Re: Cygwin performance (was [ANN] PW32 the...)

Re: Cygwin performance (was [ANN] PW32 the...)

[Chris Nappi]

I know that I would GREATLY appreciate it if someone would put some sort of stat_lite into ls.  I work with cygwin almost exclusively over a Samba connection, and have been forced to use the ls from the Microsoft Unix toolkit because the speed on ls -l, ls -CF and ls --color is so slow over a network drive.
Chris
Heribert Dahms <heribert_dahms@icon-gmbh.de> wrote:

> are you hardwired to 'ls', 'ls -l'  or (like me) 'll'?
> My stock b20 'ls' spits out only filenames!

>I usually use 'ls -l', but what I really want (usually) is the filenames,
>sizes and mod times (and sometimes permissions). And, if I want that, I'll
>have to take the stat() penalty, I guess. So...I was curious...what is it in
>he stat() call that can only be obtained by opening the file? Is it just the
>ygnus stat() emulation that does that? One of these days, maybe I'll try out
> native-Win32 perl 'ls' wrapper script...

>cott.Blachowicz@seaslug.org




--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: Cygwin performance (was [ANN] PW32 the...)

[Larry Hall (RFK Partners, Inc)]

At 01:26 PM 3/16/00, Chris Nappi wrote:
>I know that I would GREATLY appreciate it if someone would put some sort of stat_lite into ls.  I work with cygwin almost exclusively over a Samba connection, and have been forced to use the ls from the Microsoft Unix toolkit because the speed on ls -l, ls -CF and ls --color is so slow over a network drive.
>Chris


Just so things don't get confused, in order to determine information that 
ls -CF and ls --color need, files need to be open.  The Win32 API doesn't
provide other options.  So, these features don't allow you to have it both
ways (i.e. you either wait or you don't use those features).  However, all 
of this I'm addressing in the context of what happens in the Cygwin 
implementation of stat().  If one is going to take the time to actually 
modify ls, one might as well remove the calls to stat() and replace them 
with whatever one feels is an appropriate substitute (assuming
one feels that decreased functionality for these options is adequate).
Making such a change only requires access to the code for ls (from the 
Cygwin source or any GNU site) and a bit of programming.  No Cygwin (stat() 
or otherwise) needed!;-)



Larry Hall                              lhall@rfk.com
RFK Partners, Inc.                      http://www.rfk.com
118 Washington Street                   (508) 893-9779 - RFK Office
Holliston, MA 01746                     (508) 893-9889 - FAX
                                        (508) 560-1285 - cell phone



--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: Cygwin performance (was [ANN] PW32 the...)

[Chris Faylor]

On Thu, Mar 16, 2000 at 09:32:03AM -0800, Scott Blachowicz wrote:
>Heribert Dahms <heribert_dahms@icon-gmbh.de> wrote:
>> are you hardwired to 'ls', 'ls -l'  or (like me) 'll'?
>> My stock b20 'ls' spits out only filenames!
>
>I usually use 'ls -l', but what I really want (usually) is the filenames,
>sizes and mod times (and sometimes permissions). And, if I want that, I'll
>have to take the stat() penalty, I guess. So...I was curious...what is it in
>the stat() call that can only be obtained by opening the file? Is it just the
>cygnus stat() emulation that does that? One of these days, maybe I'll try out
>a native-Win32 perl 'ls' wrapper script...

The Windows API dictates that much of the information required to fill in a
stat structure can only come by querying an open file handle.

cgf

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: Cygwin performance (was [ANN] PW32 the...)

[Scott Blachowicz]

Heribert Dahms <heribert_dahms@icon-gmbh.de> wrote:

> are you hardwired to 'ls', 'ls -l'  or (like me) 'll'?
> My stock b20 'ls' spits out only filenames!

I usually use 'ls -l', but what I really want (usually) is the filenames,
sizes and mod times (and sometimes permissions). And, if I want that, I'll
have to take the stat() penalty, I guess. So...I was curious...what is it in
the stat() call that can only be obtained by opening the file? Is it just the
cygnus stat() emulation that does that? One of these days, maybe I'll try out
a native-Win32 perl 'ls' wrapper script...

Scott.Blachowicz@seaslug.org

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

RE: Cygwin performance (was [ANN] PW32 the...)

