Cute 32-bit floatingpoint coprocessor (expensive)

Go To Last Post
7 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Just stumbled across this on another board

http://www.micromegacorp.com/umf...

http://www.micromegacorp.com/dow...

But its expensive , suggested retail $14.95

/Bingo

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Interesting. What do you think it is? A pic? an avr? How many floating point operations per second does it do? Looks like the clk is 1mhz, so it must take 32 usec to load a fp number and another 32 to read it out..... I have measured 60,000 fp ops a sec on a 14.7456 mhz mega128... thats 16usec an operation.... so right off the bat, I'm wondering why do we want it?

Imagecraft compiler user

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

He..He

I dont know why we would want it :lol:

But its cute isn't it :wink:

Well if its just a Pic in disguice , then its a good marketing scam :? :?

/Bingo

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Seems there is some timing here

http://www.micromegacorp.com/dow...

That might not be impressive :oops:

Sorry guyzz

Why the FSCK does they make such a thing :? :?

/Bingo

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
What do you think it is?

SparkFun claims it's a CY8C27143 @ 24MHz.
Quote:
so it must take 32 usec to load a fp number

It actually takes 80us (4x20) due to what are presumably
buffering delays. A FADD takes 115-160us.

Maybe someone should put Bob's routines on a tiny13 or something,
priced at, say, $14.94 (:-)).

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
Why the FSCK does they make such a thing

If you were using a PIC it might be a big improvement.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

My fp bench was the regular old iccavr library routines... nothing tweaked or optimized....

Imagecraft compiler user