TRANLE@INTELLICORP.COM (Minh Tran-Le) (01/09/90)
why is the mit R4 server on a PS2 AIX 1.1 so much slower than the ibm's R3 server when you run the ico demo program: ico -dbl -faces -colors red pink orange It is at least a factor x10 slower. Did I compile it wrong ?. It seems to behave like the ibm's R3 server when you don't have a 80387 floating point chip. Thanks, Tran-Le. arpanet: tranle@intellicorp.com uucp: ..sun!icmv!mtranle -------
rws@EXPO.LCS.MIT.EDU (Bob Scheifler) (01/10/90)
why is the mit R4 server on a PS2 AIX 1.1 so much slower than the ibm's
R3 server when you run the ico demo program:
ico -dbl -faces -colors red pink orange
It's using a multi-buffering extension, with a real time delay, rather
than "run as fast as you can". Try it with "-softdbl" instead.pniy@vax5.CIT.CORNELL.EDU (01/10/90)
The new version of ico in double buffer mode draws the entire window each time the icosahedron moves, rather than just a square region around the icosahedron as the old version did. Since there are many more pixels to copy over, it's bound to be slower. Does anyone know why it was changed? I can't see the reason. Arthur Smith
jsw@xhead.ESD.SGI.COM (Jeff Weinstein) (01/10/90)
In article <9001090556.AA11528@expo.lcs.mit.edu>, TRANLE@INTELLICORP.COM (Minh Tran-Le) writes: > why is the mit R4 server on a PS2 AIX 1.1 so much slower than the ibm's > R3 server when you run the ico demo program: > ico -dbl -faces -colors red pink orange > > It is at least a factor x10 slower. Did I compile it wrong ?. It seems to > behave like the ibm's R3 server when you don't have a 80387 floating point > chip. What you are probably seeing is the effect of the R4 ico trying to use the double buffering extension. On the ibm product server ico doesn't find the double buffer extension, so it does its fake colormap double buffering. On the R4 server ico finds the double buffer extension, and proceeds to use it. The problem is that the double buffering is done in software, and copies from memory to the screen are pretty slow because of the architecture of the displays on the ps/2. So, if you want double buffered ico to go fast you should disable the double buffer extension in the server. The ddx code on the R4 tape is pretty much exactly what we shipped in the ibm product server. It may have a few more bugs fixed, a few bugs introduced by the R4 integration, and perhaps a few performance enhancements. You should see exactly the same or better performance in the R4 server than the ibm product R3 server. --Jeff Formerly IBM server hacker, now... Jeff Weinstein Silicon Graphics, Inc., Entry Systems Division, Window Systems jsw@xhead.esd.sgi.com Any opinions expressed above are mine, not sgi's.
rws@EXPO.LCS.MIT.EDU (Bob Scheifler) (01/10/90)
So, if you want double buffered ico to go fast
you should disable the double buffer extension in the server.
No, RTFM. Use -softdbl.