//-------------------------------------------------------------------------------------
// DirectXMeshUtil.cpp
//  
// DirectX Mesh Geometry Library - Utilities
//
// THIS CODE AND INFORMATION IS PROVIDED "AS IS" WITHOUT WARRANTY OF
// ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING BUT NOT LIMITED TO
// THE IMPLIED WARRANTIES OF MERCHANTABILITY AND/OR FITNESS FOR A
// PARTICULAR PURPOSE.
//
// Copyright (c) Microsoft Corporation. All rights reserved.
//
// http://go.microsoft.com/fwlink/?LinkID=324981
//-------------------------------------------------------------------------------------

#include "DirectXMeshP.h"

using namespace DirectX;

#if defined(_XBOX_ONE) && defined(_TITLE)
static_assert(XBOX_DXGI_FORMAT_R10G10B10_SNORM_A2_UNORM == DXGI_FORMAT_R10G10B10_SNORM_A2_UNORM, "Xbox One XDK mismatch detected");
#endif


namespace DirectX
{

//=====================================================================================
// DXGI Format Utilities
//=====================================================================================

//-------------------------------------------------------------------------------------
// Returns bytes-per-element for a given DXGI format, or 0 on failure
//-------------------------------------------------------------------------------------
_Use_decl_annotations_
size_t BytesPerElement( DXGI_FORMAT fmt )
{
    // This list only includes those formats that are valid for use by IB or VB

    switch( static_cast<int>(fmt) )
    {
    case DXGI_FORMAT_R32G32B32A32_FLOAT:
    case DXGI_FORMAT_R32G32B32A32_UINT:
    case DXGI_FORMAT_R32G32B32A32_SINT:
        return 16;

    case DXGI_FORMAT_R32G32B32_FLOAT:
    case DXGI_FORMAT_R32G32B32_UINT:
    case DXGI_FORMAT_R32G32B32_SINT:
        return 12;

    case DXGI_FORMAT_R16G16B16A16_FLOAT:
    case DXGI_FORMAT_R16G16B16A16_UNORM:
    case DXGI_FORMAT_R16G16B16A16_UINT:
    case DXGI_FORMAT_R16G16B16A16_SNORM:
    case DXGI_FORMAT_R16G16B16A16_SINT:
    case DXGI_FORMAT_R32G32_FLOAT:
    case DXGI_FORMAT_R32G32_UINT:
    case DXGI_FORMAT_R32G32_SINT:
        return 8;

    case DXGI_FORMAT_R10G10B10A2_UNORM:
    case DXGI_FORMAT_R10G10B10A2_UINT:
    case DXGI_FORMAT_R11G11B10_FLOAT:
    case DXGI_FORMAT_R8G8B8A8_UNORM:
    case DXGI_FORMAT_R8G8B8A8_UINT:
    case DXGI_FORMAT_R8G8B8A8_SNORM:
    case DXGI_FORMAT_R8G8B8A8_SINT:
    case DXGI_FORMAT_R16G16_FLOAT:
    case DXGI_FORMAT_R16G16_UNORM:
    case DXGI_FORMAT_R16G16_UINT:
    case DXGI_FORMAT_R16G16_SNORM:
    case DXGI_FORMAT_R16G16_SINT:
    case DXGI_FORMAT_R32_FLOAT:
    case DXGI_FORMAT_R32_UINT:
    case DXGI_FORMAT_R32_SINT:
    case DXGI_FORMAT_B8G8R8A8_UNORM:
    case DXGI_FORMAT_B8G8R8X8_UNORM:
    case XBOX_DXGI_FORMAT_R10G10B10_SNORM_A2_UNORM:
        return 4;

    case DXGI_FORMAT_R8G8_UNORM:
    case DXGI_FORMAT_R8G8_UINT:
    case DXGI_FORMAT_R8G8_SNORM:
    case DXGI_FORMAT_R8G8_SINT:
    case DXGI_FORMAT_R16_FLOAT:
    case DXGI_FORMAT_R16_UNORM:
    case DXGI_FORMAT_R16_UINT:
    case DXGI_FORMAT_R16_SNORM:
    case DXGI_FORMAT_R16_SINT:
    case DXGI_FORMAT_B5G6R5_UNORM:
    case DXGI_FORMAT_B5G5R5A1_UNORM:
        return 2;

    case DXGI_FORMAT_R8_UNORM:
    case DXGI_FORMAT_R8_UINT:
    case DXGI_FORMAT_R8_SNORM:
    case DXGI_FORMAT_R8_SINT:
        return 1;

    case DXGI_FORMAT_B4G4R4A4_UNORM:
        return 2;

    default:
        // No BC, sRGB, XRBias, SharedExp, Typeless, Depth, or Video formats
        return 0;
    }
}


