Re: [Intel-gfx] [PATCH] drm/i915/gtt: Trust the uncached store to flush wcb

2018-05-11 Thread Mika Kuoppala
Chris Wilson writes: > Quoting Mika Kuoppala (2018-05-08 13:03:13) >> Not all architectures guarantee that uncached read will >> flush the write combining buffer. So marking it explicitly >> is recommended [1]. >> >> However we know the architecture we are operating on >> and can avoid wmb as th

[Intel-gfx] [PATCH] drm/i915/gtt: Trust the uncached store to flush wcb

2018-05-08 Thread Mika Kuoppala
Not all architectures guarantee that uncached read will flush the write combining buffer. So marking it explicitly is recommended [1]. However we know the architecture we are operating on and can avoid wmb as the UC store will flush the wcb [2]. Omit the wmb() before invalidate as redudant. v2:

Re: [Intel-gfx] [PATCH] drm/i915/gtt: Trust the uncached store to flush wcb

2018-05-08 Thread Chris Wilson
Quoting Mika Kuoppala (2018-05-08 13:03:13) > Not all architectures guarantee that uncached read will > flush the write combining buffer. So marking it explicitly > is recommended [1]. > > However we know the architecture we are operating on > and can avoid wmb as the UC store will flush the wcb [

[Intel-gfx] [PATCH] drm/i915/gtt: Trust the uncached store to flush wcb

2018-05-08 Thread Mika Kuoppala
Not all architectures guarantee that uncached read will flush the write combining buffer. So marking it explicitly is recommended [1]. However we know the architecture we are operating on and can avoid wmb as the UC store will flush the wcb [2]. Omit the wmb() before invalidate as redudant. v2: