On Wed, Sep 02, 2015 at 05:13:29PM +0200, Daniel Vetter wrote: > On Wed, Sep 02, 2015 at 02:46:41PM +0100, Chris Wilson wrote: > > On Wed, Sep 02, 2015 at 02:40:03PM +0100, Michel Thierry wrote: > > > On 9/1/2015 10:06 AM, Michał Winiarski wrote: > > > >On each call to gen8_alloc_va_range_3lvl we're allocating temporary > > > >bitmaps needed for error handling. Unfortunately, when we increase > > > >address space size (48b ppgtt) we do additional (512 - 4) calls to > > > >kcalloc, increasing latency between exec and actual start of execution > > > >on the GPU. Let's just do a single kcalloc, we can also drop the size > > > >from free_gen8_temp_bitmaps since it's no longer used. > > > > > > > >v2: Use GFP_TEMPORARY to make the allocations reclaimable. > > > >v3: Drop the 2D array, just allocate a single block. > > > > > > > >Cc: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> > > > >Cc: Mika Kuoppala <mika.kuoppala@xxxxxxxxxxxxxxx> > > > >Cc: Michel Thierry <michel.thierry@xxxxxxxxx> > > > >Signed-off-by: Michał Winiarski <michal.winiarski@xxxxxxxxx> > > > > > > Unless Chris thinks otherwise, I see Michał already addressed his comments. > > > > I think I spotted a misaligned bracket... ;) > > checkpatch didn't spot it and neither me ... > > > Reviewed-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> > > Queued for -next, thanks for the patch. Strike that, patch doesn't compile too well on plain dinq. What's going on here? And there doesn't seem to be anything in -nightly which isn't in dinq ... -Daniel -- Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
