random memory access and bank conflict

Question

in these days, i'm trying program on mobile gpu(adreno)

the algorithm what i use for image processing has 'randomness' for memory access.

it refers some pixels in 'fixed' range for filtering.

BUT, i cant know exactly which pixel will be referred(depends on image)

as far as i understood. if multiple thread access local memory bank it causes bank conflict. so in my case it should make bank conflict.

MY question: Can i eliminate bank conflict at random memory access?

or can i reduce them?

adreno (qualcomm) has nothing to do with CUDA, removing CUDA tag. — Robert Crovella
– Robert Crovella, Commented May 9, 2015 at 3:10

Christian · Accepted Answer · 2015-05-09 08:03:16Z

0

Assuming that the distances of your randomly accessed pixels is somehow normal distributed, you could think of tiling your image into subimages.

What I mean: instead of working with a (lets say) 1024x1024 image, you might have 4x4 images of size 256x256. Each of them is kept together in memory, so "near" pixel access stays within the same image object. Only the far distance operations need to access different subimages.

A second option: instead of using CLImage objects, try to save your data into an array. The data in the array can be stored in a Z-order curve sorting. This also leads to a reduced spatially distribution (compared to row-order-sorting)

But of course, this depends strongly on your image size.

answered May 9, 2015 at 8:03

Christian

3952 silver badges13 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Jason Newton Over a year ago

the dependency is very strong, infact it often slows down things for larger images - AMD uses this 2d strategy with their CLImage 2d types as does Nvidia. Always benchmark to see if such strategies hurt or help! It may not do what you expect.

Tim Child Over a year ago

I agree that you always benchmark a memory access pattern. But is there any hard data to to support the claim of 'infact it often slows down things for larger images '? What threshold determines the largeness of the image? Or is this just a supposition?

Jason Newton · Accepted Answer · 2015-05-09 03:11:16Z

0

There are a variety of ways to deal with bank conflicts - the size of the elements you are working with, the strides between lines and shifting the coordinates around to different memory addresses. It's never going to be as good as non-random / conflict free though and so what you will notice is depending on the image - you will see significantly different compute times.

See http://cuda-programming.blogspot.com/2013/02/bank-conflicts-in-shared-memory-in-cuda.html

answered May 9, 2015 at 3:11

Jason Newton

1,22110 silver badges13 bronze badges

Collectives™ on Stack Overflow

random memory access and bank conflict

2 Answers 2

2 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related