libraries/gfx/gfx_ParticleShape.cpp

/*---------------------------------------------------------------------------*
  Project:  NintendoWare
  File:     gfx_ParticleShape.cpp

  Copyright (C)2009-2011 Nintendo/HAL Laboratory, Inc.  All rights reserved.

  These coded instructions, statements, and computer programs contain proprietary
  information of Nintendo and/or its licensed developers and are protected by
  national and international copyright laws. They may not be disclosed to third
  parties or copied or duplicated in any form, in whole or in part, without the
  prior written consent of Nintendo.

  The content herein is highly confidential and should be handled accordingly.

  $Revision: 31311 $
 *---------------------------------------------------------------------------*/

#include "precompiled.h"

#include <nn/gx.h>

#include <nw/gfx/gfx_ParticleShape.h>
#include <nw/gfx/gfx_ISceneVisitor.h>
#include <nw/gfx/res/gfx_ResVertex.h>

#include <nw/ut/ut_ResUtil.h>
#include <nw/ut/ut_ResDictionary.h>
#include <nw/ut/ut_Foreach.h>

namespace nw
{
namespace gfx
{

NW_UT_RUNTIME_TYPEINFO_DEFINITION(ParticleShape, SceneObject);

static const int MAX_ATTRIBUTES_NUM = 12;

//----------------------------------------
static void
SetupParticleVertexAttributeCommand(
    ParticleShape* shape,
    ParticleSet* particleSet,
    ParticleBuffer bufferSide)
{
    ParticleCollection* collection = particleSet->GetParticleCollection();

    enum
    {
        REG_VTX_SHADER_ATTR_NUM   = 0x2b9, // [3:0] 頂点属性数 - 1
        REG_VTX_SHADER_ATTR_NUM_2 = 0x242, // [3:0] 頂点族整数 - 1
        REG_VTX_MAP_0             = 0x2bb, // [31:0] 入力レジスタのインデックス
        REG_VTX_MAP_1             = 0x2bc, // [15:0] 入力レジスタのインデックス
        REG_VTX_STREAM_BASE       = 0x200, // [28:1] 頂点アレイのベースアドレス
        REG_VTX_ARRAY_OFFSET      = 0x203, // [27:0] ロードアレイ0のアドレスオフセット
        REG_VTX_PARAM_INDEX       = 0x232, // [3:0] 頂点シェーダの入力番号指定
        REG_GEOM_MAP_0            = 0x28b, // [31:0] 入力レジスタのインデックス
        REG_GEOM_MAP_1            = 0x28c  // [15:0] 入力レジスタのインデックス
    };

    u32 baseAddr = nngxGetPhysicalAddr(nn::gx::GetVramStartAddr(nn::gx::MEM_VRAMA));

    s32 vtxAttrNum = shape->GetVertexAttributesCount();

    u32* command = reinterpret_cast<u32*>(shape->m_CommandCache[bufferSide]);

    const u32 HEADER_VTX_SHADER_ATTR_NUM   = internal::MakeCommandHeader(REG_VTX_SHADER_ATTR_NUM, 1, false, 0xb);
    const u32 HEADER_VTX_SHADER_ATTR_NUM_2 = internal::MakeCommandHeader(REG_VTX_SHADER_ATTR_NUM_2, 1, false, 0x1);
    const u32 HEADER_VTX_MAP_0             = internal::MakeCommandHeader(REG_VTX_MAP_0, 2, true, 0xF);


    command[0] = (vtxAttrNum - 1) | 0xa0000000;
    command[1] = HEADER_VTX_SHADER_ATTR_NUM;

    command[2] = (vtxAttrNum - 1);
    command[3] = HEADER_VTX_SHADER_ATTR_NUM_2;

    command[4] = 0;
    command[5] = HEADER_VTX_MAP_0;
    command[6] = 0;
    command[7] = 0;

    NW_ASSERT((baseAddr & 0x0f) == 0);
    command[8] = baseAddr >> 3;
    command[9] = internal::MakeCommandHeader(REG_VTX_STREAM_BASE, 3 + vtxAttrNum * 3, true, 0xF);
    command[10] = 0;
    command[11] = static_cast<u32>(vtxAttrNum - 1) << 28;

    u32* inputTable  = &command[4];
    u32* inputFormat = &command[10];
    u32* vertexParamMask = &command[11];

    int inputIndex = 0; // 内部頂点属性番号
    u32 commandIndex = 12;

    u32 usedFlag = 0;

