Newsgroups: comp.sys.cbm Path: utzoo!utgpu!jarvis.csri.toronto.edu!godzilla.ele.toronto.edu!leblanc From: leblanc@godzilla.ele.toronto.edu (Marcel LeBlanc) Subject: Re: Burst Mode on C-64 ? Message-ID: <8806010545.AA22977@godzilla.ele.toronto.edu> Summary: High bit rates through software Organization: EECG, University of Toronto References: <1988May27.175829.5648@ziebmef.uucp> <4241@killer.UUCP> Date: Wed, 1 Jun 88 00:25:16 EDT In article <4241@killer.UUCP> elg@killer.UUCP (Eric Green) writes: >in article <1988May27.175829.5648@ziebmef.uucp>, ross@ziebmef.uucp (Ross Ridge) says: >> On a local BBS we had a discussion about burst mode on the 64. We thought the >> biggest problem would be be the hardware modifications, but our goal was to >> be able to do it through software only. The CIA's (6526) serial port *could* >> be emulated through software (side note: the 64 can go as high ar 4800 baud >> or 7200 half-duplex if you completely re-write the RS232 routines) however >> burst mode used serial line that wasn't connected on the 64... > >Note that bit-rate for each byte during burst mode is 250,000 baud, not 4800 >baud. Burst mode is FAST. With the 1581, which reads data off disk at a >reasonable rate, the 128 is almost as useful as a real computer. 8-) It IS a real computer ! 8-) ^^^^^^^^^^^^^ >You'll never do it in software. In particular, the 1581, in fast mode or burst >mode, clocks data out at 250,000 baud. No software in the world is going to be >able to sample fast enough to cope with that. Hardware is necessary. Period. I agree that software can't compete with hardware for raw speed, but it is indeed possible to transfer data at very high rates over the commodore serial bus (even at BURST speeds!). The first thing thing that must be pointed out is that the 250,000 bps BURST transfer rate is nowhere near the effective file transfer rate using present commodore burst software. Using the commodore burst protocol, the CLK bit must be toggled (by software) after each byte transfer. During this time, the bus is idle. The result is a very short period of very high speed transmission (for a single byte), followed by a relatively longer idle period where only the CLK bit is toggled. Also consider that although the 1581 is very fast because it makes use of a disk cache, some time is required to read each half track into the cache, then transfer sectors from the cache to the file buffers before each sector is transmitted to the C128 (assuming LOAD operation). It turns out that the effective transfer rate for LOAD operations using a 1581 is only about 50,000 bps, or 6Kbytes/sec (still much higher than 4800 baud). The point of this posting is to point out that the SAME speed or faster can be achieved on a vanilla C64 using only special software. The 1581 turbo LOAD routines on Super Snapshot V3.0 (using a C64) are the same speed as BURST loading with a C128 & 1581 (turbo SAVE is faster with SS V3.0 than with C128 BURST save). In more concrete terms, this means that you can load a 202 block file on a C128 or C64+SS(V3.0) from a 1581 in 9 seconds. ... >If you are doing true burst mode, and not just a software fastload/fastsave >routine. The C128 kernal only supports BURST mode for loads and saves. For other types of transfers, either fast serial or slow serial transfers are supported (in the case of the 1581, this will always be fast serial). I don't find that fast serial is a significant speed improvement over slow serial. It doesn't even come close to the speed of BURST transfers. Although other specialized types of BURST transfers are supported, these require custom software. > Eric Lee Green {cuae2,ihnp4}!killer!elg Marcel A. LeBlanc University of Toronto -- Toronto, Canada also: LMS Technologies Ltd, Fredericton, NB, Canada CSNET: leblanc@godzilla.ele.toronto.edu CDNNET: <...>.toronto.cdn UUCP: {decvax,ihnp4,linus,utzoo,uw-beaver}!utcsri!godzilla.ele!leblanc ARPA: leblanc%godzilla.ele.toronto.edu@relay.cs.net BITNET: leblanc@godzilla.ele.utoronto (may not work from all sites)