertex, setting the new color.
		for( uint32 i = 0; i < m_num_vertices; ++i )
		{
			// An index of zero means no update for this vert.
			uint32 index	= mp_vc_wibble_data[i];
			if( index > 0 )
			{
				*p_color	= p_color_array[index - 1];
			}
			p_color		= (D3DCOLOR*)((BYTE*)p_color + m_vertex_stride );
		}
	}
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::CreateDuplicateVertexBuffers( int n )
{
	// Ensure this hasn't already been called.
	Dbg_Assert( mp_vertex_buffer[0] != NULL );
	Dbg_Assert( mp_vertex_buffer[1] == NULL );
	Dbg_Assert(( n > 0 ) && ( n < MAX_VERTEX_BUFFERS ));

	// Lock the source buffer.
	BYTE *p_byte;
	if( D3D_OK != mp_vertex_buffer[0]->Lock( 0, 0, &p_byte, D3DLOCK_READONLY ))
	{
		exit( 0 );
	}
	
	for( int i = 0; i < n; ++i )
	{
		mp_vertex_buffer[i + 1] = AllocateVertexBuffer( m_vertex_stride * m_num_vertices );

		// Lock the destination buffer and copy the contents of the original buffer into the new buffer.
		BYTE *p_byte2;
		if( D3D_OK != mp_vertex_buffer[i + 1]->Lock( 0, 0, &p_byte2, 0 ))
		{
			exit( 0 );
		}
		CopyMemory( p_byte2, p_byte, m_vertex_stride * m_num_vertices );
	}

	m_num_vertex_buffers = 1 + n;
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::SetPosition( Mth::Vector &pos )
{
	// Create a transform if one doesn't exist yet.
	if( mp_transform == NULL )
	{
		mp_transform = new Mth::Matrix();
		mp_transform->Ident();
	}
	
	// Figure what we need to add to each vertex, based on current position.
	Mth::Vector offset(	pos[X] - mp_transform->GetPos()[X],
						pos[Y] - mp_transform->GetPos()[Y],
						pos[Z] - mp_transform->GetPos()[Z] );

	mp_transform->SetPos( pos );
	
	for( uint32 vb = 0; vb < m_num_vertex_buffers; ++vb )
	{
		BYTE *p_byte;
		mp_vertex_buffer[vb]->Lock( 0, 0, &p_byte, 0 );

		for( uint32 v = 0; v < m_num_vertices; ++v )
		{
			((D3DVECTOR*)p_byte )->x += offset[X];
			((D3DVECTOR*)p_byte )->y += offset[Y];
			((D3DVECTOR*)p_byte )->z += offset[Z];
			p_byte += m_vertex_stride;
		}
	}

	// We also need to adjust the bounding box and sphere information for this mesh.
	m_sphere_center.x += offset[X];
	m_sphere_center.y += offset[Y];
	m_sphere_center.z += offset[Z];
}
	

	
/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::GetPosition( Mth::Vector *p_pos )
{
	if( mp_transform == NULL )
	{
		p_pos->Set( 0.0f, 0.0f, 0.0f );
	}
	else
	{
		*p_pos = mp_transform->GetPos();
	}
}
	

	
/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::SetYRotation( Mth::ERot90 rot )
{
	if( rot > Mth::ROT_0 )
	{
		// Create a transform if one doesn't exist yet.
		if( mp_transform == NULL )
		{
			mp_transform = new Mth::Matrix();
			mp_transform->Ident();
		}

		for( uint32 vb = 0; vb < m_num_vertex_buffers; ++vb )
		{
			BYTE *p_byte;
			mp_vertex_buffer[vb]->Lock( 0, 0, &p_byte, 0 );

			switch( rot )
			{
				case Mth::ROT_90:
				{
					for( uint32 v = 0; v < m_num_vertices; ++v )
					{
						float x = ((D3DVECTOR*)p_byte )->x - mp_transform->GetPos()[X];
						float z = ((D3DVECTOR*)p_byte )->z - mp_transform->GetPos()[Z];
						((D3DVECTOR*)p_byte )->x = z + mp_transform->GetPos()[X];
						((D3DVECTOR*)p_byte )->z = -x + mp_transform->GetPos()[Z];
						p_byte += m_vertex_stride;
					}

					// Adjust the bounding sphere information for this mesh.
					m_sphere_center.x	-= mp_transform->GetPos()[X];
					m_sphere_center.z	-= mp_transform->GetPos()[Z];
					float t				= m_sphere_center.x;
					m_sphere_center.x	= m_sphere_center.z + mp_transform->GetPos()[X];
					m_sphere_center.z	= -t + mp_transform->GetPos()[Z];
					break;
				}
				case Mth::ROT_180:
				{
					for( uint32 v = 0; v < m_num_vertices; ++v )
					{
						float x = ((D3DVECTOR*)p_byte )->x - mp_transform->GetPos()[X];
						float z = ((D3DVECTOR*)p_byte )->z - mp_transform->GetPos()[Z];
						((D3DVECTOR*)p_byte )->x = -x + mp_transform->GetPos()[X];
						((D3DVECTOR*)p_byte )->z = -z + mp_transform->GetPos()[Z];
						p_byte += m_vertex_stride;
					}

