With the current code there shouldn't be a distinction - however with an
upcoming change we intend to allocate a vma much earlier, before it's
actually bound anywhere.
To do this we have to check node allocation as well for the _bound()
check.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: move list_del(&vma->vma_link) from vma_unbind to vma_destroy,
again fallout from the loss of "rm/i915: Cleanup more of VMA in
destroy".]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
fixup for drm/i915: Add vma to list at creation
formerly: "drm/i915: Create VMAs (part 4) - Error capture"
Since the active/inactive lists are per VM, we need to modify the error
capture code to be aware of this, and also extend it to capture the
buffers from all the VMs. For now all the code assumes only 1 VM, but it
will become more generic over the next few patches.
NOTE: If the number of VMs in a real world system grows significantly
we'll have to focus on only capturing the guilty VM, or else it's likely
there won't be enough space for error capture.
v2: Squashed in the "part 6" which had dependencies on the mm_list
change. Since I've moved the mm_list change to an earlier point in the
series, we were able to accomplish it here and now.
v3: Rebased over new error capture
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
formerly: "drm/i915: Create VMAs (part 5) - move mm_list"
The mm_list is used for the active/inactive LRUs. Since those LRUs are
per address space, the link should be per VMx .
Because we'll only ever have 1 VMA before this point, it's not incorrect
to defer this change until this point in the patch series, and doing it
here makes the change much easier to understand.
Shamelessly manipulated out of Daniel:
"active/inactive stuff is used by eviction when we run out of address
space, so needs to be per-vma and per-address space. Bound/unbound otoh
is used by the shrinker which only cares about the amount of memory used
and not one bit about in which address space this memory is all used in.
Of course to actual kick out an object we need to unbind it from every
address space, but for that we have the per-object list of vmas."
v2: only bump GGTT LRU in i915_gem_object_set_to_gtt_domain (Chris)
v3: Moved earlier in the series
v4: Add dropped message from v3
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Frob patch to apply and use vma->node.size directly as
discused with Ben. Also drop a needles BUG_ON before move_to_inactive,
the function itself has the same check.]
[danvet 2nd: Rebase on top of the lost "drm/i915: Cleanup more of VMA
in destroy", specifically unlink the vma from the mm_list in
vma_unbind (to keep it symmetric with bind_to_vm) instead of
vma_destroy.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
formerly: "drm/i915: Create VMAs (part 3.5) - map and fenceable
tracking"
The map_and_fenceable tracking is per object. GTT mapping, and fences
only apply to global GTT. As such, object operations which are not
performed on the global GTT should not effect mappable or fenceable
characteristics.
Functionally, this commit could very well be squashed in to a previous
patch which updated object operations to take a VM argument. This
commit is split out because it's a bit tricky (or at least it was for
me).
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Drop the bogus hunk in i915_vma_unbind as discussed with
Ben.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
drivers/gpu/drm/i915/i915_debugfs.c:2136:3: warning: symbol
'i915_debugfs_files' was not declared. Should it be static?
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We're going to use the 1/2 vs. 5/6 split option already on IVB so the
HSW name is not proper. Just give it an intel_ prefix and move it to
i915_drv.h so that we can use it there later.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We don't need to store the FBC WM enabled status in each watermark
level. We anyway have to reduce it down to a single boolean, so just
delay checking the FBC WM limit until we're computing the final
value.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Refactor the watermarks computation for one level to a separate
function. This function will now set the ->enable flag to true,
even if the watermark level wasn't actually checked yet. In the
future we will delay the checking so we must consider all unchecked
watermarks as possibly valid.
v2: Preserve comment about latency units
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Let's be consistent and always call our variables 'enabled' insted of
the occasional 'enable'.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
[danvet: Spelling fix in the commit message, spotted by Chris.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
set_frame() wraps the write_frame() vfunc. Be consistent and name the
wrapping function like the vfunc being called.
It's doubly confusing as we also have a set_infoframes() vfunc and
set_infoframe() doesn't wrap it.
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I cannot find any evidence what we shouldn't try to set those fields
when setting a non-CEA mode on an HDMI sink. So just kill that return.
