Cuda SDK Mandelbrot example - HW Double Precission

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,
i have one question about mentioned mandelbrot example:

Do u guys have any problems when selecting hw double precision via context menu? whe i do this i can see no output like it just stopped calculation :( other options work just fine. i got no errors in debug build so i am wondering what can be wrong.

i am using cuda 2.0beta on WinXP with my 9600gt

thanks in advance :D

2

You don’t have a GPU that supports double precision.

3

9600gt doesn’t support dp in hardware.

Paulius

4

thanks, i thought so :(

can ju tell me what current GPUs ( i assume tesla can handle it ) support double precision?

thanks

5

There are no currently available GPUs that support double precision.

6

I wrote the original Mandelbrot sample. I was interested to see that NVidia had added an option for double precision hardware calculation. I have a feeling this was added to compare performance between single precision, emulated double precision and hardware double precision floating point. I am looking forward to seeing the hardware that this code was designed to run on.

-Mark Granger

7

I looked briefly at the sm13 header file. It indeed looked to contain double-precision-emulation code.

8

Ah, that would be me, hope you don’t mind. I also made the code support window resizes.

Perhaps it was a bit premature to release the updated code in the beta SDK! :)