[Mesa-dev] [PATCH 1/7] i965/bufmgr: Explicitly wait instead of using I915_GEM_SET_DOMAIN.

[Chris Wilson]

Quoting Kenneth Graunke (2017-07-22 00:17:41)
> With the advent of asynchronous maps, domain tracking doesn't make a
> whole lot of sense.  Buffers can be in use on both the CPU and GPU at
> the same time.  In order to avoid blocking, we stopped using set_domain
> for asynchronous mappings, which means that the kernel's tracking has
> lies.  We can't properly track it in userspace either, as the kernel
> can change domains on us spontaneously (for example, when un-swapping).
> 
> I915_GEM_SET_DOMAIN combines two aspects: cache flushing, and waiting
> for a buffer to be idle.  In order to stop using it, we'll need to do
> clflushing ourselves, and use I915_GEM_WAIT to wait on buffers.
> 
> For cache-coherent buffers (most on LLC systems), we don't need to do
> any clflushing - the CPU and GPU views are coherent.  For non-coherent
> buffers (most on non-LLC systems), we currently only use the CPU for
> read-only maps, and we explicitly clflush when necessary.
> 
> One downside to this is that we'll stall unnecessarily if we do a
> read-only mapping of a buffer that the GPU is reading.  I believe
> that's pretty uncommon.

Pretty please add some of the warning comments about GTT from earlier.
For bonus paranoia, the *gtt_map trick as well.  :)

You should be able demonstrate the issue if you do
*brw_bo_map_gtt(bo) = 1;
assert(*brw_bo_map_cpu(bo) == 1);

Can we add unit tests to this file?
-Chris