OpenSubdiv/opensubdiv/osd/ompKernel.cpp

//
//   Copyright 2013 Pixar
//
//   Licensed under the Apache License, Version 2.0 (the "Apache License")
//   with the following modification; you may not use this file except in
//   compliance with the Apache License and the following modification to it:
//   Section 6. Trademarks. is deleted and replaced with:
//
//   6. Trademarks. This License does not grant permission to use the trade
//      names, trademarks, service marks, or product names of the Licensor
//      and its affiliates, except as required to comply with Section 4(c) of
//      the License and to reproduce the content of the NOTICE file.
//
//   You may obtain a copy of the Apache License at
//
//       http://www.apache.org/licenses/LICENSE-2.0
//
//   Unless required by applicable law or agreed to in writing, software
//   distributed under the Apache License with the above modification is
//   distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
//   KIND, either express or implied. See the Apache License for the specific
//   language governing permissions and limitations under the Apache License.
//

#include "../osd/ompKernel.h"
#include "../osd/vertexDescriptor.h"

#include <algorithm>
#include <cmath>
#include <cstdlib>
#include <omp.h>

namespace OpenSubdiv {
namespace OPENSUBDIV_VERSION {

static inline void
clear(float *dst, OsdVertexBufferDescriptor const &desc) {

    if (dst) {
        memset(dst, 0, desc.length*sizeof(float));
    }
}

static inline void
addWithWeight(float *dst, const float *srcOrigin, int srcIndex, float weight,
              OsdVertexBufferDescriptor const &desc) {

    if (srcOrigin && dst) {
        const float *src = srcOrigin + srcIndex * desc.stride;
        for (int k = 0; k < desc.length; ++k) {
            dst[k] += src[k] * weight;
        }
    }
}

static inline void
copy(float *dstOrigin, const float *src, int dstIndex,
     OsdVertexBufferDescriptor const &desc) {

    if (dstOrigin && src) {
        float *dst = dstOrigin + dstIndex * desc.stride;
        memcpy(dst, src, desc.length*sizeof(float));
    }
}

void OsdOmpComputeFace(
    float * vertex, float * varying,
    OsdVertexBufferDescriptor const &vertexDesc,
    OsdVertexBufferDescriptor const &varyingDesc,
    const int *F_IT, const int *F_ITa, int offset, int tableOffset, int start, int end) {

    int numThreads = omp_get_max_threads();
    float *vertexResultsArray = (float*)alloca(vertexDesc.length * sizeof(float) * numThreads);
    float *varyingResultsArray = (float*)alloca(varyingDesc.length * sizeof(float) * numThreads);

#pragma omp parallel for
    for (int i = start + tableOffset; i < end + tableOffset; i++) {
        int h = F_ITa[2*i];
        int n = F_ITa[2*i+1];

        float weight = 1.0f/n;
        int dstIndex = offset + i - tableOffset;

        int threadId = omp_get_thread_num();
        float *vertexResults = vertexResultsArray +
            vertexDesc.length * threadId;
        float *varyingResults = varyingResultsArray +
            varyingDesc.length * threadId;

        // clear
        clear(vertexResults, vertexDesc);
        clear(varyingResults, varyingDesc);

        for (int j = 0; j < n; ++j) {
            int index = F_IT[h+j];
            addWithWeight(vertexResults, vertex, index, weight, vertexDesc);
            addWithWeight(varyingResults, varying, index, weight, varyingDesc);
        }

        // write results
        copy(vertex, vertexResults, dstIndex, vertexDesc);
        copy(varying, varyingResults, dstIndex, varyingDesc);
    }
}

void OsdOmpComputeEdge(
    float * vertex, float * varying,
    OsdVertexBufferDescriptor const &vertexDesc,
    OsdVertexBufferDescriptor const &varyingDesc,
    const int *E_IT, const float *E_W, int offset, int tableOffset, int start, int end) {

    int numThreads = omp_get_max_threads();
    float *vertexResultsArray = (float*)alloca(vertexDesc.length * sizeof(float) * numThreads);
    float *varyingResultsArray = (float*)alloca(varyingDesc.length * sizeof(float) * numThreads);

