OpenSubdiv/opensubdiv/osd/cpuKernel.h

132 lines
3.9 KiB
C
Raw Normal View History

//
// 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
2013-07-18 21:19:50 +00:00
//
// 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_CPU_KERNEL_H
#define OPENSUBDIV3_OSD_CPU_KERNEL_H
#include "../version.h"
#include <cstring>
namespace OpenSubdiv {
namespace OPENSUBDIV_VERSION {
namespace Osd {
2013-08-14 23:41:35 +00:00
struct BufferDescriptor;
void
CpuEvalStencils(float const * src, BufferDescriptor const &srcDesc,
float * dst, BufferDescriptor const &dstDesc,
int const * sizes,
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
int const * offsets,
int const * indices,
float const * weights,
int start, int end);
void
CpuEvalStencils(float const * src, BufferDescriptor const &srcDesc,
float * dst, BufferDescriptor const &dstDesc,
float * dstDu, BufferDescriptor const &dstDuDesc,
float * dstDv, BufferDescriptor const &dstDvDesc,
int const * sizes,
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
int const * offsets,
int const * indices,
float const * weights,
float const * duWeights,
float const * dvWeights,
int start, int end);
2013-08-14 23:41:35 +00:00
//
// SIMD ICC optimization of the stencil kernel
//
2013-08-14 23:41:35 +00:00
#if defined ( __INTEL_COMPILER ) or defined ( __ICC )
#define __ALIGN_DATA __declspec(align(32))
#else
#define __ALIGN_DATA
#endif
2013-08-14 23:41:35 +00:00
// Note : this function is re-used in the TBB Compute kernel
template <int numElems> void
ComputeStencilKernel(float const * vertexSrc,
float * vertexDst,
int const * sizes,
int const * indices,
float const * weights,
int start,
int end) {
__ALIGN_DATA float result[numElems],
result1[numElems];
2013-08-14 23:41:35 +00:00
float const * src;
float * dst, weight;
2013-08-14 23:41:35 +00:00
for (int i=start; i<end; ++i) {
2013-08-14 23:41:35 +00:00
// Clear
#if defined ( __INTEL_COMPILER ) or defined ( __ICC )
#pragma simd
#pragma vector aligned
#endif
for (int k = 0; k<numElems; ++k)
result[k] = 0.0f;
2013-08-14 23:41:35 +00:00
for (int j=0; j<sizes[i]; ++j, ++indices, ++weights) {
2013-08-14 23:41:35 +00:00
src = vertexSrc + (*indices)*numElems;
weight = *weights;
2013-08-14 23:41:35 +00:00
// AddWithWeight
#if defined ( __INTEL_COMPILER ) or defined ( __ICC )
#pragma simd
#pragma vector aligned
#endif
for (int k=0; k<numElems; ++k) {
result[k] += src[k] * weight;
}
2013-08-14 23:41:35 +00:00
}
#if defined ( __INTEL_COMPILER ) or defined ( __ICC )
#pragma simd
#pragma vector aligned
#endif
for (int k=0; k<numElems; ++k) {
result1[k] = result[k];
2013-08-14 23:41:35 +00:00
}
dst = vertexDst + i*numElems;
memcpy(dst, result1, numElems*sizeof(float));
2013-08-14 23:41:35 +00:00
}
}
} // end namespace Osd
} // end namespace OPENSUBDIV_VERSION
using namespace OPENSUBDIV_VERSION;
} // end namespace OpenSubdiv
#endif // OPENSUBDIV3_OSD_CPU_KERNEL_H