| Message ID | 20161107154628.32474-1-chris@chris-wilson.co.uk (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, Nov 07, 2016 at 03:46:28PM +0000, Chris Wilson wrote: > On LLC, or even snooped, machines rendering via the GPU ends up in the CPU > cache. This cacheline dirt also needs to be flushed to main memory when > moving to an incoherent domain, such as the display's scanout engine. > Mostly, this happens because either the object is marked as dirty from > its first use or is avoided by setting the object into the display > domain from the start. > > Fixes: 0f71979ab7fb ("drm/i915: Performed deferred clflush inside set-cache-level") > References: https://bugs.freedesktop.org/show_bug.cgi?id=95414 > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> > Cc: Jani Nikula <jani.nikula@linux.intel.com> > Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> > Cc: <stable@vger.kernel.org> # v4.0+ > --- > drivers/gpu/drm/i915/i915_gem_execbuffer.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c > index 30f6eb516ca3..30b684fecf09 100644 > --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c > +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c > @@ -1282,6 +1282,7 @@ void i915_vma_move_to_active(struct i915_vma *vma, > > /* update for the implicit flush after a batch */ > obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS; > + obj->cache_dirty |= obj->cache_level != I915_CACHE_NONE; I915_CACHE_WT shouldn't dirty the cache either, so could be excluded as well. Everything else makes sense Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> > } > > if (flags & EXEC_OBJECT_NEEDS_FENCE) > -- > 2.10.2
On Mon, Nov 07, 2016 at 05:15:49PM -0000, Patchwork wrote: > == Series Details == > > Series: drm/i915: Mark CPU cache as dirty when used for rendering (rev2) > URL : https://patchwork.freedesktop.org/series/14930/ > State : success An indication that we are missing some coverage. Pushed. -Chris
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index 30f6eb516ca3..30b684fecf09 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -1282,6 +1282,7 @@ void i915_vma_move_to_active(struct i915_vma *vma, /* update for the implicit flush after a batch */ obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS; + obj->cache_dirty |= obj->cache_level != I915_CACHE_NONE; } if (flags & EXEC_OBJECT_NEEDS_FENCE)
On LLC, or even snooped, machines rendering via the GPU ends up in the CPU cache. This cacheline dirt also needs to be flushed to main memory when moving to an incoherent domain, such as the display's scanout engine. Mostly, this happens because either the object is marked as dirty from its first use or is avoided by setting the object into the display domain from the start. Fixes: 0f71979ab7fb ("drm/i915: Performed deferred clflush inside set-cache-level") References: https://bugs.freedesktop.org/show_bug.cgi?id=95414 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Jani Nikula <jani.nikula@linux.intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: <stable@vger.kernel.org> # v4.0+ --- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 1 + 1 file changed, 1 insertion(+)