It’s not clear to me, and I couldn’t find an answer online, whether or not GTX 1050 supports nVidia’s unified memory model, or if PGI community compilers support it on this device. When I compile my OpenACC code using -ta=tesla:managed I get at run time: The accelerator does not match the profile for which this program was compiled.
Hi lmeira83810,
Unified Memory should work on this device. The problem here is that since PGI 2017 defaults to using CUDA 7.5, code targeting a Pascal based GPU, such as your GTX1060, isn’t being generated. Try using “-ta=tesla:cc60,managed” where “cc60” stands for compute capability 6.0.
Hope this helps,
Mat