If you apply the kernel patch I mentioned here a few days ago
you can build floating point emulation into the kernel.
Otherwise there is a link to an old version of libfloat on my
web page; I keep meaning to make a newer one including Scott's
changes but I haven't had time yet.
I've applied the patch and compiled a new kernel. The new kernel is able
to run floating point SW but the result of the FP operations are
incorrect. In the case of a division the binary exits (i guess with a
division by zero) prematurely. (Test case appended)
I was unable to download libfloat, the link was not working.
Thanks,
Mark.
-------------------------------
/* The followin test case produces a zero for the first number printed
with a + and -. When a / is used the result is not printed.
When a * is used the result is incorrect 1->-0.25, 2->-2, 3->-6 ...
Compiled with: arm-linux-gcc -o fp fp.cpp -static
*/
#include <stdio.h>
#include <stdlib.h>
main(int argc,char **argv)
{
double d=atof(argv[1]);
printf("result %lf %lf\n",d+3,d);
}
unsubscribe: body of `unsubscribe linux-arm' to [EMAIL PROTECTED]