					// Adjust the bounding sphere information for this mesh.
					m_sphere_center.x	-= mp_transform->GetPos()[X];
					m_sphere_center.z	-= mp_transform->GetPos()[Z];
					m_sphere_center.x	= -m_sphere_center.x + mp_transform->GetPos()[X];
					m_sphere_center.z	= -m_sphere_center.z + mp_transform->GetPos()[Z];
					break;
				}
				case Mth::ROT_270:
				{
					for( uint32 v = 0; v < m_num_vertices; ++v )
					{
						float x = ((D3DVECTOR*)p_byte )->x - mp_transform->GetPos()[X];
						float z = ((D3DVECTOR*)p_byte )->z - mp_transform->GetPos()[Z];
						((D3DVECTOR*)p_byte )->x = -z + mp_transform->GetPos()[X];
						((D3DVECTOR*)p_byte )->z = x + mp_transform->GetPos()[Z];
						p_byte += m_vertex_stride;
					}

					// Adjust the bounding sphere information for this mesh.
					m_sphere_center.x	-= mp_transform->GetPos()[X];
					m_sphere_center.z	-= mp_transform->GetPos()[Z];
					float t				= m_sphere_center.x;
					m_sphere_center.x	= -m_sphere_center.z + mp_transform->GetPos()[X];
					m_sphere_center.z	= t + mp_transform->GetPos()[Z];
					break;
				}
			}
		}
	}
}




/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::SwapVertexBuffers( void )
{
	if( m_num_vertex_buffers > 1 )
	{
		m_current_write_vertex_buffer = ( m_current_write_vertex_buffer + 1 ) % m_num_vertex_buffers;
	}
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::HandleColorOverride( void )
{
	static float constants[16];

	Dbg_Assert( m_flags & sMesh::MESH_FLAG_MATERIAL_COLOR_OVERRIDE );

	// Re-jig the pixel shader constants for the material color override.
	CopyMemory( constants, EngineGlobals.pixel_shader_constants, sizeof( float ) * 16 );
			
	for( uint32 p = 0; p < mp_material->m_passes; ++p )
	{
		if( !( mp_material->m_flags[p] & MATFLAG_PASS_COLOR_LOCKED ))
		{
//			EngineGlobals.pixel_shader_constants[( p * 4 ) + 0]	*= m_material_color_override[0];
//			EngineGlobals.pixel_shader_constants[( p * 4 ) + 1]	*= m_material_color_override[1];
//			EngineGlobals.pixel_shader_constants[( p * 4 ) + 2]	*= m_material_color_override[2];
			EngineGlobals.pixel_shader_constants[( p * 4 ) + 0]	= m_material_color_override[0];
			EngineGlobals.pixel_shader_constants[( p * 4 ) + 1]	= m_material_color_override[1];
			EngineGlobals.pixel_shader_constants[( p * 4 ) + 2]	= m_material_color_override[2];
		}
	}
	EngineGlobals.upload_pixel_shader_constants = true;

	// Set the pixel shader (this will upload the new constants).
	set_pixel_shader( m_pixel_shader, mp_material->m_passes );

	// Restore the pixel shader constants and flag as needing a reload.
	CopyMemory( EngineGlobals.pixel_shader_constants, constants, sizeof( float ) * 16 );
	EngineGlobals.upload_pixel_shader_constants = true;
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::Submit( void )
{
	DWORD	stage_zero_minfilter;
	DWORD	zwrite;
	
	// Pointless submitting a mesh with zero indices.
	if( m_num_indices == 0 )
		return;

	// Deal with vertex color wibbling.
	wibble_vc();
	
	// Set vertex shader.
	set_vertex_shader( m_vertex_shader[0] );

	if( m_flags & sMesh::MESH_FLAG_MATERIAL_COLOR_OVERRIDE )
	{
		HandleColorOverride();
	}
	else
	{
		// Just set the pixel shader.
		set_pixel_shader( m_pixel_shader, mp_material->m_passes );
	}

	// Deal with meshes that set no anisotropic filtering.
	if( m_flags & MESH_FLAG_NO_ANISOTROPIC )
	{
		D3DDevice_GetTextureStageState( 0, D3DTSS_MINFILTER, &stage_zero_minfilter );
		if( stage_zero_minfilter != D3DTEXF_LINEAR )
		{
			D3DDevice_SetTextureStageState( 0, D3DTSS_MINFILTER, D3DTEXF_LINEAR );
		}
	}
	
	// Deal with meshes that set no z-write.
	if( m_flags & MESH_FLAG_NO_ZWRITE )
	{
		D3DDevice_GetRenderState( D3DRS_ZWRITEENABLE, &zwrite );
		if( zwrite == TRUE )
		{
			D3DDevice_SetRenderState( D3DRS_ZWRITEENABLE, FALSE );
		}
	}
	
	// Get the vertex buffer to submit (which is the one we have potentially just been writing to).
	IDirect3DVertexBuffer8*	p_submit_buffer = mp_vertex_buffer[m_current_write_vertex_buffer];
	
	// Swap multiple vertex buffers if present.
	SwapVertexBuffers();

	// Set the stream source.
	D3DDevice_SetStreamSource( 0, p_submit_buffer, m_vertex_stride );

	// See if we have index LOD data, in which case we need to figure distance and select the correct LOD.
	if( mp_index_lod_data )
	{
		// Figure distance from this mesh to the camera. This is *not* an efficient way to do it.
		frustum_check_sphere( &m_sphere_center, m_sphere_radius );
		float dist = get_bounding_sphere_nearest_z();
		for( int idx = 0; idx < MAX_INDEX_BUFFERS; ++idx )
		{
			if( mp_index_buffer[idx] == NULL )
			{
				// We have got to the end of the set without drawing anything. Just use the last valid index set.
				if( idx > 0 )
				{
					--idx;
					D3DDevice_DrawIndexedVertices( m_primitive_type, m_num_indices[idx], mp_index_buffer[idx] );
				}
				else
				{
					Dbg_Assert( 0 );
				}
				break;
			}

			if( dist < mp_index_lod_data[idx] )
			{
				// This is the index set we want.
				D3DDevice_DrawIndexedVertices( m_primitive_type, m_num_indices[idx], mp_index_buffer[idx] );
				break;
			}
		}
	}
	else
	{
		// Submit.
		D3DDevice_DrawIndexedVertices( m_primitive_type, m_num_indices[0], mp_index_buffer[0] );
	}

