GNOME Bugzilla – Bug 755171
libs: initializes miniobject memory to zero
Last modified: 2015-10-19 16:29:12 UTC
This patch replaces the call to g_slice_alloc() to gst_slice_alloc0(), which initializes the allocated memory to 0, making it safer to use. Doing a perf measurement with a simple H264 1080p video, the overhead of this call goes from 0.05% to 0.09%
Created attachment 311572 [details] [review] libs: initializes miniobject memory to zero
(In reply to Víctor Manuel Jáquez Leal from comment #0) > This patch replaces the call to g_slice_alloc() to gst_slice_alloc0(), > which initializes the allocated memory to 0, making it safer to use. > > Doing a perf measurement with a simple H264 1080p video, the overhead of > this call goes from 0.05% to 0.09% Just curious, did you hit any bug related with this?
(In reply to sreerenj from comment #2) > (In reply to Víctor Manuel Jáquez Leal from comment #0) > > This patch replaces the call to g_slice_alloc() to gst_slice_alloc0(), > > which initializes the allocated memory to 0, making it safer to use. > > > > Doing a perf measurement with a simple H264 1080p video, the overhead of > > this call goes from 0.05% to 0.09% > > Just curious, did you hit any bug related with this? Just as a generalization of bug 754845
As the main advantage of gstvaapiminiobject is the reduction of overhead compared with GObject and GstMiniObject, this patch would make no sense. I'm closing it as not a bug.