Tom, your understanding is correct per OpenMP SIMD model. Our implementation behaves as you stated.
Which is not part of LLVM main trunk yet.
Is that the implementation that is based on the intrinsics in the RFC you and Hal Finkel had sent out to the list? Or is it a different implementation (and if so, is there some plan to merge the two)?
// Interim Head, Department of Computer Science
// Donald B. Gillies Professor of Computer Science
// University of Illinois at Urbana-Champaign
// Admin Assistant: Amanda Foley - email@example.com
// Google Hangouts: firstname.lastname@example.org || Skype: vikramsadve
// Research page: http://vikram.cs.illinois.edu/>