Re: [PATCH v2] blkdev: Replace blksize_bits() with ilog2()

[Greg KH]

On Tue, Jun 02, 2020 at 07:51:52AM +0200, Christoph Hellwig wrote:
> On Mon, Jun 01, 2020 at 10:44:26AM +0200, Greg KH wrote:
> > But does this code path actually show up anywhere that is actually
> > measurable as mattering?
> > 
> > If so, please show that benchmark results.
> 
> I think the requests are starting to be a bit unreasonable.  Tao is
> replacing a reimplementation of a standard function with that standard
> function / compiler builtin.  We don't put such a high burden on that.

That's fine, but to say it is "faster" usually means we want to see it
actually going faster somehow :)

> And once the proper existing fields are used where possible as shown
> in my reply just replacing the rest seems totally obvious - quite
> contrary I think keeping a reimplementation would need a high bar.

Your patch makes sense, I was not objecting to that.

thanks,

greg k-h