[Larry Hall (RFK Partners, Inc)]

This is a good point and one we shouldn't loose sight of.  I, like many
others, don't use the "default" ls (my ls is aliased to ls -CF).  At least
in my case, the listing is slowed down because all files need to be opened
to determine their type.  From my recollection, ls without any bells and 
whistles does not require this and therefore any performance degradation 
noticed here on network drives is the result of just network overhead.  That
doesn't mean that this overhead couldn't be lessened nor that it wouldn't
be good to find ways to make these embellished accesses work more quickly,
across the network or otherwise.  However, it does seem prudent to be 
specific about what causes what to be slow.  Operations that require the 
files to be opened on a local or network disk will always be slower than
those that do not.

Larry


At 04:41 PM 3/15/00, Heribert Dahms wrote:
>Hi Scott,
>
>are you hardwired to 'ls', 'ls -l'  or (like me) 'll'?
>My stock b20 'ls' spits out only filenames!
>
>Bye, Heribert (heribert_dahms@icon-gmbh.de)
>
> > -----Original Message-----
> > From: Scott Blachowicz [SMTP:scott@sabami.seaslug.org]
> > Sent: Tuesday, March 15, 1994 01:29
> > To:   cygwin@sourceware.cygnus.com
> > Subject:      Re: Cygwin performance (was [ANN] PW32 the...) 
> > 
> > Geoffrey Noer <noer@cygnus.com> wrote:
> > 
> > > ...
> > > Interesting. We have been trying to improve (and succeeding in
> > improving)
> > > Cygwin's runtime performance but that's been done comparing Cygwin
> > to
> > > Cygwin-past and not so much by doing benchmarks against other
> > systems I
> > > think.
> > 
> > Great! Have you found any way to improve the performance of commands
> > like 'ls'
> > against remotely mounted file systems? I frequently have things like
> > 
> >  NET USE * \\SERVER\SHARE
> > 
> > where SERVER is located on the far end of a PPTP link to a system a
> > few
> > thousand miles (18-22 hops over the Internet via an ISDN connection on
> > my end)
> > and doing an 'ls' is unuseably slow (and I think I've tried various
> > releases
> > from b17 to b20.1). So, I usually try to remember to use the "command
> > prompt"
> > and the DIR command which works just fine. I also wave perl scripts
> > over the
> > remote directories (scripts that do file globbing and file system
> > traversals)
> > and they run fine...but they don't try to get all the file info that
> > an 'ls
> > -l' would - ought to try out an 'ls' command from the Perl Power Tools
> > set
> > sometime...
> > 
> > At any rate...since 'ls' is hardwired into my fingers and I wander
> > into these
> > directories often enough, using cygwin can be painful, so I haven't
> > gotten
> > fully into playing with it yet.
> > 
> > > Have people run any benchmarks comparing Cygwin, Uwin, NuTcracker,
> > Interix,
> > > anything else out there?
> > 
> > That would be useful info!
> > 
> > Scott.Blachowicz@seaslug.org
> > 
> > --
> > Want to unsubscribe from this list?
> > Send a message to cygwin-unsubscribe@sourceware.cygnus.com
>
>--
>Want to unsubscribe from this list?
>Send a message to cygwin-unsubscribe@sourceware.cygnus.com



--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

RE: Cygwin performance (was [ANN] PW32 the...)

[Heribert Dahms]

Hi Scott,

are you hardwired to 'ls', 'ls -l'  or (like me) 'll'?
My stock b20 'ls' spits out only filenames!

Bye, Heribert (heribert_dahms@icon-gmbh.de)