    // 頂点ストリームを番号の若い順に設定
    for (s32 i = 0; i < shape->GetVertexAttributesCount(); ++i)
    {
        if (shape->IsVertexStream(i))
        {
            s32 usage = shape->GetVertexAttributeUsage(i);
            NW_ASSERT(0 <= usage && usage < 12);
            inputTable[ (inputIndex / 8) * 2 ] |= (usage & 0xF) << (4 * (inputIndex % 8));
            usedFlag |= 0x1 << usage;

            u32 format    = shape->GetVertexAttributeFormatType(i);
            u32 dimension = shape->GetVertexAttributeDimension(i);

            inputFormat[ inputIndex / 8 ] |= internal::CommandCacheHelper::GetVertexFormat(dimension, format) << ((inputIndex % 8) * 4);

            // ロードアレイの設定
            u8* streamPtr = shape->GetVertexStreamPtr(i, bufferSide);
            NW_NULL_ASSERT(streamPtr);
            u32 bufferAddr = nngxGetPhysicalAddr(reinterpret_cast<uptr>(streamPtr));

            shape->SetVertexAttributeCommandPtr(i, bufferSide, &command[commandIndex]);

            // このクラスでは interleave 形式には対応しないので、(0番目の要素 == 内部頂点属性 index 番目)
            NW_ASSERT((bufferAddr - baseAddr) < 0x10000000);
            command[commandIndex++] = bufferAddr - baseAddr;
            command[commandIndex++] = inputIndex;
            command[commandIndex++] = (internal::CommandCacheHelper::GetVertexSize(dimension, format) << 16) + (1 <<  28);

            ++inputIndex;
        }
    }

    // パラメータの分を設定
    for (s32 i = inputIndex; i < vtxAttrNum; ++i)
    {
        command[commandIndex++] = 0;
        command[commandIndex++] = 0;
        command[commandIndex++] = 0;
    }

    // 64bit単位になるようにパディング
    if ((commandIndex & 1) == 1)
    {
        command[commandIndex++] = 0;
    }

    // 頂点ストリームの後に頂点パラメータを設定。
    for (s32 i = 0; i < shape->GetVertexAttributesCount(); ++i)
    {
        if (!shape->IsVertexStream(i))
        {
            s32 usage = shape->GetVertexAttributeUsage(i);
            NW_ASSERT(0 <= usage && usage < 12);
            inputTable[ (inputIndex / 8) * 2 ] |= (usage & 0xF) << (4 * (inputIndex % 8));
            usedFlag |= 0x1 << usage;

            // 頂点パラメータの場合の処理

            u32 data[4] = { 0, 0, 0, 0 };

            int count = shape->GetVertexAttributeDimension(i);
            f32* fdata = shape->GetVertexParameter(i);

            for (int j = 0; j < count; ++j)
            {
                data[j] = ut::Float24::Float32ToBits24(fdata[j]);
            }

            const u32 HEADER_VTX_PARAM_INDEX = internal::MakeCommandHeader(REG_VTX_PARAM_INDEX, 4, true, 0xF);

            shape->SetVertexAttributeCommandPtr(i, bufferSide, &command[commandIndex + 2]);

            command[commandIndex]     = inputIndex;
            command[commandIndex + 1] = HEADER_VTX_PARAM_INDEX;
            command[commandIndex + 2] = (data[3] <<  8) | (data[2] >> 16);
            command[commandIndex + 3] = (data[2] << 16) | (data[1] >> 8);
            command[commandIndex + 4] = (data[1] << 24) | (data[0]);
            command[commandIndex + 5] = 0;

            vertexParamMask[0] |= 1 << (16 + inputIndex);

