Re: Speeding up XCode?
Re: Speeding up XCode?
- Subject: Re: Speeding up XCode?
- From: Steve Checkoway <email@hidden>
- Date: Wed, 19 Oct 2005 05:09:36 -0700
On Oct 18, 2005, at 2:57 PM, Robert Dell wrote:
let's take this one step at a time.
1) code that doesn't branch takes less time than code that branches
2) calling functions invoke branches in the code
3) my code doesn't call other routines (other than built-in math
and locating the position of a certain index in an unknown location)
Unless, of course, your branchless code is so long that it doesn't
fit into cache or it knocks something else important out of cache.
Then again, with all of the indirect loads through arrays and
structs, it's just as likely that your data is missing in the cache
so those 17 instructions could take possibly hundreds of cycles
depending how far back in the cache you missed.
With deep, super-scalar, out-of-order pipelines, it's pretty
difficult to tell just what the performance of any given sequence of
machine instructions will actually be _before_ you consider the
memory hierarchy. This is not like the MIPS R2000 where you can guess
more or less what happens as each instruction travels down the
classic 5 stage pipeline.
With simple functions to make the code easier to read and maintain,
you say that this will decrease performance. It might, that's true.
Or the compiler might notice that you are calling a leaf procedure
and decide to inline everything.
I could continue but there are enough ifs and mights and the like in
this already. All I'm getting at with this is, you just never know
what is going to happen (especially if your code is run on multiple
generations of hardware).
- Steve
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Xcode-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden