[PATCH] D45982: [TTI, AArch64] Add transpose shuffle kind

[Matthew Simpson via Phabricator via llvm-commits]

mssimpso added inline comments.


================
Comment at: include/llvm/Analysis/TargetTransformInfo.h:642
   enum ShuffleKind {
-    SK_Broadcast,       ///< Broadcast element 0 to all other elements.
-    SK_Reverse,         ///< Reverse the order of the vector.
-    SK_Alternate,       ///< Choose alternate elements from vector.
-    SK_InsertSubvector, ///< InsertSubvector. Index indicates start offset.
-    SK_ExtractSubvector,///< ExtractSubvector Index indicates start offset.
-    SK_PermuteTwoSrc,   ///< Merge elements from two source vectors into one
-                        ///< with any shuffle mask.
-    SK_PermuteSingleSrc ///< Shuffle elements of single source vector with any
-                        ///< shuffle mask.
+    SK_Broadcast,        ///< Broadcast element 0 to all other elements.
+    SK_Reverse,          ///< Reverse the order of the vector.
----------------
javed.absar wrote:
> Not sure why indentation change is needed
Ah, clang-format decided to do that. I will undo the formatting for now.


================
Comment at: lib/Analysis/TargetTransformInfo.cpp:707
+  // matrix is stored in two result vectors such that one vector contains
+  // elements from all the even-numbered rows and the other vector contains
+  // elements from all the odd-numbered rows. For example, a 2x4 matrix can be
----------------
javed.absar wrote:
> would it be more accurate to say '..such that one vector contains _interleaved_ elements from all the even numbered rows ..."
Sounds good to me.


================
Comment at: lib/Analysis/TargetTransformInfo.cpp:737
+
+  // 4. The difference between consecutive even-numbered elements must be equal
+  // to two.
----------------
javed.absar wrote:
> Test 4 and 5 could be perhaps combined into one loop?
Yes, that would probably be a little more straightforward. Thanks!


Repository:
  rL LLVM

https://reviews.llvm.org/D45982