[Intel-gfx] [PATCH 11/66] drm/i915: Preallocate stashes for vma page-directories

[Thomas Hellström (Intel)]

On 7/28/20 4:42 PM, Chris Wilson wrote:
> Quoting Thomas Hellström (Intel) (2020-07-23 15:33:20)
>> On 2020-07-15 13:50, Chris Wilson wrote:
>>> We need to make the DMA allocations used for page directories to be
>>> performed up front so that we can include those allocations in our
>>> memory reservation pass. The downside is that we have to assume the
>>> worst case, even before we know the final layout, and always allocate
>>> enough page directories for this object, even when there will be overlap.
>>> This unfortunately can be quite expensive, especially as we have to
>>> clear/reset the page directories and DMA pages, but it should only be
>>> required during early phases of a workload when new objects are being
>>> discovered, or after memory/eviction pressure when we need to rebind.
>>> Once we reach steady state, the objects should not be moved and we no
>>> longer need to preallocating the pages tables.
>>>
>>> It should be noted that the lifetime for the page directories DMA is
>>> more or less decoupled from individual fences as they will be shared
>>> across objects across timelines.
>>>
>>> v2: Only allocate enough PD space for the PTE we may use, we do not need
>>> to allocate PD that will be left as scratch.
>>> v3: Store the shift unto the first PD level to encapsulate the different
>>> PTE counts for gen6/gen8.
>>>
>>> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
>>> Cc: Matthew Auld <matthew.auld at intel.com>
>>> ---
>>>    .../gpu/drm/i915/gem/i915_gem_client_blt.c    | 11 +--
>>>    drivers/gpu/drm/i915/gt/gen6_ppgtt.c          | 40 ++++-----
>>>    drivers/gpu/drm/i915/gt/gen8_ppgtt.c          | 78 +++++------------
>>>    drivers/gpu/drm/i915/gt/intel_ggtt.c          | 60 ++++++--------
>>>    drivers/gpu/drm/i915/gt/intel_gtt.h           | 46 ++++++----
>>>    drivers/gpu/drm/i915/gt/intel_ppgtt.c         | 83 ++++++++++++++++---
>>>    drivers/gpu/drm/i915/i915_vma.c               | 27 +++---
>>>    drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 60 ++++++++------
>>>    drivers/gpu/drm/i915/selftests/mock_gtt.c     | 22 ++---
>>>    9 files changed, 237 insertions(+), 190 deletions(-)
>> Hi, Chris,
>>
>> Overall looks good, but a question: Why can't we perform page-table
>> memory allocation on demand when needed?
> We need to allocate device memory for the page tables. The intention
> here is gather up all the resource requirements for an operation and
> reserve them in a single pass.
>   
>> Are we then under a mutex that we also take during reclaim?
> Yes, the vm->mutex is used during the shrinker to revoke the GPU
> bindings before returning memory to the system.
> -Chris

Reviewed-by: Thomas Hellström <thomas.hellstrom at intel.com>