home *** CD-ROM | disk | FTP | other *** search
- > TI's DSP handle _FLOAT_ numbers, in 40 bits. 40 bits are few for floating
- > numbers, but they are better than 48 bits for fixed numbers.
-
- That depends on what you want to do with it of course.
- In many signal processing cases the fixed point could be preferable.
-
- > I've read in the PFW that TI's DSP has got two specials instructions, to
- > compute inverse (1/x) and square root (x^0.5). These instructions are
- > _VERY_ important, and you can normalize a vector with the square root.
-
- Such instructions can indeed be very useful. They are likely to have rather
- long latency, though.
-
- > Yann Lossouarn said TI's DSP can do 8 instructions in 1 cycle. I don't
-
- He actually said 'operations per cycle', which definitely is not the same
- as 'instructions per cycle' when it comes to DSP marketing.
-
- > I think there is a Motorola DSP56002 DSP, that handles floating numbers,
-
- Motorolas floating point (32 bit) DSP is the 96002.
- The 56002 is just an enhanced 56001, running at up to 80MHz.
-
- --
- Chalmers University | Why are these | e-mail: rand@cd.chalmers.se
- of Technology | .signatures | johan@rand.thn.htu.se
- | so hard to do | WWW/ftp: rand.thn.htu.se
- Gothenburg, Sweden | well? | (MGIFv5, QLem, BAD MOOD)
-
-