================
Comment at: lib/CodeGen/CGStmtOpenMP.cpp:67
@@ +66,3 @@
+ if (SeparateIter) {
+ // TODO: Update lastprivates if the SeparateIter flag is true.
+ }
----------------
Is there something missing for you to support this? Looks like everything is in
place below.
================
Comment at: lib/Sema/SemaOpenMP.cpp:2534
@@ +2533,3 @@
+ //
+ // #define NumIterations (NI * ((NJ - J0 - 1 + 2) / 2))
+ // for (decltype(NumIterations) IV = 0; IV < NumIterations; ++IV ) {
----------------
The OpenMP specification says:
The integer type (or kind, for Fortran) used to compute the iteration count
for the collapsed loop is implementation defined.
And I'm afraid that, especially if the iteration variables are originally
32-bits, with the current scheme we'll overflow when calculating the total
number of iterations. Can we always use a larger integer type?
http://reviews.llvm.org/D5184
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
