Stockfish 16

Official release version of Stockfish 16

Bench: 2593605

---

Stockfish 16

A new major release of Stockfish is now available at

https://stockfishchess.org/download/

*Quality of chess play*

Stockfish continues to demonstrate its ability to discover superior moves
with remarkable speed. In self-play against Stockfish 15, this new
release gains up to 50 Elo[1] and wins up to 12 times more game pairs[2]
than it loses. In major chess engine tournaments, Stockfish reliably tops
the rankings[3] winning the TCEC season 24 Superfinal, Swiss, Fischer
Random, and Double Random Chess tournaments and the CCC 19 Bullet,
20 Blitz, and 20 Rapid competitions. Leela Chess Zero[4] was the
challenger in most finals, putting top-engine chess now firmly in the
hands of teams embracing free and open-source software.

*Progress made*

This updated version of Stockfish introduces several enhancements,
including an upgraded neural net architecture (SFNNv6)[5], improved
implementation, and refined parameterization. The ongoing utilization
of Leela’s data combined with a novel inference approach exploiting
sparsity[6], and network compression[7] ensure a speedy evaluation and
modest binary sizes while allowing for more weights and higher accuracy.
The search has undergone more optimization, leading to improved
performance, particularly in longer analyses[8]. Additionally,
the Fishtest framework has been improved and is now able to run the
tests needed to validate new ideas with 10000s of CPU cores.

*Usability improvements*

Stockfish now comes with documentation, found in the wiki folder when
downloading it or on GitHub[9]. Additionally, Stockfish now includes
a clear and consistent forced tablebase win score, displaying a value
of 200 minus the number of plies required to reach a tablebase win[10].
Furthermore, the UCI_Elo option, to reduce its strength, has been
calibrated[11]. It is worth noting that the evaluation system remains
consistent with Stockfish 15.1[12], maintaining the choice that 100cp
means a 50% chance of winning the game against an equal opponent[13].
Finally, binaries of our latest development version are now provided
continuously as pre-releases on GitHub making it easier for
enthusiasts to download the latest and strongest version of
the program[14], we thank Roman Korba for having provided a similar
service for a long time.

*Thank you*

The success of the Stockfish project relies on the vibrant community
of passionate enthusiasts (we appreciate each and every one of you!)
who generously contribute their knowledge, time, and resources.
Together, this dedicated community works towards the common goal of
developing a powerful, freely accessible, and open-source chess engine.
We invite all chess enthusiasts to join the Fishtest testing framework
and contribute to the project[15].

The Stockfish team

[1] https://tests.stockfishchess.org/tests/view/649409f0dc7002ce609c99cc
[2] https://tests.stockfishchess.org/tests/view/649409d7dc7002ce609c99c6
[3] https://en.wikipedia.org/wiki/Stockfish_(chess)#Competition_results
[4] https://lczero.org/
[5] https://github.com/official-stockfish/Stockfish/commit/c1fff71
[6] https://github.com/official-stockfish/Stockfish/commit/38e6166
[7] https://github.com/official-stockfish/Stockfish/commit/a46087e
[8] https://github.com/official-stockfish/Stockfish/commit/472e726
[9] https://github.com/official-stockfish/Stockfish/wiki/
[10] https://github.com/official-stockfish/Stockfish/commit/def2966
[11] https://github.com/official-stockfish/Stockfish/commit/a08b8d4
[12] https://github.com/official-stockfish/Stockfish/commit/52e84e4
[13] https://github.com/official-stockfish/Stockfish/wiki/Stockfish-FAQ#interpretation-of-the-stockfish-evaluation
[14] https://github.com/official-stockfish/Stockfish/releases?q=prerelease%3Atrue
[15] https://stockfishchess.org/get-involved/

.github/workflows/stockfish.yml

@@ -1,6 +1,8 @@
 name: Stockfish
 on:
   push:
+    tags:        
+      - '*' 
     branches:
       - master
       - tools
@@ -10,6 +12,31 @@ on:
       - master
       - tools
 jobs:
+  Prerelease:
+    if: github.ref == 'refs/heads/master'
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v3
+        with:
+          fetch-depth: 0
+
+      # returns null if no pre-release exists
+      - name: Get Commit SHA of Latest Pre-release
+        run: |
+          # Install required packages
+          sudo apt-get update
+          sudo apt-get install -y curl jq
+
+          echo "COMMIT_SHA=$(jq -r 'map(select(.prerelease)) | first | .tag_name' <<< $(curl -s https://api.github.com/repos/${{ github.repository_owner }}/Stockfish/releases))" >> $GITHUB_ENV
+
+      # delete old previous pre-release and tag
+      - uses: dev-drprasad/delete-tag-and-release@v0.2.1
+        if: env.COMMIT_SHA != 'null'
+        with:
+          tag_name: ${{ env.COMMIT_SHA }}
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
   Sanitizers:
     uses: ./.github/workflows/stockfish_sanitizers.yml
   Tests:
@@ -17,5 +44,8 @@ jobs:
   Compiles:
     uses: ./.github/workflows/stockfish_compile_test.yml
   Binaries:
-    if: github.ref == 'refs/heads/master'
+    if: github.ref == 'refs/heads/master' || (startsWith(github.ref_name, 'sf_') && github.ref_type == 'tag')
     uses: ./.github/workflows/stockfish_binaries.yml
+  ARM_Binaries:
+    if: github.ref == 'refs/heads/master' || (startsWith(github.ref_name, 'sf_') && github.ref_type == 'tag')
+    uses: ./.github/workflows/stockfish_arm_binaries.yml

.github/workflows/stockfish_arm_binaries.yml

@@ -0,0 +1,158 @@
+name: Stockfish
+on:
+  workflow_call:
+jobs:
+  Stockfish:
+    name: ${{ matrix.config.name }} ${{ matrix.binaries }}
+    runs-on: ${{ matrix.config.os }}
+    env:
+      COMPILER: ${{ matrix.config.compiler }}
+      COMP: ${{ matrix.config.comp }}
+      EMU: ${{ matrix.config.emu }}
+      EXT: ${{ matrix.config.ext }}
+      OS: ${{ matrix.config.os }}
+      BINARY: ${{ matrix.binaries }}
+    strategy:
+      matrix:
+        config:
+          - name: Android NDK aarch64
+            os: ubuntu-22.04
+            compiler: aarch64-linux-android21-clang++
+            emu: qemu-aarch64
+            comp: ndk
+            shell: bash {0}
+          - name: Android NDK arm
+            os: ubuntu-22.04
+            compiler: armv7a-linux-androideabi21-clang++
+            emu: qemu-arm
+            comp: ndk
+            shell: bash {0}
+        binaries:
+          - armv8
+          - armv7
+          - armv7-neon
+        exclude:
+          - binaries: armv8
+            config: {compiler: armv7a-linux-androideabi21-clang++}
+          - binaries: armv7
+            config: {compiler: aarch64-linux-android21-clang++}
+          - binaries: armv7-neon
+            config: {compiler: aarch64-linux-android21-clang++}
+    defaults:
+      run:
+        working-directory: src
+        shell: ${{ matrix.config.shell }}
+    steps:
+      - uses: actions/checkout@v3
+        with:
+          fetch-depth: 0
+
+      - name: Download required linux packages
+        if: runner.os == 'Linux'
+        run: |
+          sudo apt update
+          sudo apt install qemu-user
+
+      - name: Install NDK
+        if: runner.os == 'Linux'
+        run: |
+          if [ $COMP == ndk ]; then
+            NDKV="21.4.7075529"
+            ANDROID_ROOT=/usr/local/lib/android
+            ANDROID_SDK_ROOT=$ANDROID_ROOT/sdk
+            SDKMANAGER=$ANDROID_SDK_ROOT/cmdline-tools/latest/bin/sdkmanager
+            echo "y" | $SDKMANAGER "ndk;$NDKV"
+            ANDROID_NDK_ROOT=$ANDROID_SDK_ROOT/ndk/$NDKV
+            ANDROID_NDK_BIN=$ANDROID_NDK_ROOT/toolchains/llvm/prebuilt/linux-x86_64/bin
+            echo "ANDROID_NDK_BIN=$ANDROID_NDK_BIN" >> $GITHUB_ENV
+          fi
+
+      - name: Download the used network from the fishtest framework
+        run: make net
+
+      - name: Check compiler
+        run: |
+          if [ $COMP == ndk ]; then
+            export PATH=${{ env.ANDROID_NDK_BIN }}:$PATH
+          fi
+          $COMPILER -v
+
+      - name: Test help target
+        run: make help
+
+      - name: Check git
+        run: git --version
+
+      # Compile profile guided builds
+
+      - name: Compile ${{ matrix.binaries }} build
+        run: |
+          if [ $COMP == ndk ]; then
+            export PATH=${{ env.ANDROID_NDK_BIN }}:$PATH
+            export LDFLAGS="-static -Wno-unused-command-line-argument"
+          fi
+          make clean
+          make -j2 profile-build ARCH=$BINARY COMP=$COMP WINE_PATH=$EMU
+          make strip ARCH=$BINARY COMP=$COMP
+          mv ./stockfish$EXT ../stockfish-android-$BINARY$EXT
+
+      - name: Remove non src files
+        run: rm -f *.o .depend *.nnue
+
+      - name: Download wiki
+        run: |
+          git clone https://github.com/official-stockfish/Stockfish.wiki.git ../wiki
+          cd ../wiki
+          rm -rf .git
+
+      - name: Create tar archive.
+        run: |
+          cd ..
+          mkdir stockfish
+          cp -r wiki stockfish/
+          cp -r src stockfish/
+          cp stockfish-android-$BINARY$EXT stockfish/
+          cp "Top CPU Contributors.txt" stockfish/
+          cp Copying.txt stockfish/
+          cp AUTHORS stockfish/
+          cp CITATION.cff stockfish/
+          cp README.md stockfish/
+          tar -cvf stockfish-android-$BINARY.tar stockfish
+
+      - name: Upload binaries
+        uses: actions/upload-artifact@v3
+        with:
+          name: stockfish-android-${{ matrix.binaries }}
+          path: stockfish-android-${{ matrix.binaries }}.tar
+
+      - name: Release
+        if: startsWith(github.ref_name, 'sf_') && github.ref_type == 'tag'
+        uses: softprops/action-gh-release@v1
+        with:
+          files: stockfish-android-${{ matrix.binaries }}.tar
+
+      - name: Get last commit sha
+        id: last_commit
+        run: echo "COMMIT_SHA=$(git rev-parse HEAD | cut -c 1-8)" >> $GITHUB_ENV
+
+      - name: Get commit date
+        id: commit_date
+        run: echo "COMMIT_DATE=$(git show -s --date=format:'%Y%m%d' --format=%cd HEAD)" >> $GITHUB_ENV
+
+      # Make sure that an old ci which still runs on master doesn't recreate a prerelease
+      - name: Check Pullable Commits
+        id: check_commits
+        run: |
+          git fetch
+          CHANGES=$(git rev-list HEAD..origin/master --count)
+          echo "CHANGES=$CHANGES" >> $GITHUB_ENV
+
+      - name: Prerelease
+        if: github.ref_name == 'master' && env.CHANGES == '0'
+        continue-on-error: true
+        uses: softprops/action-gh-release@v1
+        with:
+          name: Stockfish dev-${{ env.COMMIT_DATE }}-${{ env.COMMIT_SHA }}
+          tag_name: stockfish-dev-${{ env.COMMIT_DATE }}-${{ env.COMMIT_SHA }}
+          prerelease: true
+          files: stockfish-android-${{ matrix.binaries }}.tar

.github/workflows/stockfish_binaries.yml

@@ -9,35 +9,35 @@ jobs:
       COMPILER: ${{ matrix.config.compiler }}
       COMP: ${{ matrix.config.comp }}
       EXT: ${{ matrix.config.ext }}
-      OS: ${{ matrix.config.os }}
+      NAME: ${{ matrix.config.simple_name }}
       BINARY: ${{ matrix.binaries }}
     strategy:
       matrix:
         config:
-          - {
+          - name: Ubuntu 20.04 GCC
-              name: "Ubuntu 20.04 GCC",
+            os: ubuntu-20.04
-              os: ubuntu-20.04,
+            simple_name: ubuntu
-              compiler: g++,
+            compiler: g++
-              comp: gcc,
+            comp: gcc
-              shell: 'bash {0}'
+            shell: bash {0}
-            }
+            archive_ext: tar
-          - {
+          - name: MacOS 12 Apple Clang
-              name: "MacOS 12 Apple Clang",
+            os: macos-12
-              os: macos-12,
+            simple_name: macos
-              compiler: clang++,
+            compiler: clang++
-              comp: clang,
+            comp: clang
-              shell: 'bash {0}'
+            shell: bash {0}
-            }
+            archive_ext: tar
-          - {
+          - name: Windows 2022 Mingw-w64 GCC x86_64
-              name: "Windows 2022 Mingw-w64 GCC x86_64",
+            os: windows-2022
-              os: windows-2022,
+            simple_name: windows
-              compiler: g++,
+            compiler: g++
-              comp: mingw,
+            comp: mingw
-              msys_sys: 'mingw64',
+            msys_sys: mingw64
-              msys_env: 'x86_64-gcc',
+            msys_env: x86_64-gcc
-              shell: 'msys2 {0}',
+            shell: msys2 {0}
-              ext: .exe
+            ext: .exe
-            }
+            archive_ext: zip
         binaries:
           - x86-64
           - x86-64-modern
@@ -56,55 +56,113 @@ jobs:
 
       - name: Download required linux packages
         if: runner.os == 'Linux'
-        run: |
+        run: sudo apt update
-          sudo apt update
 
       - name: Setup msys and install required packages
         if: runner.os == 'Windows'
         uses: msys2/setup-msys2@v2
         with:
-          msystem: ${{matrix.config.msys_sys}}
+          msystem: ${{ matrix.config.msys_sys }}
-          install: mingw-w64-${{matrix.config.msys_env}} make git expect
+          install: mingw-w64-${{ matrix.config.msys_env }} make git zip
 
       - name: Download the used network from the fishtest framework
-        run: |
+        run: make net
-          make net
 
       - name: Check compiler
-        run: |
+        run: $COMPILER -v
-          export PATH=$PATH:$ANDROID_NDK_HOME/toolchains/llvm/prebuilt/linux-x86_64/bin
-          $COMPILER -v
 
       - name: Test help target
-        run: |
+        run: make help
-          make help
+
+      - name: Check git
+        run: git --version
 
       # Compile profile guided builds
 
       - name: Compile ${{ matrix.binaries }} build
         run: |
-          make clean
           make -j2 profile-build ARCH=$BINARY COMP=$COMP
-          strip ./stockfish$EXT
+          make strip ARCH=$BINARY COMP=$COMP
-          mv ./stockfish$EXT ../stockfish-$OS-$BINARY$EXT
+          mv ./stockfish$EXT ../stockfish-$NAME-$BINARY$EXT
 
       - name: Remove non src files
-        run: rm -f *.o .depend *.nnue
+        run: git clean -fx
 
-      - name: Create tar archive.
+      - name: Download wiki
+        run: |
+          git clone https://github.com/official-stockfish/Stockfish.wiki.git ../wiki
+          rm -rf ../wiki/.git
+
+      - name: Create directory.
         run: |
           cd ..
           mkdir stockfish
+          cp -r wiki stockfish/
           cp -r src stockfish/
-          cp stockfish-$OS-$BINARY$EXT stockfish/
+          cp stockfish-$NAME-$BINARY$EXT stockfish/
           cp "Top CPU Contributors.txt" stockfish/
           cp Copying.txt stockfish/
           cp AUTHORS stockfish/
-          tar -cvf stockfish-$OS-$BINARY.tar stockfish
+          cp CITATION.cff stockfish/
+          cp README.md stockfish/
+
+      - name: Create tar
+        if: runner.os != 'Windows'
+        run: |
+          cd ..
+          tar -cvf stockfish-$NAME-$BINARY.tar stockfish
+
+      - name: Create zip
+        if: runner.os == 'Windows'
+        run: |
+          cd ..
+          zip -r stockfish-$NAME-$BINARY.zip stockfish
 
       - name: Upload binaries
+        if: runner.os != 'Windows'
         uses: actions/upload-artifact@v3
         with:
           name: stockfish-${{ matrix.config.os }}-${{ matrix.binaries }}
-          path: |
+          path: stockfish-${{ matrix.config.simple_name }}-${{ matrix.binaries }}.tar
-            stockfish-${{ matrix.config.os }}-${{ matrix.binaries }}.tar
+
+      # Artifacts automatically get zipped
+      # to avoid double zipping, we use the unzipped directory
+      - name: Upload binaries
+        if: runner.os == 'Windows'
+        uses: actions/upload-artifact@v3
+        with:
+          name: stockfish-${{ matrix.config.os }}-${{ matrix.binaries }}
+          path: stockfish
+
+      - name: Release
+        if: startsWith(github.ref_name, 'sf_') && github.ref_type == 'tag'
+        uses: softprops/action-gh-release@v1
+        with:
+          files: stockfish-${{ matrix.config.simple_name }}-${{ matrix.binaries }}.${{ matrix.config.archive_ext }}
+
+      - name: Get last commit sha
+        id: last_commit
+        run: echo "COMMIT_SHA=$(git rev-parse HEAD | cut -c 1-8)" >> $GITHUB_ENV
+
+      - name: Get commit date
+        id: commit_date
+        run: echo "COMMIT_DATE=$(git show -s --date=format:'%Y%m%d' --format=%cd HEAD)" >> $GITHUB_ENV
+
+      # Make sure that an old ci which still runs on master doesn't recreate a prerelease
+      - name: Check Pullable Commits
+        id: check_commits
+        run: |
+          git fetch
+          CHANGES=$(git rev-list HEAD..origin/master --count)
+          echo "CHANGES=$CHANGES" >> $GITHUB_ENV
+
+      - name: Prerelease
+        if: github.ref_name == 'master' && env.CHANGES == '0'
+        continue-on-error: true
+        uses: softprops/action-gh-release@v1
+        with:
+          name: Stockfish dev-${{ env.COMMIT_DATE }}-${{ env.COMMIT_SHA }}
+          tag_name: stockfish-dev-${{ env.COMMIT_DATE }}-${{ env.COMMIT_SHA }}
+          prerelease: true
+          files: stockfish-${{ matrix.config.simple_name }}-${{ matrix.binaries }}.${{ matrix.config.archive_ext }}
+

.github/workflows/stockfish_compile_test.yml

@@ -11,52 +11,40 @@ jobs:
     strategy:
       matrix:
         config:
-          - {
+          - name: Ubuntu 20.04 GCC
-              name: "Ubuntu 20.04 GCC",
+            os: ubuntu-20.04
-              os: ubuntu-20.04,
+            compiler: g++
-              compiler: g++,
+            comp: gcc
-              comp: gcc,
+            shell: bash {0}
-              shell: 'bash {0}'
+          - name: Ubuntu 20.04 Clang
-            }
+            os: ubuntu-20.04
-          - {
+            compiler: clang++
-              name: "Ubuntu 20.04 Clang",
+            comp: clang
-              os: ubuntu-20.04,
+            shell: bash {0}
-              compiler: clang++,
+          - name: MacOS 12 Apple Clang
-              comp: clang,
+            os: macos-12
-              shell: 'bash {0}'
+            compiler: clang++
-            }
+            comp: clang
-          - {
+            shell: bash {0}
-              name: "MacOS 12 Apple Clang",
+          - name: MacOS 12 GCC 11
-              os: macos-12,
+            os: macos-12
-              compiler: clang++,
+            compiler: g++-11
-              comp: clang,
+            comp: gcc
-              shell: 'bash {0}'
+            shell: bash {0}
-            }
+          - name: Windows 2022 Mingw-w64 GCC x86_64
-          - {
+            os: windows-2022
-              name: "MacOS 12 GCC 11",
+            compiler: g++
-              os: macos-12,
+            comp: mingw
-              compiler: g++-11,
+            msys_sys: mingw64
-              comp: gcc,
+            msys_env: x86_64-gcc
-              shell: 'bash {0}'
+            shell: msys2 {0}
-            }
+          - name: Windows 2022 Mingw-w64 Clang x86_64
-          - {
+            os: windows-2022
-              name: "Windows 2022 Mingw-w64 GCC x86_64",
+            compiler: clang++
-              os: windows-2022,
+            comp: clang
-              compiler: g++,
+            msys_sys: clang64
-              comp: mingw,
+            msys_env: clang-x86_64-clang
-              msys_sys: 'mingw64',
+            shell: msys2 {0}
-              msys_env: 'x86_64-gcc',
-              shell: 'msys2 {0}'
-            }
-          - {
-              name: "Windows 2022 Mingw-w64 Clang x86_64",
-              os: windows-2022,
-              compiler: clang++,
-              comp: clang,
-              msys_sys: 'clang64',
-              msys_env: 'clang-x86_64-clang',
-              shell: 'msys2 {0}'
-            }
 
     defaults:
       run:
@@ -72,20 +60,19 @@ jobs:
         uses: msys2/setup-msys2@v2
         with:
           msystem: ${{matrix.config.msys_sys}}
-          install: mingw-w64-${{matrix.config.msys_env}} make git expect
+          install: mingw-w64-${{matrix.config.msys_env}} make git
 
       - name: Download the used network from the fishtest framework
-        run: |
+        run: make net
-          make net
 
       - name: Check compiler
-        run: |
+        run: $COMPILER -v
-          export PATH=$PATH:$ANDROID_NDK_HOME/toolchains/llvm/prebuilt/linux-x86_64/bin
-          $COMPILER -v
 
       - name: Test help target
-        run: |
+        run: make help
-          make help
+
+      - name: Check git
+        run: git --version
 
       # x86-64 with newer extensions tests
 

.github/workflows/stockfish_sanitizers.yml

@@ -12,34 +12,24 @@ jobs:
     strategy:
       matrix:
         config:
-          - {
+          - name: Ubuntu 20.04 GCC
-              name: "Ubuntu 20.04 GCC",
+            os: ubuntu-20.04
-              os: ubuntu-20.04,
+            compiler: g++
-              compiler: g++,
+            comp: gcc
-              comp: gcc,
+            shell: bash {0}
-              shell: 'bash {0}'
-            }
         sanitizers:
-          - {
+          - name: Run with thread sanitizer
-              name: Run with thread sanitizer,
+            make_option: sanitize=thread
-              make_option: sanitize=thread,
+            instrumented_option: sanitizer-thread
-              instrumented_option: sanitizer-thread
+          - name: Run with UB sanitizer
-            }
+            make_option: sanitize=undefined
-          - {
+            instrumented_option: sanitizer-undefined
-              name: Run with UB sanitizer,
+          - name: Run under valgrind
-              make_option: sanitize=undefined,
+            make_option: ""
-              instrumented_option: sanitizer-undefined
+            instrumented_option: valgrind
-            }
+          - name: Run under valgrind-thread
-          - {
+            make_option: ""
-              name: Run under valgrind,
+            instrumented_option: valgrind-thread
-              make_option: "",
-              instrumented_option: valgrind
-            }
-          - {
-              name: Run under valgrind-thread,
-              make_option: "",
-              instrumented_option: valgrind-thread
-            }
     defaults:
       run:
         working-directory: src
@@ -52,20 +42,19 @@ jobs:
       - name: Download required linux packages
         run: |
           sudo apt update
-          sudo apt install expect valgrind g++-multilib qemu-user
+          sudo apt install expect valgrind g++-multilib
 
       - name: Download the used network from the fishtest framework
-        run: |
+        run: make net
-          make net
 
       - name: Check compiler
-        run: |
+        run: $COMPILER -v
-          export PATH=$PATH:$ANDROID_NDK_HOME/toolchains/llvm/prebuilt/linux-x86_64/bin
-          $COMPILER -v
 
       - name: Test help target
-        run: |
+        run: make help
-          make help
+
+      - name: Check git
+        run: git --version
 
       # Sanitizers
 

.github/workflows/stockfish_test.yml

@@ -12,95 +12,68 @@ jobs:
     strategy:
       matrix:
         config:
-          - {
+          - name: Ubuntu 20.04 GCC
-              name: "Ubuntu 20.04 GCC",
+            os: ubuntu-20.04
-              os: ubuntu-20.04,
+            compiler: g++
-              compiler: g++,
+            comp: gcc
-              comp: gcc,
+            run_32bit_tests: true
-              run_32bit_tests: true,
+            run_64bit_tests: true
-              run_64bit_tests: true,
+            shell: bash {0}
-              shell: 'bash {0}'
+          - name: Ubuntu 20.04 Clang
-            }
+            os: ubuntu-20.04
-          - {
+            compiler: clang++
-              name: "Ubuntu 20.04 Clang",
+            comp: clang
-              os: ubuntu-20.04,
+            run_32bit_tests: true
-              compiler: clang++,
+            run_64bit_tests: true
-              comp: clang,
+            shell: bash {0}
-              run_32bit_tests: true,
+          - name: Android NDK aarch64
-              run_64bit_tests: true,
+            os: ubuntu-22.04
-              shell: 'bash {0}'
+            compiler: aarch64-linux-android21-clang++
-            }
+            comp: ndk
-          - {
+            run_armv8_tests: true
-              name: "Ubuntu 20.04 NDK armv8",
+            shell: bash {0}
-              os: ubuntu-20.04,
+          - name: Android NDK arm
-              compiler: aarch64-linux-android21-clang++,
+            os: ubuntu-22.04
-              comp: ndk,
+            compiler: armv7a-linux-androideabi21-clang++
-              run_armv8_tests: false,
+            comp: ndk
-              shell: 'bash {0}'
+            run_armv7_tests: true
-            }
+            shell: bash {0}
-          - {
+          - name: MacOS 12 Apple Clang
-              name: "Ubuntu 20.04 NDK armv7",
+            os: macos-12
-              os: ubuntu-20.04,
+            compiler: clang++
-              compiler: armv7a-linux-androideabi21-clang++,
+            comp: clang
-              comp: ndk,
+            run_64bit_tests: true
-              run_armv7_tests: false,
+            shell: bash {0}
-              shell: 'bash {0}'
+          - name: MacOS 12 GCC 11
-            }
+            os: macos-12
-          - {
+            compiler: g++-11
-              name: "MacOS 12 Apple Clang",
+            comp: gcc
-              os: macos-12,
+            run_64bit_tests: true
-              compiler: clang++,
+            shell: bash {0}
-              comp: clang,
+          - name: Windows 2022 Mingw-w64 GCC x86_64
-              run_64bit_tests: true,
+            os: windows-2022
-              shell: 'bash {0}'
+            compiler: g++
-            }
+            comp: mingw
-          - {
+            run_64bit_tests: true
-              name: "MacOS 12 GCC 11",
+            msys_sys: mingw64
-              os: macos-12,
+            msys_env: x86_64-gcc
-              compiler: g++-11,
+            shell: msys2 {0}
-              comp: gcc,
+          - name: Windows 2022 Mingw-w64 GCC i686
-              run_64bit_tests: true,
+            os: windows-2022
-              shell: 'bash {0}'
+            compiler: g++
-            }
+            comp: mingw
-          - {
+            run_32bit_tests: true
-              name: "Windows 2022 Mingw-w64 GCC x86_64",
+            msys_sys: mingw32
-              os: windows-2022,
+            msys_env: i686-gcc
-              compiler: g++,
+            shell: msys2 {0}
-              comp: mingw,
+          - name: Windows 2022 Mingw-w64 Clang x86_64
-              run_64bit_tests: true,
+            os: windows-2022
-              msys_sys: 'mingw64',
+            compiler: clang++
-              msys_env: 'x86_64-gcc',
+            comp: clang
-              shell: 'msys2 {0}'
+            run_64bit_tests: true
-            }
+            msys_sys: clang64
-          - {
+            msys_env: clang-x86_64-clang
-              name: "Windows 2022 Mingw-w64 GCC i686",
+            shell: msys2 {0}
-              os: windows-2022,
-              compiler: g++,
-              comp: mingw,
-              run_32bit_tests: true,
-              msys_sys: 'mingw32',
-              msys_env: 'i686-gcc',
-              shell: 'msys2 {0}'
-            }
-          - {
-              name: "Windows 2022 Mingw-w64 Clang x86_64",
-              os: windows-2022,
-              compiler: clang++,
-              comp: clang,
-              run_64bit_tests: true,
-              msys_sys: 'clang64',
-              msys_env: 'clang-x86_64-clang',
-              shell: 'msys2 {0}'
-            }
-        exclude:
-          - config: 
-              {
-                name: "Ubuntu 20.04 NDK armv7"
-              }
-          - config:
-              {
-                name: "Ubuntu 20.04 NDK armv8"
-              }
     defaults:
       run:
         working-directory: src
@@ -116,30 +89,47 @@ jobs:
           sudo apt update
           sudo apt install expect valgrind g++-multilib qemu-user
 