            ++inputIndex;
            commandIndex += 6;
        }
    }

    const u32 HEADER_GEOM_MAP_0 = internal::MakeCommandHeader(REG_GEOM_MAP_0, 2, true, 0xF);

    command[commandIndex + 0] = 0x76543210;
    command[commandIndex + 1] = HEADER_GEOM_MAP_0;
    command[commandIndex + 2] = 0xfedcba98;
    command[commandIndex + 3] = 0;
    commandIndex += 4;

    shape->m_CommandCacheSize[bufferSide] = commandIndex * sizeof(u32);
}


//----------------------------------------
static void
SetupDeactivateParticleVertexAttributeCommand(
    ParticleShape* shape)
{
    enum
    {
        REG_VTX_ARRAY_OFFSET      = 0x203, // [27:0] ロードアレイ0のアドレスオフセット
        REG_VTX_PARAM_INDEX       = 0x232  // [3:0] 頂点シェーダの入力番号指定
    };

    s32 vtxAttrNum = shape->GetVertexAttributesCount();

    u32* command = reinterpret_cast<u32*>(shape->m_DeactivateVertexCommandCache);

    shape->m_DeactivateVertexCommandCache = command;

    int inputIndex = 0; // 内部頂点属性番号
    u32 commandIndex = 0;

    // 頂点ストリームを番号の若い順に設定
    for (s32 i = 0; i < vtxAttrNum; ++i)
    {
        if (shape->IsVertexStream(i))
        {
            // ロードアレイの無効化
            // このクラスでは interleave 形式には対応しないので、(0番目の要素 == 内部頂点属性 index 番目)
            command[commandIndex++] = 0;
            command[commandIndex++] = internal::MakeCommandHeader(REG_VTX_ARRAY_OFFSET + 2 + 3 * inputIndex, 1, false, 0xF);

            ++inputIndex;
        }
    }

    const u32 HEADER_VTX_PARAM_INDEX = internal::MakeCommandHeader(REG_VTX_PARAM_INDEX, 4, true, 0xF);

    // (0,0,0,0) の固定属性を設定
    for ( int i = 0; i < inputIndex; ++i )
    {
        command[commandIndex++] = i;
        command[commandIndex++] = HEADER_VTX_PARAM_INDEX;
        command[commandIndex++] = 0;
        command[commandIndex++] = 0;
        command[commandIndex++] = 0;
        command[commandIndex++] = 0;
    }

    shape->m_DeactivateVertexCommandCacheSize = commandIndex * sizeof(u32);

    NW_ASSERT( shape->m_DeactivateVertexCommandCacheSize <=  sizeof(u32) * (MAX_ATTRIBUTES_NUM * 8) );
}

static const int PrimitiveCommandSize = 26;

//----------------------------------------
static void
CreatePrimitiveCommandCache(
    ParticleShape* shape,
    ResParticleSet resParticleSet)
{
    enum
    {
        REG_VERTEX_UNIFORM_BOOL     = 0x2b0,

        REG_ELEMENTS_MODE           = 0x229, // [23:16]にはバイトイネーブルでアクセスしてはいけない。
        REG_ELEMENTS_MODE_2         = 0x253, // [31:16]にはバイトイネーブルでアクセスしてはいけない。
        REG_ELEMENTS_MODE_3         = 0x25e, // [31:16]にはバイトイネーブルでアクセスしてはいけない。
        REG_TRIANGLE_INDEX_RESET    = 0x25f,
        REG_DRAW_READY              = 0x245,
        REG_DRAW_KICK               = 0x22f,
        REG_VERTEX_CACHE_CLEAR      = 0x231,
        REG_COLOR_DEPTH_CACHE_CLEAR = 0x110,
        REG_COLOR_DEPTH_CACHE_FLUSH = 0x111
    };

    u32* command = reinterpret_cast<u32*>(shape->m_PrimitiveCommandCache);

    u32 commandIndex = 0;

    s32 type;
    switch (resParticleSet.GetParticleShapeBuilder().GetTypeInfo())
    {
    case ResParticleBillboardShapeBuilder::TYPE_INFO:
        type = 0;
        break;
    case ResParticleWorldBillboardShapeBuilder::TYPE_INFO:
        type = 8;
        break;
    case ResParticleYBillboardShapeBuilder::TYPE_INFO:
        type = 2 | 1;
        break;
    case ResParticleXyPlaneShapeBuilder::TYPE_INFO:
        type = 4 | 1;
        break;
    default:
        NW_FATAL_ERROR("unknown shapebuilder type");
        break;
    }

