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

Rahul Siddharthan rsidd120 at gmail.com
Mon Jun 4 04:58:18 PDT 2007


"km b" <kmb810 at gmail.com> wrote:
>> Where is the generalisation?  I said "my own code", which is what I'm
>> interested in. 
>
>You're claiming that all 32-bit binaries are faster than 64-bit
>binaries based on your specific experiments which is nothing but
>generalization.

I made no claim about "all" binaries, and I claimed the exact opposite
of what you say I claimed for my own binaries: I said the 32-bit
binaries are 50% slower than the 64-bit binaries.  (Eg, running time
25 secs for 64-bit, 38 secs for 32-bit.)

>> They are all 64-bit OSs (Linux or Solaris), but a 32-bit binary links
>> with 32-bit libraries.  I don't see why you think the cache usage
>> would be worse for a 32-bit binary.
>
>Simple answer would be cache line size and how processor utilizes
>cache for given virtual address.  You would see this effect more if
>you were just on the edge of thrashing the cache in 32-bit mode,
>64-bit mode would completely thrash.

You seem to be under the odd misconception that I'm claiming 32-bit is
faster.  For my code (and other scientific code), 32-bit is *slower*.
And that can't be explained by thrashing cache -- as you rightly point
out, a 64-bit binary would be *more* likely to suffer such problems,
yet it's faster.

Please read mail properly before shooting off angry replies.

In fact I can believe that code that only uses 32-bit ints, but has
large memory requirements, would perform worse in 64-bit mode.  But
the question is who cares about such code (indeed, how much such code
even exists that's speed-critical).  The sort of job that takes hours
to run -- scientific, multimedia, etc -- would most certainly run
better in 64-bit mode.

Rahul





More information about the Kernel mailing list