diff --git a/aksetup_helper.py b/aksetup_helper.py
index 32f80cde..311be305 100644
--- a/aksetup_helper.py
+++ b/aksetup_helper.py
@@ -25,7 +25,7 @@ def count_down_delay(delay):
         sys.stdout.flush()
         delay -= 1
         sleep(1)
-    print("")
+    print()
 
 
 DASH_SEPARATOR = 75 * "-"
@@ -297,7 +297,6 @@ def get_default_config(self):
         return dict((opt.name, opt.default) for opt in self.options)
 
     def read_config_from_pyfile(self, filename):
-        result = {}
         filevars = {}
         infile = open(filename)
         try:
@@ -307,11 +306,7 @@ def read_config_from_pyfile(self, filename):
 
         exec(compile(contents, filename, "exec"), filevars)
 
-        for key, value in filevars.items():
-            if key in self.optdict:
-                result[key] = value
-
-        return result
+        return {key: value for key, value in filevars.items() if key in self.optdict}
 
     def update_conf_file(self, filename, config):
         result = {}
@@ -324,16 +319,13 @@ def update_conf_file(self, filename, config):
 
         filevars.pop("__builtins__", None)
 
-        for key, value in config.items():
-            if value is not None:
-                filevars[key] = value
+        filevars.update({key: value for key, value in config.items() if value is not None})
 
         keys = list(filevars.keys())
         keys.sort()
 
         outf = open(filename, "w")
-        for key in keys:
-            outf.write("%s = %s\n" % (key, repr(filevars[key])))
+        outf.writelines("%s = %s\n" % (key, repr(filevars[key])) for key in keys)
         outf.close()
 
         return result
@@ -520,8 +512,7 @@ def take_from_configparser(self, options):
                 import re
                 sep = re.compile(r"(?<!\\),")
                 result = sep.split(opt)
-                result = [i.replace(r"\,", ",") for i in result]
-                return result
+                return [i.replace(r"\,", ",") for i in result]
             else:
                 return []
 
@@ -588,9 +579,9 @@ def set_up_shipped_boost_if_requested(project_name, conf, source_path=None,
             print("(The files should be under %s/.)" % source_path)
             print(DASH_SEPARATOR)
             print("If you got this package from git, you probably want to do")
-            print("")
+            print()
             print(" $ git submodule update --init")
-            print("")
+            print()
             print("to fetch what you are presently missing. If you got this from")
             print("a distributed package on the net, that package is broken and")
             print("should be fixed. For now, I will turn off 'USE_SHIPPED_BOOST'")
@@ -798,7 +789,7 @@ def _run_git_command(cmd):
         print(DASH_SEPARATOR)
         print("The package directory appears to be a git repository, but I could")
         print("not invoke git to check whether my submodules are up to date.")
-        print("")
+        print()
         print("The error was:")
         print(git_error)
         print("Hit Ctrl-C now if you'd like to think about the situation.")
@@ -855,20 +846,20 @@ def check_git_submodules():
         print("git submodules are not up-to-date or in odd state")
         print(DASH_SEPARATOR)
         print("If this makes no sense, you probably want to say")
-        print("")
+        print()
         print(" $ git submodule update --init")
-        print("")
+        print()
         print("to fetch what you are presently missing and "
                 "move on with your life.")
         print("If you got this from a distributed package on the "
                 "net, that package is")
         print("broken and should be fixed. Please inform whoever "
                 "gave you this package.")
-        print("")
+        print()
         print("These issues were found:")
         for w in pkg_warnings:
             print("  %s" % w)
-        print("")
+        print()
         print("I will try to initialize the submodules for you "
                 "after a short wait.")
         print(DASH_SEPARATOR)
@@ -898,12 +889,12 @@ def check_pybind11():
         print("Pybind11 is not installed.")
         print(DASH_SEPARATOR)
         print("Very likely, the build process after this message will fail.")
-        print("")
+        print()
         print("Simply press Ctrl+C and type")
         print("python -m pip install pybind11")
         print("to fix this. If you don't, the build will continue ")
         print("in a few seconds.")
-        print("")
+        print()
         print("[1] https://pybind11.readthedocs.io/en/stable/")
         print(DASH_SEPARATOR)
 
diff --git a/examples/cai_numba.py b/examples/cai_numba.py
index f42df68c..7f693d6e 100644
--- a/examples/cai_numba.py
+++ b/examples/cai_numba.py
@@ -7,7 +7,7 @@
 
