clEvaluator.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 OPENSUBDIV_OPENSUBDIV3_OSD_CL_EVALUATOR_H
#define OPENSUBDIV_OPENSUBDIV3_OSD_CL_EVALUATOR_H

#include "../version.h"

#include "../osd/opencl.h"
#include "../osd/types.h"
#include "../osd/bufferDescriptor.h"

namespace OpenSubdiv {
namespace OPENSUBDIV_VERSION {

namespace Far {
    class StencilTable;
}

namespace Osd {

class CLStencilTable {
public:
    template <typename DEVICE_CONTEXT>
    static CLStencilTable *Create(Far::StencilTable const *stencilTable,
                                  DEVICE_CONTEXT context) {
        return new CLStencilTable(stencilTable, context->GetContext());
    }

    template <typename DEVICE_CONTEXT>
    static CLStencilTable *Create(
        Far::LimitStencilTable const *limitStencilTable,
        DEVICE_CONTEXT context) {
        return new CLStencilTable(limitStencilTable, context->GetContext());
    }

    CLStencilTable(Far::StencilTable const *stencilTable,
                   cl_context clContext);
    CLStencilTable(Far::LimitStencilTable const *limitStencilTable,
                   cl_context clContext);
    ~CLStencilTable();

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

private:
    cl_mem _sizes;
    cl_mem _offsets;
    cl_mem _indices;
    cl_mem _weights;
    cl_mem _duWeights;
    cl_mem _dvWeights;
    int _numStencils;
};

// ---------------------------------------------------------------------------

class CLEvaluator {
public:
    typedef bool Instantiatable;
    CLEvaluator(cl_context context, cl_command_queue queue);

    ~CLEvaluator();

    template <typename DEVICE_CONTEXT>
    static CLEvaluator *Create(BufferDescriptor const &srcDesc,
                               BufferDescriptor const &dstDesc,
                               BufferDescriptor const &duDesc,
                               BufferDescriptor const &dvDesc,
                               DEVICE_CONTEXT deviceContext) {
        return Create(srcDesc, dstDesc, duDesc, dvDesc,
                      deviceContext->GetContext(),
                      deviceContext->GetCommandQueue());
    }

    static CLEvaluator * Create(BufferDescriptor const &srcDesc,
                                BufferDescriptor const &dstDesc,
                                BufferDescriptor const &duDesc,
                                BufferDescriptor const &dvDesc,
                                cl_context clContext,
                                cl_command_queue clCommandQueue) {
        CLEvaluator *kernel = new CLEvaluator(clContext, clCommandQueue);
        if (kernel->Compile(srcDesc, dstDesc, duDesc, dvDesc)) return kernel;
        delete kernel;
        return NULL;
    }


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

        if (instance) {
            return instance->EvalStencils(srcBuffer, srcDesc,
                                          dstBuffer, dstDesc,
                                          stencilTable);
        } else {
            // Create an instance on demand (slow)
            instance = Create(srcDesc, dstDesc,
                              BufferDescriptor(),
                              BufferDescriptor(),
                              deviceContext);
            if (instance) {
                bool r = instance->EvalStencils(srcBuffer, srcDesc,
                                                dstBuffer, dstDesc,
                                                stencilTable);
                delete instance;
                return r;
            }
            return false;
        }
    }

    template <typename SRC_BUFFER, typename DST_BUFFER,
              typename STENCIL_TABLE, typename DEVICE_CONTEXT>
    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,
        CLEvaluator const *instance,
        DEVICE_CONTEXT deviceContext) {

        if (instance) {
            return instance->EvalStencils(srcBuffer, srcDesc,
                                          dstBuffer, dstDesc,
                                          duBuffer,  duDesc,
                                          dvBuffer,  dvDesc,
                                          stencilTable);
        } else {
            // Create an instance on demand (slow)
            instance = Create(srcDesc, dstDesc, duDesc, dvDesc,
                              deviceContext);
            if (instance) {
                bool r = instance->EvalStencils(srcBuffer, srcDesc,
                                                dstBuffer, dstDesc,
                                                duBuffer,  duDesc,
                                                dvBuffer,  dvDesc,
                                                stencilTable);
                delete instance;
                return r;
            }
            return false;
        }
    }

    template <typename SRC_BUFFER, typename DST_BUFFER, typename STENCIL_TABLE>
    bool EvalStencils(
        SRC_BUFFER *srcBuffer, BufferDescriptor const &srcDesc,
        DST_BUFFER *dstBuffer, BufferDescriptor const &dstDesc,
        STENCIL_TABLE const *stencilTable) const {
        return EvalStencils(srcBuffer->BindCLBuffer(_clCommandQueue), srcDesc,
                            dstBuffer->BindCLBuffer(_clCommandQueue), dstDesc,
                            stencilTable->GetSizesBuffer(),
                            stencilTable->GetOffsetsBuffer(),
                            stencilTable->GetIndicesBuffer(),
                            stencilTable->GetWeightsBuffer(),
                            0,
                            stencilTable->GetNumStencils());
    }

    template <typename SRC_BUFFER, typename DST_BUFFER, typename STENCIL_TABLE>
    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 {
        return EvalStencils(srcBuffer->BindCLBuffer(_clCommandQueue), srcDesc,
                            dstBuffer->BindCLBuffer(_clCommandQueue), dstDesc,
                            duBuffer->BindCLBuffer(_clCommandQueue), duDesc,
                            dvBuffer->BindCLBuffer(_clCommandQueue), dvDesc,
                            stencilTable->GetSizesBuffer(),
                            stencilTable->GetOffsetsBuffer(),
                            stencilTable->GetIndicesBuffer(),
                            stencilTable->GetWeightsBuffer(),
                            stencilTable->GetDuWeightsBuffer(),
                            stencilTable->GetDvWeightsBuffer(),
                            0,
                            stencilTable->GetNumStencils());
    }

