]> git.karo-electronics.de Git - karo-tx-linux.git/commit
drm/i915: Only bind each object rather than for every execbuffer
authorDaniel Vetter <daniel.vetter@ffwll.ch>
Fri, 14 Feb 2014 13:01:21 +0000 (14:01 +0100)
committerDaniel Vetter <daniel.vetter@ffwll.ch>
Fri, 14 Feb 2014 13:18:38 +0000 (14:18 +0100)
commit8ea99c928787ba1712b7506b4c56c948c45d84b1
treee491adb718426b723cdd29fe2239f29b0501ab4e
parent262de1453184f65e5ccfe45790f93d41f7339d49
drm/i915: Only bind each object rather than for every execbuffer

One side-effect of the introduction of ppgtt was that we needed to
rebind the object into the appropriate vm (and global gtt in some
peculiar cases). For simplicity this was done twice for every object on
every call to execbuffer. However, that adds a tremendous amount of CPU
overhead (rewriting all the PTE for all objects into WC memory) per
draw. The fix is to push all the decision about which vm to bind into
and when down into the low-level bind routines through hints rather than
performing the bind unconditionally in the execbuffer routine.

Note that this is a regression introduced in the full ppgtt feature
branch, before this we've only done re-bound objects when the relevant
has_(aliasing_ppgtt|global_gtt)_mapping flag was clear. But since
that's per-object and not per-vma that optimization broke.

v2: Split out prep work and unrelated changes.

v3: Bring back functional change around PIN_GLOBAL that I've
accidentally split out.

v4: Remove the temporary hack for the old binding logic to avoid
bisection issues.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72906
Tested-by: jianx.zhou@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> (v1)
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
drivers/gpu/drm/i915/i915_gem.c
drivers/gpu/drm/i915/i915_gem_execbuffer.c