TSX lock elision for glibc v2

[Torvald Riegel]

On Thu, 2013-01-10 at 12:19 -0800, Andi Kleen wrote:
> - Internal locks in glibc (like malloc or stdio) do not elide at this
>   point.

Perhaps this would in fact be an easier target for a first use of lock
elision?  There's likely more knowledge about the workload (e.g., you
know how long the critical sections roughly are, compared to when trying
to tune for application-side critical sections with unknown
characteristics), and we don't have any issues with having to maintain
certain lock interface semantics.

Have you started to look at potential candidates in detail yet?


Torvald