Re: sin(x) performance on x64 and win32
- From: "Victor Bazarov" <v.Abazarov@xxxxxxxxxxxx>
- Date: Wed, 3 May 2006 08:23:10 -0400
Tim Roberts wrote:
"Phil Frisbie, Jr." <phil@xxxxxxxxxxxx> wrote:
satol wrote:
Real-world code that's highly dependent on the performance of
sin(x) is, no doubt, quite rare.
We live in the different worlds apparently. atan2 is 30% slower when
compiled to x64 with VS2005. I just started to play with x64 to see
the difference with win32.
You have real world applications that heavily use sin(x)? In most
applications trig functions amount to less that 2-5% of the actual
CPU usage, so even two times slower trig functions would increase
CPU usage by only 2-5%.
I think you are dramatically underestimating the size of the universe
of scientific computing. There's an awful lot of analytical code in
the world that relies heavily on trig functions.
Yes, an awful lot. Should it really change the fact that optimizing too
soon is a waste of time? Could it be that most of the scientific code
is written by essentially amateurs in programming and it relies heavily
on exactly that, premature optimization, and can be drastically improved
by optimizing the portions that do matter? I am not arguing the ability
of computer scientists inventing and perfecting algorithms to do their
jobs for creating the fastest and the slickest code fragments. I am
talking optimization of the entire application instead of worrying of
a few cpu ticks at the lowest level.
Also, while we're on the subject of estimating, do you have any data to
suggest that in any application (aside from the contrived example by
the OP), and I mean a real world application, the use of trigonometric
functions goes beyond 5%? How far beyond? What is the real difference
in performance of that application in 64 bits vs 32 bits? What is the
source of your information?
It's not the size of the universe that matters. It's what comprises it.
V
--
Please remove capital 'A's when replying by e-mail
I do not respond to top-posted replies, please don't ask
.
- Follow-Ups:
- Re: sin(x) performance on x64 and win32
- From: Tim Roberts
- Re: sin(x) performance on x64 and win32
- From: Simon Trew
- Re: sin(x) performance on x64 and win32
- References:
- sin(x) performance on x64 and win32
- From: satol
- Re: sin(x) performance on x64 and win32
- From: Carl Daniel [VC++ MVP]
- Re: sin(x) performance on x64 and win32
- From: satol
- Re: sin(x) performance on x64 and win32
- From: Phil Frisbie, Jr.
- Re: sin(x) performance on x64 and win32
- From: Tim Roberts
- sin(x) performance on x64 and win32
- Prev by Date: Re: bizarre behaviour: VS2005 debug/release mode
- Next by Date: Re: Strangest VC8 bug yet
- Previous by thread: Re: sin(x) performance on x64 and win32
- Next by thread: Re: sin(x) performance on x64 and win32
- Index(es):
Relevant Pages
|