//=====================================================================================
// Input Layout Descriptor Utilities
//=====================================================================================

//-------------------------------------------------------------------------------------
// Validates a Direct3D Input Layout
//-------------------------------------------------------------------------------------

#if defined(__d3d11_h__) || defined(__d3d11_x_h__)
_Use_decl_annotations_
bool IsValid( const D3D11_INPUT_ELEMENT_DESC* vbDecl, size_t nDecl )
{
    if ( !vbDecl || !nDecl )
    {
        // Note that 0 is allowed by the runtime for degenerate cases, but is not defined for DirectXMesh
        return false;
    }

    if ( nDecl > D3D11_IA_VERTEX_INPUT_STRUCTURE_ELEMENT_COUNT )
    {
        // The upper-limit depends on feature level, so we assume highest value here
        return false;
    }

    for( size_t j = 0; j < nDecl; ++j )
    {
        size_t bpe = BytesPerElement( vbDecl[ j ].Format );
        if ( !bpe )
        {
            // Not a valid DXGI format or it's not valid for VB usage
            return false;
        }

        uint32_t alignment;

        if ( bpe == 1 )
            alignment = 1;
        else if ( bpe == 2 )
            alignment = 2;
        else
            alignment = 4;

        if ( ( vbDecl[ j ].AlignedByteOffset != D3D11_APPEND_ALIGNED_ELEMENT )
             && ( vbDecl[ j ].AlignedByteOffset % alignment ) != 0 )
        {
            // Invalid alignment for element
            return false;
        }

        if ( vbDecl[ j ].InputSlot >= D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT )
        {
            // The upper-limit depends on feature level, so we assume highest value here
            return false;
        }

        switch( vbDecl[ j ].InputSlotClass )
        {
        case D3D11_INPUT_PER_VERTEX_DATA:
            if ( vbDecl[ j ].InstanceDataStepRate != 0 )
            {
                return false;
            }
            break;

        case D3D11_INPUT_PER_INSTANCE_DATA:
            break;

        default:
            return false;
        }

        if ( !vbDecl[ j ].SemanticName )
        {
            return false;
        }

        // Debug layer also checks for trailing digit in the semantic name, and checks
        // for inconsistent semantic name/slot assignment.
    }

    return true;
}
#endif

#if defined(__d3d12_h__) || defined(__d3d12_x_h__)
bool IsValid( const D3D12_INPUT_LAYOUT_DESC& vbDecl )
{
    if ( !vbDecl.pInputElementDescs || !vbDecl.NumElements )
    {
        // Note that 0 is allowed by the runtime for degenerate cases, but is not defined for DirectXMesh
        return false;
    }

    if ( vbDecl.NumElements > D3D12_IA_VERTEX_INPUT_STRUCTURE_ELEMENT_COUNT )
    {
        return false;
    }

    for( size_t j = 0; j < vbDecl.NumElements; ++j )
    {
        size_t bpe = BytesPerElement( vbDecl.pInputElementDescs[ j ].Format );
        if ( !bpe )
        {
            // Not a valid DXGI format or it's not valid for VB usage
            return false;
        }

        uint32_t alignment;

        if ( bpe == 1 )
            alignment = 1;
        else if ( bpe == 2 )
            alignment = 2;
        else
            alignment = 4;

        if ( ( vbDecl.pInputElementDescs[ j ].AlignedByteOffset != D3D12_APPEND_ALIGNED_ELEMENT )
             && ( vbDecl.pInputElementDescs[ j ].AlignedByteOffset % alignment ) != 0 )
        {
            // Invalid alignment for element
            return false;
        }