	// Deal with meshes that set no anisotropic filtering.
	if( m_flags & MESH_FLAG_NO_ANISOTROPIC )
	{
		if( stage_zero_minfilter != D3DTEXF_LINEAR )
		{
			D3DDevice_SetTextureStageState( 0, D3DTSS_MINFILTER, stage_zero_minfilter );
		}
	}

	// Deal with meshes that set no z-write.
	if( m_flags & MESH_FLAG_NO_ZWRITE )
	{
		if( zwrite == TRUE )
		{
			D3DDevice_SetRenderState( D3DRS_ZWRITEENABLE, zwrite );
		}
	}
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
sMesh *sMesh::Clone( bool instance )
{
	sMesh *p_clone = new sMesh();

	// Copy over basic details.
	CopyMemory( p_clone, this, sizeof( sMesh ));
	
	if( instance )
	{
		p_clone->m_flags |= MESH_FLAG_IS_INSTANCE;
	}
	else
	{
		// Build new vertex and index lists.
		p_clone->mp_vertex_buffer[0] = AllocateVertexBuffer( p_clone->m_vertex_stride * p_clone->m_num_vertices );

		BYTE *p_byte_src, *p_byte_dest;
		if( D3D_OK != mp_vertex_buffer[0]->Lock( 0, 0, &p_byte_src, D3DLOCK_READONLY ))
		{
			return NULL;
		}
		if( D3D_OK != p_clone->mp_vertex_buffer[0]->Lock( 0, 0, &p_byte_dest, 0 ))
		{
			return NULL;
		}

		// Copy over vertex information.
		CopyMemory( p_byte_dest, p_byte_src, p_clone->m_vertex_stride * p_clone->m_num_vertices );

		// Create index buffer(s) and copy over index information.
		for( int ib = 0; ib < MAX_INDEX_BUFFERS; ++ib )
		{
			if( p_clone->m_num_indices[ib] > 0 )
			{
				p_clone->mp_index_buffer[ib] = new uint16[p_clone->m_num_indices[ib]];
				CopyMemory( p_clone->mp_index_buffer[ib], mp_index_buffer[ib], sizeof( uint16 ) * p_clone->m_num_indices[ib] );
			}
		}

		// Handle duplicate vertex buffers if they exist.
		if( m_num_vertex_buffers > 1 )
		{
			p_clone->mp_vertex_buffer[1] = NULL;
			p_clone->CreateDuplicateVertexBuffers( m_num_vertex_buffers - 1 );
		}
	}
	return p_clone;
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
IDirect3DVertexBuffer8 *sMesh::AllocateVertexBuffer( uint32 size )
{
	uint8					*p_vb		= new uint8[sizeof( IDirect3DVertexBuffer8 ) + size];
	IDirect3DVertexBuffer8	*p_vb_ret	= (IDirect3DVertexBuffer8*)p_vb;
	
	XGSetVertexBufferHeader( 0, 0, 0, 0, p_vb_ret, 0 );
	p_vb_ret->Register( p_vb + sizeof( IDirect3DVertexBuffer8 ));

	return p_vb_ret;
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::Crunch( void )
{
	uint16 *p_indices = mp_index_buffer[0];

	uint32 i0 = p_indices[0];
	uint32 i1 = p_indices[1];

	uint32	invalid			= 0;
	uint32	total_invalid	= 0;
	bool	crunch			= false;

	for( uint32 i = 2; i < m_num_indices[0]; ++i )
	{
		uint32 i2 = p_indices[i];

		if(( i0 == i1 ) || ( i0 == i2 ) || ( i1 == i2 ))
		{
			++invalid;
		}
		else
		{
			if( invalid > 5 )
			{
				if(( invalid & 1 ) == 0 )
				{
//					printf( "Crunching %d indices (even)\n", invalid - 4 );

					// Ensure the leading and trailing degenerate indices are correct.
					p_indices[i - 3]		= p_indices[i - 2];
					p_indices[i - invalid]	= p_indices[i - invalid - 1];

					// With an even number of invalid entries, the wind order won't change during crunch.
					MoveMemory( p_indices + i - invalid + 1, p_indices + i - 3, sizeof( uint16 ) * ( m_num_indices[0] - i + 3 ));

					m_num_indices[0]	-= (uint16)( invalid - 4 );
					i					-= invalid - 4;
				}
				else
				{
//					printf( "Crunching %d indices (odd)\n", invalid - 5 );

					// Ensure the leading and trailing degenerate indices are correct.
					p_indices[i - 3]			= p_indices[i - 2];
					p_indices[i - invalid]		= p_indices[i - invalid - 1];
					p_indices[i - invalid + 1]	= p_indices[i - invalid];

					// With an odd number of invalid entries, the wind order will change during crunch, so use one extra index.
					MoveMemory( p_indices + i - invalid + 2, p_indices + i - 3, sizeof( uint16 ) * ( m_num_indices[0] - i + 3 ));
					m_num_indices[0]	-= (uint16)( invalid - 5 );
					i					-= invalid - 5;
				}
			}
			invalid = 0;
		}
		
		i0 = i1;
		i1 = i2;
	}
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::SetBillboardData( uint32 type, Mth::Vector & pivot_pos, Mth::Vector & pivot_axis )
{
	Dbg_Assert( mp_billboard_data == NULL );

	// Create the billboard data.
	mp_billboard_data					= new sBillboardData;

