[PATCH] D136661: [InstCombine] Fold series of instructions into mull for more types

Paweł Bylica via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Oct 25 00:05:31 PDT 2022


chfast requested changes to this revision.
chfast added inline comments.
This revision now requires changes to proceed.


================
Comment at: llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp:1275
+  // Skip the odd bitwidth types.
+  if ((BitWidth & 0x1))
     return nullptr;
----------------



================
Comment at: llvm/test/Transforms/InstCombine/mul_full_64.ll:451
 
-
+; Negative case: https://alive2.llvm.org/ce/z/jH4kU7
 define i64 @mullo(i64 %x, i64 %y) {
----------------
There is typo in the link: last instruction should be `add` not `and`. https://alive2.llvm.org/ce/z/y26zaW
Please update the comment to TODO if we don't plan to match the `or` pattern right now.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D136661/new/

https://reviews.llvm.org/D136661



More information about the llvm-commits mailing list