#pragma omp parallel for
    for (int i = start + tableOffset; i < end + tableOffset; i++) {
        int eidx0 = E_IT[4*i+0];
        int eidx1 = E_IT[4*i+1];
        int eidx2 = E_IT[4*i+2];
        int eidx3 = E_IT[4*i+3];

        float vertWeight = E_W[i*2+0];
        int dstIndex = offset + i - tableOffset;

        int threadId = omp_get_thread_num();
        float *vertexResults = vertexResultsArray +
            vertexDesc.length * threadId;
        float *varyingResults = varyingResultsArray +
            varyingDesc.length * threadId;

        // clear
        clear(vertexResults, vertexDesc);
        clear(varyingResults, varyingDesc);

        addWithWeight(vertexResults, vertex, eidx0, vertWeight, vertexDesc);
        addWithWeight(vertexResults, vertex, eidx1, vertWeight, vertexDesc);

        if (eidx2 != -1) {
            float faceWeight = E_W[i*2+1];

            addWithWeight(vertexResults, vertex, eidx2, faceWeight, vertexDesc);
            addWithWeight(vertexResults, vertex, eidx3, faceWeight, vertexDesc);
        }

        addWithWeight(varyingResults, varying, eidx0, 0.5f, varyingDesc);
        addWithWeight(varyingResults, varying, eidx1, 0.5f, varyingDesc);

        copy(vertex, vertexResults, dstIndex, vertexDesc);
        copy(varying, varyingResults, dstIndex, varyingDesc);
    }
}

void OsdOmpComputeVertexA(
    float * vertex, float * varying,
    OsdVertexBufferDescriptor const &vertexDesc,
    OsdVertexBufferDescriptor const &varyingDesc,
    const int *V_ITa, const float *V_W,
    int offset, int tableOffset, int start, int end, int pass) {

    int numThreads = omp_get_max_threads();
    float *vertexResultsArray = (float*)alloca(vertexDesc.length * sizeof(float) * numThreads);
    float *varyingResultsArray = (float*)alloca(varyingDesc.length * sizeof(float) * numThreads);

#pragma omp parallel for
    for (int i = start + tableOffset; i < end + tableOffset; i++) {
        int n     = V_ITa[5*i+1];
        int p     = V_ITa[5*i+2];
        int eidx0 = V_ITa[5*i+3];
        int eidx1 = V_ITa[5*i+4];

        float weight = (pass == 1) ? V_W[i] : 1.0f - V_W[i];

        // In the case of fractional weight, the weight must be inverted since
        // the value is shared with the k_Smooth kernel (statistically the
        // k_Smooth kernel runs much more often than this one)
        if (weight > 0.0f && weight < 1.0f && n > 0)
            weight = 1.0f - weight;

        int dstIndex = offset + i - tableOffset;

        int threadId = omp_get_thread_num();
        float *vertexResults = vertexResultsArray +
            vertexDesc.length * threadId;
        float *varyingResults = varyingResultsArray +
            varyingDesc.length * threadId;

        clear(vertexResults, vertexDesc);
        clear(varyingResults, varyingDesc);
        if (pass) {
            // copy previous results
            addWithWeight(vertexResults, vertex, dstIndex, 1.0f, vertexDesc);
        }

        if (eidx0 == -1 || (pass == 0 && (n == -1))) {
            addWithWeight(vertexResults, vertex, p, weight, vertexDesc);
        } else {
            addWithWeight(vertexResults, vertex, p, weight * 0.75f, vertexDesc);
            addWithWeight(vertexResults, vertex, eidx0, weight * 0.125f, vertexDesc);
            addWithWeight(vertexResults, vertex, eidx1, weight * 0.125f, vertexDesc);
        }

        copy(vertex, vertexResults, dstIndex, vertexDesc);
        if (not pass) {
            addWithWeight(varyingResults, varying, p, 1.0f, varyingDesc);
            copy(varying, varyingResults, dstIndex, varyingDesc);
        }
    }
}

void OsdOmpComputeVertexB(
    float * vertex, float * varying,
    OsdVertexBufferDescriptor const &vertexDesc,
    OsdVertexBufferDescriptor const &varyingDesc,
    const int *V_ITa, const int *V_IT, const float *V_W,
    int offset, int tableOffset, int start, int end) {

    int numThreads = omp_get_max_threads();
    float *vertexResultsArray = (float*)alloca(vertexDesc.length * sizeof(float) * numThreads);
    float *varyingResultsArray = (float*)alloca(varyingDesc.length * sizeof(float) * numThreads);

