randy@tessa.iaf.uiowa.edu (randy frank) (07/30/90)
This is in general reply to Matt Machlis' questions about the IBM RS/6000. (the machine was a 3xx? 16Meg ram) I had a demo machine for about 3 weeks earlier this year. I ported about 15 applications to it and demoed it at our site. The apps were all IRIS GL code, mostly scientific FORTRAN and included p3d, NCSA HCVIS, NCSA HDF, FRODO, TGIF, some radiation planning software, a solids modeller, and basic image and 3d object display/ manipulation software and a GL based 128k Apple //e emulator. (the latter few being locally developed workspace applications.) Porting was fairly simple especially for our sw as it had been written using only base GL calls which could be isolated and have libs written around them for porting to other platforms. IE we plan to port as we code, including graphics. Graphically the speed was about the same as our 4D20. A noted exception is that the mapcolor routine was INCREDIBLY SLOW!!!. Also the implementation of popup menus left a LOT to be desired. A last graphics note, I was unable to get any of my rectread based software to work. IE I could NOT reliably read an image from the buffer. Most of the stuff I saw will probably be corrected in a future OS release. Basically it's a 4D20 graphically. The CPU was quick. Probably in the R3000 25Mhz range especially in floating point. If your code is integer only the machine was not much faster than a 4D25. The floating point stuff was difficult to measure as its performance was wildly erratic. Some stuff ran very fast but as the numbers got small (iterative fitting) it slowed down. I found the OS to be fairly stable for a prerelease and I liked their system administration tool (SMIT) very much. In a nutshell the machine was a 4D20-25 that was a little erratic. Its UNIX takes a little getting used to but is ok. I would like to see another release of the OS before I bought one (unless the discount were over 60%). Lots of potential... Note: Many of the ports were deleted and I don't know if I could provide the source to the port although I could provide hints. My porting success rate was about 80% in two weeks. I believe 90-95% could be ported. -- rjf. Randy Frank, Engineer | (319) 335-6712 University of Iowa, Image Analysis Facility | 73 EMRB randy@tessa.iaf.uiowa.edu | Iowa City, IA 52242
sergio@sergio.uucp (Sergio Perrone/30000) (07/31/90)
In reference to comments concerning GL on the IBM RS/6000... Problems with the speed of 'mapcolor' can be alleviated by using 'mapcolors' instead - you can load a complete 4096-entry colormap in the time it takes to execute one mapcolor command. This speeds things up considerably. The implementation of popup menus is different, because the RS/6000 is an X-based system. However, the functionality is all there. Also, I've been using the rectread/rectwrite subroutines extensively and, so far as I can see, everything works fine. Floating point performance for the model 320 is rated at more than four times that of a DECstation 3100, and integer (MIPS) performance is about twice that of a DS3100. Marc --- Marc Andreessen, IBM AWD Austin, sergio@sergio.austin.ibm.com --- --- Words and ideas contained herein are independent of IBM policy. ---
spencer@eecs.umich.edu (Spencer W. Thomas) (07/31/90)
A couple of things we noted were missing from the IBM RS/6000 GL library were gversion() and foreground(). Will these be provided in the future? -- =Spencer W. Thomas EECS Dept, U of Michigan, Ann Arbor, MI 48109 spencer@eecs.umich.edu 313-936-2616 (8-6 E[SD]T M-F)