as far as I know, HP compiler team should be your natural partner. That team has extensive experience in ia64, and is the official compiler team inside HP for ia64 linux, I think. Tuning for ia64 needs a lot of know how with the microarchitecture (having been part of the ia64 compiler at Intel for some time), imho. Sun
On Tue, Aug 3, 2010 at 6:52 PM, Anton Shterenlikht <me...@bristol.ac.uk> wrote: > Hi > > I'm a lecturer in Solid Mechanics with Mech > Eng Dept at University of Bristol, UK. > My research interests are in computational > mechanics. > > I'm preparing a proposal to EPSRC (Engineering > and Physical Sciences Research Council - a major > UK government funding body for engineering, > including computer science and IT) under > HPC software development call 2010/2011. > Here's a link to the description of this call: > http://www.epsrc.ac.uk/funding/calls/open/hpcsoftwaredev/Pages/default1.aspx > > I propose to develop an optimising compiler > with MPI/OpenMP support for ia64 FreeBSD > platform. Please see my justification at the end. > > PathScale (Christopher Bergstrom) expressed > an interest in my proposal, in principle. > > NAG (Numerical Algorithms Group, nag.co.uk) - a major > UK HPC player, who currently provide user support > on HeCTOR (http://www.hector.ac.uk/), UK national > supercomputer - have been very supportive so far. > > I'm writing to ask if Open64 team would be interested, > in principle, to collaborate on this proposal. > In any case I'd be very grateful for any feedback. > > > Justification: > > ia64 accounts for only 5 system in the top500 > list. This is mainly due to compliers unable > to generate efficient code for VLIW (EPIC). > (Of course, there are other issues to do > with marketing, etc. but this in beyond the point). > > If such compiler, with MPI/OpenMP support, > were available, this would > significantly speed-up HPC on ia64 platform. > > In terms of scientific outputs - clearly > all current HPC codes we use will benefit. > However, some specific coding guidelines for > ia64 might be developed to help the compiler > produce efficient code. > > A specific engineering application, which > will be used to benchmark performance > is a set of multi-scale fracture codes > we have at Bristol. The codes include > a finite element part (floating point) and > a cellular automata part (integer). I'm interested > in simulating very large models, i.e. using > large memories, as well as speeding up the > execution. > > FreeBSD (or, for that matter *BSD systems) > represent at the very least a viable alternative > to linux. Things like better security > and more predictable and systematic development > and release schedule, and more permissible BSD > licence (compared to GPLv3) might make FreeBSD more > attractive for HPC in future. > I think having an alternative to linux > is beneficial to academia and HPC > industry. > > many thanks > anton > > > -- > Anton Shterenlikht > Room 2.6, Queen's Building > Mech Eng Dept > Bristol University > University Walk, Bristol BS8 1TR, UK > Tel: +44 (0)117 331 5944 > Fax: +44 (0)117 929 4423 > > ------------------------------------------------------------------------------ > The Palm PDK Hot Apps Program offers developers who use the > Plug-In Development Kit to bring their C/C++ apps to Palm for a share > of $1 Million in cash or HP Products. Visit us here for more details: > http://p.sf.net/sfu/dev2dev-palm > _______________________________________________ > Open64-devel mailing list > Open64-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/open64-devel > ------------------------------------------------------------------------------ The Palm PDK Hot Apps Program offers developers who use the Plug-In Development Kit to bring their C/C++ apps to Palm for a share of $1 Million in cash or HP Products. Visit us here for more details: http://p.sf.net/sfu/dev2dev-palm _______________________________________________ Open64-devel mailing list Open64-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/open64-devel
