File size: 7,278 Bytes
d844fc8 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 | #include <algorithm>
#include <chrono>
#include <cstdint>
#include <iomanip>
#include <iostream>
#include <set>
#include <utility>
#include <vector>
#include <boost/graph/adjacency_list.hpp>
#include "velographx/csr_graph.hpp"
#include "velographx/incremental/bfs.hpp"
#include "velographx/storage/dynamic_graph.hpp"
namespace boost_adapter {
using velographx::UpdateBatch;
using velographx::VertexId;
using Storage = boost::adjacency_list<boost::vecS, boost::vecS, boost::undirectedS>;
struct Graph {
explicit Graph(std::size_t vertices) : storage(vertices) {}
Storage storage;
std::uint64_t version{0};
};
std::size_t vx_vertex_count(const Graph& graph) { return boost::num_vertices(graph.storage); }
bool vx_is_directed(const Graph&) { return false; }
std::uint64_t vx_version(const Graph& graph) { return graph.version; }
template <class Fn>
void vx_for_each_neighbor(const Graph& graph, VertexId u, Fn&& fn) {
const auto [begin, end] = boost::adjacent_vertices(u, graph.storage);
for (auto it = begin; it != end; ++it) fn(static_cast<VertexId>(*it));
}
bool vx_has_edge(const Graph& graph, VertexId u, VertexId v) {
return boost::edge(u, v, graph.storage).second;
}
void vx_apply_updates(Graph& graph, const UpdateBatch& batch) {
if (batch.empty()) return;
for (const auto& op : batch.updates) {
if (op.src >= boost::num_vertices(graph.storage) ||
op.dst >= boost::num_vertices(graph.storage)) {
continue;
}
if (op.add) {
if (!boost::edge(op.src, op.dst, graph.storage).second) {
boost::add_edge(op.src, op.dst, graph.storage);
}
} else {
boost::remove_edge(op.src, op.dst, graph.storage);
}
}
++graph.version;
}
} // namespace boost_adapter
namespace {
using velographx::BasicIncrementalBFS;
using velographx::CsrGraph;
using velographx::DynamicGraph;
using velographx::UpdateBatch;
using velographx::VertexId;
std::vector<std::pair<VertexId, VertexId>> make_graph(std::size_t vertices) {
std::set<std::pair<VertexId, VertexId>> unique;
constexpr VertexId offsets[] = {1, 7, 31, 127};
for (VertexId u = 0; u < vertices; ++u) {
for (auto offset : offsets) {
const auto v = static_cast<VertexId>((static_cast<std::size_t>(u) + offset) % vertices);
if (u == v) continue;
const auto edge = std::minmax(u, v);
unique.insert({edge.first, edge.second});
}
}
return {unique.begin(), unique.end()};
}
std::vector<UpdateBatch> make_incremental_batches(std::size_t vertices) {
std::vector<UpdateBatch> batches;
batches.reserve(5);
for (VertexId i = 0; i < 5; ++i) {
UpdateBatch batch;
const auto u = static_cast<VertexId>((i * 17) % vertices);
const auto ring = static_cast<VertexId>((static_cast<std::size_t>(u) + 1) % vertices);
const auto chord = static_cast<VertexId>((static_cast<std::size_t>(u) + 3) % vertices);
batch.remove(std::min(u, ring), std::max(u, ring), i * 2);
batch.add(std::min(u, chord), std::max(u, chord), i * 2 + 1);
batches.push_back(std::move(batch));
}
return batches;
}
template <class Graph>
double median_recompute_us(Graph& graph, VertexId source, std::vector<std::uint32_t>& distances) {
BasicIncrementalBFS<Graph> bfs(graph, source);
std::vector<double> samples;
samples.reserve(5);
for (int rep = 0; rep < 5; ++rep) {
const auto begin = std::chrono::steady_clock::now();
bfs.recompute();
const auto end = std::chrono::steady_clock::now();
samples.push_back(std::chrono::duration<double, std::micro>(end - begin).count());
