Merge branch 'main' into main

Author: electron271

.github/scripts/build-cuda.sh

@@ -11,14 +11,20 @@ if [[ -v cuda_targets ]]; then
 elif [ "${build_arch}" = "aarch64" ]; then
     build_capability="75;80;90"
 
-    # CUDA 12.8+: Add sm100/sm120
+    # CUDA 12.8-12.9: Add sm100/sm120
     [[ "${cuda_version}" == 12.8.* || "${cuda_version}" == 12.9.* ]] && build_capability="75;80;90;100;120"
+
+    # CUDA 13.0+: Add sm100/sm110/sm120
+    [[ "${cuda_version}" == 13.*.* ]] && build_capability="75;80;90;100;110;120"
 else
-    # By default, target Maxwell through Hopper.
-    build_capability="50;60;70;75;80;86;89;90"
+    # By default, target Pascal through Hopper.
+    build_capability="60;70;75;80;86;89;90"
 
     # CUDA 12.8+: Add sm100 and sm120; remove < sm70 to align with PyTorch 2.8+cu128 minimum
     [[ "${cuda_version}" == 12.8.* || "${cuda_version}" == 12.9.* ]] && build_capability="70;75;80;86;89;90;100;120"
+
+    # CUDA 13.0+: Remove < sm75 to align with PyTorch 2.9+cu130 minimum
+    [[ "${cuda_version}" == 13.*.* ]] && build_capability="75;80;86;89;90;100;120"
 fi
 
 [[ "${build_os}" = windows-* ]] && python3 -m pip install ninja
@@ -29,8 +35,8 @@ if [ "${build_os:0:6}" == ubuntu ]; then
     echo "Using image $image"
 
     docker run -i -w /src -v "$PWD:/src" "$image" bash -c \
-        "dnf update -y \
-        && dnf install cmake gcc-toolset-11 -y \
+        "dnf -y --refresh update --security \
+        && dnf -y install cmake gcc-toolset-11 --setopt=install_weak_deps=False --setopt=tsflags=nodocs \
         && source scl_source enable gcc-toolset-11 \
         && cmake -DCOMPUTE_BACKEND=cuda -DCOMPUTE_CAPABILITY=\"${build_capability}\" . \
         && cmake --build . --config Release"

.github/scripts/build-xpu.sh

@@ -0,0 +1,22 @@
+#!/bin/bash
+declare build_os
+
+set -xeuo pipefail
+
+# We currently only build XPU on Linux.
+if [ "${build_os:0:6}" == ubuntu ]; then
+    # TODO: We might want to pre-build this as our own customized image in the future.
+    image=intel/deep-learning-essentials:2025.1.3-0-devel-ubuntu22.04
+    echo "Using image $image"
+    docker run --rm -i \
+        -w /src -v "$PWD:/src" "$image" sh -c \
+        "apt-get update \
+      && DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
+        cmake bison intel-fw-gpu intel-ocloc \
+      && cmake -DCOMPUTE_BACKEND=xpu . \
+      && cmake --build . --config Release"
+fi
+
+output_dir="output/${build_os}/x86_64"
+mkdir -p "${output_dir}"
+(shopt -s nullglob && cp bitsandbytes/*.{so,dylib,dll} "${output_dir}")

.github/workflows/python-package.yml

@@ -6,12 +6,13 @@ on:
     branches: [main]
     paths:
       - ".github/workflows/python-package.yml"
+      - ".github/scripts/**"
       - "bitsandbytes/**"
       - "csrc/**"
       - "include/**"
       - "tests/**"
       - "CMakeLists.txt"
-      - "requirements*.txt"
+      - "MANIFEST.in"
       - "setup.py"
       - "pyproject.toml"
   release:
@@ -25,19 +26,19 @@ concurrency:
 