> -----Original Message-----
> From:	Scott Blachowicz [SMTP:scott@sabami.seaslug.org]
> Sent:	Tuesday, March 15, 1994 01:29
> To:	cygwin@sourceware.cygnus.com
> Subject:	Re: Cygwin performance (was [ANN] PW32 the...) 
> 
> Geoffrey Noer <noer@cygnus.com> wrote:
> 
> > ...
> > Interesting. We have been trying to improve (and succeeding in
> improving)
> > Cygwin's runtime performance but that's been done comparing Cygwin
> to
> > Cygwin-past and not so much by doing benchmarks against other
> systems I
> > think.
> 
> Great! Have you found any way to improve the performance of commands
> like 'ls'
> against remotely mounted file systems? I frequently have things like
> 
>  NET USE * \\SERVER\SHARE
> 
> where SERVER is located on the far end of a PPTP link to a system a
> few
> thousand miles (18-22 hops over the Internet via an ISDN connection on
> my end)
> and doing an 'ls' is unuseably slow (and I think I've tried various
> releases
> from b17 to b20.1). So, I usually try to remember to use the "command
> prompt"
> and the DIR command which works just fine. I also wave perl scripts
> over the
> remote directories (scripts that do file globbing and file system
> traversals)
> and they run fine...but they don't try to get all the file info that
> an 'ls
> -l' would - ought to try out an 'ls' command from the Perl Power Tools
> set
> sometime...
> 
> At any rate...since 'ls' is hardwired into my fingers and I wander
> into these
> directories often enough, using cygwin can be painful, so I haven't
> gotten
> fully into playing with it yet.
> 
> > Have people run any benchmarks comparing Cygwin, Uwin, NuTcracker,
> Interix,
> > anything else out there?
> 
> That would be useful info!
> 
> Scott.Blachowicz@seaslug.org
> 
> --
> Want to unsubscribe from this list?
> Send a message to cygwin-unsubscribe@sourceware.cygnus.com

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: Cygwin performance (was [ANN] PW32 the...)

[Chris Faylor]

On Mon, Mar 14, 1994 at 04:28:47PM -0800, Scott Blachowicz wrote:
>Geoffrey Noer <noer@cygnus.com> wrote:
>
>> ...
>> Interesting. We have been trying to improve (and succeeding in improving)
>> Cygwin's runtime performance but that's been done comparing Cygwin to
>> Cygwin-past and not so much by doing benchmarks against other systems I
>> think.
>
>Great! Have you found any way to improve the performance of commands like 'ls'
>against remotely mounted file systems? I frequently have things like

I think Geoff was referring to improvements that are already in Cygwin.  If you
don't see any changes then there aren't any improvements.

cgf

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: Cygwin performance (was [ANN] PW32 the...)

[Scott Blachowicz]

Geoffrey Noer <noer@cygnus.com> wrote:

> ...
> Interesting. We have been trying to improve (and succeeding in improving)
> Cygwin's runtime performance but that's been done comparing Cygwin to
> Cygwin-past and not so much by doing benchmarks against other systems I
> think.

Great! Have you found any way to improve the performance of commands like 'ls'
against remotely mounted file systems? I frequently have things like

 NET USE * \\SERVER\SHARE

where SERVER is located on the far end of a PPTP link to a system a few
thousand miles (18-22 hops over the Internet via an ISDN connection on my end)
and doing an 'ls' is unuseably slow (and I think I've tried various releases
from b17 to b20.1). So, I usually try to remember to use the "command prompt"
and the DIR command which works just fine. I also wave perl scripts over the
remote directories (scripts that do file globbing and file system traversals)
and they run fine...but they don't try to get all the file info that an 'ls
-l' would - ought to try out an 'ls' command from the Perl Power Tools set
sometime...

At any rate...since 'ls' is hardwired into my fingers and I wander into these
directories often enough, using cygwin can be painful, so I haven't gotten
fully into playing with it yet.

> Have people run any benchmarks comparing Cygwin, Uwin, NuTcracker, Interix,
> anything else out there?

That would be useful info!

Scott.Blachowicz@seaslug.org

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Cygwin performance (was [ANN] PW32 the...)

[Geoffrey Noer]

On Mon, Mar 13, 2000, Jeff Sturm wrote:
[...]
> Don't be so sure.  I've used GCC on Interix for a while... let's just
> say that Cygwin rocks.
> 
> The Interix product relies heavily on the POSIX subsystem, instead of
> shared memory as Cygwin does.  For many of my user applications the
> subsystem process actually consumes more CPU time than the application!
[...]

Interesting.  We have been trying to improve (and succeeding in
improving) Cygwin's runtime performance but that's been done comparing
Cygwin to Cygwin-past and not so much by doing benchmarks against
other systems I think.

Have people run any benchmarks comparing Cygwin, Uwin, NuTcracker,
Interix, anything else out there?

-- 
Geoffrey Noer				Email: noer@cygnus.com
Cygnus Solutions, a Red Hat company	http://www.redhat.com/

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com