Re: Hello - and where to begin?
"Jeff Fox" <[email protected]>
| Newsgroups | gmane.comp.lang.forth.colorforth |
|---|---|
| Message-ID | <[email protected]> |
> On 20/01/2008, Jeff Fox <[email protected]> wrote: >> Did you follow the details of the synthesis that was demonstrated >> at Forth Day on the SEAforth chips and the performance comparisons >> to Pentium and Yamaha synthesizer chip implementations? > > No, I didn't. Is there anything to which you can refer me? I don't think IntellaSys has posted any pages about it yet. Sam Falvo's report in his blog mentioned what he remembered http://www.falvotech.com/blog/index.php?/archives/200-Forth-Day-Report.html From Sam's page: "Next up was Michael Montvelishsky, who spoke on the topic of music synthesis using the SeaForth processor array. Michael demonstrated with live hardware: one core was dedicated to the 6-bit DAC oversampled and filtered with what appeared to be a trivial R-C low-pass filter, to get 18-bits of equivalent resolution, another 20 cores were dedicated to the synthesis (with 2 cores per voice, there was a total of 10 voices operating inside the chip), and another core or two was dedicated to playback of some music by Bach. With a quick reprogram of the chip, those same cores were responsible for real-time, fractal music composition. The algorithm used on the chip was waveguide synthesis, a technique known for its computational expense; certainly not something you'd expect to find on an 18-bit, integer-only processor that provides only an 18-bit multiply step (not even a full, single-cycle multiply!) and addition as two separate instructions with only 128 bytes equivalent of word-addressed RAM. The fact that it was doing things in real-time just shows you that the claim of 18,000 MIPS is no joke; it is as real as the keyboard I'm typing this with, and the monitor you're using to read it. I was there. I heard it with my own ears. The parameters chosen enabled the chip to accurately reproduce the sound of several nylon, 12-string guitars with absolute ease. And, to think, there were 10 voices available." It was a good example of how a very small, low power, and cheap chip could outperform expensive dedicated special purpose chips as well as expensive general purpose chips like Pentium. Michael played both human composed and machine generated music a Forth Day. Since then he hooked up a midi keyboard as well. The latest ROMs will include the wave generator in a few words of code and will allow more voices to be programmed in the RAMs. There will also be a bigger multiply step. Best Wishes