RFC: SIMD pragma independent of Cilk Plus / OpenMPv4

Torvald Riegel triegel@redhat.com
Wed Sep 11 07:02:00 GMT 2013


On Mon, 2013-09-09 at 10:18 -0400, Tim Prince wrote:
> Current Intel implementations of safelen will fail to vectorize and give 
> notice if the value is set unnecessarily large.  It's been agreed that 
> increasing the safelen value beyond the optimum level should not turn 
> off vectorization.

Yes, it definitely shouldn't.  safelen() or similar constructs are
supposed to be a programmer assertion about correctness, not a
performance hint.

Torvald



More information about the Gcc mailing list