Suggested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Acked-by: Dave Airlie <airlied@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Suggested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Acked-by: Dave Airlie <airlied@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
All the HDMI infoframe code has been ported to use video/hdmi.c, so it's
time to say bye bye to this code.
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni at intel.com>
Signed-off-by: Thierry Reding <thierry.reding at avionic-design.de>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Let's use the drivers/video/hmdi.c and drm infoframe helpers to build
our infoframes.
v2: Simplify the logic to compute the buffer size. We can just take the
maximum infoframe size rounded to 32, which happens to be what the
hardware let us write anyway.
v3: Remove unnecessary memset() (Ville Syrjälä)
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
First step in the move to the shared infoframe infrastructure, let's
move the different infoframe helpers and the write_infoframe() vfunc to
a type (enum hdmi_infoframe_type) and a buffer + len instead of using
our struct dip_infoframe.
v2: constify the infoframe pointer and don't mix signs (Ville Syrjälä)
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni at intel.com>
Signed-off-by: Thierry Reding <thierry.reding at avionic-design.de>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
From CEA-861:
Data Byte 1, bit A0 indicates whether Active Format Data is present in
Data Byte 2 bits R3 through R0. A source device shall set A0=1 when
any of the AFD bits are set.
ie. if we want to set active_aspect, we need to set the
active_info_valid bit to 1 as well.
Cc: Thierry Reding <thierry.reding@avionic-design.de>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Acked-by: Dave Airlie <airlied@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
If the user if this API is providing a bigger buffer than the infoframe
size, it could be for a could reason. For instance it could be because
it gives the buffer that will be written to the hardware, up to the
maximum of an infoframe size.
Instead of just zeroing up to the infoframe size, let's zero the whole
incoming buffer as those extra bytes are also used to compute the
ECC and need to be 0.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Acked-by: Dave Airlie <airlied@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
And a way to pack hdmi_infoframe generically.
Cc: Thierry Reding <thierry.reding@avionic-design.de>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Acked-by: Dave Airlie <airlied@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In some places, we want to know if an object is bound in any address
space, and not just the global GTT. This often applies when there is a
single global resource (object, pages, etc.)
function | reason
--------------------------------------------------
i915_gem_object_is_inactive | global object
i915_gem_object_put_pages | object's pages
915_gem_object_unpin | global object
i915_gem_execbuffer_unreserve_object | temporary until we plumb vma
pread/pwrite | see the note below
Note: set_to_gtt_domain in pwrite/pread is abused as a wait_rendering
call - but that once only worked if the object is bound. We really
should replace this with a plain wait_rendering call, which would have
the upside that in pread it would be clearer that we actually only
wait for oustanding gpu writes.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Explain the set_to_gtt_domain in pwrite/pread and volunteer
Ben to replace those with wait_rendering calls.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Eviction code, like the rest of the converted code needs to be aware of
the address space for which it is evicting (or the everything case, all
addresses). With the updated bind/unbind interfaces of the last patch,
we can now safely move the eviction code over.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
As alluded to in several patches, and it will be reiterated later... A
VMA is an abstraction for a GEM BO bound into an address space.
Therefore it stands to reason, that the existing bind, and unbind are
the ones which will be the most impacted. This patch implements this,
and updates all callers which weren't already updated in the series
(because it was too messy).
This patch represents the bulk of an earlier, larger patch. I've pulled
out a bunch of things by the request of Daniel. The history is preserved
for posterity with the email convention of ">" One big change from the
original patch aside from a bunch of cropping is I've created an
i915_vma_unbind() function. That is because we always have the VMA
anyway, and doing an extra lookup is useful. There is a caveat, we
retain an i915_gem_object_ggtt_unbind, for the global cases which might
not talk in VMAs.
> drm/i915: plumb VM into object operations
>
> This patch was formerly known as:
> "drm/i915: Create VMAs (part 3) - plumbing"
>
> This patch adds a VM argument, bind/unbind, and the object
> offset/size/color getters/setters. It preserves the old ggtt helper
> functions because things still need, and will continue to need them.
>
> Some code will still need to be ported over after this.
