regression test for relaxed memory order/missing barrier?

[Michael Haberler]

hi,

I'd appreciate hints for a regression test which can reliably trigger on missing barriers, in particular on multicore ARM processors

maybe something like the Dekker's algorithm cited on page 78 ? :http://www.engbloms.se/presentations/engblom-esc2007-slides-as-shown.pdf

thanks in advance,

- Michael

background: we're porting an multithreaded application from x86 to ARM multicores, and need to plug memory ordering issues without killing performance