Welcome to Fractal Forums

Here's a crazy but true observation about Mandelbulb3D;  It seems to actually run more accurately on a Mac, using Crossover (I believe Crossover is WINE Based - but regardless of that, it's a program that allows certain Windows programs to run on a Mac). 

I was running an animation test, increasing my variable by 0.01 with each frame.  On both of my Windows computers it quickly slipped and the values became X.XX99999 and remained there, never evening out to X.XX0000.  But on the Mac, the increments stayed at a "Clean" 0.01   

Curious to say the least!  Images took longer to render on Windows too...

Just Bizarre!

Anybody have any ideas why that should be?

What Mac were you using?

Intel Macs should produce the exact same results if running Windows software on them (assuming the code is running native on the CPU with Windows environment support from your Wine based system.)

If you are using an older PowerPC Mac, then results might differ because of the different FPU in the processor.  Intel has always used 80 bit values to store intermediate values for floating point calculations.  This means that an Intel CPU's rounding error for floating point will be different than other 64 bit CPUs (like PowerPC) which use 64 bit intermediaries when doing 64 bit (double) floating point operations if that makes any sense.

This 80 bit intermediate step doesn't make the math more accurate, if anything it makes the results more unpredictable and less suitable to science because values can vary depending on what order the CPU executes sequences of instructions.

I don't know if something like this could be the cause, it is just the first thing that popped into my head.

Here is some more info:  http://en.allexperts.com/e/s/ss/sse2.htm

I'm using a Macbook Pro running OSX 10.5.2 with a 2.4 GHz Intel Core Duo Processor.  I thought it had a pretty Animal Name for the OS, but can't find reference to it in my "About This Mac" info thing, so maybe it's Breedless?

I'm shamefully unaware of all this modern computer lingo (I learned to program in the 1970's punching cards in Fortran - unless Tandy's TI Basic counts - then a few years earlier!).  Then I took better than 20 years off of using computers at all, and only got myself one and went online in 2002!

Missed the whole introduction of Object Oriented Programming  :/  I've been trying to fill n the gaps, but my "Old Ways" just wont let go and learning the "New Ways" hurts my brain...

So I may have misused the term "accurately" there referring to calculations.  I was only referring to the slip to X.XX9999 vs staying squarely on X.XX0000 thing.  I found it quite surprising! 

I can only assume it's running natively on the CPU, but I'm not even sure how else it could run  (I don't think I have a GPU, and am not even sure how to tell!!).

I think it is just a matter of the rounding of the double-float value for showing, intern it should make no difference.