Further work with sensekey and ASCII input (in Csound) on the OLPC has led 
me to the following tentative conclusions:

1) The substantial CPU load penalty is not due primarily to sensekey, but 
rather to the way in which the OLPC processes ASCII input.

2) Even with only MIDI controller input, spurious ASCII keystrokes can cause 
a substantial penalty (from a quiescent 3%, up to 20-35% load).

3) Suppressing keyboard echo *may* cut the CPU penalty in about half 
(contrary to what I said before); but this is not a useful solution, as one 
is working blind from the terminal in that case.

If anyone could come up with a way of reducing the "ASCII CPU penalty" on 
the OLPC, I'd surely like to hear it. This penalty is nonexistent in 
Windows.

Also, does anyone know if it exists in the Mac world? (This "penalty" would 
apply in realtime whenever sensekey is present [even if it is effectively 
doing nothing], and keys are being pressed.)

Art Hunkins

----- Original Message ----- 
From: "Andres Cabrera" 
To: 
Sent: Tuesday, January 13, 2009 3:32 PM
Subject: [Csnd] Re: Re: Re: Re: Re: sensekey and Linux/OLPC


Thanks for the info Art. Now it's probably down to the way csound
grabs the keys from the shell. Since TamTam works fine, I suppose
there's some way to capture key events from python and pass them to
Csound more effcieintly.

Cheers,
Andrés

On Tue, Jan 13, 2009 at 12:41 PM, Art Hunkins  wrote:
> Andres,
>
> Thanks for the lead.
>
> I did google, and found:
> stty -echo
> and
> stty echo
>
> That works, and effectively kills ASCII display associated with sensekey
> during a realtime rendering. The code could of course be placed in a 
> script
> - as otherwise it fails to display, for example, your command to run 
> csound.
>
> More importantly, however, is the fact that it doesn't change CPU usage at
> all. (One more theory shot down.) So, what's happening here is still a
> mystery.
>
> Art Hunkins
>
> ----- Original Message ----- From: "Andres Cabrera" 
> 
> To: 
> Sent: Tuesday, January 13, 2009 8:31 AM
> Subject: [Csnd] Re: Re: Re: sensekey and Linux/OLPC
>
>
> Hi Art,
>
> I suppose there's a way to prevent the shell from echoing the keys
> pressed, but I don't know it. You may want to google along the lines
> bash-key-echo, etc. Let me know if you find it.
>
> Cheers,
> Andrés
>
> On Mon, Jan 12, 2009 at 10:13 PM, Art Hunkins  wrote:
>>
>> Andres,
>>
>> The problem occurs with both, though it is slightly less severe with the
>> shell (ctrl-alt-F1).
>>
>> Here are some data:
>> On my relatively-current Windows XP desktop, running sensekey alone in
>> real-time (command-line), gives a csound CPU load of 5% and a total CPU
>> load
>> of 9%. This is for ASCII keystrokes or no keystrokes (probably because
>> nothing is being printed to the console).
>>
>> On the OLPC, either as a terminal activity or a shell, the same run
>> averages
>> < 1% both for csound CPU and total CPU *with no ASCII input* (this in
>> itself
>> is rather amazing). However, with keyboard input, csound CPU usage can
>> easily reach 11% and total CPU, 22%. Recall, this is running the sensekey
>> opcode alone, on a single instrument. The only difference is in the ASCII
>> keystrokes and their printing to the console.
>>
>> So the problem to me seems to point to how the OLPC and sensekey are
>> together processing ASCII input, and printing it to the console.
>>
>> I'd much appreciate all developers' input on this issue - including Mac
>> behavior in this circumstance. (Does it print ASCII to the console when
>> sensekey is processing it?)
>>
>> FWIW, my testing was done at SR=44100 and a k-rate of 10, all messages
>> off.
>>
>> Art Hunkins
>>
>> ----- Original Message ----- From: "Andres Cabrera"
>> 
>> To: 
>> Sent: Friday, January 09, 2009 7:42 AM
>> Subject: [Csnd] Re: sensekey and Linux/OLPC
>>
>>
>> Hi Art,
>> Are you using the terminal activity or the shell that you get with
>> ctrl-alt-F1?
>> Does the problem occur with both?
>>
>> Cheers,
>> Andrés
>>
>> On Thu, Jan 8, 2009 at 11:22 PM, Art Hunkins  wrote:
>>>
>>> I continue to have issues of extremely heavy CPU usage with ASCII
>>> keyboard
>>> input on the OLPC.
>>>
>>> In Windows, there is a *modest* CPU penalty incurred with keyboard input
>>> (via sensekey), but with Linux/OLPC it is major.
>>>
>>> Of course, the OLPC prints the sensekey input to the console, whereas
>>> Windows does not, and this may well be part of the problem.
>>>
>>> Indeed if one watches CPU usage via top in Linux, you can see usage jump
>>> significantly (both overall and within Csound) with keyboard input.
>>>
>>> Can someone have a look at sensekey to determine:
>>> 1) if there's any way of keeping it from printing to the terminal (in
>>> Linux); 2) if there's any other way in which Linux overhead could be
>>> minimized with ASCII input?
>>>
>>> The penalty is so bad that I may not be able to use sensekey messaging
>>> with
>>> the OLPC, but rather use only MIDI controllers - which require no such
>>> overhead at all.
>>>
>>> Art Hunkins
>>>
>>>
>>>
>>> Send bugs reports to this list.
>>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe
>>> csound"
>>>
>>
>>
>>
>> --
>>
>>
>> Andrés
>>
>>
>> Send bugs reports to this list.
>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe
>> csound"=
>>
>>
>> Send bugs reports to this list.
>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe
>> csound"
>>
>
>
>
> --
>
>
> Andrés
>
>
> Send bugs reports to this list.
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe
> csound"=
>
>
> Send bugs reports to this list.
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe
> csound"
>



-- 


Andrés


Send bugs reports to this list.
To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe 
csound"=