mirror of
https://github.com/slsdetectorgroup/aare.git
synced 2026-10-13 09:10:25 +02:00
Optimized pedestal tracking with FastPedestal and performance improvements to the cluster finder. - ~2x faster hitting 14k FPS on benchmark dataset on 16 core threadripper pro **Changes** - No bounds check on interior of frame. (only within half of a cluster) (~15% improvement) - Cache threshold (std*n_rms) Biggest improvement comes from not computing std for each access - Cache pedestal subtracted frame - Switched to FastPedestal which assumes we reached steady state **Things tried but rejected:** - Always store cluster values, commit on val==max (~15% drop in frame rate) **Options** - using 16 bit clusters and 16 bit pedestal. (slows down the single threaded case but allow us to reach 12k with multi threading) **Other notes on performance** - Passing in memory frames from python and doing nothing: 0.8us/frame - Passing + pedestal subtraction: 28us/frame - Passing + pedestal + cluster finding (no store): 652us/frame - Passing + pedestal + cluster finding: 886 us/frame
231 lines
7.5 KiB
C++
231 lines
7.5 KiB
C++
// SPDX-License-Identifier: MPL-2.0
|
|
#pragma once
|
|
#include "aare/Cluster.hpp" //TODO maybe store in seperate file !!!
|
|
#include <algorithm>
|
|
#include <array>
|
|
#include <cstddef>
|
|
#include <cstdint>
|
|
#include <numeric>
|
|
#include <vector>
|
|
|
|
#include <fmt/core.h>
|
|
|
|
#include "aare/Cluster.hpp"
|
|
#include "aare/NDView.hpp"
|
|
|
|
namespace aare {
|
|
|
|
template <typename ClusterType,
|
|
typename = std::enable_if_t<is_cluster_v<ClusterType>>>
|
|
class ClusterVector; // Forward declaration
|
|
|
|
/**
|
|
* @brief ClusterVector is a container for clusters of various sizes. It
|
|
* uses a contiguous memory buffer to store the clusters. It is templated on
|
|
* the data type and the coordinate type of the clusters.
|
|
* @note push_back can invalidate pointers to elements in the container
|
|
* @warning ClusterVector is currently move only to catch unintended copies,
|
|
* but this might change since there are probably use cases where copying is
|
|
* needed.
|
|
* @tparam T data type of the pixels in the cluster
|
|
* @tparam CoordType data type of the x and y coordinates of the cluster
|
|
* (normally uint16_t)
|
|
*/
|
|
template <typename T, uint8_t ClusterSizeX, uint8_t ClusterSizeY,
|
|
typename CoordType>
|
|
class ClusterVector<Cluster<T, ClusterSizeX, ClusterSizeY, CoordType>> {
|
|
|
|
std::vector<Cluster<T, ClusterSizeX, ClusterSizeY, CoordType>> m_data{};
|
|
int32_t m_frame_number{0}; // TODO! Check frame number size and type
|
|
|
|
public:
|
|
using value_type = T;
|
|
using ClusterType = Cluster<T, ClusterSizeX, ClusterSizeY, CoordType>;
|
|
|
|
/**
|
|
* @brief Construct a new ClusterVector object
|
|
* @param capacity initial capacity of the buffer in number of clusters
|
|
* @param frame_number frame number of the clusters. Default is 0, which is
|
|
* also used to indicate that the clusters come from many frames
|
|
*/
|
|
ClusterVector(size_t capacity = 1024, uint64_t frame_number = 0)
|
|
: m_frame_number(frame_number) {
|
|
m_data.reserve(capacity);
|
|
}
|
|
|
|
ClusterVector(ClusterVector &&other) noexcept = default;
|
|
ClusterVector &operator=(ClusterVector &&other) noexcept = default;
|
|
|
|
/**
|
|
* @brief Create a copy of the clustervector by filtering clusters in the
|
|
* ClusterVector using a boolean mask.
|
|
* @param mask boolean 1d mask
|
|
* @return ClusterVector containing only the clusters where the mask is
|
|
* true
|
|
*/
|
|
ClusterVector operator()(NDView<bool, 1> mask) {
|
|
if (static_cast<size_t>(mask.size()) != m_data.size()) {
|
|
throw std::runtime_error(
|
|
LOCATION + "Mask size does not match number of clusters");
|
|
}
|
|
ClusterVector result(capacity(), frame_number());
|
|
for (size_t i = 0; i < m_data.size(); ++i) {
|
|
if (mask(i)) {
|
|
result.push_back(m_data[i]);
|
|
}
|
|
}
|
|
return result;
|
|
}
|
|
|
|
// // Move assignment operator
|
|
// ClusterVector &operator=(ClusterVector &&other) noexcept {
|
|
// if (this != &other) {
|
|
// m_data = other.m_data;
|
|
// m_frame_number = other.m_frame_number;
|
|
// other.m_data.clear();
|
|
// other.m_frame_number = 0;
|
|
// }
|
|
// return *this;
|
|
// }
|
|
|
|
/**
|
|
* @brief Sum the pixels in each cluster
|
|
* @return std::vector<T> vector of sums for each cluster
|
|
*/
|
|
std::vector<T> sum() {
|
|
std::vector<T> sums(m_data.size());
|
|
|
|
std::transform(
|
|
m_data.begin(), m_data.end(), sums.begin(),
|
|
[](const ClusterType &cluster) { return cluster.sum(); });
|
|
|
|
return sums;
|
|
}
|
|
|
|
/**
|
|
* @brief Sum the pixels in the 2x2 subcluster with the biggest pixel sum in
|
|
* each cluster
|
|
* @return vector of sums index pairs for each cluster
|
|
*/
|
|
std::vector<Sum_index_pair<T, corner>> sum_2x2() {
|
|
std::vector<Sum_index_pair<T, corner>> sums_2x2(m_data.size());
|
|
|
|
std::transform(
|
|
m_data.begin(), m_data.end(), sums_2x2.begin(),
|
|
[](const ClusterType &cluster) { return cluster.max_sum_2x2(); });
|
|
|
|
return sums_2x2;
|
|
}
|
|
|
|
/**
|
|
* @brief Reserve space for at least capacity clusters
|
|
* @param capacity number of clusters to reserve space for
|
|
* @note If capacity is less than the current capacity, the function does
|
|
* nothing.