+      - name: Install NDK
+        if: runner.os == 'Linux'
+        run: |
+          if [ $COMP == ndk ]; then
+            NDKV="21.4.7075529"
+            ANDROID_ROOT=/usr/local/lib/android
+            ANDROID_SDK_ROOT=$ANDROID_ROOT/sdk
+            SDKMANAGER=$ANDROID_SDK_ROOT/cmdline-tools/latest/bin/sdkmanager
+            echo "y" | $SDKMANAGER "ndk;$NDKV"
+            ANDROID_NDK_ROOT=$ANDROID_SDK_ROOT/ndk/$NDKV
+            ANDROID_NDK_BIN=$ANDROID_NDK_ROOT/toolchains/llvm/prebuilt/linux-x86_64/bin
+            echo "ANDROID_NDK_BIN=$ANDROID_NDK_BIN" >> $GITHUB_ENV
+          fi
+
       - name: Setup msys and install required packages
         if: runner.os == 'Windows'
         uses: msys2/setup-msys2@v2
         with:
-          msystem: ${{matrix.config.msys_sys}}
+          msystem: ${{ matrix.config.msys_sys }}
-          install: mingw-w64-${{matrix.config.msys_env}} make git expect
+          install: mingw-w64-${{ matrix.config.msys_env }} make git expect
 
       - name: Download the used network from the fishtest framework
-        run: |
+        run: make net
-          make net
 
       - name: Extract the bench number from the commit history
         run: |
-          git log HEAD | grep "\b[Bb]ench[ :]\+[0-9]\{7\}" | head -n 1 | sed "s/[^0-9]*\([0-9]*\).*/\1/g" > git_sig
+          git log HEAD | grep -o "\b[Bb]ench[ :]\+[1-9][0-9]\{5,9\}\b" | head -n 1 | sed "s/[^0-9]//g" > git_sig
           [ -s git_sig ] && echo "benchref=$(cat git_sig)" >> $GITHUB_ENV && echo "Reference bench:" $(cat git_sig) || echo "No bench found"
 
       - name: Check compiler
         run: |
-          export PATH=$PATH:$ANDROID_NDK_HOME/toolchains/llvm/prebuilt/linux-x86_64/bin
+          if [ $COMP == ndk ]; then
+            export PATH=${{ env.ANDROID_NDK_BIN }}:$PATH
+          fi
           $COMPILER -v
 
       - name: Test help target
-        run: |
+        run: make help
-          make help
+
+      - name: Check git
+        run: git --version
 
       # x86-32 tests
 
@@ -229,13 +219,7 @@ jobs:
       - name: Test armv8 build
         if: ${{ matrix.config.run_armv8_tests }}
         run: |
-          ANDROID_ROOT=/usr/local/lib/android
+          export PATH=${{ env.ANDROID_NDK_BIN }}:$PATH
-          ANDROID_SDK_ROOT=${ANDROID_ROOT}/sdk
-          SDKMANAGER=${ANDROID_SDK_ROOT}/cmdline-tools/latest/bin/sdkmanager
-          echo "y" | $SDKMANAGER "ndk;21.4.7075529"
-          ANDROID_NDK_ROOT=${ANDROID_SDK_ROOT}/ndk-bundle
-          ln -sfn $ANDROID_SDK_ROOT/ndk/21.4.7075529 $ANDROID_NDK_ROOT
-          export PATH=$ANDROID_NDK_HOME/toolchains/llvm/prebuilt/linux-x86_64/bin:$PATH
           export LDFLAGS="-static -Wno-unused-command-line-argument"
           make clean
           make -j2 ARCH=armv8 build
@@ -246,13 +230,7 @@ jobs:
       - name: Test armv7 build
         if: ${{ matrix.config.run_armv7_tests }}
         run: |
-          ANDROID_ROOT=/usr/local/lib/android
+          export PATH=${{ env.ANDROID_NDK_BIN }}:$PATH
-          ANDROID_SDK_ROOT=${ANDROID_ROOT}/sdk
-          SDKMANAGER=${ANDROID_SDK_ROOT}/cmdline-tools/latest/bin/sdkmanager
-          echo "y" | $SDKMANAGER "ndk;21.4.7075529"
-          ANDROID_NDK_ROOT=${ANDROID_SDK_ROOT}/ndk-bundle
-          ln -sfn $ANDROID_SDK_ROOT/ndk/21.4.7075529 $ANDROID_NDK_ROOT
-          export PATH=$ANDROID_NDK_HOME/toolchains/llvm/prebuilt/linux-x86_64/bin:$PATH
           export LDFLAGS="-static -Wno-unused-command-line-argument"
           make clean
           make -j2 ARCH=armv7 build
@@ -261,13 +239,7 @@ jobs:
       - name: Test armv7-neon build
         if: ${{ matrix.config.run_armv7_tests }}
         run: |
-          ANDROID_ROOT=/usr/local/lib/android
+          export PATH=${{ env.ANDROID_NDK_BIN }}:$PATH
-          ANDROID_SDK_ROOT=${ANDROID_ROOT}/sdk
-          SDKMANAGER=${ANDROID_SDK_ROOT}/cmdline-tools/latest/bin/sdkmanager
-          echo "y" | $SDKMANAGER "ndk;21.4.7075529"
-          ANDROID_NDK_ROOT=${ANDROID_SDK_ROOT}/ndk-bundle
-          ln -sfn $ANDROID_SDK_ROOT/ndk/21.4.7075529 $ANDROID_NDK_ROOT
-          export PATH=$ANDROID_NDK_HOME/toolchains/llvm/prebuilt/linux-x86_64/bin:$PATH
           export LDFLAGS="-static -Wno-unused-command-line-argument"
           make clean
           make -j2 ARCH=armv7-neon build

AUTHORS

@@ -1,17 +1,15 @@
-# List of authors for Stockfish
+# Founders of the Stockfish project and Fishtest infrastructure
-
-# Founders of the Stockfish project and fishtest infrastructure
 Tord Romstad (romstad)
 Marco Costalba (mcostalba)
 Joona Kiiski (zamar)
 Gary Linscott (glinscott)
 
-# Authors and inventors of NNUE, training, NNUE port
+# Authors and inventors of NNUE, training, and NNUE port
 Yu Nasu (ynasu87)
 Motohiro Isozaki (yaneurao)
 Hisayori Noda (nodchip)
 
-# all other authors of the code in alphabetical order
+# All other authors of Stockfish code (in alphabetical order)
 Aditya (absimaldata)
 Adrian Petrescu (apetresc)
 Ajith Chandy Jose (ajithcj)
@@ -21,6 +19,7 @@ Alexander Kure
 Alexander Pagel (Lolligerhans)
 Alfredo Menezes (lonfom169)
 Ali AlZhrani (Cooffe)
+Andreas Matthies (Matthies)
 Andrei Vetrov (proukornew)
 Andrew Grant (AndyGrant)
 Andrey Neporada (nepal)
@@ -35,6 +34,7 @@ Ben Chaney (Chaneybenjamini)
 Ben Koshy (BKSpurgeon)
 Bill Henry (VoyagerOne)
 Bojun Guo (noobpwnftw, Nooby)
+borg323
 Boštjan Mejak (PedanticHacker)
 braich
 Brian Sheppard (SapphireBrand, briansheppard-toast)
@@ -46,12 +46,12 @@ Chess13234
 Chris Cain (ceebo)
 clefrks
 Dale Weiler (graphitemaster)
-Dan Schmidt (dfannius)
 Daniel Axtens (daxtens)
 Daniel Dugovic (ddugovic)
+Dan Schmidt (dfannius)
 Dariusz Orzechowski (dorzechowski)
-David Zar
 David (dav1312)
+David Zar
 Daylen Yang (daylen)
 Deshawn Mohan-Smith (GoldenRare)
 Dieter Dobbelaere (ddobbelaere)
@@ -65,7 +65,6 @@ Eelco de Groot (KingDefender)
 Elvin Liu (solarlight2)
 erbsenzaehler
 Ernesto Gatti
-Linmiao Xu (linrock)
 Fabian Beuke (madnight)
 Fabian Fichter (ianfab)
 Fanael Linithien (Fanael)
@@ -78,32 +77,35 @@ George Sobala (gsobala)
 gguliash
 Giacomo Lorenzetti (G-Lorenz)
 Gian-Carlo Pascutto (gcp)
+Goh CJ (cj5716)
 Gontran Lemaire (gonlem)
 Goodkov Vasiliy Aleksandrovich (goodkov)
 Gregor Cramer
 GuardianRM
-Günther Demetz (pb00067, pb00068)
 Guy Vreuls (gvreuls)
+Günther Demetz (pb00067, pb00068)
 Henri Wiechers
 Hiraoka Takuya (HiraokaTakuya)
 homoSapiensSapiens
 Hongzhi Cheng
 Ivan Ivec (IIvec)
 Jacques B. (Timshel)
+Jake Senne (w1wwwwww)
 Jan Ondruš (hxim)
 Jared Kish (Kurtbusch, kurt22i)
 Jarrod Torriero (DU-jdto)
-Jean Gauthier (OuaisBla)
 Jean-Francois Romang (jromang)
+Jean Gauthier (OuaisBla)
 Jekaa
 Jerry Donald Watson (jerrydonaldwatson)
 jjoshua2
-Jonathan Calovski (Mysseno)
 Jonathan Buladas Dumale (SFisGOD)
+Jonathan Calovski (Mysseno)
+Jonathan McDermid (jonathanmcdermid)
 Joost VandeVondele (vondele)
-Jörg Oster (joergoster)
 Joseph Ellis (jhellis3)
 Joseph R. Prostko
+Jörg Oster (joergoster)
 Julian Willemer (NightlyKing)
 jundery
 Justin Blanchard (UncombedCoconut)
@@ -117,6 +119,7 @@ Krystian Kuzniarek (kuzkry)
 Leonardo Ljubičić (ICCF World Champion)
 Leonid Pechenik (lp--)
 Liam Keegan (lkeegan)
+Linmiao Xu (linrock)
 Linus Arver (listx)
 loco-loco
 Lub van den Berg (ElbertoOne)
@@ -131,6 +134,7 @@ Matt Ginsberg (mattginsberg)
 Matthew Lai (matthewlai)
 Matthew Sullivan (Matt14916)
 Max A. (Disservin)
+Maxim Masiutin (maximmasiutin)
 Maxim Molchanov (Maxim)
 Michael An (man)
 Michael Byrne (MichaelB7)
@@ -145,16 +149,18 @@ Mira
 Miroslav Fontán (Hexik)
 Moez Jellouli (MJZ1977)
 Mohammed Li (tthsqe12)
+Muzhen J (XInTheDark)
 Nathan Rugg (nmrugg)
-Nick Pelling (nickpelling)
+Nguyen Pham (nguyenpham)
 Nicklas Persson (NicklasPersson)
+Nick Pelling (nickpelling)
 Niklas Fiekas (niklasf)
 Nikolay Kostov (NikolayIT)
-Nguyen Pham (nguyenpham)
 Norman Schmidt (FireFather)
 notruck
 Ofek Shochat (OfekShochat, ghostway)
 Ondrej Mosnáček (WOnder93)
+Ondřej Mišina (AndrovT)
 Oskar Werkelin Ahlin
 Pablo Vazquez
 Panthee
@@ -163,7 +169,9 @@ Pasquale Pigazzini (ppigazzini)
 Patrick Jansen (mibere)
 Peter Schneider (pschneider1968)
 Peter Zsifkovits (CoffeeOne)
+PikaCat
 Praveen Kumar Tummala (praveentml)
+Prokop Randáček (ProkopRandacek)
 Rahul Dsilva (silversolver1)
 Ralph Stößer (Ralph Stoesser)
 Raminder Singh
@@ -172,8 +180,8 @@ Reuven Peleg (R-Peleg)
 Richard Lloyd (Richard-Lloyd)
 Rodrigo Exterckötter Tjäder
 Rodrigo Roim (roim)
-Ron Britvich (Britvich)
 Ronald de Man (syzygy1, syzygy)
+Ron Britvich (Britvich)
 rqs
 Rui Coelho (ruicoelhopedro)
 Ryan Schmitt
@@ -184,21 +192,22 @@ Sergei Antonov (saproj)
 Sergei Ivanov (svivanov72)
 Sergio Vieri (sergiovieri)
 sf-x
+Shahin M. Shahin (peregrine)
 Shane Booth (shane31)
 Shawn Varghese (xXH4CKST3RXx)
 Siad Daboul (Topologist)
 Stefan Geschwentner (locutus2)
 Stefano Cardanobile (Stefano80)
+Stefano Di Martino (StefanoD)
 Steinar Gunderson (sesse)
 Stéphane Nicolet (snicolet)
 Syine Mineta (MinetaS)
-Prokop Randáček (ProkopRandacek)
 Thanar2
 thaspel
 theo77186
+Tomasz Sobczyk (Sopel97)
 Tom Truscott
 Tom Vijlbrief (tomtor)
-Tomasz Sobczyk (Sopel97)
 Torsten Franz (torfranz, tfranzer)
 Torsten Hellwig (Torom)
 Tracey Emery (basepr1me)
@@ -206,11 +215,12 @@ tttak
 Unai Corzo (unaiic)
 Uri Blass (uriblass)
 Vince Negri (cuddlestmonkey)
+Viren
+windfishballad
 xefoci7612
 zz4032
 
-
 # Additionally, we acknowledge the authors and maintainers of fishtest,
-# an amazing and essential framework for the development of Stockfish!
+# an amazing and essential framework for Stockfish development!
 #
 # https://github.com/glinscott/fishtest/blob/master/AUTHORS

CITATION.cff

@@ -0,0 +1,23 @@
+# This CITATION.cff file was generated with cffinit.
+# Visit https://bit.ly/cffinit to generate yours today!
+
+cff-version: 1.2.0
+title: Stockfish
+message: >-
+  Please cite this software using the metadata from this
+  file.
+type: software
+authors:
+  - name: The Stockfish developers (see AUTHORS file)
+repository-code: 'https://github.com/official-stockfish/Stockfish'
+url: 'https://stockfishchess.org/'
+repository-artifact: 'https://stockfishchess.org/download/'
+abstract: Stockfish is a free and strong UCI chess engine.
+keywords:
+  - chess
+  - artificial intelligence (AI)
+  - tree search
+  - alpha-beta search
+  - neural networks (NN)
+  - efficiently updatable neural networks (NNUE)
+license: GPL-3.0

README.md

@@ -2,6 +2,21 @@
 
   [![Stockfish][stockfish128-logo]][website-link]
 
+  <h3>Stockfish</h3>
+
+  A free and strong UCI chess engine.
+  <br>
+  <strong>[Explore Stockfish docs »][wiki-link]</strong>
+  <br>
+  <br>
+  [Report bug][issue-link]
+  ·
+  [Open a discussion][discussions-link]
+  ·
+  [Discord][discord-link]
+  ·
+  [Blog][website-blog-link]
+
   [![Build][build-badge]][build-link]
   [![License][license-badge]][license-link]
   <br>
@@ -16,19 +31,16 @@
 
 ## Overview
 
-[Stockfish][website-link] is a free, powerful UCI chess engine derived from
+[Stockfish][website-link] is a **free and strong UCI chess engine** derived from
-Glaurung 2.1. Stockfish is not a complete chess program and requires a UCI-compatible
+Glaurung 2.1 that analyzes chess positions and computes the optimal moves.
-graphical user interface (GUI) (e.g. XBoard with PolyGlot, Scid, Cute Chess, eboard,
+
-Arena, Sigma Chess, Shredder, Chess Partner or Fritz) in order to be used comfortably.
+Stockfish **does not include a graphical user interface** (GUI) that is required
-Read the documentation for your GUI of choice for information about how to use
+to display a chessboard and to make it easy to input moves. These GUIs are
+developed independently from Stockfish and are available online. **Read the
+documentation for your GUI** of choice for information about how to use
 Stockfish with it.
 
-The Stockfish engine features two evaluation functions for chess. The efficiently
+See also the Stockfish [documentation][wiki-usage-link] for further usage help.
-updatable neural network (NNUE) based evaluation is the default and by far the strongest.
-The classical evaluation based on handcrafted terms remains available. The strongest
-network is integrated in the binary and downloaded automatically during the build process.
-The NNUE evaluation benefits from the vector intrinsics available on most CPUs (sse2,
-avx2, neon, or similar).
 
 ## Files
 
@@ -36,340 +48,114 @@ This distribution of Stockfish consists of the following files:
 
   * [README.md][readme-link], the file you are currently reading.
 
-  * [Copying.txt][license-link], a text file containing the GNU General Public License
+  * [Copying.txt][license-link], a text file containing the GNU General Public
-    version 3.
+    License version 3.
 
   * [AUTHORS][authors-link], a text file with the list of authors for the project.
 
-  * [src][src-link], a subdirectory containing the full source code, including a Makefile
+  * [src][src-link], a subdirectory containing the full source code, including a
-    that can be used to compile Stockfish on Unix-like systems.
+    Makefile that can be used to compile Stockfish on Unix-like systems.
 
-  * a file with the .nnue extension, storing the neural network for the NNUE evaluation.
+  * a file with the .nnue extension, storing the neural network for the NNUE
-    Binary distributions will have this file embedded.
+    evaluation. Binary distributions will have this file embedded.
 
-## The UCI protocol and available options
+## The UCI protocol
 
-The Universal Chess Interface (UCI) is a standard protocol used to communicate with
+The [Universal Chess Interface][uci-link] (UCI) is a standard text-based protocol
-a chess engine, and is the recommended way to do so for typical graphical user interfaces
+used to communicate with a chess engine and is the recommended way to do so for
-(GUI) or chess tools. Stockfish implements the majority of its options as described
+typical graphical user interfaces (GUI) or chess tools. Stockfish implements the
-in [the UCI protocol][uci-link].
+majority of its options.
 
-Developers can see the default values for UCI options available in Stockfish by typing
+Developers can see the default values for the UCI options available in Stockfish
-`./stockfish uci` in a terminal, but the majority of users will typically see them and
+by typing `./stockfish uci` in a terminal, but most users should typically use a
-change them via a chess GUI. This is a list of available UCI options in Stockfish:
+chess GUI to interact with Stockfish.
 
-  * #### Threads
+For more information on UCI or debug commands, see our [documentation][wiki-commands-link].
-    The number of CPU threads used for searching a position. For best performance, set
-    this equal to the number of CPU cores available.
 
-  * #### Hash
+## Compiling Stockfish
-    The size of the hash table in MB. It is recommended to set Hash after setting Threads.
 
-  * #### Clear Hash
+Stockfish has support for 32 or 64-bit CPUs, certain hardware instructions,
-    Clear the hash table.
+big-endian machines such as Power PC, and other platforms.
 
-  * #### Ponder
+On Unix-like systems, it should be easy to compile Stockfish directly from the
-    Let Stockfish ponder its next move while the opponent is thinking.
+source code with the included Makefile in the folder `src`. In general, it is
-
+recommended to run `make help` to see a list of make targets with corresponding
-  * #### MultiPV
+descriptions.
-    Output the N best lines (principal variations, PVs) when searching.
-    Leave at 1 for best performance.
-
-  * #### Use NNUE
-    Toggle between the NNUE and classical evaluation functions. If set to "true",
-    the network parameters must be available to load from file (see also EvalFile),
-    if they are not embedded in the binary.
-
-  * #### EvalFile
-    The name of the file of the NNUE evaluation parameters. Depending on the GUI the
-    filename might have to include the full path to the folder/directory that contains
-    the file. Other locations, such as the directory that contains the binary and the
-    working directory, are also searched.
-
-  * #### UCI_AnalyseMode
-    An option handled by your GUI.
-
-  * #### UCI_Chess960
-    An option handled by your GUI. If true, Stockfish will play Chess960.
-
-  * #### UCI_ShowWDL
-    If enabled, show approximate WDL statistics as part of the engine output.
-    These WDL numbers model expected game outcomes for a given evaluation and
-    game ply for engine self-play at fishtest LTC conditions (60+0.6s per game).
-
-  * #### UCI_LimitStrength
-    Enable weaker play aiming for an Elo rating as set by UCI_Elo. This option overrides Skill Level.
-
-  * #### UCI_Elo
-    If enabled by UCI_LimitStrength, aim for an engine strength of the given Elo.
-    This Elo rating has been calibrated at a time control of 60s+0.6s and anchored to CCRL 40/4.
-
-  * #### Skill Level
-    Lower the Skill Level in order to make Stockfish play weaker (see also UCI_LimitStrength).
-    Internally, MultiPV is enabled, and with a certain probability depending on the Skill Level a
-    weaker move will be played.
-
-  * #### SyzygyPath
-    Path to the folders/directories storing the Syzygy tablebase files. Multiple
-    directories are to be separated by ";" on Windows and by ":" on Unix-based
-    operating systems. Do not use spaces around the ";" or ":".
-
-    Example: `C:\tablebases\wdl345;C:\tablebases\wdl6;D:\tablebases\dtz345;D:\tablebases\dtz6`
-
-    It is recommended to store .rtbw files on an SSD. There is no loss in storing
-    the .rtbz files on a regular HDD. It is recommended to verify all md5 checksums
-    of the downloaded tablebase files (`md5sum -c checksum.md5`) as corruption will
-    lead to engine crashes.
-
-  * #### SyzygyProbeDepth
-    Minimum remaining search depth for which a position is probed. Set this option
-    to a higher value to probe less aggressively if you experience too much slowdown
-    (in terms of nps) due to tablebase probing.
-
-  * #### Syzygy50MoveRule
-    Disable to let fifty-move rule draws detected by Syzygy tablebase probes count
-    as wins or losses. This is useful for ICCF correspondence games.
-
-  * #### SyzygyProbeLimit
-    Limit Syzygy tablebase probing to positions with at most this many pieces left
-    (including kings and pawns).
-
-  * #### Move Overhead
-    Assume a time delay of x ms due to network and GUI overheads. This is useful to
-    avoid losses on time in those cases.
-
-  * #### Slow Mover
-    Lower values will make Stockfish take less time in games, higher values will
-    make it think longer.
-
-  * #### nodestime
-    Tells the engine to use nodes searched instead of wall time to account for
-    elapsed time. Useful for engine testing.
-
-  * #### Debug Log File
-    Write all communication to and from the engine into a text file.
-
-For developers the following non-standard commands might be of interest, mainly useful for debugging:
-
-  * #### bench *ttSize threads limit fenFile limitType evalType*
-    Performs a standard benchmark using various options. The signature of a version
-    (standard node count) is obtained using all defaults. `bench` is currently
-    `bench 16 1 13 default depth mixed`.
-
-  * #### compiler
-    Give information about the compiler and environment used for building a binary.
-
-  * #### d
-    Display the current position, with ascii art and fen.
-
-  * #### eval
-    Return the evaluation of the current position.
-
-  * #### export_net [filename]
-    Exports the currently loaded network to a file.
-    If the currently loaded network is the embedded network and the filename
-    is not specified then the network is saved to the file matching the name
-    of the embedded network, as defined in evaluate.h.
-    If the currently loaded network is not the embedded network (some net set
-    through the UCI setoption) then the filename parameter is required and the
-    network is saved into that file.
-
-  * #### flip
-    Flips the side to move.
-
-
-## A note on classical evaluation versus NNUE evaluation
-
-Both approaches assign a value to a position that is used in alpha-beta (PVS) search
-to find the best move. The classical evaluation computes this value as a function
-of various chess concepts, handcrafted by experts, tested and tuned using fishtest.
-The NNUE evaluation computes this value with a neural network based on basic
-inputs (e.g. piece positions only). The network is optimized and trained
-on the evaluations of millions of positions at moderate search depth.
-
-The NNUE evaluation was first introduced in shogi, and ported to Stockfish afterward.
-It can be evaluated efficiently on CPUs, and exploits the fact that only parts
-of the neural network need to be updated after a typical chess move.
-[The nodchip repository][nodchip-link] provided the first version of the needed tools
-to train and develop the NNUE networks. Today, more advanced training tools are
-available in [the nnue-pytorch repository][pytorch-link], while data generation tools
-are available in [a dedicated branch][tools-link].
-
-On CPUs supporting modern vector instructions (avx2 and similar), the NNUE evaluation
-results in much stronger playing strength, even if the nodes per second computed by
-the engine is somewhat lower (roughly 80% of nps is typical).
-
-Notes:
-
-1) the NNUE evaluation depends on the Stockfish binary and the network parameter file
-(see the EvalFile UCI option). Not every parameter file is compatible with a given
-Stockfish binary, but the default value of the EvalFile UCI option is the name of a
-network that is guaranteed to be compatible with that binary.
-
-2) to use the NNUE evaluation, the additional data file with neural network parameters
-needs to be available. Normally, this file is already embedded in the binary or it can
-be downloaded. The filename for the default (recommended) net can be found as the default
-value of the `EvalFile` UCI option, with the format `nn-[SHA256 first 12 digits].nnue`
-(for instance, `nn-c157e0a5755b.nnue`). This file can be downloaded from
-```
-https://tests.stockfishchess.org/api/nn/[filename]
-```
-replacing `[filename]` as needed.
-
-## What to expect from the Syzygy tablebases?
-
-If the engine is searching a position that is not in the tablebases (e.g.
-a position with 8 pieces), it will access the tablebases during the search.
-If the engine reports a very large score (typically 153.xx), this means
-it has found a winning line into a tablebase position.
-
-If the engine is given a position to search that is in the tablebases, it
-will use the tablebases at the beginning of the search to preselect all
-good moves, i.e. all moves that preserve the win or preserve the draw while
-taking into account the 50-move rule.
-It will then perform a search only on those moves. **The engine will not move
-immediately**, unless there is only a single good move. **The engine likely
-will not report a mate score, even if the position is known to be won.**
-
-It is therefore clear that this behaviour is not identical to what one might
-be used to with Nalimov tablebases. There are technical reasons for this
-difference, the main technical reason being that Nalimov tablebases use the
-DTM metric (distance-to-mate), while the Syzygy tablebases use a variation of the
-DTZ metric (distance-to-zero, zero meaning any move that resets the 50-move
-counter). This special metric is one of the reasons that the Syzygy tablebases are
-more compact than Nalimov tablebases, while still storing all information
-needed for optimal play and in addition being able to take into account
-the 50-move rule.
-
-## Large Pages
-
-Stockfish supports large pages on Linux and Windows. Large pages make
-the hash access more efficient, improving the engine speed, especially
-on large hash sizes. Typical increases are 5..10% in terms of nodes per
-second, but speed increases up to 30% have been measured. The support is
-automatic. Stockfish attempts to use large pages when available and
-will fall back to regular memory allocation when this is not the case.
-
-### Support on Linux
-
-Large page support on Linux is obtained by the Linux kernel
-transparent huge pages functionality. Typically, transparent huge pages
-are already enabled, and no configuration is needed.
-
-### Support on Windows
-
-The use of large pages requires "Lock Pages in Memory" privilege. See
-[Enable the Lock Pages in Memory Option (Windows)][lockpages-link]
-on how to enable this privilege, then run [RAMMap][rammap-link]
-to double-check that large pages are used. We suggest that you reboot
-your computer after you have enabled large pages, because long Windows
-sessions suffer from memory fragmentation, which may prevent Stockfish
-from getting large pages: a fresh session is better in this regard.
-
-## Compiling Stockfish yourself from the sources
-
-Stockfish has support for 32 or 64-bit CPUs, certain hardware
-instructions, big-endian machines such as Power PC, and other platforms.
-
-On Unix-like systems, it should be easy to compile Stockfish
-directly from the source code with the included Makefile in the folder
-`src`. In general it is recommended to run `make help` to see a list of make
-targets with corresponding descriptions.
 
 ```
-    cd src
+cd src
-    make help
+make -j build ARCH=x86-64-modern
-    make net
-    make build ARCH=x86-64-modern
 ```
 
-When not using the Makefile to compile (for instance, with Microsoft MSVC) you
+Detailed compilation instructions for all platforms can be found in our
-need to manually set/unset some switches in the compiler command line; see
+[documentation][wiki-compile-link].
-file *types.h* for a quick reference.
 
-When reporting an issue or a bug, please tell us which Stockfish version
+## Contributing
-and which compiler you used to create your executable. This information
-can be found by typing the following command in a console:
-
-```
-    ./stockfish compiler
-```
-
-## Understanding the code base and participating in the project
-
-Stockfish's improvement over the last decade has been a great community
-effort. There are a few ways to help contribute to its growth.
 
 ### Donating hardware
 
-Improving Stockfish requires a massive amount of testing. You can donate
+Improving Stockfish requires a massive amount of testing. You can donate your
-your hardware resources by installing the [Fishtest Worker][worker-link]
+hardware resources by installing the [Fishtest Worker][worker-link] and viewing
-and view the current tests on [Fishtest][fishtest-link].
+the current tests on [Fishtest][fishtest-link].
 
 ### Improving the code
 
-If you want to help improve the code, there are several valuable resources:
+In the [chessprogramming wiki][programming-link], many techniques used in
-
-* [In this wiki,][programming-link] many techniques used in
 Stockfish are explained with a lot of background information.
+The [section on Stockfish][programmingsf-link] describes many features
+and techniques used by Stockfish. However, it is generic rather than
+focused on Stockfish's precise implementation.
 
-* [The section on Stockfish][programmingsf-link]
-describes many features and techniques used by Stockfish. However, it is
-generic rather than being focused on Stockfish's precise implementation.
-Nevertheless, a helpful resource.
-
-* The latest source can always be found on [GitHub][github-link].
-Discussions about Stockfish take place these days mainly in the [FishCooking][fishcooking-link]
-group and on the [Stockfish Discord channel][discord-link].
 The engine testing is done on [Fishtest][fishtest-link].
 If you want to help improve Stockfish, please read this [guideline][guideline-link]
 first, where the basics of Stockfish development are explained.
 
+Discussions about Stockfish take place these days mainly in the Stockfish
+[Discord server][discord-link]. This is also the best place to ask questions
+about the codebase and how to improve it.
 
