search

lattice / quda

QUDA is a library for performing calculations in lattice QCD on GPUs.
https://lattice.github.io/quda
Other
279 stars 94 forks source link

Develop branch plaquette code and dslash_staggered #367

Closed nmrcardoso closed 8 years ago

nmrcardoso commented 8 years ago

Hi, I tried to compile (cuda 7.5) the last code in the develop branch and it breaks in files: 1) dslash_staggered.cu dslash_core/staggered_dslash_core.h(717): error: no suitable user-defined conversion from "float2" to "float4" exists 2) gauge_plaq.cu ptxas error : Entry function '_ZN4quda11computePlaqILi960EfNS_11FloatNOrderIfLi18ELi4ELi8EEEEEvNS_12GaugePlaqArgIT1_EE' uses too much shared data (0xc320 bytes, 0xc000 max)(...)

mathiaswagner commented 8 years ago

I think Kate ran into the same or a similar gauge_plaq error when compiling for Fermi. For which architecture did you try to compile?

nmrcardoso commented 8 years ago

Hi Mathias, I tried for fermi architecture. Now I've tested also for kepler and maxwell and this happens only for fermi. Do you know if there are a workaround for this?

On Tue, Sep 29, 2015 at 12:44 PM, Mathias Wagner notifications@github.com wrote:

I think Kate ran into the same or a similar gauge_plaq error when compiling for Fermi. For which architecture did you try to compile?

— Reply to this email directly or view it on GitHub https://github.com/lattice/quda/issues/367#issuecomment-144034930.

mathiaswagner commented 8 years ago

I have not yet looked into this. Maybe @mikeaclark has some more insight already. I am not sure whether anyone ever tried to compile develop for Fermi recently with a different CUDA version.

maddyscientist commented 8 years ago

The problem with the plaquette has been fixed in my feature/improved_staggered_rhmc branch. I will look at the other issue later today and make the fix there so it goes in pull #366.

maddyscientist commented 8 years ago

The staggered dslash problem has now been fixed also.

nmrcardoso commented 8 years ago

thanks @mikeaclark. Fixed in #366.