|
|
*/
|
|
void reserve(size_t capacity) { m_data.reserve(capacity); }
|
|
|
|
void resize(size_t size) { m_data.resize(size); }
|
|
|
|
void push_back(const ClusterType &cluster) { m_data.push_back(cluster); }
|
|
|
|
ClusterVector &operator+=(const ClusterVector &other) {
|
|
m_data.insert(m_data.end(), other.begin(), other.end());
|
|
|
|
return *this;
|
|
}
|
|
|
|
/**
|
|
* @brief Return the number of clusters in the vector
|
|
*/
|
|
size_t size() const { return m_data.size(); }
|
|
|
|
/**
|
|
* @brief Check if the vector is empty
|
|
*/
|
|
bool empty() const { return m_data.empty(); }
|
|
|
|
uint8_t cluster_size_x() const { return ClusterSizeX; }
|
|
|
|
uint8_t cluster_size_y() const { return ClusterSizeY; }
|
|
|
|
/**
|
|
* @brief Return the capacity of the buffer in number of clusters. This is
|
|
* the number of clusters that can be stored in the current buffer without
|
|
* reallocation.
|
|
*/
|
|
size_t capacity() const { return m_data.capacity(); }
|
|
|
|
auto begin() const { return m_data.begin(); }
|
|
|
|
auto end() const { return m_data.end(); }
|
|
|
|
/**
|
|
* @brief Return the size in bytes of a single cluster
|
|
*/
|
|
size_t item_size() const {
|
|
return sizeof(ClusterType); // 2 * sizeof(CoordType) + ClusterSizeX *
|
|
// ClusterSizeY * sizeof(T);
|
|
}
|
|
|
|
ClusterType *data() { return m_data.data(); }
|
|
ClusterType const *data() const { return m_data.data(); }
|
|
|
|
/**
|
|
* @brief Return a reference to the i-th cluster casted to type V
|
|
* @tparam V type of the cluster
|
|
*/
|
|
ClusterType &operator[](size_t i) { return m_data[i]; }
|
|
|
|
const ClusterType &operator[](size_t i) const { return m_data[i]; }
|
|
|
|
/**
|
|
* @brief Return the frame number of the clusters. 0 is used to indicate
|
|
* that the clusters come from many frames
|
|
*/
|
|
int32_t frame_number() const { return m_frame_number; }
|
|
|
|
void set_frame_number(int32_t frame_number) {
|
|
m_frame_number = frame_number;
|
|
}
|
|
};
|
|
|
|
/**
|
|
* @brief Reduce a cluster to a 2x2 cluster by selecting the 2x2 block with the
|
|
* highest sum.
|
|
* @param cv Clustervector containing clusters to reduce
|
|
* @return Clustervector with reduced clusters
|
|
* @note The cluster is filled using row major ordering starting at the top-left
|
|
* (thus for a max subcluster in the top left cornern the photon hit is at
|
|
* the fourth position)
|
|
*/
|
|
template <typename T, uint8_t ClusterSizeX, uint8_t ClusterSizeY,
|
|
typename CoordType>
|
|
ClusterVector<Cluster<T, 2, 2, CoordType>> reduce_to_2x2(
|
|
const ClusterVector<Cluster<T, ClusterSizeX, ClusterSizeY, CoordType>>
|
|
&cv) {
|
|
ClusterVector<Cluster<T, 2, 2, CoordType>> result;
|
|
for (const auto &c : cv) {
|
|
result.push_back(reduce_to_2x2(c));
|
|
}
|
|
return result;
|
|
}
|
|
|
|
/**
|
|
* @brief Reduce a cluster to a 3x3 cluster
|
|
* @param cv Clustervector containing clusters to reduce
|
|
* @return Clustervector with reduced clusters
|
|
*/
|
|
template <typename T, uint8_t ClusterSizeX, uint8_t ClusterSizeY,
|
|
typename CoordType>
|
|
ClusterVector<Cluster<T, 3, 3, CoordType>> reduce_to_3x3(
|
|
const ClusterVector<Cluster<T, ClusterSizeX, ClusterSizeY, CoordType>>
|
|
&cv) {
|
|
ClusterVector<Cluster<T, 3, 3, CoordType>> result;
|
|
for (const auto &c : cv) {
|
|
result.push_back(reduce_to_3x3(c));
|
|
}
|
|
return result;
|
|
}
|
|
|
|
} // namespace aare
|