Mail Archives: cygwin/1998/03/07/11:10:17
> I don't know anything about anything but here's a guess.
> If B19 used slightly more ram than B18, and your system
> was starved for RAM, IE you had ~8-24 MBytes, then a
> 10% increase in ram consumption could result in a non-linear
> decrease in performance 2x, even 3x.
> A good test would be to compare mem use under B18 vs B19
> by looking at 'non-paged pool' (which os are you in 95 or NT?)
> Please specifiy your system configureation when mentioning
> performance problems
Sorry about that. I am running NT4 sp3 on a Intel PII 300Mhz with 64
Megs of RAM and a 75 Meg Swap file. Total Kernal memory used
while running is 9.1 Megs (6.8 paged and 2.3 non-paged). The
peak memory utilization with all processes running (ie tcsh, emacs,
ediff etc) is 45Megs well below the limit of 139 Megs. I strongly
doubt that it is a memory problem. I seems more like a
spawn/fork or a IPC/pipe problem. The coolview patch helping a
bit leads me to belive that a bit more.
I have no problems with helping with the debug, but I have no clue
where to start looking. "make" not displaying the stderr for failures
is the most pressing problem (see below). The slow responce, I can
live with for a little while.
Any help and/or suggestions will be greatly appreciated.
-raj
>At 06:37 PM 3/4/98 -0700, you wrote:
>>Hello,
>>
>>I have seen a couple of other post about similar problems, but
>>no solutions that worked for me. I just upgraded to B19 easily.
>>All gnuwin32 programs run significantly slower than in B18 and
>>error info in makes disappears.
>>
>>A make on my project that used to take about 45 seconds in B18
>>now takes over 3 minutes. Exitting bash or tcsh takes about
>>10-30 seconds (sometime 10, sometimes 20, sometime 30).
>>I added to coolview patches and things got a little better, but
>>not by much.
>>
>>Also, when I run ediff in emacs, refining regions take 8 seconds
>>with b19 when it used to happen almost instantly unde b18.
>>
>>Also, I set CYGWIN32 to tty (from the control pannel) for all users.
>>Still, stderr from b19 programs seem to disappear in bash, tcsh
>>and cmd.exe. They do show up correctly when I run the compiler
>>by hand outside make.
>>
>>Also, I have to run make with the win32 setting as some of the NT
based
>>compilers I use return a code that causes make to think
>>it is an error (ie. reaps a losing child instead of a winning child).
>>It seems like "make --unix" does not like the return value of the
>>compiler in a bash/tcsh/cmd window but "make --win32" will
>>accept it just fine. All this used to work in B18.
>>
>>Any ideas?
>>
>>-raj
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request AT cygnus DOT com" with one line of text: "help".
- Raw text -