use syrk for boolean dtypes when possible

Author: Vahid Tavanashad

dpnp/dpnp_utils/dpnp_utils_linearalgebra.py

@@ -986,7 +986,10 @@ def dpnp_multiplication(
         if _is_syrk_compatible(x1, x2):
             call_flag = "syrk"
             res_dtype_orig = res_dtype
-            if dpnp.issubdtype(res_dtype, dpnp.integer):
+            # for exact dtypes, use syrk implementation unlike general approach
+            # where dpctl implementation is used for exact dtypes for better
+            # performance
+            if not dpnp.issubdtype(res_dtype, dpnp.inexact):
                 res_dtype = dpnp.default_float_type(x1.device)
     elif x1_base_is_1D:
         # TODO: implement gemv_batch to use it here with transpose

dpnp/tests/test_product.py

@@ -1171,7 +1171,7 @@ def test_special_case(self, dt_out, shape1, shape2):
         result = dpnp.matmul(ia, ib, out=iout)
         assert_dtype_allclose(result, expected)
 
-    @pytest.mark.parametrize("dt", get_all_dtypes())
+    @pytest.mark.parametrize("dt", get_all_dtypes(no_none=True))
     def test_syrk(self, dt):
         a = generate_random_numpy_array((6, 9), dtype=dt, low=-5, high=5)
         ia = dpnp.array(a)