[PATCH] New numbers in the benchtests.

[Patrick McGehearty]

On 12/20/2017 5:42 PM, Joseph Myers wrote:
> On Wed, 20 Dec 2017, Patrick McGehearty wrote:
>
>> The tgamma failures disappeared, giving strong support to Joseph's hypothesis.
>> For x86, there is roughly a 5% performance cost which would be tolerable.
>> Unfortunately, for Sparc, there is a 76% performance cost which is
>> less tolerable. When Sparc changes the rounding mode, the instruction pipeline
> But is presumably still better than the existing code (as you mentioned a
> 5x improvement), so is a reasonable incremental step.
>
>> 3) Define a macro either within e_exp.c or in an include file that selects
>> get_rounding_mode and libc_fesetround for all platforms except x86.
>> It selects SET_RESTORE_ROUND for x86.
>> Putting platform specific macros inside ieee754 branch seems
>> undesirable, but I thought I should mention it as a possibility.
> The correct thing to do is as I said: add libc_fegetround,
> libc_fegetroundf and libc_fegetroundl to the large set of math_private.h /
> fenv_private.h libc_fe* macros.  All of these would default to using
> get_rounding_mode, but sysdeps/i386/fpu/fenv_private.h would, in the
> __SSE_MATH__ case, use the SSE rounding mode for libc_fegetroundf, and in
> the __SSE2_MATH__ case use it also for libc_fegetround.  Then you could
> use libc_fegetround where you previously used get_rounding_mode.
>
> However, using SET_RESTORE_ROUND as an incremental step still makes sense
> before adding libc_fegetround* as an improvement on top of that.
>
While investigating SET_RESTORE_ROUND, I found two cases which were
missing the Rounding mode correction. Fixing those removed the cexp()
test case failures. I also increased the table size to 256 from 64
to reduce the 1 ulp failures to less than 1 in 1000 by informal testing.
Those remaining failures are likely to be in the range of 0.51 ulp.
Investigating those issues delayed my revised submission until now.

I agree it would be desirable to add libc_fegetround* where appropriate,
but I also agree it is better to get the revised patch in the code
base first, then see about the rounding changes.

- Patrick McGehearty