 ## Terms of use
 
-Stockfish is free, and distributed under the **GNU General Public License version 3**
+Stockfish is free and distributed under the
-(GPL v3). Essentially, this means you are free to do almost exactly
+[**GNU General Public License version 3**][license-link] (GPL v3). Essentially,
-what you want with the program, including distributing it among your
+this means you are free to do almost exactly what you want with the program,
-friends, making it available for download from your website, selling
+including distributing it among your friends, making it available for download
-it (either by itself or as part of some bigger software package), or
+from your website, selling it (either by itself or as part of some bigger
-using it as the starting point for a software project of your own.
+software package), or using it as the starting point for a software project of
+your own.
 
-The only real limitation is that whenever you distribute Stockfish in
+The only real limitation is that whenever you distribute Stockfish in some way,
-some way, you MUST always include the license and the full source code
+you MUST always include the license and the full source code (or a pointer to
-(or a pointer to where the source code can be found) to generate the
+where the source code can be found) to generate the exact binary you are
-exact binary you are distributing. If you make any changes to the
+distributing. If you make any changes to the source code, these changes must
-source code, these changes must also be made available under the GPL v3.
+also be made available under GPL v3.
-
-For full details, read the copy of the GPL v3 found in the file named
-[*Copying.txt*][license-link].
 
 
 [authors-link]:       https://github.com/official-stockfish/Stockfish/blob/master/AUTHORS
 [build-link]:         https://github.com/official-stockfish/Stockfish/actions/workflows/stockfish.yml
 [commits-link]:       https://github.com/official-stockfish/Stockfish/commits/master
 [discord-link]:       https://discord.gg/GWDRS3kU6R
-[fishcooking-link]:   https://groups.google.com/g/fishcooking
+[issue-link]:         https://github.com/official-stockfish/Stockfish/issues/new?assignees=&labels=&template=BUG-REPORT.yml
+[discussions-link]:   https://github.com/official-stockfish/Stockfish/discussions/new
 [fishtest-link]:      https://tests.stockfishchess.org/tests
-[github-link]:        https://github.com/official-stockfish/Stockfish
 [guideline-link]:     https://github.com/glinscott/fishtest/wiki/Creating-my-first-test
 [license-link]:       https://github.com/official-stockfish/Stockfish/blob/master/Copying.txt
-[lockpages-link]:     https://docs.microsoft.com/en-us/sql/database-engine/configure-windows/enable-the-lock-pages-in-memory-option-windows
-[nodchip-link]:       https://github.com/nodchip/Stockfish
 [programming-link]:   https://www.chessprogramming.org/Main_Page
 [programmingsf-link]: https://www.chessprogramming.org/Stockfish
-[pytorch-link]:       https://github.com/glinscott/nnue-pytorch
-[rammap-link]:        https://docs.microsoft.com/en-us/sysinternals/downloads/rammap
 [readme-link]:        https://github.com/official-stockfish/Stockfish/blob/master/README.md
 [release-link]:       https://github.com/official-stockfish/Stockfish/releases/latest
 [src-link]:           https://github.com/official-stockfish/Stockfish/tree/master/src
 [stockfish128-logo]:  https://stockfishchess.org/images/logo/icon_128x128.png
-[tools-link]:         https://github.com/official-stockfish/Stockfish/tree/tools
+[uci-link]:           https://backscattering.de/chess/uci/
-[uci-link]:           https://www.shredderchess.com/download/div/uci.zip
 [website-link]:       https://stockfishchess.org
-[worker-link]:        https://github.com/glinscott/fishtest/wiki/Running-the-worker:-overview
+[website-blog-link]:  https://stockfishchess.org/blog/
+[wiki-link]:          https://github.com/official-stockfish/Stockfish/wiki
+[wiki-usage-link]:    https://github.com/official-stockfish/Stockfish/wiki/Download-and-usage
+[wiki-compile-link]:  https://github.com/official-stockfish/Stockfish/wiki/Compiling-from-source
+[wiki-commands-link]: https://github.com/official-stockfish/Stockfish/wiki/Commands
+[worker-link]:        https://github.com/glinscott/fishtest/wiki/Running-the-worker
 
-[build-badge]:        https://img.shields.io/github/workflow/status/official-stockfish/Stockfish/Stockfish?style=for-the-badge&label=stockfish&logo=github
+[build-badge]:        https://img.shields.io/github/actions/workflow/status/official-stockfish/Stockfish/stockfish.yml?branch=master&style=for-the-badge&label=stockfish&logo=github
 [commits-badge]:      https://img.shields.io/github/commits-since/official-stockfish/Stockfish/latest?style=for-the-badge
 [discord-badge]:      https://img.shields.io/discord/435943710472011776?style=for-the-badge&label=discord&logo=Discord
 [fishtest-badge]:     https://img.shields.io/website?style=for-the-badge&down_color=red&down_message=Offline&label=Fishtest&up_color=success&up_message=Online&url=https%3A%2F%2Ftests.stockfishchess.org%2Ftests%2Ffinished

Top CPU Contributors.txt

@@ -1,160 +1,167 @@
-Contributors to Fishtest with >10,000 CPU hours, as of 2022-11-19.
+Contributors to Fishtest with >10,000 CPU hours, as of 2023-06-20.
 Thank you!
 
 Username                                CPU Hours     Games played
 ------------------------------------------------------------------
-noobpwnftw                               36475307       2748033975
+noobpwnftw                               37457426       2850540907
-technologov                              14570711        760073590
+technologov                              14135647        742892808
+linrock                                   4423514        303254809
 mlang                                     3026000        200065824
-dew                                       1689222        100034318
+dew                                       1689162        100033738
-grandphish2                               1442171         86798057
+okrout                                    1578136        148855886
-okrout                                    1439985        133471766
+pemo                                      1508508         48814305
-pemo                                      1405374         44189811
+grandphish2                               1461406         91540343
-linrock                                   1299003         28382783
+TueRens                                   1194790         70400852
-TueRens                                   1163420         71159522
+JojoM                                      947612         61773190
-JojoM                                      897158         55177114
 tvijlbrief                                 796125         51897690
+sebastronomy                               742434         38218524
 mibere                                     703840         46867607
-gvreuls                                    635982         40652394
+gvreuls                                    651026         42988582
-oz                                         590763         41201352
+oz                                         543438         39314736
-sebastronomy                               581517         23307132
+cw                                         517858         34869755
-cw                                         517915         34865769
+fastgm                                     503862         30260818
-fastgm                                     504266         30264740
+leszek                                     467278         33514883
-CSU_Dynasty                                479901         31846710
+CSU_Dynasty                                464940         31177118
-ctoks                                      433503         28180725
+ctoks                                      434416         28506889
 crunchy                                    427035         27344275
-leszek                                     416883         27493447
+maximmasiutin                              424795         26577722
-bcross                                     409982         28062127
+bcross                                     415722         29060963
-velislav                                   345954         22232274
+olafm                                      395922         32268020
+rpngn                                      348378         24560289
+velislav                                   342567         22138992
 Fisherman                                  327231         21829379
+mgrabiak                                   300612         20608380
 Dantist                                    296386         18031762
-mgrabiak                                   288928         18869896
+nordlandia                                 246201         16189678
-rpngn                                      259965         16281463
+robal                                      241300         15656382
-robal                                      237653         15148350
+marrco                                     234581         17714473
-ncfish1                                    231764         15275003
+ncfish1                                    227517         15233777
-nordlandia                                 226923         14624832
 glinscott                                  208125         13277240
 drabel                                     204167         13930674
 mhoram                                     202894         12601997
 bking_US                                   198894         11876016
-thirdlife                                  198844          5453268
 Thanar                                     179852         12365359
 vdv                                        175544          9904472
-armo9494                                   168201         11136452
 spams                                      157128         10319326
-marrco                                     151599          9551115
 sqrt2                                      147963          9724586
-vdbergh                                    137690          8971569
+DesolatedDodo                              146350          9536172
+Calis007                                   143165          9478764
+vdbergh                                    138650          9064413
 CoffeeOne                                  137100          5024116
+armo9494                                   136191          9460264
 malala                                     136182          8002293
-DesolatedDodo                              135276          8657464
 xoto                                       133759          9159372
 davar                                      129023          8376525
+DMBK                                       122960          8980062
 dsmith                                     122059          7570238
 amicic                                     119661          7938029
 Data                                       113305          8220352
 BrunoBanani                                112960          7436849
 CypressChess                               108331          7759788
-skiminki                                   106518          7062598
+skiminki                                   107583          7218170
+jcAEie                                     105675          8238962
 MaZePallas                                 102823          6633619
 sterni1971                                 100532          5880772
 sunu                                       100167          7040199
 zeryl                                       99331          6221261
+thirdlife                                   99124          2242380
 ElbertoOne                                  99028          7023771
-DMBK                                        97572          6950312
+cuistot                                     98853          6069816
-Calis007                                    96779          5611552
+bigpen0r                                    94809          6529203
-cuistot                                     93111          5536500
 brabos                                      92118          6186135
-Wolfgang                                    91769          5720158
+Wolfgang                                    91939          6105872
 psk                                         89957          5984901
+sschnee                                     88235          5268000
 racerschmacer                               85805          6122790
-jcAEie                                      85527          5630616
+Fifis                                       85722          5709729
+Dubslow                                     84986          6042456
 Vizvezdenec                                 83761          5344740
-sschnee                                     83557          4853690
 0x3C33                                      82614          5271253
 BRAVONE                                     81239          5054681
-Dubslow                                     78461          5042980
 nssy                                        76497          5259388
 jromang                                     76106          5236025
 teddybaer                                   75125          5407666
-yurikvelo                                   73933          5031096
+tolkki963                                   74762          5149662
-tolkki963                                   73885          4721430
+megaman7de                                  74351          4940352
+Wencey                                      74181          4711488
 Pking_cda                                   73776          5293873
-Bobo1239                                    71675          4860987
+yurikvelo                                   73150          5004382
+markkulix                                   72607          5304642
+Bobo1239                                    70579          4794999
 solarlight                                  70517          5028306
 dv8silencer                                 70287          3883992
-Gelma                                       69304          3980932
 manap                                       66273          4121774
-megaman7de                                  65419          4120200
-markkulix                                   65331          4114860
-bigpen0r                                    64932          4683883
 tinker                                      64333          4268790
 qurashee                                    61208          3429862
-AGI                                         58325          4258646
+Mineta                                      59357          4418202
+Spprtr                                      58723          3911011
+AGI                                         58147          4325994
 robnjr                                      57262          4053117
 Freja                                       56938          3733019
 MaxKlaxxMiner                               56879          3423958
+MarcusTullius                               56746          3762951
 ttruscott                                   56010          3680085
 rkl                                         55132          4164467
 renouve                                     53811          3501516
-Spprtr                                      52736          3410019
+javran                                      53785          4627608
 finfish                                     51360          3370515
 eva42                                       51272          3599691
 eastorwest                                  51117          3454811
 rap                                         49985          3219146
-unixwizard                                  49734          2536230
+pb00067                                     49733          3298934
-pb00067                                     49727          3298270
+OuaisBla                                    48626          3445134
 ronaldjerum                                 47654          3240695
 biffhero                                    46564          3111352
-GPUex                                       45861          2926502
-Fifis                                       45843          3088497
-oryx                                        45578          3493978
 VoyagerOne                                  45476          3452465
-Wencey                                      44943          2654490
+jmdana                                      44893          3065205
+maposora                                    44597          4039578
+oryx                                        44570          3454238
 speedycpu                                   43842          3003273
 jbwiebe                                     43305          2805433
+GPUex                                       42378          3133332
 Antihistamine                               41788          2761312
 mhunt                                       41735          2691355
-olafm                                       41277          3284344
 homyur                                      39893          2850481
 gri                                         39871          2515779
-MarcusTullius                               38303          2251097
 Garf                                        37741          2999686
-kdave                                       37424          2557406
 SC                                          37299          2731694
 csnodgrass                                  36207          2688994
-jmdana                                      36157          2210661
 strelock                                    34716          2074055
+szupaw                                      34102          2880346
 EthanOConnor                                33370          2090311
 slakovv                                     32915          2021889
-gopeto                                      31669          2060958
+Gelma                                       31771          1551204
+gopeto                                      31671          2060990
+kdave                                       31157          2198362
 manapbk                                     30987          1810399
 Prcuvu                                      30377          2170122
 anst                                        30301          2190091
 jkiiski                                     30136          1904470
-spcc                                        30135          1903728
+spcc                                        29925          1901692
 hyperbolic.tom                              29840          2017394
-xwziegtm                                    29763          2347412
 chuckstablers                               29659          2093438
 Pyafue                                      29650          1902349
 belzedar94                                  28846          1811530
-OuaisBla                                    27636          1578800
 chriswk                                     26902          1868317
+xwziegtm                                    26897          2124586
 achambord                                   26582          1767323
 Patrick_G                                   26276          1801617
 yorkman                                     26193          1992080
-Ulysses                                     25289          1674274
+Ulysses                                     25288          1689730
 SFTUser                                     25182          1675689
 nabildanial                                 24942          1519409
 Sharaf_DG                                   24765          1786697
+Maxim                                       24705          1502062
 rodneyc                                     24376          1416402
 agg177                                      23890          1395014
-Ente                                        23747          1674582
+Goatminola                                  23763          1956036
-Karpovbot                                   23629          1313186
+Ente                                        23639          1671638
+Jopo12321                                   23467          1483172
 JanErik                                     23408          1703875
 Isidor                                      23388          1680691
 Norabor                                     23371          1603244
-cisco2015                                   22934          1763773
+cisco2015                                   22920          1763301
+jsys14                                      22824          1591906
 Zirie                                       22542          1472937
 team-oh                                     22272          1636708
 Roady                                       22220          1465606
@@ -165,96 +172,96 @@ xor12                                       21628          1680365
 dex                                         21612          1467203
 nesoneg                                     21494          1463031
 user213718                                  21454          1404128
-AndreasKrug                                 21227          1577833
 sphinx                                      21211          1384728
+AndreasKrug                                 21097          1634811
 jjoshua2                                    21001          1423089
+Zake9298                                    20938          1565848
 horst.prack                                 20878          1465656
-jsys14                                      20729          1221010
 0xB00B1ES                                   20590          1208666
 j3corre                                     20405           941444
 Adrian.Schmidt123                           20316          1281436
-bonsi                                       20022          1300682
 wei                                         19973          1745989
-dapper                                      19754          1167758
+notchris                                    19958          1800128
-Zake9298                                    19745          1458416
+Serpensin                                   19840          1697528
+Gaster319                                   19712          1677310
 fishtester                                  19617          1257388
 rstoesser                                   19569          1293588
 eudhan                                      19274          1283717
+votoanthuan                                 19108          1609992
 vulcan                                      18871          1729392
-Jopo12321                                   18803          1036284
+Karpovbot                                   18766          1053178
+qoo_charly_cai                              18543          1284937
 jundery                                     18445          1115855
 ville                                       17883          1384026
-5t0ckf15hTr4in3r                            17809          1105858
 chris                                       17698          1487385
-dju                                         17697           994333
 purplefishies                               17595          1092533
+dju                                         17414           981289
 iisiraider                                  17275          1049015
 DragonLord                                  17014          1162790
-Karby                                       16457          1010138
+redstone59                                  16842          1461780
-Goatminola                                  16278          1145026
+Alb11747                                    16787          1213926
 IgorLeMasson                                16064          1147232
-Gaster319                                   16056          1109070
+Karby                                       15982           979610
-redstone59                                  15953          1161664
 scuzzi                                      15757           968735
 ako027ako                                   15671          1173203
 Nikolay.IT                                  15154          1068349
 Andrew Grant                                15114           895539
 Naven94                                     15054           834762
 OssumOpossum                                14857          1007129
-qoo_charly_cai                              14490           847865
+ZacHFX                                      14783          1021842
 enedene                                     14476           905279
-szupaw                                      14252           929130
 bpfliegel                                   14233           882523
 mpx86                                       14019           759568
 jpulman                                     13982           870599
+Skiff84                                     13826           721996
 crocogoat                                   13803          1117422
 Nesa92                                      13786          1114691
 joster                                      13710           946160
 mbeier                                      13650          1044928
 Hjax                                        13535           915487
+Nullvalue                                   13468          1140498
 Dark_wizzie                                 13422          1007152
 Rudolphous                                  13244           883140
+pirt                                        13100          1009897
 Machariel                                   13010           863104
 infinigon                                   12991           943216
-pirt                                        12925           985437
-Skiff84                                     12923           649994
 mabichito                                   12903           749391
 thijsk                                      12886           722107
 AdrianSA                                    12860           804972
 Flopzee                                     12698           894821
+korposzczur                                 12606           838168
 fatmurphy                                   12547           853210
-woutboat                                    12419           836696
 SapphireBrand                               12416           969604
-Oakwen                                      12406           840961
+Oakwen                                      12399           844109
 deflectooor                                 12386           579392
 modolief                                    12386           896470
 Farseer                                     12249           694108
+Jackfish                                    12213           805008
 pgontarz                                    12151           848794
+dbernier                                    12103           860824
+getraideBFF                                 12072          1024966
 stocky                                      11954           699440
 mschmidt                                    11941           803401
-MooTheCow                                   11871           773654
+MooTheCow                                   11870           773598
-Jackfish                                    11867           773550
+FormazChar                                  11766           885707
-dbernier                                    11705           821780
 whelanh                                     11557           245188
-Maxim                                       11543           836024
+3cho                                        11494          1031076
-Nullvalue                                   11534           731410
-icewulf                                     11528           650470
-FormazChar                                  11523           861599
 infinity                                    11470           727027
 aga                                         11412           695127
 torbjo                                      11395           729145
 Thomas A. Anderson                          11372           732094
 savage84                                    11358           670860
-ali-al-zhrani                               11272           781310
 d64                                         11263           789184
-Bourbaki                                    11108           709144
+ali-al-zhrani                               11245           779246
 snicolet                                    11106           869170
-Alb11747                                    10855           696920
+dapper                                      11032           771402
+ols                                         10947           624903
+Karmatron                                   10828           677458
 basepi                                      10637           744851
 Cubox                                       10621           826448
-Karmatron                                   10616           674818
 michaelrpg                                  10509           739239
 OIVAS7572                                   10420           995586
-Garruk                                      10348           704905
+jojo2357                                    10419           929708
+WoodMan777                                  10380           873720
+Garruk                                      10365           706465
 dzjp                                        10343           732529
-ols                                         10259           570669

src/Makefile

@@ -1,5 +1,5 @@
 # Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-# Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+# Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 #
 # Stockfish is free software: you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by
@@ -69,32 +69,33 @@ VPATH = syzygy:nnue:nnue/features
 ### Section 2. High-level Configuration
 ### ==========================================================================
 #
-# flag                --- Comp switch      --- Description
+# flag                --- Comp switch        --- Description
 # ----------------------------------------------------------------------------
 #
-# debug = yes/no      --- -DNDEBUG         --- Enable/Disable debug mode
+# debug = yes/no      --- -DNDEBUG           --- Enable/Disable debug mode
 # sanitize = none/<sanitizer> ... (-fsanitize )
-#                     --- ( undefined )    --- enable undefined behavior checks
+#                     --- ( undefined )      --- enable undefined behavior checks
-#                     --- ( thread    )    --- enable threading error checks
+#                     --- ( thread    )      --- enable threading error checks
-#                     --- ( address   )    --- enable memory access checks
+#                     --- ( address   )      --- enable memory access checks
-#                     --- ...etc...        --- see compiler documentation for supported sanitizers
+#                     --- ...etc...          --- see compiler documentation for supported sanitizers
-# optimize = yes/no   --- (-O3/-fast etc.) --- Enable/Disable optimizations
+# optimize = yes/no   --- (-O3/-fast etc.)   --- Enable/Disable optimizations
-# arch = (name)       --- (-arch)          --- Target architecture
+# arch = (name)       --- (-arch)            --- Target architecture
-# bits = 64/32        --- -DIS_64BIT       --- 64-/32-bit operating system
+# bits = 64/32        --- -DIS_64BIT         --- 64-/32-bit operating system
-# prefetch = yes/no   --- -DUSE_PREFETCH   --- Use prefetch asm-instruction
+# prefetch = yes/no   --- -DUSE_PREFETCH     --- Use prefetch asm-instruction
-# popcnt = yes/no     --- -DUSE_POPCNT     --- Use popcnt asm-instruction
+# popcnt = yes/no     --- -DUSE_POPCNT       --- Use popcnt asm-instruction
-# pext = yes/no       --- -DUSE_PEXT       --- Use pext x86_64 asm-instruction
+# pext = yes/no       --- -DUSE_PEXT         --- Use pext x86_64 asm-instruction
-# sse = yes/no        --- -msse            --- Use Intel Streaming SIMD Extensions
+# sse = yes/no        --- -msse              --- Use Intel Streaming SIMD Extensions
-# mmx = yes/no        --- -mmmx            --- Use Intel MMX instructions
+# mmx = yes/no        --- -mmmx              --- Use Intel MMX instructions
-# sse2 = yes/no       --- -msse2           --- Use Intel Streaming SIMD Extensions 2
+# sse2 = yes/no       --- -msse2             --- Use Intel Streaming SIMD Extensions 2
-# ssse3 = yes/no      --- -mssse3          --- Use Intel Supplemental Streaming SIMD Extensions 3
+# ssse3 = yes/no      --- -mssse3            --- Use Intel Supplemental Streaming SIMD Extensions 3
-# sse41 = yes/no      --- -msse4.1         --- Use Intel Streaming SIMD Extensions 4.1
+# sse41 = yes/no      --- -msse4.1           --- Use Intel Streaming SIMD Extensions 4.1
-# avx2 = yes/no       --- -mavx2           --- Use Intel Advanced Vector Extensions 2
+# avx2 = yes/no       --- -mavx2             --- Use Intel Advanced Vector Extensions 2
-# avxvnni = yes/no    --- -mavxvnni        --- Use Intel Vector Neural Network Instructions AVX
+# avxvnni = yes/no    --- -mavxvnni          --- Use Intel Vector Neural Network Instructions AVX
-# avx512 = yes/no     --- -mavx512bw       --- Use Intel Advanced Vector Extensions 512
+# avx512 = yes/no     --- -mavx512bw         --- Use Intel Advanced Vector Extensions 512
-# vnni256 = yes/no    --- -mavx512vnni     --- Use Intel Vector Neural Network Instructions 256
+# vnni256 = yes/no    --- -mavx256vnni       --- Use Intel Vector Neural Network Instructions 512 with 256bit operands
-# vnni512 = yes/no    --- -mavx512vnni     --- Use Intel Vector Neural Network Instructions 512
+# vnni512 = yes/no    --- -mavx512vnni       --- Use Intel Vector Neural Network Instructions 512
-# neon = yes/no       --- -DUSE_NEON       --- Use ARM SIMD architecture
+# neon = yes/no       --- -DUSE_NEON         --- Use ARM SIMD architecture
+# dotprod = yes/no    --- -DUSE_NEON_DOTPROD --- Use ARM advanced SIMD Int8 dot product instructions
 #
 # Note that Makefile is space sensitive, so when adding new architectures
 # or modifying existing flags, you have to make sure there are no extra spaces
@@ -116,7 +117,7 @@ ifeq ($(ARCH), $(filter $(ARCH), \
                  x86-64-vnni512 x86-64-vnni256 x86-64-avx512 x86-64-avxvnni x86-64-bmi2 \
                  x86-64-avx2 x86-64-sse41-popcnt x86-64-modern x86-64-ssse3 x86-64-sse3-popcnt \
                  x86-64 x86-32-sse41-popcnt x86-32-sse2 x86-32 ppc-64 ppc-32 e2k \
-                 armv7 armv7-neon armv8 apple-silicon general-64 general-32 riscv64))
+                 armv7 armv7-neon armv8 armv8-dotprod apple-silicon general-64 general-32 riscv64))
    SUPPORTED_ARCH=true
 else
    SUPPORTED_ARCH=false
@@ -140,6 +141,7 @@ avx512 = no
 vnni256 = no
 vnni512 = no
 neon = no
+dotprod = no
 arm_version = 0
 STRIP = strip
 
@@ -308,11 +310,21 @@ ifeq ($(ARCH),armv8)
 	arm_version = 8
 endif
 
+ifeq ($(ARCH),armv8-dotprod)
+	arch = armv8
+	prefetch = yes
+	popcnt = yes
+	neon = yes
+	dotprod = yes
+	arm_version = 8
+endif
+
 ifeq ($(ARCH),apple-silicon)
 	arch = arm64
 	prefetch = yes
 	popcnt = yes
 	neon = yes
+	dotprod = yes
 	arm_version = 8
 endif
 
@@ -366,7 +378,7 @@ endif
 ifeq ($(COMP),gcc)
 	comp=gcc
 	CXX=g++
-	CXXFLAGS += -pedantic -Wextra -Wshadow
+	CXXFLAGS += -pedantic -Wextra -Wshadow -Wmissing-declarations
 
 	ifeq ($(arch),$(filter $(arch),armv7 armv8 riscv64))
 		ifeq ($(OS),Android)
@@ -410,13 +422,14 @@ ifeq ($(COMP),mingw)
 			CXX=i686-w64-mingw32-c++-posix
 		endif
 	endif
-	CXXFLAGS += -pedantic -Wextra -Wshadow
+	CXXFLAGS += -pedantic -Wextra -Wshadow -Wmissing-declarations
 endif
 
-ifeq ($(COMP),icc)
+ifeq ($(COMP),icx)
-	comp=icc
+	comp=icx
-	CXX=icpc
+	CXX=icpx
-	CXXFLAGS += -diag-disable 1476,10120 -Wcheck -Wabi -Wdeprecated -strict-ansi
+	CXXFLAGS += --intel -pedantic -Wextra -Wshadow -Wmissing-prototypes \
+		-Wconditional-uninitialized -Wabi -Wdeprecated
 endif
 
 ifeq ($(COMP),clang)
@@ -426,7 +439,8 @@ ifeq ($(COMP),clang)
 		CXX=x86_64-w64-mingw32-clang++
 	endif
 
-	CXXFLAGS += -pedantic -Wextra -Wshadow
+	CXXFLAGS += -pedantic -Wextra -Wshadow -Wmissing-prototypes \
+	            -Wconditional-uninitialized
 
 	ifeq ($(filter $(KERNEL),Darwin OpenBSD FreeBSD),)
 	ifeq ($(target_windows),)
@@ -486,9 +500,9 @@ ifeq ($(COMP),ndk)
 	LDFLAGS += -static-libstdc++ -pie -lm -latomic
 endif
 
-ifeq ($(comp),icc)
+ifeq ($(comp),icx)
-	profile_make = icc-profile-make
+	profile_make = icx-profile-make
-	profile_use = icc-profile-use
+	profile_use = icx-profile-use
 else ifeq ($(comp),clang)
 	profile_make = clang-profile-make
 	profile_use = clang-profile-use
@@ -512,7 +526,7 @@ endif
 
 ### Sometimes gcc is really clang
 ifeq ($(COMP),gcc)
-	gccversion = $(shell $(CXX) --version)
+	gccversion = $(shell $(CXX) --version 2>/dev/null)
 	gccisclang = $(findstring clang,$(gccversion))
 	ifneq ($(gccisclang),)
 		profile_make = clang-profile-make
@@ -559,7 +573,7 @@ ifeq ($(optimize),yes)
 	endif
 
 	ifeq ($(KERNEL),Darwin)
-		ifeq ($(comp),$(filter $(comp),clang icc))
+		ifeq ($(comp),$(filter $(comp),clang icx))
 			CXXFLAGS += -mdynamic-no-pic
 		endif
 
@@ -571,7 +585,10 @@ ifeq ($(optimize),yes)
 	endif
 
 	ifeq ($(comp),clang)
-		CXXFLAGS += -fexperimental-new-pass-manager
+		clangmajorversion = $(shell $(CXX) -dumpversion 2>/dev/null | cut -f1 -d.)
+		ifeq ($(shell expr $(clangmajorversion) \< 16),1)
+			CXXFLAGS += -fexperimental-new-pass-manager
+		endif
 	endif
 endif
 
@@ -592,8 +609,6 @@ endif
 ifeq ($(popcnt),yes)
 	ifeq ($(arch),$(filter $(arch),ppc64 armv7 armv8 arm64))
 		CXXFLAGS += -DUSE_POPCNT
-	else ifeq ($(comp),icc)
-		CXXFLAGS += -msse3 -DUSE_POPCNT
 	else
 		CXXFLAGS += -msse3 -mpopcnt -DUSE_POPCNT
 	endif
@@ -602,63 +617,63 @@ endif
 ### 3.6 SIMD architectures
 ifeq ($(avx2),yes)
 	CXXFLAGS += -DUSE_AVX2
-	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
 		CXXFLAGS += -mavx2 -mbmi
 	endif
 endif
 
 ifeq ($(avxvnni),yes)
 	CXXFLAGS += -DUSE_VNNI -DUSE_AVXVNNI
-	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
 		CXXFLAGS += -mavxvnni
 	endif
 endif
 
 ifeq ($(avx512),yes)
 	CXXFLAGS += -DUSE_AVX512
-	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
 		CXXFLAGS += -mavx512f -mavx512bw
 	endif
 endif
 
 ifeq ($(vnni256),yes)
 	CXXFLAGS += -DUSE_VNNI
-	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
 		CXXFLAGS += -mavx512f -mavx512bw -mavx512vnni -mavx512dq -mavx512vl -mprefer-vector-width=256
 	endif
 endif
 
 ifeq ($(vnni512),yes)
 	CXXFLAGS += -DUSE_VNNI
-	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
-		CXXFLAGS += -mavx512vnni -mavx512dq -mavx512vl
+		CXXFLAGS += -mavx512f -mavx512bw -mavx512vnni -mavx512dq -mavx512vl -mprefer-vector-width=512
 	endif
 endif
 
 ifeq ($(sse41),yes)
 	CXXFLAGS += -DUSE_SSE41
-	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
 		CXXFLAGS += -msse4.1
 	endif
 endif
 
 ifeq ($(ssse3),yes)
 	CXXFLAGS += -DUSE_SSSE3
-	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
 		CXXFLAGS += -mssse3
 	endif
 endif
 
 ifeq ($(sse2),yes)
 	CXXFLAGS += -DUSE_SSE2
-	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
 		CXXFLAGS += -msse2
 	endif
 endif
 
 ifeq ($(mmx),yes)
 	CXXFLAGS += -DUSE_MMX
-	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
 		CXXFLAGS += -mmmx
 	endif
 endif
@@ -674,24 +689,28 @@ ifeq ($(neon),yes)
 	endif
 endif
 
