[LAD] Denormals / subnormals (again)

Carl Hetherington lists at carlh.net
Tue Jan 3 17:40:22 UTC 2012


On Mon, 2 Jan 2012, Gabriel M. Beddingfield wrote:

> On 01/02/2012 07:22 PM, Carl Hetherington wrote:
>> If you compile your code (e.g. a plugin) without -msse and -mfpmath=sse
>> on the GCC command line you get *no* protection from denormals from the
>> CPU. If they occur in your code, they will be very much slower than
>
> -ffast-math doesn't work?

It does indeed help, in some cases; thanks for pointing that out.

Current results are somewhat haphazardly compiled at
http://carlh.net/software/denormals/

>> If you compile your code with -msse and -mfpmath=sse, you have Ardour's
>> protection from denormals. If the user's CPU supports it, you get there
>> is no significant slowdown with denormals using this mode. However CPU
>> support is "some later processors with SSE2", according to Intel.
>
> Doesn't that translate to something like: Pentium 4, Pentium M, or Later.

I'm not entirely sure ...

Thanks

Carl



More information about the Linux-audio-dev mailing list