[llvm] [InterleavedAccessPass] Get round the unsupported large scalarize vectors (PR #88643)

[Paul Walker via llvm-commits]

================
@@ -16433,7 +16433,9 @@ bool AArch64TargetLowering::lowerDeinterleaveIntrinsicToLoad(
   if (UseScalable && !VTy->isScalableTy())
     return false;
 
-  unsigned NumLoads = getNumInterleavedAccesses(VTy, DL, UseScalable);
+  // Get around for the missing of split large size scalarize vectors.
+  // TODO: Support the custom split of large size scalarize vectors
+  unsigned NumLoads = getNumInterleavedAccesses(VTy, DL, false);
----------------
paulwalker-arm wrote:

Sorry, I should have been clearer.  I meant the code within `getNumInterleavedAccesses()` needs to be changed because that is where the bug exists.

https://github.com/llvm/llvm-project/pull/88643