Hi Ville,
>
> On Fri, Oct 29, 2021 at 12:43:03AM -0700, Vivek Kasireddy wrote:
> > On platforms capable of allowing 8K (7680 x 4320) modes, pinning 2 or
> > more framebuffers/scanout buffers results in only one that is mappable/
> > fenceable. Therefore, pageflipping between these 2 FBs where only one
> > is mappable/fenceable creates latencies large enough to miss alternate
> > vblanks thereby producing less optimal framerate.
> >
> > This mainly happens because when i915_gem_object_pin_to_display_plane()
> > is called to pin one of the FB objs, the associated vma is identified
> > as misplaced and therefore i915_vma_unbind() is called which unbinds and
> > evicts it. This misplaced vma gets subseqently pinned only when
> > i915_gem_object_ggtt_pin_ww() is called without the mappable flag. This
> > results in a latency of ~10ms and happens every other vblank/repaint cycle.
> >
> > Testcase:
> > Running Weston and weston-simple-egl on an Alderlake_S (ADLS) platform
> > with a 8K@60 mode results in only ~40 FPS. Since upstream Weston submits
> > a frame ~7ms before the next vblank, the latencies seen between atomic
> > commit and flip event are 7, 24 (7 + 16.66), 7, 24..... suggesting that
> > it misses the vblank every other frame.
> >
> > Here is the ftrace snippet that shows the source of the ~10ms latency:
> > i915_gem_object_pin_to_display_plane() {
> > 0.102 us | i915_gem_object_set_cache_level();
> > i915_gem_object_ggtt_pin_ww() {
> > 0.390 us | i915_vma_instance();
> > 0.178 us | i915_vma_misplaced();
> > i915_vma_unbind() {
> > __i915_active_wait() {
> > 0.082 us | i915_active_acquire_if_busy();
> > 0.475 us | }
> > intel_runtime_pm_get() {
> > 0.087 us | intel_runtime_pm_acquire();
> > 0.259 us | }
> > __i915_active_wait() {
> > 0.085 us | i915_active_acquire_if_busy();
> > 0.240 us | }
> > __i915_vma_evict() {
> > ggtt_unbind_vma() {
> > gen8_ggtt_clear_range() {
> > 10507.255 us | }
> > 10507.689 us | }
> > 10508.516 us | }
> >
> > v2: Instead of using bigjoiner checks, determine whether a scanout
> > buffer is too big by checking to see if it is possible to map
> > two of them into the ggtt.
> >
> > v3 (Ville):
> > - Count how many fb objects can be fit into the available holes
> > instead of checking for a hole twice the object size.
> > - Take alignment constraints into account.
> > - Limit this large scanout buffer check to >= Gen 12 platforms.
> >
> > Cc: Ville Syrjälä <ville.syrj...@linux.intel.com>
> > Cc: Maarten Lankhorst <maarten.lankho...@linux.intel.com>
> > Cc: Tvrtko Ursulin <tvrtko.ursu...@linux.intel.com>
> > Cc: Manasi Navare <manasi.d.nav...@intel.com>
> > Signed-off-by: Vivek Kasireddy <vivek.kasire...@intel.com>
> > ---
> > drivers/gpu/drm/i915/i915_gem.c | 65 ++++++++++++++++++++++++++++-----
> > drivers/gpu/drm/i915/i915_vma.c | 2 +-
> > 2 files changed, 57 insertions(+), 10 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_gem.c
> > b/drivers/gpu/drm/i915/i915_gem.c
> > index 981e383d1a5d..761dc385fbfc 100644
> > --- a/drivers/gpu/drm/i915/i915_gem.c
> > +++ b/drivers/gpu/drm/i915/i915_gem.c
> > @@ -866,6 +866,61 @@ static void discard_ggtt_vma(struct i915_vma *vma)
> > spin_unlock(&obj->vma.lock);
> > }
> >
> > +static bool i915_gem_obj_too_big(struct drm_i915_gem_object *obj,
> > + u64 alignment)
> > +{
> > + struct drm_i915_private *i915 = to_i915(obj->base.dev);
> > + struct i915_ggtt *ggtt = &i915->ggtt;
> > + struct drm_mm_node *hole;
> > + u64 hole_start, hole_end, start, end;
> > + u64 fence_size, fence_alignment;
> > + unsigned int count = 0;
> > +
> > + /*
> > + * If the required space is larger than the available
> > + * aperture, we will not able to find a slot for the
> > + * object and unbinding the object now will be in
> > + * vain. Worse, doing so may cause us to ping-pong
> > + * the object in and out of the Global GTT and
> > + * waste a lot of cycles under the mutex.
