[Intel-gfx] [PATCH] i915/gem_flink_race: Fix error in buffer usage

[Hampson, Steven T]

Chris,

Is this acceptable?  Can it be merged?

-----Original Message-----
From: Hampson, Steven T 
Sent: Wednesday, November 18, 2020 12:41 PM
To: 'Chris Wilson' <chris at chris-wilson.co.uk>
Cc: intel-gfx@ <lists.freedesktop.org intel-gfx at lists.freedesktop.org>
Subject: RE: [Intel-gfx] [PATCH] i915/gem_flink_race: Fix error in buffer usage

The problem is that the machine it was running on had 32 cpus, so one set of numbers per cpu filled the buffer.

-----Original Message-----
From: Chris Wilson <chris at chris-wilson.co.uk>
Sent: Wednesday, November 18, 2020 2:52 AM
To: Hampson, Steven T <steven.t.hampson at intel.com>
Cc: intel-gfx@ <lists.freedesktop.org intel-gfx at lists.freedesktop.org>
Subject: Re: [Intel-gfx] [PATCH] i915/gem_flink_race: Fix error in buffer usage

Quoting Hampson, Steven T (2020-11-17 23:45:23)
> 
> 
> Sent from my iPhone
> 
> > On Nov 17, 2020, at 2:28 PM, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> > 
> > Quoting Steve Hampson (2020-11-17 22:23:08)
> >> A buffer in function test_flink_name was both too small and never 
> >> checked for overflow.  Both errors are fixed.
> > 
> > That many numbers is not interesting. Show the range and median instead.
> > -Chris
> 
> I don’t understand what you are talking about.  

The reason I printed the individual numbers was so that we could see the distribution in case one thread was being starved or not. That is fine for a few numbers, but beyond that we can summarise with statistics.
-Chris