phil@mit-amt (a phil sohn) (08/26/88)
Has anyone tried to milk a SCSI drive for all it's speed? I would like to get about 400 Kbytes/sec. Does anyone have any suggestions? How low of level do I have to write my code? Does anyone know about using async FileManager calls? How much CPU time is lost when the calls are async? What is the optimal buffer size? We are using it to interface to a 56000 board that we built to do real time sound out put from disk. phil
bryant@oakhill.UUCP (Bryant Wilder) (08/28/88)
In article <2972@mit-amt>, phil@mit-amt (a phil sohn) writes: > > Has anyone tried to milk a SCSI drive for all > it's speed? I would like to get about 400 Kbytes/sec. > Does anyone have any suggestions? How low of level do > I have to write my code? > Does anyone know about using async FileManager calls? How > much CPU time is lost when the calls are async? What is the > optimal buffer size? > We are using it to interface to a 56000 board > that we built to do real time sound out put from disk. > > phil i don't know whether this suggestion will help, but when we were developing the Mac II version of the DSP56000ADS (the Applications Development System hardware for the DSP56001R20 digital signal processor chip), we bypassed the FILEMANAGER in the Mac II operating system to do real time reads/writes to the NuBus cards. using the Aztec C compiler we were able to read/write directly to the Mac II interface card for the DSP56000ADS. it is very simple and uses peek/poke calls similar to basic. it can be done on byte, word, or long word and is probably the fastest you're going to get on the Mac II. hope this helps. bryant wilder motorola dsp operation in austin, texas