> > + */
> > + if (obj->base.size > ggtt->mappable_end)
> > + return true;
> > +
> > + if (HAS_GMCH(i915) || DISPLAY_VER(i915) < 11 ||
> > + !i915_gem_object_is_framebuffer(obj))
> > + return false;
> > +
> > + fence_size = i915_gem_fence_size(i915, obj->base.size,
> > + i915_gem_object_get_tiling(obj),
> > + i915_gem_object_get_stride(obj));
> > + fence_alignment = i915_gem_fence_alignment(i915, obj->base.size,
> > + i915_gem_object_get_tiling(obj),
> > + i915_gem_object_get_stride(obj));
> > + alignment = max_t(u64, alignment, fence_alignment);
> > +
> > + /*
> > + * Assuming this object is a large scanout buffer, we try to find
> > + * out if there is room to map at-least two of them. There could
> > + * be space available to map one but to be consistent, we try to
> > + * avoid mapping/fencing any of them.
> > + */
> > + drm_mm_for_each_hole(hole, &ggtt->vm.mm, hole_start, hole_end) {
> > + do {
> > + start = round_up(hole_start, alignment);
> > + end = min_t(u64, hole_end, ggtt->mappable_end);
> > +
> > + if (range_overflows(start, fence_size, end))
> > + break;
> > +
> > + count++;
> > + hole_start = start + fence_size;
> > + } while (1);
> > + }
> > +
> > + return count < 2;
> > +}
> > +
> > struct i915_vma *
> > i915_gem_object_ggtt_pin_ww(struct drm_i915_gem_object *obj,
> > struct i915_gem_ww_ctx *ww,
> > @@ -879,15 +934,7 @@ i915_gem_object_ggtt_pin_ww(struct drm_i915_gem_object
> *obj,
> >
> > if (flags & PIN_MAPPABLE &&
> > (!view || view->type == I915_GGTT_VIEW_NORMAL)) {
> > - /*
> > - * If the required space is larger than the available
> > - * aperture, we will not able to find a slot for the
> > - * object and unbinding the object now will be in
> > - * vain. Worse, doing so may cause us to ping-pong
> > - * the object in and out of the Global GTT and
> > - * waste a lot of cycles under the mutex.
> > - */
> > - if (obj->base.size > ggtt->mappable_end)
> > + if (i915_gem_obj_too_big(obj, alignment))
> > return ERR_PTR(-E2BIG);
> >
> > /*
>
> As I pointed out we already have the current ping-pong heuristic
> right here. You should adjust that instead of adding yet another
> heuristic in parallel.
[Kasireddy, Vivek] I think the heuristic you are referring to is:
obj->base.size > ggtt->mappable_end / 2
Are you saying that I should do something like
obj->base.size > ggtt->mappable_end / 4
and return early? I guess I could do that and it would work in this case
but I thought it would make more sense to look at the available space
dynamically given that both the 8K FBs fail the
obj->base.size > ggtt->mappable_end / 2 check.
Thanks,
Vivek
>
> > diff --git a/drivers/gpu/drm/i915/i915_vma.c
> > b/drivers/gpu/drm/i915/i915_vma.c
> > index 90546fa58fc1..551644dbfa8a 100644
> > --- a/drivers/gpu/drm/i915/i915_vma.c
> > +++ b/drivers/gpu/drm/i915/i915_vma.c
> > @@ -977,7 +977,7 @@ int i915_vma_pin_ww(struct i915_vma *vma, struct
> i915_gem_ww_ctx *ww,
> > if (err)
> > goto err_active;
> >
> > - if (i915_is_ggtt(vma->vm))
> > + if (i915_is_ggtt(vma->vm) && flags & PIN_MAPPABLE)
> > __i915_vma_set_map_and_fenceable(vma);
> > }
> >
> > --
> > 2.31.1
>
> --
> Ville Syrjälä
> Intel
