Merge pull request #237 from NVIDIA/ksimpson/add_program_options

Add the options data class to program

Author: leofang

cuda_core/cuda/core/experimental/__init__.py

@@ -7,7 +7,7 @@
 from cuda.core.experimental._event import EventOptions
 from cuda.core.experimental._launcher import LaunchConfig, launch
 from cuda.core.experimental._linker import Linker, LinkerOptions
-from cuda.core.experimental._program import Program
+from cuda.core.experimental._program import Program, ProgramOptions
 from cuda.core.experimental._stream import Stream, StreamOptions
 from cuda.core.experimental._system import System
 

cuda_core/cuda/core/experimental/_linker.py

@@ -9,6 +9,7 @@
 from dataclasses import dataclass
 from typing import List, Optional
 
+from cuda.core.experimental._device import Device
 from cuda.core.experimental._module import ObjectCode
 from cuda.core.experimental._utils import check_or_create_options, driver, handle_return
 
@@ -91,10 +92,10 @@ class LinkerOptions:
 
     Attributes
     ----------
-    arch : str
-        Pass the SM architecture value, such as ``-arch=sm_<CC>`` (for generating CUBIN) or
-        ``compute_<CC>`` (for generating PTX).
-        This is a required option.
+    arch : str, optional
+        Pass the SM architecture value, such as ``sm_<CC>`` (for generating CUBIN) or
+        ``compute_<CC>`` (for generating PTX). If not provided, the current device's architecture
+        will be used.
     max_register_count : int, optional
         Maximum register count.
         Maps to: ``-maxrregcount=<N>``.
@@ -172,7 +173,7 @@ class LinkerOptions:
         Default: False.
     """
 
-    arch: str
+    arch: Optional[str] = None
     max_register_count: Optional[int] = None
     time: Optional[bool] = None
     verbose: Optional[bool] = None
@@ -204,6 +205,8 @@ def __post_init__(self):
     def _init_nvjitlink(self):
         if self.arch is not None:
             self.formatted_options.append(f"-arch={self.arch}")
+        else:
+            self.formatted_options.append("-arch=sm_" + "".join(f"{i}" for i in Device().compute_capability))
         if self.max_register_count is not None:
             self.formatted_options.append(f"-maxrregcount={self.max_register_count}")
         if self.time is not None:

cuda_core/cuda/core/experimental/_program.py

@@ -5,6 +5,7 @@
 import functools
 import importlib.metadata
 from collections import namedtuple
+from collections.abc import Sequence
 from typing import Callable, Dict
 
 try:
@@ -93,6 +94,13 @@ def check_or_create_options(cls, options, options_description, *, keep_none=Fals
     return options
 
 
+def _handle_boolean_option(option: bool) -> str:
+    """
+    Convert a boolean option to a string representation.
+    """
+    return "true" if bool(option) else "false"
+
+
 def precondition(checker: Callable[..., None], what: str = "") -> Callable:
     """
     A decorator that adds checks to ensure any preconditions are met.
@@ -142,6 +150,20 @@ def get_device_from_ctx(ctx_handle) -> int:
     return device_id
 
 
+def is_sequence(obj):
+    """
+    Check if the given object is a sequence (list or tuple).
+    """
+    return isinstance(obj, Sequence)
+
+
+def is_nested_sequence(obj):
+    """
+    Check if the given object is a nested sequence (list or tuple with atleast one list or tuple element).
+    """
+    return is_sequence(obj) and any(is_sequence(elem) for elem in obj)
+
+
 def get_binding_version():
     try:
         major_minor = importlib.metadata.version("cuda-bindings").split(".")[:2]

cuda_core/cuda/core/experimental/_utils.py

@@ -35,6 +35,7 @@ CUDA compilation toolchain
 
    :template: dataclass.rst
 
+   ProgramOptions
    LinkerOptions
 
 

cuda_core/docs/source/api.rst

@@ -8,7 +8,7 @@ and other functionalities.
    :maxdepth: 2
    :caption: Contents:
 
-   release.md
+   release.rst
    install.md
    interoperability.rst
    api.rst

cuda_core/docs/source/index.rst

@@ -0,0 +1,9 @@
+Release Notes
+=============
+
+.. toctree::
+   :maxdepth: 3
+
+   release/0.2.0-notes
+   release/0.1.1-notes
+   release/0.1.0-notes

cuda_core/docs/source/release.md

@@ -1,17 +1,21 @@
-# `cuda.core` v0.1.0 Release notes
+``cuda.core`` 0.1.0 Release Notes
+=================================
 
 Released on Nov 8, 2024
 
-## Hightlights
+Highlights
+----------
+
 - Initial beta release
 - Supports all platforms that CUDA is supported
 - Supports all CUDA 11.x/12.x drivers
 - Supports all CUDA 11.x/12.x Toolkits
 - Pythonic CUDA runtime and other core functionalities
 
-## Limitations
+Limitations
+-----------
 
 - All APIs are currently *experimental* and subject to change without deprecation notice.
-  Please kindly share your feedbacks with us so that we can make `cuda.core` better!
+  Please kindly share your feedback with us so that we can make ``cuda.core`` better!
 - Source code release only; `pip`/`conda` support is coming in a future release
-- Windows TCC mode is [not yet supported](https://github.com/NVIDIA/cuda-python/issues/206)
+- Windows TCC mode is `not yet supported <https://github.com/NVIDIA/cuda-python/issues/206>`_