Re: Fortran compilers for multi-core chips
From: Steve Lionel (Steve.Lionel_at_REMOVEintelME.com)
Date: 12/08/04
- Next message: sarah: "Re: Intel and pgi compiler difference"
- Previous message: Phillip Helbig---remove CLOTHES to reply: "Re: skip an item to WRITE using a FORMAT"
- In reply to: beliavsky_at_aol.com: "Fortran compilers for multi-core chips"
- Next in thread: ejko123_at_yahoo.com: "Re: Fortran compilers for multi-core chips"
- Reply: ejko123_at_yahoo.com: "Re: Fortran compilers for multi-core chips"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Wed, 08 Dec 2004 13:45:22 -0500
On 8 Dec 2004 06:07:31 -0800, beliavsky@aol.com wrote:
>A Wall Street Journal article (page B8, Dec 8 2004) says that Intel
>aims for a tenfold boost in chip performance by 2008 by using dual and
>multi-core chips. An article from EWeek is at
>http://www.eweek.com/article2/0,1759,1737050,00.asp .
>
>Will current Windows and Linux Fortran 95 compilers from vendors like
>Intel, Absoft, and Lahey/Fujitsu be able to exploit the new chips, or
>will one need to buy new versions of them?
>
>Will Fortran programs need to be written differently to maximize speed?
>I have heard of terms like OpenMP and MPI but am not knowledgable about
>them. What is a good book on small-scale parallel programming for
>Fortran programmers?
Let me also recommend "Programming with Hyper-Threading Technology" by Richard
Gerber and Andrew Binstock - Intel Press (www.intel.com/press), ISBN
0-9717861-4-3. While not Fortran oriented, it does discuss many general
concepts that are useful for writing efficient parallel applications in a
tightly-coupled environment.
Intel's compilers already do support multithreaded development, with support
for OpenMP and automatic parallelization. We have a number of things we're
looking at to further exploit multi-core designs. Some designs have
independent caches per core, so they look more like a traditional SMP design.
Others have some form of shared cache or other high-speed communication among
the cores, leading to the NUMA concepts of "local" and "distant" memory.
I was at a "birds of a feather" session on development environments last month
at Supercomputing, and the host asked the audience which programming model
they wanted to use for multicore. Nearly everyone wanted to treat multicore
the same as multiprocessor, recognizing that programming for a NUMA
environment is difficult and not always rewarding. (I recall that, while I was
with DEC, we put NUMA features into our Tru64 UNIX Fortran compiler to take
advantage of the new AlphaServer GS series, but couldn't get anyone to even
try it out. Everyone wanted to treat the system as a traditional SMP. So we
never released the feature.)
Steve Lionel
Software Products Division
Intel Corporation
Nashua, NH
User communities for Intel Software Development Products
http://softwareforums.intel.com/
Intel Fortran Support
http://developer.intel.com/software/products/support/
- Next message: sarah: "Re: Intel and pgi compiler difference"
- Previous message: Phillip Helbig---remove CLOTHES to reply: "Re: skip an item to WRITE using a FORMAT"
- In reply to: beliavsky_at_aol.com: "Fortran compilers for multi-core chips"
- Next in thread: ejko123_at_yahoo.com: "Re: Fortran compilers for multi-core chips"
- Reply: ejko123_at_yahoo.com: "Re: Fortran compilers for multi-core chips"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|