drm/i915: Replace execbuf vma ht with an idr
authorChris Wilson <chris@chris-wilson.co.uk>
Wed, 16 Aug 2017 08:52:08 +0000 (09:52 +0100)
committerChris Wilson <chris@chris-wilson.co.uk>
Fri, 18 Aug 2017 10:59:02 +0000 (11:59 +0100)
commitd1b48c1e7184d9bc4ae6d7f9fe2eed9efed11ffc
treecbb1cb9f87c16ba6c17cbd47279979c823555d70
parent170fa29b14fadf2deb361589cefe6a78b21b1b22
drm/i915: Replace execbuf vma ht with an idr

This was the competing idea long ago, but it was only with the rewrite
of the idr as an radixtree and using the radixtree directly ourselves,
along with the realisation that we can store the vma directly in the
radixtree and only need a list for the reverse mapping, that made the
patch performant enough to displace using a hashtable. Though the vma ht
is fast and doesn't require any extra allocation (as we can embed the node
inside the vma), it does require a thread for resizing and serialization
and will have the occasional slow lookup. That is hairy enough to
investigate alternatives and favour them if equivalent in peak performance.
One advantage of allocating an indirection entry is that we can support a
single shared bo between many clients, something that was done on a
first-come first-serve basis for shared GGTT vma previously. To offset
the extra allocations, we create yet another kmem_cache for them.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170816085210.4199-5-chris@chris-wilson.co.uk
drivers/gpu/drm/i915/i915_debugfs.c
drivers/gpu/drm/i915/i915_drv.h
drivers/gpu/drm/i915/i915_gem.c
drivers/gpu/drm/i915/i915_gem_context.c
drivers/gpu/drm/i915/i915_gem_context.h
drivers/gpu/drm/i915/i915_gem_execbuffer.c
drivers/gpu/drm/i915/i915_gem_object.h
drivers/gpu/drm/i915/i915_vma.c
drivers/gpu/drm/i915/i915_vma.h
drivers/gpu/drm/i915/selftests/mock_context.c
lib/radix-tree.c