 # We use autoprimaryctx instead of autoinit because Numba can only operate on a
 # primary context
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 
 
 # Create a PyCUDA gpuarray
diff --git a/examples/demo.py b/examples/demo.py
index 3d49c92f..0a51158a 100644
--- a/examples/demo.py
+++ b/examples/demo.py
@@ -42,7 +42,7 @@
 
 # part 2 ----------------------------------------------------------------------
 
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 
 
 a_gpu = gpuarray.to_gpu(numpy.random.randn(4, 4).astype(numpy.float32))
diff --git a/examples/demo_elementwise.py b/examples/demo_elementwise.py
index 8c361332..3157321a 100644
--- a/examples/demo_elementwise.py
+++ b/examples/demo_elementwise.py
@@ -2,7 +2,7 @@
 
 import numpy
 
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.curandom import rand as curand
 
 
diff --git a/examples/fill_gpu_with_nans.py b/examples/fill_gpu_with_nans.py
index 928b5133..31c7d07d 100644
--- a/examples/fill_gpu_with_nans.py
+++ b/examples/fill_gpu_with_nans.py
@@ -3,7 +3,7 @@
 import numpy
 
 import pycuda.driver as cuda
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 
 
 free_bytes, total_bytes = cuda.mem_get_info()
diff --git a/examples/from-wiki/2d_fft.py b/examples/from-wiki/2d_fft.py
index 07dd2875..168ee3f6 100644
--- a/examples/from-wiki/2d_fft.py
+++ b/examples/from-wiki/2d_fft.py
@@ -68,7 +68,7 @@ def run(self):
 num = 2  # I KNOW THIS IS A BAD PRACTISE, BUT I COUNDN'T FIND ANY OTHER WAY(INIT CANNOT BE USED HERE)
 
 # THE TRANSPOSE-SPLIT ALGORITHM FOR FFT
-for _t in range(0, trans):
+for _t in range(trans):
     for i in range(m):
         a2_1[i, :] = a2[i*p/m:(i+1)*p/m, :].flatten()  # DIVIDE AND RESHAPE THE INPUT IMAGE INTO 1D ARRAY
 
diff --git a/examples/from-wiki/arithmetic_example.py b/examples/from-wiki/arithmetic_example.py
index c70023d0..91b36170 100644
--- a/examples/from-wiki/arithmetic_example.py
+++ b/examples/from-wiki/arithmetic_example.py
@@ -5,7 +5,7 @@
 import numpy as np
 
 import pycuda.driver as cuda
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.compiler import SourceModule
 
 
diff --git a/examples/from-wiki/c++_function_templates.py b/examples/from-wiki/c++_function_templates.py
index 61fcfd49..c0e7fcb7 100644
--- a/examples/from-wiki/c++_function_templates.py
+++ b/examples/from-wiki/c++_function_templates.py
@@ -3,7 +3,7 @@
 
 import numpy as np
 
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.compiler import SourceModule
 
 
diff --git a/examples/from-wiki/demo_complex.py b/examples/from-wiki/demo_complex.py
index a704f4f1..c3932142 100644
--- a/examples/from-wiki/demo_complex.py
+++ b/examples/from-wiki/demo_complex.py
@@ -3,7 +3,7 @@
 
 import numpy
 
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 
 
 a = (numpy.random.randn(400)
diff --git a/examples/from-wiki/distance_element_wise3d.py b/examples/from-wiki/distance_element_wise3d.py
index 64a15f74..5ebbff31 100644
--- a/examples/from-wiki/distance_element_wise3d.py
+++ b/examples/from-wiki/distance_element_wise3d.py
@@ -6,7 +6,7 @@
 import numpy
 
 import pycuda.driver as cuda
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.elementwise import ElementwiseKernel as Elementwise
 
 
diff --git a/examples/from-wiki/game_of_life.py b/examples/from-wiki/game_of_life.py
index 8663b216..596be4e4 100644
--- a/examples/from-wiki/game_of_life.py
+++ b/examples/from-wiki/game_of_life.py
@@ -12,9 +12,9 @@
 
 import matplotlib.pyplot as plt
 import numpy as np
-from pylab import cm as cm
+from pylab import cm
 
