DragonFly kernel List (threaded) for 2007-06
DragonFly BSD
DragonFly kernel List (threaded) for 2007-06
[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]

Re: implemented features (Re: Decision time....)


From: Oliver Fromme <check+jj5pa700rspr9xfw@xxxxxxxxxx>
Date: 05 Jun 2007 10:58:05 GMT

Michel Talon wrote:
 > Rahul Siddharthan wrote:
 > > Where is the generalisation?  I said "my own code", which is what I'm
 > > interested in.  But it also appears to be true of GSL (the GNU
 > > scientific library), at least the parts that I'm interested in.  I
 > > don't think, even if my claim was made "generally", it would be at all
 > > controversial.  Floating-point performance is much better in 64-bit
 > > mode, and there are twice as many general-purpose registers available.
 > 
 > I can confirm what Rahul is saying, perhaps this is no suprise since we are
 > both theoretical physicists. I see 64 bits machines   being *twice* faster 
 > than 32 bits machines (same machine, one in 32 bits Linux, the other in
 > 64 bits Linux) on my computations, and particularly in symbolic maths
 > computations, e.g. running maple 32 bits and maple 64 bits, or floating
 > points computations. With this experience, i don't give any credence to the
 > computer people who pretend that there is no difference between the two
 > modes, or that you need >4 Gigs memory to see the difference. Our machines
 > have 2 Gigs and i see an enormous difference. 

I'm not a theoretical physicist, but I can confirm what you
are saying.  :-)   On certain algorithms amd64 machines are
considerably faster than i386 machines.  For example I have
written a small Sudoku solver which is significantly faster
when compiled and run on an amd64 machine.  I also believe
that it's because of the greater number of CPU registers
available in "long mode" of amd64 processors.  (The machine
in question only has 1 GB of RAM, and the Sudoku processes
have a virtual size of only a few MB even, so the 4GB limit
is completely irrelevant here.)

It is also a popular fallacy that you need > 4 GB of RAM to
be able to use the larger address space of a 64bit machine.
That's wrong.  The virtual address space of a process is
limited by the width of a pointer (which is 32 bit in i386
mode and 64 bit in amd64 mode), not by the amount of RAM.
You can have processes that are 8 GB in size on a machine
with 2 GB of RAM.  Only in 64bit mode, of course.

Having said that, I appreciate all the work that Matt and
the others are doing.  But there's only a limited number
of things you can work on with a (relatively) small team,
compared to the other BSDs.  Frankly, I'm astonished how
much has been achieved in DF.  Matt has done a tremendous
amount of work.

I'm sure that DF will grow amd64 support sooner or later,
it's probably not very difficult but just requires a
certain amount of time and effort.  In fact it would be
a pity if Matt wasted his time on it.

Just my 2 cents.  :-)

Best regards
   Oliver

PS:  I would really love to help DF myself, but my spare
time is extremely limited since I left university and went
into the "real business life" ...  And even worse since I
got married.  :)

-- 
Oliver Fromme,  secnetix GmbH & Co. KG, Marktplatz 29, 85567 Grafing
Dienstleistungen mit Schwerpunkt FreeBSD: http://www.secnetix.de/bsd

Any opinions expressed in this message may be personal to the author
and may not necessarily reflect the opinions of secnetix in any way.



[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]