        if ( vbDecl.pInputElementDescs[ j ].InputSlot >= D3D12_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT )
        {
            // The upper-limit depends on feature level, so we assume highest value here
            return false;
        }

        switch( vbDecl.pInputElementDescs[ j ].InputSlotClass )
        {
        case D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA:
            if ( vbDecl.pInputElementDescs[ j ].InstanceDataStepRate != 0 )
            {
                return false;
            }
            break;

        case D3D12_INPUT_CLASSIFICATION_PER_INSTANCE_DATA:
            break;

        default:
            return false;
        }

        if ( !vbDecl.pInputElementDescs[ j ].SemanticName )
        {
            return false;
        }

        // Debug layer also checks for trailing digit in the semantic name, and checks
        // for inconsistent semantic name/slot assignment.
    }

    return true;
}
#endif

//-------------------------------------------------------------------------------------
// Compute the offsets to each element, and total stride of each slot
//-------------------------------------------------------------------------------------

#if defined(__d3d11_h__) || defined(__d3d11_x_h__)
_Use_decl_annotations_
void ComputeInputLayout( const D3D11_INPUT_ELEMENT_DESC* vbDecl, size_t nDecl, uint32_t* offsets, uint32_t* strides )
{
    assert( IsValid( vbDecl, nDecl ) );

    if ( offsets )
        memset( offsets, 0, sizeof(uint32_t) * nDecl );

    if ( strides )
        memset( strides, 0, sizeof(uint32_t) * D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT );

    uint32_t prevABO[ D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT ];
    memset( prevABO, 0, sizeof(uint32_t) * D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT );

    for( size_t j = 0; j < nDecl; ++j )
    {
        uint32_t slot = vbDecl[ j ].InputSlot;
        if ( slot >= D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT )
        {
            // ignore bad input slots
            continue;
        }

        size_t bpe = BytesPerElement( vbDecl[ j ].Format );
        if ( !bpe )
        {
            // ignore invalid format
            continue;
        }

        uint32_t alignment;

        if ( bpe == 1 )
            alignment = 1;
        else if ( bpe == 2 )
            alignment = 2;
        else
            alignment = 4;

        uint32_t alignedByteOffset = vbDecl[ j ].AlignedByteOffset;

        if ( alignedByteOffset == D3D11_APPEND_ALIGNED_ELEMENT )
        {
            alignedByteOffset = prevABO[ slot ];
        }

        if ( offsets )
        {
            offsets[ j ] = alignedByteOffset;
        }
        
        if( strides )
        {
            uint32_t istride = uint32_t( alignedByteOffset + bpe );
            strides[ slot ] = std::max<uint32_t>( strides[ slot ], istride );
        }

        prevABO[ slot ] = uint32_t( alignedByteOffset + bpe + ( bpe % alignment ) );
    }
}
#endif

#if defined(__d3d12_h__) || defined(__d3d12_x_h__)
_Use_decl_annotations_
void ComputeInputLayout( const D3D12_INPUT_LAYOUT_DESC& vbDecl, uint32_t* offsets, uint32_t* strides )
{
    assert( IsValid( vbDecl ) );

    if ( offsets )
        memset( offsets, 0, sizeof(uint32_t) * vbDecl.NumElements);

    if ( strides )
        memset( strides, 0, sizeof(uint32_t) * D3D12_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT );

    uint32_t prevABO[ D3D12_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT ];
    memset( prevABO, 0, sizeof(uint32_t) * D3D12_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT );

    for( size_t j = 0; j < vbDecl.NumElements; ++j )
    {
        uint32_t slot = vbDecl.pInputElementDescs[ j ].InputSlot;
        if ( slot >= D3D12_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT )
        {
            // ignore bad input slots
            continue;
        }

        size_t bpe = BytesPerElement( vbDecl.pInputElementDescs[ j ].Format );
        if ( !bpe )
        {
            // ignore invalid format
            continue;
        }

        uint32_t alignment;

        if ( bpe == 1 )
            alignment = 1;
        else if ( bpe == 2 )
            alignment = 2;
        else
            alignment = 4;

