[PATCH] PowerPC: Trigonometric optimizations for POWER

[Ryan S. Arnold]

On Tue, Aug 9, 2011 at 12:46 PM, Adhemerval Zanella
<azanella@linux.vnet.ibm.com> wrote:
> These POWER optimizations remove most of the FP->INT conversions in
> hypot/hypotf and sinf/cosf on POWER making the computation done with
> FP operations instead.  This eliminates Load-Hit-Store (LHS) stalls,
> increasing performance of hypot/hypotf (by about 100% on POWER7 and
> 12% on POWER6) and sinf/cosf (by 80% on POWER7 and 30% on POWER6).
>
>
> 2011-08-01  Adhemerval Zanella  <azanella@linux.vnet.ibm.com>
>
>        * sysdeps/powerpc/fpu/e_hypot.c: New file: hypot optimized for POWER.
>        * sysdeps/powerpc/fpu/e_hypotf.c: New file: hypotf optimized for POWER.
>        * sysdeps/powerpc/fpu/e_rem_pio2f.c: New file: optimized for POWER.
>        * sysdeps/powerpc/fpu/k_rem_pio2f.c: Likewise.
>        * sysdeps/powerpc/fpu/k_cosf.c: Likewise.
>        * sysdeps/powerpc/fpu/k_sinf.c: Likewise.
>        * sysdeps/powerpc/fpu/s_cosf.c: Likewise.
>        * sysdeps/powerpc/fpu/s_sinf.c: Likewise.
>        * sysdeps/powerpc/fpu/s_scalbnf.c: Likewise.
>        * sysdeps/powerpc/fpu/s_float_bitwise.h: new file: bitwise operation over
>        floats optimized for POWER7.

Ack!

Please consider committing this patch.  I've been running with this
patch on my ibm/2.13 branch for a month.
git commit 0c6ff7937931f1171a6aba3970c4aea7978a1b86

Ryan S. Arnold