Re: long FIR filters
Re: long FIR filters
- Subject: Re: long FIR filters
- From: "Cor Jansen" <email@hidden>
- Date: Fri, 2 Feb 2007 19:12:51 +0100
Thanks for the very interesting information.
I would like very little latency.
I was wondering if with the current processor speeds and architecture (SIMD
instructions?), it would be possible
to get good results with a direct form FIR implementation.
E.g: for a 44.1Kc samplerate stereo signal, and with 10.000 taps, I would
need about 880 miljoen multiply-accumulations.
With a 1 GHz processor I will need about one multiply-add per clock.
I suppose the Intel processor of the apple computers has some kind of
Altivec (MMX?) architecture to do several
multiple-adds each clock? Am I correcect?
Does anybody know how to work this out? Preferrable without writing assembly
instructions.
Cor
----- Original Message -----
From: <email@hidden>
To: "coreaudio" <email@hidden>
Sent: Friday, February 02, 2007 6:48 PM
Subject: Re: long FIR filters
Unfortunately this technique has been patented (#6625629)
http://www.google.com/patents?vid=USPAT6625629&id=GgAPAAAAEBAJ&dq=Efficient+Convolution+without+Input-Output+Delay
Only for low-latency partitioned convolution with unequal partition sizes
(and even that is a bit moot). The plain partitioned convolution (which
can still give usefully low latency) is ad lib (see "BruteFIR" for one
example, "pconvolve" opcode in Csound for another).
Richard Dobson
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden