[Intel-gfx] [PATCH] drm/i915: Use memset64() to prefill the GTT page

[Chris Wilson]

Quoting Tvrtko Ursulin (2017-09-26 11:30:50)
> 
> On 26/09/2017 10:53, Chris Wilson wrote:
> > Take advantage of optimised memset64() instead of open coding it to
> > prefill the GTT pages.
> > 
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > ---
> > Needs backmerge from 4.14-rc1, but it's tantalisingly close.
> > ---
> >   drivers/gpu/drm/i915/i915_gem_gtt.c | 4 +---
> >   1 file changed, 1 insertion(+), 3 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
> > index 0477949805e1..74062897d331 100644
> > --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
> > +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
> > @@ -486,10 +486,8 @@ static void fill_page_dma(struct i915_address_space *vm,
> >                         const u64 val)
> >   {
> >       u64 * const vaddr = kmap_atomic(p->page);
> > -     int i;
> >   
> > -     for (i = 0; i < 512; i++)
> > -             vaddr[i] = val;
> > +     memset64(vaddr, val, PAGE_SIZE / sizeof(val));
> >   
> >       kunmap_atomic(vaddr);
> >   }
> > 
> 
> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> 
> I wonder if for 32-bit builds (who cares) memset32 would be better in 
> fill_page_dma_32 or it makes no difference?

I considered it, but I thought the unification would still be to our
advantage, and even then that arch may still have a fancy memset64?

One day, we should build a 32bit bdw/skl just for fun and see if
profiles any differently. As well as fixing all the u32/u64 bugs.
-Chris