On Wed, Jan 18, 2017 at 11:21:40AM -0600, Segher Boessenkool wrote:
> On Tue, Jan 17, 2017 at 08:50:08PM -0500, Michael Meissner wrote:
> > I have checked this on a little endian power8 system (64-bit only), a big
> > endian power8 system (64-bit only), and a big endian power7 system (both
> > 32-bit
> > and 64-bit), and there were no regressions. Can I check this into the
> > trunk?
>
> Yes please.
>
> One comment:
>
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRQ_V16QI,
> > + RS6000_BTI_unsigned_V16QI, RS6000_BTI_unsigned_V16QI, 0, 0 },
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRQ_V16QI,
> > + RS6000_BTI_V16QI, RS6000_BTI_V16QI, 0, 0 },
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRQ_V1TI,
> > + RS6000_BTI_unsigned_V1TI, RS6000_BTI_unsigned_V1TI, 0, 0 },
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRQ_V1TI,
> > + RS6000_BTI_V1TI, RS6000_BTI_V1TI, 0, 0 },
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRD_V2DI,
> > + RS6000_BTI_unsigned_V2DI, RS6000_BTI_unsigned_V2DI, 0, 0 },
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRD_V2DI,
> > + RS6000_BTI_V2DI, RS6000_BTI_V2DI, 0, 0 },
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRD_V2DF,
> > + RS6000_BTI_V2DF, RS6000_BTI_V2DF, 0, 0 },
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRW_V4SI,
> > + RS6000_BTI_unsigned_V4SI, RS6000_BTI_unsigned_V4SI, 0, 0 },
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRW_V4SI,
> > + RS6000_BTI_V4SI, RS6000_BTI_V4SI, 0, 0 },
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRW_V4SF,
> > + RS6000_BTI_V4SF, RS6000_BTI_V4SF, 0, 0 },
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRH_V8HI,
> > + RS6000_BTI_unsigned_V8HI, RS6000_BTI_unsigned_V8HI, 0, 0 },
> > + { P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRH_V8HI,
> > + RS6000_BTI_V8HI, RS6000_BTI_V8HI, 0, 0 },
>
> This is much harder to review (and read in general) than needed, because
> it uses alphapetical ordering instead of something logical (like, all
> integer together and ordered by size; all float together and ordered by
> size).
I ordered it by the size of the vector element being swapped, i.e. XXBRQ comes
first, because it is swapping the bytes of one element; XXBRD comes next
swapping bytes in vector long/vector double; XXBRW comes next swapping bytes in
vector int/vector short, and finally XXBRH swapping bytes in vector short.
--
Michael Meissner, IBM
IBM, M/S 2506R, 550 King Street, Littleton, MA 01460-6245, USA
email: meiss...@linux.vnet.ibm.com, phone: +1 (978) 899-4797
