Mail Archives: djgpp/1997/01/26/08:29:23
On 24 Jan 1997 15:47:33 GMT, Fabrice Francès <frances AT ensica DOT fr> wrote:
[timer rate under NT]
>sorry, but this can not be true... if it were, my program would run
>50/18.2=2.7 times
>slower... and I have checked it runs only 1.4 times slower...
Hmm.. OK, soryr, you might be correct - I didn't check this very
carefully, I just noticed that the timer won't operate correctly. Anyway,
what I do know for a fact is that you can't access the sound card hardware
under NT, at least with the normal drivers, and that's basically all that
matters to me :)
>> So, yeah, this renders NT pretty useless for running many DOS
>> applications, but that's what DOS is for :)
>
>I know, but don't focus on DOS (nor on Open-DOS) or you will soon be
>outdated...
I don't - I'm mainly doing my development for Win32 platforms nowadays.
It's just that people looking for a good performance shouldn't be running
DOS programs under WinNT or Win95 - native Win32 applications are much
more suitable for that. Depending on your application, a native Win32 port
might be pretty easy or very difficult to do.
>This would mean NT has some tasks running, which grab nearly 30% of the CPU
>?
>I can't believe that... under Linux, my program runs near "real-time", and
>all the
>daemons sleep most of the time...
Not normally, yes, but I'm not sure if WinNT gives all the available CPU
time to DOS programs. In addition, hardware virtualization might be
slowing it down quite a bit.
For example, at the moment I have a bunch of programs open, and still have
only 3-4% of CPU in use. And my Pentium 90 certainly isn't a speed-demon.
You might want to check the free CPU time before starting your tests
though - programs like Netscape have a nasty habid of eating a lot of CPU
time even if they should be inactive.
Petteri
- Raw text -