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.compiler import SourceModule
 
 
diff --git a/examples/from-wiki/gl_interop.py b/examples/from-wiki/gl_interop.py
index 37268afe..3de9969c 100644
--- a/examples/from-wiki/gl_interop.py
+++ b/examples/from-wiki/gl_interop.py
@@ -126,7 +126,7 @@ def keyPressed(*args):
         print("Closing..")
         destroy_PBOs()
         destroy_texture()
-        exit()
+        sys.exit()
     elif args[0] == "a":
         print("toggling animation")
         animate = not animate
diff --git a/examples/from-wiki/gpu_scalar_mult.py b/examples/from-wiki/gpu_scalar_mult.py
index 11e133cd..159986e2 100644
--- a/examples/from-wiki/gpu_scalar_mult.py
+++ b/examples/from-wiki/gpu_scalar_mult.py
@@ -4,7 +4,7 @@
 import numpy
 
 import pycuda.driver as drv
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.tools import context_dependent_memoize
 
 
diff --git a/examples/from-wiki/mandelbrot.py b/examples/from-wiki/mandelbrot.py
index 8c214cc7..b76cb020 100644
--- a/examples/from-wiki/mandelbrot.py
+++ b/examples/from-wiki/mandelbrot.py
@@ -16,7 +16,7 @@
 import numpy as nm
 
 import pycuda.driver as drv
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 
 
 # set width and height of window, more pixels take longer to calculate
@@ -42,8 +42,7 @@ def calculate_z_gpu(q, maxiter, z):
     # we bring the iterations_gpu array back to determine pixel colours later
     complex_gpu(z_gpu, q_gpu, iterations_gpu, maxiter)
 
-    iterations = iterations_gpu.get()
-    return iterations
+    return iterations_gpu.get()
 
 
 def calculate_z_numpy_gpu(q, maxiter, z):
@@ -73,8 +72,7 @@ def calculate_z_numpy_gpu(q, maxiter, z):
         zg = gpuarray.if_positive(comparison_result, cmplx0sg, zg)
         outputg = gpuarray.if_positive(comparison_result, iterg, outputg)
         iterg = iterg + 1
-    output = outputg.get()
-    return output
+    return outputg.get()
 
 
 def calculate_z_numpy(q, maxiter, z):
diff --git a/examples/from-wiki/mandelbrot_interactive.py b/examples/from-wiki/mandelbrot_interactive.py
index 4b81767c..65a2a491 100644
--- a/examples/from-wiki/mandelbrot_interactive.py
+++ b/examples/from-wiki/mandelbrot_interactive.py
@@ -33,7 +33,7 @@
 import numpy as np
 from matplotlib.patches import Rectangle
 from matplotlib.widgets import RectangleSelector
-from pylab import cm as cm
+from pylab import cm
 
 import pycuda.driver as drv
 from pycuda.compiler import SourceModule
diff --git a/examples/from-wiki/manhattan_distance_for_2D_array.py b/examples/from-wiki/manhattan_distance_for_2D_array.py
index 6a71d7d9..80fd3aa9 100644
--- a/examples/from-wiki/manhattan_distance_for_2D_array.py
+++ b/examples/from-wiki/manhattan_distance_for_2D_array.py
@@ -53,7 +53,7 @@ def run(res, step):
 
 
 res   = numpy.array([[0
-                        for _ in range(0, w)]
-                        for _ in range(0, w)], dtype="int32")
+                        for _ in range(w)]
+                        for _ in range(w)], dtype="int32")
 print(res)
 run(res, 0)
diff --git a/examples/from-wiki/matrix_transpose.py b/examples/from-wiki/matrix_transpose.py
index a2e66d25..ff4b9d9f 100644
--- a/examples/from-wiki/matrix_transpose.py
+++ b/examples/from-wiki/matrix_transpose.py
@@ -10,7 +10,7 @@
 
 import pycuda.autoinit
 import pycuda.driver as cuda
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.compiler import SourceModule
 from pycuda.tools import context_dependent_memoize
 
diff --git a/examples/from-wiki/measure_gpuarray_speed_random.py b/examples/from-wiki/measure_gpuarray_speed_random.py
index 51786340..005a1a29 100644
--- a/examples/from-wiki/measure_gpuarray_speed_random.py
+++ b/examples/from-wiki/measure_gpuarray_speed_random.py
@@ -8,12 +8,12 @@
 
 from pytools import Table
 
-import pycuda.curandom as curandom
 import pycuda.driver as drv
+from pycuda import curandom
 
 
 def main():
-    import pycuda.gpuarray as gpuarray
+    from pycuda import gpuarray
 
     sizes = []
     times = []
diff --git a/examples/from-wiki/plot_random_data.py b/examples/from-wiki/plot_random_data.py
index 9f0ef419..e1fb2a67 100644
--- a/examples/from-wiki/plot_random_data.py
+++ b/examples/from-wiki/plot_random_data.py
@@ -2,7 +2,7 @@
 # simple module to show the plotting of random data
 from __future__ import annotations
 
