RFC: What inputs to use for sqrt perf testing

[Joseph S. Myers]

On Fri, 4 Oct 2013, Siddhesh Poyarekar wrote:

> > Would it make sense to start off with all the inputs used by
> > libm-test.inc e.g. sqrt_test_data?
> 
> Oh yeah, why not.  In fact, it won't be a bad idea to do that for all
> functions.  That would be a better starting point than using 42.

Test data for libm functions is typically biased to special cases, whereas 
benchmarks should be biased to more ordinary inputs and certainly don't 
need thorough coverage of many variations on special cases the way the 
test data does.  (But it's a good idea for any value used as a benchmark 
input also to be in libm-test.inc, to verify that the results are correct 
for those arguments.)

-- 
Joseph S. Myers
joseph@codesourcery.com