[NNPACK] Add check for NNPACK being available (`nnp_initialize()` succeeding) (#2119)

This fixes issues with failing tests on PowerPC.

python/tvm/contrib/nnpack.py

@@ -5,16 +5,11 @@ from .. import api as _api
 from .. import intrin as _intrin
 from .._ffi.function import _init_api
 
-def config(nthreads):
-    """Configure the nnpack library.
-
-    Parameters
-    ----------
-    nthreads : int
-        The threads number of nnpack thread pool, must be a nonnegative.
-
+def is_available():
+    """Check whether NNPACK is available, that is, `nnp_initialize()`
+    returns `nnp_status_success`.
     """
-    _Config(nthreads)
+    return _initialize() == 0
 
 def fully_connected_inference(lhs, rhs, nthreads=1):
     """Create an extern op that compute fully connected of 1D tensor lhs and

src/contrib/nnpack/nnpack_utils.cc

@@ -38,9 +38,10 @@ bool NNPackConfig(uint64_t nthreads) {
 }
 
 
-TVM_REGISTER_GLOBAL("contrib.nnpack._Config")
+TVM_REGISTER_GLOBAL("contrib.nnpack._initialize")
 .set_body([](TVMArgs args, TVMRetValue *ret) {
-    CHECK(NNPackConfig(args[0]));
+    *ret = nnp_initialize();
   });
+
 }  // namespace contrib
 }  // namespace tvm

tests/python/contrib/test_nnpack.py

@@ -21,7 +21,9 @@ def test_fully_connected_output():
         if not tvm.get_global_func("tvm.contrib.nnpack.fully_connected_output", True):
             print("skip because extern function is not available")
             return
-        return
+        if not nnpack.is_available():
+            return
+
         ctx = tvm.cpu(0)
         f = tvm.build(s, [A, B, D, bias], target)
         a = tvm.nd.array(np.random.uniform(size=(n, l)).astype(A.dtype), ctx)
@@ -52,7 +54,9 @@ def test_fully_connected_inference():
         if not tvm.get_global_func("tvm.contrib.nnpack.fully_connected_inference", True):
             print("skip because extern function is not available")
             return
-        return
+        if not nnpack.is_available():
+            return
+
         ctx = tvm.cpu(0)
         f = tvm.build(s, [A, B, D, bias], target)
         a = tvm.nd.array(np.random.uniform(size=(l)).astype(A.dtype), ctx)
@@ -130,7 +134,9 @@ def test_convolution_inference():
         if not tvm.get_global_func("tvm.contrib.nnpack.fully_connected_inference", True):
             print("skip because extern function is not available")
             return
-        return
+        if not nnpack.is_available():
+            return
+
         ctx = tvm.cpu(0)
         output = nnpack.convolution_inference(
             data, kernel, bias if with_bias else None,
@@ -192,7 +198,9 @@ def test_convolution_inference_without_weight_transform():
         if not tvm.get_global_func("tvm.contrib.nnpack.fully_connected_inference", True):
             print("skip because extern function is not available")
             return
-        return
+        if not nnpack.is_available():
+            return
+
         ctx = tvm.cpu(0)
         transformed_kernel = nnpack.convolution_inference_weight_transform(
             kernel, algorithm=algorithm)
@@ -249,7 +257,9 @@ def test_convolution_output():
         if not tvm.get_global_func("tvm.contrib.nnpack.fully_connected_inference", True):
             print("skip because extern function is not available")
             return
-        return
+        if not nnpack.is_available():
+            return
+
         ctx = tvm.cpu(0)
         f = tvm.build(s, [data, kernel, bias, output], target)