https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112465
john henning <mailboxnotfound at yahoo dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |mailboxnotfound at yahoo dot
com
--- Comment #3 from john henning <mailboxnotfound at yahoo dot com> ---
(In reply to Wilco from comment #2)
> it is best to use -mcpu=native and
> only use -mcmodel=large if there is no other option.
For the record if anyone else running large OpenMP programs (such as McCalpin
Stream) should find their way here, the first two below work, the last two
fail:
DOIT="/export/GCC/gcc-14.2.0/bin/gcc -O3 -DSTREAM_ARRAY_SIZE=1073741824
-fopenmp stream.c "
$DOIT -mcmodel=large -mcpu=native
$DOIT -mcmodel=large -mno-outline-atomics
$DOIT -mcpu=native
$DOIT -mno-outline-atomics
The above use gcc 14.2 for target=aarch64-unknown-linux-gnu configured with
--enable-languages=c,c++,fortran --enable-lto
plus
unmodified stream.c,v 5.10 2013/01/17 16:01:06 mccalpin
For both of the failing cases, the symptoms are:
stream.c:(.text+0x nnnn ): relocation truncated to fit:
R_AARCH64_ADR_PREL_PG_HI21 against `.bss'
The above repeats for several nnn, and ends with
stream.c:(.text+0x nnn ): additional relocation overflows omitted from the
output
