cudaEvaluator.h

Go to the documentation of this file.

//
//   Copyright 2015 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.
//

#ifndef OPENSUBDIV3_OSD_CUDA_EVALUATOR_H
#define OPENSUBDIV3_OSD_CUDA_EVALUATOR_H

#include "../version.h"

#include <vector>
#include "../osd/bufferDescriptor.h"
#include "../osd/types.h"

namespace OpenSubdiv {
namespace OPENSUBDIV_VERSION {

namespace Far {
    class PatchTable;
    class StencilTable;
    class LimitStencilTable;
}

namespace Osd {

class CudaStencilTable {
public:
    static CudaStencilTable *Create(Far::StencilTable const *stencilTable,
                                    void *deviceContext = NULL) {
        (void)deviceContext;  // unused
        return new CudaStencilTable(stencilTable);
    }
    static CudaStencilTable *Create(Far::LimitStencilTable const *limitStencilTable,
                                    void *deviceContext = NULL) {
        (void)deviceContext;  // unused
        return new CudaStencilTable(limitStencilTable);
    }

    explicit CudaStencilTable(Far::StencilTable const *stencilTable);
    explicit CudaStencilTable(Far::LimitStencilTable const *limitStencilTable);
    ~CudaStencilTable();

    // interfaces needed for CudaCompute
    void *GetSizesBuffer() const { return _sizes; }
    void *GetOffsetsBuffer() const { return _offsets; }
    void *GetIndicesBuffer() const { return _indices; }
    void *GetWeightsBuffer() const { return _weights; }
    void *GetDuWeightsBuffer() const { return _duWeights; }
    void *GetDvWeightsBuffer() const { return _dvWeights; }
    int GetNumStencils() const { return _numStencils; }

private:
    void * _sizes,
         * _offsets,
         * _indices,
         * _weights,
         * _duWeights,
         * _dvWeights;
    int _numStencils;
};

class CudaEvaluator {
public:

    template <typename SRC_BUFFER, typename DST_BUFFER, typename STENCIL_TABLE>
    static bool EvalStencils(
        SRC_BUFFER *srcBuffer, BufferDescriptor const &srcDesc,
        DST_BUFFER *dstBuffer, BufferDescriptor const &dstDesc,
        STENCIL_TABLE const *stencilTable,
        const void *instance = NULL,
        void * deviceContext = NULL) {

        (void)instance;  // unused
        (void)deviceContext;  // unused
        return EvalStencils(srcBuffer->BindCudaBuffer(), srcDesc,
                            dstBuffer->BindCudaBuffer(), dstDesc,
                            (int const *)stencilTable->GetSizesBuffer(),
                            (int const *)stencilTable->GetOffsetsBuffer(),
                            (int const *)stencilTable->GetIndicesBuffer(),
                            (float const *)stencilTable->GetWeightsBuffer(),
                            /*start = */ 0,
                            /*end   = */ stencilTable->GetNumStencils());
    }

    static bool EvalStencils(
        const float *src, BufferDescriptor const &srcDesc,
        float *dst,       BufferDescriptor const &dstDesc,
        const int * sizes,
        const int * offsets,
        const int * indices,
        const float * weights,
        int start, int end);

    template <typename SRC_BUFFER, typename DST_BUFFER, typename STENCIL_TABLE>
    static bool EvalStencils(
        SRC_BUFFER *srcBuffer, BufferDescriptor const &srcDesc,
        DST_BUFFER *dstBuffer, BufferDescriptor const &dstDesc,
        DST_BUFFER *duBuffer,  BufferDescriptor const &duDesc,
        DST_BUFFER *dvBuffer,  BufferDescriptor const &dvDesc,
        STENCIL_TABLE const *stencilTable,
        const CudaEvaluator *instance = NULL,
        void * deviceContext = NULL) {