	// Determine the billboard type.
	if( type == 1 )
	{
		mp_billboard_data->m_type = sBillboardData::vBILLBOARD_TYPE_SCREEN_ALIGNED;
	}
	else if( type == 2 )
	{
		// Axial aligned. See if this is y axis.
		if( fabsf( Mth::DotProduct( pivot_axis, Mth::Vector( 0.0f, 1.0f, 0.0f ))) > 0.99f )
		{
			mp_billboard_data->m_type = sBillboardData::vBILLBOARD_TYPE_Y_AXIS_ALIGNED;
		}
		else
		{
			mp_billboard_data->m_type = sBillboardData::vBILLBOARD_TYPE_ARBITRARY_AXIS_ALIGNED;
		}
	}
	else
	{
		Dbg_Assert( 0 );
	}
	
	mp_billboard_data->m_pivot_pos		= pivot_pos;
	if( mp_billboard_data->m_type == sBillboardData::vBILLBOARD_TYPE_ARBITRARY_AXIS_ALIGNED )
	{
		mp_billboard_data->m_pivot_axis	= pivot_axis;
	}

	// We need to go through at a low level and rebuild the vertex buffer. In all cases the
	// mesh won't have been exported with normals, which we use for billboards to store the
	// pivot-relative position, so we need to recalculate the vertex stride.
	Dbg_Assert( m_normal_offset == 0 );
	Dbg_Assert( m_diffuse_offset > 0 );
	Dbg_Assert( m_uv0_offset > 0 );
	Dbg_Assert( m_num_vertices == 4 );
	Dbg_Assert( m_num_indices[0] == 4 );

	// Add size of normal to vertex size.
	int old_vertex_stride = m_vertex_stride;
	int new_vertex_stride = old_vertex_stride + ( sizeof( float ) * 3 );

	IDirect3DVertexBuffer8*	p_old_buffer = mp_vertex_buffer[0];
	IDirect3DVertexBuffer8*	p_new_buffer = AllocateVertexBuffer( new_vertex_stride * 4 );

	// Lock old buffer (read) and new buffer (write).
	BYTE *p_old_vb_data;
	BYTE *p_new_vb_data;
	p_old_buffer->Lock( 0, 0, &p_old_vb_data, D3DLOCK_READONLY | D3DLOCK_NOFLUSH );
	p_new_buffer->Lock( 0, 0, &p_new_vb_data, 0 );

	// Calculate the normal of the billboard, using the first tri.
	float		*p_vert;
	uint16		indices[4];
	indices[0]			= mp_index_buffer[0][0];
	indices[1]			= mp_index_buffer[0][1];
	indices[2]			= mp_index_buffer[0][2];
	indices[3]			= mp_index_buffer[0][3];

	p_vert				= (float*)( p_old_vb_data + ( indices[0] * old_vertex_stride ));
	Mth::Vector v0( p_vert[0], p_vert[1], p_vert[2] );
	p_vert				= (float*)( p_old_vb_data + ( indices[1] * old_vertex_stride ));
	Mth::Vector v1( p_vert[0], p_vert[1], p_vert[2] );
	p_vert				= (float*)( p_old_vb_data + ( indices[2] * old_vertex_stride ));
	Mth::Vector v2( p_vert[0], p_vert[1], p_vert[2] );
	Mth::Vector	normal	= Mth::CrossProduct(( v1 - v0 ), ( v2 - v0 )).Normalize();

	// Given the normal, calculate the local right and up (u and v) vectors, based on the billboard type.
	Mth::Vector u, v;

	switch( mp_billboard_data->m_type )
	{
		case sBillboardData::vBILLBOARD_TYPE_SCREEN_ALIGNED:
		case sBillboardData::vBILLBOARD_TYPE_Y_AXIS_ALIGNED:
		case sBillboardData::vBILLBOARD_TYPE_WORLD_ORIENTED:
		{
			// Use the world 'up' vector to generate the 'u' vector.
			u = Mth::CrossProduct( normal, Mth::Vector( 0.0f, 1.0f, 0.0f )).Normalize();

			// Use the 'u' vector and the normal vector to generate the 'v' vector.
			v = Mth::CrossProduct( u, normal ).Normalize();
			break;
		}
		case sBillboardData::vBILLBOARD_TYPE_ARBITRARY_AXIS_ALIGNED:
		{
			// Use the pivot axis and the normal vector to generate the 'u' vector.
			u = Mth::CrossProduct( normal, pivot_axis ).Normalize();

			// Use the 'u' vector and the normal vector to generate the 'v' vector.
			v = Mth::CrossProduct( u, normal ).Normalize();
			break;
		}
	}

	for( int i = 0; i < 4; ++i )
	{
		// The new position is actually the position of the pivot point for the billboard.
		float *p_pos_old	= (float*)( p_old_vb_data + ( i * old_vertex_stride ));
		float *p_pos_new	= (float*)( p_new_vb_data + ( i * new_vertex_stride ));
		p_pos_new[0]		= pivot_pos[X];
		p_pos_new[1]		= pivot_pos[Y];
		p_pos_new[2]		= pivot_pos[Z];

		// Introduce normal (which is actually the position of the vertex relative to the pivot).
		Mth::Vector pos_relative_to_pivot( p_pos_old[0] - pivot_pos[X], p_pos_old[1] - pivot_pos[Y], p_pos_old[2] - pivot_pos[Z] );

		p_pos_new[3]		= Mth::DotProduct( pos_relative_to_pivot, u );
		p_pos_new[4]		= Mth::DotProduct( pos_relative_to_pivot, v );
		p_pos_new[5]		= Mth::DotProduct( pos_relative_to_pivot, normal );

