xobjects 0.2.7__tar.gz → 0.2.9__tar.gz

{xobjects-0.2.7/xobjects.egg-info → xobjects-0.2.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: xobjects
-Version: 0.2.7
+Version: 0.2.9
 Summary: In-memory serialization and code generator for CPU and GPU
 Home-page: https://xsuite.readthedocs.io/
 Download-URL: https://pypi.python.org/pypi/xobjects

xobjects-0.2.9/xobjects/_version.py

@@ -0,0 +1 @@
+__version__ = "0.2.9"

{xobjects-0.2.7 → xobjects-0.2.9}/xobjects/context_cpu.py

@@ -16,6 +16,8 @@ from .general import _print
 import numpy as np
 import scipy as sp
 
+_forbid_compile = False
+
 from .context import (
     Kernel,
     ModuleNotAvailable,
@@ -301,6 +303,9 @@ class ContextCpu(XContext):
             cdefs = "\n".join(cls._gen_c_decl({}) for cls in classes)
             cdefs += "\n" + extra_cdef
 
+            if _forbid_compile:
+                raise RuntimeError("Compilation is forbidden")
+
             so_file = self.compile_kernel(
                 module_name,
                 kernel_descriptions,
@@ -407,6 +412,12 @@ class ContextCpu(XContext):
             xtr_compile_args.append("-fopenmp")
             xtr_link_args.append("-fopenmp")
 
+            # https://mac.r-project.org/openmp/
+            # on macos comment the above and uncomment the below flags to compile OpenMP with Xcode clang:
+            #xtr_compile_args.append("-Xclang")
+            #xtr_compile_args.append("-fopenmp")
+            #xtr_link_args.append("-lomp")
+
         if os.name == "nt":  # windows
             # TODO: to be handled properly
             xtr_compile_args = []

{xobjects-0.2.7 → xobjects-0.2.9}/xobjects/context_cupy.py

@@ -392,13 +392,14 @@ class ContextCupy(XContext):
     def linked_array_type(self):
         return LinkedArrayCupy
 
-    def __init__(self, default_block_size=256, device=None):
+    def __init__(self, default_block_size=256, default_shared_mem_size_bytes=0, device=None):
         if device is not None:
             cupy.cuda.Device(device).use()
 
         super().__init__()
 
         self.default_block_size = default_block_size
+        self.default_shared_mem_size_bytes = default_shared_mem_size_bytes
 
     def _make_buffer(self, capacity):
         return BufferCupy(capacity=capacity, context=self)
@@ -417,7 +418,6 @@ class ContextCupy(XContext):
     ) -> Dict[Tuple[str, tuple], "KernelCupy"]:
         if not compile:
             raise NotImplementedError("compile=False available only on CPU.")
-
         classes = list(classes_from_kernels(kernel_descriptions))
         classes += list(extra_classes)
         classes = sort_classes(classes)
@@ -455,12 +455,12 @@ class ContextCupy(XContext):
         for pyname, kernel in kernel_descriptions.items():
             if kernel.c_name is None:
                 kernel.c_name = pyname
-
             out_kernels[pyname] = KernelCupy(
                 function=module.get_function(kernel.c_name),
                 description=kernel,
                 block_size=self.default_block_size,
                 context=self,
+                shared_mem_size_bytes=self.default_shared_mem_size_bytes
             )
 
             out_kernels[pyname].source = source
@@ -633,11 +633,13 @@ class KernelCupy(object):
         description,
         block_size,
         context,
+        shared_mem_size_bytes
     ):
         self.function = function
         self.description = description
         self.block_size = block_size
         self.context = context
+        self.shared_mem_size_bytes = shared_mem_size_bytes
 
     def to_function_arg(self, arg, value):
         if arg.pointer:
@@ -682,7 +684,7 @@ class KernelCupy(object):
             n_threads = self.description.n_threads
 
         grid_size = int(np.ceil(n_threads / self.block_size))
-        self.function((grid_size,), (self.block_size,), arg_list)
+        self.function((grid_size,), (self.block_size,), arg_list, shared_mem=self.shared_mem_size_bytes)
 
 
 class FFTCupy(object):

{xobjects-0.2.7 → xobjects-0.2.9}/xobjects/hybrid_class.py

@@ -301,7 +301,13 @@ class HybridClass(metaclass=MetaHybridClass):
 
         if hasattr(obj, "_store_in_to_dict"):
             for nn in obj._store_in_to_dict:
-                out[nn] = getattr(obj, nn)
+                ww = getattr(obj, nn)
+                if hasattr(ww, "to_dict"):
+                    out[nn] = ww.to_dict()
+                elif hasattr(ww, "_to_dict"):
+                    out[nn] = ww._to_dict()
+                else:
+                    out[nn] = ww
 
         return out
 

{xobjects-0.2.7 → xobjects-0.2.9/xobjects.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: xobjects
-Version: 0.2.7
+Version: 0.2.9
 Summary: In-memory serialization and code generator for CPU and GPU
 Home-page: https://xsuite.readthedocs.io/
 Download-URL: https://pypi.python.org/pypi/xobjects

xobjects-0.2.7/xobjects/_version.py

@@ -1 +0,0 @@
-__version__ = "0.2.7"