}
std::sort(samples.begin(), samples.end());
distances = bfs.distances();
return samples[samples.size() / 2];
}
std::uint64_t digest(const std::vector<std::uint32_t>& distances) {
std::uint64_t h = 1469598103934665603ULL;
for (auto d : distances) {
h ^= static_cast<std::uint64_t>(d);
h *= 1099511628211ULL;
}
return h;
}
} // namespace
int main(int argc, char** argv) {
const std::size_t vertices = argc > 1 ? static_cast<std::size_t>(std::stoull(argv[1])) : 8192;
const VertexId source = 0;
const auto edges = make_graph(vertices);
DynamicGraph dynamic(vertices, false);
dynamic.bulk_load_edges(edges);
CsrGraph csr(edges, false);
boost_adapter::Graph boost_graph(vertices);
for (const auto& [u, v] : edges) boost::add_edge(u, v, boost_graph.storage);
std::vector<std::uint32_t> dynamic_dist;
std::vector<std::uint32_t> csr_dist;
std::vector<std::uint32_t> boost_dist;
const auto dynamic_us = median_recompute_us(dynamic, source, dynamic_dist);
const auto csr_us = median_recompute_us(csr, source, csr_dist);
const auto boost_us = median_recompute_us(boost_graph, source, boost_dist);
const bool recompute_exact = dynamic_dist == csr_dist && dynamic_dist == boost_dist;
BasicIncrementalBFS<DynamicGraph> dynamic_incremental(dynamic, source);
BasicIncrementalBFS<boost_adapter::Graph> boost_incremental(boost_graph, source);
std::vector<double> dynamic_apply_samples;
std::vector<double> boost_apply_samples;
bool incremental_exact = true;
for (const auto& batch : make_incremental_batches(vertices)) {
auto begin = std::chrono::steady_clock::now();
dynamic_incremental.apply(batch);
auto end = std::chrono::steady_clock::now();
dynamic_apply_samples.push_back(std::chrono::duration<double, std::micro>(end - begin).count());
begin = std::chrono::steady_clock::now();
boost_incremental.apply(batch);
end = std::chrono::steady_clock::now();
boost_apply_samples.push_back(std::chrono::duration<double, std::micro>(end - begin).count());
incremental_exact = incremental_exact &&
dynamic_incremental.distances() == boost_incremental.distances();
}
std::sort(dynamic_apply_samples.begin(), dynamic_apply_samples.end());
std::sort(boost_apply_samples.begin(), boost_apply_samples.end());
const auto dynamic_apply_us = dynamic_apply_samples[dynamic_apply_samples.size() / 2];
const auto boost_apply_us = boost_apply_samples[boost_apply_samples.size() / 2];
const bool exact = recompute_exact && incremental_exact;
std::cout << std::fixed << std::setprecision(3)
<< "{\"schema_version\":1,"
<< "\"benchmark\":\"same-algorithm-storage-bfs\","
<< "\"backend\":\"boost-adjacency-list\","
<< "\"algorithm\":\"BasicIncrementalBFS\","
<< "\"vertices\":" << vertices << ','
<< "\"edges\":" << edges.size() << ','
<< "\"source\":" << source << ','
<< "\"repetitions\":5,"
<< "\"incremental_batches\":5,"
<< "\"exact\":" << (exact ? "true" : "false") << ','
<< "\"recompute_exact\":" << (recompute_exact ? "true" : "false") << ','
<< "\"incremental_exact\":" << (incremental_exact ? "true" : "false") << ','
<< "\"digest\":" << digest(dynamic_incremental.distances()) << ','
<< "\"dynamic_graph_median_us\":" << dynamic_us << ','
<< "\"csr_graph_median_us\":" << csr_us << ','
<< "\"boost_median_us\":" << boost_us << ','
<< "\"dynamic_incremental_median_us\":" << dynamic_apply_us << ','
<< "\"boost_incremental_median_us\":" << boost_apply_us
<< "}\n";
return exact ? 0 : 2;
}
|