		// Copy color.
		D3DCOLOR *p_col_old	= (D3DCOLOR*)( p_old_vb_data + ( i * old_vertex_stride ) + m_diffuse_offset );
		D3DCOLOR *p_col_new	= (D3DCOLOR*)( p_new_vb_data + ( i * new_vertex_stride ) + m_diffuse_offset + ( sizeof( float ) * 3 ));
		p_col_new[0]		= p_col_old[0];

		// Copy uv0...
		float *p_uv0_old	= (float*)( p_old_vb_data + ( i * old_vertex_stride ) + m_uv0_offset );
		float *p_uv0_new	= (float*)( p_new_vb_data + ( i * new_vertex_stride ) + m_uv0_offset + ( sizeof( float ) * 3 ));
		p_uv0_new[0]		= p_uv0_old[0];
		p_uv0_new[1]		= p_uv0_old[1];

		// ...and additional uv's if present.
		if(( m_vertex_shader[0] & D3DFVF_TEXCOUNT_MASK ) > D3DFVF_TEX1 )
		{
			p_uv0_new[2]		= p_uv0_old[2];
			p_uv0_new[3]		= p_uv0_old[3];
		}
		if(( m_vertex_shader[0] & D3DFVF_TEXCOUNT_MASK ) > D3DFVF_TEX2 )
		{
			p_uv0_new[4]		= p_uv0_old[4];
			p_uv0_new[5]		= p_uv0_old[5];
		}
		if(( m_vertex_shader[0] & D3DFVF_TEXCOUNT_MASK ) > D3DFVF_TEX3 )
		{
			p_uv0_new[6]		= p_uv0_old[6];
			p_uv0_new[7]		= p_uv0_old[7];
		}
	}

	// Now fix up the mesh. Flag the mesh as being a billboard (stop the mesh being rendered by the regular pathway).
	m_flags |= sMesh::MESH_FLAG_BILLBOARD;

	// Switch vertex buffers, deleting the old one.
	mp_vertex_buffer[0]	= p_new_buffer;
	delete p_old_buffer;

	// Set the new vertex stride, diffuse and uv0 offset (and normal offset, just to be complete).
	m_vertex_stride		= new_vertex_stride;
	m_diffuse_offset	+= sizeof( float ) * 3;
	m_uv0_offset		+= sizeof( float ) * 3;
	m_normal_offset		= sizeof( float ) * 3;

	// Copy the new vertex buffer into existing buffered buffers if m_num_vertex_buffers > 1.
	if( m_num_vertex_buffers > 1 )
	{
		BYTE *p_buffer0;
		BYTE *p_bufferN;
		mp_vertex_buffer[0]->Lock( 0, 0, &p_buffer0, D3DLOCK_READONLY | D3DLOCK_NOFLUSH );
		for( int vb = 1; vb < m_num_vertex_buffers; ++vb )
		{
			delete mp_vertex_buffer[vb];
			mp_vertex_buffer[vb] = AllocateVertexBuffer( new_vertex_stride * 4 );
			mp_vertex_buffer[vb]->Lock( 0, 0, &p_bufferN, 0 );
			CopyMemory( p_bufferN, p_buffer0, new_vertex_stride * 4 );
		}
	}

	// Set the new vertex shader.
	m_vertex_shader[0]	= 999;
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::SetBoundingData( Mth::Vector & sphere_center, float radius, Mth::Vector & bb_min, Mth::Vector & bb_max )
{
//	m_bbox.Set( bb_min, bb_max );

	m_sphere_center = D3DXVECTOR3( sphere_center[X], sphere_center[Y], sphere_center[Z] );
	m_sphere_radius = radius;
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::Initialize( int				num_vertices,
						float			*p_positions,
						float			*p_normals,
						float			*p_tex_coords,
						int				num_tc_sets,
						DWORD			*p_colors,
						int				num_index_sets,			// How many sets of indices there are (usually 1 set)
						int				*p_num_indices,			// Pointer to an array of ints containing number of indices per set
						uint16			**pp_indices,			// Pointer to an array of pointers to the actual indices
						unsigned long	material_checksum,
						void			*p_scene,
						uint16			*p_matrix_indices,
						uint32			*p_weights,
						char			*p_vc_wibble_anims )
{
	// First thing to do is grab the material pointer for this mesh.
	mp_material	= ((sScene*)p_scene )->GetMaterial( material_checksum );
	if( mp_material == NULL )
	{
		Dbg_Assert( 0 );
		return;
	}
	
	if(( num_index_sets == 0 ) || ( p_num_indices[0] == 0 ))
	{
		return;
	}

	uint16 min_index	= ( pp_indices[0] )[0];
	uint16 max_index	= ( pp_indices[0] )[0];
	for( int i = 0; i < p_num_indices[0]; ++i )
	{
		if(( pp_indices[0] )[i] > max_index )
		{
			max_index = ( pp_indices[0] )[i];
		}
		else if(( pp_indices[0] )[i] < min_index )
		{
			min_index = ( pp_indices[0] )[i];
		}
	}

	if( max_index >= num_vertices )
	{
		// Error!
		Dbg_Assert( 0 );
		return;
	}

	// Grab top-down heap memory for the mesh workspace buffer. This will need to be as big as the maximum vertex indexed.
	int16 *p_mesh_workspace_array = new (Mem::Manager::sHandle().TopDownHeap()) int16[max_index + 1];

	// Setup workspace buffer.
	memset( p_mesh_workspace_array, 1, sizeof( int16 ) * ( max_index + 1 ));
	for( int i = 0; i < p_num_indices[0]; ++i )
	{
		p_mesh_workspace_array[( pp_indices[0] )[i]] = 0;
	}
	
