RFC: S/390 Transactional memory support - save/restore of FPRs

[Richard Henderson]

On 05/22/2013 02:23 AM, Andreas Krebbel wrote:
> Mmmh ok. Where do you think the other approach (clobber in abort code + abnormal edge) could break?


It's mostly about the infrastructure of maintaining the edges.  It's quite a
lot of code to maintain normal EH + TM edges.  We don't even bother maintaining
edges for setjmp/longjmp.  There's really no infrastructure set up for special
backend abnormal edges.

And since we hope that everyone will be using the language-level TM, it seems
like a waste to implement those special backend edges.



r~