Re: [PATCH v4] eliminate mutex in fast path of __register_frame

["H.J. Lu" <hjl.tools@gmail.com>]

On Mon, Nov 21, 2022 at 3:49 AM Jakub Jelinek via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> On Mon, Nov 21, 2022 at 12:22:32PM +0100, Thomas Neumann via Gcc-patches wrote:
> > > When dynamically linking a fast enough machine hides the latency, but when
> > > Statically linking or on slower devices this change caused a 5x increase in
> > > Instruction count and 2x increase in cycle count before getting to main.
> > >
> > > This has been quite noticeable on smaller devices.  Is there a reason the btree
> > > can't be initialized lazily? It seems a bit harsh to pay the cost of unwinding at
> > > startup even when you don't throw exceptions..
> >
> > we cannot easily do that lazily because otherwise we need a mutex for lazy
> > initialization, which is exactly what we wanted to get rid of.
> >
> > Having said that, I am surprised that you saw a noticeable difference. On
> > most platforms there should not be dynamic frame registration at all, as the
> > regular frames are directly read from the ELF data.
> >
> > Can you please send me an precise description on how to reproduce the issue?
> > (Platform, tools, a VM if you have one would be great). I will then debug
> > this to improve the startup time.
>
> I can see it being called as well for -static linked binaries.
> -static links in crtbeginT.o which is libgcc/crtstuff.c built with
> CRTSTUFFT_O macro being defined among other things, and that disables
> USE_PT_GNU_EH_FRAME:
> #if defined(OBJECT_FORMAT_ELF) \
>     && !defined(OBJECT_FORMAT_FLAT) \
>     && defined(HAVE_LD_EH_FRAME_HDR) \
>     && !defined(inhibit_libc) && !defined(CRTSTUFFT_O) \
>     && defined(__GLIBC__) && __GLIBC__ >= 2
> #include <link.h>
> /* uClibc pretends to be glibc 2.2 and DT_CONFIG is defined in its link.h.
>    But it doesn't use PT_GNU_EH_FRAME ELF segment currently.  */
> # if !defined(__UCLIBC__) \
>      && (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ > 2) \
>      || (__GLIBC__ == 2 && __GLIBC_MINOR__ == 2 && defined(DT_CONFIG)))
> #  define USE_PT_GNU_EH_FRAME
> # endif
> #endif
>
> I think .eh_frame_hdr was never used for statically linked programs,
> see already https://gcc.gnu.org/legacy-ml/gcc-patches/2001-12/msg01383.html
> We don't pass --eh-frame-hdr when linking statically and dl_iterate_phdr
> doesn't handle those.
> Now, if -static -Wl,--eh-frame-hdr is passed when linking to the driver,
> .eh_frame_hdr section is created and __GNU_EH_FRAME_HDR symbol points to
> the start of that section, so at least that section could be found
> if something in the crt files and libgcc is adjusted.  But e.g.
> i?86, nios2, frv and bfin we also need to find the got.  Also, would it
> work even for static PIEs?
>
>         Jakub
>

There is

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54568

-- 
H.J.