[PATCH] D39976: [AArch64] Consider the cost model when folding loads and stores

[Jun Bum Lim via Phabricator via llvm-commits]

junbuml added a comment.

For me, the heuristic seems pretty reasonable, but I'm not fully sure if this can make a right decision target-independently.   Someone else who have better idea in hardware level may need to review this.



================
Comment at: llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp:691
+  long UopN = SCN->NumMicroOps,
+       UopA = TSM.getNumMicroOps(&MIA, SCA),
+       UopB = TSM.getNumMicroOps(&MIB, SCB);
----------------
Isn't it possible to do just TSM.getNumMicroOps(&MIA)   and remove  above  *SCA = TSM.resolveSchedClass(&MIA) in line 678.


================
Comment at: llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp:707
+  else if (UopA <= 1 && UopB <= 1) {
+    if (LatDif < -UopDif) {
+      DEBUG(dbgs() << "Evaluating instructions: replacement is faster "
----------------
I'm not sure if it's okay to compare values from  computeInstrLatency() and getNumMicroOps() ? 


Repository:
  rL LLVM

https://reviews.llvm.org/D39976