[PATCH] D13779: [Polly] Full/partial tile separation for vectorization

[Roman via llvm-commits]

gareevroman added a comment.

Thank you for the comments!


================
Comment at: lib/Transform/ScheduleOptimizer.cpp:193
@@ +192,3 @@
+/// 2. Extend it to a set, which has exactly VectorWidth iterations for
+///    an every prefix from the previous step.
+/// 3. Subtract loop domain from it, project out the vector loop dimension and
----------------
grosser wrote:
> "an every" ?  
Is "for any" better?

================
Comment at: lib/Transform/ScheduleOptimizer.cpp:210
@@ +209,3 @@
+  ExtentConstraints = isl_set_drop_constraints_involving_dims(
+      ExtentConstraints, isl_dim_param, 0, DimsParam);
+  ExtentConstraints =
----------------
jdoerfert wrote:
> As Tobias once told me, drop_constraints_* is a dangerous function and it is usally not what we want. I do not fully understand why it is needed here but maybe we can use project_out or nothing?
I’ve tried to get rid of this by explicit allocation of constraints in a new version of the patch.

================
Comment at: lib/Transform/ScheduleOptimizer.cpp:229
@@ +228,3 @@
+
+static __isl_give isl_schedule_node *
+isolateFullPartialTiles(__isl_take isl_schedule_node *Node) {
----------------
jdoerfert wrote:
> Can we make this a (static) member of the optimizer class? I would probably use it without the scheduling pass at some point.
Should we make this a public static member?


http://reviews.llvm.org/D13779