>
> v2: Fix purge to pick an object and unbind all vmas
> This was doable because of the global bound list change.
>
> v3: With the commit to actually pin/unpin pages in place, there is no
> longer a need to check if unbind succeeded before calling put_pages().
> Make put_pages only BUG() after checking pin count.
>
> v4: Rebased on top of the new hangcheck work by Mika
> plumbed eb_destroy also
> Many checkpatch related fixes
>
> v5: Very large rebase
>
> v6:
> Change BUG_ON to WARN_ON (Daniel)
> Rename vm to ggtt in preallocate stolen, since it is always ggtt when
> dealing with stolen memory. (Daniel)
> list_for_each will short-circuit already (Daniel)
> remove superflous space (Daniel)
> Use per object list of vmas (Daniel)
> Make obj_bound_any() use obj_bound for each vm (Ben)
> s/bind_to_gtt/bind_to_vm/ (Ben)
>
> Fixed up the inactive shrinker. As Daniel noticed the code could
> potentially count the same object multiple times. While it's not
> possible in the current case, since 1 object can only ever be bound into
> 1 address space thus far - we may as well try to get something more
> future proof in place now. With a prep patch before this to switch over
> to using the bound list + inactive check, we're now able to carry that
> forward for every address space an object is bound into.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Rebase on top of the loss of "drm/i915: Cleanup more of VMA
in destroy".]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In order to do this for all VMs, it's convenient to rework the logic a
bit. This should have no functional impact.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Rather than open-code the teardown of a framebuffer, export the routine
from intel_display.c. This then make intel_fbdev symmetric in its use of
the common intel_framebuffer routines to initialise and clean up the
struct intel_framebuffer. (And new features need only be added in one
location!)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
MLC_LLC was never validated for Sandybridge and was superseded by a new
level of cacheing for the GPU in Ivybridge. Update our names to be
consistent with usage, and in the process stop setting the unwanted bit
on Sandybridge.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
[danvet: s/BUG/WARN_ON(1) bikeshed.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We set the mode based on the port, and we already pass the port as an
argument.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
These messages are not really useful since it's very easy to check
which mode is used for each port: The values programmed are based on
the port type, then assigned to the ddi_translations variable.
Currently we use DP mode for ports A-D and FDI mode for port E.
Also, when we add the code to enable/disable PC8+,
intel_prepare_ddi_buffers will be called more often and will eat your
dmesg buffers.
While at it, fix the coding style of the "for" statement above.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
[danvet: Pimp commit message with Paulo's more detailed explanation of
how the ddi translation buffer settings are computed, to answer a
question from Chris.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The code itself is no longer accurate without updating once we have
multiple address space since clearing the domains of every object
requires scanning the inactive list for all VMs.
"This code is dead. Just remove it rather than port it to vma." - Chris
Wilson
Recommended-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
All the ILK+ WM compute functions take the latency values in 0.1us
units. Add a few comments to remind people about that.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Adjust the current ILK/SNB/IVB watermark codepaths to use the
pre-populated latency values from dev_priv instead of reading
them out from the registers every time.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Return UINT_MAX for the calculated WM level if the latency is zero.
This will lead to marking the WM level as disabled.
I'm not sure if latency==0 should mean that we want to disable the
level. But that's the implication I got from the fact that we don't
even enable the watermark code of the SSKDP register is 0.
v2: Use WARN() to scare people
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Seeing the watermark latency values in dmesg might help sometimes.
v2: Use DRM_ERROR() when expected latency values are missing
Note: We might hit the DRM_ERROR added in this patch and apparently
there's not much we can do about that. But I think it'd be interesting
to figure out whether that actually happens in the real world, so I
didn't apply a s/DRM_ERROR/DRM_DEBUG_KMS/ bikeshed while applying.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
[danvet: Add note about new error dmesg output.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Rather than pass around the plane latencies, just grab them from
dev_priv nearer to where they're needed. Do the same for cursor
latencies.
v2: Add some comments about latency units
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Rather than having to read the latency values out every time, just
store them in dev_priv.