 jobs:
   ##
-  # This job matrix builds the non-CUDA versions of the libraries for all supported platforms.
+  # This job matrix builds the CPU versions of the libraries for all supported platforms.
   ##
-  build-shared-libs:
+  build-cpu:
     strategy:
       matrix:
         include:
           - os: ubuntu-22.04
             arch: x86_64
           - os: ubuntu-22.04-arm
             arch: aarch64
-          - os: windows-latest
+          - os: windows-2025
             arch: x86_64
-          - os: macos-latest
+          - os: macos-15
             arch: arm64
     runs-on: ${{ matrix.os }}
     steps:
@@ -56,37 +57,40 @@ jobs:
           name: shared_library_${{ matrix.os }}_${{ matrix.arch }}
           path: output/*
           retention-days: 7
+
   ##
   # This job matrix builds the CUDA versions of the libraries for platforms that support CUDA (Linux x64/aarch64 + Windows x64)
   ##
-  build-shared-libs-cuda:
+  build-cuda:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-22.04, ubuntu-22.04-arm, windows-latest]
+        os: [ubuntu-22.04, ubuntu-22.04-arm, windows-2025]
         include:
           - os: ubuntu-22.04
             arch: x86_64
           - os: ubuntu-22.04-arm
             arch: aarch64
-          - os: windows-latest
+          - os: windows-2025
             arch: x86_64
         cuda_version:
-          ["11.8.0", "12.0.1", "12.1.1", "12.2.2", "12.3.2", "12.4.1", "12.5.1", "12.6.3", "12.8.1", "12.9.1"]
+          ["11.8.0", "12.0.1", "12.1.1", "12.2.2", "12.3.2", "12.4.1", "12.5.1", "12.6.3", "12.8.1", "12.9.1", "13.0.1"]
     runs-on: ${{ matrix.os }}
     steps:
       - uses: actions/checkout@v4
         # Windows: We install Cuda on the agent (slow)
-      - uses: Jimver/cuda-toolkit@c35baa1a18fd1fc9dcf47c5bd839bf30559c0bc3 # v0.2.24
+      - uses: N-Storm/cuda-toolkit@d68ba29a800229200a2c3f572f9e816d7f67cdb4 # v0.2.24m
         if: startsWith(matrix.os, 'windows')
         id: cuda-toolkit
         with:
-          # Temporary: Use CUDA 12.9.0 for Windows until 12.9.1 is supported with this action.
-          cuda: ${{ matrix.cuda_version == '12.9.1' && '12.9.0' || matrix.cuda_version }}
+          # Temporary: Use CUDA 13.0.0 for Windows until 13.0.1 is supported with this action.
+          cuda: ${{ matrix.cuda_version == '13.0.1' && '13.0.0' || matrix.cuda_version }}
           method: "network"
-          sub-packages: '["nvcc","cudart","cusparse","cublas","thrust","nvrtc_dev","cublas_dev","cusparse_dev"]'
-          linux-local-args: '["--toolkit"]'
+          # The "crt" "nvvm" and "nvptxcompiler" components are added for CUDA 13.
+          sub-packages: ${{ format('["nvcc"{0},"cudart","cusparse","cublas","thrust","cublas_dev","cusparse_dev"]', startsWith(matrix.cuda_version, '13.') && ',"crt","nvvm","nvptxcompiler"' || '') }}
           use-github-cache: false
+          use-local-cache: false
+          log-file-suffix: ${{matrix.os}}-${{matrix.cuda_version}}.txt
       - name: Setup MSVC
         if: startsWith(matrix.os, 'windows')
         uses: ilammy/[email protected] # to use cl
@@ -103,18 +107,34 @@ jobs:
           path: output/*
           retention-days: 7
 
-  build-shared-libs-rocm:
+  build-xpu:
+    strategy:
+      matrix:
+        os: [ubuntu-22.04]
+    runs-on: ${{ matrix.os }}
+    steps:
+      - uses: actions/checkout@v4
+      - name: Build C++
+        run: bash .github/scripts/build-xpu.sh
+        env:
+          build_os: ${{ matrix.os }}
+      - name: Upload build artifact
+        uses: actions/upload-artifact@v4
+        with:
+          name: shared_library_xpu_${{ matrix.os }}_x86_64
+          path: output/*
+          retention-days: 7
+
+  build-rocm:
     strategy:
       matrix:
         os: [ubuntu-22.04]
         arch: [x86_64]
         rocm_version:
-          ["6.1.2", "6.2.4", "6.3.2"]
+          ["6.1.2", "6.2.4", "6.3.4", "6.4.4", "7.0"]
     runs-on: ${{ matrix.os }}
     steps:
       - uses: actions/checkout@v4
-      - name: Set up Docker multiarch
-        uses: docker/setup-qemu-action@v3
       - name: Clean up disk space
         run: |
           sudo rm -rf \
@@ -149,23 +169,24 @@ jobs:
 
   build-wheels:
     needs:
-      - build-shared-libs
-      - build-shared-libs-cuda
-      - build-shared-libs-rocm
+      - build-cpu
+      - build-cuda
+      - build-rocm
+      - build-xpu
     strategy:
       matrix:
-        os: [ubuntu-22.04, ubuntu-22.04-arm, windows-latest, macos-latest]
+        os: [ubuntu-22.04, ubuntu-22.04-arm, windows-2025, macos-15]
         include:
           - os: ubuntu-22.04
             arch: x86_64
           - os: ubuntu-22.04-arm
             arch: aarch64
-          - os: windows-latest
+          - os: windows-2025
             arch: x86_64
-          - os: macos-latest
+          - os: macos-15
             arch: arm64
         # The specific Python version is irrelevant in this context as we are only packaging non-C extension
-        # code. This ensures compatibility across Python versions, including Python 3.9, as compatibility is
+        # code. This ensures compatibility across Python versions, as compatibility is
         # dictated by the packaged code itself, not the Python version used for packaging.
         python-version: ["3.10"]
     runs-on: ${{ matrix.os }}