-import pycuda.curandom as curandom
+from pycuda import curandom
 
 
 size = 1000
diff --git a/examples/from-wiki/select_to_list.py b/examples/from-wiki/select_to_list.py
index ff3d34c5..f60a8096 100644
--- a/examples/from-wiki/select_to_list.py
+++ b/examples/from-wiki/select_to_list.py
@@ -8,7 +8,7 @@
 import numpy
 
 import pycuda.driver as cuda
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.compiler import SourceModule
 
 
diff --git a/examples/from-wiki/simple_speed_test.py b/examples/from-wiki/simple_speed_test.py
index 6f603628..a8e9bf91 100644
--- a/examples/from-wiki/simple_speed_test.py
+++ b/examples/from-wiki/simple_speed_test.py
@@ -48,8 +48,8 @@
 import pycuda.autoinit
 import pycuda.cumath
 import pycuda.driver as drv
-import pycuda.gpuarray as gpuarray
 import pycuda.tools
+from pycuda import gpuarray
 from pycuda.compiler import SourceModule
 from pycuda.elementwise import ElementwiseKernel
 
diff --git a/examples/from-wiki/sobel_filter.py b/examples/from-wiki/sobel_filter.py
index b3c51cce..57c03642 100644
--- a/examples/from-wiki/sobel_filter.py
+++ b/examples/from-wiki/sobel_filter.py
@@ -514,13 +514,13 @@ def keyboard(key, x=0, y=0):
     global mode, scale
     if key == "q":
         sys.exit(0)
-    elif key == "I" or key == "i":
+    elif key in {"I", "i"}:
         mode = 0
-    elif key == "T" or key == "t":
+    elif key in {"T", "t"}:
         mode = 1
-    elif key == "S" or key == "s":
+    elif key in {"S", "s"}:
         mode = 2
-    elif key == "D" or key == "d":
+    elif key in {"D", "d"}:
         mode = 3
     elif key == "-":
         scale -= 0.1
diff --git a/examples/from-wiki/sparse_solve.py b/examples/from-wiki/sparse_solve.py
index 38505249..0e94a0cc 100644
--- a/examples/from-wiki/sparse_solve.py
+++ b/examples/from-wiki/sparse_solve.py
@@ -5,7 +5,7 @@
 import numpy.linalg as la
 
 import pycuda.driver as drv
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 
 
 def main_cg():
diff --git a/examples/from-wiki/thrust_interop.py b/examples/from-wiki/thrust_interop.py
index 1f19a096..7ce54a5b 100644
--- a/examples/from-wiki/thrust_interop.py
+++ b/examples/from-wiki/thrust_interop.py
@@ -6,7 +6,7 @@
 from codepy.bpl import BoostPythonModule
 from codepy.cuda import CudaModule
 
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 
 
 # Make a host_module, compiled for CPU
diff --git a/pycuda/autoinit.py b/pycuda/autoinit.py
index b628b3cd..34e00a94 100644
--- a/pycuda/autoinit.py
+++ b/pycuda/autoinit.py
@@ -11,7 +11,6 @@
 from pycuda.tools import make_default_context
 
 
-global context
 context = make_default_context()
 device = context.get_device()
 
diff --git a/pycuda/autoprimaryctx.py b/pycuda/autoprimaryctx.py
index 5a860b73..988511e8 100644
--- a/pycuda/autoprimaryctx.py
+++ b/pycuda/autoprimaryctx.py
@@ -17,7 +17,6 @@ def _retain_primary_context(dev):
     return context
 
