Merge pull request #33 from gunrock/muhosama/improvements

Build system improvements.

Author: neoblizz

.github/workflows/clang-format.yml

@@ -9,9 +9,9 @@ on:
 jobs:
   formatting-check:
     name: Format
-    runs-on: ubuntu-22.04
+    runs-on: ubuntu-latest
     steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v5
       with:
         ref: ${{ github.event.pull_request.head.ref }}
 

.github/workflows/ubuntu-22.04.yml

@@ -25,20 +25,20 @@ jobs:
 
     # Steps represent a sequence of tasks that will be executed as part of the job
     steps:
-      # Fetch CUDA toolkit using Jimver/cuda-toolkit@v0.2.7
+      # Fetch CUDA toolkit using Jimver/cuda-toolkit
       - name: Fetch CUDA toolkit
-        uses: Jimver/cuda-toolkit@v0.2.7
+        uses: Jimver/cuda-toolkit@v0.2.27
         id: cuda-toolkit
         with:
-          cuda: '11.7.0'
+          cuda: '13.0.0'
           linux-local-args: '["--toolkit"]'
 
       # Runs a single command using the runners shell
       - name: Check nvcc version
         run: nvcc -V
 
       # Checks-out your repository under $GITHUB_WORKSPACE, so your job can access it
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v5
 
       # Configure CMake in a 'build' subdirectory. `CMAKE_BUILD_TYPE` is only required if you are using a single-configuration generator such as make.
       # See https://cmake.org/cmake/help/latest/variable/CMAKE_BUILD_TYPE.html?highlight=cmake_build_type

.github/workflows/ubuntu-20.04.yml .github/workflows/ubuntu-latest.yml.github/workflows/ubuntu-20.04.yml renamed to .github/workflows/ubuntu-latest.yml

@@ -1,4 +1,4 @@
-name: ubuntu-20.04
+name: ubuntu-latest
 
 # Controls when the workflow will run
 on:
@@ -21,24 +21,24 @@ jobs:
   build:
     # https://github.blog/changelog/2021-02-08-github-actions-skip-pull-request-and-push-workflows-with-skip-ci/
     if: "!contains(github.event.commits[0].message, '[skip build]')"
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
 
     # Steps represent a sequence of tasks that will be executed as part of the job
     steps:
-      # Fetch CUDA toolkit using Jimver/cuda-toolkit@v0.2.7
+      # Fetch CUDA toolkit using Jimver/cuda-toolkit
       - name: Fetch CUDA toolkit
-        uses: Jimver/cuda-toolkit@v0.2.7
+        uses: Jimver/cuda-toolkit@v0.2.27
         id: cuda-toolkit
         with:
-          cuda: '11.7.0'
+          cuda: '13.0.0'
           linux-local-args: '["--toolkit"]'
 
       # Runs a single command using the runners shell
       - name: Check nvcc version
         run: nvcc -V
 
       # Checks-out your repository under $GITHUB_WORKSPACE, so your job can access it
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v5
 
       # Configure CMake in a 'build' subdirectory. `CMAKE_BUILD_TYPE` is only required if you are using a single-configuration generator such as make.
       # See https://cmake.org/cmake/help/latest/variable/CMAKE_BUILD_TYPE.html?highlight=cmake_build_type

.github/workflows/windows-2019.yml

@@ -1,4 +1,4 @@
-name: windows-2022
+name: windows-latest
 
 # Controls when the workflow will run
 on:
@@ -21,24 +21,24 @@ jobs:
   build:
     # https://github.blog/changelog/2021-02-08-github-actions-skip-pull-request-and-push-workflows-with-skip-ci/
     if: "!contains(github.event.commits[0].message, '[skip build]')"
-    runs-on: windows-2022
+    runs-on: windows-latest
 
     # Steps represent a sequence of tasks that will be executed as part of the job
     steps:
       # Fetch CUDA toolkit using Jimver/cuda-toolkit@v0.2.7
       - name: Fetch CUDA toolkit
-        uses: Jimver/cuda-toolkit@v0.2.7
+        uses: Jimver/cuda-toolkit@v0.2.27
         id: cuda-toolkit
         with:
-          cuda: '11.7.0'
+          cuda: '13.0.0'
           linux-local-args: '["--toolkit"]'
 
       # Runs a single command using the runners shell
       - name: Check nvcc version
         run: nvcc -V
 
       # Checks-out your repository under $GITHUB_WORKSPACE, so your job can access it
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v5
 
       # Configure CMake in a 'build' subdirectory. `CMAKE_BUILD_TYPE` is only required if you are using a single-configuration generator such as make.
       # See https://cmake.org/cmake/help/latest/variable/CMAKE_BUILD_TYPE.html?highlight=cmake_build_type

.github/workflows/windows-2022.yml .github/workflows/windows-latest.yml.github/workflows/windows-2022.yml renamed to .github/workflows/windows-latest.yml

@@ -9,19 +9,21 @@ Prior to the work we present and propose here, the only way to unleash the GPU's
 With our open-source framework, we hope to not only improve programmers' productivity when developing irregular-parallel algorithms on the GPU but also improve the overall performance characteristics for such applications by allowing a quick path to experimentation with a variety of existing load-balancing techniques. Consequently, we also hope that by separating the concerns of load-balancing from work processing within our abstraction, managing and extending existing code to future architectures becomes easier.
 
