mirror of
https://github.com/ceres-solver/ceres-solver.git
synced 2026-08-29 08:34:37 +08:00
Dmitriy Korchemkin
dc7a859752
As pointed out by several users, introduction of parallel operations on vectors severely impacts solver performance on small problems, with time consumption increasing with the number of threads. In order to minimize overhead of trying to execute small tasks using a large number of threads, task scheduling mechanism was changed to avoid scheduling all tasks at once. However, there is still a large difference in exectuion time because the main thread always launches the next thread before starting doing the work. This leads to several orders of magnitude slowdown when going from a single-threaded execution (which follows a fast-forward path to a single loop over all indices, without any synchronization involved) to a two-thread execution: /bin/parallel_vector_operations_benchmark ------------------------------------------- Benchmark Time ------------------------------------------- SetZero/128 12.8 ns SetZeroParallel/128/1 16.6 ns SetZeroParallel/128/2 2211 ns In order to eliminate this effect, we limit the block-size of parallel execution of vector operations to 2^16 elements (thus, starting parallel execution only for vectors of at least 2^17 elements). Threshold of 2^16 elements was choosen by evaluating thresholds from 2^10 to 2^20 (only powers of 2), with 2^14..2^20 significantly reducing worst-case runtime degradation. Details can be found in discussion of the issue at https://github.com/ceres-solver/ceres-solver/issues/1016 Change-Id: I555c882d63ee53323ceb426743b970f989b65503
91 lines
3.8 KiB
C++
91 lines
3.8 KiB
C++
// Ceres Solver - A fast non-linear least squares minimizer
|
|
// Copyright 2023 Google Inc. All rights reserved.
|
|
// http://ceres-solver.org/
|
|
//
|
|
// Redistribution and use in source and binary forms, with or without
|
|
// modification, are permitted provided that the following conditions are met:
|
|
//
|
|
// * Redistributions of source code must retain the above copyright notice,
|
|
// this list of conditions and the following disclaimer.
|
|
// * Redistributions in binary form must reproduce the above copyright notice,
|
|
// this list of conditions and the following disclaimer in the documentation
|
|
// and/or other materials provided with the distribution.
|
|
// * Neither the name of Google Inc. nor the names of its contributors may be
|
|
// used to endorse or promote products derived from this software without
|
|
// specific prior written permission.
|
|
//
|
|
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
|
|
// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
// ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
|
|
// LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
|
|
// CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
|
|
// SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
|
|
// INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
|
|
// CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
|
|
// ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
// POSSIBILITY OF SUCH DAMAGE.
|
|
//
|
|
// Authors: vitus@google.com (Michael Vitus),
|
|
// dmitriy.korchemkin@gmail.com (Dmitriy Korchemkin)
|
|
|
|
#ifndef CERES_INTERNAL_PARALLEL_VECTOR_OPS_H_
|
|
#define CERES_INTERNAL_PARALLEL_VECTOR_OPS_H_
|
|
|
|
#include <mutex>
|
|
#include <vector>
|
|
|
|
#include "ceres/context_impl.h"
|
|
#include "ceres/internal/eigen.h"
|
|
#include "ceres/internal/export.h"
|
|
#include "ceres/parallel_for.h"
|
|
|
|
namespace ceres::internal {
|
|
|
|
// Lower bound on block size for parallel vector operations.
|
|
// Operations with vectors of less than kMinBlockSizeParallelVectorOps elements
|
|
// will be executed in a single thread.
|
|
constexpr int kMinBlockSizeParallelVectorOps = 1 << 16;
|
|
// Evaluate vector expression in parallel
|
|
// Assuming LhsExpression and RhsExpression are some sort of column-vector
|
|
// expression, assignment lhs = rhs is eavluated over a set of contiguous blocks
|
|
// in parallel. This is expected to work well in the case of vector-based
|
|
// expressions (since they typically do not result into temporaries). This
|
|
// method expects lhs to be size-compatible with rhs
|
|
template <typename LhsExpression, typename RhsExpression>
|
|
void ParallelAssign(ContextImpl* context,
|
|
int num_threads,
|
|
LhsExpression& lhs,
|
|
const RhsExpression& rhs) {
|
|
static_assert(LhsExpression::ColsAtCompileTime == 1);
|
|
static_assert(RhsExpression::ColsAtCompileTime == 1);
|
|
CHECK_EQ(lhs.rows(), rhs.rows());
|
|
const int num_rows = lhs.rows();
|
|
ParallelFor(
|
|
context,
|
|
0,
|
|
num_rows,
|
|
num_threads,
|
|
[&lhs, &rhs](const std::tuple<int, int>& range) {
|
|
auto [start, end] = range;
|
|
lhs.segment(start, end - start) = rhs.segment(start, end - start);
|
|
},
|
|
kMinBlockSizeParallelVectorOps);
|
|
}
|
|
|
|
// Set vector to zero using num_threads
|
|
template <typename VectorType>
|
|
void ParallelSetZero(ContextImpl* context,
|
|
int num_threads,
|
|
VectorType& vector) {
|
|
ParallelSetZero(context, num_threads, vector.data(), vector.rows());
|
|
}
|
|
void ParallelSetZero(ContextImpl* context,
|
|
int num_threads,
|
|
double* values,
|
|
int num_values);
|
|
|
|
} // namespace ceres::internal
|
|
|
|
#endif // CERES_INTERNAL_PARALLEL_FOR_H_
|