Hello folks at PGI:
I am planning to run a large problem, with a single process (better, a single matrix)being larger than 2GB. I have two questions:
On kernel 2.6 series is there no memory limitation address for 64bits machines, right? Is there a limit on the userland process?
And what limits does the pgcc/pgCC/pgf90 have on single process allocation? I understand that the -Mlarge_arrays flag will abilitate the 64bits addressing scheme, so my process can grow larger than 2GB. But this not happened on a Opteron with 8GB of memory, so i think that maybe the kernel is limitating factor.
How can I overcome this? On 2.6 kernels on 32 bits and 64 bits?
Thank you very much!