tenny@z.dec.com (Dave Tenny) (12/10/90)
In trying to choose which language I'll be doing in my NeXT development, one thing I have to ask myself about use of Obj-c is how portable the application will be. (Mine is not a NeXT specific application). If the implementation of Obj-c by GNU means GNU support of Obj-c, then this would imply that the Gnu Obj-c implementations will ultimately end up running everywhere that gcc/g++ do. Is this true? If not, then Obj-c isn't going to be quite as portable (let alone free) as using g++. While I'm here asking about languages, here's one for the NeXT Lispers. While Allegro CL on the old cubes compares favorably with other Lisps on similiar speed platforms, I'm finding it just a bit slow for what I'd hoped to use it for. I can't tell if it's entirely paging (on my 8mb system), or also a general throughput bottleneck on the next. The CPU figures during perceived bottlenecks are generally in the 50-70% range. Can people who really use ACL tell me if first hand experience with 68040 based NeXT machines alleviate the performance problem? How about another 8mb memory? I need to decide which way to allocate my pennies. I guess my real concern is that ACL is big enough, and with code generation characteristics that it may be something which the throughput-mindedness of NeXT machines can't deal with. (For instance, too wide-spread code characteristics for effective cache use, or 100ns memory). (Can some hardware expert tell me if 100ns memory is fast enough for effective use of the 68040? Sounds awful slow to me). Thanks for tips on Lisp of GNu Obj-c directions. Dave
eps@toaster.SFSU.EDU (Eric P. Scott) (12/11/90)
In article <1990Dec10.130953.12476@engage.enet.dec.com> tenny@z.dec.com (Dave Tenny) writes: >If the implementation of Obj-c by GNU means GNU support of Obj-c, >then this would imply that the Gnu Obj-c implementations will ultimately >end up running everywhere that gcc/g++ do. Is this true? The GPL requires NeXT to make their source code available. I'm led to believe that the ObjC extensions will find their way into a future "official" gcc distribution. >While I'm here asking about languages, here's one for the NeXT Lispers. >While Allegro CL on the old cubes compares favorably with other Lisps >on similiar speed platforms, I'm finding it just a bit slow for what >I'd hoped to use it for. I can't tell if it's entirely paging (on my 8mb >system), or also a general throughput bottleneck on the next. You're trying to run cl on an 8MB machine? Good grief! No, it's not a throughput bottleneck, you're RAM starved and paging like crazy. 16MB is reasonable for most things--but CL and Mathematica are PIGS and anyone seriously interested in either of these should consider a 32MB configuration. "No kidding" -=EPS=-
jacob@gore.com (Jacob Gore) (12/15/90)
/ comp.sys.next / tenny@z.dec.com (Dave Tenny) / Dec 10, 1990 / > If the implementation of Obj-c by GNU means GNU support of Obj-c, > then this would imply that the Gnu Obj-c implementations will ultimately > end up running everywhere that gcc/g++ do. Is this true? The 2.0 "gcc" compiler will compile C, C++, and Objective-C (the latter using work done at NeXT). However, there is as yet no run-time support for Objective-C (the messager library) for the GNU distribution. FSF is looking for a volunteer to write one. Jacob -- Jacob Gore Jacob@Gore.Com boulder!gore!jacob
jacob@gore.com (Jacob Gore) (12/15/90)
/ comp.sys.next / jacob@gore.com (Jacob Gore) / Dec 14, 1990 / > there is as yet no run-time support for > Objective-C (the messager library) for the GNU distribution. FSF is > looking for a volunteer to write one. I've just been informed that somebody is already working on the run-time support. Jacob -- Jacob Gore Jacob@Gore.Com boulder!gore!jacob
glang@Autodesk.COM (Gary Lang) (12/16/90)
>I've just been informed that somebody is already working on the run-time >support. Great. At FSF? - g