 ## Requirements
-- **OS:** Ubuntu 18.04, 20.04, Windows
-- **Hardware:** NVIDIA GPU (Volta or newer)
+
+- **OS:** Ubuntu 24.04, Windows
+- **Hardware:** NVIDIA GPU (Turing or newer)
 - **Software:** CUDA 11.7 or above and cmake 3.20.1 or above.
 - **CUDA Architecture:** SM 70 or above (see [GPUs supported](https://en.wikipedia.org/wiki/CUDA#GPUs_supported)), this is specified using cmake's command: `-DCMAKE_CUDA_ARCHITECTURES=70`. Alternatively, set the CUDA architecture version in the `CMakeLists.txt` file directly: [CMakeLists.txt#72](https://github.com/gunrock/loops/blob/main/CMakeLists.txt#L72).
 
 ## Getting Started
+
 Before building `loops` make sure you have CUDA Toolkit and cmake installed on your system, and exported in `PATH` of your system. Other external dependencies such as `NVIDIA/thrust`, `NVIDIA/cub`, etc. are automatically fetched using cmake.
 
 ```bash
 git clone https://github.com/gunrock/loops.git
 cd loops
 mkdir build && cd build
-cmake -DCMAKE_CUDA_ARCHITECTURES=70 .. # Volta = 70, Turing = 75, ...
+cmake -DCMAKE_CUDA_ARCHITECTURES=75 .. # Turing = 75, ...
 make -j$(nproc)
 bin/loops.spmv.merge_path -m ../datasets/chesapeake/chesapeake.mtx
 ```
@@ -31,7 +33,9 @@ bin/loops.spmv.merge_path -m ../datasets/chesapeake/chesapeake.mtx
 ```bash
 make loops.spmv.<algorithm>
 ```
+
 Replaced the `<algorithm>` with one of the following algorithm names to build a specific SpMV algorithm instead of all of them:
+
 - `original`
 - `thread_mapped`
 - `group_mapped`
@@ -44,6 +48,7 @@ An example of the above: `make loops.spmv.merge_path`.
 
 To download the SuiteSparse Matrix Collection[^1], simply run the following command. We recommend using a `tmux` session, because downloading the entire collection can take a significant time. Uncompress the dataset by running the following command in the dataset's directory `find . -name '*.tar.gz' -execdir tar -xzvf '{}' \;
 ` The total downloaded size of the dataset is nontrivial: uncompressed + compressed = 887GB.
+
 ```bash
 wget --recursive --no-parent --force-directories -l inf -X RB,mat \ 
 --accept "*.tar.gz" "https://suitesparse-collection-website.herokuapp.com/"
@@ -59,15 +64,20 @@ wget --recursive --no-parent --force-directories -l inf -X RB,mat \
 [^1]: Timothy A. Davis and Yifan Hu. 2011. The University of Florida Sparse Matrix Collection. ACM Transactions on Mathematical Software 38, 1, Article 1 (December 2011), 25 pages. DOI: https://doi.org/10.1145/2049662.2049663
 
 ## Experimentation
+
 If CUDA and cmake are already setup, follow the [Getting Started](#getting-started) instructions. Or, you may prefer to set up the entire project using docker, and for that we have provided a docker file and instructions on how to use it in [/docker](https://github.com/gunrock/loops/tree/main/docker) directory.
 
 ### Sanity Check
+
 Run the following command in the cmake's `build` folder:
+
 ```bash
 bin/loops.spmv.merge_path -m ../datasets/chesapeake/chesapeake.mtx \
  --validate -v
 ```
+
 You should approximately see the following output:
+
 ```bash
 ~/loops/build$ bin/loops.spmv.merge_path \
  -m ../datasets/chesapeake/chesapeake.mtx --validate -v
@@ -76,7 +86,9 @@ Matrix:         chesapeake.mtx
 Dimensions:     39 x 39 (340)
 Errors:         0
 ```
+
 ## Reproducing Results
+
 > Find pre-generated results in [plots/](https://github.com/gunrock/loops/blob/main/plots/) directory along with `performance_evaluation.ipynb` notebook to recreate the plots (labeled figures) found in the paper.
 
 1. In the run script, update the `DATASET_DIR` to point to the path of all the downloaded datasets (set to the path of the directory containing `MM` directory, and inside the `MM` it has subdirectories with `.mtx` files): [scripts/run.sh](https://github.com/gunrock/loops/blob/main/scripts/run.sh). Additionally, you may change the path to `DATASET_FILES_NAME` containing the list of all the datasets (default points to [datasets/suitesparse.txt](https://github.com/gunrock/loops/blob/main/datasets/suitesparse.txt)).

README.md

@@ -0,0 +1,32 @@
+services:
+  loops-dev:
+    build:
+      context: .
+      dockerfile: Dockerfile
+    image: loops-dev:latest
+    container_name: loops-dev
+    network_mode: "host"
+    devices:
+      - /dev/dri
+    group_add:
+      - video
+    cap_add:
+      - SYS_PTRACE
+    security_opt:
+      - seccomp=unconfined
+    shm_size: "16g"
+    ulimits:
+      memlock:
+        soft: -1
+        hard: -1
+      stack:
+        soft: 67108864
+        hard: 67108864
+    volumes:
+      - ${HOME}:${HOME}
+      - ./:/workspace
+    environment:
+      - GIT_SSH_COMMAND=ssh -o StrictHostKeyChecking=no
+    working_dir: /workspace
+    tty: true
+    stdin_open: true

docker-compose.yml

@@ -1,6 +1,4 @@
-# Or directly get from docker:
-# docker pull neoblizz/loops:v0.1
-FROM nvidia/cuda:11.7.0-devel-ubuntu20.04
+FROM nvidia/cuda:13.0.0-devel-ubuntu22.04
 
 # Install dependencies
 RUN apt-get update