This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

More: wish84 incredibly slow


Hello,

This was first raised at
http://www.cygwin.com/ml/cygwin/2004-06/msg00326.html and one or two people
tried to help during June, mainly by scanning straces, without much that was
useful emerging. It is driving me absolutely insane. If I provide a bit more
information I wonder whether this will trigger any ideas.

Scenario: running a completely up-to-date Cygwin on an astonishingly slow
Toshiba laptop. Allowing for that sloth, the whole system is adequately
fast. The single exception that I have found, more easily measured on the
Toshiba but evident on other machinery, is /bin/wish84.exe. E.g:

	$ time echo exit | /usr/bin/wish84
		real    0m28.279s
		user    0m0.230s
		sys    0m0.500s

This is a system that does _not_ require the explicit setting

	$ export TCL_LIBRARY=/usr/share/tcl8.4 

to function properly, though I have needed it in the past and I have a
friend running what we both think should be (and in all respects appears to
be) an identical Cygwin setup on his machine, and he _does_ need it.
(Incidentally, I deduce from some threads on this topic that even Cygwin
gurus don't seem fully to understand this strange inconsistency between
different setups?) Anyway, using the slow Toshiba because it usefully
exhibits measurable waiting times, but remembering that similar behaviour is
to be identified on other machines, here is what happens in a sequence of
repeated commands:

	~> time echo exit | /usr/bin/wish84
		real    0m28.279s
		user    0m0.230s
		sys     0m0.500s
	~> time echo exit | /usr/bin/wish84
		real    0m18.745s
		user    0m0.260s
		sys     0m0.530s
	~> time echo exit | /usr/bin/wish84
		real    0m8.515s
		user    0m0.410s
		sys     0m0.430s
	~> time echo exit | /usr/bin/wish84
		real    0m4.474s
		user    0m0.330s
		sys     0m0.540s
	~> time echo exit | /usr/bin/wish84
		real    0m4.467s
		user    0m0.290s
		sys     0m0.550s
	~> time echo exit | /usr/bin/wish84
		real    0m4.476s
		user    0m0.280s
		sys     0m0.530s

To within a few decimal points I have been able to mimic this behaviour time
and time again: "real" times steadily reduce with successive calls to wish84
until a steady state is reached at the 3rd or 4th call. (E.g. the attempt
after this one: real times were 26.148s, 18.261s, 10.144s, 4.406s, 4.399s,
...)

I attach cygcheck.out.

Thank you. Sorry for bringing this up again.

Fergus

Attachment: cygcheck.out
Description: Binary data

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]