https://community.intel.com/t5/Intel-Moderncode-for-Parallel/About-lockfree-mpmc/m-p/809744#M940 <BR />Hello again,<BR /><BR /><BR />Sorry, as you have noticied there is loads on<BR />setObject(lastTail,tm) between line [2] and line [4].<BR /><BR />And loads are not reordered with older loads on x86<BR /><BR />So i think there is no problem with lockfree_mpmc.<BR /><BR /><BR />Thank you.<BR />Amine Moulay Ramdne<BR /><BR /> Sun, 03 Jun 2012 19:23:02 GMT aminer10 2012-06-03T19:23:02Z https://community.intel.com/t5/Intel-Moderncode-for-Parallel/About-lockfree-mpmc/m-p/809742#M938 <BR />Hello,<BR /><BR />I think i have forgot something with lockfree_mpmc...<BR /><BR />Please follow with me... <BR /><BR />The Intel x86 memory model, detailed in Intel 64 Architecture Memory Ordering White Paper <BR />and the AMD spec, AMD64 Architecture Programmers Manual, list a lot of memory ordering <BR />guarantees, among them:<BR /><BR />Loads are not reordered with other loads. <BR />Stores are not reordered with other stores. <BR />Stores are not reordered with older loads. <BR />In a multiprocessor system, memory ordering obeys causality (memory ordering respects transitive visibility). <BR />In a multiprocessor system, stores to the same location have a total order. <BR />In a multiprocessor system, locked instructions have a total order. <BR />Loads and stores are not reordered with locked instructions. <BR /><DIV><BR />But since on x86 Loads may be reordered with older stores to different locations <BR /><BR /><BR />So please take a look at the following lockfree_mpmc code: <BR /><BR />---<BR /><BR />function TLockfree_MPMC.push(tm : tNodeQueue):boolean;<BR />var lasttail,newtemp:long;<BR />i,j:integer;<BR />begin<BR /><BR />if getlength &gt;= fsize <BR /> then <BR /> begin<BR /> result:=false;<BR /> exit;<BR /> end; <BR />result:=true;<BR /><BR />[1]newTemp:=LockedIncLong(temp);<BR /><BR />[2] lastTail:=newTemp-1;<BR />setObject(lastTail,tm);<BR /><BR />[3]<BR /><BR />repeat<BR /><BR />[4] if CAS(tail,lasttail,newtemp) <BR /> then <BR /> begin<BR /> exit; <BR /> end;<BR />sleep(0);<BR />until false;<BR /><BR />end;<BR />---<BR /><BR /><BR />So the the loads of lasttail and newtemp in [4] line can be reordered <BR />with older stores of lastail and newtemp in [1] [2], so in this case must <BR />i insert an mfence in line [3] to respect the logic of the program?...<BR /><BR /></DIV><DIV></DIV><DIV>Thank you,<BR />Amine Moulay Ramdane.</DIV><DIV><BR /><BR /></DIV> Sun, 03 Jun 2012 18:59:41 GMT https://community.intel.com/t5/Intel-Moderncode-for-Parallel/About-lockfree-mpmc/m-p/809742#M938 aminer10 2012-06-03T18:59:41Z https://community.intel.com/t5/Intel-Moderncode-for-Parallel/About-lockfree-mpmc/m-p/809743#M939 <SPAN style="font-size: x-small;"><P><BR /><BR />Hello again,<BR /><BR /><BR />Sorry, as you have noticied there is loads on<BR />setObject(lastTail,tm)between line [2] and line [4].<BR /><BR />And loads are not reordered with older loads on x86 <BR /><BR />So i think there is no problemwith lockfree_mpmc.<BR /><BR /><BR />Thank you.<BR />Amine Moulay Ramdne<BR /><BR /><BR /><BR /></P></SPAN> Sun, 03 Jun 2012 19:19:13 GMT https://community.intel.com/t5/Intel-Moderncode-for-Parallel/About-lockfree-mpmc/m-p/809743#M939 aminer10 2012-06-03T19:19:13Z https://community.intel.com/t5/Intel-Moderncode-for-Parallel/About-lockfree-mpmc/m-p/809744#M940 <BR />Hello again,<BR /><BR /><BR />Sorry, as you have noticied there is loads on<BR />setObject(lastTail,tm) between line [2] and line [4].<BR /><BR />And loads are not reordered with older loads on x86<BR /><BR />So i think there is no problem with lockfree_mpmc.<BR /><BR /><BR />Thank you.<BR />Amine Moulay Ramdne<BR /><BR /> Sun, 03 Jun 2012 19:23:02 GMT https://community.intel.com/t5/Intel-Moderncode-for-Parallel/About-lockfree-mpmc/m-p/809744#M940 aminer10 2012-06-03T19:23:02Z