[llvm] [X86][AMX] Combine constant zero vector and AMX cast to tilezero (PR #92384)

[Simon Pilgrim via llvm-commits]

================
@@ -1198,6 +1219,14 @@ bool X86LowerAMXCast::combineLdSt(SmallVectorImpl<Instruction *> &Casts) {
       for (auto *Store : DeadStores)
         Store->eraseFromParent();
     } else { // x86_cast_vector_to_tile
+      //  %19 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> zeroinitializer)
+      //  -->
+      //  %19 = tail call x86_amx @llvm.x86.tilezero.internal(i16 %row, i16 %col)
+      if (dyn_cast<ConstantAggregateZero>(Cast->getOperand(0))) {
----------------
RKSimon wrote:

dyn_cast<> -> isa<>

https://github.com/llvm/llvm-project/pull/92384