TIB: Lets Beat a horse
[Prev][Next][Index][Thread]
TIB: Lets Beat a horse
Sorry to beat a dead horse but...
--- begin forwarded text
Here's an article I picked up recently regarding DOS in Windows 95:
Windows 95 is DOS 7 + Windows 4. Thats it. Nothing more to it. I don't
see why people would need to debate that. Its not like it's a bad thing.
As an example: When Windows 95 boots press F8 to get the screen where you
can boot it in DOS-prompt only mode. Do that, and then go into your
Windows 3.11 directory. Type WIN and, hey presto! your running Windows
3.11. Exit Windows 3.11 and go to the root. type Win, and your starting
Windows 95. Then do select Shutdown and select "Restart the computer in
MS-DOS mode". Does the computer restart? Nope, it exits Windows. No
restart whatsoever.
Not really. The difference really boils down to Win95 handles more things
with VxD's than Win3 did. The architecture is about the same. It has to
be...otherwise you would not be able to load DOS drivers and TSR's before
Windows and have them appear in all DOS sessions. OS/2 can run DOS
drivers and TSR's, but they only appear in the session in which they are
loaded because there is no "underlying" DOS.
Windows 3+ has always run DOS in V86 mode, and could always run multiple
DOS sessions. The multitasking wasn't very good, but it could run
multiple DOS sessions if you had patience.
You're right, DOS is not used for all that much under Win95... but it
wasn't under Windows for Workgroups either if you had 32-bit file access
enabled.
MS is following a strategy of gradual evolution. With each release of
Windows, DOS becomes less important. But Win95 is not "all new". It
relies quite heavily on 16-bit Windows 3 code, even for 32-bit programs.
Many functions in USER32 and GDI32 are "thunked" down to the USER16 and
GDI16. That's the origin of the "Win16MutEx" sema- phore that protects
such code from being reentered. All programs, even 32-bit ones, get a DOS
PSP structure. Which makes sense, as the old 16-bit Windows code relies
on having that.
Frankly, with all the thunking, the long interrupt handler chains, the
mode transitions, etc, I am amazed that it performs as well as it does.
But maybe it's not all that great...
One article in Windows/DOS Journal this month is a study comparing the
context-switch performance of '95 and NT. One would think that '95 would
be faster, as it is Intel-specific and assembler-optimized, while NT is
portable and written in C and C++.
One would be wrong. According to the article, NT context-switches about
twice as fast. The author speculates that it is all the thunking and such
that kills performance. He points out that '95 has to go all the way down
to DOS to context-switch, because it has to switch DOS PSP's. Going "down
to DOS" involves quite a large number of traps and mode transitions, a
lot of overhead.
Conclusion is that if you are targeting Win95, you should think carefully
before using threads. Using too many might make your application much
slower than it would be otherwise.
So, if DOS isn't used, why is all this thunking going on? Why are they
giving each program a PSP, and switching PSP's on a context switch?
Answer--DOS _is_ still being used for crucial functions. Whether this is
good or bad is a different argument.
To unsubscribe, send a message to: listmaster@marlowe.net and include the
following in the body of the message:
leave macmarines your_email_address
(substitute the email address you used when joining for "youremailaddress")
--- end forwarded text
---
I pledge allegiance to the Mac of Apple Computer Incorporated, and to the
developers for which it stands, one platform, under Guy, indestructible,
with creativity and multimedia for all.
Follow-Ups: