GuBV4_Internal.h

// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions
// are met:
//  * Redistributions of source code must retain the above copyright
//    notice, this list of conditions and the following disclaimer.
//  * Redistributions in binary form must reproduce the above copyright
//    notice, this list of conditions and the following disclaimer in the
//    documentation and/or other materials provided with the distribution.
//  * Neither the name of NVIDIA CORPORATION nor the names of its
//    contributors may be used to endorse or promote products derived
//    from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ''AS IS'' AND ANY
// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
// PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
// OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Copyright (c) 2008-2022 NVIDIA Corporation. All rights reserved.
// Copyright (c) 2004-2008 AGEIA Technologies, Inc. All rights reserved.
// Copyright (c) 2001-2004 NovodeX AG. All rights reserved.  
 
#ifndef GU_BV4_INTERNAL_H
#define GU_BV4_INTERNAL_H
 
#include "foundation/PxFPU.h"
 
    // PT: the general structure is that there is a root "process stream" function which is the entry point for the query.
    // It then calls "process node" functions for each traversed node, except for the Slabs-based raycast versions that deal
    // with 4 nodes at a time within the "process stream" function itself. When a leaf is found, "doLeafTest" functors
    // passed to the "process stream" entry point are called.
#ifdef GU_BV4_USE_SLABS
    // PT: Linux tries to compile templates even when they're not used so I had to wrap them all with defines to avoid build errors. Blame the only platform that does this.
    #ifdef GU_BV4_PROCESS_STREAM_NO_ORDER
        template<class LeafTestT, class ParamsT>
        PX_FORCE_INLINE PxIntBool processStreamNoOrder(const BV4Tree& tree, ParamsT* PX_RESTRICT params)
        {
            if(tree.mQuantized)
                return BV4_ProcessStreamSwizzledNoOrderQ<LeafTestT, ParamsT>(reinterpret_cast<const BVDataPackedQ*>(tree.mNodes), tree.mInitData, params);
            else
                return BV4_ProcessStreamSwizzledNoOrderNQ<LeafTestT, ParamsT>(reinterpret_cast<const BVDataPackedNQ*>(tree.mNodes), tree.mInitData, params);
        }
    #endif
 
    #ifdef GU_BV4_PROCESS_STREAM_ORDERED
        template<class LeafTestT, class ParamsT>
        PX_FORCE_INLINE void processStreamOrdered(const BV4Tree& tree, ParamsT* PX_RESTRICT params)
        {
            if(tree.mQuantized)
                BV4_ProcessStreamSwizzledOrderedQ<LeafTestT, ParamsT>(reinterpret_cast<const BVDataPackedQ*>(tree.mNodes), tree.mInitData, params);
            else
                BV4_ProcessStreamSwizzledOrderedNQ<LeafTestT, ParamsT>(reinterpret_cast<const BVDataPackedNQ*>(tree.mNodes), tree.mInitData, params);
        }
    #endif
 
    #ifdef GU_BV4_PROCESS_STREAM_RAY_NO_ORDER
        template<int inflateT, class LeafTestT, class ParamsT>
        PX_FORCE_INLINE PxIntBool processStreamRayNoOrder(const BV4Tree& tree, ParamsT* PX_RESTRICT params)
        {
            if(tree.mQuantized)
                return BV4_ProcessStreamKajiyaNoOrderQ<inflateT, LeafTestT, ParamsT>(reinterpret_cast<const BVDataPackedQ*>(tree.mNodes), tree.mInitData, params);
            else
                return BV4_ProcessStreamKajiyaNoOrderNQ<inflateT, LeafTestT, ParamsT>(reinterpret_cast<const BVDataPackedNQ*>(tree.mNodes), tree.mInitData, params);
        }
    #endif
 
    #ifdef GU_BV4_PROCESS_STREAM_RAY_ORDERED
        template<int inflateT, class LeafTestT, class ParamsT>
        PX_FORCE_INLINE void processStreamRayOrdered(const BV4Tree& tree, ParamsT* PX_RESTRICT params)
        {
            if(tree.mQuantized)
                BV4_ProcessStreamKajiyaOrderedQ<inflateT, LeafTestT, ParamsT>(reinterpret_cast<const BVDataPackedQ*>(tree.mNodes), tree.mInitData, params);
            else
                BV4_ProcessStreamKajiyaOrderedNQ<inflateT, LeafTestT, ParamsT>(reinterpret_cast<const BVDataPackedNQ*>(tree.mNodes), tree.mInitData, params);
        }
    #endif
#else
    #define processStreamNoOrder            BV4_ProcessStreamNoOrder
    #define processStreamOrdered            BV4_ProcessStreamOrdered2
    #define processStreamRayNoOrder(a, b)   BV4_ProcessStreamNoOrder<b>
    #define processStreamRayOrdered(a, b)   BV4_ProcessStreamOrdered2<b>
#endif
 
