v23.03.00

v23.03.00

Author: marcinz

.github/workflows/require-labels.yml

@@ -6,13 +6,10 @@ jobs:
   label:
     runs-on: ubuntu-latest
     steps:
-      - name: Delay checking labels if PR is just created
-        if: ${{ github.event.action == 'opened' }}
-        run: sleep 300s
-        shell: bash
       - name: Check Labels
-        uses: mheap/github-action-required-labels@v2
+        uses: mheap/github-action-required-labels@v3
         with:
           mode: exactly
           count: 1
-          labels: "category:new-feature, category:improvement, category:bug-fix, category:task"
+          labels: "category:new-feature, category:improvement, category:bug-fix, category:task, category:documentation"
+          

.gitignore

@@ -19,7 +19,6 @@ core
 *.bbl
 *.blg
 *.aux
-*.json
 *.dylib
 legate_python
 openblas
@@ -41,3 +40,4 @@ cunumeric/install_info.py
 /dist
 /_skbuild
 /_cmake_test_compile
+.ipynb_checkpoints

.pre-commit-config.yaml

@@ -1,17 +1,17 @@
 repos:
     - repo: https://github.com/pre-commit/mirrors-mypy
-      rev: 'v0.991'
+      rev: 'v1.0.1'
       hooks:
         - id: mypy
           language: system
           pass_filenames: false
           args: ['cunumeric']
     - repo: https://github.com/PyCQA/isort
-      rev: 5.11.4
+      rev: 5.12.0
       hooks:
             - id: isort
     - repo: https://github.com/psf/black
-      rev: 22.12.0
+      rev: 23.1.0
       hooks:
             - id: black
     - repo: https://github.com/PyCQA/flake8
@@ -26,10 +26,10 @@ repos:
           types_or: []
     - repo: local
       hooks:
-        - id: enforce-pytest-main
-          name: Enforce pytest.main
-          description: Make sure tests use sys.ext(pytest.main(...))
-          entry: python scripts/hooks/enforce_pytest_main.py
+        - id: enforce-boilerplate
+          name: Enforce boilerplate
+          description: Make sure tests include required boilerplate
+          entry: python scripts/hooks/enforce_boilerplate.py
           language: python
           pass_filenames: false
 

BUILD.md

@@ -15,7 +15,10 @@ limitations under the License.
 
 -->
 
-# Basic build
+(building-from-source)=
+# Building from source
+
+## Basic build
 
 Users must have a working installation of the
 [Legate Core](https://github.com/nv-legate/legate.core)
@@ -30,9 +33,9 @@ Once Legate Core is installed, you can simply invoke `./install.py` from the
 cuNumeric top-level directory. The build will automatically pick up the
 configuration used when building Legate Core (e.g. the CUDA Toolkit directory).
 
-# Advanced topics
+## Advanced topics
 
-## Building through pip & cmake
+### Building through pip & cmake
 
 cuNumeric uses the same cmake/scikit-build-based build workflow as Legate Core.
 See the
@@ -74,4 +77,4 @@ The Python source tree and CMake build tree are now available with the environme
 for running cuNumeric programs. The diagram below illustrates the
 complete workflow for building both Legate core and cuNumeric.
 
-<img src="docs/figures/developer-build.png" alt="drawing" width="600"/>
+<img src="https://raw.githubusercontent.com/nv-legate/cunumeric/branch-23.03/docs/figures/developer-build.png" alt="drawing" width="600"/>

CMakeLists.txt

@@ -37,7 +37,7 @@ endif()
 # - Download and initialize RAPIDS CMake helpers -----------------------------
 
 if(NOT EXISTS ${CMAKE_BINARY_DIR}/RAPIDS.cmake)
-  file(DOWNLOAD https://raw.githubusercontent.com/rapidsai/rapids-cmake/branch-22.08/RAPIDS.cmake
+  file(DOWNLOAD https://raw.githubusercontent.com/rapidsai/rapids-cmake/branch-23.02/RAPIDS.cmake
        ${CMAKE_BINARY_DIR}/RAPIDS.cmake)
 endif()
 include(${CMAKE_BINARY_DIR}/RAPIDS.cmake)
@@ -47,7 +47,7 @@ include(rapids-cuda)
 include(rapids-export)
 include(rapids-find)
 
-set(cunumeric_version 22.12.00)
+set(cunumeric_version 23.03.00)
 
 # For now we want the optimization flags to match on both normal make and cmake
 # builds so we override the cmake defaults here for release, this changes

