Changeset dd7fc12 in sasmodels for sasmodels/kernelcl.py

Timestamp:

Apr 15, 2016 11:11:43 AM (8 years ago)

Author:

Paul Kienzle <pkienzle@…>

Branches:

master, core_shell_microgels, costrafo411, magnetic_model, release_v0.94, release_v0.95, ticket-1257-vesicle-product, ticket_1156, ticket_1265_superball, ticket_822_more_unit_tests

Children:

3599d36

Parents:

b151003

Message:

fix kerneldll dtype problem; more type hinting

File:

• sasmodels/kernelcl.py (modified) (22 diffs)

sasmodels/kernelcl.py

@@ -103 +103 @@
 ENV = None
 def environment():
+    # type: () -> "GpuEnvironment"
     """
     Returns a singleton :class:`GpuEnvironment`.
@@ -114 +115 @@
 
 def has_type(device, dtype):
+    # type: (cl.Device, np.dtype) -> bool
     """
     Return true if device supports the requested precision.
@@ -127 +129 @@
 
 def get_warp(kernel, queue):
+    # type: (cl.Kernel, cl.CommandQueue) -> int
     """
     Return the size of an execution batch for *kernel* running on *queue*.
@@ -135 +138 @@
 
 def _stretch_input(vector, dtype, extra=1e-3, boundary=32):
+    # type: (np.ndarray, np.dtype, float, int) -> np.ndarray
     """
     Stretch an input vector to the correct boundary.
@@ -157 +161 @@
 
 def compile_model(context, source, dtype, fast=False):
+    # type: (cl.Context, str, np.dtype, bool) -> cl.Program
     """
     Build a model to run on the gpu.
@@ -192 +197 @@
     """
     def __init__(self):
+        # type: () -> None
         # find gpu context
         #self.context = cl.create_some_context()
@@ -210 +216 @@
 
     def has_type(self, dtype):
+        # type: (np.dtype) -> bool
         """
         Return True if all devices support a given type.
         """
-        dtype = generate.F32 if dtype == 'fast' else np.dtype(dtype)
         return any(has_type(d, dtype)
                    for context in self.context
@@ -219 +225 @@
 
     def get_queue(self, dtype):
+        # type: (np.dtype) -> cl.CommandQueue
         """
         Return a command queue for the kernels of type dtype.
@@ -227 +234 @@
 
     def get_context(self, dtype):
+        # type: (np.dtype) -> cl.Context
         """
         Return a OpenCL context for the kernels of type dtype.
@@ -235 +243 @@
 
     def _create_some_context(self):
+        # type: () -> cl.Context
         """
         Protected call to cl.create_some_context without interactivity.  Use
@@ -248 +257 @@
 
     def compile_program(self, name, source, dtype, fast=False):
+        # type: (str, str, np.dtype, bool) -> cl.Program
         """
         Compile the program for the device in the given context.
@@ -261 +271 @@
 
     def release_program(self, name):
+        # type: (str) -> None
         """
         Free memory associated with the program on the device.
@@ -269 +280 @@
 
 def _get_default_context():
+    # type: () -> cl.Context
     """
     Get an OpenCL context, preferring GPU over CPU, and preferring Intel
@@ -334 +346 @@
     that the compiler is allowed to take shortcuts.
     """
-    def __init__(self, source, model_info, dtype=generate.F32):
+    def __init__(self, source, model_info, dtype=generate.F32, fast=False):
+        # type: (str, ModelInfo, np.dtype, bool) -> None
         self.info = model_info
         self.source = source
-        self.dtype = generate.F32 if dtype == 'fast' else np.dtype(dtype)
-        self.fast = (dtype == 'fast')
+        self.dtype = dtype
+        self.fast = fast
         self.program = None # delay program creation
 
     def __getstate__(self):
+        # type: () -> Tuple[ModelInfo, str, np.dtype, bool]
         return self.info, self.source, self.dtype, self.fast
 
     def __setstate__(self, state):
+        # type: (Tuple[ModelInfo, str, np.dtype, bool]) -> None
         self.info, self.source, self.dtype, self.fast = state
         self.program = None
 
     def make_kernel(self, q_vectors):
+        # type: (List[np.ndarray]) -> "GpuKernel"
         if self.program is None:
             compiler = environment().compile_program
@@ -356 +372 @@
         kernel_name = generate.kernel_name(self.info, is_2d)
         kernel = getattr(self.program, kernel_name)
-        return GpuKernel(kernel, self.info, q_vectors, self.dtype)
+        return GpuKernel(kernel, self.info, q_vectors)
 
     def release(self):
+        # type: () -> None
         """
         Free the resources associated with the model.
@@ -367 +384 @@
 
     def __del__(self):
+        # type: () -> None
         self.release()
 
@@ -390 +408 @@
     """
     def __init__(self, q_vectors, dtype=generate.F32):
+        # type: (List[np.ndarray], np.dtype) -> None
         # TODO: do we ever need double precision q?
         env = environment()
@@ -419 +438 @@
 
     def release(self):
+        # type: () -> None
         """
         Free the memory.
@@ -427 +447 @@
 
     def __del__(self):
+        # type: () -> None
         self.release()
 
@@ -450 +471 @@
     """
     def __init__(self, kernel, model_info, q_vectors):
-        # type: (KernelModel, ModelInfo, List[np.ndarray]) -> None
+        # type: (cl.Kernel, ModelInfo, List[np.ndarray]) -> None
         max_pd = model_info.parameters.max_pd
         npars = len(model_info.parameters.kernel_parameters)-2
@@ -505 +526 @@
 
     def release(self):
+        # type: () -> None
         """
         Release resources associated with the kernel.
@@ -513 +535 @@
 
     def __del__(self):
+        # type: () -> None
         self.release()