Mercurial > hg > sonic-annotator
comparison runner/FeatureExtractionManager.cpp @ 177:795d67a5e266
Change internal blocksize from 16384 to 1024 (see comment in FeatureExtractionManager ctor)
author | Chris Cannam |
---|---|
date | Thu, 16 Oct 2014 10:32:16 +0100 |
parents | 447230267c0d |
children | 3f7c65f26559 |
comparison
equal
deleted
inserted
replaced
176:59abb58b1855 | 177:795d67a5e266 |
---|---|
52 #include <QFileInfo> | 52 #include <QFileInfo> |
53 | 53 |
54 FeatureExtractionManager::FeatureExtractionManager() : | 54 FeatureExtractionManager::FeatureExtractionManager() : |
55 m_summariesOnly(false), | 55 m_summariesOnly(false), |
56 // We can read using an arbitrary fixed block size -- | 56 // We can read using an arbitrary fixed block size -- |
57 // PluginBufferingAdapter handles this for us. It's likely to be | 57 // PluginBufferingAdapter handles this for us. But while this |
58 // quicker to use larger sizes than smallish ones like 1024 | 58 // doesn't affect the step and block size actually passed to the |
59 m_blockSize(16384), | 59 // plugin, it does affect the overall time range of the audio |
60 // input (which gets rounded to the nearest block boundary). So | |
61 // although a larger blocksize will normally run faster, and we | |
62 // used a blocksize of 16384 in earlier releases of Sonic | |
63 // Annotator for that reason, a smaller blocksize produces | |
64 // "better" results and this is particularly relevant now we | |
65 // support the start and duration flags for a transform. | |
66 m_blockSize(1024), | |
60 m_defaultSampleRate(0), | 67 m_defaultSampleRate(0), |
61 m_sampleRate(0), | 68 m_sampleRate(0), |
62 m_channels(0), | 69 m_channels(0), |
63 m_normalise(false) | 70 m_normalise(false) |
64 { | 71 { |