Quoting Chris Wilson (2019-01-02 09:41:37) > Now that we have allocated ourselves a cacheline to store a breadcrumb, > we can emit a write from the GPU into the timeline's HWSP of the > per-context seqno as we complete each request. This drops the mirroring > of the per-engine HWSP and allows each context to operate independently. > We do not need to unwind the per-context timeline, and so requests are > always consistent with the timeline breadcrumb, greatly simplifying the > completion checks as we no longer need to be concerned about the > global_seqno changing mid check. What I forgot about when using STORE_DWORD_IMM are the generations where that either explodes (snb + vcs) or where it only supports physical addresses (gen2 + early gen3). Sigh. One idea is to make the timeline use the real HWSP for gen2-6, still keeping the per-context timeline for gen7 so that one day we can do request shuffling. Alternatively, use physical address for gen2, and MI_FLUSH_DW for gen6/xcs. -Chris _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
