1 /* 2 PTEX SOFTWARE 3 Copyright 2014 Disney Enterprises, Inc. All rights reserved 4 5 Redistribution and use in source and binary forms, with or without 6 modification, are permitted provided that the following conditions are 7 met: 8 9 * Redistributions of source code must retain the above copyright 10 notice, this list of conditions and the following disclaimer. 11 12 * Redistributions in binary form must reproduce the above copyright 13 notice, this list of conditions and the following disclaimer in 14 the documentation and/or other materials provided with the 15 distribution. 16 17 * The names "Disney", "Walt Disney Pictures", "Walt Disney Animation 18 Studios" or the names of its contributors may NOT be used to 19 endorse or promote products derived from this software without 20 specific prior written permission from Walt Disney Pictures. 21 22 Disclaimer: THIS SOFTWARE IS PROVIDED BY WALT DISNEY PICTURES AND 23 CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, 24 BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS 25 FOR A PARTICULAR PURPOSE, NONINFRINGEMENT AND TITLE ARE DISCLAIMED. 26 IN NO EVENT SHALL WALT DISNEY PICTURES, THE COPYRIGHT HOLDER OR 27 CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 28 EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 29 PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 30 PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND BASED ON ANY 31 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 32 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 33 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. 34 */ 35 #include "PtexPlatform.h" 36 #include "PtexUtils.h" 37 #include "PtexHalf.h" 38 #include "PtexSeparableKernel.h" 39 40 PTEX_NAMESPACE_BEGIN 41 42 namespace { 43 // apply to 1..4 channels (unrolled channel loop) of packed data (nTxChan==nChan) 44 template<class T, int nChan> Apply(PtexSeparableKernel & k,float * result,void * data,int,int)45 void Apply(PtexSeparableKernel& k, float* result, void* data, int /*nChan*/, int /*nTxChan*/) 46 { 47 float* rowResult = (float*) alloca(nChan*sizeof(float)); 48 int rowlen = k.res.u() * nChan; 49 int datalen = k.uw * nChan; 50 int rowskip = rowlen - datalen; 51 float* kvp = k.kv; 52 T* p = static_cast<T*>(data) + (k.v * k.res.u() + k.u) * nChan; 53 T* pEnd = p + k.vw * rowlen; 54 while (p != pEnd) 55 { 56 float* kup = k.ku; 57 T* pRowEnd = p + datalen; 58 // just mult and copy first element 59 PtexUtils::VecMult<T,nChan>()(rowResult, p, *kup++); 60 p += nChan; 61 // accumulate remaining elements 62 while (p != pRowEnd) { 63 // rowResult[i] = p[i] * ku[u] for i in {0..n-1} 64 PtexUtils::VecAccum<T,nChan>()(rowResult, p, *kup++); 65 p += nChan; 66 } 67 // result[i] += rowResult[i] * kv[v] for i in {0..n-1} 68 PtexUtils::VecAccum<float,nChan>()(result, rowResult, *kvp++); 69 p += rowskip; 70 } 71 } 72 73 // apply to 1..4 channels (unrolled channel loop) w/ pixel stride 74 template<class T, int nChan> ApplyS(PtexSeparableKernel & k,float * result,void * data,int,int nTxChan)75 void ApplyS(PtexSeparableKernel& k, float* result, void* data, int /*nChan*/, int nTxChan) 76 { 77 float* rowResult = (float*) alloca(nChan*sizeof(float)); 78 int rowlen = k.res.u() * nTxChan; 79 int datalen = k.uw * nTxChan; 80 int rowskip = rowlen - datalen; 81 float* kvp = k.kv; 82 T* p = static_cast<T*>(data) + (k.v * k.res.u() + k.u) * nTxChan; 83 T* pEnd = p + k.vw * rowlen; 84 while (p != pEnd) 85 { 86 float* kup = k.ku; 87 T* pRowEnd = p + datalen; 88 // just mult and copy first element 89 PtexUtils::VecMult<T,nChan>()(rowResult, p, *kup++); 90 p += nTxChan; 91 // accumulate remaining elements 92 while (p != pRowEnd) { 93 // rowResult[i] = p[i] * ku[u] for i in {0..n-1} 94 PtexUtils::VecAccum<T,nChan>()(rowResult, p, *kup++); 95 p += nTxChan; 96 } 97 // result[i] += rowResult[i] * kv[v] for i in {0..n-1} 98 PtexUtils::VecAccum<float,nChan>()(result, rowResult, *kvp++); 99 p += rowskip; 100 } 101 } 102 103 // apply to N channels (general case) 104 template<class T> ApplyN(PtexSeparableKernel & k,float * result,void * data,int nChan,int nTxChan)105 void ApplyN(PtexSeparableKernel& k, float* result, void* data, int nChan, int nTxChan) 106 { 107 float* rowResult = (float*) alloca(nChan*sizeof(float)); 108 int rowlen = k.res.u() * nTxChan; 109 int datalen = k.uw * nTxChan; 110 int rowskip = rowlen - datalen; 111 float* kvp = k.kv; 112 T* p = static_cast<T*>(data) + (k.v * k.res.u() + k.u) * nTxChan; 113 T* pEnd = p + k.vw * rowlen; 114 while (p != pEnd) 115 { 116 float* kup = k.ku; 117 T* pRowEnd = p + datalen; 118 // just mult and copy first element 119 PtexUtils::VecMultN<T>()(rowResult, p, nChan, *kup++); 120 p += nTxChan; 121 // accumulate remaining elements 122 while (p != pRowEnd) { 123 // rowResult[i] = p[i] * ku[u] for i in {0..n-1} 124 PtexUtils::VecAccumN<T>()(rowResult, p, nChan, *kup++); 125 p += nTxChan; 126 } 127 // result[i] += rowResult[i] * kv[v] for i in {0..n-1} 128 PtexUtils::VecAccumN<float>()(result, rowResult, nChan, *kvp++); 129 p += rowskip; 130 } 131 } 132 } 133 134 135 136 PtexSeparableKernel::ApplyFn 137 PtexSeparableKernel::applyFunctions[] = { 138 // nChan == nTxChan 139 ApplyN<uint8_t>, ApplyN<uint16_t>, ApplyN<PtexHalf>, ApplyN<float>, 140 Apply<uint8_t,1>, Apply<uint16_t,1>, Apply<PtexHalf,1>, Apply<float,1>, 141 Apply<uint8_t,2>, Apply<uint16_t,2>, Apply<PtexHalf,2>, Apply<float,2>, 142 Apply<uint8_t,3>, Apply<uint16_t,3>, Apply<PtexHalf,3>, Apply<float,3>, 143 Apply<uint8_t,4>, Apply<uint16_t,4>, Apply<PtexHalf,4>, Apply<float,4>, 144 145 // nChan != nTxChan (need pixel stride) 146 ApplyN<uint8_t>, ApplyN<uint16_t>, ApplyN<PtexHalf>, ApplyN<float>, 147 ApplyS<uint8_t,1>, ApplyS<uint16_t,1>, ApplyS<PtexHalf,1>, ApplyS<float,1>, 148 ApplyS<uint8_t,2>, ApplyS<uint16_t,2>, ApplyS<PtexHalf,2>, ApplyS<float,2>, 149 ApplyS<uint8_t,3>, ApplyS<uint16_t,3>, ApplyS<PtexHalf,3>, ApplyS<float,3>, 150 ApplyS<uint8_t,4>, ApplyS<uint16_t,4>, ApplyS<PtexHalf,4>, ApplyS<float,4>, 151 }; 152 153 PTEX_NAMESPACE_END 154