OpenSubdiv/opensubdiv/osd/clEvaluator.h

231 lines
8.6 KiB
C
Raw Normal View History

Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
//
// 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
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
#include "../version.h"
#include "../osd/opencl.h"
#include "../osd/vertexDescriptor.h"
namespace OpenSubdiv {
namespace OPENSUBDIV_VERSION {
namespace Far {
class StencilTables;
}
namespace Osd {
/// \brief OpenCL stencil tables
///
/// This class is an OpenCL buffer representation of Far::StencilTables.
///
/// CLCompute consumes this table to apply stencils
///
///
class CLStencilTables {
public:
template <typename DEVICE_CONTEXT>
static CLStencilTables *Create(Far::StencilTables const *stencilTables,
DEVICE_CONTEXT context) {
return new CLStencilTables(stencilTables, context->GetContext());
}
CLStencilTables(Far::StencilTables const *stencilTables,
cl_context clContext);
~CLStencilTables();
// 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; }
int GetNumStencils() const { return _numStencils; }
private:
cl_mem _sizes;
cl_mem _offsets;
cl_mem _indices;
cl_mem _weights;
int _numStencils;
};
// ---------------------------------------------------------------------------
/// \brief OpenCL stencil kernel
///
///
class CLEvaluator {
public:
typedef bool Instantiatable;
/// Constructor.
CLEvaluator(cl_context context, cl_command_queue queue);
/// Desctructor.
~CLEvaluator();
/// Generic creator template.
template <typename DEVICE_CONTEXT>
static CLEvaluator *Create(VertexBufferDescriptor const &srcDesc,
VertexBufferDescriptor const &dstDesc,
DEVICE_CONTEXT deviceContext) {
return Create(srcDesc, dstDesc,
deviceContext->GetContext(),
deviceContext->GetCommandQueue());
}
static CLEvaluator * Create(VertexBufferDescriptor const &srcDesc,
VertexBufferDescriptor const &dstDesc,
cl_context clContext,
cl_command_queue clCommandQueue) {
CLEvaluator *kernel = new CLEvaluator(clContext, clCommandQueue);
if (kernel->Compile(srcDesc, dstDesc)) return kernel;
delete kernel;
return NULL;
}
/// \brief Generic static compute function. This function has a same
/// signature as other device kernels have so that it can be called
/// transparently from OsdMesh template interface.
///
/// @param srcBuffer Input primvar buffer.
/// must have BindCLBuffer() method returning a
/// const float pointer for read
///
/// @param srcDesc vertex buffer descriptor for the input buffer
///
/// @param dstBuffer Output primvar buffer
/// must have BindCLBuffer() method returning a
/// float pointer for write
///
/// @param dstDesc vertex buffer descriptor for the output buffer
///
/// @param stencilTables stencil table to be applied. The table must have
/// OpenCL memory interfaces.
///
/// @param instance cached compiled instance. Clients are supposed to
/// pre-compile an instance of this class and provide
/// to this function. If it's null the kernel still
/// compute by instantiating on-demand kernel although
/// it may cause a performance problem.
///
/// @param deviceContext client providing context class which supports
/// cL_context GetContext()
/// cl_command_queue GetCommandQueue()
/// methods.
///
template <typename VERTEX_BUFFER, typename STENCIL_TABLE,
typename DEVICE_CONTEXT>
static bool EvalStencils(VERTEX_BUFFER *srcVertexBuffer,
VertexBufferDescriptor const &srcDesc,
VERTEX_BUFFER *dstVertexBuffer,
VertexBufferDescriptor const &dstDesc,
STENCIL_TABLE const *stencilTable,
CLEvaluator const *instance,
DEVICE_CONTEXT deviceContext) {
if (instance) {
return instance->EvalStencils(srcVertexBuffer, srcDesc,
dstVertexBuffer, dstDesc,
stencilTable);
} else {
// Create an instance on demand (slow)
instance = Create(srcDesc, dstDesc, deviceContext);
if (instance) {
bool r = instance->EvalStencils(srcVertexBuffer, srcDesc,
dstVertexBuffer, dstDesc,
stencilTable);
delete instance;
return r;
}
return false;
}
}
/// Generic compute function.
/// Dispatch the CL compute kernel asynchronously.
/// Returns false if the kernel hasn't been compiled yet.
template <typename VERTEX_BUFFER, typename STENCIL_TABLE>
bool EvalStencils(VERTEX_BUFFER *srcVertexBuffer,
VertexBufferDescriptor const &srcDesc,
VERTEX_BUFFER *dstVertexBuffer,
VertexBufferDescriptor const &dstDesc,
STENCIL_TABLE const *stencilTable) const {
return EvalStencils(srcVertexBuffer->BindCLBuffer(_clCommandQueue),
srcDesc,
dstVertexBuffer->BindCLBuffer(_clCommandQueue),
dstDesc,
stencilTable->GetSizesBuffer(),
stencilTable->GetOffsetsBuffer(),
stencilTable->GetIndicesBuffer(),
stencilTable->GetWeightsBuffer(),
0,
stencilTable->GetNumStencils());
}
/// Dispatch the CL compute kernel asynchronously.
/// returns false if the kernel hasn't been compiled yet.
bool EvalStencils(cl_mem src,
VertexBufferDescriptor const &srcDesc,
cl_mem dst,
VertexBufferDescriptor const &dstDesc,
cl_mem sizes,
cl_mem offsets,
cl_mem indices,
cl_mem weights,
int start,
int end) const;
/// Configure OpenCL kernel.
/// Returns false if it fails to compile the kernel.
bool Compile(VertexBufferDescriptor const &srcDesc,
VertexBufferDescriptor const &dstDesc);
/// Wait the OpenCL kernels finish.
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 _stencilsKernel;
};
} // end namespace Osd
} // end namespace OPENSUBDIV_VERSION
using namespace OPENSUBDIV_VERSION;
} // end namespace OpenSubdiv
#endif // OPENSUBDIV_OPENSUBDIV3_OSD_CL_EVALUATOR_H