[PATCH] D25020: [ARM] Fix 26% performance regression on Cortex-A9 caused by not using VMLA/VMLS

[Diana Picus via llvm-commits]

rovka added a comment.

Thanks for getting the performance results.
What about the vector support?



================
Comment at: lib/Target/ARM/ARMISelDAGToDAG.cpp:443
+    break;
+  }
+
----------------
This is now checking only that the node can be lowered to VMLx. What happened to the part checking if forwarding can be used (i.e. mac following multiply or mac) and all the other checks?


https://reviews.llvm.org/D25020