[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Discuss-gnuradio] SIMD optimizations on FIR filters
From: |
Eric Blossom |
Subject: |
Re: [Discuss-gnuradio] SIMD optimizations on FIR filters |
Date: |
Mon, 14 Jul 2003 14:02:54 -0700 |
User-agent: |
Mutt/1.4i |
On Mon, Jul 14, 2003 at 12:51:18AM +0100, S?rgio Rui Silva wrote:
> I'm using to kinds of FIR filters in my program: SCC and FFF.
> When I start my program I get the message:
>
> >>> gr_fir_SCC: using 3DNow!Ext
>
> but I don't get the message: >>> gr_fir_FFF: using "something".
> Does this mean that the FIR FFF filter is not using any of the
> available SIMD optimizations? From what I saw in the code I
> should be getting such message.
Are you sure you're not getting it? It only prints once.
> Is it preferable to use 3DNow!Ext or 3DNow! instead of SSE?
> (benchmark_dotprod seems to point to that...)
On the Athlons the 3DNow code is a lot faster than SSE. That's why
it's selected when it's available.
> S?rgio Rui Silva
>
> P.S.: I'm running a dual Athlon MP 2400+ processor system and
> the functions has_3dnow() and has_sse() return true...
You could set a breakpoint in gr_fir_sysconfig_x86::create_gr_fir_FFF
Eric