+ifeq ($(dotprod),yes)
+	CXXFLAGS += -march=armv8.2-a+dotprod -DUSE_NEON_DOTPROD
+endif
+
 ### 3.7 pext
 ifeq ($(pext),yes)
 	CXXFLAGS += -DUSE_PEXT
-	ifeq ($(comp),$(filter $(comp),gcc clang mingw))
+	ifeq ($(comp),$(filter $(comp),gcc clang mingw icx))
 		CXXFLAGS += -mbmi2
 	endif
 endif
 
 ### 3.7.1 Try to include git commit sha for versioning
-GIT_SHA = $(shell git rev-parse --short HEAD 2>/dev/null)
+GIT_SHA = $(shell git rev-parse HEAD 2>/dev/null | cut -c 1-8)
 ifneq ($(GIT_SHA), )
-	CXXFLAGS += -DGIT_SHA=\"$(GIT_SHA)\"
+	CXXFLAGS += -DGIT_SHA=$(GIT_SHA)
 endif
 
 ### 3.7.2 Try to include git commit date for versioning
 GIT_DATE = $(shell git show -s --date=format:'%Y%m%d' --format=%cd HEAD 2>/dev/null)
 ifneq ($(GIT_DATE), )
-	CXXFLAGS += -DGIT_DATE=\"$(GIT_DATE)\"
+	CXXFLAGS += -DGIT_DATE=$(GIT_DATE)
 endif
 
 ### 3.8 Link Time Optimization
@@ -699,8 +718,11 @@ endif
 ### needs access to the optimization flags.
 ifeq ($(optimize),yes)
 ifeq ($(debug), no)
-	ifeq ($(comp),clang)
+	ifeq ($(comp),$(filter $(comp),clang icx))
 		CXXFLAGS += -flto=full
+		ifeq ($(comp),icx)
+			CXXFLAGS += -fwhole-program-vtables
+                endif
 		ifeq ($(target_windows),yes)
 			CXXFLAGS += -fuse-ld=lld
 		endif
@@ -747,19 +769,19 @@ help:
 	@echo "Supported targets:"
 	@echo ""
 	@echo "help                    > Display architecture details"
-	@echo "build                   > Standard build"
+	@echo "profile-build           > standard build with profile-guided optimization"
+	@echo "build                   > skip profile-guided optimization"
 	@echo "net                     > Download the default nnue net"
-	@echo "profile-build           > Faster build (with profile-guided optimization)"
 	@echo "strip                   > Strip executable"
 	@echo "install                 > Install executable"
 	@echo "clean                   > Clean up"
 	@echo ""
 	@echo "Supported archs:"
 	@echo ""
-	@echo "x86-64-vnni512          > x86 64-bit with vnni support 512bit wide"
+	@echo "x86-64-vnni512          > x86 64-bit with vnni 512bit support"
-	@echo "x86-64-vnni256          > x86 64-bit with vnni support 256bit wide"
+	@echo "x86-64-vnni256          > x86 64-bit with vnni 512bit support, limit operands to 256bit wide"
 	@echo "x86-64-avx512           > x86 64-bit with avx512 support"
-	@echo "x86-64-avxvnni          > x86 64-bit with avxvnni support"
+	@echo "x86-64-avxvnni          > x86 64-bit with vnni 256bit support"
 	@echo "x86-64-bmi2             > x86 64-bit with bmi2 support"
 	@echo "x86-64-avx2             > x86 64-bit with avx2 support"
 	@echo "x86-64-sse41-popcnt     > x86 64-bit with sse41 and popcnt support"
@@ -775,6 +797,7 @@ help:
 	@echo "armv7                   > ARMv7 32-bit"
 	@echo "armv7-neon              > ARMv7 32-bit with popcnt and neon"
 	@echo "armv8                   > ARMv8 64-bit with popcnt and neon"
+	@echo "armv8-dotprod           > ARMv8 64-bit with popcnt, neon and dot product support"
 	@echo "e2k                     > Elbrus 2000"
 	@echo "apple-silicon           > Apple silicon ARM64"
 	@echo "general-64              > unspecified 64-bit"
@@ -786,17 +809,18 @@ help:
 	@echo "gcc                     > Gnu compiler (default)"
 	@echo "mingw                   > Gnu compiler with MinGW under Windows"
 	@echo "clang                   > LLVM Clang compiler"
-	@echo "icc                     > Intel compiler"
+	@echo "icx                     > Intel oneAPI DPC++/C++ Compiler"
 	@echo "ndk                     > Google NDK to cross-compile for Android"
 	@echo ""
-	@echo "Simple examples. If you don't know what to do, you likely want to run: "
+	@echo "Simple examples. If you don't know what to do, you likely want to run one of: "
 	@echo ""
-	@echo "make -j build ARCH=x86-64  (A portable, slow compile for 64-bit systems)"
+	@echo "make -j profile-build ARCH=x86-64-avx2    # typically a fast compile for common systems "
-	@echo "make -j build ARCH=x86-32  (A portable, slow compile for 32-bit systems)"
+	@echo "make -j profile-build ARCH=x86-64-modern  # A more portable compile for 64-bit systems "
+	@echo "make -j profile-build ARCH=x86-64         # A portable compile for 64-bit systems "
 	@echo ""
-	@echo "Advanced examples, for experienced users looking for performance: "
+	@echo "Advanced examples, for experienced users: "
 	@echo ""
-	@echo "make    help  ARCH=x86-64-bmi2"
+	@echo "make -j profile-build ARCH=x86-64-bmi2"
 	@echo "make -j profile-build ARCH=x86-64-bmi2 COMP=gcc COMPCXX=g++-9.0"
 	@echo "make -j build ARCH=x86-64-ssse3 COMP=clang"
 	@echo ""
@@ -811,8 +835,10 @@ endif
 
 
 .PHONY: help build profile-build strip install clean net objclean profileclean \
-        config-sanity icc-profile-use icc-profile-make gcc-profile-use gcc-profile-make \
+	config-sanity \
-        clang-profile-use clang-profile-make FORCE
+	icx-profile-use icx-profile-make \
+	gcc-profile-use gcc-profile-make \
+	clang-profile-use clang-profile-make FORCE
 
 build: net config-sanity
 	$(MAKE) ARCH=$(ARCH) COMP=$(COMP) all
@@ -852,7 +878,7 @@ net:
 	$(eval nnuedownloadurl2 := https://github.com/official-stockfish/networks/raw/master/$(nnuenet))
 	$(eval curl_or_wget := $(shell if hash curl 2>/dev/null; then echo "curl -skL"; elif hash wget 2>/dev/null; then echo "wget -qO-"; fi))
 	@if [ "x$(curl_or_wget)" = "x" ]; then \
-	    echo "Automatic download failed: neither curl nor wget is installed. Install one of these tools or download the net manually"; exit 1; \
+	    echo "Neither curl nor wget is installed. Install one of these tools unless the net has been downloaded manually"; \
         fi
 	$(eval shasum_command := $(shell if hash shasum 2>/dev/null; then echo "shasum -a 256 "; elif hash sha256sum 2>/dev/null; then echo "sha256sum "; fi))
 	@if [ "x$(shasum_command)" = "x" ]; then \
@@ -863,7 +889,9 @@ net:
 	      echo "$(nnuenet) available."; \
 	   else \
 	      if [ "x$(curl_or_wget)" != "x" ]; then \
-	         echo "Downloading $${nnuedownloadurl}"; $(curl_or_wget) $${nnuedownloadurl} > $(nnuenet);\
+                 echo "Downloading $${nnuedownloadurl}"; $(curl_or_wget) $${nnuedownloadurl} > $(nnuenet);\
+              else \
+                 echo "No net found and download not possible"; exit 1;\
 	      fi; \
 	   fi; \
 	   if [ "x$(shasum_command)" != "x" ]; then \
@@ -925,7 +953,9 @@ config-sanity: net
 	@echo "vnni256: '$(vnni256)'"
 	@echo "vnni512: '$(vnni512)'"
 	@echo "neon: '$(neon)'"
+	@echo "dotprod: '$(dotprod)'"
 	@echo "arm_version: '$(arm_version)'"
+	@echo "target_windows: '$(target_windows)'"
 	@echo ""
 	@echo "Flags:"
 	@echo "CXX: $(CXX)"
@@ -954,7 +984,7 @@ config-sanity: net
 	@test "$(vnni256)" = "yes" || test "$(vnni256)" = "no"
 	@test "$(vnni512)" = "yes" || test "$(vnni512)" = "no"
 	@test "$(neon)" = "yes" || test "$(neon)" = "no"
-	@test "$(comp)" = "gcc" || test "$(comp)" = "icc" || test "$(comp)" = "mingw" || test "$(comp)" = "clang" \
+	@test "$(comp)" = "gcc" || test "$(comp)" = "icx" || test "$(comp)" = "mingw" || test "$(comp)" = "clang" \
 	|| test "$(comp)" = "armv7a-linux-androideabi16-clang"  || test "$(comp)" = "aarch64-linux-android21-clang"
 
 $(EXE): $(OBJS)
@@ -992,18 +1022,22 @@ gcc-profile-use:
 	EXTRALDFLAGS='-lgcov' \
 	all
 
-icc-profile-make:
+icx-profile-make:
-	@mkdir -p profdir
 	$(MAKE) ARCH=$(ARCH) COMP=$(COMP) \
-	EXTRACXXFLAGS='-prof-gen=srcpos -prof_dir ./profdir' \
+	EXTRACXXFLAGS='-fprofile-instr-generate ' \
+	EXTRALDFLAGS=' -fprofile-instr-generate' \
 	all
 
-icc-profile-use:
+icx-profile-use:
+	$(XCRUN) llvm-profdata merge -output=stockfish.profdata *.profraw
 	$(MAKE) ARCH=$(ARCH) COMP=$(COMP) \
-	EXTRACXXFLAGS='-prof_use -prof_dir ./profdir' \
+	EXTRACXXFLAGS='-fprofile-instr-use=stockfish.profdata' \
+	EXTRALDFLAGS='-fprofile-use ' \
 	all
 
 .depend: $(SRCS)
 	-@$(CXX) $(DEPENDFLAGS) -MM $(SRCS) > $@ 2> /dev/null
 
+ifeq (, $(filter $(MAKECMDGOALS), help strip install clean net objclean profileclean config-sanity))
 -include .depend
+endif

src/benchmark.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -16,6 +16,8 @@
   along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */
 
+#include "benchmark.h"
+
 #include <fstream>
 #include <iostream>
 #include <istream>

src/benchmark.h

@@ -0,0 +1,34 @@
+/*
+  Stockfish, a UCI chess playing engine derived from Glaurung 2.1
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
+
+  Stockfish is free software: you can redistribute it and/or modify
+  it under the terms of the GNU General Public License as published by
+  the Free Software Foundation, either version 3 of the License, or
+  (at your option) any later version.
+
+  Stockfish is distributed in the hope that it will be useful,
+  but WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+  GNU General Public License for more details.
+
+  You should have received a copy of the GNU General Public License
+  along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#ifndef BENCHMARK_H_INCLUDED
+#define BENCHMARK_H_INCLUDED
+
+#include <iosfwd>
+#include <string>
+#include <vector>
+
+namespace Stockfish {
+
+class Position;
+
+std::vector<std::string> setup_bench(const Position&, std::istream&);
+
+} // namespace Stockfish
+
+#endif // #ifndef BENCHMARK_H_INCLUDED

src/bitbase.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/bitboard.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -27,7 +27,6 @@ namespace Stockfish {
 uint8_t PopCnt16[1 << 16];
 uint8_t SquareDistance[SQUARE_NB][SQUARE_NB];
 
-Bitboard SquareBB[SQUARE_NB];
 Bitboard LineBB[SQUARE_NB][SQUARE_NB];
 Bitboard BetweenBB[SQUARE_NB][SQUARE_NB];
 Bitboard PseudoAttacks[PIECE_TYPE_NB][SQUARE_NB];
@@ -82,9 +81,6 @@ void Bitboards::init() {
   for (unsigned i = 0; i < (1 << 16); ++i)
       PopCnt16[i] = uint8_t(std::bitset<16>(i).count());
 
-  for (Square s = SQ_A1; s <= SQ_H8; ++s)
-      SquareBB[s] = (1ULL << s);
-
   for (Square s1 = SQ_A1; s1 <= SQ_H8; ++s1)
       for (Square s2 = SQ_A1; s2 <= SQ_H8; ++s2)
           SquareDistance[s1][s2] = std::max(distance<File>(s1, s2), distance<Rank>(s1, s2));

src/bitboard.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -74,7 +74,6 @@ constexpr Bitboard KingFlank[FILE_NB] = {
 extern uint8_t PopCnt16[1 << 16];
 extern uint8_t SquareDistance[SQUARE_NB][SQUARE_NB];
 
-extern Bitboard SquareBB[SQUARE_NB];
 extern Bitboard BetweenBB[SQUARE_NB][SQUARE_NB];
 extern Bitboard LineBB[SQUARE_NB][SQUARE_NB];
 extern Bitboard PseudoAttacks[PIECE_TYPE_NB][SQUARE_NB];
@@ -108,7 +107,7 @@ extern Magic BishopMagics[SQUARE_NB];
 
 inline Bitboard square_bb(Square s) {
   assert(is_ok(s));
-  return SquareBB[s];
+  return (1ULL << s);
 }
 
 

src/endgame.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/endgame.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/evaluate.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -36,7 +36,7 @@
 #include "timeman.h"
 #include "uci.h"
 #include "incbin/incbin.h"
-
+#include "nnue/evaluate_nnue.h"
 
 // Macro to embed the default efficiently updatable neural network (NNUE) file
 // data in the engine binary (using incbin.h, by Dale Weiler).
@@ -82,20 +82,18 @@ namespace Eval {
         eval_file = EvalFileDefaultName;
 
     #if defined(DEFAULT_NNUE_DIRECTORY)
-    #define stringify2(x) #x
-    #define stringify(x) stringify2(x)
     vector<string> dirs = { "<internal>" , "" , CommandLine::binaryDirectory , stringify(DEFAULT_NNUE_DIRECTORY) };
     #else
     vector<string> dirs = { "<internal>" , "" , CommandLine::binaryDirectory };
     #endif
 
-    for (string directory : dirs)
+    for (const string& directory : dirs)
         if (currentEvalFileName != eval_file)
         {
             if (directory != "<internal>")
             {
                 ifstream stream(directory + eval_file, ios::binary);
-                if (load_eval(eval_file, stream))
+                if (NNUE::load_eval(eval_file, stream))
                     currentEvalFileName = eval_file;
             }
 
@@ -111,7 +109,7 @@ namespace Eval {
                 (void) gEmbeddedNNUEEnd; // Silence warning on unused variable
 
                 istream stream(&buffer);
-                if (load_eval(eval_file, stream))
+                if (NNUE::load_eval(eval_file, stream))
                     currentEvalFileName = eval_file;
             }
         }
@@ -159,24 +157,24 @@ namespace Trace {
 
   Score scores[TERM_NB][COLOR_NB];
 
-  double to_cp(Value v) { return double(v) / UCI::NormalizeToPawnValue; }
+  static double to_cp(Value v) { return double(v) / UCI::NormalizeToPawnValue; }
 
-  void add(int idx, Color c, Score s) {
+  static void add(int idx, Color c, Score s) {
     scores[idx][c] = s;
   }
 
-  void add(int idx, Score w, Score b = SCORE_ZERO) {
+  static void add(int idx, Score w, Score b = SCORE_ZERO) {
     scores[idx][WHITE] = w;
     scores[idx][BLACK] = b;
   }
 
-  std::ostream& operator<<(std::ostream& os, Score s) {
+  static std::ostream& operator<<(std::ostream& os, Score s) {
     os << std::setw(5) << to_cp(mg_value(s)) << " "
        << std::setw(5) << to_cp(eg_value(s));
     return os;
   }
 
-  std::ostream& operator<<(std::ostream& os, Term t) {
+  static std::ostream& operator<<(std::ostream& os, Term t) {
 
     if (t == MATERIAL || t == IMBALANCE || t == WINNABLE || t == TOTAL)
         os << " ----  ----"    << " | " << " ----  ----";
@@ -193,8 +191,8 @@ using namespace Trace;
 namespace {
 
   // Threshold for lazy and space evaluation
-  constexpr Value LazyThreshold1    =  Value(3631);
+  constexpr Value LazyThreshold1    =  Value(3622);
-  constexpr Value LazyThreshold2    =  Value(2084);
+  constexpr Value LazyThreshold2    =  Value(1962);
   constexpr Value SpaceThreshold    =  Value(11551);
 
   // KingAttackWeights[PieceType] contains king attack weights by piece type
@@ -388,10 +386,10 @@ namespace {
   template<Tracing T> template<Color Us, PieceType Pt>
   Score Evaluation<T>::pieces() {
 
-    constexpr Color     Them = ~Us;
+    constexpr Color Them = ~Us;
-    constexpr Direction Down = -pawn_push(Us);
+    [[maybe_unused]] constexpr Direction Down = -pawn_push(Us);
-    constexpr Bitboard OutpostRanks = (Us == WHITE ? Rank4BB | Rank5BB | Rank6BB
+    [[maybe_unused]] constexpr Bitboard OutpostRanks = (Us == WHITE ? Rank4BB | Rank5BB | Rank6BB
-                                                   : Rank5BB | Rank4BB | Rank3BB);
+                                                                    : Rank5BB | Rank4BB | Rank3BB);
     Bitboard b1 = pos.pieces(Us, Pt);
     Bitboard b, bb;
     Score score = SCORE_ZERO;
@@ -430,7 +428,7 @@ namespace {
         int mob = popcount(b & mobilityArea[Us]);
         mobility[Us] += MobilityBonus[Pt - 2][mob];
 
-        if (Pt == BISHOP || Pt == KNIGHT)
+        if constexpr (Pt == BISHOP || Pt == KNIGHT)
         {
             // Bonus if the piece is on an outpost square or can reach one
             // Bonus for knights (UncontestedOutpost) if few relevant targets
@@ -1048,52 +1046,41 @@ make_v:
 /// evaluate() is the evaluator for the outer world. It returns a static
 /// evaluation of the position from the point of view of the side to move.
 
-Value Eval::evaluate(const Position& pos, int* complexity) {
+Value Eval::evaluate(const Position& pos) {
+
+  assert(!pos.checkers());
 
   Value v;
   Value psq = pos.psq_eg_stm();
 
   // We use the much less accurate but faster Classical eval when the NNUE
   // option is set to false. Otherwise we use the NNUE eval unless the
-  // PSQ advantage is decisive and several pieces remain. (~3 Elo)
+  // PSQ advantage is decisive. (~4 Elo at STC, 1 Elo at LTC)
-  bool useClassical = !useNNUE || (pos.count<ALL_PIECES>() > 7 && abs(psq) > 1760);
+  bool useClassical = !useNNUE || abs(psq) > 2048;
 
   if (useClassical)
       v = Evaluation<NO_TRACE>(pos).value();
   else
   {
       int nnueComplexity;
-      int scale = 1064 + 106 * pos.non_pawn_material() / 5120;
+      int npm = pos.non_pawn_material() / 64;
 
       Color stm = pos.side_to_move();
       Value optimism = pos.this_thread()->optimism[stm];
 
       Value nnue = NNUE::evaluate(pos, true, &nnueComplexity);
 
-      // Blend nnue complexity with (semi)classical complexity
+      // Blend optimism with nnue complexity and (semi)classical complexity
-      nnueComplexity = (  416 * nnueComplexity
+      optimism += optimism * (nnueComplexity + abs(psq - nnue)) / 512;
-                        + 424 * abs(psq - nnue)
+      v = (nnue * (945 + npm) + optimism * (150 + npm)) / 1024;
-                        + (optimism  > 0 ? int(optimism) * int(psq - nnue) : 0)
-                        ) / 1024;
-
-      // Return hybrid NNUE complexity to caller
-      if (complexity)
-          *complexity = nnueComplexity;
-
-      optimism = optimism * (269 + nnueComplexity) / 256;
-      v = (nnue * scale + optimism * (scale - 754)) / 1024;
   }
 
   // Damp down the evaluation linearly when shuffling
-  v = v * (195 - pos.rule50_count()) / 211;
+  v = v * (200 - pos.rule50_count()) / 214;
 
   // Guarantee evaluation does not hit the tablebase range
   v = std::clamp(v, VALUE_TB_LOSS_IN_MAX_PLY + 1, VALUE_TB_WIN_IN_MAX_PLY - 1);
 
-  // When not using NNUE, return classical complexity to caller
-  if (complexity && (!useNNUE || useClassical))
-      *complexity = abs(v - psq);
-
   return v;
 }
 

src/evaluate.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -31,7 +31,7 @@ class Position;
 namespace Eval {
 
   std::string trace(Position& pos);
-  Value evaluate(const Position& pos, int* complexity = nullptr);
+  Value evaluate(const Position& pos);
 
   extern bool useNNUE;
   extern std::string currentEvalFileName;
@@ -39,20 +39,13 @@ namespace Eval {
   // The default net name MUST follow the format nn-[SHA256 first 12 digits].nnue
   // for the build process (profile-build and fishtest) to work. Do not change the
   // name of the macro, as it is used in the Makefile.
-  #define EvalFileDefaultName   "nn-ad9b42354671.nnue"
+  #define EvalFileDefaultName   "nn-5af11540bbfe.nnue"
 
   namespace NNUE {
 
-    std::string trace(Position& pos);
-    Value evaluate(const Position& pos, bool adjusted = false, int* complexity = nullptr);
-
     void init();
     void verify();
 
-    bool load_eval(std::string name, std::istream& stream);
-    bool save_eval(std::ostream& stream);
-    bool save_eval(const std::optional<std::string>& filename);
-
   } // namespace NNUE
 
 } // namespace Eval

src/main.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/material.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/material.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -28,7 +28,7 @@ namespace Stockfish::Material {
 
 /// Material::Entry contains various information about a material configuration.
 /// It contains a material imbalance evaluation, a function pointer to a special
-/// endgame evaluation function (which in most cases is NULL, meaning that the
+/// endgame evaluation function (which in most cases is nullptr, meaning that the
 /// standard evaluation function will be used), and scale factors.
 ///
 /// The scale factors are used to scale the evaluation score up or down. For
@@ -62,7 +62,7 @@ struct Entry {
   uint8_t factor[COLOR_NB];
 };
 
-typedef HashTable<Entry, 8192> Table;
+using Table = HashTable<Entry, 8192>;
 
 Entry* probe(const Position& pos);
 

src/misc.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -32,21 +32,26 @@
 // the calls at compile time), try to load them at runtime. To do this we need
 // first to define the corresponding function pointers.
 extern "C" {
-typedef bool(*fun1_t)(LOGICAL_PROCESSOR_RELATIONSHIP,
+using fun1_t = bool(*)(LOGICAL_PROCESSOR_RELATIONSHIP,
-                      PSYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX, PDWORD);
+                       PSYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX, PDWORD);
-typedef bool(*fun2_t)(USHORT, PGROUP_AFFINITY);
+using fun2_t = bool(*)(USHORT, PGROUP_AFFINITY);
-typedef bool(*fun3_t)(HANDLE, CONST GROUP_AFFINITY*, PGROUP_AFFINITY);
+using fun3_t = bool(*)(HANDLE, CONST GROUP_AFFINITY*, PGROUP_AFFINITY);
-typedef bool(*fun4_t)(USHORT, PGROUP_AFFINITY, USHORT, PUSHORT);
+using fun4_t = bool(*)(USHORT, PGROUP_AFFINITY, USHORT, PUSHORT);
-typedef WORD(*fun5_t)();
+using fun5_t = WORD(*)();
+using fun6_t = bool(*)(HANDLE, DWORD, PHANDLE);
+using fun7_t = bool(*)(LPCSTR, LPCSTR, PLUID);
+using fun8_t = bool(*)(HANDLE, BOOL, PTOKEN_PRIVILEGES, DWORD, PTOKEN_PRIVILEGES, PDWORD);
 }
 #endif
 
+#include <cmath>
+#include <cstdlib>
 #include <fstream>
 #include <iomanip>
 #include <iostream>
 #include <sstream>
+#include <string_view>
 #include <vector>
-#include <cstdlib>
 
 #if defined(__linux__) && !defined(__ANDROID__)
 #include <stdlib.h>
@@ -68,7 +73,7 @@ namespace Stockfish {
 namespace {
 
 /// Version number or dev.
-const string version = "15.1";
+constexpr string_view version = "16";
 
 /// Our fancy logging facility. The trick here is to replace cin.rdbuf() and
 /// cout.rdbuf() with two Tie objects that tie cin and cout to a file stream. We
@@ -151,13 +156,13 @@ string engine_info(bool to_uci) {
   stringstream ss;
   ss << "Stockfish " << version << setfill('0');
 
-  if (version == "dev")
+  if constexpr (version == "dev")
   {
       ss << "-";
       #ifdef GIT_DATE
-      ss << GIT_DATE;
+      ss << stringify(GIT_DATE);
       #else
-      const string months("Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec");
+      constexpr string_view months("Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec");
       string month, day, year;
       stringstream date(__DATE__); // From compiler, format is "Sep 21 2008"
 
@@ -168,7 +173,7 @@ string engine_info(bool to_uci) {
       ss << "-";
 
       #ifdef GIT_SHA
-      ss << GIT_SHA;
+      ss << stringify(GIT_SHA);
       #else
       ss << "nogit";
       #endif
@@ -185,8 +190,6 @@ string engine_info(bool to_uci) {
 
 std::string compiler_info() {
 
-  #define stringify2(x) #x
-  #define stringify(x) stringify2(x)
   #define make_version_string(major, minor, patch) stringify(major) "." stringify(minor) "." stringify(patch)
 
 /// Predefined macros hell:
@@ -298,21 +301,94 @@ std::string compiler_info() {
 
 
 /// Debug functions used mainly to collect run-time statistics
-static std::atomic<int64_t> hits[2], means[2];
+constexpr int MaxDebugSlots = 32;
 
-void dbg_hit_on(bool b) { ++hits[0]; if (b) ++hits[1]; }
+namespace {
-void dbg_hit_on(bool c, bool b) { if (c) dbg_hit_on(b); }
+
-void dbg_mean_of(int v) { ++means[0]; means[1] += v; }
+template<size_t N>
+struct DebugInfo {
+    std::atomic<int64_t> data[N] = { 0 };
+
+    constexpr inline std::atomic<int64_t>& operator[](int index) { return data[index]; }
+};
+
+DebugInfo<2> hit[MaxDebugSlots];
+DebugInfo<2> mean[MaxDebugSlots];
+DebugInfo<3> stdev[MaxDebugSlots];
+DebugInfo<6> correl[MaxDebugSlots];
+
+}  // namespace
+
+void dbg_hit_on(bool cond, int slot) {
+
+    ++hit[slot][0];
+    if (cond)
+        ++hit[slot][1];
+}
+
+void dbg_mean_of(int64_t value, int slot) {
+
+    ++mean[slot][0];
+    mean[slot][1] += value;
+}
+
+void dbg_stdev_of(int64_t value, int slot) {
+
+    ++stdev[slot][0];
+    stdev[slot][1] += value;
+    stdev[slot][2] += value * value;
+}
+
+void dbg_correl_of(int64_t value1, int64_t value2, int slot) {
+
+    ++correl[slot][0];
+    correl[slot][1] += value1;
+    correl[slot][2] += value1 * value1;
+    correl[slot][3] += value2;
+    correl[slot][4] += value2 * value2;
+    correl[slot][5] += value1 * value2;
+}
 
 void dbg_print() {
 
-  if (hits[0])
+    int64_t n;
-      cerr << "Total " << hits[0] << " Hits " << hits[1]
+    auto E   = [&n](int64_t x) { return double(x) / n; };
-           << " hit rate (%) " << 100 * hits[1] / hits[0] << endl;
+    auto sqr = [](double x) { return x * x; };
 
-  if (means[0])
+    for (int i = 0; i < MaxDebugSlots; ++i)
-      cerr << "Total " << means[0] << " Mean "
+        if ((n = hit[i][0]))
-           << (double)means[1] / means[0] << endl;
+            std::cerr << "Hit #" << i
+                      << ": Total " << n << " Hits " << hit[i][1]
+                      << " Hit Rate (%) " << 100.0 * E(hit[i][1])
+                      << std::endl;
+
+    for (int i = 0; i < MaxDebugSlots; ++i)
+        if ((n = mean[i][0]))
+        {
+            std::cerr << "Mean #" << i
+                      << ": Total " << n << " Mean " << E(mean[i][1])
+                      << std::endl;
+        }
+
+    for (int i = 0; i < MaxDebugSlots; ++i)
+        if ((n = stdev[i][0]))
+        {
+            double r = sqrtl(E(stdev[i][2]) - sqr(E(stdev[i][1])));
+            std::cerr << "Stdev #" << i
+                      << ": Total " << n << " Stdev " << r
+                      << std::endl;
+        }
+
+    for (int i = 0; i < MaxDebugSlots; ++i)
+        if ((n = correl[i][0]))
+        {
+            double r = (E(correl[i][5]) - E(correl[i][1]) * E(correl[i][3]))
+                       / (  sqrtl(E(correl[i][2]) - sqr(E(correl[i][1])))
+                          * sqrtl(E(correl[i][4]) - sqr(E(correl[i][3]))));
+            std::cerr << "Correl. #" << i
+                      << ": Total " << n << " Coefficient " << r
+                      << std::endl;
+        }
 }
 
 
@@ -373,8 +449,10 @@ void* std_aligned_alloc(size_t alignment, size_t size) {
 #if defined(POSIXALIGNEDALLOC)
   void *mem;
   return posix_memalign(&mem, alignment, size) ? nullptr : mem;
-#elif defined(_WIN32)
+#elif defined(_WIN32) && !defined(_M_ARM) && !defined(_M_ARM64)
   return _mm_malloc(size, alignment);
+#elif defined(_WIN32)
+  return _aligned_malloc(size, alignment);
 #else
   return std::aligned_alloc(alignment, size);
 #endif
@@ -384,8 +462,10 @@ void std_aligned_free(void* ptr) {
 
 #if defined(POSIXALIGNEDALLOC)
   free(ptr);
-#elif defined(_WIN32)
+#elif defined(_WIN32) && !defined(_M_ARM) && !defined(_M_ARM64)
   _mm_free(ptr);
+#elif defined(_WIN32)
+  _aligned_free(ptr);
 #else
   free(ptr);
 #endif
@@ -409,11 +489,30 @@ static void* aligned_large_pages_alloc_windows([[maybe_unused]] size_t allocSize
   if (!largePageSize)
       return nullptr;
 
-  // We need SeLockMemoryPrivilege, so try to enable it for the process
+  // Dynamically link OpenProcessToken, LookupPrivilegeValue and AdjustTokenPrivileges
-  if (!OpenProcessToken(GetCurrentProcess(), TOKEN_ADJUST_PRIVILEGES | TOKEN_QUERY, &hProcessToken))
+
+  HMODULE hAdvapi32 = GetModuleHandle(TEXT("advapi32.dll"));
+
+  if (!hAdvapi32)
+      hAdvapi32 = LoadLibrary(TEXT("advapi32.dll"));
+
+  auto fun6 = (fun6_t)(void(*)())GetProcAddress(hAdvapi32, "OpenProcessToken");
+  if (!fun6)
+      return nullptr;
+  auto fun7 = (fun7_t)(void(*)())GetProcAddress(hAdvapi32, "LookupPrivilegeValueA");
+  if (!fun7)
+      return nullptr;
+  auto fun8 = (fun8_t)(void(*)())GetProcAddress(hAdvapi32, "AdjustTokenPrivileges");
+  if (!fun8)
       return nullptr;
 
-  if (LookupPrivilegeValue(NULL, SE_LOCK_MEMORY_NAME, &luid))
+  // We need SeLockMemoryPrivilege, so try to enable it for the process
+  if (!fun6( // OpenProcessToken()
+      GetCurrentProcess(), TOKEN_ADJUST_PRIVILEGES | TOKEN_QUERY, &hProcessToken))
+          return nullptr;
+
+  if (fun7( // LookupPrivilegeValue(nullptr, SE_LOCK_MEMORY_NAME, &luid)
+      nullptr, "SeLockMemoryPrivilege", &luid))
   {
       TOKEN_PRIVILEGES tp { };
       TOKEN_PRIVILEGES prevTp { };
@@ -425,17 +524,18 @@ static void* aligned_large_pages_alloc_windows([[maybe_unused]] size_t allocSize
 
       // Try to enable SeLockMemoryPrivilege. Note that even if AdjustTokenPrivileges() succeeds,
       // we still need to query GetLastError() to ensure that the privileges were actually obtained.
-      if (AdjustTokenPrivileges(
+      if (fun8( // AdjustTokenPrivileges()
               hProcessToken, FALSE, &tp, sizeof(TOKEN_PRIVILEGES), &prevTp, &prevTpLen) &&
           GetLastError() == ERROR_SUCCESS)
       {
           // Round up size to full pages and allocate
           allocSize = (allocSize + largePageSize - 1) & ~size_t(largePageSize - 1);
           mem = VirtualAlloc(
-              NULL, allocSize, MEM_RESERVE | MEM_COMMIT | MEM_LARGE_PAGES, PAGE_READWRITE);
+              nullptr, allocSize, MEM_RESERVE | MEM_COMMIT | MEM_LARGE_PAGES, PAGE_READWRITE);
 
           // Privilege no longer needed, restore previous state
-          AdjustTokenPrivileges(hProcessToken, FALSE, &prevTp, 0, NULL, NULL);
+          fun8( // AdjustTokenPrivileges ()
+              hProcessToken, FALSE, &prevTp, 0, nullptr, nullptr);
       }
   }
 
@@ -453,7 +553,7 @@ void* aligned_large_pages_alloc(size_t allocSize) {
 
   // Fall back to regular, page aligned, allocation if necessary
   if (!mem)
-      mem = VirtualAlloc(NULL, allocSize, MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE);
+      mem = VirtualAlloc(nullptr, allocSize, MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE);
 
   return mem;
 }
@@ -517,7 +617,7 @@ void bindThisThread(size_t) {}
 /// API and returns the best node id for the thread with index idx. Original
 /// code from Texel by Peter Österlund.
 
-int best_node(size_t idx) {
+static int best_node(size_t idx) {
 
   int threads = 0;
   int nodes = 0;
@@ -526,7 +626,7 @@ int best_node(size_t idx) {
   DWORD byteOffset = 0;
 
   // Early exit if the needed API is not available at runtime
-  HMODULE k32 = GetModuleHandle("Kernel32.dll");
+  HMODULE k32 = GetModuleHandle(TEXT("Kernel32.dll"));
   auto fun1 = (fun1_t)(void(*)())GetProcAddress(k32, "GetLogicalProcessorInformationEx");
   if (!fun1)
       return -1;
@@ -596,7 +696,7 @@ void bindThisThread(size_t idx) {
       return;
 
   // Early exit if the needed API are not available at runtime
-  HMODULE k32 = GetModuleHandle("Kernel32.dll");
+  HMODULE k32 = GetModuleHandle(TEXT("Kernel32.dll"));
   auto fun2 = (fun2_t)(void(*)())GetProcAddress(k32, "GetNumaNodeProcessorMaskEx");
   auto fun3 = (fun3_t)(void(*)())GetProcAddress(k32, "SetThreadGroupAffinity");
   auto fun4 = (fun4_t)(void(*)())GetProcAddress(k32, "GetNumaNodeProcessorMask2");

src/misc.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -28,6 +28,9 @@
 
 #include "types.h"
 
+#define stringify2(x) #x
+#define stringify(x) stringify2(x)
+
 namespace Stockfish {
 
 std::string engine_info(bool to_uci = false);
@@ -39,12 +42,13 @@ void std_aligned_free(void* ptr);
 void* aligned_large_pages_alloc(size_t size); // memory aligned by page size, min alignment: 4096 bytes
 void aligned_large_pages_free(void* mem); // nop if mem == nullptr
 
-void dbg_hit_on(bool b);
+void dbg_hit_on(bool cond, int slot = 0);
-void dbg_hit_on(bool c, bool b);
+void dbg_mean_of(int64_t value, int slot = 0);
-void dbg_mean_of(int v);
+void dbg_stdev_of(int64_t value, int slot = 0);
+void dbg_correl_of(int64_t value1, int64_t value2, int slot = 0);
 void dbg_print();
 
-typedef std::chrono::milliseconds::rep TimePoint; // A value in milliseconds
+using TimePoint = std::chrono::milliseconds::rep; // A value in milliseconds
 static_assert(sizeof(TimePoint) == sizeof(int64_t), "TimePoint should be 64 bits");
 inline TimePoint now() {
   return std::chrono::duration_cast<std::chrono::milliseconds>
@@ -85,32 +89,6 @@ static inline const union { uint32_t i; char c[4]; } Le = { 0x01020304 };
 static inline const bool IsLittleEndian = (Le.c[0] == 4);
 
 
-// RunningAverage : a class to calculate a running average of a series of values.
-// For efficiency, all computations are done with integers.
-class RunningAverage {
-  public:
-
-      // Reset the running average to rational value p / q
-      void set(int64_t p, int64_t q)
-        { average = p * PERIOD * RESOLUTION / q; }
-
-      // Update average with value v
-      void update(int64_t v)
-        { average = RESOLUTION * v + (PERIOD - 1) * average / PERIOD; }
-
-      // Test if average is strictly greater than rational a / b
-      bool is_greater(int64_t a, int64_t b) const
-        { return b * average > a * (PERIOD * RESOLUTION); }
-
-      int64_t value() const
-        { return average / (PERIOD * RESOLUTION); }
-
-  private :
-      static constexpr int64_t PERIOD     = 4096;
-      static constexpr int64_t RESOLUTION = 1024;
-      int64_t average;
-};
-
 template <typename T, std::size_t MaxSize>
 class ValueList {
 
@@ -164,7 +142,7 @@ public:
 
 inline uint64_t mul_hi64(uint64_t a, uint64_t b) {
 #if defined(__GNUC__) && defined(IS_64BIT)
-    __extension__ typedef unsigned __int128 uint128;
+    __extension__ using uint128 = unsigned __int128;
     return ((uint128)a * (uint128)b) >> 64;
 #else
     uint64_t aL = (uint32_t)a, aH = a >> 32;

src/movegen.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -25,13 +25,21 @@ namespace Stockfish {
 
 namespace {
 
-  template<GenType Type, Direction D>
+  template<GenType Type, Direction D, bool Enemy>
-  ExtMove* make_promotions(ExtMove* moveList, Square to) {
+  ExtMove* make_promotions(ExtMove* moveList, [[maybe_unused]] Square to) {
 
-    if (Type == CAPTURES || Type == EVASIONS || Type == NON_EVASIONS)
+    if constexpr (Type == CAPTURES || Type == EVASIONS || Type == NON_EVASIONS)
+    {
         *moveList++ = make<PROMOTION>(to - D, to, QUEEN);
+        if constexpr (Enemy && Type == CAPTURES)
+        {
+            *moveList++ = make<PROMOTION>(to - D, to, ROOK);
+            *moveList++ = make<PROMOTION>(to - D, to, BISHOP);
+            *moveList++ = make<PROMOTION>(to - D, to, KNIGHT);
+        }
+    }
 
-    if (Type == QUIETS || Type == EVASIONS || Type == NON_EVASIONS)
+    if constexpr ((Type == QUIETS && !Enemy) || Type == EVASIONS || Type == NON_EVASIONS)
     {
         *moveList++ = make<PROMOTION>(to - D, to, ROOK);
         *moveList++ = make<PROMOTION>(to - D, to, BISHOP);
@@ -60,18 +68,18 @@ namespace {
     Bitboard pawnsNotOn7 = pos.pieces(Us, PAWN) & ~TRank7BB;
 
     // Single and double pawn pushes, no promotions
-    if (Type != CAPTURES)
+    if constexpr (Type != CAPTURES)
     {
         Bitboard b1 = shift<Up>(pawnsNotOn7)   & emptySquares;
         Bitboard b2 = shift<Up>(b1 & TRank3BB) & emptySquares;
 
-        if (Type == EVASIONS) // Consider only blocking squares
+        if constexpr (Type == EVASIONS) // Consider only blocking squares
         {
             b1 &= target;
             b2 &= target;
         }
 
-        if (Type == QUIET_CHECKS)
+        if constexpr (Type == QUIET_CHECKS)
         {
             // To make a quiet check, you either make a direct check by pushing a pawn
             // or push a blocker pawn that is not on the same file as the enemy king.
@@ -102,21 +110,21 @@ namespace {
         Bitboard b2 = shift<UpLeft >(pawnsOn7) & enemies;
         Bitboard b3 = shift<Up     >(pawnsOn7) & emptySquares;
 
-        if (Type == EVASIONS)
+        if constexpr (Type == EVASIONS)
             b3 &= target;
 
         while (b1)
-            moveList = make_promotions<Type, UpRight>(moveList, pop_lsb(b1));
+            moveList = make_promotions<Type, UpRight, true>(moveList, pop_lsb(b1));
 
         while (b2)
-            moveList = make_promotions<Type, UpLeft >(moveList, pop_lsb(b2));
+            moveList = make_promotions<Type, UpLeft, true>(moveList, pop_lsb(b2));
 
         while (b3)
-            moveList = make_promotions<Type, Up     >(moveList, pop_lsb(b3));
+            moveList = make_promotions<Type, Up,    false>(moveList, pop_lsb(b3));
     }
 
     // Standard and en passant captures
-    if (Type == CAPTURES || Type == EVASIONS || Type == NON_EVASIONS)
+    if constexpr (Type == CAPTURES || Type == EVASIONS || Type == NON_EVASIONS)
     {
         Bitboard b1 = shift<UpRight>(pawnsNotOn7) & enemies;
         Bitboard b2 = shift<UpLeft >(pawnsNotOn7) & enemies;
@@ -264,7 +272,7 @@ ExtMove* generate<LEGAL>(const Position& pos, ExtMove* moveList) {
   moveList = pos.checkers() ? generate<EVASIONS    >(pos, moveList)
                             : generate<NON_EVASIONS>(pos, moveList);
   while (cur != moveList)
-      if (  ((pinned && pinned & from_sq(*cur)) || from_sq(*cur) == ksq || type_of(*cur) == EN_PASSANT)
+      if (  ((pinned & from_sq(*cur)) || from_sq(*cur) == ksq || type_of(*cur) == EN_PASSANT)
           && !pos.legal(*cur))
           *cur = (--moveList)->move;
       else

src/movegen.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/movepick.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -69,7 +69,6 @@ MovePicker::MovePicker(const Position& p, Move ttm, Depth d, const ButterflyHist
 
   stage = (pos.checkers() ? EVASION_TT : MAIN_TT) +
           !(ttm && pos.pseudo_legal(ttm));
-  threatenedPieces = 0;
 }
 
 /// MovePicker constructor for quiescence search
@@ -93,20 +92,20 @@ MovePicker::MovePicker(const Position& p, Move ttm, Value th, const CapturePiece
 {
   assert(!pos.checkers());
 
-  stage = PROBCUT_TT + !(ttm && pos.capture(ttm)
+  stage = PROBCUT_TT + !(ttm && pos.capture_stage(ttm)
                              && pos.pseudo_legal(ttm)
                              && pos.see_ge(ttm, threshold));
 }
 
 /// MovePicker::score() assigns a numerical value to each move in a list, used
 /// for sorting. Captures are ordered by Most Valuable Victim (MVV), preferring
-/// captures with a good history. Quiets moves are ordered using the histories.
+/// captures with a good history. Quiets moves are ordered using the history tables.
 template<GenType Type>
 void MovePicker::score() {
 
   static_assert(Type == CAPTURES || Type == QUIETS || Type == EVASIONS, "Wrong type");
 
-  [[maybe_unused]] Bitboard threatenedByPawn, threatenedByMinor, threatenedByRook;
+  [[maybe_unused]] Bitboard threatenedByPawn, threatenedByMinor, threatenedByRook, threatenedPieces;
   if constexpr (Type == QUIETS)
   {
       Color us = pos.side_to_move();
@@ -123,8 +122,8 @@ void MovePicker::score() {
 
   for (auto& m : *this)
       if constexpr (Type == CAPTURES)
-          m.value =  6 * int(PieceValue[MG][pos.piece_on(to_sq(m))])
+          m.value =  (7 * int(PieceValue[MG][pos.piece_on(to_sq(m))])
-                   +     (*captureHistory)[pos.moved_piece(m)][to_sq(m)][type_of(pos.piece_on(to_sq(m)))];
+                   +     (*captureHistory)[pos.moved_piece(m)][to_sq(m)][type_of(pos.piece_on(to_sq(m)))]) / 16;
 
       else if constexpr (Type == QUIETS)
           m.value =  2 * (*mainHistory)[pos.side_to_move()][from_to(m)]
@@ -141,7 +140,7 @@ void MovePicker::score() {
                    +     bool(pos.check_squares(type_of(pos.moved_piece(m))) & to_sq(m)) * 16384;
       else // Type == EVASIONS
       {
-          if (pos.capture(m))
+          if (pos.capture_stage(m))
               m.value =  PieceValue[MG][pos.piece_on(to_sq(m))]
                        - Value(type_of(pos.moved_piece(m)))
                        + (1 << 28);
@@ -158,7 +157,7 @@ Move MovePicker::select(Pred filter) {
 
   while (cur < endMoves)
   {
-      if (T == Best)
+      if constexpr (T == Best)
           std::swap(*cur, *std::max_element(cur, endMoves));
 
       if (*cur != ttMove && filter())
@@ -197,7 +196,7 @@ top:
 
   case GOOD_CAPTURE:
       if (select<Next>([&](){
-                       return pos.see_ge(*cur, Value(-69 * cur->value / 1024)) ?
+                       return pos.see_ge(*cur, Value(-cur->value)) ?
                               // Move losing capture to endBadCaptures to be tried later
                               true : (*endBadCaptures++ = *cur, false); }))
           return *(cur - 1);
@@ -216,7 +215,7 @@ top:
 
   case REFUTATION:
       if (select<Next>([&](){ return    *cur != MOVE_NONE
-                                    && !pos.capture(*cur)
+                                    && !pos.capture_stage(*cur)
                                     &&  pos.pseudo_legal(*cur); }))
           return *(cur - 1);
       ++stage;

src/movepick.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -62,14 +62,14 @@ public:
 template <typename T, int D, int Size, int... Sizes>
 struct Stats : public std::array<Stats<T, D, Sizes...>, Size>
 {
-  typedef Stats<T, D, Size, Sizes...> stats;
+  using stats = Stats<T, D, Size, Sizes...>;
 
   void fill(const T& v) {
 
     // For standard-layout 'this' points to first struct member
     assert(std::is_standard_layout<stats>::value);
 
-    typedef StatsEntry<T, D> entry;
+    using entry = StatsEntry<T, D>;
     entry* p = reinterpret_cast<entry*>(this);
     std::fill(p, p + sizeof(*this) / sizeof(entry), v);
   }
@@ -87,23 +87,23 @@ enum StatsType { NoCaptures, Captures };
 /// ordering decisions. It uses 2 tables (one for each color) indexed by
 /// the move's from and to squares, see www.chessprogramming.org/Butterfly_Boards
 /// (~11 elo)
-typedef Stats<int16_t, 7183, COLOR_NB, int(SQUARE_NB) * int(SQUARE_NB)> ButterflyHistory;
+using ButterflyHistory = Stats<int16_t, 7183, COLOR_NB, int(SQUARE_NB) * int(SQUARE_NB)>;
 
 /// CounterMoveHistory stores counter moves indexed by [piece][to] of the previous
 /// move, see www.chessprogramming.org/Countermove_Heuristic
-typedef Stats<Move, NOT_USED, PIECE_NB, SQUARE_NB> CounterMoveHistory;
+using CounterMoveHistory = Stats<Move, NOT_USED, PIECE_NB, SQUARE_NB>;
 
 /// CapturePieceToHistory is addressed by a move's [piece][to][captured piece type]
-typedef Stats<int16_t, 10692, PIECE_NB, SQUARE_NB, PIECE_TYPE_NB> CapturePieceToHistory;
+using CapturePieceToHistory = Stats<int16_t, 10692, PIECE_NB, SQUARE_NB, PIECE_TYPE_NB>;
 
 /// PieceToHistory is like ButterflyHistory but is addressed by a move's [piece][to]
-typedef Stats<int16_t, 29952, PIECE_NB, SQUARE_NB> PieceToHistory;
+using PieceToHistory = Stats<int16_t, 29952, PIECE_NB, SQUARE_NB>;
 
 /// ContinuationHistory is the combined history of a given pair of moves, usually
 /// the current one given a previous one. The nested history table is based on
 /// PieceToHistory instead of ButterflyBoards.
 /// (~63 elo)
-typedef Stats<PieceToHistory, NOT_USED, PIECE_NB, SQUARE_NB> ContinuationHistory;
+using ContinuationHistory = Stats<PieceToHistory, NOT_USED, PIECE_NB, SQUARE_NB>;
 
 
 /// MovePicker class is used to pick one pseudo-legal move at a time from the
@@ -131,8 +131,6 @@ public:
   MovePicker(const Position&, Move, Value, const CapturePieceToHistory*);
   Move next_move(bool skipQuiets = false);
 
-  Bitboard threatenedPieces;
-
 private:
   template<PickType T, typename Pred> Move select(Pred);
   template<GenType> void score();

src/nnue/evaluate_nnue.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -18,15 +18,15 @@
 
 // Code for calculating NNUE evaluation function
 
+#include <fstream>
+#include <iomanip>
 #include <iostream>
 #include <set>
 #include <sstream>
-#include <iomanip>
+#include <string_view>
-#include <fstream>
 
 #include "../evaluate.h"
 #include "../position.h"
-#include "../misc.h"
 #include "../uci.h"
 #include "../types.h"
 
@@ -83,7 +83,7 @@ namespace Stockfish::Eval::NNUE {
   }  // namespace Detail
 
   // Initialize the evaluation function parameters
-  void initialize() {
+  static void initialize() {
 
     Detail::initialize(featureTransformer);
     for (std::size_t i = 0; i < LayerStacks; ++i)
@@ -91,7 +91,7 @@ namespace Stockfish::Eval::NNUE {
   }
 
   // Read network header
-  bool read_header(std::istream& stream, std::uint32_t* hashValue, std::string* desc)
+  static bool read_header(std::istream& stream, std::uint32_t* hashValue, std::string* desc)
   {
     std::uint32_t version, size;
 
@@ -105,7 +105,7 @@ namespace Stockfish::Eval::NNUE {
   }
 
   // Write network header
-  bool write_header(std::ostream& stream, std::uint32_t hashValue, const std::string& desc)
+  static bool write_header(std::ostream& stream, std::uint32_t hashValue, const std::string& desc)
   {
     write_little_endian<std::uint32_t>(stream, Version);
     write_little_endian<std::uint32_t>(stream, hashValue);
@@ -115,7 +115,7 @@ namespace Stockfish::Eval::NNUE {
   }
 
   // Read network parameters
-  bool read_parameters(std::istream& stream) {
+  static bool read_parameters(std::istream& stream) {
 
     std::uint32_t hashValue;
     if (!read_header(stream, &hashValue, &netDescription)) return false;
@@ -127,7 +127,7 @@ namespace Stockfish::Eval::NNUE {
   }
 
   // Write network parameters
-  bool write_parameters(std::ostream& stream) {
+  static bool write_parameters(std::ostream& stream) {
 
     if (!write_header(stream, HashValue, netDescription)) return false;
     if (!Detail::write_parameters(stream, *featureTransformer)) return false;
@@ -136,6 +136,11 @@ namespace Stockfish::Eval::NNUE {
     return (bool)stream;
   }
 
+  void hint_common_parent_position(const Position& pos) {
+    if (Eval::useNNUE)
+        featureTransformer->hint_common_access(pos);
+  }
+
   // Evaluation function. Perform differential calculation.
   Value evaluate(const Position& pos, bool adjusted, int* complexity) {
 
@@ -143,7 +148,7 @@ namespace Stockfish::Eval::NNUE {
     // overaligning stack variables with alignas() doesn't work correctly.
 
     constexpr uint64_t alignment = CacheLineSize;
-    int delta = 24 - pos.non_pawn_material() / 9560;
+    constexpr int delta = 24;
 
 #if defined(ALIGNAS_ON_STACK_VARIABLES_BROKEN)
     TransformedFeatureType transformedFeaturesUnaligned[
@@ -211,7 +216,7 @@ namespace Stockfish::Eval::NNUE {
     return t;
   }
 
-  static const std::string PieceToChar(" PNBRQK  pnbrqk");
+  constexpr std::string_view PieceToChar(" PNBRQK  pnbrqk");
 
 
   // format_cp_compact() converts a Value into (centi)pawns and writes it in a buffer.
@@ -245,14 +250,15 @@ namespace Stockfish::Eval::NNUE {
   }
 
 
-  // format_cp_aligned_dot() converts a Value into (centi)pawns and writes it in a buffer,
+  // format_cp_aligned_dot() converts a Value into (centi)pawns, always keeping two decimals.
-  // always keeping two decimals. The buffer must have capacity for at least 7 chars.
+  static void format_cp_aligned_dot(Value v, std::stringstream &stream) {
-  static void format_cp_aligned_dot(Value v, char* buffer) {
+    const double cp = 1.0 * std::abs(int(v)) / UCI::NormalizeToPawnValue;
 
-    buffer[0] = (v < 0 ? '-' : v > 0 ? '+' : ' ');
+    stream << (v < 0 ? '-' : v > 0 ? '+' : ' ')
-
+           << std::setiosflags(std::ios::fixed)
-    double cp = 1.0 * std::abs(int(v)) / UCI::NormalizeToPawnValue;
+           << std::setw(6)
-    sprintf(&buffer[1], "%6.2f", cp);
+           << std::setprecision(2)
+           << cp;
   }
 
 
@@ -332,17 +338,10 @@ namespace Stockfish::Eval::NNUE {
 
     for (std::size_t bucket = 0; bucket < LayerStacks; ++bucket)
     {
-      char buffer[3][8];
+      ss <<  "|  " << bucket    << "        ";
-      std::memset(buffer, '\0', sizeof(buffer));
+      ss << " |  "; format_cp_aligned_dot(t.psqt[bucket], ss); ss << "  "
-
+         << " |  "; format_cp_aligned_dot(t.positional[bucket], ss); ss << "  "
-      format_cp_aligned_dot(t.psqt[bucket], buffer[0]);
+         << " |  "; format_cp_aligned_dot(t.psqt[bucket] + t.positional[bucket], ss); ss << "  "
-      format_cp_aligned_dot(t.positional[bucket], buffer[1]);
-      format_cp_aligned_dot(t.psqt[bucket] + t.positional[bucket], buffer[2]);
-
-      ss <<  "|  " << bucket    << "        "
-         << " |  " << buffer[0] << "  "
-         << " |  " << buffer[1] << "  "
-         << " |  " << buffer[2] << "  "
          << " |";
       if (bucket == t.correctBucket)
           ss << " <-- this bucket is used";

src/nnue/evaluate_nnue.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -31,6 +31,7 @@ namespace Stockfish::Eval::NNUE {
   constexpr std::uint32_t HashValue =
       FeatureTransformer::get_hash_value() ^ Network::get_hash_value();
 
+
   // Deleter for automating release of memory area
   template <typename T>
   struct AlignedDeleter {
@@ -54,6 +55,14 @@ namespace Stockfish::Eval::NNUE {
   template <typename T>
   using LargePagePtr = std::unique_ptr<T, LargePageDeleter<T>>;
 
+  std::string trace(Position& pos);
+  Value evaluate(const Position& pos, bool adjusted = false, int* complexity = nullptr);
+  void hint_common_parent_position(const Position& pos);
+
+  bool load_eval(std::string name, std::istream& stream);
+  bool save_eval(std::ostream& stream);
+  bool save_eval(const std::optional<std::string>& filename);
+
 }  // namespace Stockfish::Eval::NNUE
 
 #endif // #ifndef NNUE_EVALUATE_NNUE_H_INCLUDED

src/nnue/features/half_ka_v2_hm.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/nnue/features/half_ka_v2_hm.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/nnue/layers/affine_transform.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -31,7 +31,7 @@
   This file contains the definition for a fully connected layer (aka affine transform).
   Two approaches are employed, depending on the sizes of the transform.
 
