dylan@cs.washington.edu (Dylan McNamee) (03/10/90)
Thanks to many people on the net, here is a SetCPU file entry for using CardROM option with the 2091: 0x202 0x03 0x10000 0x2000 0x3FFF CBM_2091_SCSI_Controller These are, in order, Manufacturer, Product (may be 0x02, use Dave Haynie's CacheCard Program to check,) Card size, Base Address of ROM, size of ROM. I found the size of the ROM by guessing and checking. The system crashed when I was wrong. Be sure to use the verbose option of setcpu, to be sure your file was accepted, and always use the cardrom option along with the Fastrom option. Now for the wierd part; I'm not getting _any_ performance improvement. It's about 382K/sec Reads, and 335K/sec Writes (32KBits at a time) using DiskSpeed. Has anyone had any better luck? About what should I be getting? Anyone? Anyone? dylan mcnamee dylan@cs.washington.edu
ammrk@swbatl.sbc.com (Mike R. Kraml) (03/12/90)
In article <11051@june.cs.washington.edu> dylan@june.cs.washington.edu (Dylan McNamee) writes: > >Thanks to many people on the net, here is a SetCPU file entry for using >CardROM option with the 2091: > >0x202 0x03 0x10000 0x2000 0x3FFF CBM_2091_SCSI_Controller > >These are, in order, Manufacturer, Product (may be 0x02, use Dave Haynie's >CacheCard Program to check,) Card size, Base Address of ROM, size of ROM. > >I found the size of the ROM by guessing and checking. The system crashed >when I was wrong. Be sure to use the verbose option of setcpu, to be sure >your file was accepted, and always use the cardrom option along with the >Fastrom option. > >Now for the wierd part; I'm not getting _any_ performance improvement. >It's about 382K/sec Reads, and 335K/sec Writes (32KBits at a time) using >DiskSpeed. Has anyone had any better luck? About what should I be getting? >Anyone? Anyone? NO! I did the same for my 2090a without any real improvments. I think I got 3k per second improvments on the top end (about 460k/sec with buffering, vs 457k/sec), but they were minor to say the least. > >dylan mcnamee >dylan@cs.washington.edu Just thought I would put in my 2 cents worth. Take it easy, Mike... -- ============================================================================= Mike Kraml - Manager-Separations MECHANIZATION - SWBT - (The Techies) UUCP: {uunet, bellcore, texbell}...!swbatl!slims!ammrk =============================================================================
stan@teroach.UUCP (Stan Fisher) (03/13/90)
In article <11051@june.cs.washington.edu> dylan@june.cs.washington.edu (Dylan McNamee) writes: > >Thanks to many people on the net, here is a SetCPU file entry for using >CardROM option with the 2091: > >0x202 0x03 0x10000 0x2000 0x3FFF CBM_2091_SCSI_Controller > >These are, in order, Manufacturer, Product (may be 0x02, use Dave Haynie's >CacheCard Program to check,) Card size, Base Address of ROM, size of ROM. > >I found the size of the ROM by guessing and checking. The system crashed >when I was wrong. Be sure to use the verbose option of setcpu, to be sure >your file was accepted, and always use the cardrom option along with the >Fastrom option. > >Now for the wierd part; I'm not getting _any_ performance improvement. >It's about 382K/sec Reads, and 335K/sec Writes (32KBits at a time) using >DiskSpeed. Has anyone had any better luck? About what should I be getting? >Anyone? Anyone? > >dylan mcnamee >dylan@cs.washington.edu GREAT!! thanks for the CardROM line, I was about 95% of the way to the same goal myself. As far as performance improvements, I'll quote Joe Augenbraun (designed 2091/590 hardware), (actually paraphrase) "Don't expect faster disk access BUT more idle CPU time for other tasks to use during disk transfers." I had had the same thoughts initially, until I talked to Joe. Having the code in 32 bit ram will give the rest of the system more processor use during heavy disk accesses, since the 2091 is DMA. Did I get that right Joe? Stan Fisher - stan@teroach.phx.mcd.mot.com - asuvax!mcdphx!teroach!stan Motorola Microcomputer Division, Tempe, Arizona - Voice (602) 438-3228 Call our User Group BBS "M.E.C.C.A." running Atredes 1.1 @ (602) 893-0804
daveh@cbmvax.commodore.com (Dave Haynie) (03/13/90)
In article <11051@june.cs.washington.edu> dylan@june.cs.washington.edu (Dylan McNamee) writes: >I found the size of the ROM by guessing and checking. The system crashed >when I was wrong. That's not a bad way to find it, actually. You aren't guaranteed to get the whole ROM by randomly guessing, and you aren't guaranteed to not get any registers by running Wack or Metascope to poke around for the ROM. But it is a foregone conclusion that mapping the I/O into RAM will crash or hang the system. >Now for the wierd part; I'm not getting _any_ performance improvement. I didn't get much of a speedup on the 2090A version at first. The 2091 code is supposedly tighter, and alot of it depends on where your bottleneck is -- if you spend all the time waiting for the drive, the code efficiency (what CardROM helps out with) matters less and less. The most striking performance difference in my home machine was when I started running the 2090A in conjunction with 3 and 4 bitplane hires displays. >dylan mcnamee -- Dave Haynie Commodore-Amiga (Systems Engineering) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy Too much of everything is just enough