11:03 <itoral> apinheiro[m]1: I am working on plugging nops after a thrsw:

11:04 <itoral> inst-and-stalls in affected programs: 12297868 -> 12024882 (-2.22%)

11:04 <itoral> HURT: 17977

11:04 <itoral> total inst-and-stalls in shared programs: 13753165 -> 13480179 (-1.98%)

11:04 <itoral> helped: 65960

11:04 <itoral> that's actually pretty good

11:04 <itoral> almost a 2% improvement with a single patch

11:04 <itoral> going through CI now

11:04 <itoral> and I think I can actually do a bit better than that

11:04 <itoral> in the final versionn

11:09 <apinheiro[m]1> itoral: nice

11:09 <apinheiro[m]1> btw, taking into accoun that most of your late work is for the broadcom compiler

11:10 <apinheiro[m]1> perhaps it would be good to also do some perf measuref (before vs after) with OpenGL apps

11:10 <apinheiro[m]1> so far we only did that with Vulkan apps

11:10 <itoral> meh, I don't think that is necessary, gains should be on similar levels

11:35 <apinheiro[m]1> ok, just pure curiosity

12:33 <itoral> I mean, it is fine to collect stats if you're curious, I just expect to see similar results since well, they use the same compiler code :)

12:34 <itoral> the only difference could be that in the case of OpenGL there is more CPU overhead and that may make some of the benfits smaller in some games

12:34 <itoral> but it would not be related to the compiler opts at all