    const u32 HEADER_VERTEX_UNIFORM_BOOL     = internal::MakeCommandHeader(REG_VERTEX_UNIFORM_BOOL, 5, true, 0xF);

    const u32 HEADER_ELEMENTS_MODE           = internal::MakeCommandHeader(REG_ELEMENTS_MODE, 1, false, 0x2);
    const u32 HEADER_ELEMENTS_MODE_2         = internal::MakeCommandHeader(REG_ELEMENTS_MODE_2, 1, false, 0x2);
    const u32 HEADER_ELEMENTS_MODE_3         = internal::MakeCommandHeader(REG_ELEMENTS_MODE_3, 1, false, 0x2);
    const u32 HEADER_TRIANGLE_INDEX_RESET    = internal::MakeCommandHeader(REG_TRIANGLE_INDEX_RESET, 1, false, 0xF);
    const u32 HEADER_DRAW_READY              = internal::MakeCommandHeader(REG_DRAW_READY, 1, false, 0xF);
    const u32 HEADER_DRAW_KICK               = internal::MakeCommandHeader(REG_DRAW_KICK, 1, false, 0xF);
    const u32 HEADER_VERTEX_CACHE_CLEAR      = internal::MakeCommandHeader(REG_VERTEX_CACHE_CLEAR, 1, false, 0xF);
    const u32 HEADER_COLOR_DEPTH_CACHE_FLUSH = internal::MakeCommandHeader(REG_COLOR_DEPTH_CACHE_FLUSH, 1, false, 0xF);
    const u32 HEADER_COLOR_DEPTH_CACHE_CLEAR = internal::MakeCommandHeader(REG_COLOR_DEPTH_CACHE_CLEAR, 1, false, 0xF);

    command[commandIndex++] = 0x7fff0000 | (type << 1); // bit0は予約されているため、bit1から配置する
    command[commandIndex++] = HEADER_VERTEX_UNIFORM_BOOL;
    command[commandIndex++] = 0;
    command[commandIndex++] = 0;
    command[commandIndex++] = 0;
    command[commandIndex++] = 0;

    command[commandIndex++] = 0;
    command[commandIndex++] = HEADER_ELEMENTS_MODE;
    command[commandIndex++] = 0;
    command[commandIndex++] = HEADER_ELEMENTS_MODE_2;
    command[commandIndex++] = 3 << 8;
    command[commandIndex++] = HEADER_ELEMENTS_MODE_3;
    command[commandIndex++] = 1;
    command[commandIndex++] = HEADER_TRIANGLE_INDEX_RESET;
    command[commandIndex++] = 0;
    command[commandIndex++] = HEADER_DRAW_READY;
    command[commandIndex++] = 1;
    command[commandIndex++] = HEADER_DRAW_KICK;
    command[commandIndex++] = 1;
    command[commandIndex++] = HEADER_DRAW_READY;
    command[commandIndex++] = 1;
    command[commandIndex++] = HEADER_VERTEX_CACHE_CLEAR;
    command[commandIndex++] = 1;
    command[commandIndex++] = HEADER_COLOR_DEPTH_CACHE_FLUSH;
    command[commandIndex++] = 1;
    command[commandIndex++] = HEADER_COLOR_DEPTH_CACHE_CLEAR;

    NW_ASSERT(commandIndex == PrimitiveCommandSize);

    shape->m_PrimitiveCommandCacheSize = commandIndex * sizeof(u32);
}

//----------------------------------------
void
ParticleShape::CreateCommandCache(
    ParticleSet* particleSet
)
{
    for (int side = 0; side < 2; ++side)
    {
        SetupParticleVertexAttributeCommand(
            this,
            particleSet,
            (ParticleBuffer)side);
    }

    SetupDeactivateParticleVertexAttributeCommand(this);

