kubuntu-users Digest, Vol 19, Issue 4

[jakykong at theanythingbox.com]

>jakykong at theanythingbox.com dedi ki:
>
>> As far is i know (doing exactly that -- packages built for 64-bit are too
>> limited and the lagrest advantage of 64-bit is that it can handle more
>> memory, that's about it as far as i know), there is little or no
>> performance issues with 32-bit.
>
>As I've stated earlier, amd64 has more instructions, plus more (x2)
>registers and larger (x2) register width. As it is hardware-wise backward
>compatible with i386 and up, it will run programs compiled for i386 and
>up - but will use only i386 subset of its full architecture. I.e. an app
>compiled for i386 will not be able to use half the registers available, do
>juggling to handle 64-bit data while half of the 64-bit registers are
>unused, and resort to software routines to accomplish some tasks instead
>of doing it with a single (new) assembly instruction.
>
>Presuming an optimizing compiler is being used, the same app compiled for
>native amd64 will automatically tap into all the extras of amd64
>architecture.
>
>Dug a bit of the past discussions I've had over this subject and found
>these links:
>
>http://www.linuxhardware.org/article.pl?sid=04/09/17/1453239 (Jump
>to "Performance" header, just past half way the article)
>
>http://www.pcstats.com/articleview.cfm?articleID=1665 (See benchmarks at
>page 6,7, and 8)
>
>By the way, welcome! :)
>-- 
>Abdullah Ramazanoglu
>aramazan ÄT myrealbox D0T cöm

I wasn't here to know what you stated earlier :-) (didn't think to check the
archives... looked like a topic starter)

I won't call myself an expert (i'm not, i'm only 16, no classes, no
certifications (yet), and i program assembly for intel IA32 only, never used
64-bit, so i have relatively patchy knowlege of the subject), but my
understanding was that 64 bit differed from 32 bit in the same way that 32
differed from 16 bit: yes more registers to handle the 32-bit capabilities of
the new chip (for example, we got EAX instead of just AX :D), more memory
(since the number of bits listed for the processor is the width of the data
bus... or was it address bus? i'd have to look that up...) capacity, and some
new instructions. 
Most of the new instructions that i know of (i don't know all the
instructions, by far) that were introduced in 32-bit were 32-bit versions of
the old 16-bit instructions and some memory management instructions.

Again, i'm open-minded, i'd rather know the truth then be egotistic, so
correct me if i'm wrong, but if 64-bit continued the pattern, most new
instructions would be little more then 64-bit versions of the older 32-bit
instructions.

However, to the topic at hand: performance issues depends if you're talking
about running a different port of the program on the 2 systems. What i meant
to come across was that if you run the application on a 32-bit machine, and
run the same application on a 64-bit machine in 32-bit mode, there should be
no performance issues. Due to the fact that data can be moved back and forth
faster on 64-bit computers, i would expect that 64-bit applications running
on a 64-bit machine would run faster then the same application compiled for
32-bit, but again, that could just be my fuzzy logic programming goin' at it
again :-P.