-  Approach 1:
+  Approach 1 (a specialization for large inputs):
     - used when the PaddedInputDimensions >= 128
     - uses AVX512 if possible
     - processes inputs in batches of 2*InputSimdWidth
@@ -42,9 +42,8 @@
       depends on the architecture (the amount of registers)
     - accumulate + hadd is used
 
-  Approach 2:
+  Approach 2 (a specialization for small inputs):
     - used when the PaddedInputDimensions < 128
-    - does not use AVX512
     - expected use-case is for when PaddedInputDimensions == 32 and InputDimensions <= 32.
       - that's why AVX512 is hard to implement
     - expected use-case is small layers
@@ -72,6 +71,10 @@ namespace Stockfish::Eval::NNUE::Layers {
     const __m64 Zeros = _mm_setzero_si64();
     const auto inputVector = reinterpret_cast<const __m64*>(input);
 
+# elif defined(USE_NEON_DOTPROD)
+    constexpr IndexType NumChunks = ceil_to_multiple<IndexType>(InputDimensions, 16) / 16;
+    const auto inputVector = reinterpret_cast<const int8x16_t*>(input);
+
 # elif defined(USE_NEON)
     constexpr IndexType NumChunks = ceil_to_multiple<IndexType>(InputDimensions, 16) / 16;
     const auto inputVector = reinterpret_cast<const int8x8_t*>(input);
@@ -123,6 +126,14 @@ namespace Stockfish::Eval::NNUE::Layers {
       sum = _mm_add_pi32(sum, _mm_unpackhi_pi32(sum, sum));
       output[i] = _mm_cvtsi64_si32(sum);
 
+# elif defined(USE_NEON_DOTPROD)
+      int32x4_t sum = {biases[i]};
+      const auto row = reinterpret_cast<const int8x16_t*>(&weights[offset]);
+      for (IndexType j = 0; j < NumChunks; ++j) {
+        sum = vdotq_s32(sum, inputVector[j], row[j]);
+      }
+      output[i] = vaddvq_s32(sum);
+
 # elif defined(USE_NEON)
       int32x4_t sum = {biases[i]};
       const auto row = reinterpret_cast<const int8x8_t*>(&weights[offset]);
@@ -157,7 +168,7 @@ namespace Stockfish::Eval::NNUE::Layers {
   constexpr IndexType LargeInputSize = std::numeric_limits<IndexType>::max();
 #endif
 
-  // A specialization for large inputs.
+  // A specialization for large inputs
   template <IndexType InDims, IndexType OutDims>
   class AffineTransform<InDims, OutDims, std::enable_if_t<(ceil_to_multiple<IndexType>(InDims, MaxSimdWidth) >= LargeInputSize)>> {
    public:
@@ -176,36 +187,39 @@ namespace Stockfish::Eval::NNUE::Layers {
 
     using OutputBuffer = OutputType[PaddedOutputDimensions];
 
-    static_assert(PaddedInputDimensions >= LargeInputSize, "Something went wrong. This specialization should not have been chosen.");
+    static_assert(PaddedInputDimensions >= LargeInputSize, "Something went wrong. This specialization (for large inputs) should not have been chosen.");
 
 #if defined (USE_AVX512)
-    static constexpr const IndexType InputSimdWidth = 64;
+    static constexpr IndexType InputSimdWidth = 64;
-    static constexpr const IndexType MaxNumOutputRegs = 16;
+    static constexpr IndexType MaxNumOutputRegs = 16;
 #elif defined (USE_AVX2)
-    static constexpr const IndexType InputSimdWidth = 32;
+    static constexpr IndexType InputSimdWidth = 32;
-    static constexpr const IndexType MaxNumOutputRegs = 8;
+    static constexpr IndexType MaxNumOutputRegs = 8;
 #elif defined (USE_SSSE3)
-    static constexpr const IndexType InputSimdWidth = 16;
+    static constexpr IndexType InputSimdWidth = 16;
-    static constexpr const IndexType MaxNumOutputRegs = 8;
+    static constexpr IndexType MaxNumOutputRegs = 8;
+#elif defined (USE_NEON_DOTPROD)
+    static constexpr IndexType InputSimdWidth = 16;
+    static constexpr IndexType MaxNumOutputRegs = 8;
 #elif defined (USE_NEON)
-    static constexpr const IndexType InputSimdWidth = 8;
+    static constexpr IndexType InputSimdWidth = 8;
-    static constexpr const IndexType MaxNumOutputRegs = 8;
+    static constexpr IndexType MaxNumOutputRegs = 8;
 #else
     // The fallback implementation will not have permuted weights.
     // We define these to avoid a lot of ifdefs later.
-    static constexpr const IndexType InputSimdWidth = 1;
+    static constexpr IndexType InputSimdWidth = 1;
-    static constexpr const IndexType MaxNumOutputRegs = 1;
+    static constexpr IndexType MaxNumOutputRegs = 1;
 #endif
 
     // A big block is a region in the weight matrix of the size [PaddedInputDimensions, NumOutputRegs].
     // A small block is a region of size [InputSimdWidth, 1]
 
-    static constexpr const IndexType NumOutputRegs = std::min(MaxNumOutputRegs, OutputDimensions);
+    static constexpr IndexType NumOutputRegs = std::min(MaxNumOutputRegs, OutputDimensions);
-    static constexpr const IndexType SmallBlockSize = InputSimdWidth;
+    static constexpr IndexType SmallBlockSize = InputSimdWidth;
-    static constexpr const IndexType BigBlockSize = NumOutputRegs * PaddedInputDimensions;
+    static constexpr IndexType BigBlockSize = NumOutputRegs * PaddedInputDimensions;
-    static constexpr const IndexType NumSmallBlocksInBigBlock = BigBlockSize / SmallBlockSize;
+    static constexpr IndexType NumSmallBlocksInBigBlock = BigBlockSize / SmallBlockSize;
-    static constexpr const IndexType NumSmallBlocksPerOutput = PaddedInputDimensions / SmallBlockSize;
+    static constexpr IndexType NumSmallBlocksPerOutput = PaddedInputDimensions / SmallBlockSize;
-    static constexpr const IndexType NumBigBlocks = OutputDimensions / NumOutputRegs;
+    static constexpr IndexType NumBigBlocks = OutputDimensions / NumOutputRegs;
 
     static_assert(OutputDimensions % NumOutputRegs == 0);
 
@@ -241,8 +255,7 @@ namespace Stockfish::Eval::NNUE::Layers {
 
     // Read network parameters
     bool read_parameters(std::istream& stream) {
-      for (IndexType i = 0; i < OutputDimensions; ++i)
+      read_little_endian<BiasType>(stream, biases, OutputDimensions);
-        biases[i] = read_little_endian<BiasType>(stream);
 
       for (IndexType i = 0; i < OutputDimensions * PaddedInputDimensions; ++i)
         weights[get_weight_index(i)] = read_little_endian<WeightType>(stream);
@@ -252,8 +265,7 @@ namespace Stockfish::Eval::NNUE::Layers {
 
     // Write network parameters
     bool write_parameters(std::ostream& stream) const {
-      for (IndexType i = 0; i < OutputDimensions; ++i)
+      write_little_endian<BiasType>(stream, biases, OutputDimensions);
-          write_little_endian<BiasType>(stream, biases[i]);
 
       for (IndexType i = 0; i < OutputDimensions * PaddedInputDimensions; ++i)
         write_little_endian<WeightType>(stream, weights[get_weight_index(i)]);
@@ -292,6 +304,15 @@ namespace Stockfish::Eval::NNUE::Layers {
       #define vec_add_dpbusd_32x2 Simd::m128_add_dpbusd_epi32x2
       #define vec_hadd Simd::m128_hadd
       #define vec_haddx4 Simd::m128_haddx4
+#elif defined (USE_NEON_DOTPROD)
+      using acc_vec_t = int32x4_t;
+      using bias_vec_t = int32x4_t;
+      using weight_vec_t = int8x16_t;
+      using in_vec_t = int8x16_t;
+      #define vec_zero {0}
+      #define vec_add_dpbusd_32x2 Simd::dotprod_m128_add_dpbusd_epi32x2
+      #define vec_hadd Simd::neon_m128_hadd
+      #define vec_haddx4 Simd::neon_m128_haddx4
 #elif defined (USE_NEON)
       using acc_vec_t = int32x4_t;
       using bias_vec_t = int32x4_t;
@@ -374,6 +395,7 @@ namespace Stockfish::Eval::NNUE::Layers {
     alignas(CacheLineSize) WeightType weights[OutputDimensions * PaddedInputDimensions];
   };
 
+  // A specialization for small inputs
   template <IndexType InDims, IndexType OutDims>
   class AffineTransform<InDims, OutDims, std::enable_if_t<(ceil_to_multiple<IndexType>(InDims, MaxSimdWidth) < LargeInputSize)>> {
    public:
@@ -393,12 +415,7 @@ namespace Stockfish::Eval::NNUE::Layers {
 
     using OutputBuffer = OutputType[PaddedOutputDimensions];
 
-    static_assert(PaddedInputDimensions < LargeInputSize, "Something went wrong. This specialization should not have been chosen.");
+    static_assert(PaddedInputDimensions < LargeInputSize, "Something went wrong. This specialization (for small inputs) should not have been chosen.");
-
-#if defined (USE_SSSE3)
-    static constexpr const IndexType OutputSimdWidth = SimdWidth / 4;
-    static constexpr const IndexType InputSimdWidth = SimdWidth;
-#endif
 
     // Hash value embedded in the evaluation file
     static constexpr std::uint32_t get_hash_value(std::uint32_t prevHash) {
@@ -428,8 +445,7 @@ namespace Stockfish::Eval::NNUE::Layers {
 
     // Read network parameters
     bool read_parameters(std::istream& stream) {
-      for (IndexType i = 0; i < OutputDimensions; ++i)
+      read_little_endian<BiasType>(stream, biases, OutputDimensions);
-        biases[i] = read_little_endian<BiasType>(stream);
       for (IndexType i = 0; i < OutputDimensions * PaddedInputDimensions; ++i)
         weights[get_weight_index(i)] = read_little_endian<WeightType>(stream);
 
@@ -438,8 +454,7 @@ namespace Stockfish::Eval::NNUE::Layers {
 
     // Write network parameters
     bool write_parameters(std::ostream& stream) const {
-      for (IndexType i = 0; i < OutputDimensions; ++i)
+      write_little_endian<BiasType>(stream, biases, OutputDimensions);
-        write_little_endian<BiasType>(stream, biases[i]);
 
       for (IndexType i = 0; i < OutputDimensions * PaddedInputDimensions; ++i)
         write_little_endian<WeightType>(stream, weights[get_weight_index(i)]);
@@ -450,29 +465,34 @@ namespace Stockfish::Eval::NNUE::Layers {
     const OutputType* propagate(
         const InputType* input, OutputType* output) const {
 
-#if defined (USE_AVX2)
+#if defined (USE_AVX512)
+      using vec_t = __m512i;
+      #define vec_setzero _mm512_setzero_si512
+      #define vec_set_32 _mm512_set1_epi32
+      #define vec_add_dpbusd_32 Simd::m512_add_dpbusd_epi32
+      #define vec_add_dpbusd_32x2 Simd::m512_add_dpbusd_epi32x2
+      #define vec_hadd Simd::m512_hadd
+#elif defined (USE_AVX2)
       using vec_t = __m256i;
       #define vec_setzero _mm256_setzero_si256
       #define vec_set_32 _mm256_set1_epi32
       #define vec_add_dpbusd_32 Simd::m256_add_dpbusd_epi32
       #define vec_add_dpbusd_32x2 Simd::m256_add_dpbusd_epi32x2
-      #define vec_add_dpbusd_32x4 Simd::m256_add_dpbusd_epi32x4
       #define vec_hadd Simd::m256_hadd
-      #define vec_haddx4 Simd::m256_haddx4
 #elif defined (USE_SSSE3)
       using vec_t = __m128i;
       #define vec_setzero _mm_setzero_si128
       #define vec_set_32 _mm_set1_epi32
       #define vec_add_dpbusd_32 Simd::m128_add_dpbusd_epi32
       #define vec_add_dpbusd_32x2 Simd::m128_add_dpbusd_epi32x2
-      #define vec_add_dpbusd_32x4 Simd::m128_add_dpbusd_epi32x4
       #define vec_hadd Simd::m128_hadd
-      #define vec_haddx4 Simd::m128_haddx4
 #endif
 
 #if defined (USE_SSSE3)
       const auto inputVector = reinterpret_cast<const vec_t*>(input);
 
+      static constexpr IndexType OutputSimdWidth = sizeof(vec_t) / sizeof(OutputType);
+
       static_assert(OutputDimensions % OutputSimdWidth == 0 || OutputDimensions == 1);
 
       if constexpr (OutputDimensions % OutputSimdWidth == 0)
@@ -518,9 +538,7 @@ namespace Stockfish::Eval::NNUE::Layers {
 # undef vec_set_32
 # undef vec_add_dpbusd_32
 # undef vec_add_dpbusd_32x2
-# undef vec_add_dpbusd_32x4
 # undef vec_hadd
-# undef vec_haddx4
 #else
       // Use old implementation for the other architectures.
       affine_transform_non_ssse3<

src/nnue/layers/affine_transform_sparse_input.h

@@ -0,0 +1,286 @@
+/*
+  Stockfish, a UCI chess playing engine derived from Glaurung 2.1
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
+
+  Stockfish is free software: you can redistribute it and/or modify
+  it under the terms of the GNU General Public License as published by
+  the Free Software Foundation, either version 3 of the License, or
+  (at your option) any later version.
+
+  Stockfish is distributed in the hope that it will be useful,
+  but WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+  GNU General Public License for more details.
+
+  You should have received a copy of the GNU General Public License
+  along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+// Definition of layer AffineTransformSparseInput of NNUE evaluation function
+
+#ifndef NNUE_LAYERS_AFFINE_TRANSFORM_SPARSE_INPUT_H_INCLUDED
+#define NNUE_LAYERS_AFFINE_TRANSFORM_SPARSE_INPUT_H_INCLUDED
+
+#include <iostream>
+#include <algorithm>
+#include <array>
+#include <type_traits>
+#include "../nnue_common.h"
+#include "affine_transform.h"
+#include "simd.h"
+
+/*
+  This file contains the definition for a fully connected layer (aka affine transform) with block sparse input.
+*/
+
+namespace Stockfish::Eval::NNUE::Layers {
+#if defined(__GNUC__)  // GCC, Clang, ICC
+
+  static inline IndexType lsb_(std::uint32_t b) {
+    assert(b);
+    return IndexType(__builtin_ctzl(b));
+  }
+
+#elif defined(_MSC_VER)  // MSVC
+
+  static inline IndexType lsb_(std::uint32_t b) {
+    assert(b);
+    unsigned long idx;
+    _BitScanForward(&idx, b);
+    return (IndexType) idx;
+  }
+
+#else  // Compiler is neither GCC nor MSVC compatible
+
+#error "Compiler not supported."
+
+#endif
+
+
+#if defined(USE_SSSE3)
+  alignas(CacheLineSize) static inline const std::array<std::array<std::uint16_t, 8>, 256> lookup_indices = [](){
+    std::array<std::array<std::uint16_t, 8>, 256> v{};
+    for (int i = 0; i < 256; ++i)
+    {
+      int j = i;
+      int k = 0;
+      while(j)
+      {
+        const IndexType lsbIndex = lsb_(std::uint32_t(j));
+        j &= j - 1;
+        v[i][k] = lsbIndex;
+        ++k;
+      }
+    }
+    return v;
+  }();
+  alignas(CacheLineSize) static inline const std::array<unsigned, 256> lookup_count = [](){
+    std::array<unsigned, 256> v;
+    for (int i = 0; i < 256; ++i)
+    {
+      int j = i;
+      int k = 0;
+      while(j)
+      {
+        j &= j - 1;
+        ++k;
+      }
+      v[i] = k;
+    }
+    return v;
+  }();
+
+  // Find indices of nonzero numbers in an int32_t array
+  template<const IndexType InputDimensions>
+  void find_nnz(const std::int32_t* input, std::uint16_t* out, IndexType& count_out) {
+#if defined (USE_AVX512)
+    using vec_t = __m512i;
+    #define vec_nnz(a) _mm512_cmpgt_epi32_mask(a, _mm512_setzero_si512())
+#elif defined (USE_AVX2)
+    using vec_t = __m256i;
+    #define vec_nnz(a) _mm256_movemask_ps(_mm256_castsi256_ps(_mm256_cmpgt_epi32(a, _mm256_setzero_si256())))
+#elif defined (USE_SSSE3)
+    using vec_t = __m128i;
+    #define vec_nnz(a) _mm_movemask_ps(_mm_castsi128_ps(_mm_cmpgt_epi32(a, _mm_setzero_si128())))
+#endif
+    constexpr IndexType InputSimdWidth = sizeof(vec_t) / sizeof(std::int32_t);
+    // Inputs are processed InputSimdWidth at a time and outputs are processed 8 at a time so we process in chunks of max(InputSimdWidth, 8)
+    constexpr IndexType ChunkSize = std::max<IndexType>(InputSimdWidth, 8);
+    constexpr IndexType NumChunks = InputDimensions / ChunkSize;
+    constexpr IndexType InputsPerChunk = ChunkSize / InputSimdWidth;
+    constexpr IndexType OutputsPerChunk = ChunkSize / 8;
+
+    const auto inputVector = reinterpret_cast<const vec_t*>(input);
+    IndexType count = 0;
+    __m128i base = _mm_set1_epi16(0);
+    __m128i increment = _mm_set1_epi16(8);
+    for (IndexType i = 0; i < NumChunks; ++i)
+    {
+      // bitmask of nonzero values in this chunk
+      unsigned nnz = 0;
+      for (IndexType j = 0; j < InputsPerChunk; ++j)
+      {
+        const vec_t inputChunk = inputVector[i * InputsPerChunk + j];
+        nnz |= (unsigned)vec_nnz(inputChunk) << (j * InputSimdWidth);
+      }
+      for (IndexType j = 0; j < OutputsPerChunk; ++j)
+      {
+        const auto lookup = (nnz >> (j * 8)) & 0xFF;
+        const auto offsets = _mm_loadu_si128(reinterpret_cast<const __m128i*>(&lookup_indices[lookup]));
+        _mm_storeu_si128(reinterpret_cast<__m128i*>(out + count), _mm_add_epi16(base, offsets));
+        count += lookup_count[lookup];
+        base = _mm_add_epi16(base, increment);
+      }
+    }
+    count_out = count;
+  }
+# undef vec_nnz
+#endif
+
+  // Sparse input implementation
+  template <IndexType InDims, IndexType OutDims>
+  class AffineTransformSparseInput {
+   public:
+    // Input/output type
+    // Input/output type
+    using InputType = std::uint8_t;
+    using OutputType = std::int32_t;
+
+    // Number of input/output dimensions
+    static constexpr IndexType InputDimensions = InDims;
+    static constexpr IndexType OutputDimensions = OutDims;
+
+    static_assert(OutputDimensions % 16 == 0, "Only implemented for OutputDimensions divisible by 16.");
+
+    static constexpr IndexType PaddedInputDimensions =
+      ceil_to_multiple<IndexType>(InputDimensions, MaxSimdWidth);
+    static constexpr IndexType PaddedOutputDimensions =
+      ceil_to_multiple<IndexType>(OutputDimensions, MaxSimdWidth);
+
+#if defined (USE_SSSE3)
+    static constexpr IndexType ChunkSize = 4;
+#else
+    static constexpr IndexType ChunkSize = 1;
+#endif
+
+    using OutputBuffer = OutputType[PaddedOutputDimensions];
+
+    // Hash value embedded in the evaluation file
+    static constexpr std::uint32_t get_hash_value(std::uint32_t prevHash) {
+      std::uint32_t hashValue = 0xCC03DAE4u;
+      hashValue += OutputDimensions;
+      hashValue ^= prevHash >> 1;
+      hashValue ^= prevHash << 31;
+      return hashValue;
+    }
+
+    static IndexType get_weight_index_scrambled(IndexType i)
+    {
+      return
+        (i / ChunkSize) % (PaddedInputDimensions / ChunkSize) * OutputDimensions * ChunkSize +
+        i / PaddedInputDimensions * ChunkSize +
+        i % ChunkSize;
+    }
+
+    static IndexType get_weight_index(IndexType i)
+    {
+#if defined (USE_SSSE3)
+      return get_weight_index_scrambled(i);
+#else
+      return i;
+#endif
+    }
+
+    // Read network parameters
+    bool read_parameters(std::istream& stream) {
+      read_little_endian<BiasType>(stream, biases, OutputDimensions);
+      for (IndexType i = 0; i < OutputDimensions * PaddedInputDimensions; ++i)
+        weights[get_weight_index(i)] = read_little_endian<WeightType>(stream);
+
+      return !stream.fail();
+    }
+
+    // Write network parameters
+    bool write_parameters(std::ostream& stream) const {
+      write_little_endian<BiasType>(stream, biases, OutputDimensions);
+
+      for (IndexType i = 0; i < OutputDimensions * PaddedInputDimensions; ++i)
+        write_little_endian<WeightType>(stream, weights[get_weight_index(i)]);
+
+      return !stream.fail();
+    }
+    // Forward propagation
+    const OutputType* propagate(
+        const InputType* input, OutputType* output) const {
+
+#if defined (USE_SSSE3)
+#if defined (USE_AVX512)
+      using vec_t = __m512i;
+      #define vec_setzero _mm512_setzero_si512
+      #define vec_set_32 _mm512_set1_epi32
+      #define vec_add_dpbusd_32 Simd::m512_add_dpbusd_epi32
+#elif defined (USE_AVX2)
+      using vec_t = __m256i;
+      #define vec_setzero _mm256_setzero_si256
+      #define vec_set_32 _mm256_set1_epi32
+      #define vec_add_dpbusd_32 Simd::m256_add_dpbusd_epi32
+#elif defined (USE_SSSE3)
+      using vec_t = __m128i;
+      #define vec_setzero _mm_setzero_si128
+      #define vec_set_32 _mm_set1_epi32
+      #define vec_add_dpbusd_32 Simd::m128_add_dpbusd_epi32
+#endif
+      static constexpr IndexType OutputSimdWidth = sizeof(vec_t) / sizeof(OutputType);
+
+      constexpr IndexType NumChunks = ceil_to_multiple<IndexType>(InputDimensions, 8) / ChunkSize;
+      constexpr IndexType NumRegs = OutputDimensions / OutputSimdWidth;
+      std::uint16_t nnz[NumChunks];
+      IndexType count;
+
+      const auto input32 = reinterpret_cast<const std::int32_t*>(input);
+
+      // Find indices of nonzero 32bit blocks
+      find_nnz<NumChunks>(input32, nnz, count);
+
+      const vec_t* biasvec = reinterpret_cast<const vec_t*>(biases);
+      vec_t acc[NumRegs];
+      for (IndexType k = 0; k < NumRegs; ++k)
+        acc[k] = biasvec[k];
+
+      for (IndexType j = 0; j < count; ++j)
+      {
+        const auto i = nnz[j];
+        const vec_t in = vec_set_32(input32[i]);
+        const auto col = reinterpret_cast<const vec_t*>(&weights[i * OutputDimensions * ChunkSize]);
+        for (IndexType k = 0; k < NumRegs; ++k)
+          vec_add_dpbusd_32(acc[k], in, col[k]);
+      }
+
+      vec_t* outptr = reinterpret_cast<vec_t*>(output);
+      for (IndexType k = 0; k < NumRegs; ++k)
+        outptr[k] = acc[k];
+# undef vec_setzero
+# undef vec_set_32
+# undef vec_add_dpbusd_32
+#else
+      // Use dense implementation for the other architectures.
+      affine_transform_non_ssse3<
+        InputDimensions,
+        PaddedInputDimensions,
+        OutputDimensions>(output, weights, biases, input);
+#endif
+
+      return output;
+    }
+
+   private:
+    using BiasType = OutputType;
+    using WeightType = std::int8_t;
+
+    alignas(CacheLineSize) BiasType biases[OutputDimensions];
+    alignas(CacheLineSize) WeightType weights[OutputDimensions * PaddedInputDimensions];
+  };
+
+}  // namespace Stockfish::Eval::NNUE::Layers
+
+#endif // #ifndef NNUE_LAYERS_AFFINE_TRANSFORM_SPARSE_INPUT_H_INCLUDED

src/nnue/layers/clipped_relu.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/nnue/layers/simd.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -153,7 +153,7 @@ namespace Stockfish::Simd {
       asm(
         "vpdpbusd %[b0], %[a0], %[acc]\n\t"
         "vpdpbusd %[b1], %[a1], %[acc]\n\t"
-        : [acc]"+v"(acc)
+        : [acc]"+&v"(acc)
         : [a0]"v"(a0), [b0]"vm"(b0), [a1]"v"(a1), [b1]"vm"(b1)
       );
 #   else
@@ -165,18 +165,19 @@ namespace Stockfish::Simd {
       __m512i tmp0 = _mm512_maddubs_epi16(a0, b0);
       __m512i tmp1 = _mm512_maddubs_epi16(a1, b1);
       asm(
-          "vpaddsw     %[tmp0], %[tmp1], %[tmp0]\n\t"
           "vpmaddwd    %[tmp0], %[ones], %[tmp0]\n\t"
+          "vpmaddwd    %[tmp1], %[ones], %[tmp1]\n\t"
+          "vpaddd      %[tmp0], %[tmp1], %[tmp0]\n\t"
           "vpaddd      %[acc], %[tmp0], %[acc]\n\t"
-          : [acc]"+v"(acc), [tmp0]"+&v"(tmp0)
+          : [acc]"+v"(acc), [tmp0]"+&v"(tmp0), [tmp1]"+&v"(tmp1)
-          : [tmp1]"v"(tmp1), [ones]"v"(_mm512_set1_epi16(1))
+          : [ones]"v"(_mm512_set1_epi16(1))
       );
 #   else
       __m512i product0 = _mm512_maddubs_epi16(a0, b0);
       __m512i product1 = _mm512_maddubs_epi16(a1, b1);
-      product0 = _mm512_adds_epi16(product0, product1);
       product0 = _mm512_madd_epi16(product0, _mm512_set1_epi16(1));
-      acc = _mm512_add_epi32(acc, product0);
+      product1 = _mm512_madd_epi16(product1, _mm512_set1_epi16(1));
+      acc = _mm512_add_epi32(acc, _mm512_add_epi32(product0, product1));
 #   endif
 # endif
     }
@@ -249,7 +250,7 @@ namespace Stockfish::Simd {
       asm(
         VNNI_PREFIX "vpdpbusd %[b0], %[a0], %[acc]\n\t"
         VNNI_PREFIX "vpdpbusd %[b1], %[a1], %[acc]\n\t"
-        : [acc]"+v"(acc)
+        : [acc]"+&v"(acc)
         : [a0]"v"(a0), [b0]"vm"(b0), [a1]"v"(a1), [b1]"vm"(b1)
       );
 #   else
@@ -261,18 +262,19 @@ namespace Stockfish::Simd {
       __m256i tmp0 = _mm256_maddubs_epi16(a0, b0);
       __m256i tmp1 = _mm256_maddubs_epi16(a1, b1);
       asm(
-          "vpaddsw     %[tmp0], %[tmp1], %[tmp0]\n\t"
           "vpmaddwd    %[tmp0], %[ones], %[tmp0]\n\t"
+          "vpmaddwd    %[tmp1], %[ones], %[tmp1]\n\t"
+          "vpaddd      %[tmp0], %[tmp1], %[tmp0]\n\t"
           "vpaddd      %[acc], %[tmp0], %[acc]\n\t"
-          : [acc]"+v"(acc), [tmp0]"+&v"(tmp0)
+          : [acc]"+v"(acc), [tmp0]"+&v"(tmp0), [tmp1]"+&v"(tmp1)
-          : [tmp1]"v"(tmp1), [ones]"v"(_mm256_set1_epi16(1))
+          : [ones]"v"(_mm256_set1_epi16(1))
       );
 #   else
       __m256i product0 = _mm256_maddubs_epi16(a0, b0);
       __m256i product1 = _mm256_maddubs_epi16(a1, b1);
-      product0 = _mm256_adds_epi16(product0, product1);
       product0 = _mm256_madd_epi16(product0, _mm256_set1_epi16(1));
-      acc = _mm256_add_epi32(acc, product0);
+      product1 = _mm256_madd_epi16(product1, _mm256_set1_epi16(1));
+      acc = _mm256_add_epi32(acc, _mm256_add_epi32(product0, product1));
 #   endif
 # endif
     }
@@ -326,23 +328,37 @@ namespace Stockfish::Simd {
       __m128i tmp0 = _mm_maddubs_epi16(a0, b0);
       __m128i tmp1 = _mm_maddubs_epi16(a1, b1);
       asm(
-          "paddsw     %[tmp1], %[tmp0]\n\t"
           "pmaddwd    %[ones], %[tmp0]\n\t"
+          "pmaddwd    %[ones], %[tmp1]\n\t"
+          "paddd      %[tmp1], %[tmp0]\n\t"
           "paddd      %[tmp0], %[acc]\n\t"
-          : [acc]"+v"(acc), [tmp0]"+&v"(tmp0)
+          : [acc]"+v"(acc), [tmp0]"+&v"(tmp0), [tmp1]"+&v"(tmp1)
-          : [tmp1]"v"(tmp1), [ones]"v"(_mm_set1_epi16(1))
+          : [ones]"v"(_mm_set1_epi16(1))
       );
 #   else
       __m128i product0 = _mm_maddubs_epi16(a0, b0);
       __m128i product1 = _mm_maddubs_epi16(a1, b1);
-      product0 = _mm_adds_epi16(product0, product1);
       product0 = _mm_madd_epi16(product0, _mm_set1_epi16(1));
-      acc = _mm_add_epi32(acc, product0);
+      product1 = _mm_madd_epi16(product1, _mm_set1_epi16(1));
+      acc = _mm_add_epi32(acc, _mm_add_epi32(product0, product1));
 #   endif
     }
 
 #endif
 
+#if defined (USE_NEON_DOTPROD)
+
+    [[maybe_unused]] static void dotprod_m128_add_dpbusd_epi32x2(
+        int32x4_t& acc,
+        int8x16_t a0, int8x16_t b0,
+        int8x16_t a1, int8x16_t b1) {
+
+        acc = vdotq_s32(acc, a0, b0);
+        acc = vdotq_s32(acc, a1, b1);
+    }
+
+#endif
+
 #if defined (USE_NEON)
 
