Code runs 3x times faster on X260 than on tesla c1060

[cbuchner1]

How embarassing, I am about ready to consult an optometrist. ;)

Christian

[Sylvain_Collange]

I heard a whole lecture about CUDA, learnt so much about coalesced mem access, about divergence, about shared memory bank conflicts, but I havent heard anything about these memory partitions. Do you have any literature I can read it up to understand this issue a bit more?

Would be really happy if it is explained somewhere and thanks for the help!

Currently, the official reference on Partition Camping is the whitepaper of the transposeNew sample from the CUDA SDK.

You will also find a long but interesting discussion in [topic=“106924”]topic 106924[/topic]

It would be great to have a discussion about these effects in the CUDA manual or Best Practices Guide…

How embarassing, I am about ready to consult an optometrist. ;)

I should do the same, took me a while to spot the difference too. :)