        uint32_t alignedByteOffset = vbDecl.pInputElementDescs[ j ].AlignedByteOffset;

        if ( alignedByteOffset == D3D12_APPEND_ALIGNED_ELEMENT )
        {
            alignedByteOffset = prevABO[ slot ];
        }

        if ( offsets )
        {
            offsets[ j ] = alignedByteOffset;
        }
        
        if( strides )
        {
            uint32_t istride = uint32_t( alignedByteOffset + bpe );
            strides[ slot ] = std::max<uint32_t>( strides[ slot ], istride );
        }

        prevABO[ slot ] = uint32_t( alignedByteOffset + bpe + ( bpe % alignment ) );
    }
}
#endif

//=====================================================================================
// Attribute Utilities
//=====================================================================================
_Use_decl_annotations_
std::vector<std::pair<size_t,size_t>> ComputeSubsets( const uint32_t* attributes, size_t nFaces )
{
    std::vector<std::pair<size_t,size_t>> subsets;

    if ( !nFaces )
        return subsets;

    if ( !attributes )
    {
        subsets.emplace_back( std::pair<size_t,size_t>( 0, nFaces ) );
        return subsets;
    }

    uint32_t lastAttr = attributes[ 0 ];
    size_t offset = 0;
    size_t count = 1;

    for( size_t j = 1; j < nFaces; ++j )
    {
        if ( attributes[ j ] != lastAttr )
        {
            subsets.emplace_back( std::pair<size_t,size_t>( offset, count ) );
            lastAttr = attributes[ j ];
            offset = j;
            count = 1;
        }
        else
        {
            count += 1;
        }
    }

    if ( count > 0 )
    {
        subsets.emplace_back( std::pair<size_t,size_t>( offset, count ) );
    }

    return subsets;
}

} // namespace DirectX


//=====================================================================================
// Mesh Optimization Utilities
//=====================================================================================

namespace
{

template<class index_t>
void _ComputeVertexCacheMissRate( _In_reads_(nFaces*3) const index_t* indices, size_t nFaces, size_t nVerts, size_t cacheSize,
                                  float& acmr, float& atvr )
{
    acmr = -1.f;
    atvr = -1.f;

    if ( !indices || !nFaces || !nVerts || !cacheSize )
        return;

    if ( ( uint64_t(nFaces) * 3 ) >= UINT32_MAX )
        return;

    if ( nVerts >= index_t(-1) )
        return;

    size_t misses = 0;

    std::unique_ptr<uint32_t[]> fifo( new uint32_t[ cacheSize ] );
    size_t tail = 0;
    
    memset( fifo.get(), 0xff, sizeof(uint32_t) * cacheSize );

    for( size_t j = 0; j < (nFaces * 3); ++j )
    {
        if ( indices[ j ] == index_t(-1) )
            continue;

        bool found = false;

        for( size_t ptr = 0; ptr < cacheSize; ++ptr )
        {
            if ( fifo[ ptr ] == indices[ j ] )
            {
                found = true;
                break;
            }
        }

        if ( !found )
        {
            ++misses;
            fifo[ tail ] = indices[ j ];
            ++tail;
            if ( tail == cacheSize ) tail = 0;
        }
    }

    // ideal is 0.5, individual tris have 3.0
    acmr = float( misses ) / float( nFaces );

    // ideal is 1.0, worst case is 6.0
    atvr = float( misses ) / float( nVerts );
}

} // namespace

namespace DirectX
{

//-------------------------------------------------------------------------------------
_Use_decl_annotations_
void ComputeVertexCacheMissRate( const uint16_t* indices, size_t nFaces, size_t nVerts, size_t cacheSize,
                                 float& acmr, float& atvr )
{
    _ComputeVertexCacheMissRate<uint16_t>( indices, nFaces, nVerts, cacheSize, acmr, atvr );
}

_Use_decl_annotations_
void ComputeVertexCacheMissRate( const uint32_t* indices, size_t nFaces, size_t nVerts, size_t cacheSize,
                                 float& acmr, float& atvr )
{
    _ComputeVertexCacheMissRate<uint32_t>( indices, nFaces, nVerts, cacheSize, acmr, atvr );
}

} // namespace // DirectX