diff --git a/pycuda/compiler.py b/pycuda/compiler.py
index 96aa1b19e1b84a16854feb197731cebf6ffc2b69..369b6375774677965437d7f0d33c815d37f58a0b 100644
--- a/pycuda/compiler.py
+++ b/pycuda/compiler.py
@@ -211,7 +211,7 @@ def _find_pycuda_include_path():
 
     return os.path.abspath(
         os.path.join(importlib.util.find_spec("pycuda").origin,
-                     os.path.pardir, "cuda"))
+                     os.path.join(os.path.pardir, "cuda")))
 
 
 DEFAULT_NVCC_FLAGS = [
diff --git a/pycuda/elementwise.py b/pycuda/elementwise.py
index 7d633011ce727f549b5a7ce929a0f5594de2ad3a..83fd779a19d79fc375c387e8852359a851bc8683 100644
--- a/pycuda/elementwise.py
+++ b/pycuda/elementwise.py
@@ -565,7 +565,7 @@ def get_binary_func_scalar_kernel(func, dtype_x, dtype_y, dtype_z):
 
 
 def get_binary_minmax_kernel(func, dtype_x, dtype_y, dtype_z, use_scalar):
-    if np.float64 not in [dtype_x, dtype_y]:
+    if (np.float64 not in [dtype_x, dtype_y]) and (bool not in [dtype_x, dtype_y]):
         func = func + "f"
 
     if any(dt.kind == "f" for dt in [dtype_x, dtype_y, dtype_z]):
diff --git a/pycuda/gpuarray.py b/pycuda/gpuarray.py
index 100d21cc7a75325d93295bdc8a89d0bf3e439b01..893a802e97ec4f8e35036060d8f5bc4342ba16f9 100644
--- a/pycuda/gpuarray.py
+++ b/pycuda/gpuarray.py
@@ -2019,7 +2019,17 @@ def where(criterion, then_, else_, out=None, stream=None):
 
 def _make_binary_minmax_func(which):
     def f(a, b, out=None, stream=None):
-        if isinstance(a, GPUArray) and isinstance(b, GPUArray):
+        allocator = (
+            getattr(a, "allocator", None)
+            or getattr(b, "allocator", None)
+            or drv.mem_alloc)
+
+        if np.isscalar(a) and np.isscalar(b):
+            if out is None:
+                out = GPUArray(shape=(), dtype=np.bool_, allocator=allocator)
+            import pycuda.gpuarray as gpuarray
+            out[...] = gpuarray.to_gpu(np.array(getattr(np, "f"+which)(a, b)))
+        elif isinstance(a, GPUArray) and isinstance(b, GPUArray):
             if out is None:
                 out = empty_like(a)
             func = elementwise.get_binary_minmax_kernel(
diff --git a/test/test_gpuarray.py b/test/test_gpuarray.py
index 7091fee9481c52f4109ea1085a0d2cb0fa9765dc..62791e7d9d7d52693b50f944b634926040180214 100644
--- a/test/test_gpuarray.py
+++ b/test/test_gpuarray.py
@@ -11,6 +11,7 @@ import pycuda.gpuarray as gpuarray
 import pycuda.driver as drv
 from pycuda.compiler import SourceModule
 import pytest
+import itertools
 
 
 @pytest.fixture(autouse=True)
@@ -809,6 +810,14 @@ class TestGPUArray:
         assert la.norm(max_a_b_gpu.get() - np.maximum(a, b)) == 0
         assert la.norm(min_a_b_gpu.get() - np.minimum(a, b)) == 0
 
+    @pytest.mark.parametrize("func", ["minimum", "maximum"])
+    def test_min_max_elemwise_on_scalars(self, func):
+
+        for a, b in itertools.product([False, True], [False, True]):
+            result_ref = getattr(np, func)(a, b)
+            result = getattr(gpuarray, func)(a, b)
+            np.testing.assert_array_equal(result.get(), result_ref)
+
     def test_take_put(self):
         for n in [5, 17, 333]:
             one_field_size = 8