    CreatePrimitiveCommandCache(
        this,
        particleSet->GetResParticleSet());
}

void
ParticleShape::GetMemorySizeInternal(
    os::MemorySizeCalculator* pSize,
    int capacity)
{
    os::MemorySizeCalculator& size = *pSize;

    const int streamSize = (capacity + internal::PARTICLE_SIMD_WIDTH_MAX) * sizeof(u16);
    const int drawCommandCacheSize = sizeof(u32) * (12 + MAX_ATTRIBUTES_NUM * 6 + MAX_ATTRIBUTES_NUM * 4 + 4);
    const int primitiveCommandCacheSize = sizeof(u32) * PrimitiveCommandSize;
    const int deactivateVertexCommandCacheSize = sizeof(u32) * (MAX_ATTRIBUTES_NUM * 8);

    int deviceMemorySize = 0;
    if (size.GetAlignment() < 32)
    {
        deviceMemorySize += 32 - size.GetAlignment(); // 最悪値
    }

    deviceMemorySize = ut::RoundUp(deviceMemorySize, 32);
    deviceMemorySize += streamSize;
    deviceMemorySize = ut::RoundUp(deviceMemorySize, 32);
    deviceMemorySize += streamSize;

    int nodeMemorySize = sizeof(ParticleShape);
    nodeMemorySize = ut::RoundUp(nodeMemorySize, 32);
    nodeMemorySize += drawCommandCacheSize;
    nodeMemorySize = ut::RoundUp(nodeMemorySize, 32);
    nodeMemorySize += drawCommandCacheSize;
    nodeMemorySize = ut::RoundUp(nodeMemorySize, 32);
    nodeMemorySize += deactivateVertexCommandCacheSize;
    nodeMemorySize = ut::RoundUp(nodeMemorySize, 32);
    nodeMemorySize += primitiveCommandCacheSize;
    nodeMemorySize = ut::RoundUp(nodeMemorySize, 32);

    size.Add(nodeMemorySize, 32);
}

void
ParticleShape::GetDeviceMemorySizeInternal(
    os::MemorySizeCalculator* pSize,
    int capacity)
{
    os::MemorySizeCalculator& size = *pSize;

    const int streamSize = (capacity + internal::PARTICLE_SIMD_WIDTH_MAX) * sizeof(u16);
    const int drawCommandCacheSize = sizeof(u32) * (12 + MAX_ATTRIBUTES_NUM * 6 + MAX_ATTRIBUTES_NUM * 4 + 4);
    const int primitiveCommandCacheSize = sizeof(u32) * PrimitiveCommandSize;
    const int deactivateVertexCommandCacheSize = sizeof(u32) * (MAX_ATTRIBUTES_NUM * 8);

    int deviceMemorySize = 0;
    if (size.GetAlignment() < 32)
    {
        deviceMemorySize += 32 - size.GetAlignment(); // 最悪値
    }

    deviceMemorySize = ut::RoundUp(deviceMemorySize, 32);
    deviceMemorySize += streamSize;
    deviceMemorySize = ut::RoundUp(deviceMemorySize, 32);
    deviceMemorySize += streamSize;

    size.Add(deviceMemorySize, 32);
}

//----------------------------------------
ParticleShape*
ParticleShape::Create(
    ResSceneObject resource,
    int capacity,
    os::IAllocator* mainAllocator,
    os::IAllocator* deviceAllocator
)
{
    NW_NULL_ASSERT(mainAllocator);
    NW_NULL_ASSERT(deviceAllocator);

    ResParticleShape resNode = ResDynamicCast<ResParticleShape>(resource);
    NW_ASSERT(resNode.IsValid());

    const int streamSize = (capacity + internal::PARTICLE_SIMD_WIDTH_MAX) * sizeof(u16);
    const int drawCommandCacheSize = sizeof(u32) * (12 + MAX_ATTRIBUTES_NUM * 6 + MAX_ATTRIBUTES_NUM * 4 + 4);
    const int primitiveCommandCacheSize = sizeof(u32) * PrimitiveCommandSize;
    const int deactivateVertexCommandCacheSize = sizeof(u32) * (MAX_ATTRIBUTES_NUM * 8);

    int deviceMemorySize = 0;
    deviceMemorySize = ut::RoundUp(deviceMemorySize, 32);
    deviceMemorySize += streamSize;
    deviceMemorySize = ut::RoundUp(deviceMemorySize, 32);
    deviceMemorySize += streamSize;

