[PD] [OT] SSE/MMX tips?

[Charles Henry]

On Thu, Sep 8, 2011 at 1:02 AM, Mathieu Bouchard <matju at artengine.ca> wrote:
> On Wed, 7 Sep 2011, Charles Henry wrote:
>>
>> On Wed, Sep 7, 2011 at 7:59 PM, Mathieu Bouchard <matju at artengine.ca>
>> wrote:
>>>
>>> On Wed, 7 Sep 2011, Mathieu Bouchard wrote:
>>>>
>>>> On Wed, 7 Sep 2011, Bill Gribble wrote:
>>>>>
>>>>> So far iteration on plain floats seems to be the best I can come up
>>>>> with,
>>>>> but HADDPS is tantalizingly close to what I want to do.  Any hints?
>>
>> Sorry, what's HADDPS?
>
> http://www.rz.uni-karlsruhe.de/rz/docs/VTune/reference/HADDPS--Packed_Single-FP_Horizontal_Add.htm
>
>> This is really interesting.  Your compiler probably knows how to
>> optimize this kind of information.
>
> How can you tell that ? I bet it doesn't...

Yeah, I thought it over.  I was wrong.
I was also wrong about SSE4.2--AVX is the new instruction set with
256-bit wide operations.

> What could it be doing about a scan like this, anyway ?
fft->multiply by 2*pi*i*f->ifft and fall over...

I dunno, but I'm working on it a bit.