cancel
Showing results for 
Search instead for 
Did you mean: 

Archives Discussions

boxerab
Challenger

GCN: local memory barrier and work group size Part II

A year and a half ago I asked a question here on the forum about the need for local memory barriers for

work groups with local size <= 64 (size of wave front).

The answer at that time was that the compiler would remove these barriers anyways, so they didn't need

to be added to the code, even though the spec requires them.

I can verify now that for Polaris arch, they are needed. This change has been causing a hard crash on my application

when targeting Polaris.

Lesson is to follow the spec Or, re-validate assumptions with every new architecture.

Aaron

0 Likes
1 Solution

Hi Aaron,

Yes, I've got the same confirmation that barriers are still removed by the compiler even on Polaris.

Regards,

View solution in original post

0 Likes
3 Replies
boxerab
Challenger

Correction:  this seems to be a bug in my kernels.  Situation remains unchanged - don't need a local memory barrier

for local work group size <= 64

0 Likes

Hi Aaron,

Yes, I've got the same confirmation that barriers are still removed by the compiler even on Polaris.

Regards,

0 Likes

Thanks!

0 Likes