Kaffe on Win95 (was bounced mail)
John D. Gwinner
gwinner at northnet.org
Tue Dec 9 14:12:19 PST 1997
> Kaffe internal threads would be Many-to-One implementation,
> Win32 would be One-to-One and latest, preferred on Solaris,
> implementation would be Many-to-Many. Checking other
> articles I am also given to try one Many-to-Many implementation
> on Win32 for Kaffe, but there are still too many issues unclear to
> me at this time.
Strange .. I'm not sure why they think Win32 is that way; there's a lot of
ways to go and it's more full featured than most Unix's. (it's also not
POSIX beyond L1, frustrating).
>> Consequently, many implementations of the one-to-one model, such as
Windows NT and the OS/2® threads package, limit the number of threads
supported on the system. <<
Not strictly true: Win32 threads are much lighter weight than most Unix
threads. Of course, some Unix thread libraries may be more efficient.
Under Win32 (95 or NT), there's a big difference between a thread and a
process; I think Unix has something called "lightweight threads", but not
Also, what's strange is that in their table they say that the kernel is
multithreaded on NT, but in the article they define it as One to One or
Many to One (not MT). Inconsistent.
I don't mind reading these articles as I learn more about Unix, but it
just bugs me at the inaccuracies that I find sometimes, especially when a
comparison is being made between OS's. I don't mind calling a spade a
spade, and there are a lot of things I just wish MS had done differently,
but I don't like calling a spade a club; it messes up the card game <G>.
> Oh, yes, JRE for Win32 allows you to distribute it in virgin form :-)
That *IS* a big change... at least for commercial licensees; anyway, I
take back last month's notes <G>.
> My situation is different. For me Kaffe on PC is just a testing/learning
> ground for implementing Kaffe JVM on a custom embedded system.
> Good luck ( I'll need it too ) and regards,
Wow, sounds like you will. C compilers for embedded systems aren't cheap;
a JVM would be a cool thing. I don't know though, it seems like there'd
be a lot of overhead, but I guess it depends on which embedded CPU; I'd
love to see a JavaVM for a 68HC11 <G>
== John ==
More information about the kaffe