Merge branch 'master' into port-polyeval-cpu

mops/include/mops/sap.h

@@ -1,5 +1,5 @@
-#ifndef MOPS_OUTER_PRODUCT_SCATTER_ADD_H
-#define MOPS_OUTER_PRODUCT_SCATTER_ADD_H
+#ifndef MOPS_SPARSE_ACCUMULATION_OF_PRODUCTS_H
+#define MOPS_SPARSE_ACCUMULATION_OF_PRODUCTS_H
 
 #include "mops/exports.h"
 #include "mops/tensor.h"

mops/include/mops/sap.hpp

@@ -1,5 +1,5 @@
-#ifndef MOPS_OUTER_PRODUCT_SCATTER_ADD_HPP
-#define MOPS_OUTER_PRODUCT_SCATTER_ADD_HPP
+#ifndef MOPS_SPARSE_ACCUMULATION_OF_PRODUCTS_HPP
+#define MOPS_SPARSE_ACCUMULATION_OF_PRODUCTS_HPP
 
 #include <cstddef>
 #include <cstdint>

python/mops/src/mops/__init__.py

@@ -1,4 +1,3 @@
 from .hpe import homogeneous_polynomial_evaluation  # noqa
 from .sap import sparse_accumulation_of_products  # noqa
 from .opsa import outer_product_scatter_add  # noqa
-

python/mops/src/mops/_c_api.py

@@ -17,7 +17,7 @@ class mops_tensor_2d_f64_t(ctypes.Structure):
 
 class mops_tensor_1d_f32_t(ctypes.Structure):
     _fields_ = [
-        ("data", ctypes.POINTER(ctypes.c_double)),
+        ("data", ctypes.POINTER(ctypes.c_float)),
         ("shape", ctypes.ARRAY(ctypes.c_int64, 1)),
     ]
 
@@ -159,3 +159,48 @@ def setup_functions(lib):
         mops_tensor_1d_i32_t,
     ]
     lib.mops_cuda_outer_product_scatter_add_f64.restype = _check_status
+
+    # sparse_accumulation_of_products
+    lib.mops_sparse_accumulation_of_products_f32.argtypes = [
+        mops_tensor_2d_f32_t,
+        mops_tensor_2d_f32_t,
+        mops_tensor_2d_f32_t,
+        mops_tensor_1d_f32_t,
+        mops_tensor_1d_i32_t,
+        mops_tensor_1d_i32_t,
+        mops_tensor_1d_i32_t,
+    ]
+    lib.mops_sparse_accumulation_of_products_f32.restype = _check_status
+
+    lib.mops_sparse_accumulation_of_products_f64.argtypes = [
+        mops_tensor_2d_f64_t,
+        mops_tensor_2d_f64_t,
+        mops_tensor_2d_f64_t,
+        mops_tensor_1d_f64_t,
+        mops_tensor_1d_i32_t,
+        mops_tensor_1d_i32_t,
+        mops_tensor_1d_i32_t,
+    ]
+    lib.mops_sparse_accumulation_of_products_f64.restype = _check_status
+
+    lib.mops_cuda_sparse_accumulation_of_products_f32.argtypes = [
+        mops_tensor_2d_f32_t,
+        mops_tensor_2d_f32_t,
+        mops_tensor_2d_f32_t,
+        mops_tensor_1d_f32_t,
+        mops_tensor_1d_i32_t,
+        mops_tensor_1d_i32_t,
+        mops_tensor_1d_i32_t,
+    ]
+    lib.mops_cuda_sparse_accumulation_of_products_f32.restype = _check_status
+
+    lib.mops_cuda_sparse_accumulation_of_products_f64.argtypes = [
+        mops_tensor_2d_f64_t,
+        mops_tensor_2d_f64_t,
+        mops_tensor_2d_f64_t,
+        mops_tensor_1d_f64_t,
+        mops_tensor_1d_i32_t,
+        mops_tensor_1d_i32_t,
+        mops_tensor_1d_i32_t,
+    ]
+    lib.mops_cuda_sparse_accumulation_of_products_f64.restype = _check_status

python/mops/tests/opsa.py

@@ -35,5 +35,8 @@ def test_opsa_no_neighbors():
 
 def test_opsa_wrong_type():
 
-    with pytest.raises(ValueError):
+    with pytest.raises(
+        ValueError,
+        match="A must be a 2D array in opsa, got a 1D array"
+    ):
         opsa(np.array([1]), 2, 3, 4)

tox.ini

@@ -37,20 +37,17 @@ deps =
     wheel
     cmake
     twine
+    build
 
 allowlist_externals =
     bash
 
 commands =
     # check building sdist and wheels from a checkout
-    python setup.py sdist
-    python setup.py bdist_wheel
+    python -m build .  --outdir dist
     twine check dist/*.tar.gz
     twine check dist/*.whl
 
-    # check building wheels from the sdist
-    bash -c "python -m pip wheel --verbose dist/mops-*.tar.gz -w dist/test"
-
 ; [testenv:torch-cxx-tests]
 ; package = skip
 ; passenv = *