README.md

@@ -22,7 +22,7 @@ that aims to provide a distributed and accelerated drop-in replacement for the
 [NumPy API](https://numpy.org/doc/stable/reference/) on top of the
 [Legion](https://legion.stanford.edu) runtime. Using cuNumeric you do things like run
 [the final example of the Python CFD course](https://github.com/barbagroup/CFDPython/blob/master/lessons/15_Step_12.ipynb)
-completely unmodified on 2048 A100 GPUs in a [DGX SuperPOD](https://github.com/barbagroup/CFDPython/blob/master/lessons/15_Step_12.ipynb) and achieve good weak scaling.
+completely unmodified on 2048 A100 GPUs in a [DGX SuperPOD](https://www.nvidia.com/en-us/data-center/dgx-superpod/) and achieve good weak scaling.
 
 <img src="docs/figures/cfd-demo.png" alt="drawing" width="500"/>
 
@@ -43,8 +43,12 @@ cuNumeric is available [on conda](https://anaconda.org/legate/cunumeric):
 conda install -c nvidia -c conda-forge -c legate cunumeric
 ```
 
-The conda package is compatible with CUDA >= 11.4 (CUDA driver version >= r470),
-and Volta or later GPU architectures.
+Only linux-64 packages are available at the moment.
+
+The default package contains GPU support, and is compatible with CUDA >= 11.4
+(CUDA driver version >= r470), and Volta or later GPU architectures. There are
+also CPU-only packages available, and will be automatically selected by `conda`
+when installing on a machine without GPUs.
 
 See [BUILD.md](BUILD.md) for instructions on building cuNumeric from source.
 
@@ -76,8 +80,9 @@ users can supply the `--nodes` option. For execution with GPUs, users can use th
 to familiarize themselves with these resource flags as described in the Legate Core
 documentation or simply by passing `--help` to the `legate` driver script.
 
-You can use `test.py` to run the test suite. This script will invoke the `legate`
-driver script automatically. Check out `test.py --help` for further options.
+You can use `test.py` to run the test suite. Invoke the script directly or through
+standard `python`; the script will invoke the `legate` driver script internally.
+Check out `test.py --help` for further options.
 
 ## Supported and Planned Features
 
@@ -100,8 +105,8 @@ installing legate.core with a larger `--max-dim`.
 
 ## Documentation
 
-A complete list of available features can is provided in the [API
-reference](https://nv-legate.github.io/cunumeric/api/index.html).
+The cuNumeric documentation can be found
+[here](https://nv-legate.github.io/cunumeric).
 
 ## Future Directions
 

cmake/Modules/cuda_arch_helpers.cmake

@@ -44,6 +44,9 @@ function(set_cuda_arch_from_names)
   if(CMAKE_CUDA_ARCHITECTURES MATCHES "ampere")
     list(APPEND cuda_archs 80)
   endif()
+  if(CMAKE_CUDA_ARCHITECTURES MATCHES "hopper")
+    list(APPEND cuda_archs 90)
+  endif()
 
   if(cuda_archs)
     list(LENGTH cuda_archs num_archs)
@@ -83,6 +86,7 @@ function(add_cuda_architecture_defines defs)
   add_def_if_arch_enabled("70" "VOLTA_ARCH")
   add_def_if_arch_enabled("75" "TURING_ARCH")
   add_def_if_arch_enabled("80" "AMPERE_ARCH")
+  add_def_if_arch_enabled("90" "HOPPER_ARCH")
 
   set(${defs} ${_defs} PARENT_SCOPE)
 endfunction()

cmake/thirdparty/get_legate_core.cmake

@@ -1,5 +1,5 @@
 #=============================================================================
-# Copyright 2022 NVIDIA Corporation
+# Copyright 2022-2023 NVIDIA Corporation
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -21,6 +21,18 @@ function(find_or_configure_legate_core)
   include("${rapids-cmake-dir}/export/detail/parse_version.cmake")
   rapids_export_parse_version(${PKG_VERSION} legate_core PKG_VERSION)
 
