This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: x86 fp fix
On Sat, Mar 25, 2000 at 02:57:29PM +1100, Alan Modra wrote:
> +#ifdef FP_PARANOIA
> + /* Even if we do not want to check the inputs, this documents input
> + constraints. Which helps in understanding the following code. */
Feel free to use ENABLE_CHECKING here.
> + /* know operands[0] == operands[1] */
And a real if/abort here.
> if (STACK_TOP_P (operands[0]))
> - p = "p\t{%0,%2|%2, %0}";
> + /* How is it that we are storing to a dead operand[2]?
Certainly this shouldn't happen when optimizing. When
not optimizing, I'm not so sure it can't happen. Things
are certainly ok if %2 is st(1).
> - p = "p\t{%2,%0|%0, %2}";
> + p = "p\t{%2,%0|%0, %2}"; /* st(r1) = st(r1) op st(0); pop */
As a nit, could you fix up the whitespace on both sides
while you're at it? I.e. {%2, %0|%0, %2}.
r~