| Message ID | 1254177668-4190-1-git-send-email-brian@xyzw.org (mailing list archive) |
|---|---|
| State | Accepted |
| Headers | show |
On Mon, 28 Sep 2009 15:41:08 -0700 Brian Rogers <brian@xyzw.org> wrote: > Commit 74dff282 exposed this unnecessary call by causing a change in > the failure path on i965 where framebuffer compression will be turned > on and off on every cursor update. If you don't have the > xf86-video-intel fix to avoid the blinking cursor effect, this is > very slow. > > Symptoms were a far more noticeable cursor blink with every cursor > image change combined with severe slowdown for animated cursors. > > Signed-off-by: Brian Rogers <brian@xyzw.org> > Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org> Should also go to stable. Eric can you cc stable when you apply this? Thanks, Jesse
On Mon, 28 Sep 2009 16:25:11 -0700 Jesse Barnes <jbarnes@virtuousgeek.org> wrote: > On Mon, 28 Sep 2009 15:41:08 -0700 > Brian Rogers <brian@xyzw.org> wrote: > > > Commit 74dff282 exposed this unnecessary call by causing a change in > > the failure path on i965 where framebuffer compression will be > > turned on and off on every cursor update. If you don't have the > > xf86-video-intel fix to avoid the blinking cursor effect, this is > > very slow. > > > > Symptoms were a far more noticeable cursor blink with every cursor > > image change combined with severe slowdown for animated cursors. > > > > Signed-off-by: Brian Rogers <brian@xyzw.org> > > Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org> > > Should also go to stable. Eric can you cc stable when you apply this? Ignore this; this code isn't in 2.6.31 so stable cc isn't needed. Jesse
Jesse Barnes schrieb: > On Mon, 28 Sep 2009 16:25:11 -0700 > Jesse Barnes <jbarnes@virtuousgeek.org> wrote: > >> On Mon, 28 Sep 2009 15:41:08 -0700 >> Brian Rogers <brian@xyzw.org> wrote: >> >>> Commit 74dff282 exposed this unnecessary call by causing a change in >>> the failure path on i965 where framebuffer compression will be >>> turned on and off on every cursor update. If you don't have the >>> xf86-video-intel fix to avoid the blinking cursor effect, this is >>> very slow. >>> >>> Symptoms were a far more noticeable cursor blink with every cursor >>> image change combined with severe slowdown for animated cursors. >>> >>> Signed-off-by: Brian Rogers <brian@xyzw.org> >>> Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org> >> Should also go to stable. Eric can you cc stable when you apply this? > > Ignore this; this code isn't in 2.6.31 so stable cc isn't needed. > > Jesse This fix my "regression on vanilla kernel "drm/i915: framebuffer compression for pre-GM45" issue. Thanks Alexey
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 93ff6c0..7a5fb79 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -3095,7 +3095,6 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc, struct drm_gem_object *bo; struct drm_i915_gem_object *obj_priv; int pipe = intel_crtc->pipe; - int plane = intel_crtc->plane; uint32_t control = (pipe == 0) ? CURACNTR : CURBCNTR; uint32_t base = (pipe == 0) ? CURABASE : CURBBASE; uint32_t temp = I915_READ(control); @@ -3182,9 +3181,6 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc, drm_gem_object_unreference(intel_crtc->cursor_bo); } - if ((IS_I965G(dev) || plane == 0)) - intel_update_fbc(crtc, &crtc->mode); - mutex_unlock(&dev->struct_mutex); intel_crtc->cursor_addr = addr;