169984 2017-03-22 17:48:49 -0700 [jsc][mips] study whether using SYNC_MB for memoryFence/storeFence works and improves performances 2021-03-05 09:46:48 -0800 1 1 1 Unclassified WebKit JavaScriptCore Other Other Linux NEW P2 Normal --- 1 guijemont webkit-unassigned 3bq8_bay oldest_to_newest 1290283 0 guijemont 2017-03-22 17:48:49 -0700 For MacroAssemblerMIPS::memoryFence() and ::storeFence(), we use a sync instruction with SYNC (0) in the stype field. We should check whether the optional SYNC_MB (0x10) stype can work for our needs. It is defined as forcing the order of memory accesses, but unlike SYNC, SYNC_MB does not require all older instructions to be completed when the sync instruction completes, but only that they reach the load/store ordering point before the sync instruction completes, which is a slightly lighter constraint that should still be good enough for us, while potentially having a lower performance hit. 1290286 1 guijemont 2017-03-22 17:51:02 -0700 See also the discussion in https://bugs.webkit.org/show_bug.cgi?id=169705