phd_ivo@gsbacd.uchicago.edu (07/15/90)
>Also, even though the 56001 is a relatively inexpensive chip (<$50), it's >still much cheaper for a company to include it with their motherboard than >to add it in as a separate processor board later. What really startles me is how little general-purpose software uses the DSP. I am not interested in music or electronics. IMHO, NeXT should bundle a simple speech recognition system, so that Informix can have businessmen talk to their spreadsheets. Can you say "Up-down-1.24-enter-1.23-next-help"? This would make businesspeople buy a NeXT. Not the fact that IB is neat, or that the interface is better or worse than a Mac. Something that is substantially different, new and unique to every package to be written on NeXT. Why isn't there such a novelty, yet? On the other hand, many academics buy a NeXT because it's the best workstation for the money (particularly after the 68040 upgrade). /ivo welch ivo@next.agsm.ucla.edu
declan@portia.Stanford.EDU (Declan McCullagh) (07/16/90)
Ivo Welch (ivo@next.agsm.ucla.edu) writes: >What really startles me is how little general-purpose software uses the DSP. I'd say the problem is that at least in the current version of the operating system, the DSP can't be shared - it's a "single user" chip. So if you're using it for some giant nifty sound manipulation or FFTs, and a spreadsheet wants to use it, that's too bad. So far, I don't really see that as a problem, since if a lot of NeXT software wants to use the DSP, you'll have some conflict over it... >I am not interested in music or electronics. IMHO, NeXT should bundle a simple >speech recognition system, so that Informix can have businessmen talk to their >spreadsheets. Can you say "Up-down-1.24-enter-1.23-next-help"? Have you seen the Sphinx spreadsheet demo? It does exactly that (albeit a bit slowly)... >This would make businesspeople buy a NeXT. I'm not so sure. Didn't a PC made by Texas Instruments have some simple voice recognition software bundled with it - about five years ago? It didn't seem to be selling too well... >On the other hand, many academics buy a NeXT because it's the best workstation >for the money (particularly after the 68040 upgrade). Exactly! It would be wonderful to find a (high-end) niche for NeXT to target, something that would allow them to move into the corporate world through the legendary back door... ... Carlos Salinas (carlos@eeyore.caltech.edu) writes: >From what I've read about Unix's scheduling algorithm, CPU intensive >activities are scheduled with a low priority, I/O intensive activities are >scheduled with a higher priority. Why? To increase the responsiveness of a >terminal login (I/O intensive.) At first, Unix systems were assumed to have very little physical memory and fast hard drives, which led to this I/O intensive approach to things (early PDP-11s supported a maximum of only 248K of main memory). Also, since the Vax systems 4BSD was developed on were CPU limited but with excess I/O capacity, this approach remained viable. Now, the NeXT needs a perky, responsive interface, and that may require a substantially different approach to scheduling... After all, Unix wasn't really designed to support a graphics interface. >Luckily NeXT is not entirely Unix, the core system is Mach. This leaves the >door open for better schedulers. A few messages ago, Robert Lin posted that NextStep v2.0 might have a substantially revised scheduler, so we can at least hope... -Declan ------------------------------------------------------------------------------ Olympic Technologies / Registered NeXT Developers \ declan@portia.stanford.edu ------------------------------------------------------------------------------
daugher@cs.tamu.edu (Dr. Walter C. Daugherity) (07/17/90)
In article <1990Jul15.202600.17995@portia.Stanford.EDU> declan@portia.Stanford.EDU (Declan McCullagh) writes: > >At first, Unix systems were assumed to have very little physical memory and >fast hard drives, which led to this I/O intensive approach to things (early >PDP-11s supported a maximum of only 248K of main memory). Don't you mean 28K 16-bit words (with the last 4KW for memory-mapped I/O)? ------------------------------------------------------------------------------- Walter C. Daugherity Internet, NeXTmail: daugher@cs.tamu.edu Knowledge Systems Research Center uucp: uunet!cs.tamu.edu!daugher Texas A & M University BITNET: DAUGHER@TAMVENUS College Station, TX 77843-3112 CSNET: daugher%cs.tamu.edu@RELAY.CS.NET ---Not an official document of Texas A&M---
garton@cunixa.cc.columbia.edu (Bradford Garton) (07/17/90)
In article <1990Jul15.202600.17995@portia.Stanford.EDU> declan@portia.Stanford.EDU (Declan McCullagh) writes: >Ivo Welch (ivo@next.agsm.ucla.edu) writes: > >>What really startles me is how little general-purpose software uses the DSP. > >I'd say the problem is that at least in the current version of the operating >system, the DSP can't be shared - it's a "single user" chip. I think another big problem with the current OS is the lack of DMA. Hardly makes the overhead of using it worthwhile for tasks which cannot be run entirely on the DSP. Rumor has it that DSP DMA will be in place for 2.0, however. Brad Garton Columbia University brad@woof.columbia.edu