On Thu, Aug 29, 2002 at 01:09:05PM +1000, Rusty Russell wrote:
> In message <200208282131.g7SLVVGx024191@siamese.dyndns.org> you write:
> > Here is a patch that does the same as what Keith Owens did in
> > his patch recently.
> >
> > Message-ID: <fa.iks3ohv.1flge08@ifi.uio.no>
> > From: Keith Owens <kaos@ocs.com.au>
> > Subject: [patch] 2.4.19 Generate better code for nfs_sillyrename
> > Date: Wed, 28 Aug 2002 07:08:17 GMT
> >
> > Using strlen() generates an unnecessary inline function expansion plus
> > dynamic stack adjustment. For constant strings, strlen() == sizeof()-1
> > and the object code is better.
>
> Disagree. If you really care make strlen use __builtin_constant_p().
> Then authors don't have to sacrifice readability.
>
> #define strlen(x) (__builtin_constant_p(x) ? sizeof(x)-1 : __strlen(x))
Also disagree; besides, the evidence implies that Keith is wrong. GCC
2.95.3:
drow@nevyn:~% cat strlen.c
int foo() { return strlen ("baz"); }
int bar() { return sizeof ("baz") - 1; }
drow@nevyn:~% cat strlen.s
.file "strlen.c"
.version "01.01"
gcc2_compiled.:
.text
.align 4
.globl foo
.type foo,@function
foo:
pushl %ebp
movl %esp,%ebp
movl $3,%eax
leave
ret
.Lfe1:
.size foo,.Lfe1-foo
.align 4
.globl bar
.type bar,@function
bar:
pushl %ebp
movl %esp,%ebp
movl $3,%eax
leave
ret
.Lfe2:
.size bar,.Lfe2-bar
.ident "GCC: (GNU) 2.95.4 20011002 (Debian prerelease)"
3.2 does the same thing. With -fomit-frame-pointer the results are as
expected, just a move and a return.
If you include headers that define strlen, that's another problem.
-- Daniel Jacobowitz MontaVista Software Debian GNU/Linux Developer - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Aug 31 2002 - 22:00:25 EST