 
-global context
 context = make_default_context(_retain_primary_context)
 device = context.get_device()
 
diff --git a/pycuda/characterize.py b/pycuda/characterize.py
index 74e6c9c7..6a9e707d 100644
--- a/pycuda/characterize.py
+++ b/pycuda/characterize.py
@@ -43,7 +43,7 @@ def sizeof(type_name, preamble=""):
         no_extern_c=True,
     )
 
-    import pycuda.gpuarray as gpuarray
+    from pycuda import gpuarray
 
     output = gpuarray.empty((), dtype=np.uintp)
     mod.get_function("write_size")(output, block=(1, 1, 1), grid=(1, 1))
diff --git a/pycuda/compiler.py b/pycuda/compiler.py
index 5715462e..a645812f 100644
--- a/pycuda/compiler.py
+++ b/pycuda/compiler.py
@@ -5,6 +5,7 @@
 import sys
 from os import unlink
 from tempfile import mkstemp
+from warnings import warn
 
 from pytools import memoize
 from pytools.prefork import call_capture_output
@@ -16,8 +17,6 @@ def get_nvcc_version(nvcc):
     result, stdout, _stderr = call_capture_output(cmdline)
 
     if result != 0 or not stdout:
-        from warnings import warn
-
         warn("NVCC version could not be determined.", stacklevel=2)
         stdout = b"nvcc unknown version"
 
@@ -105,8 +104,8 @@ def compile_plain(source, options, keep, nvcc, cache_dir, target="cubin"):
             with open(cache_path, "rb") as cache_file:
                 return cache_file.read()
 
-        except Exception:
-            pass
+        except Exception as e:
+            warn(f"Could not read cache file '{cache_path}': {e!s}", stacklevel=1)
 
     from tempfile import mkdtemp
 
@@ -139,8 +138,6 @@ def compile_plain(source, options, keep, nvcc, cache_dir, target="cubin"):
 
     if result != 0 or (no_output and (stdout or stderr)):
         if result == 0:
-            from warnings import warn
-
             warn(
                 "PyCUDA: nvcc exited with status 0, but appears to have "
                 "encountered an error", stacklevel=2
@@ -156,8 +153,6 @@ def compile_plain(source, options, keep, nvcc, cache_dir, target="cubin"):
 
     if stdout or stderr:
         lcase_err_text = (stdout + stderr).decode("utf-8", "replace").lower()
-        from warnings import warn
-
         if "demoted" in lcase_err_text or "demoting" in lcase_err_text:
             warn(
                 "nvcc said it demoted types in source code it "
@@ -304,14 +299,12 @@ def _check_arch(self, arch):
 
             capability = Context.get_device().compute_capability()
             if tuple(map(int, tuple(arch.split("_")[1]))) > capability:
-                from warnings import warn
-
                 warn(
                     "trying to compile for a compute capability "
                     "higher than selected GPU", stacklevel=2
                 )
-        except Exception:
-            pass
+        except Exception as e:
+            warn(f"failed to check compute capability for module: {e!s}", stacklevel=1)
 
     def _bind_module(self):
         self.get_global = self.module.get_global
@@ -411,7 +404,7 @@ def __init__(
             include_dirs = []
         compute_capability = Context.get_device().compute_capability()
         if compute_capability < (3, 5):
-            raise Exception(
+            raise RuntimeError(
                 "Minimum compute capability for dynamic parallelism is 3.5 (found: %u.%u)!"
                 % (compute_capability[0], compute_capability[1])
             )
diff --git a/pycuda/compyte b/pycuda/compyte
index f1010ef6..80ed45de 160000
--- a/pycuda/compyte
+++ b/pycuda/compyte
@@ -1 +1 @@
-Subproject commit f1010ef64b5d4559ca4bb7690d4b3394091477b6
+Subproject commit 80ed45de98b5432341763b9fa52a00fdac870b89
diff --git a/pycuda/cumath.py b/pycuda/cumath.py
index 4ac71d58..5110105c 100644
--- a/pycuda/cumath.py
+++ b/pycuda/cumath.py
@@ -4,8 +4,7 @@
 
 import numpy as np
 
-import pycuda.elementwise as elementwise
-import pycuda.gpuarray as gpuarray
+from pycuda import elementwise, gpuarray
 from pycuda.driver import Stream
 
