OpenCL DemoFractal on NVIDIA compile error

Started by MikOfClassX, March 30, 2011, 07:44:03

Previous topic - Next topic

MikOfClassX

Win7-64, NVidia 267.59

OpenCL COMPILER OPTIONS: -D USE_TEXTURE -D DOUBLE_FP
BUILD LOG: <program source>:74:53: error: can't convert between vector values of different size ('float4' and 'double')
            write_imagef(output, (int2)(ix, iy), oc / _255);

But works perfectly with doublePrecision=false (runs at 173 fps)

So, looks like the isDoubleFPAvailable() gives TRUE but it's not really supported by hardware.. I'm confused.

Cheers,

Mik

spasi

Could you grab Mandelbrot.cl from SVN and see if it's fixed now?

MikOfClassX

Yes, it is! 58FPS in double precision.. Well done!

I'm a bit confused about the speed of OpenCL. I was expecting different numbers..

Cheers!