     [[maybe_unused]] static int neon_m128_reduce_add_epi32(int32x4_t s) {

src/nnue/layers/sqr_clipped_relu.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -106,7 +106,7 @@ namespace Stockfish::Eval::NNUE::Layers {
 
       for (IndexType i = Start; i < InputDimensions; ++i) {
         output[i] = static_cast<OutputType>(
-            // realy should be /127 but we need to make it fast
+            // really should be /127 but we need to make it fast
             // needs to be accounted for in the trainer
             std::max(0ll, std::min(127ll, (((long long)input[i] * input[i]) >> (2 * WeightScaleBits)) / 128)));
       }

src/nnue/nnue_accumulator.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/nnue/nnue_architecture.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -27,6 +27,7 @@
 
 #include "features/half_ka_v2_hm.h"
 
+#include "layers/affine_transform_sparse_input.h"
 #include "layers/affine_transform.h"
 #include "layers/clipped_relu.h"
 #include "layers/sqr_clipped_relu.h"
@@ -39,7 +40,7 @@ namespace Stockfish::Eval::NNUE {
 using FeatureSet = Features::HalfKAv2_hm;
 
 // Number of input feature dimensions after conversion
-constexpr IndexType TransformedFeatureDimensions = 1024;
+constexpr IndexType TransformedFeatureDimensions = 1536;
 constexpr IndexType PSQTBuckets = 8;
 constexpr IndexType LayerStacks = 8;
 
@@ -48,7 +49,7 @@ struct Network
   static constexpr int FC_0_OUTPUTS = 15;
   static constexpr int FC_1_OUTPUTS = 32;
 
-  Layers::AffineTransform<TransformedFeatureDimensions, FC_0_OUTPUTS + 1> fc_0;
+  Layers::AffineTransformSparseInput<TransformedFeatureDimensions, FC_0_OUTPUTS + 1> fc_0;
   Layers::SqrClippedReLU<FC_0_OUTPUTS + 1> ac_sqr_0;
   Layers::ClippedReLU<FC_0_OUTPUTS + 1> ac_0;
   Layers::AffineTransform<FC_0_OUTPUTS * 2, FC_1_OUTPUTS> fc_1;
@@ -72,22 +73,20 @@ struct Network
 
   // Read network parameters
   bool read_parameters(std::istream& stream) {
-    if (!fc_0.read_parameters(stream)) return false;
+    return   fc_0.read_parameters(stream)
-    if (!ac_0.read_parameters(stream)) return false;
+          && ac_0.read_parameters(stream)
-    if (!fc_1.read_parameters(stream)) return false;
+          && fc_1.read_parameters(stream)
-    if (!ac_1.read_parameters(stream)) return false;
+          && ac_1.read_parameters(stream)
-    if (!fc_2.read_parameters(stream)) return false;
+          && fc_2.read_parameters(stream);
-    return true;
   }
 
-  // Read network parameters
+  // Write network parameters
   bool write_parameters(std::ostream& stream) const {
-    if (!fc_0.write_parameters(stream)) return false;
+    return   fc_0.write_parameters(stream)
-    if (!ac_0.write_parameters(stream)) return false;
+          && ac_0.write_parameters(stream)
-    if (!fc_1.write_parameters(stream)) return false;
+          && fc_1.write_parameters(stream)
-    if (!ac_1.write_parameters(stream)) return false;
+          && ac_1.write_parameters(stream)
-    if (!fc_2.write_parameters(stream)) return false;
+          && fc_2.write_parameters(stream);
-    return true;
   }
 
   std::int32_t propagate(const TransformedFeatureType* transformedFeatures)

src/nnue/nnue_common.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -57,6 +57,9 @@ namespace Stockfish::Eval::NNUE {
   // Size of cache line (in bytes)
   constexpr std::size_t CacheLineSize = 64;
 
+  constexpr const char Leb128MagicString[] = "COMPRESSED_LEB128";
+  constexpr const std::size_t Leb128MagicStringSize = sizeof(Leb128MagicString) - 1;
+
   // SIMD width (in bytes)
   #if defined(USE_AVX2)
   constexpr std::size_t SimdWidth = 32;
@@ -159,6 +162,80 @@ namespace Stockfish::Eval::NNUE {
               write_little_endian<IntType>(stream, values[i]);
   }
 
+  template <typename IntType>
+  inline void read_leb_128(std::istream& stream, IntType* out, std::size_t count) {
+      static_assert(std::is_signed_v<IntType>, "Not implemented for unsigned types");
+      char leb128MagicString[Leb128MagicStringSize];
+      stream.read(leb128MagicString, Leb128MagicStringSize);
+      assert(strncmp(Leb128MagicString, leb128MagicString, Leb128MagicStringSize) == 0);
+      const std::uint32_t BUF_SIZE = 4096;
+      std::uint8_t buf[BUF_SIZE];
+      auto bytes_left = read_little_endian<std::uint32_t>(stream);
+      std::uint32_t buf_pos = BUF_SIZE;
+      for (std::size_t i = 0; i < count; ++i) {
+          IntType result = 0;
+          size_t shift = 0;
+          do {
+              if (buf_pos == BUF_SIZE) {
+                  stream.read(reinterpret_cast<char*>(buf), std::min(bytes_left, BUF_SIZE));
+                  buf_pos = 0;
+              }
+              std::uint8_t byte = buf[buf_pos++];
+              --bytes_left;
+              result |= (byte & 0x7f) << shift;
+              shift += 7;
+              if ((byte & 0x80) == 0) {
+                  out[i] = sizeof(IntType) * 8 <= shift || (byte & 0x40) == 0 ? result : result | ~((1 << shift) - 1);
+                  break;
+              }
+          } while (shift < sizeof(IntType) * 8);
+      }
+      assert(bytes_left == 0);
+  }
+
+  template <typename IntType>
+  inline void write_leb_128(std::ostream& stream, const IntType* values, std::size_t count) {
+      static_assert(std::is_signed_v<IntType>, "Not implemented for unsigned types");
+      stream.write(Leb128MagicString, Leb128MagicStringSize);
+      std::uint32_t byte_count = 0;
+      for (std::size_t i = 0; i < count; ++i) {
+          IntType value = values[i];
+          std::uint8_t byte;
+          do {
+              byte = value & 0x7f;
+              value >>= 7;
+              ++byte_count;
+          } while ((byte & 0x40) == 0 ? value != 0 : value != -1);
+      }
+      write_little_endian(stream, byte_count);
+      const std::uint32_t BUF_SIZE = 4096;
+      std::uint8_t buf[BUF_SIZE];
+      std::uint32_t buf_pos = 0;
+      auto flush = [&]() {
+          if (buf_pos > 0) {
+              stream.write(reinterpret_cast<char*>(buf), buf_pos);
+              buf_pos = 0;
+          }
+      };
+      auto write = [&](std::uint8_t byte) {
+          buf[buf_pos++] = byte;
+          if (buf_pos == BUF_SIZE) flush();
+      };
+      for (std::size_t i = 0; i < count; ++i) {
+          IntType value = values[i];
+          while (true) {
+              std::uint8_t byte = value & 0x7f;
+              value >>= 7;
+              if ((byte & 0x40) == 0 ? value == 0 : value == -1) {
+                  write(byte);
+                  break;
+              }
+              write(byte | 0x80);
+          }
+      }
+      flush();
+  }
+
 }  // namespace Stockfish::Eval::NNUE
 
 #endif // #ifndef NNUE_COMMON_H_INCLUDED

src/nnue/nnue_feature_transformer.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -25,6 +25,7 @@
 #include "nnue_architecture.h"
 
 #include <cstring> // std::memset()
+#include <utility> // std::pair
 
 namespace Stockfish::Eval::NNUE {
 
@@ -41,8 +42,8 @@ namespace Stockfish::Eval::NNUE {
     "Per feature PSQT values cannot be processed at granularity lower than 8 at a time.");
 
   #ifdef USE_AVX512
-  typedef __m512i vec_t;
+  using vec_t = __m512i;
-  typedef __m256i psqt_vec_t;
+  using psqt_vec_t = __m256i;
   #define vec_load(a) _mm512_load_si512(a)
   #define vec_store(a,b) _mm512_store_si512(a,b)
   #define vec_add_16(a,b) _mm512_add_epi16(a,b)
@@ -65,8 +66,8 @@ namespace Stockfish::Eval::NNUE {
   #define MaxChunkSize 64
 
   #elif USE_AVX2
-  typedef __m256i vec_t;
+  using vec_t = __m256i;
-  typedef __m256i psqt_vec_t;
+  using psqt_vec_t = __m256i;
   #define vec_load(a) _mm256_load_si256(a)
   #define vec_store(a,b) _mm256_store_si256(a,b)
   #define vec_add_16(a,b) _mm256_add_epi16(a,b)
@@ -89,8 +90,8 @@ namespace Stockfish::Eval::NNUE {
   #define MaxChunkSize 32
 
   #elif USE_SSE2
-  typedef __m128i vec_t;
+  using vec_t = __m128i;
-  typedef __m128i psqt_vec_t;
+  using psqt_vec_t = __m128i;
   #define vec_load(a) (*(a))
   #define vec_store(a,b) *(a)=(b)
   #define vec_add_16(a,b) _mm_add_epi16(a,b)
@@ -110,8 +111,8 @@ namespace Stockfish::Eval::NNUE {
   #define MaxChunkSize 16
 
   #elif USE_MMX
-  typedef __m64 vec_t;
+  using vec_t = __m64;
-  typedef __m64 psqt_vec_t;
+  using psqt_vec_t = __m64;
   #define vec_load(a) (*(a))
   #define vec_store(a,b) *(a)=(b)
   #define vec_add_16(a,b) _mm_add_pi16(a,b)
@@ -138,8 +139,8 @@ namespace Stockfish::Eval::NNUE {
   #define MaxChunkSize 8
 
   #elif USE_NEON
-  typedef int16x8_t vec_t;
+  using vec_t = int16x8_t;
-  typedef int32x4_t psqt_vec_t;
+  using psqt_vec_t = int32x4_t;
   #define vec_load(a) (*(a))
   #define vec_store(a,b) *(a)=(b)
   #define vec_add_16(a,b) vaddq_s16(a,b)
@@ -252,9 +253,9 @@ namespace Stockfish::Eval::NNUE {
     // Read network parameters
     bool read_parameters(std::istream& stream) {
 
-      read_little_endian<BiasType      >(stream, biases     , HalfDimensions                  );
+      read_leb_128<BiasType      >(stream, biases     , HalfDimensions                  );
-      read_little_endian<WeightType    >(stream, weights    , HalfDimensions * InputDimensions);
+      read_leb_128<WeightType    >(stream, weights    , HalfDimensions * InputDimensions);
-      read_little_endian<PSQTWeightType>(stream, psqtWeights, PSQTBuckets    * InputDimensions);
+      read_leb_128<PSQTWeightType>(stream, psqtWeights, PSQTBuckets    * InputDimensions);
 
       return !stream.fail();
     }
@@ -262,9 +263,9 @@ namespace Stockfish::Eval::NNUE {
     // Write network parameters
     bool write_parameters(std::ostream& stream) const {
 
-      write_little_endian<BiasType      >(stream, biases     , HalfDimensions                  );
+      write_leb_128<BiasType      >(stream, biases     , HalfDimensions                  );
-      write_little_endian<WeightType    >(stream, weights    , HalfDimensions * InputDimensions);
+      write_leb_128<WeightType    >(stream, weights    , HalfDimensions * InputDimensions);
-      write_little_endian<PSQTWeightType>(stream, psqtWeights, PSQTBuckets    * InputDimensions);
+      write_leb_128<PSQTWeightType>(stream, psqtWeights, PSQTBuckets    * InputDimensions);
 
       return !stream.fail();
     }
@@ -332,27 +333,16 @@ namespace Stockfish::Eval::NNUE {
 #endif
 
       return psqt;
+    } // end of function transform()
 
-   } // end of function transform()
+    void hint_common_access(const Position& pos) const {
-
+      hint_common_access_for_perspective<WHITE>(pos);
-
+      hint_common_access_for_perspective<BLACK>(pos);
+    }
 
    private:
     template<Color Perspective>
-    void update_accumulator(const Position& pos) const {
+    [[nodiscard]] std::pair<StateInfo*, StateInfo*> try_find_computed_accumulator(const Position& pos) const {
-
-      // The size must be enough to contain the largest possible update.
-      // That might depend on the feature set and generally relies on the
-      // feature set's update cost calculation to be correct and never
-      // allow updates with more added/removed features than MaxActiveDimensions.
-
-  #ifdef VECTOR
-      // Gcc-10.2 unnecessarily spills AVX2 registers if this array
-      // is defined in the VECTOR code below, once in each branch
-      vec_t acc[NumRegs];
-      psqt_vec_t psqt[NumPsqtRegs];
-  #endif
-
       // Look for a usable accumulator of an earlier position. We keep track
       // of the estimated gain in terms of features to be added/subtracted.
       StateInfo *st = pos.state(), *next = nullptr;
@@ -367,218 +357,313 @@ namespace Stockfish::Eval::NNUE {
         next = st;
         st = st->previous;
       }
+      return { st, next };
+    }
 
-      if (st->accumulator.computed[Perspective])
+    // NOTE: The parameter states_to_update is an array of position states, ending with nullptr.
-      {
+    //       All states must be sequential, that is states_to_update[i] must either be reachable
-        if (next == nullptr)
+    //       by repeatedly applying ->previous from states_to_update[i+1] or states_to_update[i] == nullptr.
-          return;
+    //       computed_st must be reachable by repeatedly applying ->previous on states_to_update[0], if not nullptr.
+    template<Color Perspective, size_t N>
+    void update_accumulator_incremental(const Position& pos, StateInfo* computed_st, StateInfo* states_to_update[N]) const {
+      static_assert(N > 0);
+      assert(states_to_update[N-1] == nullptr);
 
-        // Update incrementally in two steps. First, we update the "next"
-        // accumulator. Then, we update the current accumulator (pos.state()).
-
-        // Gather all features to be updated.
-        const Square ksq = pos.square<KING>(Perspective);
-        FeatureSet::IndexList removed[2], added[2];
-        FeatureSet::append_changed_indices<Perspective>(
-          ksq, next->dirtyPiece, removed[0], added[0]);
-        for (StateInfo *st2 = pos.state(); st2 != next; st2 = st2->previous)
-          FeatureSet::append_changed_indices<Perspective>(
-            ksq, st2->dirtyPiece, removed[1], added[1]);
-
-        // Mark the accumulators as computed.
-        next->accumulator.computed[Perspective] = true;
-        pos.state()->accumulator.computed[Perspective] = true;
-
-        // Now update the accumulators listed in states_to_update[], where the last element is a sentinel.
-        StateInfo *states_to_update[3] =
-          { next, next == pos.state() ? nullptr : pos.state(), nullptr };
   #ifdef VECTOR
-        for (IndexType j = 0; j < HalfDimensions / TileHeight; ++j)
+      // Gcc-10.2 unnecessarily spills AVX2 registers if this array
+      // is defined in the VECTOR code below, once in each branch
+      vec_t acc[NumRegs];
+      psqt_vec_t psqt[NumPsqtRegs];
+  #endif
+
+      if (states_to_update[0] == nullptr)
+        return;
+
+      // Update incrementally going back through states_to_update.
+
+      // Gather all features to be updated.
+      const Square ksq = pos.square<KING>(Perspective);
+
+      // The size must be enough to contain the largest possible update.
+      // That might depend on the feature set and generally relies on the
+      // feature set's update cost calculation to be correct and never
+      // allow updates with more added/removed features than MaxActiveDimensions.
+      FeatureSet::IndexList removed[N-1], added[N-1];
+
+      {
+        int i = N-2; // last potential state to update. Skip last element because it must be nullptr.
+        while (states_to_update[i] == nullptr)
+          --i;
+
+        StateInfo *st2 = states_to_update[i];
+
+        for (; i >= 0; --i)
         {
-          // Load accumulator
+          states_to_update[i]->accumulator.computed[Perspective] = true;
-          auto accTile = reinterpret_cast<vec_t*>(
-            &st->accumulator.accumulation[Perspective][j * TileHeight]);
-          for (IndexType k = 0; k < NumRegs; ++k)
-            acc[k] = vec_load(&accTile[k]);
 
-          for (IndexType i = 0; states_to_update[i]; ++i)
+          StateInfo* end_state = i == 0 ? computed_st : states_to_update[i - 1];
-          {
-            // Difference calculation for the deactivated features
-            for (const auto index : removed[i])
-            {
-              const IndexType offset = HalfDimensions * index + j * TileHeight;
-              auto column = reinterpret_cast<const vec_t*>(&weights[offset]);
-              for (IndexType k = 0; k < NumRegs; ++k)
-                acc[k] = vec_sub_16(acc[k], column[k]);
-            }
 
-            // Difference calculation for the activated features
+          for (; st2 != end_state; st2 = st2->previous)
-            for (const auto index : added[i])
+            FeatureSet::append_changed_indices<Perspective>(
-            {
+              ksq, st2->dirtyPiece, removed[i], added[i]);
-              const IndexType offset = HalfDimensions * index + j * TileHeight;
-              auto column = reinterpret_cast<const vec_t*>(&weights[offset]);
-              for (IndexType k = 0; k < NumRegs; ++k)
-                acc[k] = vec_add_16(acc[k], column[k]);
-            }
-
-            // Store accumulator
-            accTile = reinterpret_cast<vec_t*>(
-              &states_to_update[i]->accumulator.accumulation[Perspective][j * TileHeight]);
-            for (IndexType k = 0; k < NumRegs; ++k)
-              vec_store(&accTile[k], acc[k]);
-          }
         }
+      }
 
-        for (IndexType j = 0; j < PSQTBuckets / PsqtTileHeight; ++j)
+      StateInfo* st = computed_st;
-        {
-          // Load accumulator
-          auto accTilePsqt = reinterpret_cast<psqt_vec_t*>(
-            &st->accumulator.psqtAccumulation[Perspective][j * PsqtTileHeight]);
-          for (std::size_t k = 0; k < NumPsqtRegs; ++k)
-            psqt[k] = vec_load_psqt(&accTilePsqt[k]);
 
-          for (IndexType i = 0; states_to_update[i]; ++i)
+      // Now update the accumulators listed in states_to_update[], where the last element is a sentinel.
-          {
+#ifdef VECTOR
-            // Difference calculation for the deactivated features
+      for (IndexType j = 0; j < HalfDimensions / TileHeight; ++j)
-            for (const auto index : removed[i])
+      {
-            {
+        // Load accumulator
-              const IndexType offset = PSQTBuckets * index + j * PsqtTileHeight;
+        auto accTile = reinterpret_cast<vec_t*>(
-              auto columnPsqt = reinterpret_cast<const psqt_vec_t*>(&psqtWeights[offset]);
+          &st->accumulator.accumulation[Perspective][j * TileHeight]);
-              for (std::size_t k = 0; k < NumPsqtRegs; ++k)
+        for (IndexType k = 0; k < NumRegs; ++k)
-                psqt[k] = vec_sub_psqt_32(psqt[k], columnPsqt[k]);
+          acc[k] = vec_load(&accTile[k]);
-            }
 
-            // Difference calculation for the activated features
-            for (const auto index : added[i])
-            {
-              const IndexType offset = PSQTBuckets * index + j * PsqtTileHeight;
-              auto columnPsqt = reinterpret_cast<const psqt_vec_t*>(&psqtWeights[offset]);
-              for (std::size_t k = 0; k < NumPsqtRegs; ++k)
-                psqt[k] = vec_add_psqt_32(psqt[k], columnPsqt[k]);
-            }
-
-            // Store accumulator
-            accTilePsqt = reinterpret_cast<psqt_vec_t*>(
-              &states_to_update[i]->accumulator.psqtAccumulation[Perspective][j * PsqtTileHeight]);
-            for (std::size_t k = 0; k < NumPsqtRegs; ++k)
-              vec_store_psqt(&accTilePsqt[k], psqt[k]);
-          }
-        }
-
-  #else
         for (IndexType i = 0; states_to_update[i]; ++i)
         {
-          std::memcpy(states_to_update[i]->accumulator.accumulation[Perspective],
-              st->accumulator.accumulation[Perspective],
-              HalfDimensions * sizeof(BiasType));
-
-          for (std::size_t k = 0; k < PSQTBuckets; ++k)
-            states_to_update[i]->accumulator.psqtAccumulation[Perspective][k] = st->accumulator.psqtAccumulation[Perspective][k];
-
-          st = states_to_update[i];
-
           // Difference calculation for the deactivated features
           for (const auto index : removed[i])
           {
-            const IndexType offset = HalfDimensions * index;
+            const IndexType offset = HalfDimensions * index + j * TileHeight;
-
+            auto column = reinterpret_cast<const vec_t*>(&weights[offset]);
-            for (IndexType j = 0; j < HalfDimensions; ++j)
+            for (IndexType k = 0; k < NumRegs; ++k)
-              st->accumulator.accumulation[Perspective][j] -= weights[offset + j];
+              acc[k] = vec_sub_16(acc[k], column[k]);
-
-            for (std::size_t k = 0; k < PSQTBuckets; ++k)
-              st->accumulator.psqtAccumulation[Perspective][k] -= psqtWeights[index * PSQTBuckets + k];
           }
 
           // Difference calculation for the activated features
           for (const auto index : added[i])
-          {
-            const IndexType offset = HalfDimensions * index;
-
-            for (IndexType j = 0; j < HalfDimensions; ++j)
-              st->accumulator.accumulation[Perspective][j] += weights[offset + j];
-
-            for (std::size_t k = 0; k < PSQTBuckets; ++k)
-              st->accumulator.psqtAccumulation[Perspective][k] += psqtWeights[index * PSQTBuckets + k];
-          }
-        }
-  #endif
-      }
-      else
-      {
-        // Refresh the accumulator
-        auto& accumulator = pos.state()->accumulator;
-        accumulator.computed[Perspective] = true;
-        FeatureSet::IndexList active;
-        FeatureSet::append_active_indices<Perspective>(pos, active);
-
-  #ifdef VECTOR
-        for (IndexType j = 0; j < HalfDimensions / TileHeight; ++j)
-        {
-          auto biasesTile = reinterpret_cast<const vec_t*>(
-              &biases[j * TileHeight]);
-          for (IndexType k = 0; k < NumRegs; ++k)
-            acc[k] = biasesTile[k];
-
-          for (const auto index : active)
           {
             const IndexType offset = HalfDimensions * index + j * TileHeight;
             auto column = reinterpret_cast<const vec_t*>(&weights[offset]);
-
+            for (IndexType k = 0; k < NumRegs; ++k)
-            for (unsigned k = 0; k < NumRegs; ++k)
               acc[k] = vec_add_16(acc[k], column[k]);
           }
 
-          auto accTile = reinterpret_cast<vec_t*>(
+          // Store accumulator
-              &accumulator.accumulation[Perspective][j * TileHeight]);
+          accTile = reinterpret_cast<vec_t*>(
-          for (unsigned k = 0; k < NumRegs; k++)
+            &states_to_update[i]->accumulator.accumulation[Perspective][j * TileHeight]);
+          for (IndexType k = 0; k < NumRegs; ++k)
             vec_store(&accTile[k], acc[k]);
         }
+      }
 
-        for (IndexType j = 0; j < PSQTBuckets / PsqtTileHeight; ++j)
+      for (IndexType j = 0; j < PSQTBuckets / PsqtTileHeight; ++j)
+      {
+        // Load accumulator
+        auto accTilePsqt = reinterpret_cast<psqt_vec_t*>(
+          &st->accumulator.psqtAccumulation[Perspective][j * PsqtTileHeight]);
+        for (std::size_t k = 0; k < NumPsqtRegs; ++k)
+          psqt[k] = vec_load_psqt(&accTilePsqt[k]);
+
+        for (IndexType i = 0; states_to_update[i]; ++i)
         {
-          for (std::size_t k = 0; k < NumPsqtRegs; ++k)
+          // Difference calculation for the deactivated features
-            psqt[k] = vec_zero_psqt();
+          for (const auto index : removed[i])
-
-          for (const auto index : active)
           {
             const IndexType offset = PSQTBuckets * index + j * PsqtTileHeight;
             auto columnPsqt = reinterpret_cast<const psqt_vec_t*>(&psqtWeights[offset]);
+            for (std::size_t k = 0; k < NumPsqtRegs; ++k)
+              psqt[k] = vec_sub_psqt_32(psqt[k], columnPsqt[k]);
+          }
 
+          // Difference calculation for the activated features
+          for (const auto index : added[i])
+          {
+            const IndexType offset = PSQTBuckets * index + j * PsqtTileHeight;
+            auto columnPsqt = reinterpret_cast<const psqt_vec_t*>(&psqtWeights[offset]);
             for (std::size_t k = 0; k < NumPsqtRegs; ++k)
               psqt[k] = vec_add_psqt_32(psqt[k], columnPsqt[k]);
           }
 
-          auto accTilePsqt = reinterpret_cast<psqt_vec_t*>(
+          // Store accumulator
-            &accumulator.psqtAccumulation[Perspective][j * PsqtTileHeight]);
+          accTilePsqt = reinterpret_cast<psqt_vec_t*>(
+            &states_to_update[i]->accumulator.psqtAccumulation[Perspective][j * PsqtTileHeight]);
           for (std::size_t k = 0; k < NumPsqtRegs; ++k)
             vec_store_psqt(&accTilePsqt[k], psqt[k]);
         }
+      }
 
-  #else
+#else
-        std::memcpy(accumulator.accumulation[Perspective], biases,
+      for (IndexType i = 0; states_to_update[i]; ++i)
+      {
+        std::memcpy(states_to_update[i]->accumulator.accumulation[Perspective],
+            st->accumulator.accumulation[Perspective],
             HalfDimensions * sizeof(BiasType));
 
         for (std::size_t k = 0; k < PSQTBuckets; ++k)
-          accumulator.psqtAccumulation[Perspective][k] = 0;
+          states_to_update[i]->accumulator.psqtAccumulation[Perspective][k] = st->accumulator.psqtAccumulation[Perspective][k];
 
-        for (const auto index : active)
+        st = states_to_update[i];
+
+        // Difference calculation for the deactivated features
+        for (const auto index : removed[i])
         {
           const IndexType offset = HalfDimensions * index;
 
           for (IndexType j = 0; j < HalfDimensions; ++j)
-            accumulator.accumulation[Perspective][j] += weights[offset + j];
+            st->accumulator.accumulation[Perspective][j] -= weights[offset + j];
 
           for (std::size_t k = 0; k < PSQTBuckets; ++k)
-            accumulator.psqtAccumulation[Perspective][k] += psqtWeights[index * PSQTBuckets + k];
+            st->accumulator.psqtAccumulation[Perspective][k] -= psqtWeights[index * PSQTBuckets + k];
+        }
+
+        // Difference calculation for the activated features
+        for (const auto index : added[i])
+        {
+          const IndexType offset = HalfDimensions * index;
+
+          for (IndexType j = 0; j < HalfDimensions; ++j)
+            st->accumulator.accumulation[Perspective][j] += weights[offset + j];
+
+          for (std::size_t k = 0; k < PSQTBuckets; ++k)
+            st->accumulator.psqtAccumulation[Perspective][k] += psqtWeights[index * PSQTBuckets + k];
         }
-  #endif
       }
+#endif
 
