-
Going forward we are no longer going to support the Tesla architecture (sm_1x). There are multiple hacks throughout QUDA that have been necessary to keep Tesla support on life support. E.g.,
``` c+…
-
During compilation in a local cluster in Cyprus using Fermi I came accross with this error:
make[1]: se ingresa al directorio `/fhgfs/prometheus/home/avaquero/git/test/quda-develop/lib'
/fhgfs/promet…
-
Aof now, QUDA is mostly oriented towards a CPU-GPU interface, where the data is read from CPU, used to compute something in the GPU and then written back in CPU. There are, though, some applications t…
-
While not really relevant for users it might be nice in the development process to have QUDA build its object files in a build directory. One motivation for that is to have separate build dirs for dif…
-
This is a remnant from before QUDA went C++.
-
Due to the number of actions and tasks that QUDA now supports, the time required to compile the code from scratch has become pretty significant. Oftentimes, particularly when debugging, we only want t…
-
Not just for the components as right now to get a total time the application spend in QUDA and the total time spend in compute, H2D, ...
-
Looking at the object files generated by quda this turns out to be among the bigger ones.
See (only files larger than 1MB)
```
-rw-rw-r-- 1 mathias mathias 32M Oct 5 03:03 copy_color_spinor.o
-rw-…
-
the official site http://lattice.github.io/quda/ redirects now to the develop branch (instead of master, as it should).
-
Just for curiosity, I run a test and after a minor tweaks I has able to successfully run it.
Problems running in ARM7:
-> -m32 flags must be removed
-> cudaHostRegister() does not work, in cuda 6.5 to…