	// Now figure the wasted space.
	int wasted_verts = 0;
	for( int i = min_index; i <= max_index; ++i )
	{
		if( p_mesh_workspace_array[i] != 0 )
			++wasted_verts;
	}

	// Now figure the total number of vertices required for this mesh, to span the min->max indices.
	uint16 vertices_for_this_mesh	= ( max_index - min_index + 1 ) - wasted_verts;
	m_num_vertices					= vertices_for_this_mesh;

	// Create the index buffer(s). (Should be 16byte aligned for best performance).
	for( int ib = 0; ib < num_index_sets; ++ib )
	{
		mp_index_buffer[ib]	= new uint16[p_num_indices[ib]];
		m_num_indices[ib]	= p_num_indices[ib];
	}
	
	// Use the material flags to figure the vertex format.
	int vertex_size			= 3 * sizeof( float );

	// Include weights (for weighted animation) if present.
	uint32 biggest_index_used = 0;
	if( p_weights )
	{
		Dbg_Assert( p_matrix_indices );

		// Calculate the biggest weight used.
		uint32*	p_weight_read	= p_weights + min_index;
		for( int v = min_index; v <= max_index; ++v )
		{
			if( p_mesh_workspace_array[v] == 0 )
			{
				// This vertex is used.
				uint32 w2 = (( p_weight_read[0] >> 22 ) & 0x3FF );
				if( w2 > 0 )
				{
					biggest_index_used = 2;
					break;
				}
				else
				{
					uint32 w1 = (( p_weight_read[0] >> 11 ) & 0x7FF );
					if( w1 > 0 )
					{
						biggest_index_used = 1;
					}
				}
			}
			++p_weight_read;
		}
		vertex_size	+= sizeof( uint32 );
	}

	// Include indices (for weighted animation) if present.
	if( p_matrix_indices )
	{
		Dbg_Assert( p_weights );
		vertex_size	+= sizeof( uint16 ) * 4;
	}
	
	// Texture coordinates.
	uint32	tex_coord_pass	= 0;
	bool	env_mapped		= false;
	if( p_tex_coords )
	{
		for( uint32 pass = 0; pass < mp_material->m_passes; ++pass )
		{
			if( mp_material->m_flags[pass] & MATFLAG_ENVIRONMENT )
			{
				env_mapped		= true;
			}

			// Only need UV's for this stage if it is *not* environment mapped.
			if(( mp_material->mp_tex[pass] ) && ( !( mp_material->m_flags[pass] & MATFLAG_ENVIRONMENT )))
			{
				// Will need uv for this pass and all before it.
				tex_coord_pass	= pass + 1; 
			}
		}
	}
	else
	{
		for( uint32 pass = 0; pass < mp_material->m_passes; ++pass )
		{
			if( mp_material->m_flags[pass] & MATFLAG_ENVIRONMENT )
			{
				env_mapped		= true;
			}
		}
	}

	if( tex_coord_pass > 0 )
	{
		vertex_size			+= 2 * sizeof( float ) * tex_coord_pass;
	}

	// Assume no normals for now, unless weight information indicates an animating mesh.
	bool use_normals		= false;
	bool use_packed_normals	= false;

	if( p_normals || p_weights || env_mapped )
	{
		// Need to include normals. They will be packed differently for weighted meshes.
		use_normals	= true;
		if( p_weights )
		{
			use_packed_normals = true;
			vertex_size	+= sizeof( uint32 );
		}
		else
		{
			vertex_size	+= sizeof( float ) * 3;
		}
	}

	bool use_colors = false;
	if( p_colors )
	{
		// The raw vertex data does contain vertex colors.
		vertex_size	+= sizeof( D3DCOLOR );
		use_colors	= true;
	}


	// Create the vertex buffer.
	m_vertex_stride	= vertex_size;

	// One allocation for the header and the data buffer.	
	mp_vertex_buffer[0] = AllocateVertexBuffer( vertex_size * vertices_for_this_mesh );

	// Lock the vertex buffer.
	BYTE* p_byte;
	if( D3D_OK != mp_vertex_buffer[0]->Lock(	0,					// Offset to lock.
												0,					// Size to lock ( 0 means all).
												&p_byte,			// Pointer to data.
												D3DLOCK_NOFLUSH ))	// Flags.
	{
		Dbg_Assert( 0 );
		return;
	}
	
	// Copy in vertex position data (for vertices that are used).
	uint32		byte_write_offset	= 0;
	float*		p_read				= p_positions + ( min_index * 3 );
	float*		p_write				= (float*)p_byte;

	for( int v = min_index; v <= max_index; ++v )
	{
		if( p_mesh_workspace_array[v] == 0 )
		{
			// This vertex is used.
			p_write[0]	= p_read[0];
			p_write[1]	= p_read[1];
			p_write[2]	= p_read[2];
			p_write	= (float*)((char*)p_write + vertex_size );
		}
		p_read += 3;
	}

	byte_write_offset	+= sizeof( float ) * 3;
	m_vertex_shader[0]	|= D3DFVF_XYZ;

	// Copy in vertex weight data.
	if( p_weights )
	{
		uint32*	p_weight_read	= p_weights + min_index;
		uint32*	p_weight_write	= (uint32*)((char*)p_byte + byte_write_offset );

		for( int v = min_index; v <= max_index; ++v )
		{
			if( p_mesh_workspace_array[v] == 0 )
			{
				// This vertex is used.
				p_weight_write[0]	= p_weight_read[0];
				p_weight_write		= (uint32*)((char*)p_weight_write + vertex_size );
			}
			++p_weight_read;
		}
		byte_write_offset += sizeof( uint32 );