        (void)instance;   // unused
        (void)deviceContext;   // unused

        return EvalStencils(srcBuffer->BindCudaBuffer(), srcDesc,
                            dstBuffer->BindCudaBuffer(), dstDesc,
                            duBuffer->BindCudaBuffer(),  duDesc,
                            dvBuffer->BindCudaBuffer(),  dvDesc,
                            (int const *)stencilTable->GetSizesBuffer(),
                            (int const *)stencilTable->GetOffsetsBuffer(),
                            (int const *)stencilTable->GetIndicesBuffer(),
                            (float const *)stencilTable->GetWeightsBuffer(),
                            (float const *)stencilTable->GetDuWeightsBuffer(),
                            (float const *)stencilTable->GetDvWeightsBuffer(),
                            /*start = */ 0,
                            /*end   = */ stencilTable->GetNumStencils());
    }

    static bool EvalStencils(
        const float *src, BufferDescriptor const &srcDesc,
        float *dst,       BufferDescriptor const &dstDesc,
        float *du,        BufferDescriptor const &duDesc,
        float *dv,        BufferDescriptor const &dvDesc,
        const int * sizes,
        const int * offsets,
        const int * indices,
        const float * weights,
        const float * duWeights,
        const float * dvWeights,
        int start, int end);


    template <typename SRC_BUFFER, typename DST_BUFFER,
              typename PATCHCOORD_BUFFER, typename PATCH_TABLE>
    static bool EvalPatches(
        SRC_BUFFER *srcBuffer, BufferDescriptor const &srcDesc,
        DST_BUFFER *dstBuffer, BufferDescriptor const &dstDesc,
        int numPatchCoords,
        PATCHCOORD_BUFFER *patchCoords,
        PATCH_TABLE *patchTable,
        CudaEvaluator const *instance,
        void * deviceContext = NULL) {

        (void)instance;   // unused
        (void)deviceContext;   // unused

        return EvalPatches(srcBuffer->BindCudaBuffer(), srcDesc,
                           dstBuffer->BindCudaBuffer(), dstDesc,
                           numPatchCoords,
                           (const PatchCoord *)patchCoords->BindCudaBuffer(),
                           (const PatchArray *)patchTable->GetPatchArrayBuffer(),
                           (const int *)patchTable->GetPatchIndexBuffer(),
                           (const PatchParam *)patchTable->GetPatchParamBuffer());
    }

    template <typename SRC_BUFFER, typename DST_BUFFER,
              typename PATCHCOORD_BUFFER, typename PATCH_TABLE>
    static bool EvalPatches(
        SRC_BUFFER *srcBuffer, BufferDescriptor const &srcDesc,
        DST_BUFFER *dstBuffer, BufferDescriptor const &dstDesc,
        DST_BUFFER *duBuffer,  BufferDescriptor const &duDesc,
        DST_BUFFER *dvBuffer,  BufferDescriptor const &dvDesc,
        int numPatchCoords,
        PATCHCOORD_BUFFER *patchCoords,
        PATCH_TABLE *patchTable,
        CudaEvaluator const *instance,
        void * deviceContext = NULL) {

        (void)instance;   // unused
        (void)deviceContext;   // unused

        return EvalPatches(srcBuffer->BindCudaBuffer(), srcDesc,
                           dstBuffer->BindCudaBuffer(), dstDesc,
                           duBuffer->BindCudaBuffer(),  duDesc,
                           dvBuffer->BindCudaBuffer(),  dvDesc,
                           numPatchCoords,
                           (const PatchCoord *)patchCoords->BindCudaBuffer(),
                           (const PatchArray *)patchTable->GetPatchArrayBuffer(),
                           (const int *)patchTable->GetPatchIndexBuffer(),
                           (const PatchParam *)patchTable->GetPatchParamBuffer());
    }

    static bool EvalPatches(
        const float *src, BufferDescriptor const &srcDesc,
        float *dst,       BufferDescriptor const &dstDesc,
        int numPatchCoords,
        const PatchCoord *patchCoords,
        const PatchArray *patchArrays,
        const int *patchIndices,
        const PatchParam *patchParams);

    static bool EvalPatches(
        const float *src, BufferDescriptor const &srcDesc,
        float *dst,       BufferDescriptor const &dstDesc,
        float *du,        BufferDescriptor const &duDesc,
        float *dv,        BufferDescriptor const &dvDesc,
        int numPatchCoords,
        const PatchCoord *patchCoords,
        const PatchArray *patchArrays,
        const int *patchIndices,
        const PatchParam *patchParams);

    static void Synchronize(void *deviceContext = NULL);
};


}  // end namespace Osd

}  // end namespace OPENSUBDIV_VERSION
using namespace OPENSUBDIV_VERSION;

}  // end namespace OpenSubdiv


#endif  // OPENSUBDIV3_OSD_CUDA_EVALUATOR_H