Hello all. The company I work for is planning on buying a 4-way AMD opteron based system with about 8GiB of RAM. It'll be running SuSE Enterprise Server 8 for x86-64. We currently have have some scientific code written in Fortran77 that processes data stored in a couple of very large statically allocated arrays. A couple of weeks back I had an opertunity to test some of our code on an 8GiB opteron system, and had problems when trying to compile the code when the static arrays grew above 2GiB in size. I can't remember the exact error but is an assembler error complaining that "<large negative number> is an invalid array size" or something similar. The gcc in question was 3.3. What I'm wondering is, if it is possible, how do I get an environment whereby the f77 code can allocate huge static arrays (6GiB+)? I'm quite happy to recompile the kernel, binutils, glibc, gcc etc. I just want to check that this can actually be done under Linux. I'd very much appreciate any help / advice that anyone can give on this matter. Cheers. -- Chris Sykes <chris.sykes@xxxxxxxxxxxxxx> Sysadmin/Programmer - OHM Ltd. UK. http://www.ohmsurveys.com Phone: +44 (0)870 429 6594 Fax: +44 (0)870 429 6582