		// No fvf flag setting, since it will be determined at the end.
	}

	// Copy in vertex matrix index data.
	if( p_matrix_indices )
	{
		uint16*	p_index_read	= p_matrix_indices + ( min_index * 4 );
		uint16*	p_index_write	= (uint16*)((char*)p_byte + byte_write_offset );
		for( int v = min_index; v <= max_index; ++v )
		{
			if( p_mesh_workspace_array[v] == 0 )
			{
				// Have to multiply the indices by three to get the correct register offset, since there are 3 registers
				// per matrix.
				p_index_write[0]	= p_index_read[0] * 3;
				p_index_write[1]	= p_index_read[1] * 3;
				p_index_write[2]	= p_index_read[2] * 3;
				p_index_write[3]	= p_index_read[3] * 3;
				p_index_write		= (uint16*)((char*)p_index_write + vertex_size );
			}
			p_index_read += 4;
		}
		byte_write_offset += sizeof( uint16 ) * 4;

		// No fvf flag setting, since it will be determined at the end.
	}

	// Copy in normals data.
	if( use_normals )
	{
		m_normal_offset		= (uint8)byte_write_offset;
		if( use_packed_normals )
		{
			float *p_read	= p_normals + ( min_index * 3 );
			uint32 *p_write	= (uint32*)((char*)p_byte + byte_write_offset );
			for( int v = min_index; v <= max_index; ++v )
			{
				if( p_mesh_workspace_array[v] == 0 )
				{
					// The packed normal format is as follows:
					// 31                                             0
					// |----- 10 -----|----- 11 ------|----- 11 ------|
					// |       z      |       y       |       x       |
					uint32 snx	= Ftoi_ASM( p_read[0] * 1023.0f );
					uint32 sny	= Ftoi_ASM( p_read[1] * 1023.0f );
					uint32 snz	= Ftoi_ASM( p_read[2] * 511.0f );
					p_write[0]	= ( snx & 0x7FF ) | (( sny & 0x7FF ) << 11 ) | (( snz & 0x3FF ) << 22 );
					p_write		= (uint32*)((char*)p_write + vertex_size );
				}
				p_read += 3;
			}
			byte_write_offset += sizeof( uint32 );
		}
		else
		{
			float*	p_read	= p_normals + ( min_index * 3 );
			float*	p_write	= (float*)((char*)p_byte + byte_write_offset );
			for( int v = min_index; v <= max_index; ++v )
			{
				if( p_mesh_workspace_array[v] == 0 )
				{
					p_write[0]	= p_read[0];
					p_write[1]	= p_read[1];
					p_write[2]	= p_read[2];
					p_write		= (float*)((char*)p_write + vertex_size );
				}
				p_read += 3;
			}
			byte_write_offset += sizeof( float ) * 3;
		}
		m_vertex_shader[0]	|= D3DFVF_NORMAL;
	}

	// Copy in vertex color data.
	if( use_colors )
	{
		m_diffuse_offset	= (uint8)byte_write_offset;
		DWORD*	p_col_read	= p_colors + min_index;
		DWORD*	p_col_write	= (DWORD*)((char*)p_byte + byte_write_offset );
		for( int v = min_index; v <= max_index; ++v )
		{
			if( p_mesh_workspace_array[v] == 0 )
			{
				p_col_write[0]	= p_col_read[0];
				p_col_write		= (DWORD*)((char*)p_col_write + vertex_size );
			}
			p_col_read++;
		}
		byte_write_offset += sizeof( DWORD );
		m_vertex_shader[0] |= D3DFVF_DIFFUSE;
	}

	// Copy in vertex texture coordinate data.
	if(( tex_coord_pass > 0 ) && ( p_tex_coords != NULL ))
	{
		m_uv0_offset						= (uint8)byte_write_offset;
		p_read								= p_tex_coords + ( min_index * 2 * num_tc_sets );
		p_write								= (float*)((char*)p_byte + byte_write_offset );
		for( int v = min_index; v <= max_index; ++v )
		{
			if( p_mesh_workspace_array[v] == 0 )
			{
				for( uint32 pass = 0; pass < tex_coord_pass; ++pass )
				{
					p_write[( pass * 2 ) + 0]	= p_read[( pass * 2 ) + 0];
					p_write[( pass * 2 ) + 1]	= p_read[( pass * 2 ) + 1];
				}
				p_write	= (float*)((char*)p_write + vertex_size );
			}
			p_read = p_read + ( num_tc_sets * 2 );
		}
		byte_write_offset	+= sizeof( float ) * 2 * tex_coord_pass;

		switch( tex_coord_pass )
		{
			case 1:
			{
				m_vertex_shader[0]	|= D3DFVF_TEX1 | D3DFVF_TEXCOORDSIZE2( 0 );
				break;
			}
			case 2:
			{
				m_vertex_shader[0]	|= D3DFVF_TEX2 | D3DFVF_TEXCOORDSIZE2( 0 ) | D3DFVF_TEXCOORDSIZE2( 1 );
				break;
			}
			case 3:
			{
				m_vertex_shader[0]	|= D3DFVF_TEX3 | D3DFVF_TEXCOORDSIZE2( 0 ) | D3DFVF_TEXCOORDSIZE2( 1 ) | D3DFVF_TEXCOORDSIZE2( 2 );
				break;
			}
			case 4:
			{
				m_vertex_shader[0]	|= D3DFVF_TEX4 | D3DFVF_TEXCOORDSIZE2( 0 ) | D3DFVF_TEXCOORDSIZE2( 1 ) | D3DFVF_TEXCOORDSIZE2( 2 ) | D3DFVF_TEXCOORDSIZE2( 3 );
				break;
			}
			default:
			{
				Dbg_MsgAssert( 0, ( "Bad number of passes" ));
				break;
			}
		}
	}