#pragma omp parallel for
    for (int i = start + tableOffset; i < end + tableOffset; i++) {
        int h = V_ITa[5*i];
        int n = V_ITa[5*i+1];
        int p = V_ITa[5*i+2];

        float weight = V_W[i];
        float wp = 1.0f/static_cast<float>(n*n);
        float wv = (n-2.0f) * n * wp;

        int dstIndex = offset + i - tableOffset;

        int threadId = omp_get_thread_num();
        float *vertexResults = vertexResultsArray +
            vertexDesc.length * threadId;
        float *varyingResults = varyingResultsArray +
            varyingDesc.length * threadId;

        clear(vertexResults, vertexDesc);
        clear(varyingResults, varyingDesc);

        addWithWeight(vertexResults, vertex, p, weight * wv, vertexDesc);

        for (int j = 0; j < n; ++j) {
            addWithWeight(vertexResults, vertex, V_IT[h+j*2], weight * wp, vertexDesc);
            addWithWeight(vertexResults, vertex, V_IT[h+j*2+1], weight * wp, vertexDesc);
        }
        addWithWeight(varyingResults, varying, p, 1.0f, varyingDesc);

        copy(vertex, vertexResults, dstIndex, vertexDesc);
        copy(varying, varyingResults, dstIndex, varyingDesc);
    }
}

void OsdOmpComputeLoopVertexB(
    float * vertex, float * varying,
    OsdVertexBufferDescriptor const &vertexDesc,
    OsdVertexBufferDescriptor const &varyingDesc,
    const int *V_ITa, const int *V_IT, const float *V_W,
    int vertexOffset, int tableOffset, int start, int end) {

    int numThreads = omp_get_max_threads();
    float *vertexResultsArray = (float*)alloca(vertexDesc.length * sizeof(float) * numThreads);
    float *varyingResultsArray = (float*)alloca(varyingDesc.length * sizeof(float) * numThreads);

#pragma omp parallel for
    for (int i = start + tableOffset; i < end + tableOffset; i++) {
        int h = V_ITa[5*i];
        int n = V_ITa[5*i+1];
        int p = V_ITa[5*i+2];

        float weight = V_W[i];
        float wp = 1.0f/static_cast<float>(n);
        float beta = 0.25f * cosf(static_cast<float>(M_PI) * 2.0f * wp) + 0.375f;
        beta = beta * beta;
        beta = (0.625f - beta) * wp;

        int dstIndex = i + vertexOffset - tableOffset;

        int threadId = omp_get_thread_num();
        float *vertexResults = vertexResultsArray +
            vertexDesc.length * threadId;
        float *varyingResults = varyingResultsArray +
            varyingDesc.length * threadId;

        clear(vertexResults, vertexDesc);
        clear(varyingResults, varyingDesc);

        addWithWeight(vertexResults, vertex, p, weight * (1.0f - (beta * n)), vertexDesc);

        for (int j = 0; j < n; ++j)
            addWithWeight(vertexResults, vertex, V_IT[h+j], weight * beta, vertexDesc);

        addWithWeight(varyingResults, varying, p, 1.0f, varyingDesc);

        copy(vertex, vertexResults, dstIndex, vertexDesc);
        copy(varying, varyingResults, dstIndex, varyingDesc);
    }
}

void OsdOmpComputeBilinearEdge(
    float * vertex, float * varying,
    OsdVertexBufferDescriptor const &vertexDesc,
    OsdVertexBufferDescriptor const &varyingDesc,
    const int *E_IT, int vertexOffset, int tableOffset, int start, int end) {

    int numThreads = omp_get_max_threads();
    float *vertexResultsArray = (float*)alloca(vertexDesc.length * sizeof(float) * numThreads);
    float *varyingResultsArray = (float*)alloca(varyingDesc.length * sizeof(float) * numThreads);

