On Thu, Mar 05, 2015 at 09:51:22PM -0500, Vivek Kasireddy wrote: > If there are multiple EGL surfaces associated with one EGL context, for > any given surface, we unnecessarily have to wait for swapbuffers to > complete on other EGL surfaces because we throttle to the context. > By moving first_post_swapbuffers_batch from the brw context into > intel_renderuffer, we would no longer be throttled by other surface's > swapbuffers thereby leading to better performance.
Since you don't actually flush between drawables or track all drawables in a batch, nor is there any guarantee on the progression of render targets, this is not equivalent at all to the previous throttling scheme, and could trivially prevent throttling at all (which may be counter-intuitative but lead to lower throughput as well as hurt latency). Meanwhile it still succumbs to all the shortcomings of the previous scheme, since often it is the same render target being used over and over again.. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev