Merge remote-tracking branch 'origin/master' into searchspace_experiments

Author: fjwillemsen

INSTALL.rst

@@ -125,7 +125,7 @@ Or you could install Kernel Tuner and PyOpenCL together if you haven't done so a
 If this fails, please see the PyOpenCL installation guide (https://wiki.tiker.net/PyOpenCL/Installation)
 
 HIP and HIP Python
--------------
+------------------
 
 Before we can install HIP Python, you'll need to have the HIP runtime and compiler installed on your system.
 The HIP compiler is included as part of the ROCm software stack. Here is AMD's installation guide:

doc/source/architecture.png

@@ -89,14 +89,8 @@ kernel_tuner.runners.sequential.SimulationRunner
     :members:
 
 
-Device Interfaces
------------------
-
-kernel_tuner.core.DeviceInterface
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-.. autoclass:: kernel_tuner.core.DeviceInterface
-    :special-members: __init__
-    :members:
+Backends
+--------
 
 kernel_tuner.backends.pycuda.PyCudaFunctions
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

doc/source/design.rst

@@ -104,6 +104,7 @@ def __init__(self, device=0, iterations=7, compiler_options=None, observers=None
 
     def ready_argument_list(self, arguments):
         """Ready argument list to be passed to the HIP function.
+
         :param arguments: List of arguments to be passed to the HIP function.
             The order should match the argument list on the HIP function.
             Allowed values are np.ndarray, and/or np.int32, np.float32, and so on.

kernel_tuner/backends/hip.py

@@ -597,8 +597,8 @@ def tune_kernel(
     # check for forbidden names in tune parameters
     util.check_tune_params_list(tune_params, observers, simulation_mode=simulation_mode)
 
-    # check whether block_size_names are used as expected
-    util.check_block_size_params_names_list(block_size_names, tune_params)
+    # check whether block_size_names are used
+    block_size_names = util.check_block_size_params_names_list(block_size_names, tune_params)
 
     # ensure there is always at least three names
     util.append_default_block_size_names(block_size_names)

kernel_tuner/interface.py

@@ -12,7 +12,7 @@
 class PowerSensorObserver(BenchmarkObserver):
     """Observer that an external PowerSensor2 device to accurately measure power
 
-    Requires PowerSensor2 hardware and powersensor Python bindings.
+    Requires PowerSensor3 hardware and powersensor Python bindings.
 
     :param observables: A list of string, containing any of "ps_energy" or "ps_power".
         To measure energy in Joules or power consumption in Watt.
@@ -51,9 +51,8 @@ def after_finish(self):
             )  # Joules
             self.results["ps_energy"].append(ps_measured_e)
         if "ps_power" in self.observables:
-            ps_measured_t = (
-                    end_state.time_at_read - self.begin_state.time_at_read
-            )  # seconds
+            ps_measured_t = ((end_state.time_at_read - self.begin_state.time_at_read).microseconds / 1e6)    # Seconds
+
             self.results["ps_power"].append(ps_measured_e / ps_measured_t)  # Watt
 
     def get_results(self):

kernel_tuner/observers/powersensor.py

@@ -236,11 +236,22 @@ def check_block_size_params_names_list(block_size_names, tune_params):
                     "Block size name " + name + " is not specified in the tunable parameters list!", UserWarning
                 )
     else:  # if default block size names are used
-        if not any([k in default_block_size_names for k in tune_params.keys()]):
+        if not any([k.lower() in default_block_size_names for k in tune_params.keys()]):
             warnings.warn(
                 "None of the tunable parameters specify thread block dimensions!",
                 UserWarning,
             )
+        else:
+            # check for alternative case spelling of defaults such as BLOCK_SIZE_X or block_Size_X etc
+            result = []
+            for k in tune_params.keys():
+                if k.lower() in default_block_size_names and k not in default_block_size_names:
+                    result.append(k)
+            # ensure order of block_size_names is correct regardless of case used
+            block_size_names = sorted(result, key=str.casefold)
+
+    return block_size_names
+
 
 
 def check_restriction(restrict, params: dict) -> bool: