[lkml]   [2011]   [Nov]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH] perf_events: fix and improve x86 event scheduling
On Mon, 2011-11-07 at 13:52 +0000, Stephane Eranian wrote:
> But given we limit the number of events to that of counters,
> we do have O(c^3).

Right, but SNB without HT gives you 8 GP counters, yielding a rather big
number. Suppose you're trying to fill it with 9 cycle events (1 for the
fixed purpose thingy), that'll end up being: 9^3 = 729 = big number.

(arguably adding 9 cycle counters is a tad retarded, but hey ;-)

It would be good to try and get it down to somewhere near 81 again,
although my brain isn't currently providing any sane ideas on how.

> As for the map_idx, it's there to track the position of each event in the
> initial event list. We shuffle events between constrained and unconstrained.
> By stashing the map_idx in the hw_perf_event struct we avoid having to
> pass around yet another array.

Yeah, I saw why you needed it..

 \ /
  Last update: 2011-11-07 15:59    [W:0.070 / U:1.472 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site