On Mon, Nov 4, 2013 at 7:33 PM, Bill Schmidt
wrote:
> Hi,
>
> Here's a revised version of this patch according to Richard's
> suggestions. It differs from the previous version only in the method
> used to ensure vmulouh is generated; we now call the new
> gen_altivec_vmulouh to accomplish this.
>
Hi,
Here's a revised version of this patch according to Richard's
suggestions. It differs from the previous version only in the method
used to ensure vmulouh is generated; we now call the new
gen_altivec_vmulouh to accomplish this.
Bootstrapped and tested on powerpc64{,le}-unknown-linux-gnu with
On Mon, 2013-11-04 at 15:48 +, Richard Sandiford wrote:
> Bill Schmidt writes:
> > + /* We need this to be vmulouh for both big and little endian,
> > + but for little endian we would swap this, so avoid that. */
> > + if (BYTES_BIG_ENDIAN)
> > + emit_insn (gen_vec_widen_umult_od
Bill Schmidt writes:
> + /* We need this to be vmulouh for both big and little endian,
> + but for little endian we would swap this, so avoid that. */
> + if (BYTES_BIG_ENDIAN)
> + emit_insn (gen_vec_widen_umult_odd_v8hi (low_product, one, two));
> + else
> + emit_insn (gen_vec
Hi,
This patch contains two more fixes related to the multiply even/odd
problem. First, it changes the mulv4si3 pattern so that it always uses
the vmulouh (vector multiply odd halfword) instruction regardless of
endianness. The reason for this is that we are not really multiplying
halfwords, but
