Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!mips!ptimtc!nntp-server.caltech.edu!beyond!andrey From: andrey@beyond.caltech.edu (Andre Yew) Newsgroups: comp.sys.amiga.advocacy Subject: Re: AMIGA Message-ID: Date: 4 May 91 23:42:20 GMT References: <1991Jan10.082327.7378@rice.edu> <1991Jan10.095304.16900@mintaka.lcs.mit.edu> <1991Jan10.151816.13893@rice.edu> <1991Jan10.164423.23644@mintaka.lcs.mit.edu> Sender: news@nntp-server.caltech.edu Distribution: usa Organization: California Institute of Technology, Pasadena Lines: 46 melling@cs.psu.edu (Michael D Mellinger) writes: >In article <1991Jan10.164423.23644@mintaka.lcs.mit.edu> rjc@wookumz.ai.mit.edu (Ray Cromwell) writes: > The DSP in the NeXT is SLOW, and integer only. Even further, the NeXT >It's a 10mip processor. The Amiga 500 is less than 1 mip. Also, >Bzzzzt. The DSP does single precision floating point numbers. >Please, reread your "Don't buy the NeXT it doesn't..." propaganda >published by NCW. The DSP56001 does 24-bit fixed-point arithmetic. Internally, it has 56-bit registers to preserve precision during calculations. It has a 24-bit data bus and therefore outputs 24 bit results. > doesn't seem to share its DSP effectively. On the Amiga, the blitter > chip is shared nicely. Its possible to 'OwnBlitter()' but it won't > lock it completely. On the NeXT you'll usually see 'DSP Already in Use' > I Keep seeing people harp about the fact that the NeXT has a DSP, > but I have not seen it used much, unlike the Amiga in which a huge > majority of software shares the custom chips effectively. The DSP56001 actually has an interesting interface for talking to other processors, the host interface port. It also has pins that can be used to time DMA transfers into and out of its own memory. The host processor can activate one of 32 vectors in the DSP through the HIP and launch the DSP on some task. Perhaps the DSP is much more complicated than the Amiga blitter, such that if you allowed sharing, the context-switches would take forever. I mean, this chip is loaded -- on-chip, hardware stack, pipelines, lots of registers, so saving its state and loading another on might be impractical. Also, the 56001 has a small program memory space (64 KW, where each word is 24 bits long), so you couldn't keep multiple programs as well as the switcher program in memory and would have to DMA the program from outside, after you save its old state outside somewhere, of course. >Don't know what to tell you. You're right. Perhaps there are >technical reasons for it. That's my best guess. >-Mike Andre -- Andre Yew andrey@through.cs.caltech.edu (131.215.131.169)