 
diff --git a/pycuda/curandom.py b/pycuda/curandom.py
index 4382b888..f31ac220 100644
--- a/pycuda/curandom.py
+++ b/pycuda/curandom.py
@@ -1015,8 +1015,7 @@ def _kernels(self):
 def generate_direction_vectors(count, direction=None):
     if get_curand_version() >= (4, 0, 0):
         if (
-            direction == direction_vector_set.VECTOR_64
-            or direction == direction_vector_set.SCRAMBLED_VECTOR_64
+            direction in (direction_vector_set.VECTOR_64, direction_vector_set.SCRAMBLED_VECTOR_64)
         ):
             result = np.empty((count, 64), dtype=np.uint64)
         else:
diff --git a/pycuda/debug.py b/pycuda/debug.py
index f2973b08..5ed22d65 100644
--- a/pycuda/debug.py
+++ b/pycuda/debug.py
@@ -27,4 +27,4 @@
 sys.argv = args
 
 with open(mainpyfile) as mainpy:
-    exec(compile(mainpy.read(), mainpyfile, "exec"))
+    exec(compile(mainpy.read(), mainpyfile, "exec"))  # noqa: S102
diff --git a/pycuda/gpuarray.py b/pycuda/gpuarray.py
index f208a286..41afeca0 100644
--- a/pycuda/gpuarray.py
+++ b/pycuda/gpuarray.py
@@ -15,7 +15,7 @@
 from pytools import memoize, memoize_method
 
 import pycuda.driver as drv
-import pycuda.elementwise as elementwise
+from pycuda import elementwise
 from pycuda.characterize import has_double_support
 from pycuda.compyte.array import (
     ArrayFlags as _ArrayFlags,
@@ -1345,10 +1345,9 @@ def _array_like_helper(other_ary, dtype, order):
 
 def empty_like(other_ary, dtype=None, order="K"):
     dtype, order, strides = _array_like_helper(other_ary, dtype, order)
-    result = GPUArray(
+    return GPUArray(
         other_ary.shape, dtype, other_ary.allocator, order=order, strides=strides
     )
-    return result
 
 
 def zeros_like(other_ary, dtype=None, order="K"):
diff --git a/pycuda/scan.py b/pycuda/scan.py
index 65a9b0d6..48c3dc45 100644
--- a/pycuda/scan.py
+++ b/pycuda/scan.py
@@ -28,8 +28,7 @@
 import numpy as np
 
 import pycuda._mymako as mako
-import pycuda.driver as driver
-import pycuda.gpuarray as gpuarray
+from pycuda import driver, gpuarray
 from pycuda._cluda import CLUDA_PREAMBLE
 from pycuda.compiler import SourceModule
 from pycuda.tools import dtype_to_ctype
diff --git a/pycuda/sparse/cg.py b/pycuda/sparse/cg.py
index 4079c1fa..fbf341a6 100644
--- a/pycuda/sparse/cg.py
+++ b/pycuda/sparse/cg.py
@@ -5,7 +5,7 @@
 from pytools import memoize_method
 
 import pycuda.driver as drv
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.sparse.inner import AsyncInnerProduct
 
 
diff --git a/pycuda/sparse/coordinate.py b/pycuda/sparse/coordinate.py
index 13191a86..4b7e0a19 100644
--- a/pycuda/sparse/coordinate.py
+++ b/pycuda/sparse/coordinate.py
@@ -5,7 +5,7 @@
 from pytools import memoize_method
 
 import pycuda.driver as drv
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.compiler import SourceModule
 
 
diff --git a/pycuda/sparse/inner.py b/pycuda/sparse/inner.py
index 7e61498f..4400b0c8 100644
--- a/pycuda/sparse/inner.py
+++ b/pycuda/sparse/inner.py
@@ -3,7 +3,7 @@
 import atexit
 
 import pycuda.driver as drv
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 
 
 STREAM_POOL = []
diff --git a/pycuda/sparse/packeted.py b/pycuda/sparse/packeted.py
index c06ae322..bbbb397d 100644
--- a/pycuda/sparse/packeted.py
+++ b/pycuda/sparse/packeted.py
@@ -4,7 +4,7 @@
 
 from pytools import memoize_method
 
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.compiler import SourceModule
 
 
diff --git a/pycuda/sparse/pkt_build.py b/pycuda/sparse/pkt_build.py
index 235e1a0f..1746d187 100644
--- a/pycuda/sparse/pkt_build.py
+++ b/pycuda/sparse/pkt_build.py
@@ -2,7 +2,7 @@
 
 import numpy as np
 
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 
 
 def build_pkt_data_structure(
diff --git a/pycuda/tools.py b/pycuda/tools.py
index e37e616f..d30b1d35 100644
--- a/pycuda/tools.py
+++ b/pycuda/tools.py
@@ -1,4 +1,4 @@
-"""Miscallenous helper functionality."""
+"""Miscellaneous helper functionality."""
 from __future__ import annotations
 
