Re: Readline patch for high CPU usage

[Michael Goffioul]

On Mon, Dec 2, 2013 at 8:42 PM, Rik <address@hidden> wrote:

On 11/29/2013 03:16 PM, address@hidden wrote:
> Message: 5
> Date: Fri, 29 Nov 2013 15:29:20 -0500
> From: Michael Goffioul <address@hidden>
> To: "John W. Eaton" <address@hidden>
> Cc: Octave Maintainers List <address@hidden>
> Subject: Re: Readline patch for high CPU usage
> Message-ID:
>       <address@hidden>
> Content-Type: text/plain; charset="iso-8859-1"

>
> On Fri, Nov 29, 2013 at 2:26 PM, John W. Eaton <address@hidden> wrote:
>
>> > On 11/29/2013 11:30 AM, Michael Goffioul wrote:
>> >
>>> >> Hi,
>>> >>
>>> >> Here's a patch that solves the high CPU usage of readline under Win32
>>> >> and also allow to interrupt readline in CLI mode. This works for me with
>>> >> MSVC. Could anybody have this tested in MinGW?
>>> >>
>> >
>> > It seems to work for me.
>> >
>> > The task manager shows the CPU usage at between 4-10%.
> Really? That much...
> When I test it on my system, I get 0% CPU usage when octave is idle (WinXP
> 32 bits, Atom N470 1.6GHz, 512MB RAM). Maybe there's something not working
> as expected; 4-10% CPU usage seems way too much in idle state, as readline
> should have a "quiet" period of 100ms before looping

12/2/13

Michael,

The patch makes a big difference.  The octave executable is now in the 1-3%
range.  The csrss.exe process still consumes, on average, about 12% of CPU,
but this is still better than before and makes it workable on my Windows XP
machine.

This is still quite high CPU usage. I don't experience it with my MSVC-built octave. In idle state, both octave.exe and csrss.exe are at 0% CPU.

Michael.