]> git.karo-electronics.de Git - karo-tx-linux.git/commit
drm/i915: Invariably invalidate before ctx switch
authorBen Widawsky <benjamin.widawsky@intel.com>
Thu, 3 Apr 2014 05:30:23 +0000 (22:30 -0700)
committerDaniel Vetter <daniel.vetter@ffwll.ch>
Thu, 3 Apr 2014 09:41:39 +0000 (11:41 +0200)
commit057f6a8ad7691d028681ba443d7f5e482a6a80a4
tree1bf5ee2145c24f5a031e125df7366876d5ce1351
parenteb6008ad30fe51b5be36d2ccce2c9c0409ac88a4
drm/i915: Invariably invalidate before ctx switch

We have been setting the bit which was originally BIOS dependent since:
commit f05bb0c7b624252a5e768287e340e8e45df96e42
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Sun Jan 20 16:33:32 2013 +0000

    drm/i915: GFX_MODE Flush TLB Invalidate Mode must be '1' for scanline waits

Therefore, we do not need to try to figure it out dynamically and we can
just always invalidate the TLBs.

It's a partial revert of:
commit 12b0286f49947a6cdc9285032d918466a8c3f5f9
Author: Ben Widawsky <ben@bwidawsk.net>
Date:   Mon Jun 4 14:42:50 2012 -0700

    drm/i915: possibly invalidate TLB before context switch

The original commit attempted to only invalidate when necessary
(very much a relic from the old days). Now, we can just always invalidate.

I guess the old TODO still exists. Since we seem to have abandoned ILK
contexts however, there isn't much point in even remembering.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
drivers/gpu/drm/i915/i915_gem_context.c
drivers/gpu/drm/i915/intel_ringbuffer.c
drivers/gpu/drm/i915/intel_ringbuffer.h