Re: p = kmalloc(sizeof(*p), )

[Matthias Urlichs <smurf@smurf.noris.de>]

Hi, Linus Torvalds wrote:

> Well, to be slightly more positive: it's not a very easy feature to do
> properly.

It's apparently an easy feature to do decidedly suboptimally.

This sample code

#include <malloc.h>
struct foo {
    char fill1[1000];
    int bar;
    char fill2[1000];
    int baz;
    char fill3[1000];
};
struct foo *get_foo() {
    struct foo *res = malloc(sizeof(struct foo));
    *res = (struct foo){.bar=5, .baz=7};
    return res;
}

calls memcpy _twice_ -- it initializes the object on the stack from a
couple of mostly-zero bytes in .rodata, and then memcpy's the thing to the
heap.
 
> So considering that almost nobody does this (certainly not SpecInt), and
> it would probably require re-organizations at many levels, I'm not
> surprised it hasn't gotten a lot of attention.
> 
This is gcc 4.0. Optimization levels (I tried 0, 3, and s) don't affect
this -- which surprised me; I'd have thought that gcc would decide on
the proper trade-off between programmed and static initialization a bit
later.

-- 
Matthias Urlichs   |   {M:U} IT Design @ m-u-it.de   |  smurf@smurf.noris.de
Disclaimer: The quote was selected randomly. Really. | http://smurf.noris.de
 - -
O that my tongue were in the thunder's mouth! Then with a passion would I
shake the world.
					-- Shakespeare