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

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

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

        if (instance) {
            return instance->EvalPatches(srcBuffer, srcDesc,
                                         dstBuffer, dstDesc,
                                         numPatchCoords, patchCoords,
                                         patchTable);
        } else {
            // Create an instance on demand (slow)
            (void)deviceContext;  // unused
            instance = Create(srcDesc, dstDesc,
                              BufferDescriptor(),
                              BufferDescriptor(),
                              deviceContext);
            if (instance) {
                bool r = instance->EvalPatches(srcBuffer, srcDesc,
                                               dstBuffer, dstDesc,
                                               numPatchCoords, patchCoords,
                                               patchTable);
                delete instance;
                return r;
            }
            return false;
        }
    }

    template <typename SRC_BUFFER, typename DST_BUFFER,
              typename PATCHCOORD_BUFFER, typename PATCH_TABLE,
              typename DEVICE_CONTEXT>
    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,
        CLEvaluator const *instance,
        DEVICE_CONTEXT deviceContext) {

        if (instance) {
            return instance->EvalPatches(srcBuffer, srcDesc,
                                         dstBuffer, dstDesc,
                                         duBuffer, duDesc,
                                         dvBuffer, dvDesc,
                                         numPatchCoords, patchCoords,
                                         patchTable);
        } else {
            // Create an instance on demand (slow)
            (void)deviceContext;  // unused
            instance = Create(srcDesc, dstDesc, duDesc, dvDesc, deviceContext);
            if (instance) {
                bool r = instance->EvalPatches(srcBuffer, srcDesc,
                                               dstBuffer, dstDesc,
                                               duBuffer, duDesc,
                                               dvBuffer, dvDesc,
                                               numPatchCoords, patchCoords,
                                               patchTable);
                delete instance;
                return r;
            }
            return false;
        }
    }

    template <typename SRC_BUFFER, typename DST_BUFFER,
              typename PATCHCOORD_BUFFER, typename PATCH_TABLE>
    bool EvalPatches(
        SRC_BUFFER *srcBuffer, BufferDescriptor const &srcDesc,
        DST_BUFFER *dstBuffer, BufferDescriptor const &dstDesc,
        int numPatchCoords,
        PATCHCOORD_BUFFER *patchCoords,
        PATCH_TABLE *patchTable) const {

        return EvalPatches(srcBuffer->BindCLBuffer(_clCommandQueue), srcDesc,
                           dstBuffer->BindCLBuffer(_clCommandQueue), dstDesc,
                           0, BufferDescriptor(),
                           0, BufferDescriptor(),
                           numPatchCoords,
                           patchCoords->BindCLBuffer(_clCommandQueue),
                           patchTable->GetPatchArrayBuffer(),
                           patchTable->GetPatchIndexBuffer(),
                           patchTable->GetPatchParamBuffer());
    }

    template <typename SRC_BUFFER, typename DST_BUFFER,
              typename PATCHCOORD_BUFFER, typename PATCH_TABLE>
    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) const {

        return EvalPatches(srcBuffer->BindCLBuffer(_clCommandQueue), srcDesc,
                           dstBuffer->BindCLBuffer(_clCommandQueue), dstDesc,
                           duBuffer->BindCLBuffer(_clCommandQueue),  duDesc,
                           dvBuffer->BindCLBuffer(_clCommandQueue),  dvDesc,
                           numPatchCoords,
                           patchCoords->BindCLBuffer(_clCommandQueue),
                           patchTable->GetPatchArrayBuffer(),
                           patchTable->GetPatchIndexBuffer(),
                           patchTable->GetPatchParamBuffer());
    }

    bool EvalPatches(cl_mem src, BufferDescriptor const &srcDesc,
                     cl_mem dst, BufferDescriptor const &dstDesc,
                     cl_mem du,  BufferDescriptor const &duDesc,
                     cl_mem dv,  BufferDescriptor const &dvDesc,
                     int numPatchCoords,
                     cl_mem patchCoordsBuffer,
                     cl_mem patchArrayBuffer,
                     cl_mem patchIndexBuffer,
                     cl_mem patchParamsBuffer) const;


    bool Compile(BufferDescriptor const &srcDesc,
                 BufferDescriptor const &dstDesc,
                 BufferDescriptor const &duDesc,
                 BufferDescriptor const &dvDesc);

    template <typename DEVICE_CONTEXT>
    static void Synchronize(DEVICE_CONTEXT deviceContext) {
        Synchronize(deviceContext->GetCommandQueue());
    }

    static void Synchronize(cl_command_queue queue);

private:
    cl_context _clContext;
    cl_command_queue _clCommandQueue;
    cl_program _program;
    cl_kernel _stencilKernel;
    cl_kernel _stencilDerivKernel;
    cl_kernel _patchKernel;
};


}  // end namespace Osd

}  // end namespace OPENSUBDIV_VERSION
using namespace OPENSUBDIV_VERSION;

}  // end namespace OpenSubdiv


#endif  // OPENSUBDIV_OPENSUBDIV3_OSD_CL_EVALUATOR_H