Arg! Twice as slow as Pentium...

Compiler, eh? I wonder if that really is the problem. I should check to see if any optimizations are turned on in Project Builder. Under the normal development build, maybe it turns them off.

Incidentally that test that took the 733Mhz PIII 1.8 seconds when converted to floats took my 500Mhz G4 3.6 seconds as doubles and 3.6 seconds as floats - no change. Odd.

Yeah, it ought to be convertable to Java. It's just a bunch of additions and multiplications.

-Rob
 
Motorola's big claim with the current generation of chips is
"single cycle double precision floating point"

Perhaps that explains something? It doesn't explainthe general suck, but it might explain something.
 
and i don't have much to add to the actual value of this conversation, i will say that i tried running 10.0.4 on an unaccellerated 8600/300 (the high end model!) with 128MB of RAM, 4 MB VRAM, and um, well... it was usable to a point. :) I wouldn't say it crawled, and I just added another 128 of RAM and am going to reinstall X after 10.1 comes out.

I think perhaps they duped us all and spread all this FUD (aka 10-10.0.4) on purpose, just to unleash the horrible windows-eating beast that 10.1 will be and surprise EVERYONE, even the early adopters ;)

uh oh i think i hear Apple lawyers at my window

has anyone tried these same functions on a mac running 9? granted, it won't take advantage of any DP, but it would still be interesting ...
 
Project Builder (gcc) does have different levels of optimization. By default, during "development" the optimization is turned off to allow for debugging.

To change Optimization (in PB), go to the "Project Menu", select "Edit Active Target", change to the "Build Settings" tab, and under the Compiler Settings Group.

<B>Different Subject:</B>
I just got my new 733 G4 at work today and it shipped with 9.2 & 10.0.4. The Software Restore CD is actually 4 CDs and preloads X and 9.2 for a total of about 5 gigs of stuff.
 
I turned on Level 3 optimization - no change. I made some functions inline - no change.

I'm going to extract a piece of the RNG that's just a bunch of "real-world" math work. I'll make it a nice small .c program that we can all compile on the command line from, hopefully, many OS's. Let's see what happens.

-Rob
 
Check out the new thread about a math benchmark in this forum to download, compile, and run a quick test I put together that uses part of the RNG I've been discussing.

-Rob
 
Back
Top