Subject: Speaker bigMan - Performance
Date: Sat, 18 Aug 2001 15:20:45 -0500
Performance enhancements ideas requested.
I have been look into the Voice synthesis stuff. In going through the
examples in the image such as Speaker bigMan say: 'You can cheat, but
don''t get caught.'
I noticed that the speech pattern is slow like a 45rpm record played at
33 1/3. I tried several other examples and they all do the same. I
would hope that a 500 mhz PC should handle this but apparently not. I
evaluated the following:
TimeProfileBrowser onBlock: [Speaker bigMan say: 'You can cheat, but
don''t get caught.']
In the leaves section, a tremendous amount of time is lots in "hash" and
process resume. I am rusty on hash. Any practical suggestions on
improving the performance of hash or any of the other parts. I would
like to heard bigMan Speaker speak at a normal tempo.
' - 58 tallies, 1031 msec.
Error: this should not happenTreeError: this should not happen
69.0% {711ms} Speaker>>say:
|60.3% {622ms} CompositeEvent(VoiceEvent)>>playOn:delayed:
| |60.3% {622ms} CompositeEvent>>playOn:at:
| | 56.9% {587ms} PhoneticEvent>>playOn:at:
| | |56.9% {587ms} KlattVoice>>playPhoneticEvent:at:
| | | 56.9% {587ms} KlattVoice>>playEvent:segments:boundary:at:
| | | 36.2% {373ms} KlattVoice>>playEvent:frames:at:
| | | |15.5% {160ms} KlattSynthesizer>>samplesFromFrames:
| | | | |8.6% {89ms} primitives
| | | | |6.9% {71ms} OrderedCollection>>do:
| | | |10.3% {106ms} KlattVoice(Voice)>>playBuffer:at:
| | | | |10.3% {106ms} QueueSound(AbstractSound)>>play
| | | | | 10.3% {106ms} SoundPlayer class>>playSound:
| | | | | 10.3% {106ms} SoundPlayer class>>resumePlaying:
| | | | | 10.3% {106ms} SoundPlayer
class>>resumePlaying:quickStart:
| | | | | 10.3% {106ms} SoundPlayer
class>>startUpWithSound:
| | | | | 10.3% {106ms} SoundPlayer
class>>startPlayerProcessBu...e:rate:stereo:sound:
| | | | | 10.3% {106ms} Process>>resume
| | | |3.4% {35ms} KlattVoice>>dBFromLinear:
| | | |3.4% {35ms} KlattVoice>>linearFromdB:
| | | | 3.4% {35ms} SmallInteger(Number)>>raisedTo:
| | | 20.7% {213ms} KlattVoice>>currentFramesCount:
| | | 20.7% {213ms} KlattSegment>>left:right:speed:pattern:
| | | 15.5% {160ms} KlattSegment>>slopeWith:selector:speed:
| | | 13.8% {142ms} Dictionary>>at:
| | | 13.8% {142ms} Dictionary>>at:ifAbsent:
| | | 12.1% {125ms}
Dictionary(Set)>>findElementOrNil:
| | | 12.1% {125ms} Dictionary>>scanFor:
| | | 8.6% {89ms}
Symbol(SequenceableCollection)>>hash
| | | 6.9% {71ms} primitives
| | 3.4% {35ms} KlattVoice>>flush
| | 3.4% {35ms} KlattVoice>>playEvent:segments:boundary:at:
| | 3.4% {35ms} KlattVoice>>currentFramesCount:
| | 3.4% {35ms} KlattSegment>>left:right:speed:pattern:
|8.6% {89ms} Speaker>>eventsFromString:
| 5.2% {54ms} Clause>>accept:
| |3.4% {35ms} F0RenderingVisitor>>clause:
| 3.4% {35ms} Speaker>>clauseFromString:
| 3.4% {35ms} Speaker>>phraseFromString:
| 3.4% {35ms} Speaker>>wordFromString:
| 3.4% {35ms} PhoneticTranscriber>>transcriptionOf:
| 3.4% {35ms} PhoneticRule>>matches:at:
31.0% {320ms} Speaker class>>bigMan
22.4% {231ms} KlattVoice class(Voice class)>>new
|22.4% {231ms} KlattVoice>>initialize
| 22.4% {231ms} KlattSegmentSet class>>arpabet
| 22.4% {231ms} KlattSegmentSet>>initializeArpabet
8.6% {89ms} Speaker class>>new
8.6% {89ms} Speaker>>initialize
6.9% {71ms} PhoneticTranscriber class>>default
6.9% {71ms} PhoneticTranscriber class>>english
6.9% {71ms} PhoneticRule class>>english
Error: this should not happenLeavesError: this should not happen
10.3% {106ms} String(SequenceableCollection)>>hash
10.3% {106ms} Process>>resume
8.6% {89ms} SmallInteger>>hashMultiply
8.6% {89ms} KlattSynthesizer>>samplesFromFrames:
6.9% {71ms} Dictionary>>scanFor:
6.9% {71ms} OrderedCollection>>do:
3.4% {35ms} KlattVoice>>dBFromLinear:
3.4% {35ms} Symbol>>=
3.4% {35ms} KlattSegmentParameter>>fixed:
3.4% {35ms} String class(Object)>>hash
'
Thanks,
Phil
Note: emacspeak may use DECTalk speech servers.
