Re: ldso : dladdr support

[Rich Felker <dalias@aerifal.cx>]

[-- Attachment #1: Type: text/plain, Size: 1063 bytes --]

On Sat, Aug 25, 2012 at 09:42:38AM +0200, boris brezillon wrote:
> >> Do you want me to do some specific tests ?
> >
> > Actually, the main thing I'm interested in is whether the bloom filter
> > is ever beneficial. I took it out trying to streamline the code and
> > shaved about 8% off the lookup time for symbols in the main program,
> > but I didn't investigate how the change affects symbols not found in
> > the first file searched. Would you be interested in running some tests
> > to determine if it might be useful to try adding it back?
> I'll do some tests with multiple levels of big libraries : prog ->
> libtest -> libc -> libb -> liba ...
> How do you get your perf results (specific tools, time measurement
> inside libc code, time measurement in main program, ...)?

I forgot to attach my test code. It's x86-specific (uses rdtsc because
that's the most accurate way) but hopefully that's not a problem.

Note: if searching for symbols in the main program like it's doing as
I have it configured now, you'll need to compile with -rdynamic.

Rich

[-- Attachment #2: gnuhash.c --]
[-- Type: text/plain, Size: 356 bytes --]

#include <dlfcn.h>
#include <stdio.h>

static inline unsigned rdtsc()
{
	unsigned x, dummy;
	__asm__ __volatile__ ( "rdtsc" : "=a"(x), "=d"(dummy) );
	return x;
}

int main()
{
	unsigned i, t0, t, tmin=-1;
	void *p;
	for (i=0; i<16; i++) {
		t0 = rdtsc();
		p = dlsym(0, "main");
		t = rdtsc()-t0;
		if (t<tmin) tmin = t;
	}
	printf("%u %p\n", tmin, p);
}