[PATCH] Add limitations to loop convert user doc
Sean Silva
silvas at purdue.edu
Tue Mar 19 20:11:26 PDT 2013
Cool, I've been wondering when this would get incorporated.
================
Comment at: docs/LoopConvertTransform.rst:41-42
@@ -43,1 +40,4 @@
+See :ref:`IteratorLimitations` for an example of an incorrect transformation
+when the maximum acceptable risk level is set to `risky`.
+
----------------
"an incorrect transformation"? That section has multiple. Which one are you referring to?
================
Comment at: docs/LoopConvertTransform.rst:134-136
@@ -120,1 +133,5 @@
+Limitations
+===========
+
+Comments
----------------
Please add some introductory text explaining to the user what they can expect to learn by reading this section, and why they would want to read it.
================
Comment at: docs/LoopConvertTransform.rst:149-150
@@ +148,4 @@
+
+Iterators
+^^^^^^^^^
+
----------------
This section title is too generic. Maybe "risky loop invariance assumptions"? (and then start the section by saying that "if -risk=risky, then transformations are done assuming...")
================
Comment at: docs/LoopConvertTransform.rst:221-223
@@ +220,5 @@
+
+The following transformation is performed at the ``-risk=safe`` level. If we had
+directly used the container instead of the pointer or reference, the
+transformation would have only been applied at the ``-risk=risky`` level.
+
----------------
Can you give the user a bit more insight into why?
================
Comment at: docs/LoopConvertTransform.rst:74-75
@@ -71,1 +73,4 @@
+ for (iterator it = container.begin(), e = container.end(); it != e; it++)
+ cout << *it;
+
----------------
Please put this and the other small cleanups/additions in a separate path (you can commit it directly). No need to sneak them in this patch.
http://llvm-reviews.chandlerc.com/D552
More information about the cfe-commits
mailing list