    int nodeMemorySize = sizeof(ParticleShape);
    nodeMemorySize = ut::RoundUp(nodeMemorySize, 32);
    nodeMemorySize += drawCommandCacheSize;
    nodeMemorySize = ut::RoundUp(nodeMemorySize, 32);
    nodeMemorySize += drawCommandCacheSize;
    nodeMemorySize = ut::RoundUp(nodeMemorySize, 32);
    nodeMemorySize += deactivateVertexCommandCacheSize;
    nodeMemorySize = ut::RoundUp(nodeMemorySize, 32);
    nodeMemorySize += primitiveCommandCacheSize;
    nodeMemorySize = ut::RoundUp(nodeMemorySize, 32);

    u8* devicememory = reinterpret_cast<u8*>(deviceAllocator->Alloc(deviceMemorySize, 32));
    if (devicememory == NULL)
    {
        return NULL;
    }

    u8* nodememory = reinterpret_cast<u8*>(mainAllocator->Alloc(nodeMemorySize, 32));
    if (nodememory == NULL)
    {
        deviceAllocator->Free(devicememory);
        return NULL;
    }

    ParticleShape* node = new(nodememory) ParticleShape(
        capacity,
        mainAllocator,
        deviceAllocator,
        devicememory,
        resNode);
    nodememory += sizeof(ParticleShape);

    for (int i = 0; i < 2; ++i)
    {
        devicememory = reinterpret_cast<u8*>(ut::RoundUp(devicememory, 32));
        node->m_PrimitiveBuffer[i] = devicememory;
        devicememory += streamSize;

        //std::memset(node->m_PrimitiveBuffer[i], 0, streamSize);

        u32 baseAddr = nngxGetPhysicalAddr(nn::gx::GetVramStartAddr(nn::gx::MEM_VRAMA));
        u32 bufferAddr = nngxGetPhysicalAddr(reinterpret_cast<uptr>(node->m_PrimitiveBuffer[i]));

        NW_ASSERT((bufferAddr - baseAddr) < 0x10000000);
        node->m_PrimitiveBufferOffset[i] = bufferAddr - baseAddr;
    }

    for (int side = 0; side < 2; ++side)
    {
        nodememory = reinterpret_cast<u8*>(ut::RoundUp(nodememory, 32));
        node->m_CommandCache[side] = reinterpret_cast<u32*>(nodememory);
        nodememory += drawCommandCacheSize;
    }

    nodememory = reinterpret_cast<u8*>(ut::RoundUp(nodememory, 32));
    node->m_DeactivateVertexCommandCache = reinterpret_cast<u32*>(nodememory);
    nodememory += deactivateVertexCommandCacheSize;

    nodememory = reinterpret_cast<u8*>(ut::RoundUp(nodememory, 32));
    node->m_PrimitiveCommandCache = reinterpret_cast<u32*>(nodememory);
    nodememory += primitiveCommandCacheSize;

    return node;
}

int
ParticleShape::AddVertexStreamSize(
    u32 formatType,
    int dimension,
    int capacity,
    int prevSize
)
{
    int formatSize = 1;
    switch (formatType)
    {
    case GL_FLOAT:
        formatSize = 4;
        break;
    default:
        NW_FATAL_ERROR("unsupported formatType");
    }

    int streamSize = (capacity + internal::PARTICLE_SIMD_WIDTH_MAX) * dimension * formatSize;

    prevSize = ut::RoundUp(prevSize, 32);
    prevSize += streamSize;
    prevSize = ut::RoundUp(prevSize, 32);
    prevSize += streamSize;

    return prevSize + streamSize;
}

ParticleShape::VertexAttribute*
ParticleShape::AddVertexStream(
    s32 usage,
    u32 formatType,
    int dimension,
    int capacity,
    u8** memory
)
{
    int formatSize = 1;
    switch (formatType)
    {
    case GL_FLOAT:
        formatSize = 4;
        break;
    default:
        NW_FATAL_ERROR("unsupported formatType");
    }

