Where's ".div" implemented?

Doug Evans dje@transmeta.com
Tue Jun 3 14:25:00 GMT 2003

.div is in gcc/config/sparc/lb1spc.asm in the gcc sources.

I'm guessing no one's tried running linux on an old v7 machine
because the linux configurations don't include lb1spc.asm in libgcc.a.
[not surprising]

If you configure for sparc-elf then you'll get a libgcc.a that has .div.
There's a few other configurations that will get you .div,
but sparc-elf is, I think, the canonical one for doing embedded
sparc programs with elf.

Nick Patavalis writes:
 > I'm compiling the following program using GCC-3.2.3 configured as a
 > cross-compiler targeting "sparc-linux" and hosted on ix86-linux.
 > int main (void) { int i; i = i / 3; return 0; }
 > My target system uses a sparc-v7 compliant processor without integet
 > multiplication and division instructions, so I'm compiling with:
 >   sparc-linux-gcc -mcpu=v7 -fno-builtin -c tst.c -o tst.o
 > but when I try to link the generated object, like this:
 >   sparc-linux-ld tst.o /path/to/gcc-lib/3.2.3/libgcc.a -o tst
 > I get an error about the linker not been able to find the ".div"
 > function. I was under the impression that the integer-math functions
 > called by code generated from C integer expressions were implemented
 > in "libgcc.a". Obviously they are not!
 > Looking arround I found a ".div" function in *libc*, but for several
 > reasons, I don't want to link my program with the glibc "monster". All
 > I want is a library that supports the *compiler generated* function
 > references. Isn't "libgcc" supposed to be exactly that???
 > I finaly managed to get my program (performing integer math) link,
 > using this round-about way:
 >   mkdir libintm; cd libintm
 >   ar x /path/to/glibc/libc.a \
 >      addmul_1.o sub_n.o submul_1.o dotmul.o \
 >      mul.o mul_1.o mul_n.o umul.o \
 >      div.o udiv.o udiv_qrnnd.o sdiv.o ldiv.o lldiv.o \
 >      divmod_1.o divrem.o rem.o urem.o
 >   ar cr libintm.a *.o
 >   ranlib libintm.a
 > and then link my program with "-lintm".
 > Are the above-mentioned functions *exactly* the ones required? That
 > is, does the compiler implicitly produce refences *only* to these
 > functions, and to *all* of these? Is there a nicer way to achieve the
 > same result?
 > /npat
 > P.S. Obviously my intention is to compile a program that will run at
 > "boot"-time without the services of an operation system, or of an
 > extensive standard library.

Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com

More information about the crossgcc mailing list