+  include("${rapids-cmake-dir}/cpm/detail/package_details.cmake")
+  rapids_cpm_package_details(legate_core version git_repo git_branch shallow exclude_from_all)
+
+  set(version ${PKG_VERSION})
+  set(exclude_from_all ${PKG_EXCLUDE_FROM_ALL})
+  if(PKG_BRANCH)
+    set(git_branch "${PKG_BRANCH}")
+  endif()
+  if(PKG_REPOSITORY)
+    set(git_repo "${PKG_REPOSITORY}")
+  endif()
+
   set(FIND_PKG_ARGS
       GLOBAL_TARGETS     legate::core
       BUILD_EXPORT_SET   cunumeric-exports
@@ -35,18 +47,24 @@ function(find_or_configure_legate_core)
   if(legate_core_DIR OR legate_core_ROOT)
     set(_find_mode REQUIRED)
   endif()
-  rapids_find_package(legate_core ${PKG_VERSION} EXACT CONFIG ${_find_mode} ${FIND_PKG_ARGS})
+  rapids_find_package(legate_core ${version} EXACT CONFIG ${_find_mode} ${FIND_PKG_ARGS})
 
   if(legate_core_FOUND)
-    message(STATUS "CPM: using local package legate_core@${PKG_VERSION}")
+    message(STATUS "CPM: using local package legate_core@${version}")
   else()
     include(${CMAKE_CURRENT_SOURCE_DIR}/cmake/Modules/cpm_helpers.cmake)
-    get_cpm_git_args(legate_core_cpm_git_args REPOSITORY ${PKG_REPOSITORY} BRANCH ${PKG_BRANCH})
-    rapids_cpm_find(legate_core ${PKG_VERSION} ${FIND_PKG_ARGS}
+    get_cpm_git_args(legate_core_cpm_git_args REPOSITORY ${git_repo} BRANCH ${git_branch})
+
+    message(VERBOSE "cunumeric: legate.core version: ${version}")
+    message(VERBOSE "cunumeric: legate.core git_repo: ${git_repo}")
+    message(VERBOSE "cunumeric: legate.core git_branch: ${git_branch}")
+    message(VERBOSE "cunumeric: legate.core exclude_from_all: ${exclude_from_all}")
+
+    rapids_cpm_find(legate_core ${version} ${FIND_PKG_ARGS}
         CPM_ARGS
           ${legate_core_cpm_git_args}
           FIND_PACKAGE_ARGUMENTS EXACT
-          EXCLUDE_FROM_ALL      ${PKG_EXCLUDE_FROM_ALL}
+          EXCLUDE_FROM_ALL       ${exclude_from_all}
     )
   endif()
 
@@ -59,15 +77,26 @@ function(find_or_configure_legate_core)
   message(VERBOSE "Legion_BOUNDS_CHECKS=${Legion_BOUNDS_CHECKS}")
 endfunction()
 
-if(NOT DEFINED cunumeric_LEGATE_CORE_BRANCH)
-  set(cunumeric_LEGATE_CORE_BRANCH branch-22.10)
-endif()
+foreach(_var IN ITEMS "cunumeric_LEGATE_CORE_VERSION"
+                      "cunumeric_LEGATE_CORE_BRANCH"
+                      "cunumeric_LEGATE_CORE_REPOSITORY"
+                      "cunumeric_EXCLUDE_LEGATE_CORE_FROM_ALL")
+  if(DEFINED ${_var})
+    # Create a cunumeric_LEGATE_CORE_BRANCH variable in the current scope either from the existing
+    # current-scope variable, or the cache variable.
+    set(${_var} "${${_var}}")
+    # Remove cunumeric_LEGATE_CORE_BRANCH from the CMakeCache.txt. This ensures reconfiguring the same
+    # build dir without passing `-Dcunumeric_LEGATE_CORE_BRANCH=` reverts to the value in versions.json
+    # instead of reusing the previous `-Dcunumeric_LEGATE_CORE_BRANCH=` value.
+    unset(${_var} CACHE)
+  endif()
+endforeach()
 
-if(NOT DEFINED cunumeric_LEGATE_CORE_REPOSITORY)
-  set(cunumeric_LEGATE_CORE_REPOSITORY https://github.com/nv-legate/legate.core.git)
+if(NOT DEFINED cunumeric_LEGATE_CORE_VERSION)
+  set(cunumeric_LEGATE_CORE_VERSION "${cunumeric_VERSION}")
 endif()
 
-find_or_configure_legate_core(VERSION          ${cunumeric_VERSION}
+find_or_configure_legate_core(VERSION          ${cunumeric_LEGATE_CORE_VERSION}
                               REPOSITORY       ${cunumeric_LEGATE_CORE_REPOSITORY}
                               BRANCH           ${cunumeric_LEGATE_CORE_BRANCH}
