On 2013-09-19 19:26, Eric Blake wrote:
[…]
That's good news. Remember, my question about whether we needed the
ability for command-line tuning was conditional on whether the tuning
would make a difference - but if you aren't seeing a difference with
your benchmark, then we might as well unconditionally enable the
checking, and not worry about the complexity of exposing additional
tuning. Simpler is better if it makes no difference in end.
[Of course, if you WANT to write the patches for making it configurable,
I won't stop you; but now that you answered my question about
performance, and the answer is the desirable "no measurable impact", I
don't know that I would ever plan on using such tuning]
In fact, I have written those patches already, so I'll just send them. ;)
Max
