Enable test_optimize_thread_locality for CPU.

Signed-off-by: Ilya Enkovich <[email protected]>

python/test/unit/language/test_core.py

@@ -2468,6 +2468,7 @@ def histogram_kernel(x_ptr, z_ptr, M: tl.constexpr, N: tl.constexpr):
     assert (z_torch == z).all()
 
 
+@pytest.mark.cpu
 @pytest.mark.interpreter
 @pytest.mark.parametrize("op", ['sum', 'max', 'min'])
 @pytest.mark.parametrize("BLOCK_N", [32, 64, 128])
@@ -2513,7 +2514,7 @@ def kernel(X, Y, N, BLOCK_M: tl.constexpr, BLOCK_N: tl.constexpr, NUM_PID_N: tl.
     x = torch.randn((BLOCK_M, N), dtype=torch.float32, device=device)
     y = torch.randn((BLOCK_M, num_pid_n), dtype=torch.float32, device=device)
     h = kernel[(1, num_pid_n, 1)](x, y, N, BLOCK_M, BLOCK_N, NUM_PID_N=num_pid_n)
-    if not is_interpreter():
+    if not is_interpreter() and not is_cpu():
         assert h.asm['ttgir'].count(
             '"tt.reduce"') == 2, "tt.reduce should be called twice, otherwise the optimization didn't work"
     y_ref = numpy_op(x.cpu().numpy(), axis=1, keepdims=True)