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
|
// Copyright 2017 Yahoo Holdings. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
#include "match_master.h"
#include "docid_range_scheduler.h"
#include "match_loop_communicator.h"
#include "match_thread.h"
#include <vespa/searchlib/attribute/attribute_operation.h>
#include <vespa/searchlib/engine/trace.h>
#include <vespa/searchlib/common/featureset.h>
#include <vespa/vespalib/util/thread_bundle.h>
#include <vespa/vespalib/data/slime/inserter.h>
#include <vespa/vespalib/data/slime/inject.h>
#include <vespa/vespalib/data/slime/cursor.h>
#include <vespa/eval/eval/tensor.h>
#include <vespa/eval/eval/tensor_engine.h>
#include <vespa/vespalib/objects/nbostream.h>
#include <vespa/log/log.h>
LOG_SETUP(".proton.matching.match_master");
namespace proton::matching {
using namespace search::fef;
using search::queryeval::SearchIterator;
using search::FeatureSet;
using search::attribute::AttributeOperation;
namespace {
struct TimedMatchLoopCommunicator : IMatchLoopCommunicator {
IMatchLoopCommunicator &communicator;
fastos::StopWatch rerank_time;
TimedMatchLoopCommunicator(IMatchLoopCommunicator &com) : communicator(com) {}
double estimate_match_frequency(const Matches &matches) override {
return communicator.estimate_match_frequency(matches);
}
Hits selectBest(SortedHitSequence sortedHits) override {
auto result = communicator.selectBest(sortedHits);
rerank_time.start();
return result;
}
RangePair rangeCover(const RangePair &ranges) override {
RangePair result = communicator.rangeCover(ranges);
rerank_time.stop();
return result;
}
};
DocidRangeScheduler::UP
createScheduler(uint32_t numThreads, uint32_t numSearchPartitions, uint32_t numDocs)
{
if (numSearchPartitions == 0) {
return std::make_unique<AdaptiveDocidRangeScheduler>(numThreads, 1, numDocs);
}
if (numSearchPartitions <= numThreads) {
return std::make_unique<PartitionDocidRangeScheduler>(numThreads, numDocs);
}
return std::make_unique<TaskDocidRangeScheduler>(numThreads, numSearchPartitions, numDocs);
}
} // namespace proton::matching::<unnamed>
ResultProcessor::Result::UP
MatchMaster::match(search::engine::Trace & trace,
const MatchParams ¶ms,
vespalib::ThreadBundle &threadBundle,
const MatchToolsFactory &mtf,
ResultProcessor &resultProcessor,
uint32_t distributionKey,
uint32_t numSearchPartitions)
{
fastos::StopWatch query_latency_time;
query_latency_time.start();
vespalib::DualMergeDirector mergeDirector(threadBundle.size());
MatchLoopCommunicator communicator(threadBundle.size(), params.heapSize, mtf.createDiversifier(params.heapSize));
TimedMatchLoopCommunicator timedCommunicator(communicator);
DocidRangeScheduler::UP scheduler = createScheduler(threadBundle.size(), numSearchPartitions, params.numDocs);
std::vector<MatchThread::UP> threadState;
std::vector<vespalib::Runnable*> targets;
for (size_t i = 0; i < threadBundle.size(); ++i) {
IMatchLoopCommunicator &com = (i == 0)
? static_cast<IMatchLoopCommunicator&>(timedCommunicator)
: static_cast<IMatchLoopCommunicator&>(communicator);
threadState.emplace_back(std::make_unique<MatchThread>(i, threadBundle.size(), params, mtf, com, *scheduler,
resultProcessor, mergeDirector, distributionKey,
trace.getRelativeTime(), trace.getLevel()));
targets.push_back(threadState.back().get());
}
resultProcessor.prepareThreadContextCreation(threadBundle.size());
threadBundle.run(targets);
ResultProcessor::Result::UP reply = resultProcessor.makeReply(threadState[0]->extract_result());
query_latency_time.stop();
double query_time_s = query_latency_time.elapsed().sec();
double rerank_time_s = timedCommunicator.rerank_time.elapsed().sec();
double match_time_s = 0.0;
std::unique_ptr<vespalib::slime::Inserter> inserter;
if (trace.shouldTrace(4)) {
inserter = std::make_unique<vespalib::slime::ArrayInserter>(trace.createCursor("match_threads").setArray("threads"));
}
for (size_t i = 0; i < threadState.size(); ++i) {
const MatchThread & matchThread = *threadState[i];
match_time_s = std::max(match_time_s, matchThread.get_match_time());
_stats.merge_partition(matchThread.get_thread_stats(), i);
if (inserter && matchThread.getTrace().hasTrace()) {
vespalib::slime::inject(matchThread.getTrace().getRoot(), *inserter);
}
}
_stats.queryLatency(query_time_s);
_stats.matchTime(match_time_s - rerank_time_s);
_stats.rerankTime(rerank_time_s);
_stats.groupingTime(query_time_s - match_time_s);
_stats.queries(1);
if (mtf.match_limiter().was_limited()) {
_stats.limited_queries(1);
}
return reply;
}
FeatureSet::SP
MatchMaster::getFeatureSet(const MatchToolsFactory &mtf,
const std::vector<uint32_t> &docs, bool summaryFeatures)
{
MatchTools::UP matchTools = mtf.createMatchTools();
if (summaryFeatures) {
matchTools->setup_summary();
} else {
matchTools->setup_dump();
}
RankProgram &rankProgram = matchTools->rank_program();
std::vector<vespalib::string> featureNames;
FeatureResolver resolver(rankProgram.get_seeds(false));
featureNames.reserve(resolver.num_features());
for (size_t i = 0; i < resolver.num_features(); ++i) {
featureNames.emplace_back(resolver.name_of(i));
}
auto retval = std::make_shared<FeatureSet>(featureNames, docs.size());
if (docs.empty()) {
return retval;
}
FeatureSet &fs = *retval;
SearchIterator &search = matchTools->search();
search.initRange(docs.front(), docs.back()+1);
for (uint32_t i = 0; i < docs.size(); ++i) {
if (search.seek(docs[i])) {
uint32_t docId = search.getDocId();
search.unpack(docId);
auto * f = fs.getFeaturesByIndex(fs.addDocId(docId));
for (uint32_t j = 0; j < featureNames.size(); ++j) {
if (resolver.is_object(j)) {
auto obj = resolver.resolve(j).as_object(docId);
if (const auto *tensor = obj.get().as_tensor()) {
vespalib::nbostream buf;
tensor->engine().encode(*tensor, buf);
f[j].set_data(vespalib::Memory(buf.peek(), buf.size()));
} else {
f[j].set_double(obj.get().as_double());
}
} else {
f[j].set_double(resolver.resolve(j).as_number(docId));
}
}
} else {
LOG(debug, "getFeatureSet: Did not find hit for docid '%u'. Skipping hit", docs[i]);
}
}
if (auto onSummaryTask = mtf.createOnSummaryTask()) {
onSummaryTask->run(docs);
}
return retval;
}
}
|