 
@@ -198,7 +198,7 @@ def ctx_maker(dev):
             assert homedir is not None
             with open(os.path.join(homedir, ".cuda_device")) as devrc:
                 devn = devrc.read().strip()
-        except Exception:
+        except Exception:  # noqa: S110
             pass
 
     # If either CUDA_DEVICE or $HOME/.cuda_device is set, try to use it
@@ -314,7 +314,7 @@ def align_words(self, word_size):
     def align_bytes(self, word_size=4):
         if word_size == 4:
             return 64
-        elif word_size == 8 or word_size == 16:
+        elif word_size in {8, 16}:
             return 128
         else:
             raise ValueError("no alignment possible for fetches of size %d" % word_size)
diff --git a/pyproject.toml b/pyproject.toml
index 394b9f68..6c5c5fa3 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -45,6 +45,7 @@ extend-ignore = [
     "SIM108", # ternary instead of if-then-else
     "RUF067", # no code in __init__ *shrug*
     "RUF069", # unreliable FP comparison
+    "BLE001",
 ]
 
 [tool.ruff.lint.flake8-quotes]
@@ -67,12 +68,15 @@ required-imports = ["from __future__ import annotations"]
 extend-ignore-names = ["update_for_*"]
 
 [tool.ruff.lint.per-file-ignores]
+"doc/conf.py" = ["S102"]
 "test/test_*.py" = [
     "N806", # upper case locals
+    "S102",
 ]
 "setup.py" = [
     "N806", # upper case locals
     "SIM115",  # context manager for files
+    "S102",
 ]
 "pycuda/sparse/coordinate.py" = [
     "E501", # line length
@@ -85,7 +89,7 @@ extend-ignore-names = ["update_for_*"]
     "E501", # line length
 ]
 "examples/from-wiki/*.py" = [
-    "F", "E", "N", "B"
+    "F", "E", "N", "B", "EXE001", "PLW",
 ]
 "examples/demo_cdpSimplePrint.py" = [
     "E501", # line length
@@ -93,6 +97,6 @@ extend-ignore-names = ["update_for_*"]
 ]
 "aksetup_helper.py" = [
     # effectively unmaintained, will go away
-    "UP", "C", "E501", "B", "SIM", "RUF",
+    "UP", "C", "E501", "B", "SIM", "RUF", "S102", "S110", "TRY"
 ]
 "test/undistributed/*.py" = ["B"]
diff --git a/setup.py b/setup.py
index a15ef2b8..bf3c1c51 100644
--- a/setup.py
+++ b/setup.py
@@ -1,4 +1,3 @@
-#!/usr/bin/env python
 from __future__ import annotations
 
 from os.path import dirname, join, normpath
diff --git a/test/test_cumath.py b/test/test_cumath.py
index 4267e120..7951963a 100644
--- a/test/test_cumath.py
+++ b/test/test_cumath.py
@@ -4,9 +4,8 @@
 
 import numpy as np
 
-import pycuda.cumath as cumath
 import pycuda.driver as drv  # noqa
-import pycuda.gpuarray as gpuarray
+from pycuda import cumath, gpuarray
 from pycuda.tools import mark_cuda_test
 
 
diff --git a/test/test_driver.py b/test/test_driver.py
index 5e9befe7..cf9ca89e 100644
--- a/test/test_driver.py
+++ b/test/test_driver.py
@@ -11,7 +11,7 @@
 import pytest
 
 import pycuda.driver as drv
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.compiler import SourceModule
 from pycuda.tools import dtype_to_ctype, mark_cuda_test
 
@@ -141,7 +141,7 @@ def test_streamed_kernel(self):
     def test_gpuarray(self):
         a = np.arange(200000, dtype=np.float32)
         b = a + 17
-        import pycuda.gpuarray as gpuarray
+        from pycuda import gpuarray
 
         a_g = gpuarray.to_gpu(a)
         b_g = gpuarray.to_gpu(b)
@@ -172,7 +172,7 @@ def test_gpuarray_cai(self):
     def donottest_cublas_mixing(self):
         self.test_streamed_kernel()
 
