test: working on benchmarks with Finch as backend

.github/workflows/codspeed.yml

@@ -25,4 +25,9 @@ jobs:
       - name: Run benchmarks
         uses: CodSpeedHQ/action@v3
         with:
-          run: pytest benchmarks/ --codspeed
+          run: SPARSE_BACKEND=Numba pytest benchmarks/ --codspeed
+
+      - name: Run benchmarks
+        uses: CodSpeedHQ/action@v3
+        with:
+          run: SPARSE_BACKEND=Finch pytest benchmarks/ --codspeed

benchmarks/conftest.py

@@ -1,3 +1,5 @@
+import sparse
+
 import pytest
 
 
@@ -6,6 +8,21 @@ def seed(scope="session"):
     return 42
 
 
+def get_backend_id(param):
+    backend = param
+    return f"{backend=}"
+
+
+@pytest.fixture(params=[sparse._BACKEND], autouse=True, ids=get_backend_id)
+def backend(request):
+    return request.param
+
+
+@pytest.fixture
+def min_size(scope="session"):
+    return 100
+
+
 @pytest.fixture
 def max_size(scope="session"):
     return 2**26

benchmarks/test_benchmark_coo.py

@@ -15,16 +15,22 @@ def format_id(format):
 
 
 @pytest.mark.parametrize("format", ["coo", "gcxs"])
-def test_matmul(benchmark, sides, format, seed, max_size, ids=format_id):
+def test_matmul(benchmark, sides, seed, format, backend, min_size, max_size, ids=format_id):
+    # if backend == sparse._BackendType.Finch:
+    #    pytest.skip()
+
     m, n, p = sides
 
-    if m * n >= max_size or n * p >= max_size:
+    if m * n >= max_size or n * p >= max_size or m * n <= min_size or n * p <= min_size:
         pytest.skip()
 
     rng = np.random.default_rng(seed=seed)
     x = sparse.random((m, n), density=DENSITY, format=format, random_state=rng)
     y = sparse.random((n, p), density=DENSITY, format=format, random_state=rng)
 
+    if hasattr(sparse, "compiled"):
+        operator.matmul = sparse.compiled(operator.matmul)
+
     x @ y  # Numba compilation
 
     @benchmark
@@ -52,6 +58,10 @@ def elemwise_args(request, seed, max_size):
 @pytest.mark.parametrize("f", [operator.add, operator.mul])
 def test_elemwise(benchmark, f, elemwise_args):
     x, y = elemwise_args
+
+    if hasattr(sparse, "compiled"):
+        f = sparse.compiled(f)
+
     f(x, y)
 
     @benchmark