Nathan, I'd agree, there probably isn't as much emphasis on teaching
optimization these days. Back when we had to make an Apple II do
miracles there was a lot more need for everything to be optimal. Now
with > 1 Ghz CPUs the range of applications that need optimization is
very small (3D graphics being one of the few).
As you point out, the concern now is generating working code at a
reasonable cost. This is one reason we use Java 3D. It's more cost
effective than coding right to OpenGL or DirectX. For our purposes we
don't need top performance, we don't need the latest special effects.
We need basic 3D that can be coded quickly (and thus cheaply).
This is why I shake my head at those arguing for blistering fast
performance from Java 3D and support for all the latest hardware
features. Sigh... if you want that then use OpenGL and DirectX.
Like Justin pointed out there are too few of us that have hardware
design backgrounds (Electrical Engineering) AND programming (to provide
indepth understanding of performance optimizations). But also there
aren't enough of us with management experience AND programming (to
provide indepth understanding of budget optimization).
- John Wright
Starfire Research
Nathan Bower wrote:
>
> Justin,
>
> > You'd be surprised about how much programmers don't understand many
> > really fundamental things about computer programming. That is, they are
> > never taught (and probably rightly so) this like about how a for loop
> > gets converted to a set of CPU op-codes and then what those opcodes
> > might do in hardware (for example, how many programmers on this list
> > understand the concept of optimal path programming in 3D graphics?).
>
> I believe that the reason for us not being taught issues such as this
> (although we did go into the usual search algorithm performance thing at
> uni) is that it simply doesn't matter for most applications. People seem
> more bothered with good design (50% of all projects fail / over-run / go
> over-budget), and formally proving that a program matches a specification.
>
> I agree that this sort of thing is useful in situations such as HPC and 3D.
>
> My guess as to why the first loop is faster is that it has something to do
> with either the way cache lines are loaded, or that the compiler will
> generate a looper that is optimal for 99% of the time (i.e. while condition
> is not yet met) rather than the other way around. Wouldn't a good compiler
> catch this sort of thing anyway though?
>
> Eagerly awaiting your answers,
>
> Nathan
>
> ===========================================================================
> To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
> of the message "signoff JAVA3D-INTEREST". For general help, send email to
> [EMAIL PROTECTED] and include in the body of the message "help".
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA3D-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