#ifndef GU_BV4_USE_SLABS
#ifdef  GU_BV4_PRECOMPUTED_NODE_SORT
    // PT: see http://www.codercorner.com/blog/?p=734
 
    // PT: TODO: refactor with dup in bucket pruner (TA34704)
    PX_FORCE_INLINE PxU32 computeDirMask(const PxVec3& dir)
    {
        // XYZ
        // ---
        // --+
        // -+-
        // -++
        // +--
        // +-+
        // ++-
        // +++
 
        const PxU32 X = PX_IR(dir.x)>>31;
        const PxU32 Y = PX_IR(dir.y)>>31;
        const PxU32 Z = PX_IR(dir.z)>>31;
        const PxU32 bitIndex = Z|(Y<<1)|(X<<2);
        return 1u<<bitIndex;
    }
 
    // 0    0    0      PP PN NP NN     0 1 2 3
    // 0    0    1      PP PN NN NP     0 1 3 2
    // 0    1    0      PN PP NP NN     1 0 2 3
    // 0    1    1      PN PP NN NP     1 0 3 2
    // 1    0    0      NP NN PP PN     2 3 0 1
    // 1    0    1      NN NP PP PN     3 2 0 1
    // 1    1    0      NP NN PN PP     2 3 1 0
    // 1    1    1      NN NP PN PP     3 2 1 0
    static const PxU8 order[] = {
        0,1,2,3,
        0,1,3,2,
        1,0,2,3,
        1,0,3,2,
        2,3,0,1,
        3,2,0,1,
        2,3,1,0,
        3,2,1,0,
    };
 
    PX_FORCE_INLINE PxU32 decodePNS(const BVDataPacked* PX_RESTRICT node, const PxU32 dirMask)
    {
        const PxU32 bit0 = (node[0].decodePNSNoShift() & dirMask) ? 1u : 0;
        const PxU32 bit1 = (node[1].decodePNSNoShift() & dirMask) ? 1u : 0;
        const PxU32 bit2 = (node[2].decodePNSNoShift() & dirMask) ? 1u : 0;     //### potentially reads past the end of the stream here!
        return bit2|(bit1<<1)|(bit0<<2);
    }
#endif  // GU_BV4_PRECOMPUTED_NODE_SORT
 
    #define PNS_BLOCK(i, a, b, c, d)                                        \
        case i:                                                             \
        {                                                                   \
            if(code & (1<<a))   { stack[nb++] = node[a].getChildData(); }   \
            if(code & (1<<b))   { stack[nb++] = node[b].getChildData(); }   \
            if(code & (1<<c))   { stack[nb++] = node[c].getChildData(); }   \
            if(code & (1<<d))   { stack[nb++] = node[d].getChildData(); }   \
        }break;
 
    #define PNS_BLOCK1(i, a, b, c, d)                                       \
        case i:                                                             \
        {                                                                   \
        stack[nb] = node[a].getChildData(); nb += (code & (1<<a))?1:0;      \
        stack[nb] = node[b].getChildData(); nb += (code & (1<<b))?1:0;      \
        stack[nb] = node[c].getChildData(); nb += (code & (1<<c))?1:0;      \
        stack[nb] = node[d].getChildData(); nb += (code & (1<<d))?1:0;      \
        }break;
 
    #define PNS_BLOCK2(a, b, c, d)  {                                       \
        if(code & (1<<a))   { stack[nb++] = node[a].getChildData(); }       \
        if(code & (1<<b))   { stack[nb++] = node[b].getChildData(); }       \
        if(code & (1<<c))   { stack[nb++] = node[c].getChildData(); }       \
        if(code & (1<<d))   { stack[nb++] = node[d].getChildData(); }   }   \
 
    template<class LeafTestT, class ParamsT>
    static PxIntBool BV4_ProcessStreamNoOrder(const BVDataPacked* PX_RESTRICT node, PxU32 initData, ParamsT* PX_RESTRICT params)
    {
        const BVDataPacked* root = node;
 
        PxU32 nb=1;
        PxU32 stack[GU_BV4_STACK_SIZE];
        stack[0] = initData;
 
        do
        {
            const PxU32 childData = stack[--nb];
            node = root + getChildOffset(childData);
            const PxU32 nodeType = getChildType(childData);
 
            if(nodeType>1 && BV4_ProcessNodeNoOrder<LeafTestT, 3>(stack, nb, node, params))
                return 1;
            if(nodeType>0 && BV4_ProcessNodeNoOrder<LeafTestT, 2>(stack, nb, node, params))
                return 1;
            if(BV4_ProcessNodeNoOrder<LeafTestT, 1>(stack, nb, node, params))
                return 1;
            if(BV4_ProcessNodeNoOrder<LeafTestT, 0>(stack, nb, node, params))
                return 1;
 
        }while(nb);
 
        return 0;
    }
 
    template<class LeafTestT, class ParamsT>
    static void BV4_ProcessStreamOrdered(const BVDataPacked* PX_RESTRICT node, PxU32 initData, ParamsT* PX_RESTRICT params)
    {
        const BVDataPacked* root = node;
 
        PxU32 nb=1;
        PxU32 stack[GU_BV4_STACK_SIZE];
        stack[0] = initData;
 
        const PxU32 dirMask = computeDirMask(params->mLocalDir)<<3;
 
        do
        {
            const PxU32 childData = stack[--nb];
            node = root + getChildOffset(childData);
 
            const PxU8* PX_RESTRICT ord = order + decodePNS(node, dirMask)*4;
            const PxU32 limit = 2 + getChildType(childData);
 
            BV4_ProcessNodeOrdered<LeafTestT>(stack, nb, node, params, ord[0], limit);
            BV4_ProcessNodeOrdered<LeafTestT>(stack, nb, node, params, ord[1], limit);
            BV4_ProcessNodeOrdered<LeafTestT>(stack, nb, node, params, ord[2], limit);
            BV4_ProcessNodeOrdered<LeafTestT>(stack, nb, node, params, ord[3], limit);
        }while(Nb);
    }
 
    // Alternative, experimental version using PNS
    template<class LeafTestT, class ParamsT>
    static void BV4_ProcessStreamOrdered2(const BVDataPacked* PX_RESTRICT node, PxU32 initData, ParamsT* PX_RESTRICT params)
    {
        const BVDataPacked* root = node;
 
        PxU32 nb=1;
        PxU32 stack[GU_BV4_STACK_SIZE];
        stack[0] = initData;
 
        const PxU32 X = PX_IR(params->mLocalDir_Padded.x)>>31;
        const PxU32 Y = PX_IR(params->mLocalDir_Padded.y)>>31;
        const PxU32 Z = PX_IR(params->mLocalDir_Padded.z)>>31;
        const PxU32 bitIndex = 3+(Z|(Y<<1)|(X<<2));
        const PxU32 dirMask = 1u<<bitIndex;
 
        do
        {
            const PxU32 childData = stack[--nb];
            node = root + getChildOffset(childData);
            const PxU32 nodeType = getChildType(childData);
 
            PxU32 code = 0;
            BV4_ProcessNodeOrdered2<LeafTestT, 0>(code, node, params);
            BV4_ProcessNodeOrdered2<LeafTestT, 1>(code, node, params);
            if(nodeType>0)
                BV4_ProcessNodeOrdered2<LeafTestT, 2>(code, node, params);
            if(nodeType>1)
                BV4_ProcessNodeOrdered2<LeafTestT, 3>(code, node, params);
 
            if(code)
            {
                // PT: TODO: check which implementation is best on each platform (TA34704)
#define FOURTH_TEST // Version avoids computing the PNS index, and also avoids all non-constant shifts. Full of branches though. Fastest on Win32.
#ifdef FOURTH_TEST
                {
                    if(node[0].decodePNSNoShift() & dirMask)    //  Bit2
                    {
                        if(node[1].decodePNSNoShift() & dirMask)    // Bit1
                        {
                            if(node[2].decodePNSNoShift() & dirMask)    // Bit0
                                PNS_BLOCK2(3,2,1,0) // 7
                            else
                                PNS_BLOCK2(2,3,1,0) // 6
                        }
                        else
                        {
                            if(node[2].decodePNSNoShift() & dirMask)    // Bit0
                                PNS_BLOCK2(3,2,0,1) // 5
                            else
                                PNS_BLOCK2(2,3,0,1) // 4
                        }
                    }
                    else
                    {
                        if(node[1].decodePNSNoShift() & dirMask)    // Bit1
                        {
                            if(node[2].decodePNSNoShift() & dirMask)    // Bit0
                                PNS_BLOCK2(1,0,3,2) // 3
                            else
                                PNS_BLOCK2(1,0,2,3) // 2
                        }
                        else
                        {
                            if(node[2].decodePNSNoShift() & dirMask)    // Bit0
                                PNS_BLOCK2(0,1,3,2) // 1
                            else
                                PNS_BLOCK2(0,1,2,3) // 0
                        }
                    }
                }
#endif
            }
        }while(nb);
    }
#endif  // GU_BV4_USE_SLABS
 
#endif // GU_BV4_INTERNAL_H