#pragma omp parallel for
    for (int i = start + tableOffset; i < end + tableOffset; i++) {
        int eidx0 = E_IT[2*i+0];
        int eidx1 = E_IT[2*i+1];

        int dstIndex = i + vertexOffset - tableOffset;

        int threadId = omp_get_thread_num();
        float *vertexResults = vertexResultsArray +
            vertexDesc.length * threadId;
        float *varyingResults = varyingResultsArray +
            varyingDesc.length * threadId;

        clear(vertexResults, vertexDesc);
        clear(varyingResults, varyingDesc);

        addWithWeight(vertexResults, vertex, eidx0, 0.5f, vertexDesc);
        addWithWeight(vertexResults, vertex, eidx1, 0.5f, vertexDesc);

        addWithWeight(varyingResults, varying, eidx0, 0.5f, varyingDesc);
        addWithWeight(varyingResults, varying, eidx1, 0.5f, varyingDesc);

        copy(vertex, vertexResults, dstIndex, vertexDesc);
        copy(varying, varyingResults, dstIndex, varyingDesc);
    }
}

void OsdOmpComputeBilinearVertex(
    float * vertex, float * varying,
    OsdVertexBufferDescriptor const &vertexDesc,
    OsdVertexBufferDescriptor const &varyingDesc,
    const int *V_ITa, int vertexOffset, int tableOffset, int start, int end) {

    int numThreads = omp_get_max_threads();
    float *vertexResultsArray = (float*)alloca(vertexDesc.length * sizeof(float) * numThreads);
    float *varyingResultsArray = (float*)alloca(varyingDesc.length * sizeof(float) * numThreads);

#pragma omp parallel for
    for (int i = start + tableOffset; i < end + tableOffset; i++) {
        int p = V_ITa[i];

        int dstIndex = i + vertexOffset - tableOffset;

        int threadId = omp_get_thread_num();
        float *vertexResults = vertexResultsArray +
            vertexDesc.length * threadId;
        float *varyingResults = varyingResultsArray +
            varyingDesc.length * threadId;

        clear(vertexResults, vertexDesc);
        clear(varyingResults, varyingDesc);

        addWithWeight(vertexResults, vertex, p, 1.0f, vertexDesc);
        addWithWeight(varyingResults, varying, p, 1.0f, varyingDesc);

        copy(vertex, vertexResults, dstIndex, vertexDesc);
        copy(varying, varyingResults, dstIndex, varyingDesc);
    }
}

void OsdOmpEditVertexAdd(
    float * vertex,
    OsdVertexBufferDescriptor const &vertexDesc,
    int primVarOffset, int primVarWidth, int vertexOffset, int tableOffset,
    int start, int end,
    const unsigned int *editIndices, const float *editValues) {

#pragma omp parallel for
    for (int i = start+tableOffset; i < end+tableOffset; i++) {

        if (vertex) {
            int editIndex = editIndices[i] + vertexOffset;
            float *dst = vertex + editIndex * vertexDesc.stride + primVarOffset;

            for (int j = 0; j < primVarWidth; ++j) {
                dst[j] += editValues[j];
            }
        }
    }
}

void OsdOmpEditVertexSet(
    float * vertex,
    OsdVertexBufferDescriptor const &vertexDesc,
    int primVarOffset, int primVarWidth, int vertexOffset, int tableOffset,
    int start, int end,
    const unsigned int *editIndices, const float *editValues) {

#pragma omp parallel for
    for (int i = start+tableOffset; i < end+tableOffset; i++) {

        if (vertex) {
            int editIndex = editIndices[i] + vertexOffset;
            float *dst = vertex + editIndex * vertexDesc.stride + primVarOffset;

            for (int j = 0; j < primVarWidth; ++j) {
                dst[j] = editValues[j];
            }
        }
    }
}

}  // end namespace OPENSUBDIV_VERSION
}  // end namespace OpenSubdiv