Add Comprehensive CI Scripts and Documentation

.ci/README.md

@@ -0,0 +1,122 @@
+# CI Scripts
+
+This directory contains scripts for setting up and running the tritonparse CI environment.
+
+## Scripts Overview
+
+### `setup.sh`
+Sets up the conda environment, installs dependencies, configures CUDA, and installs cuDNN.
+
+**Environment Variables:**
+- `CONDA_ENV`: Conda environment name (default: "tritonparse")
+- `PYTHON_VERSION`: Python version (default: "3.11")
+- `CUDA_VERSION`: CUDA version (default: "12.8")
+- `CUDNN_VERSION`: cuDNN version (default: "9.10.2.21")
+
+**Usage:**
+```bash
+CONDA_ENV=tritonparse PYTHON_VERSION=3.11 bash .ci/setup.sh
+```
+
+> **Note:**
+> `setup.sh` will automatically download and execute the official PyTorch cuDNN installation script from:
+> https://github.com/pytorch/pytorch/blob/main/.ci/docker/common/install_cudnn.sh
+> There is no need to maintain a local cuDNN install script.
+
+### `install-triton.sh`
+Installs Triton from source by cloning the repository and building it.
+
+**Environment Variables:**
+- `CONDA_ENV`: Conda environment name (required)
+
+**Usage:**
+```bash
+CONDA_ENV=tritonparse bash .ci/install-triton.sh
+```
+
+### `install-project.sh`
+Installs the tritonparse project in editable mode with test dependencies.
+
+**Environment Variables:**
+- `CONDA_ENV`: Conda environment name (required)
+
+**Usage:**
+```bash
+CONDA_ENV=tritonparse bash .ci/install-project.sh
+```
+
+### `run-tests.sh`
+Runs the test suite with proper environment setup.
+
+**Environment Variables:**
+- `CONDA_ENV`: Conda environment name (required)
+- `TEST_TYPE`: Type of tests to run (default: "all")
+  - `cpu`: CPU tests only
+  - `cuda`: CUDA tests only
+  - `all`: All tests
+- `VERBOSE`: Enable verbose output (default: "true")
+- `COVERAGE`: Enable coverage reporting (default: "false")
+
+**Usage:**
+```bash
+# Run all tests
+CONDA_ENV=tritonparse bash .ci/run-tests.sh
+
+# Run CPU tests only with coverage
+CONDA_ENV=tritonparse TEST_TYPE=cpu COVERAGE=true bash .ci/run-tests.sh
+
+# Run CUDA tests only
+CONDA_ENV=tritonparse TEST_TYPE=cuda bash .ci/run-tests.sh
+```
+
+## Complete Workflow
+
+For a complete setup and test run:
+
+```bash
+# 1. Setup environment (includes cuDNN installation)
+CONDA_ENV=tritonparse PYTHON_VERSION=3.11 bash .ci/setup.sh
+
+# 2. Install Triton
+CONDA_ENV=tritonparse bash .ci/install-triton.sh
+
+# 3. Install project
+CONDA_ENV=tritonparse bash .ci/install-project.sh
+
+# 4. Run tests
+CONDA_ENV=tritonparse TEST_TYPE=all COVERAGE=true bash .ci/run-tests.sh
+```
+
+## Local Development
+
+For local development, you can use these scripts to set up the same environment as CI:
+
+```bash
+# Setup local environment (includes cuDNN installation)
+CONDA_ENV=tritonparse-local bash .ci/setup.sh
+
+# Install Triton
+CONDA_ENV=tritonparse-local bash .ci/install-triton.sh
+
+# Install project
+CONDA_ENV=tritonparse-local bash .ci/install-project.sh
+
+# Run tests
+CONDA_ENV=tritonparse-local bash .ci/run-tests.sh
+```
+
+## Script Features
+
+- **Error handling**: All scripts use `set -e` to stop on errors
+- **Environment validation**: Scripts check for required environment variables
+- **Verbose output**: Detailed logging for debugging
+- **Modular design**: Each script has a single responsibility
+- **Reusable**: Scripts can be used in different contexts (CI, local development)
+
+## Dependencies
+
+The scripts assume:
+- Linux environment
+- Git available
+- Internet access for downloading packages
+- Sufficient disk space for conda and packages

