Hey Damien,
> I already got a fair review comment from Brad Volkin on this: he proposes to
> do this instead
>
> struct i915_hw_context {
> struct i915_address_space *vm;
> struct {
> struct drm_i915_gem_object *ctx_obj;
> struct intel_ringbuffer *ringbuf;
> } engine[I915_MAX_RINGS];
> ...
> };
>
> This is: instead of creating extra contexts with the same Context ID, modify the
> current i915_hw_context to work with all engines. I agree this alternative looks
> less *hackish*, but I want to get eyes on it (several things need careful
> consideration if we do this, e.g.: should the hang_stats also be per-engine?)
After looking at the execlists code, does this also make sense to you?
Cheers,
Oscar
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
