some fixes to the documentation

docs/source/api/kernelcreation/index.rst

@@ -4,10 +4,15 @@ Kernel Creation
 
 The primary interface for creating numerical kernels in pystencils is the function `create_kernel`.
 
+Configuration
+=============
 
-.. autoclass:: pystencils.CreateKernelConfig
+.. automodule:: pystencils.config
     :members:
 
+Creation
+========
+
 .. autofunction:: pystencils.create_kernel
 
 .. autoclass:: pystencils.backend.KernelFunction

docs/source/backend/ast.rst

@@ -2,11 +2,29 @@
 Abstract Syntax Tree
 ********************
 
+Inheritance Diagramm
+====================
+
+.. inheritance-diagram:: pystencils.backend.ast.astnode.PsAstNode pystencils.backend.ast.structural pystencils.backend.ast.expressions pystencils.backend.extensions.foreign_ast
+    :top-classes: pystencils.types.PsAstNode
+    :parts: 1
+
+
+Base Classes
+============
+
 .. automodule:: pystencils.backend.ast.astnode
     :members:
 
+Structural Nodes
+================
+
 .. automodule:: pystencils.backend.ast.structural
     :members:
 
+
+Expressions
+===========
+
 .. automodule:: pystencils.backend.ast.expressions
     :members:

docs/source/backend/extensions.rst

+************************************
+Extensions and Experimental Features
+************************************
+
+.. automodule:: pystencils.backend.extensions

docs/source/backend/index.rst

@@ -16,6 +16,7 @@ who wish to customize or extend the behaviour of the code generator in their app
     platforms
     transformations
     jit
+    extensions
 
 Internal Representation
 -----------------------

src/pystencils/backend/extensions/__init__.py

+"""
+The module `pystencils.backend.extensions` contains extensions to the pystencils code generator
+beyond its core functionality.
+
+The tools and classes of this module are considered experimental;
+their support by the remaining code generator is limited.
+They can be used to model and generate code outside of the usual scope of pystencils,
+such as non-standard syntax and types.
+At the moment, the primary use case is the modelling of C++ syntax.
+
+
+Foreign Syntax Support
+======================
+
+.. automodule:: pystencils.backend.extensions.foreign_ast
+    :members:
+
+
+C++ Language Support
+====================
+
+.. automodule:: pystencils.backend.extensions.cpp
+    :members:
+
+"""
+
+from .foreign_ast import PsForeignExpression
+
+__all__ = ["PsForeignExpression"]

src/pystencils/backend/extensions/foreign_ast.py

@@ -14,12 +14,12 @@ class PsForeignExpression(PsExpression, ABC):
     
     Foreign expressions are expressions whose properties are not modelled by the pystencils AST,
     and which pystencils therefore does not understand.
-    Support for foreign expressions by the code generator is therefore very limited;
-    as a rule of thumb, only printing is supported.
-    Type checking and most transformations will fail when encountering a `PsForeignExpression`.
 
     There are many situations where non-supported expressions are needed;
     the most common use case is C++ syntax.
+    Support for foreign expressions by the code generator is therefore very limited;
+    as a rule of thumb, only printing is supported.
+    Type checking and most transformations will fail when encountering a `PsForeignExpression`.
     """
 
     __match_args__ = ("children",)

src/pystencils/config.py

@@ -121,7 +121,7 @@ class VectorizationConfig:
 
 @dataclass
 class GpuIndexingConfig:
-    """Configure index translation behaviour inside kernels generated for `Target.SYCL`."""
+    """Configure index translation behaviour for kernels generated for GPU targets."""
 
     omit_range_check: bool = False
     """If set to `True`, omit the iteration counter range check.
@@ -134,10 +134,13 @@ class GpuIndexingConfig:
     sycl_automatic_block_size: bool = True
     """If set to `True` while generating for `Target.SYCL`, let the SYCL runtime decide on the block size.
 
-    If set to `True`, the kernel is generated for execution via ``parallel_for``-dispatch using
+    If set to `True`, the kernel is generated for execution via
+    `parallel_for <https://registry.khronos.org/SYCL/specs/sycl-2020/html/sycl-2020.html#_parallel_for_invoke>`_
+    -dispatch using
     a flat `sycl::range`. In this case, the GPU block size will be inferred by the SYCL runtime.
 
-    If set to `False`, the kernel will receive an `nd_item` and has to be executed using ``parallel_for``
+    If set to `False`, the kernel will receive an `nd_item` and has to be executed using
+    `parallel_for <https://registry.khronos.org/SYCL/specs/sycl-2020/html/sycl-2020.html#_parallel_for_invoke>`_
     with an `nd_range`. This allows manual specification of the block size.
     """