On ILK and IVB there is a difference between some of the latency
values for different planes, so store the latency values for each
plane type separately, and apply the necesary fixups during init.
v2: Fix some checkpatch complaints
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
ILK has a slightly different way to read out the watermark
latency values. On ILK the LP0 latenciy values are in fact
not stored in any register, and instead we must use fixed
values.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Hangcheck, and some of the recent reset code for guilty batches need to
know which address space the object was in at the time of a hangcheck.
This is because we use offsets in the (PP|G)GTT to determine this
information, and those offsets can differ depending on which VM they are
bound into.
Since we still only have 1 VM ever, this code shouldn't yet have any
impact.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
With multiple VMs, the eviction code benefits from being able to blindly
put pages without needing to know if there are any entities still
holding on to those pages. As such it's preferable to return the -EBUSY
before the BUG.
Eviction code is the only user for now, but overall it makes sense
anyway, IMO.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
For now, objects will maintain the same cache levels amongst all address
spaces. This is to limit the risk of bugs, as playing with cacheability
in the different domains can be very error prone.
In the future, it may be optimal to allow setting domains per VMA (ie.
an object bound into an address space).
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This represents the first half of hooking up VMs to execbuf. Here we
basically pass an address space all around to the different internal
functions. It should be much more readable, and have less risk than the
second half, which begins switching over to using VMAs instead of an
obj,vm.
The overall series echoes this style of, "add a VM, then make it smart
later"
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Switch a BUG_ON to WARN_ON.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Make it aware of which domain it is bound into GGTT, or PPGTT.
While modifying the function, add a global gtt flag to the object
description. Global is more interesting than aliasing since aliasing is
the default.
v2: Access VMA directly for start/size instead of helpers (Daniel)
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Just some small cleanups, and a rename of vm->ggtt_vm requested by
Daniel.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
To verbalize it, one can say, "pin an object into the given address
space." The semantics of pinning remain the same otherwise.
Certain objects will always have to be bound into the global GTT.
Therefore, global GTT is a special case, and keep a special interface
around for it (i915_gem_obj_ggtt_pin).
v2: s/i915_gem_ggtt_pin/i915_gem_obj_ggtt_pin
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Do to the move active/inactive lists, it no longer makes sense to use
them for shrinking, since shrinking isn't VM specific (such a need may
also exist, but doesn't yet).
What we can do instead is use the global bound list to find all objects
which aren't active.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Earlier in the conversion sequence we attempted to quickly wedge in the
transitional interface as static inlines.
Now that we're sure these interfaces are sane, for easier debug and to
decrease code size (since many of these functions may be called quite a
bit), make them real functions
While at it, kill off the set_color interface. We'll always have the
VMA, or easily get to it.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
With an upcoming change to bind, to make checkpatch happy and keep the
code clean, we need to rework this code a bit.
This should have no functional impact.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Add the newline Chris requested.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Move all the similar address space (VM) initialization code to one
function. Until we have multiple VMs, there should only ever be 1 VM.
The aliasing ppgtt is a special case without it's own VM (since it
doesn't need it's own address space management).
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The default LLC age was changed:
commit 0d8ff15e9a
Author: Ben Widawsky <benjamin.widawsky@intel.com>
Date: Thu Jul 4 11:02:03 2013 -0700
drm/i915/hsw: Set correct Haswell PTE encodings.
On the surface it would seem setting a default age wouldn't matter
because all GEM BOs are aged similarly, so the order in which objects
are evicted would not be subject to aging. The current working theory as
to why this caused a regression though is that LLC is a bit special in
that it is shared with the CPU. Presumably (not verified) the CPU
fetches cachelines with age 3, and therefore recently cached GPU objects
would be evicted before similar CPU object first when the LLC is full.
It stands to reason therefore that this would negatively impact CPU
bound benchmarks - but those seem to be low on the priority list.
eLLC OTOH does not have this same property as LLC. It should be used
entirely for the GPU, and so the age really shouldn't matter.
Furthermore, we have no evidence to suggest one is better than another
on eLLC. Since we've never properly supported eLLC before no, there
should be no regression. If the GPU client really wants "younger"
objects, they should use MOCS.
v2: Drop the extra #define (Chad)
v3: Actually git add
v4: Pimped commit message
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67062
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