-        import pycuda.blas as blas
+        from pycuda import blas
 
         shape = (10,)
         a = blas.ones(shape, dtype=np.float32)
@@ -675,7 +675,7 @@ def test_large_smem(self):
 
         kernel = mod.get_function("kernel")
 
-        import pycuda.gpuarray as gpuarray
+        from pycuda import gpuarray
 
         arg = gpuarray.zeros((n,), dtype=np.float32)
 
diff --git a/test/test_gpuarray.py b/test/test_gpuarray.py
index 18516a46..630bedce 100644
--- a/test/test_gpuarray.py
+++ b/test/test_gpuarray.py
@@ -1,4 +1,3 @@
-#! /usr/bin/env python
 from __future__ import annotations
 
 import operator
@@ -9,7 +8,7 @@
 import pytest
 
 import pycuda.driver as drv
-import pycuda.gpuarray as gpuarray
+from pycuda import gpuarray
 from pycuda.characterize import has_double_support
 from pycuda.compiler import SourceModule
 from pycuda.tools import init_cuda_context_fixture
@@ -765,7 +764,7 @@ def test_reverse(self):
 
         b = a_cpu.get()
 
-        for i in range(0, 10):
+        for i in range(10):
             assert a[len(a) - 1 - i] == b[i]
 
     def test_sum(self):
@@ -996,7 +995,7 @@ def test_2d_slice_c(self):
             assert la.norm(a_gpu_slice.get() - a_slice) == 0
 
     def test_2d_slice_f(self):
-        import pycuda.gpuarray as gpuarray
+        from pycuda import gpuarray
         from pycuda.curandom import rand as curand
 
         n = 1000
@@ -1023,7 +1022,7 @@ def test_where(self):
         b = np.array([2, 2, 2])
         c = np.array([3, 3, 3])
 
-        import pycuda.gpuarray as gpuarray
+        from pycuda import gpuarray
 
         a_gpu = gpuarray.to_gpu(a)
         b_gpu = gpuarray.to_gpu(b)
@@ -1043,7 +1042,7 @@ def test_if_positive(self):
         a = a_gpu.get()
         b = b_gpu.get()
 
-        import pycuda.gpuarray as gpuarray
+        from pycuda import gpuarray
 
         max_a_b_gpu = gpuarray.maximum(a_gpu, b_gpu)
         min_a_b_gpu = gpuarray.minimum(a_gpu, b_gpu)
@@ -1479,7 +1478,7 @@ def test_minimum_maximum_scalar(self):
         a_gpu = curand((sz,))
         a = a_gpu.get()
 
-        import pycuda.gpuarray as gpuarray
+        from pycuda import gpuarray
 
         max_a0_gpu = gpuarray.maximum(a_gpu, 0)
         min_a0_gpu = gpuarray.minimum(0, a_gpu)
@@ -1546,7 +1545,7 @@ def test_copy(self):
             )
 
     def test_get_set(self):
-        import pycuda.gpuarray as gpuarray
+        from pycuda import gpuarray
 
         a = np.random.normal(0.0, 1.0, (4, 4))
         a_gpu = gpuarray.to_gpu(a)
diff --git a/test/undistributed/elwise-perf.py b/test/undistributed/elwise-perf.py
index 960c275b..de2270a4 100644
--- a/test/undistributed/elwise-perf.py
+++ b/test/undistributed/elwise-perf.py
@@ -9,7 +9,7 @@ def main():
     tbl = Table()
     tbl.add_row(("size [MiB]", "time [s]", "mem.bw [GB/s]"))
 
-    import pycuda.gpuarray as gpuarray
+    from pycuda import gpuarray
 
     # they're floats, i.e. 4 bytes each
     for power in range(10, 28):
diff --git a/test/undistributed/measure_gpuarray_speed.py b/test/undistributed/measure_gpuarray_speed.py
index 4e441fde..1edb4dc2 100755
--- a/test/undistributed/measure_gpuarray_speed.py
+++ b/test/undistributed/measure_gpuarray_speed.py
@@ -7,7 +7,7 @@
 
 
 def main():
-    import pycuda.gpuarray as gpuarray
+    from pycuda import gpuarray
 
     sizes = []
     times_gpu = []