.ci/install-project.sh

@@ -0,0 +1,33 @@
+#!/bin/bash
+
+# Install tritonparse project dependencies
+# This script installs the project in editable mode with test dependencies
+
+set -e
+
+echo "Installing tritonparse project dependencies..."
+
+# Ensure we're in the conda environment
+if [ -z "$CONDA_ENV" ]; then
+    echo "ERROR: CONDA_ENV is not set"
+    exit 1
+fi
+
+# Activate conda environment
+source /opt/miniconda3/etc/profile.d/conda.sh
+conda activate "$CONDA_ENV"
+
+# Upgrade pip
+echo "Upgrading pip..."
+python -m pip install --upgrade pip
+
+# Install project in editable mode with test dependencies
+echo "Installing tritonparse in editable mode..."
+pip install -e ".[test]"
+
+# Verify installation
+echo "Verifying installation..."
+python -c "import tritonparse; print(f'tritonparse installed successfully')"
+python -c "import coverage; print(f'coverage version: {coverage.__version__}')"
+
+echo "Project installation completed successfully!"

.ci/install-triton.sh

@@ -0,0 +1,169 @@
+#!/bin/bash
+
+# Install Triton from source
+# This script clones and installs Triton from the main repository
+
+set -e
+
+echo "🚀 Installing Triton from source..."
+START_TIME=$(date +%s)
+
+# Function to show elapsed time
+show_elapsed() {
+    CURRENT_TIME=$(date +%s)
+    ELAPSED=$((CURRENT_TIME - START_TIME))
+    echo "⏱️ Elapsed time: ${ELAPSED}s"
+}
+
+# Pre-flight checks
+echo "🔍 Running pre-flight checks..."
+
+# Set Triton version/commit for cache consistency
+TRITON_COMMIT=${TRITON_COMMIT:-"main"}
+echo "🎯 Target Triton commit/branch: $TRITON_COMMIT"
+TRITON_CACHE_DIR="/tmp/triton-cache"
+TRITON_SOURCE_DIR="/tmp/triton"
+
+# Check disk space (need at least 10GB for Triton compilation)
+AVAILABLE_SPACE=$(df /tmp | tail -1 | awk '{print $4}')
+REQUIRED_SPACE=10485760 # 10GB in KB
+if [ "$AVAILABLE_SPACE" -lt "$REQUIRED_SPACE" ]; then
+    echo "⚠️ WARNING: Low disk space. Available: $(($AVAILABLE_SPACE / 1024 / 1024))GB, Recommended: 10GB"
+else
+    echo "✅ Sufficient disk space available: $(($AVAILABLE_SPACE / 1024 / 1024))GB"
+fi
+
+# Ensure we're in the conda environment
+if [ -z "$CONDA_ENV" ]; then
+    echo "ERROR: CONDA_ENV is not set"
+    exit 1
+fi
+
+# Activate conda environment
+source /opt/miniconda3/etc/profile.d/conda.sh
+conda activate "$CONDA_ENV"
+
+# Create cache directory
+mkdir -p "$TRITON_CACHE_DIR"
+
+# Check if we have cached source with correct commit
+if [ -f "$TRITON_CACHE_DIR/commit" ] && [ -d "$TRITON_SOURCE_DIR" ]; then
+    CACHED_COMMIT=$(cat "$TRITON_CACHE_DIR/commit")
+    if [ "$CACHED_COMMIT" = "$TRITON_COMMIT" ] && [ "$TRITON_COMMIT" != "main" ]; then
+        echo "Found cached Triton source with correct commit ($CACHED_COMMIT)"
+        echo "Will use cached source and re-install to new conda environment"
+        USE_CACHED_SOURCE=true
+    elif [ "$TRITON_COMMIT" = "main" ]; then
+        echo "Target is 'main' branch (API fallback), will rebuild from scratch"
+        echo "Cached commit: $CACHED_COMMIT"
+        USE_CACHED_SOURCE=false
+    else
+        echo "Cached source commit mismatch: cached=$CACHED_COMMIT, target=$TRITON_COMMIT"
+        echo "Will rebuild from scratch"
+        USE_CACHED_SOURCE=false
+    fi
+else
+    echo "No cached source found or no commit info, will build from scratch"
+    USE_CACHED_SOURCE=false
+fi
+
+# Update libstdc++ to match system version
+# Otherwise, we get errors like:
+# ImportError: /opt/miniconda3/envs/tritonparse/bin/../lib/libstdc++.so.6:
+# version `GLIBCXX_3.4.30' not found (required by /tmp/triton/python/triton/_C/libtriton.so)
