bernie@metapro.DIALix.oz.au (Bernd Felsche) (02/23/91)
In <8651@gollum.twg.com> david@twg.com (David S. Herron) writes: >Yes, but ... comparing OS architectures says that AmigaDOS will get more >out of any CPU than any Unix-y system. That is .. in both Unix and Mach >the user process scribbling on the screen is isolated in its own virtual >address space away from the virtual address space where the screen >hardware is location. Ergo, any data to be scribbled onto the screen >has to be _copied_ between (at least) two address spaces. This, >all by itself, is quite a CPU drain. WRONG! It doesn't _have_ to be copied by the CPU. There are options provided by the supporting hardware, such as DMA (if it's in the blitter's address range), or more flexibly, mapping the address space concerned into the address using the MMU (which is even faster, but somewhat limited). UNIX raw device I/O allows for transfer of data directly from a user-defined buffer, to the device controller (as long as the device driver can handle it). True, there is a very small performance penalty, but IMHO it's a price worth paying to prevent system crashes (aka "Software Errors" aka GURUs)! I'd love to be able to run AmigaDOS as a UNIX Process. A GURU would only cause a core dump. And I could use all the reasonably priced AmigaDOS software at the same time as UNIX. -- Bernd Felsche, _--_|\ #include <std/disclaimer.h> Metapro Systems, / sale \ Fax: +61 9 472 3337 328 Albany Highway, \_.--._/ Phone: +61 9 362 9355 Victoria Park, Western Australia v Email: bernie@metapro.DIALix.oz.au