[Omp] First Touch initialization

[Federico Massaioli]

Francisco,

> We have tried it and it works indeed (Intel Fortran compiler v9 on 4-way
> Opteron),
> but I don't understand why.

Array elements are mapped in contiguous addresses in the process address 
space, which is virtual, and has nothing to do with the physical address 
space layout of your process. This is contiguous inside a memory page (4KB 
tipically), but memory pages can be scattered around in whatever order.
The CPU hardware will take care of translating contiguous virtual 
addresses to non contiguous physical ones, transparently to the program.

On NUMA machines, first-touch allocation performed in parallel has the 
effect of allocating each physical memory page on the physical memory bank 
connected to the CPU running the thread that (hopefully, depending on the 
program) will prevalently access the corresponding virtual addresses.

Federico

-----------------------------------------------------------------------
"Even if you don't always actually 'need' the highest precision, you
  can still delight in the joy of fully understanding what's going on."
              Paul Horowitz & Winfield Hill - The Art of Electronics
-----------------------------------------------------------------------
Federico Massaioli                          CASPUR
email: federico.massaioli at caspur.it         V. dei Tizii 6/b
Ph.: +39-06-44486710                        I-00185 Roma Italy
Fax: +39-06-4957083                         www.caspur.it