rbj@icst-cmr (root) (04/28/86)
> Why do so many of the micro operating systems > insist on putting arguments in registers!!!. Even MS-DOS expects > arguments in registers rather than on the stack. Surprise! Even UNIX (V6) does this! The reason is that the machine generally uses the same regs in supervisor mode. This is easier than trying to did the args out of the user's space, which entails accessing another map entry. BSD seems to use the stack however. (Root Boy) Jim Cottrell <rbj@cmr> "One man gathers what another man spills"
rbj@icst-cmr (Root Boy Jim) (05/20/86)
>I don't want to quote too much of the original article -- it was rather long. >The thrust of it was that passing parameters in registers doesn't make things >run faster. This does not fit with my experience. When we implemented >parameter passing in registers at Zilog, it was a big win. *WAS* a big win - meaning maybe it looked good on benchmarks... [...] >I will admit that this parameter passing caused difficulties. In particular, >varargs code was messy, and debuggers more difficult to implement. Was it >worth it? I believe so. Our emphasis was on speed, and that we had. 1) pdp-11 style varargs was broken - no way from C to get a those arguments that got crammed into registers. The key phrase here is `pdp-11 style'. No way from C? Just use the variable names! The compiler knows about them. 2) Because of the register assignment scheme used, you couldn't even blindly pass a list of variable arguments to a routine like printf that did know how to deal with them. Why not? And who says varargs has to be written in C? It just may have to be written in assembly some day. 3) setjmp/longjmp got broken with respect to restoring register variable parameters somewhere along the way. Probably difficult, but not impossible. Pyramids seem to work. In the long run, portability seems to have become the more important issue: Just try porting some of the code in mod.sources and net.sources to the machine and see how far you can get without having to go into the editor and start whacking at source code. Assume that software developers will have the same problems porting their wares... Agreed. No one said it would be easy. But then, if it wasn't just a little bit hard, they'd hire your kid sister instead of you. Also, note that Zilog has never delivered its ATT SVID blessed System V port, and that their next series of machines will contain ATT chips running vanilla 3Bx software... Zilog of course has other problems as well. George Robbins - now working with, uucp: {ihnp4|seismo|caip}!cbmvax!grr but no way officially representing arpa: cbmvax!grr@seismo.css.GOV Commodore, Engineering Department fone: 215-431-9255 (only by moonlite) (Root Boy) Jim Cottrell <rbj@cmr> "One man gathers what another man spills"