Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Logpoint Technologies (LGPT) -- Ignore unavailable to you. Want to Upgrade?

To: kash johal who wrote (615)	5/4/1999 3:34:00 PM
From: docsox	Read Replies (2) \| Respond to of 698

The volume/need for imprecise calculation vastly exceeds the need for precise calculation. How many individuals were affected by the Pentium bug? We don't carry our stock prices to the 15th decimal. So when I build my backyard linear accelerator, I'll be sure to avoid a computer using this technology. Just one man's opinion. David

To: kash johal who wrote (615)	5/4/1999 6:16:00 PM
From: Spyder	Read Replies (1) \| Respond to of 698

Kash, I want to short LGPT also, but the small volumes are not worth my time. I've bookmarked this thread so I can occasionally check back and maybe have a chuckle. I don't blame you at all for being frustrated with the few hypsters on this thread whom don't know accepted boundaries. It's clear from some earlier posts that they are trying to get the word out on this risky stock. I enjoyed reading some of the posts. I'm sure most on the thread are not hypsters but a few are ... full of sound and fury signifying [an imprecise number log(1)] Goodluck to you, Spyder

To: kash johal who wrote (615)	5/5/1999 5:43:00 PM
From: mnispel	Read Replies (2) \| Respond to of 698

The idea that major semiconductor or DSP companies are going to embed LGPT's algorithms into silicon is Laughable at best. The reason is that general purpose CPU's and DSP's need to be precise. And the whole reason that LGPT's algorthms (sic) work is that they are imprecise. This comment combined with the immediate response given it seemed to assume that the thread had discussed this previously, although I've been unable to find it. From the FAQ on LGPT's homepage: "3. What kind of precision can I expect using efp? Is this always going to be the same? Answer: The Soft CoProcessor implementing high performance efp computation in extremely compact (3 to 5 kbytes) software on standard 8-bit microprocessors usually provide 16, 24 or 32 significant bits (equivalent to 5, 7+ or 9+ decimal digits) of precision. Software for the 16-, 32-, and 64-bit microprocessors presently provide 20, 22, or 24 significant bits of precision in a 32-bit data format, using 17 to 55 kbytes of memory. These levels of precision meet most embedded computational needs, but 16, 24, 32, 48, or 64 significant bits of precision at similar or even higher levels of performance can be provided by using substantially larger amounts of memory (1/8 to 1/2 Mbyte). " The response by logpoint (presumeably Dr. S) speaks for itself. One probably can assume that since LGPTs technology is made up in part by look up tables, precision is also in (large) part a function of the tables. More precision requires more bits per element per table and perhaps larger tables (rows and/or columns). BTW, given the press releases we've seen it seems safe to assume as well that the exact memory figures given in the FAQ may be somewhat out of date. Thus, it appears more probable that a lack of precision should be predicated to the comments made in the referenced post rather than to LGPT. Mark N.