                               EXCLUDE_FROM_ALL ${cunumeric_EXCLUDE_LEGATE_CORE_FROM_ALL}

cmake/versions.json

@@ -0,0 +1,8 @@
+{
+  "packages" : {
+    "legate_core" : {
+      "git_url" : "https://github.com/nv-legate/legate.core.git",
+      "git_tag" : "37596159b1f8f06439fdc95d6090bbc4315c302c"
+    }
+  }
+}

conda/conda-build/build.sh

@@ -13,7 +13,8 @@ if [ -z "$CPU_ONLY" ]; then
   # cutensor, relying on the conda cutensor package
   CMAKE_ARGS+="
 -Dcutensor_DIR=$PREFIX
--DCMAKE_CUDA_ARCHITECTURES:LIST=60-real;70-real;75-real;80-real;86
+-DCMAKE_CUDA_ARCHITECTURES:LIST=60-real;70-real;75-real;80-real;90
+-DBUILD_MARCH=haswell
 "
 else
   # When we build without cuda, we need to provide the location of curand
@@ -32,7 +33,7 @@ export CUDAHOSTCXX=${CXX}
 
 echo "Build starting on $(date)"
 
-cmake -S . -B build ${CMAKE_ARGS}
+cmake -S . -B build ${CMAKE_ARGS} -DCMAKE_BUILD_PARALLEL_LEVEL=$CPU_COUNT
 cmake --build build -j$CPU_COUNT
 cmake --install build
 

conda/conda-build/conda_build_config.yaml

@@ -3,9 +3,9 @@ gpu_enabled:
   - false
 
 python:
-  - 3.8
-  - 3.9
+  - "3.9,!=3.9.7"
   - 3.10
+  - 3.11
 
 numpy_version:
   - ">=1.22"

conda/conda-build/meta.yaml

@@ -104,10 +104,7 @@ requirements:
     # the nvcc requirement is necessary because it contains crt/host_config.h used by cuda runtime. This is a packaging bug that has been reported.
     - cuda-nvcc ={{ cuda_version }}
     # libcurand is used both in CPU and GPU builds
-    # temporarily pin curand until problems are resolved
-    - libcurand-dev =10.3.0.86
-    # the following line is only necessary for pinning curand
-    - libcurand =10.3.0.86
+    - libcurand-dev
     # cudart needed for CPU and GPU builds because of curand
     - cuda-cudart-dev ={{ cuda_version }}
     - python
@@ -140,7 +137,6 @@ requirements:
 {% else %}
     - legate-core ={{ core_version }}
     - cuda-cudart >={{ cuda_version }}
-    # - libcutensor >=1.3
     - cutensor >=1.3 =*_*
     - libcublas
     - libcusolver =11.4.1.48-0

cunumeric/__init__.py

@@ -24,19 +24,22 @@
 """
 from __future__ import annotations
 
+import os
+
 import numpy as _np
 
 from cunumeric import linalg, random, fft
-from cunumeric.array import ndarray
+from cunumeric.array import maybe_convert_to_np_ndarray, ndarray
 from cunumeric.bits import packbits, unpackbits
 from cunumeric.module import *
 from cunumeric._ufunc import *
 from cunumeric.logic import *
 from cunumeric.window import bartlett, blackman, hamming, hanning, kaiser
 from cunumeric.coverage import clone_module
 
-clone_module(_np, globals())
+clone_module(_np, globals(), maybe_convert_to_np_ndarray)
 
+del maybe_convert_to_np_ndarray
 del clone_module
 del _np
 

cunumeric/_sphinxext/comparison_table.py

@@ -29,7 +29,6 @@
 
 
 class ComparisonTable(CunumericDirective):
-
     has_content = False
     required_arguments = 0
     optional_arguments = 1

cunumeric/_sphinxext/implemented_index.py

@@ -46,13 +46,11 @@ def _filter(x: Any) -> bool:
 
 
 class ImplementedIndex(CunumericDirective):
-
     has_content = False
     required_arguments = 0
     optional_arguments = 0
 
     def run(self) -> nodes.Node:
-
         refs: list[str] = []
         for ns in namespaces:
             refs += [

cunumeric/_sphinxext/missing_refs.py

@@ -53,7 +53,6 @@
 
 
 class MissingRefs(SphinxPostTransform):
-
     default_priority = 5
 
     def run(self, **kwargs: Any) -> None:

cunumeric/_ufunc/ufunc.py

@@ -178,7 +178,6 @@ def to_dtypes(chars: str) -> tuple[np.dtype[Any], ...]:
 
 
 class ufunc:
-
     _types: Dict[Any, str]
     _nin: int
     _nout: int