On Tue, Apr 04, 2017 at 02:27:25PM +1000, Dave Airlie wrote: > This series enhances my previous semaphore work on for amdgpu, > with a generic DRM sync object. (drm_syncobj). > > It first enhances sync_file to have a type/flags so we can have > different semantics for different sync files, and a wait > to retrieve the type of sync_file for userspace. > > Then it adds drm sync objects which are just a drm wrapper around > a sync_file object, allowing creation/info/destroy and import/export > of the objects. > > Next it enhances sync_file to have semaphore semantics for Vulkan. > > Finally it adds amdgpu support to it's command submission paths to > use the new code. > > I've hopefully fixed up the things pointed out in the last > review of the sync_file fence changes, I do wonder if we should > just block poll on semaphore objects as currently I've no use > case for this. Looking pretty good, some minor polish and ack from Jason that this seems ok for anv too. Pls also add igt testcases for some of the silly cases (trying to import a non-sync_file fd, bad handle for export, stuff like that). Cheers, Daniel -- Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/dri-devel
