This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Unordered SPARC fixes
On Wed, Jan 26, 2000 at 07:14:06PM +0100, Jakub Jelinek wrote:
> Yep, but I think the comparison argument did not matter on SPARC, as before
> we were passing comparison itself (e.g. GT, LT etc. eventhough we meant NE)
> and it worked. The actual branch is generated in .md.
Oh I see.
> BTW: Have you had time to look at the LONG_DOUBLE_TYPE_SIZE patch I've
> posted yesterday?
Briefly. I wanted to examine the i960 bits closer before I acked it.
> * config/sparc/sparc.c (sparc_emit_float_lib_cmp): Handle
> TARGET_ARCH32 again. Handle ORDERED, UN* and LTGT comparisons
> using _Qp_cmp/_Q_cmp and testing the return value.
> (print_operand): Call reverse_condition_maybe_unordered if
> we are handling CCFPmode or CCFPEmode.
> Handle ORDERED, UN* and LTGT comparisons.
> * config/sparc/sparc.md (cmptf): Use even on TARGET_ARCH32
> if not TARGET_HARD_QUAD.
> (seq, sne, sgt, slt, sge, sle, beq, bne, bgt, blt, bge, ble,
> bunordered, bordered, bungt, bunlt, buneq, bunge, bunle, bltgt):
> Call sparc_emit_float_lib_cmp even on TARGET_ARCH32.
> Adjust gen_b* calls so that they reflect return comparison of
> sparc_emit_float_lib_cmp.
Ok.
r~