comp.lang.idl-pvwave archive: archive » Re: Quickest method for calculation

Home » Public Forums » archive » Re: Quickest method for calculation

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: Quickest method for calculation [message #71052 is a reply to message #71051]

Mon, 24 May 2010 12:28

Karl[1]
Messages: 79
Registered: October 2005

Member

On May 24, 1:10 pm, FÖLDY Lajos <fo...@rmki.kfki.hu> wrote:
> On Mon, 24 May 2010, Karl wrote:
>> If your compiler is really good, you may be able to turn on an option
>> that generates SSE code, which may give you another 2x to 3x. The
>> above pattern is an easy one for the compiler to recognize as SIMD-
>> exploitable. If the compiler doesn't do this for you and you have the
>> time, you can write the SSE code yourself.
>
> I think this will have no effect. The bottleneck is memory access, the CPU
> is already starving on data.
>
> regards,
> lajos

Yeah, you are probably right. I did some measurements along these
lines awhile ago and had the same conclusion. It would only help if
the ratio of operations to memory accesses were a lot greater.

Report message to a moderator

[Message index]

		Re: Quickest method for calculation By: jaz on Tue, 25 May 2010 15:53
		Re: Quickest method for calculation By: David Fanning on Mon, 24 May 2010 12:37
		Re: Quickest method for calculation By: Karl[1] on Mon, 24 May 2010 12:28
		Re: Quickest method for calculation By: Foldy Lajos on Mon, 24 May 2010 12:10
		Re: Quickest method for calculation By: David Fanning on Mon, 24 May 2010 09:12
		Re: Quickest method for calculation By: Karl[1] on Mon, 24 May 2010 08:58
		Re: Quickest method for calculation By: David Fanning on Mon, 24 May 2010 07:51
		Re: Quickest method for calculation By: Gray on Mon, 24 May 2010 07:39

Previous Topic:	Landsat HDF
Next Topic:	SAVE again

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Sat Nov 29 00:19:54 PST 2025

Total time taken to generate the page: 0.16066 seconds