This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
long double instructions dump core.
- To: gcc-bugs at gcc dot gnu dot org
- Subject: long double instructions dump core.
- From: "M.V.S. Phani Narasimham" <phanimvs at wipinfo dot soft dot net>
- Date: Thu, 30 Dec 1999 15:20:47 +0530 (IST)
- Reply-To: "M.V.S. Phani Narasimham" <phanimvs at wipinfo dot soft dot net>
hi,
The follwing code dumped on sparc.
----------------------------------------------------------------
c code:-
1 main(){
2 long double a=2e0;
3 int i;
4 for(i=0;i<10000000;i++)
5 a=a+2;
6 printf("%Lfg\n",a);
7 }
----------------------------------------------------------------
The following code when it is compiled with cc1 of gcc and linked with
the assembler of sun it dumps core.
Machine details:-
cpu-ultrasparc,instruction set - sparcv9.
The s file generated has the following instructions. It has a call
to the function _Qp_add. It passes the parmaeters in %f0 and %f4.
---------------------------------------------------------------------
.LL6:
sethi %hh(.LLC0), %o0
sethi %lm(.LLC0), %o1
or %o0, %hm(.LLC0), %o0
sllx %o0, 32, %o0
add %o0, %o1, %o0
or %o0, %lo(.LLC0), %o0
ldd [%fp+2015], %f0 <---------
ldd [%fp+2023], %f2
ldd [%o0], %f4 <---------
ldd [%o0+8], %f6
call _Qp_add, 0
nop
fmovd %f0, %f4
fmovd %f2, %f6
std %f4, [%fp+2015]
std %f6, [%fp+2023]
------------------------------------------------------------------------
Where as the Sun c compiler passes a pointers to the two arguments in %o0
and %o1.
The code from the .s file
or %l0,%g1,%l0
ldd [%l0+0],%f12
ldd [%l0+8],%f14
add %fp,1991,%o0 <-------
std %f8,[%fp+1975]
std %f10,[%fp+1983]
std %f12,[%fp+1959]
std %f14,[%fp+1967]
add %fp,1975,%o1 <------
call _Qp_add
add %fp,1959,%o2
ldd [%fp+1991],%f8
ldd [%fp+1999],%f10
---------------------------------------------------------------------
Is it something the the long double operations are not ABI complaint
in gcc.
When the same code is compiled with the option -mhard-quad-float for
cc1. It generates a faddq instruction and it is quite slower. why ?
-----------------------------------------------------------------------
M.V.S. Phani Narasimham
Global R & D Solutions
1 st Block,7 th main road,No:8
Kormangla
BANGLORE-36 , India
Ph- 5530053 ext.1053
E-MAIL- phani.sesha@wipro.com
www.wipro.com
The World's First SEI CMM LEVEL 5 Software Services Company
-----------------------------------------------------------------------