+echo "Updating libstdc++ to match system version..."
+conda install -y -c conda-forge libstdcxx-ng=12.3.0
+# Check if the update was successful
+strings /opt/miniconda3/envs/tritonparse/lib/libstdc++.so.6 | grep GLIBCXX | tail -5
+
+# Uninstall existing pytorch-triton
+echo "Uninstalling existing pytorch-triton..."
+pip uninstall -y pytorch-triton || true
+pip uninstall -y triton || true
+
+# Setup Triton repository based on cache status
+if [ "$USE_CACHED_SOURCE" = "true" ]; then
+    echo "Using cached Triton source..."
+    cd "$TRITON_SOURCE_DIR"
+    ACTUAL_COMMIT=$(git rev-parse HEAD)
+    echo "Using cached Triton commit: $ACTUAL_COMMIT"
+else
+    echo "Setting up Triton repository from scratch..."
+    if [ -d "$TRITON_SOURCE_DIR" ]; then
+        echo "Removing existing source directory..."
+        rm -rf "$TRITON_SOURCE_DIR"
+    fi
+
+    echo "Cloning Triton repository..."
+    if ! git clone https://github.com/triton-lang/triton.git "$TRITON_SOURCE_DIR"; then
+        echo "❌ ERROR: Failed to clone Triton repository"
+        echo "This might be due to network issues or GitHub rate limiting"
+        exit 1
+    fi
+
+    cd "$TRITON_SOURCE_DIR"
+
+    # Checkout specific commit for reproducibility
+    echo "Checking out commit: $TRITON_COMMIT"
+    if ! git checkout "$TRITON_COMMIT"; then
+        echo "❌ ERROR: Failed to checkout commit $TRITON_COMMIT"
+        echo "This might be due to an invalid commit hash or network issues"
+        exit 1
+    fi
+
+    ACTUAL_COMMIT=$(git rev-parse HEAD)
+    echo "✅ Using Triton commit: $ACTUAL_COMMIT"
+fi
+
+# Install build dependencies
+echo "Installing build dependencies..."
+pip install ninja cmake wheel pybind11
+
+# Install Triton requirements
+echo "Installing Triton requirements..."
+pip install -r python/requirements.txt
+
+# Set environment to use clang compiler for faster compilation
+echo "Setting up clang compiler for faster compilation..."
+export CC=clang
+export CXX=clang++
+echo "Using CC: $CC"
+echo "Using CXX: $CXX"
+
+# Install Triton in editable mode with clang
+if [ "$USE_CACHED_SOURCE" = "true" ]; then
+    echo "Installing cached Triton to new conda environment..."
+    echo "This should be fast since build artifacts are cached"
+else
+    echo "Compiling and installing Triton from scratch..."
+    echo "This will take 30-50 minutes for compilation"
+fi
+pip install -e .
+show_elapsed
+
+# Verify Triton installation
+echo "Verifying Triton installation..."
+if python -c "import triton; print(f'Triton version: {triton.__version__}')" 2>/dev/null; then
+    python -c "import triton; print(f'Triton path: {triton.__file__}')"
+    echo "✅ Triton installation verified successfully"
+
+    # Only save commit info after successful verification
+    echo "$ACTUAL_COMMIT" >"$TRITON_CACHE_DIR/commit"
+    echo "✅ Cache information saved"
+
+    show_elapsed
+    echo "🎉 Triton installation completed successfully!"
+else
+    echo "❌ ERROR: Failed to import triton"
+    echo "This might be due to libstdc++ version issues"
+    echo "Checking system libstdc++ version:"
+    strings /usr/lib/x86_64-linux-gnu/libstdc++.so.6 | grep GLIBCXX | tail -5 || echo "Could not check system libstdc++"
+    echo "Checking conda libstdc++ version:"
+    strings /opt/miniconda3/envs/tritonparse/lib/libstdc++.so.6 | grep GLIBCXX | tail -5 || echo "Could not check conda libstdc++"
+
+    # Clean up cache on failure to prevent corruption
+    echo "🧹 Cleaning up cache due to installation failure..."
+    rm -f "$TRITON_CACHE_DIR/commit"
+
+    exit 1
+fi