    *memory = reinterpret_cast<u8*>(ut::RoundUp(*memory, 32));

    int streamSize = (capacity + internal::PARTICLE_SIMD_WIDTH_MAX) * dimension * formatSize;
    void* stream0 = *memory;
    *memory += streamSize;

    void* stream1 = *memory;
    *memory += streamSize;

    std::memset(stream0, 0, streamSize);
    std::memset(stream1, 0, streamSize);

    return this->AddVertexAttribute(
        usage,
        formatType,
        dimension,
        true,
        reinterpret_cast<u8*>(stream0),
        reinterpret_cast<u8*>(stream1));
}

int
ParticleShape::AddVertexParamSize(
    u32 formatType,
    int dimension,
    int prevSize
)
{
    int formatSize = 1;
    switch (formatType)
    {
    case GL_FLOAT:
        formatSize = 4;
        break;
    default:
        NW_FATAL_ERROR("unsupported formatType");
    }

    prevSize = ut::RoundUp(prevSize, 32);

    int streamSize = dimension * formatSize;

    return prevSize + streamSize;
}

ParticleShape::VertexAttribute*
ParticleShape::AddVertexParam(
    s32 usage,
    u32 formatType,
    int dimension,
    f32* parameters,
    u8** memory
)
{
    NW_UNUSED_VARIABLE(parameters);

    int formatSize = 1;
    switch (formatType)
    {
    case GL_FLOAT:
        formatSize = 4;
        break;
    default:
        NW_FATAL_ERROR("unsupported formatType");
    }

    int streamSize = dimension * formatSize;
    *memory = reinterpret_cast<u8*>(ut::RoundUp(*memory, 32));
    void* stream = *memory;
    *memory += streamSize;

    nw::os::MemCpy(stream, parameters, streamSize);

    return this->AddVertexAttribute(
        usage,
        formatType,
        dimension,
        false,
        reinterpret_cast<u8*>(stream),
        NULL);
}

//----------------------------------------
ParticleShape::ParticleShape(
    int capacity,
    os::IAllocator* allocator,
    os::IAllocator* deviceAllocator,
    void* deviceMemory,
    ResParticleShape resObj)
    : SceneObject(allocator, resObj),
    m_Capacity(capacity),
    m_BufferSide(false),
    m_ResVertexAttributeDataCount(0),
    m_DeviceAllocator(deviceAllocator),
    m_DeviceMemory(deviceMemory)
{
    m_PrimitiveBuffer[0] = NULL;
    m_PrimitiveBuffer[1] = NULL;
    m_PrimitiveBufferOffset[0] = 0;
    m_PrimitiveBufferOffset[1] = 0;

    for (int usage = 0; usage < PARTICLEUSAGE_COUNT; ++usage)
    {
        for (int side = 0; side < 2; ++side)
        {
            m_VertexAttribute[usage].m_Stream[side] = NULL;
            m_VertexAttribute[usage].m_CommandPtr[side] = NULL;
        }
    }

    for (int i = 0; i < 2; ++i)
    {
        m_CommandCache[i] = NULL;
        m_CommandCacheSize[i] = 0;
    }

    m_DeactivateVertexCommandCache = NULL;
    m_DeactivateVertexCommandCacheSize = 0;
    m_PrimitiveCommandCache = NULL;
    m_PrimitiveCommandCacheSize = 0;
}

//----------------------------------------
ParticleShape::~ParticleShape()
{
    m_DeviceAllocator->Free(m_DeviceMemory);

    for (int i = 0; i < 2; ++ i)
    {
        if (this->m_CommandCache[i] != NULL)
        {
            this->m_CommandCache[i] = NULL;
            this->m_CommandCacheSize[i] = 0;
        }
    }

    if (this->m_DeactivateVertexCommandCache != NULL)
    {
        this->m_DeactivateVertexCommandCache = NULL;
        this->m_DeactivateVertexCommandCacheSize = 0;
    }

    if (this->m_PrimitiveCommandCache != NULL)
    {
        this->m_PrimitiveCommandCache = NULL;
        this->m_PrimitiveCommandCacheSize = 0;
    }
}

void
ParticleShape::FlushBuffer()
{
    for (s32 i = 0; i < this->GetVertexAttributesCount(); ++i)
    {
        if (this->IsVertexStream(i))
        {
            nngxUpdateBuffer(
                this->GetVertexStreamPtr(i, PARTICLE_BUFFER_FRONT),
                this->GetVertexCapacity() * this->GetVertexAttributeDimension(i) * sizeof(f32));
        }
    }

    nngxUpdateBuffer(
        this->GetPrimitiveStreamPtr(PARTICLE_BUFFER_FRONT),
        this->GetVertexCapacity() * sizeof(u16));
}

} // namespace gfx
} // namespace nw