tlrmchlsmth

Large Allocation Makes Kernel Run Slowly

Discussion created by tlrmchlsmth on Jun 15, 2010
Latest reply on Jun 15, 2010 by MicahVillmow

I'm using a 5870 with Win7 and the ATI StreamSDK 2.1

I wrote a matrix multiplication kernel, and if I declare a 64x64 matrix of floats, my kernel takes approximately 3 times longer, even if I don't touch the array at all.  Does anybody know why this is?

Outcomes