On Thu, 22 Nov 2012 11:13:04 +0000, Chris Wilson <chris at chris-wilson.co.uk> wrote: > On Thu, 22 Nov 2012 12:51:29 +0200, Mika Kuoppala <mika.kuoppala at intel.com> wrote: > > If seqno already exists, add request with that seqno. Otherwise > > allocate new seqno for this request. Make i915_gem_check_olr use > > already allocated seqnos in order to avoid nesting into > > i915_gem_alloc_seqno during i915_gem_idle(). > > I really don't like that mess in i915_add_request(). I was thinking more > alongthe lines of: > > http://cgit.freedesktop.org/~ickle/linux-2.6/commit/?h=bug55984&id=87cf35086182f9ce36bb8d85a0fe4cfc136cb3e2 > -Chris Yours look much better. I got distracted by intel_ring_begin can't fail in do_execbuffer. With quick testing this also can survive hundreds of wraps. Please send yours for review. Thanks, -Mika > -- > Chris Wilson, Intel Open Source Technology Centre > _______________________________________________ > Intel-gfx mailing list > Intel-gfx at lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