	// Create the vertex color wibble array if data is present.
	if( p_vc_wibble_anims )
	{
		mp_vc_wibble_data			= new char[m_num_vertices];
		int vc_wibble_data_offset	= 0;

		for( int v = min_index; v <= max_index; ++v )
		{
			if( p_mesh_workspace_array[v] == 0 )
			{
				mp_vc_wibble_data[vc_wibble_data_offset++] = p_vc_wibble_anims[v];
			}
		}
		// Double buffer the vertex buffer.
		CreateDuplicateVertexBuffers( 1 );
	}

	// Process the workspace array.
	int offset = 0;
	for( int v = 0; v <= max_index; ++v )
	{
		if( p_mesh_workspace_array[v] == 0 )
		{
			// This vertex is used.
			p_mesh_workspace_array[v] = offset;
		}
		else
		{
			// This vertex is not used. Update the offset for the next used vertex.
			--offset;
		}
	}
	
	// Copy in index data, normalising the indices for this vertex buffer (i.e. so the lowest index will reference
	// vertex 0 in the buffer built specifically for this mesh).
	for( int ib = 0; ib < num_index_sets; ++ib )
	{
		for( int i = 0; i < p_num_indices[ib]; ++i )
		{
			uint16 idx				= ( pp_indices[ib] )[i];
			mp_index_buffer[ib][i]	= idx + p_mesh_workspace_array[idx];
		}
	}
	
	// Can now remove the mesh workspace array.
	delete [] p_mesh_workspace_array;
	
	// Set the correct vertex shader if a weighted mesh.
	// The number of indices used will be one more than the biggest index used (given 0 base).
	if( p_weights )
	{
		m_vertex_shader[0] = GetVertexShader( use_colors, ( mp_material->m_flags[0] & MATFLAG_SPECULAR ) ? true : false, biggest_index_used + 1 );
	}

	// Set the pixel shader regardless.
	GetPixelShader( mp_material, &m_pixel_shader );

	if( num_index_sets > 1 )
	{
		mp_index_lod_data = new float[num_index_sets];
		for( int d = 0; d < num_index_sets; ++d )
		{
			float dist				= ( 15.0f + ( d * 10.0f )) * 12.0f;
			mp_index_lod_data[d]	= dist;
		}
	}
}



/******************************************************************/
/*                                                                */
/*                                                                */
/******************************************************************/
void sMesh::DrawBoundingSphere( void )
{
#	ifdef __NOPT_ASSERT__
	const uint32 NUM_SPHERE_POINTS	= 64;

	static uint32 sphere_buffer[NUM_SPHERE_POINTS * 4];

	D3DXMATRIX *p_matrix = (D3DXMATRIX*)&NxXbox::EngineGlobals.view_matrix;
	Mth::Vector up( 0.0f, 1.0f, 0.0f );

	// Get the 'right' vector as the cross product of camera 'at and world 'up'.
	Mth::Vector at( p_matrix->m[0][2], p_matrix->m[1][2], p_matrix->m[2][2] );
	Mth::Vector screen_right	= Mth::CrossProduct( at, up );
	Mth::Vector screen_up		= Mth::CrossProduct( screen_right, at );

	screen_right.Normalize();
	screen_up.Normalize();

	set_vertex_shader( D3DFVF_XYZ | D3DFVF_DIFFUSE );
	set_pixel_shader( PixelShader5 );

	int		index		= 0;
	float	angle		= 0.0f;
	float	angle_step	= ( Mth::PI * 2.0f ) / (float)NUM_SPHERE_POINTS;

	// Draw the screen aligned sphere.
	for( uint32 i = 0; i < NUM_SPHERE_POINTS; ++i, angle += angle_step )
	{
		float x = m_sphere_center.x + ( m_sphere_radius * cosf( angle ) * screen_right[X] ) + ( m_sphere_radius * sinf( angle ) * screen_up[X] );
		float y = m_sphere_center.y + ( m_sphere_radius * cosf( angle ) * screen_right[Y] ) + ( m_sphere_radius * sinf( angle ) * screen_up[Y] );
		float z = m_sphere_center.z + ( m_sphere_radius * cosf( angle ) * screen_right[Z] ) + ( m_sphere_radius * sinf( angle ) * screen_up[Z] );

		sphere_buffer[index++]	= *((uint32*)&x );
		sphere_buffer[index++]	= *((uint32*)&y );
		sphere_buffer[index++]	= *((uint32*)&z );
		sphere_buffer[index++]	= 0x80800000;
	}
	D3DDevice_DrawVerticesUP( D3DPT_LINELOOP, NUM_SPHERE_POINTS, sphere_buffer, sizeof( uint32 ) * 4 );

	// Draw the xz plane sphere.
	index		= 0;
	angle		= 0.0f;
	for( uint32 i = 0; i < NUM_SPHERE_POINTS; ++i, angle += angle_step )
	{
		float x = m_sphere_center.x + ( m_sphere_radius * cosf( angle ));
		float y = m_sphere_center.y;
		float z = m_sphere_center.z + ( m_sphere_radius * sinf( angle ));

		sphere_buffer[index++]	= *((uint32*)&x );
		sphere_buffer[index++]	= *((uint32*)&y );
		sphere_buffer[index++]	= *((uint32*)&z );
		sphere_buffer[index++]	= 0x80800000;
	}
	D3DDevice_DrawVerticesUP( D3DPT_LINELOOP, NUM_SPHERE_POINTS, sphere_buffer, sizeof( uint32 ) * 4 );
#	endif
}





} // namespace NxXbox