float to int conversions

Chris Townsend ctownsend at arboretum.com
Mon Jun 8 16:23:05 EDT 1998

At 11:48 AM 6/8/98 -0700, you wrote:
>Hi Chris,
>Check out the Intel Signal Processing Library which contains a number of
>float to int conversion routines.  You can specify features such as
>saturation and rounding control.  Much faster than doing the saturation
>control and float to int in C.

I'm really trying to avoid using the Intel Signal Processing library in
this case for a number of reasons.  First I would like to maintain cross
platform compatibility if possible, although I've used the ISPL for other
things and it wasn't too difficult to have two versions of code.  More
importantly the ISPL is designed to process large blocks of data at a time,
yet I find it much more convenient to process data one sample at a time in
most cases, and we'd need to do a significant rewrite of our existing code
to do it this way.  I have been thinking about using the ISPL float to int
conversion function to process just one sample at a time, but I'm afraid
the overhead of the function call etc. will be too high.  If I can't find
any other solution I will certainly give it a try.


Chris Townsend
DSP Engineer
Arboretum Systems, Inc.
"Hyperprism is a wonderful invention" - Brian Eno

More information about the music-dsp mailing list