On Tue, Nov 14, 2017 at 11:48 AM, Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote: > On Tue, Nov 14, 2017 at 11:33:20AM +0200, Gilad Ben-Yossef wrote: >> On Mon, Nov 13, 2017 at 8:33 PM, Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote: >> > These cleanups look nice. Thanks. >> > >> > I hope you do a mass remove of likely/unlikely in a patch soon. >> > Whenever, I see one of those in a + line I always have to remind myself >> > that you're planning to do it in a later patch. >> > >> >> So, a question about that - there indeed seems to be an inflation of >> likely/unlikely in the ccree driver, but >> what stopped me from removing them was that I found out I don't have a >> clue about when it's a good idea >> to use them and when it isn't (obviously in places where you know the >> probable code flow of course). >> >> Any hints? > > They should only be included if benchmarking shows that it makes a > difference. I think they need to be about 100 right predictions to 1 > wrong prediction on a fast path. So remove them all and add them back > one at a time. > OK, that makes a lot of sense. Thanks, Gilad -- Gilad Ben-Yossef Chief Coffee Drinker "If you take a class in large-scale robotics, can you end up in a situation where the homework eats your dog?" -- Jean-Baptiste Queru
