Two thumbs up!

Two thumbs up!

Post by Charles Hann » Tue, 15 Oct 1991 21:45:04



Well, it appears that the upgrade to 3.1.6 will fix most (if not
all--though I don't know about NFS locking) of my current problems.
This is good.  Now I either have to call it in and*off my SE or
walk over to the local IBM office.  B-)

I now have amassed a pretty decent pile of literature on AIX and the
RS/6000, almost none of which is blatant hype.  In particular, I've
been reading "RISC System/6000 Technology".  All I can say is: two
thumbs up.  Now I understand why this machine blows away every other
workstation I use.  B-) (The 7 HP 720s do pretty well, but just aren't
as fast as I'd expect.  How's that for a subjective comparison?)

To all you people in the Advanced Workstation Division: consider this
a compliment, from one who rarely does so.  B-)

FYI, my RS/6K wish list:

*  Smalltalk
*  a faster X11R5
*  a Sabine adapter B-) (not really needed, but oh would it be nice!)
*  another RS/6K or three B-)/2 (only half joking, since this one is
starting to get overloaded)

--
- Mycroft, a.k.a. Charles Hannum, thinking about how to write a
Smalltalk system for the RS/6K (and realizing that he probably needs
to write some sort of Smalltalk-->C translator, since he isn't at all
interested in rewriting the XL optimizer this century).  <<Insert
standard pitch for free software.>>  Suggestions and valid pointers
(except NULL pointers, which shouldn't be valid) are, of course,
welcome.

P.S.  It's 8am, I haven't had any sleep, and I've been reading IBM
technical literature.  In other words, my brain is fried.  Please
understand if none of the above makes any sense.  B-)
--

 
 
 

Two thumbs up!

Post by Robert D. Sea » Wed, 16 Oct 1991 23:55:55



Quote:>FYI, my RS/6K wish list:
> ...
>*  a faster X11R5
>*  a Sabine adapter B-) (not really needed, but oh would it be nice!)

How about supported X11R5 on Sabines? Or UNsupported R5 on Sabines!
It's really quite frustrating having a Sabine (especially an 8-bit).
For one thing, IBM has been struggling to get R4 out.
I think they ought to punt on R4, and just write the R5 server code
for the rest of their adapters (for those who don't know, R5 from
MIT includes support for the dumb color adapter for rs/6000, and
NOTHING else). They could (barring internal software procedures and
marketing crapola) probably have the stuff to us before Halloween.  Well,
maybe Thanksgiving. Well, maybe Christmas. It's been delayed till 2Q92. Etc.
Anyway, get the base stuff to us now, and add swell IBM extensions
later (which we can ignore if we chose, and probably will, since
the executable size will quintuple).

It's also frustrating in that I can write lovely GL programs that
would probably work immediately on SGI machines; but the reverse is
almost never true. People always think you have 24 bits and z-buffer.
It'll probably get worse now that the Indigo can emulate 24 bits with
hardware and only 8 bitplanes. Oh no!

And finally, when these machines came out, the Sabine was known as the
High Performance 3D Adapter. And considering that its made by SGI,
one would be inclined to think that might be so. But, if you haven't
heard, it is not a particularly fast graphics adapter. It's OK for
me (I'm not particular), except when you have to open a window...

Sure crunches like hell, though!

rob

 
 
 

Two thumbs up!

Post by jim fro » Fri, 18 Oct 1991 01:31:53



Quote:>And finally, when these machines came out, the Sabine was known as the
>High Performance 3D Adapter. And considering that its made by SGI,
>one would be inclined to think that might be so. But, if you haven't
>heard, it is not a particularly fast graphics adapter. It's OK for
>me (I'm not particular), except when you have to open a window...

The performance problems you speak of are not in the Sabine board, but
rather in the X server.  The X server on the Sabine (even at AIX
3.1.6) is hellishly slow and buggy.  In particular opening windows and
processing keyboard events is very slow.  Running GL applications in
an already opened window without keyboard usage seems about as fast as
it was on an SGI 4D/70 (my test of "apparent speed" is running the SGI
flight simulator -- runs dandy so long as you're the only person
running anything on the system).

You've probably noticed by now that the X server is pretty sluggish.
This affects a number of things in GL applications.  Some GL calls are
translated into X calls by the GL library.  Among those are ginit(),
setcursor(), and the device queue reading functions.  Of these the
queue reading functions are the worst because they halt the graphics
pipeline when reading keyboard events.  This isn't a problem except
with highly interactive applications such as the flight simulator.

Also note that there are a number of bugs in the GL library -- ginit()
and gexit() make a lot of bogus calls to free() (these are ignored by
the IBM implementation so they are effectively invisible) and
setcursor() can spit out X protocol errors due to an unset variable
(this is rare but does happen with some applications).  These problems
and some others have been reported to IBM and they are in the process
of tracking them down.

I'm anxiously awaiting the R4 implementation that's supposed to ship
with AIX 3.2.  Hopefully it will correct the backing store and
colormap problems and speed things up a lot.

jim frost

 
 
 

1. Two Thumbs up for the AntiTrust Movie and Open Source

I just came from the first showing of the Antitrust movie,
http://www.antitrustthemovie.com , and give it a 2 thumbs up for support
for the Open Source Community. It very clearly shows what Microsoft has
and is doing and in the end, Open Source prevails!

A must see for all fellow Open Source geeks!!

2. xterm questions

3. Two thumbs-up to the Linux Community

4. / limit??????

5. Ten second thumb twiddling on X start

6. NEWBIE. / and // dirs

7. Shape Ups,Men's Shape Ups,Men's Skechers Shape Ups - new styles!

8. FreeBSD 4.4 and Eterm 0.9.1

9. UPS for two machines

10. APC BE500U 300 Watt UPS and two computers?

11. DIP and Two Dial-ups

12. diff -rcN ups-2.45.2/Makefile ups/Makefile

13. APC USB ups, Back-UPS ES series, 2.5.68