Mail Archives: djgpp/1997/11/24/09:38:10
On Mon, 24 Nov 1997, Thomas Knudsen wrote:
> My DOS properties under Win95: No EMS, all other memory settings "Auto",
> idle sensitivity set to "Low". Screen settings: 50 lines, Fast ROM Emu,
> Dynamic memory allocation. These settings have (empirically) shown to be
> optimum, wrt. speed
That's consistent with what I know (and use for my Windows 95 setup) as
well.
> Any explanations are very welcome!
I don't have any, right now. I see similar effects as well with the
DJGPP version of GNU Emacs. Sometimes, its operation is significantly
slower than usual. I have found out that the only way to solve it is to
restart Windows(!), since restarting Emacs itself isn't enough. Once
restarted, I usually get the fast operation. Since I usually leave my
machine (including Emacs) working all the week and only shut it down on
weekends, it is not too nuch of a bother to restart it once in a few
days.
One thing to test is whether the program gets less CPU cycles from the
Windows scheduler, or indeed runs slower. This is very tricky with
Emacs, but maybe easier with your program. One way to check this is to
compile and link the program with -pg switch, then compare the profiles
generated by `gprof' for a fast run and a slow run.
- Raw text -