[Mlir-commits] [mlir] [mlir][ArmSME] Use liveness information in the tile allocator (PR #90448)

[Benjamin Maxwell]

================
@@ -100,13 +142,29 @@ func.func @non_overlapping_branches(%cond: i1) {
 
 // -----
 
-// Incorrect result! Everything assigned to tile 0 (which means values that are still live are overwritten).
-//
-// CHECK-BAD-LABEL: @constant_loop_init_with_multiple_users
-// CHECK-BAD: arm_sme.zero {tile_id = 0 : i32} : vector<[4]x[4]xf32>
-// CHECK-BAD: arm_sme.move_vector_to_tile_slice {{.*}} {tile_id = 0 : i32} : vector<[4]xf32> into vector<[4]x[4]xf32>
-// CHECK-BAD: arm_sme.move_vector_to_tile_slice {{.*}} {tile_id = 0 : i32} : vector<[4]xf32> into vector<[4]x[4]xf32>
+// Here %vecA and %vecB are not merged into the same live range (as they are unknown values).
+// This means that %vecA and %vecB are both allocated to different tiles (which is not legal).
+
+// expected-note at below {{tile operand is: <block argument> of type 'vector<[4]x[4]xf32>'}}
+func.func @overlapping_branches(%cond: i1, %vecA: vector<[4]x[4]xf32>, %vecB: vector<[4]x[4]xf32>) {
+  // expected-error at below {{op tile operand allocated to different SME virtial tile (move required)}}
+  %tile = scf.if %cond -> vector<[4]x[4]xf32> {
+    scf.yield %vecA : vector<[4]x[4]xf32>
+  } else {
+    scf.yield %vecB : vector<[4]x[4]xf32>
+  }
+  "test.some_use"(%tile) : (vector<[4]x[4]xf32>) -> ()
+  return
+}
+
+// -----
+
+// CHECK-LABEL: @constant_loop_init_with_multiple_users
----------------
MacDue wrote:

There's nothing partially interesting about the live ranges here (so I omitted them). I just want to check the allocation. 

https://github.com/llvm/llvm-project/pull/90448