Hi Folks
I can give you some inputs on Log Point. I was one of the original founders. I left the company in December 94 and have had almost no contact with the management since then so I am not up to date on the business aspects of the company. The company has the potential of becoming the dominant math computing technnology. See their web site logpoint.com to get some ideas as to how much better their performance is than conventional floating point. If they could get the industry to adopt their technology Log Point would be one of Silicon Valley's "great" companies.
The technoligy was developed by Les Pickett. Les is one of the more brilliant mathmeticians in the world when it comes to digital computing. He has been working on the technology for many years and has written his won patents, not a small task in itself. Les and I were room mates in 1963-64 and he was the brightest guy I had met at the time and maby still is.
One of the other Co Founders Sam Shanks, now CEO, is a PHD in math and the recipient for the highest award ever given by NASA. Sam re-designed the Space Shuttle main engine controls. Had he not done so the Space Shuttle may never have flown. That probably isn't true but they would have been held up for a long time solvintg the problem.
The Log Point technology is based on using look up tables to produce the log of a number and then using the log and the original number for all of your computation. One dramatic example of the improvement in performance is taking the square root of a number. From what I remember taking the log of a number takes about 8,000 instrucions in floating point. But if you put the Log of the number you want to take the square root of in a register you can take the square root by just shifting the log by one digit. There are other examples for functions like multiply, divide, square, trig and other functions.... all math funcions except add and subtract see varying performance improvements.
With all the work going on in imbedded systems today there is a real need for Log Points technology. Some of the more obvious applications are:
3D Graphics, Engine Controls, Motor Controls, Non Skid Braking, Dynamic Suspension, Anti Collission, Auto matic pilot for cars... There are all kinds of applications in the military where there is a need for very fast math and the new Communications technology called wavelength division multiplexing using DSP could be used in every cell phone in the future. Log Point would be ideal for this application.
The technique Les used for converting floating point to its Log using look up tables wouldn't be unique if there was a one to one relationship between the input argument and the outrput but he figured out how to use a look up table with only 1/2 the number of digits of the input argument that means if you wanted to convert a 32 bit floating point argument into its log you would only need a 16 bit table.
The problem the company has had is taking the technology to market. Most of the users of math computation are like the scientists that read Albert Einstein's Special Theory of Relativity they just don't get it. It's a bit hard to grasp the technology and difficult to change the methods the programmers have been working with for years. None of us like to change something we are comfortable with. But one day the computing and math world will know the technology as "Picket" Math" and it will more than likely be used for all computation except those problems that require more than 64 bits of precision. There is ultimately a limit to the size of the tables that can be used. If you need precisions of 128 bits or more I don't think Log Point is the answer, of course Les may have figured out how to advance the precision by now. But really high precision doesn't represent more that one or two percent of the marketplace and these are probably math computation that do millions of iterations that must have very high resolution where speed isn't the predominent need.
The conversion tables are just one aspect of the technology. I think Les now has at least four issued patents and probably another three to four in patent offices for approval. Patents are what is needed for a small company to protect itself from the big guys who would not like to see their markets erode.
The problems the company has had so far is the financing and marketing and management that understood how to go about getting the technogy accepted as "THE STANDARD". With the right financing and personnel it is, my estimation, that Log Point could be the next Intel.
I hope this providides you some background and good luck with your investments.
Dave |
