[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [dev-serveez] solaris build
From: |
Raimi 'Raimi' Jacob |
Subject: |
Re: [dev-serveez] solaris build |
Date: |
Thu, 6 Dec 2001 10:45:51 +0100 (CET) |
On Thu, 6 Dec 2001, stefan wrote:
hi!
> Maybe we could inhibit the warning by using
> '-L/usr/gnu/lib/gcc-lib/sparc-sun-solaris2.8/2.95.3 -lgcc'? I'll give it
> a try and check it into CVS. This is because I think libtool recognises
> static archives by thefile extension ".a". I conclude this because
> libtool does not complain about the "libiberty.a" which is a static
> library, too.
> > ... this message isnt really a surprise but i think we dont have to care
> > (since the explicit link is dynamically created).
> I agree, actually we do not need to care.
so? change and check ? or just leave it as it is ? i dont care about the
warning. since gcc --print ... is somehow guaranteed to work, i'd rather
use this thatn -L..... -lgcc (might introduce other trouble).
> > p.s.: the stripped executable is 500kb, compared to the 300kb of the x86
> > version. no real suprise, either.
> Uhm. Why isn't this a surprise? I wonder a bit...
SPARC is a RISC: few opcodes, all instructions have same length (32 bit in
this case). x86 is CISC: many opcodes, even variable length: a
multiplication is (for example) 1 + 4 byte in x86 (opcode, operand) while
the Sparc doesnt have a multiplication -> uses 32 "multstep" opcodes (4
byte each). this is an extreme example. another one: "dec eax" (probably 1
byte, dont know) is a "sub %r1, 1, %r1" (32 bit for sure). "mov eax,
42" is 5 byte (1+4) but 8 on Sparc (ld and sethi).
bottom line: Risc code is usually bigger than Cisc code.
Bye,
Raimi
--
__/\ _/\ _____/\.___/\
/ / / /___/ ____/ __/\ Name : Raimi
/ / __/ __/ / __/ / _\/ Contact : address@hidden
/ /__/ / \/ /_/_/ /_/ Visit : http://www.lkcc.org
/________/___/\._\._____/_____/\ 3.141592653589793238462643383
\._______\.__\/__/\.____\.____\/ 27950288419716939937510582097