   #if defined(USE_MMX)
       _mm_empty();
   #endif
     }
 
+    template<Color Perspective>
+    void update_accumulator_refresh(const Position& pos) const {
+  #ifdef VECTOR
+      // Gcc-10.2 unnecessarily spills AVX2 registers if this array
+      // is defined in the VECTOR code below, once in each branch
+      vec_t acc[NumRegs];
+      psqt_vec_t psqt[NumPsqtRegs];
+  #endif
+
+      // Refresh the accumulator
+      // Could be extracted to a separate function because it's done in 2 places,
+      // but it's unclear if compilers would correctly handle register allocation.
+      auto& accumulator = pos.state()->accumulator;
+      accumulator.computed[Perspective] = true;
+      FeatureSet::IndexList active;
+      FeatureSet::append_active_indices<Perspective>(pos, active);
+
+#ifdef VECTOR
+      for (IndexType j = 0; j < HalfDimensions / TileHeight; ++j)
+      {
+        auto biasesTile = reinterpret_cast<const vec_t*>(
+            &biases[j * TileHeight]);
+        for (IndexType k = 0; k < NumRegs; ++k)
+          acc[k] = biasesTile[k];
+
+        for (const auto index : active)
+        {
+          const IndexType offset = HalfDimensions * index + j * TileHeight;
+          auto column = reinterpret_cast<const vec_t*>(&weights[offset]);
+
+          for (unsigned k = 0; k < NumRegs; ++k)
+            acc[k] = vec_add_16(acc[k], column[k]);
+        }
+
+        auto accTile = reinterpret_cast<vec_t*>(
+            &accumulator.accumulation[Perspective][j * TileHeight]);
+        for (unsigned k = 0; k < NumRegs; k++)
+          vec_store(&accTile[k], acc[k]);
+      }
+
+      for (IndexType j = 0; j < PSQTBuckets / PsqtTileHeight; ++j)
+      {
+        for (std::size_t k = 0; k < NumPsqtRegs; ++k)
+          psqt[k] = vec_zero_psqt();
+
+        for (const auto index : active)
+        {
+          const IndexType offset = PSQTBuckets * index + j * PsqtTileHeight;
+          auto columnPsqt = reinterpret_cast<const psqt_vec_t*>(&psqtWeights[offset]);
+
+          for (std::size_t k = 0; k < NumPsqtRegs; ++k)
+            psqt[k] = vec_add_psqt_32(psqt[k], columnPsqt[k]);
+        }
+
+        auto accTilePsqt = reinterpret_cast<psqt_vec_t*>(
+          &accumulator.psqtAccumulation[Perspective][j * PsqtTileHeight]);
+        for (std::size_t k = 0; k < NumPsqtRegs; ++k)
+          vec_store_psqt(&accTilePsqt[k], psqt[k]);
+      }
+
+#else
+      std::memcpy(accumulator.accumulation[Perspective], biases,
+          HalfDimensions * sizeof(BiasType));
+
+      for (std::size_t k = 0; k < PSQTBuckets; ++k)
+        accumulator.psqtAccumulation[Perspective][k] = 0;
+
+      for (const auto index : active)
+      {
+        const IndexType offset = HalfDimensions * index;
+
+        for (IndexType j = 0; j < HalfDimensions; ++j)
+          accumulator.accumulation[Perspective][j] += weights[offset + j];
+
+        for (std::size_t k = 0; k < PSQTBuckets; ++k)
+          accumulator.psqtAccumulation[Perspective][k] += psqtWeights[index * PSQTBuckets + k];
+      }
+#endif
+
+  #if defined(USE_MMX)
+      _mm_empty();
+  #endif
+    }
+
+    template<Color Perspective>
+    void hint_common_access_for_perspective(const Position& pos) const {
+
+      // Works like update_accumulator, but performs less work.
+      // Updates ONLY the accumulator for pos.
+
+      // Look for a usable accumulator of an earlier position. We keep track
+      // of the estimated gain in terms of features to be added/subtracted.
+      // Fast early exit.
+      if (pos.state()->accumulator.computed[Perspective])
+        return;
+
+      auto [oldest_st, _] = try_find_computed_accumulator<Perspective>(pos);
+
+      if (oldest_st->accumulator.computed[Perspective])
+      {
+        // Only update current position accumulator to minimize work.
+        StateInfo* states_to_update[2] = { pos.state(), nullptr };
+        update_accumulator_incremental<Perspective, 2>(pos, oldest_st, states_to_update);
+      }
+      else
+      {
+        update_accumulator_refresh<Perspective>(pos);
+      }
+    }
+
+    template<Color Perspective>
+    void update_accumulator(const Position& pos) const {
+
+      auto [oldest_st, next] = try_find_computed_accumulator<Perspective>(pos);
+
+      if (oldest_st->accumulator.computed[Perspective])
+      {
+        if (next == nullptr)
+          return;
+
+        // Now update the accumulators listed in states_to_update[], where the last element is a sentinel.
+        // Currently we update 2 accumulators.
+        //     1. for the current position
+        //     2. the next accumulator after the computed one
+        // The heuristic may change in the future.
+        StateInfo *states_to_update[3] =
+          { next, next == pos.state() ? nullptr : pos.state(), nullptr };
+
+        update_accumulator_incremental<Perspective, 3>(pos, oldest_st, states_to_update);
+      }
+      else
+      {
+        update_accumulator_refresh<Perspective>(pos);
+      }
+    }
+
     alignas(CacheLineSize) BiasType biases[HalfDimensions];
     alignas(CacheLineSize) WeightType weights[HalfDimensions * InputDimensions];
     alignas(CacheLineSize) PSQTWeightType psqtWeights[InputDimensions * PSQTBuckets];

src/pawns.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/pawns.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -61,7 +61,7 @@ struct Entry {
   int blockedCount;
 };
 
-typedef HashTable<Entry, 131072> Table;
+using Table = HashTable<Entry, 131072>;
 
 Entry* probe(const Position& pos);
 

src/position.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -22,6 +22,7 @@
 #include <cstring> // For std::memset, std::memcmp
 #include <iomanip>
 #include <sstream>
+#include <string_view>
 
 #include "bitboard.h"
 #include "misc.h"
@@ -46,7 +47,7 @@ namespace Zobrist {
 
 namespace {
 
-const string PieceToChar(" PNBRQK  pnbrqk");
+constexpr std::string_view PieceToChar(" PNBRQK  pnbrqk");
 
 constexpr Piece Pieces[] = { W_PAWN, W_KNIGHT, W_BISHOP, W_ROOK, W_QUEEN, W_KING,
                              B_PAWN, B_KNIGHT, B_BISHOP, B_ROOK, B_QUEEN, B_KING };
@@ -96,7 +97,7 @@ std::ostream& operator<<(std::ostream& os, const Position& pos) {
 
 // Marcel van Kervinck's cuckoo algorithm for fast detection of "upcoming repetition"
 // situations. Description of the algorithm in the following paper:
-// https://marcelk.net/2013-04-06/paper/upcoming-rep-v2.pdf
+// http://web.archive.org/web/20201107002606/https://marcelk.net/2013-04-06/paper/upcoming-rep-v2.pdf
 
 // First and second hash functions for indexing the cuckoo tables
 inline int H1(Key h) { return h & 0x1fff; }
@@ -281,7 +282,7 @@ Position& Position::set(const string& fenStr, bool isChess960, StateInfo* si, Th
 
   chess960 = isChess960;
   thisThread = th;
-  set_state(st);
+  set_state();
 
   assert(pos_is_ok());
 
@@ -312,60 +313,59 @@ void Position::set_castling_right(Color c, Square rfrom) {
 
 /// Position::set_check_info() sets king attacks to detect if a move gives check
 
-void Position::set_check_info(StateInfo* si) const {
+void Position::set_check_info() const {
 
-  si->blockersForKing[WHITE] = slider_blockers(pieces(BLACK), square<KING>(WHITE), si->pinners[BLACK]);
+  st->blockersForKing[WHITE] = slider_blockers(pieces(BLACK), square<KING>(WHITE), st->pinners[BLACK]);
-  si->blockersForKing[BLACK] = slider_blockers(pieces(WHITE), square<KING>(BLACK), si->pinners[WHITE]);
+  st->blockersForKing[BLACK] = slider_blockers(pieces(WHITE), square<KING>(BLACK), st->pinners[WHITE]);
 
   Square ksq = square<KING>(~sideToMove);
 
-  si->checkSquares[PAWN]   = pawn_attacks_bb(~sideToMove, ksq);
+  st->checkSquares[PAWN]   = pawn_attacks_bb(~sideToMove, ksq);
-  si->checkSquares[KNIGHT] = attacks_bb<KNIGHT>(ksq);
+  st->checkSquares[KNIGHT] = attacks_bb<KNIGHT>(ksq);
-  si->checkSquares[BISHOP] = attacks_bb<BISHOP>(ksq, pieces());
+  st->checkSquares[BISHOP] = attacks_bb<BISHOP>(ksq, pieces());
-  si->checkSquares[ROOK]   = attacks_bb<ROOK>(ksq, pieces());
+  st->checkSquares[ROOK]   = attacks_bb<ROOK>(ksq, pieces());
-  si->checkSquares[QUEEN]  = si->checkSquares[BISHOP] | si->checkSquares[ROOK];
+  st->checkSquares[QUEEN]  = st->checkSquares[BISHOP] | st->checkSquares[ROOK];
-  si->checkSquares[KING]   = 0;
+  st->checkSquares[KING]   = 0;
 }
 
 
 /// Position::set_state() computes the hash keys of the position, and other
 /// data that once computed is updated incrementally as moves are made.
-/// The function is only used when a new position is set up, and to verify
+/// The function is only used when a new position is set up
-/// the correctness of the StateInfo data when running in debug mode.
 
-void Position::set_state(StateInfo* si) const {
+void Position::set_state() const {
 
-  si->key = si->materialKey = 0;
+  st->key = st->materialKey = 0;
-  si->pawnKey = Zobrist::noPawns;
+  st->pawnKey = Zobrist::noPawns;
-  si->nonPawnMaterial[WHITE] = si->nonPawnMaterial[BLACK] = VALUE_ZERO;
+  st->nonPawnMaterial[WHITE] = st->nonPawnMaterial[BLACK] = VALUE_ZERO;
-  si->checkersBB = attackers_to(square<KING>(sideToMove)) & pieces(~sideToMove);
+  st->checkersBB = attackers_to(square<KING>(sideToMove)) & pieces(~sideToMove);
 
-  set_check_info(si);
+  set_check_info();
 
   for (Bitboard b = pieces(); b; )
   {
       Square s = pop_lsb(b);
       Piece pc = piece_on(s);
-      si->key ^= Zobrist::psq[pc][s];
+      st->key ^= Zobrist::psq[pc][s];
 
       if (type_of(pc) == PAWN)
-          si->pawnKey ^= Zobrist::psq[pc][s];
+          st->pawnKey ^= Zobrist::psq[pc][s];
 
       else if (type_of(pc) != KING)
-          si->nonPawnMaterial[color_of(pc)] += PieceValue[MG][pc];
+          st->nonPawnMaterial[color_of(pc)] += PieceValue[MG][pc];
   }
 
-  if (si->epSquare != SQ_NONE)
+  if (st->epSquare != SQ_NONE)
-      si->key ^= Zobrist::enpassant[file_of(si->epSquare)];
+      st->key ^= Zobrist::enpassant[file_of(st->epSquare)];
 
   if (sideToMove == BLACK)
-      si->key ^= Zobrist::side;
+      st->key ^= Zobrist::side;
 
-  si->key ^= Zobrist::castling[si->castlingRights];
+  st->key ^= Zobrist::castling[st->castlingRights];
 
   for (Piece pc : Pieces)
       for (int cnt = 0; cnt < pieceCount[pc]; ++cnt)
-          si->materialKey ^= Zobrist::psq[pc][cnt];
+          st->materialKey ^= Zobrist::psq[pc][cnt];
 }
 
 
@@ -568,8 +568,7 @@ bool Position::pseudo_legal(const Move m) const {
                         : MoveList<NON_EVASIONS>(*this).contains(m);
 
   // Is not a promotion, so promotion piece must be empty
-  if (promotion_type(m) - KNIGHT != NO_PIECE_TYPE)
+  assert(promotion_type(m) - KNIGHT == NO_PIECE_TYPE);
-      return false;
 
   // If the 'from' square is not occupied by a piece belonging to the side to
   // move, the move is obviously not legal.
@@ -765,9 +764,6 @@ void Position::do_move(Move m, StateInfo& newSt, bool givesCheck) {
       // Update board and piece lists
       remove_piece(capsq);
 
-      if (type_of(m) == EN_PASSANT)
-          board[capsq] = NO_PIECE;
-
       // Update material hash key and prefetch access to materialTable
       k ^= Zobrist::psq[captured][capsq];
       st->materialKey ^= Zobrist::psq[captured][pieceCount[captured]];
@@ -868,7 +864,7 @@ void Position::do_move(Move m, StateInfo& newSt, bool givesCheck) {
   sideToMove = ~sideToMove;
 
   // Update king attacks used for fast check detection
-  set_check_info(st);
+  set_check_info();
 
   // Calculate the repetition info. It is the ply distance from the previous
   // occurrence of the same position, negative in the 3-fold case, or zero
@@ -1020,7 +1016,7 @@ void Position::do_null_move(StateInfo& newSt) {
 
   sideToMove = ~sideToMove;
 
-  set_check_info(st);
+  set_check_info();
 
   st->repetition = 0;
 
@@ -1065,7 +1061,7 @@ Key Position::key_after(Move m) const {
 /// SEE value of move is greater or equal to the given threshold. We'll use an
 /// algorithm similar to alpha-beta pruning with a null window.
 
-bool Position::see_ge(Move m, Value threshold) const {
+bool Position::see_ge(Move m, Bitboard& occupied, Value threshold) const {
 
   assert(is_ok(m));
 
@@ -1084,7 +1080,7 @@ bool Position::see_ge(Move m, Value threshold) const {
       return true;
 
   assert(color_of(piece_on(from)) == sideToMove);
-  Bitboard occupied = pieces() ^ from ^ to;
+  occupied = pieces() ^ from ^ to; // xoring to is important for pinned piece logic
   Color stm = sideToMove;
   Bitboard attackers = attackers_to(to, occupied);
   Bitboard stmAttackers, bb;
@@ -1115,45 +1111,44 @@ bool Position::see_ge(Move m, Value threshold) const {
       // the bitboard 'attackers' any X-ray attackers behind it.
       if ((bb = stmAttackers & pieces(PAWN)))
       {
+          occupied ^= least_significant_square_bb(bb);
           if ((swap = PawnValueMg - swap) < res)
               break;
 
-          occupied ^= least_significant_square_bb(bb);
           attackers |= attacks_bb<BISHOP>(to, occupied) & pieces(BISHOP, QUEEN);
       }
 
       else if ((bb = stmAttackers & pieces(KNIGHT)))
       {
+          occupied ^= least_significant_square_bb(bb);
           if ((swap = KnightValueMg - swap) < res)
               break;
-
-          occupied ^= least_significant_square_bb(bb);
       }
 
       else if ((bb = stmAttackers & pieces(BISHOP)))
       {
+          occupied ^= least_significant_square_bb(bb);
           if ((swap = BishopValueMg - swap) < res)
               break;
 
-          occupied ^= least_significant_square_bb(bb);
           attackers |= attacks_bb<BISHOP>(to, occupied) & pieces(BISHOP, QUEEN);
       }
 
       else if ((bb = stmAttackers & pieces(ROOK)))
       {
+          occupied ^= least_significant_square_bb(bb);
           if ((swap = RookValueMg - swap) < res)
               break;
 
-          occupied ^= least_significant_square_bb(bb);
           attackers |= attacks_bb<ROOK>(to, occupied) & pieces(ROOK, QUEEN);
       }
 
       else if ((bb = stmAttackers & pieces(QUEEN)))
       {
+          occupied ^= least_significant_square_bb(bb);
           if ((swap = QueenValueMg - swap) < res)
               break;
 
-          occupied ^= least_significant_square_bb(bb);
           attackers |=  (attacks_bb<BISHOP>(to, occupied) & pieces(BISHOP, QUEEN))
                       | (attacks_bb<ROOK  >(to, occupied) & pieces(ROOK  , QUEEN));
       }
@@ -1167,6 +1162,11 @@ bool Position::see_ge(Move m, Value threshold) const {
   return bool(res);
 }
 
+bool Position::see_ge(Move m, Value threshold) const {
+    Bitboard occupied;
+    return see_ge(m, occupied, threshold);
+}
+
 
 /// Position::is_draw() tests whether the position is drawn by 50-move rule
 /// or by repetition. It does not detect stalemates.
@@ -1323,12 +1323,6 @@ bool Position::pos_is_ok() const {
           if (p1 != p2 && (pieces(p1) & pieces(p2)))
               assert(0 && "pos_is_ok: Bitboards");
 
-  StateInfo si = *st;
-  ASSERT_ALIGNED(&si, Eval::NNUE::CacheLineSize);
-
-  set_state(&si);
-  if (std::memcmp(&si, st, sizeof(StateInfo)))
-      assert(0 && "pos_is_ok: State");
 
   for (Piece pc : Pieces)
       if (   pieceCount[pc] != popcount(pieces(color_of(pc), type_of(pc)))

src/position.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -68,7 +68,7 @@ struct StateInfo {
 /// start position to the position just before the search starts). Needed by
 /// 'draw by repetition' detection. Use a std::deque because pointers to
 /// elements are not invalidated upon list resizing.
-typedef std::unique_ptr<std::deque<StateInfo>> StateListPtr;
+using StateListPtr = std::unique_ptr<std::deque<StateInfo>>;
 
 
 /// Position class stores information regarding the board representation as
@@ -92,10 +92,9 @@ public:
 
   // Position representation
   Bitboard pieces(PieceType pt) const;
-  Bitboard pieces(PieceType pt1, PieceType pt2) const;
+  template<typename ...PieceTypes> Bitboard pieces(PieceType pt, PieceTypes... pts) const;
   Bitboard pieces(Color c) const;
-  Bitboard pieces(Color c, PieceType pt) const;
+  template<typename ...PieceTypes> Bitboard pieces(Color c, PieceTypes... pts) const;
-  Bitboard pieces(Color c, PieceType pt1, PieceType pt2) const;
   Piece piece_on(Square s) const;
   Square ep_square() const;
   bool empty(Square s) const;
@@ -126,6 +125,7 @@ public:
   bool legal(Move m) const;
   bool pseudo_legal(const Move m) const;
   bool capture(Move m) const;
+  bool capture_stage(Move m) const;
   bool gives_check(Move m) const;
   Piece moved_piece(Move m) const;
   Piece captured_piece() const;
@@ -144,6 +144,7 @@ public:
 
   // Static Exchange Evaluation
   bool see_ge(Move m, Value threshold = VALUE_ZERO) const;
+  bool see_ge(Move m, Bitboard& occupied, Value threshold = VALUE_ZERO) const;
 
   // Accessing hash keys
   Key key() const;
@@ -178,8 +179,8 @@ public:
 private:
   // Initialization helpers (used while setting up a position)
   void set_castling_right(Color c, Square rfrom);
-  void set_state(StateInfo* si) const;
+  void set_state() const;
-  void set_check_info(StateInfo* si) const;
+  void set_check_info() const;
 
   // Other helpers
   void move_piece(Square from, Square to);
@@ -204,7 +205,7 @@ private:
   bool chess960;
 };
 
-extern std::ostream& operator<<(std::ostream& os, const Position& pos);
+std::ostream& operator<<(std::ostream& os, const Position& pos);
 
 inline Color Position::side_to_move() const {
   return sideToMove;
@@ -227,20 +228,18 @@ inline Bitboard Position::pieces(PieceType pt = ALL_PIECES) const {
   return byTypeBB[pt];
 }
 
-inline Bitboard Position::pieces(PieceType pt1, PieceType pt2) const {
+template<typename ...PieceTypes>
-  return pieces(pt1) | pieces(pt2);
+inline Bitboard Position::pieces(PieceType pt, PieceTypes... pts) const {
+  return pieces(pt) | pieces(pts...);
 }
 
 inline Bitboard Position::pieces(Color c) const {
   return byColorBB[c];
 }
 
-inline Bitboard Position::pieces(Color c, PieceType pt) const {
+template<typename ...PieceTypes>
-  return pieces(c) & pieces(pt);
+inline Bitboard Position::pieces(Color c, PieceTypes... pts) const {
-}
+  return pieces(c) & pieces(pts...);
-
-inline Bitboard Position::pieces(Color c, PieceType pt1, PieceType pt2) const {
-  return pieces(c) & (pieces(pt1) | pieces(pt2));
 }
 
 template<PieceType Pt> inline int Position::count(Color c) const {
@@ -383,8 +382,16 @@ inline bool Position::is_chess960() const {
 
 inline bool Position::capture(Move m) const {
   assert(is_ok(m));
-  // Castling is encoded as "king captures rook"
+  return     (!empty(to_sq(m)) && type_of(m) != CASTLING)
-  return (!empty(to_sq(m)) && type_of(m) != CASTLING) || type_of(m) == EN_PASSANT;
+          ||  type_of(m) == EN_PASSANT;
+}
+
+// returns true if a move is generated from the capture stage
+// having also queen promotions covered, i.e. consistency with the capture stage move generation
+// is needed to avoid the generation of duplicate moves.
+inline bool Position::capture_stage(Move m) const {
+  assert(is_ok(m));
+  return  capture(m) || promotion_type(m) == QUEEN;
 }
 
 inline Piece Position::captured_piece() const {

src/psqt.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/psqt.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -30,7 +30,7 @@ namespace Stockfish::PSQT
 extern Score psq[PIECE_NB][SQUARE_NB];
 
 // Fill psqt array from a set of internally linked parameters
-extern void init();
+void init();
 
 } // namespace Stockfish::PSQT
 

src/search.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -71,6 +71,7 @@ struct RootMove {
   Value score = -VALUE_INFINITE;
   Value previousScore = -VALUE_INFINITE;
   Value averageScore = -VALUE_INFINITE;
+  Value uciScore = -VALUE_INFINITE;
   bool scoreLowerbound = false;
   bool scoreUpperbound = false;
   int selDepth = 0;
@@ -79,7 +80,7 @@ struct RootMove {
   std::vector<Move> pv;
 };
 
-typedef std::vector<RootMove> RootMoves;
+using RootMoves = std::vector<RootMove>;
 
 
 /// LimitsType struct stores information sent by GUI about available time to

src/syzygy/tbprobe.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -24,9 +24,10 @@
 #include <fstream>
 #include <iostream>
 #include <list>
-#include <sstream>
-#include <type_traits>
 #include <mutex>
+#include <sstream>
+#include <string_view>
+#include <type_traits>
 
 #include "../bitboard.h"
 #include "../movegen.h"
@@ -70,7 +71,7 @@ enum TBFlag { STM = 1, Mapped = 2, WinPlies = 4, LossPlies = 8, Wide = 16, Singl
 inline WDLScore operator-(WDLScore d) { return WDLScore(-int(d)); }
 inline Square operator^(Square s, int i) { return Square(int(s) ^ i); }
 
-const std::string PieceToChar = " PNBRQK  pnbrqk";
+constexpr std::string_view PieceToChar = " PNBRQK  pnbrqk";
 
 int MapPawns[SQUARE_NB];
 int MapB1H1H7[SQUARE_NB];
@@ -141,7 +142,7 @@ struct SparseEntry {
 
 static_assert(sizeof(SparseEntry) == 6, "SparseEntry must be 6 bytes");
 
-typedef uint16_t Sym; // Huffman symbol
+using Sym = uint16_t; // Huffman symbol
 
 struct LR {
     enum Side { Left, Right };
@@ -199,13 +200,10 @@ public:
         }
     }
 
-    // Memory map the file and check it. File should be already open and will be
+    // Memory map the file and check it.
-    // closed after mapping.
     uint8_t* map(void** baseAddress, uint64_t* mapping, TBType type) {
-
+        if (is_open())
-        assert(is_open());
+            close(); // Need to re-open to get native file descriptor
-
-        close(); // Need to re-open to get native file descriptor
 
 #ifndef _WIN32
         struct stat statbuf;
@@ -236,7 +234,7 @@ public:
         }
 #else
         // Note FILE_FLAG_RANDOM_ACCESS is only a hint to Windows and as such may get ignored.
-        HANDLE fd = CreateFile(fname.c_str(), GENERIC_READ, FILE_SHARE_READ, nullptr,
+        HANDLE fd = CreateFileA(fname.c_str(), GENERIC_READ, FILE_SHARE_READ, nullptr,
                                OPEN_EXISTING, FILE_FLAG_RANDOM_ACCESS, nullptr);
 
         if (fd == INVALID_HANDLE_VALUE)
@@ -329,7 +327,7 @@ struct PairsData {
 // first access, when the corresponding file is memory mapped.
 template<TBType Type>
 struct TBTable {
-    typedef typename std::conditional<Type == WDL, WDLScore, int>::type Ret;
+    using Ret = typename std::conditional<Type == WDL, WDLScore, int>::type;
 
     static constexpr int Sides = Type == WDL ? 2 : 1;
 
@@ -1514,7 +1512,7 @@ int Tablebases::probe_dtz(Position& pos, ProbeState* result) {
 // A return value false indicates that not all probes were successful.
 bool Tablebases::root_probe(Position& pos, Search::RootMoves& rootMoves) {
 
-    ProbeState result;
+    ProbeState result = OK;
     StateInfo st;
 
     // Obtain 50-move counter for the root position
@@ -1593,7 +1591,7 @@ bool Tablebases::root_probe_wdl(Position& pos, Search::RootMoves& rootMoves) {
 
     static const int WDL_to_rank[] = { -MAX_DTZ, -MAX_DTZ + 101, 0, MAX_DTZ - 101, MAX_DTZ };
 
-    ProbeState result;
+    ProbeState result = OK;
     StateInfo st;
     WDLScore wdl;
 

src/syzygy/tbprobe.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/thread.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -60,7 +60,6 @@ void Thread::clear() {
   counterMoves.fill(MOVE_NONE);
   mainHistory.fill(0);
   captureHistory.fill(0);
-  previousDepth = 0;
 
   for (bool inCheck : { false, true })
       for (StatsType c : { NoCaptures, Captures })
@@ -73,9 +72,9 @@ void Thread::clear() {
 /// Thread::start_searching() wakes up the thread that will start the search
 
 void Thread::start_searching() {
-
+  mutex.lock();
-  std::lock_guard<std::mutex> lk(mutex);
   searching = true;
+  mutex.unlock(); // Unlock before notifying saves a few CPU-cycles
   cv.notify_one(); // Wake up the thread in idle_loop()
 }
 
@@ -125,20 +124,20 @@ void Thread::idle_loop() {
 
 void ThreadPool::set(size_t requested) {
 
-  if (size() > 0)   // destroy any existing thread(s)
+  if (threads.size() > 0)   // destroy any existing thread(s)
   {
       main()->wait_for_search_finished();
 
-      while (size() > 0)
+      while (threads.size() > 0)
-          delete back(), pop_back();
+          delete threads.back(), threads.pop_back();
   }
 
   if (requested > 0)   // create new thread(s)
   {
-      push_back(new MainThread(0));
+      threads.push_back(new MainThread(0));
 
-      while (size() < requested)
+      while (threads.size() < requested)
-          push_back(new Thread(size()));
+          threads.push_back(new Thread(threads.size()));
       clear();
 
       // Reallocate the hash with the new threadpool size
@@ -154,7 +153,7 @@ void ThreadPool::set(size_t requested) {
 
 void ThreadPool::clear() {
 
-  for (Thread* th : *this)
+  for (Thread* th : threads)
       th->clear();
 
   main()->callsCnt = 0;
@@ -187,7 +186,7 @@ void ThreadPool::start_thinking(Position& pos, StateListPtr& states,
       Tablebases::rank_root_moves(pos, rootMoves);
 
   // After ownership transfer 'states' becomes empty, so if we stop the search
-  // and call 'go' again without setting a new position states.get() == NULL.
+  // and call 'go' again without setting a new position states.get() == nullptr.
   assert(states.get() || setupStates.get());
 
   if (states.get())
@@ -198,7 +197,7 @@ void ThreadPool::start_thinking(Position& pos, StateListPtr& states,
   // be deduced from a fen string, so set() clears them and they are set from
   // setupStates->back() later. The rootState is per thread, earlier states are shared
   // since they are read-only.
-  for (Thread* th : *this)
+  for (Thread* th : threads)
   {
       th->nodes = th->tbHits = th->nmpMinPly = th->bestMoveChanges = 0;
       th->rootDepth = th->completedDepth = 0;
@@ -212,12 +211,12 @@ void ThreadPool::start_thinking(Position& pos, StateListPtr& states,
 
 Thread* ThreadPool::get_best_thread() const {
 
-    Thread* bestThread = front();
+    Thread* bestThread = threads.front();
     std::map<Move, int64_t> votes;
     Value minScore = VALUE_NONE;
 
     // Find minimum score of all threads
-    for (Thread* th: *this)
+    for (Thread* th: threads)
         minScore = std::min(minScore, th->rootMoves[0].score);
 
     // Vote according to score and depth, and select the best thread
@@ -225,10 +224,10 @@ Thread* ThreadPool::get_best_thread() const {
             return (th->rootMoves[0].score - minScore + 14) * int(th->completedDepth);
         };
 
-    for (Thread* th : *this)
+    for (Thread* th : threads)
         votes[th->rootMoves[0].pv[0]] += thread_value(th);
 
-    for (Thread* th : *this)
+    for (Thread* th : threads)
         if (abs(bestThread->rootMoves[0].score) >= VALUE_TB_WIN_IN_MAX_PLY)
         {
             // Make sure we pick the shortest mate / TB conversion or stave off mate the longest
@@ -239,7 +238,8 @@ Thread* ThreadPool::get_best_thread() const {
                  || (   th->rootMoves[0].score > VALUE_TB_LOSS_IN_MAX_PLY
                      && (   votes[th->rootMoves[0].pv[0]] > votes[bestThread->rootMoves[0].pv[0]]
                          || (   votes[th->rootMoves[0].pv[0]] == votes[bestThread->rootMoves[0].pv[0]]
-                             && thread_value(th) > thread_value(bestThread)))))
+                             &&   thread_value(th) * int(th->rootMoves[0].pv.size() > 2)
+                                > thread_value(bestThread) * int(bestThread->rootMoves[0].pv.size() > 2)))))
             bestThread = th;
 
     return bestThread;
@@ -250,8 +250,8 @@ Thread* ThreadPool::get_best_thread() const {
 
 void ThreadPool::start_searching() {
 
-    for (Thread* th : *this)
+    for (Thread* th : threads)
-        if (th != front())
+        if (th != threads.front())
             th->start_searching();
 }
 
@@ -260,8 +260,8 @@ void ThreadPool::start_searching() {
 
 void ThreadPool::wait_for_search_finished() const {
 
-    for (Thread* th : *this)
+    for (Thread* th : threads)
-        if (th != front())
+        if (th != threads.front())
             th->wait_for_search_finished();
 }
 

src/thread.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -60,16 +60,14 @@ public:
   Pawns::Table pawnsTable;
   Material::Table materialTable;
   size_t pvIdx, pvLast;
-  RunningAverage complexityAverage;
   std::atomic<uint64_t> nodes, tbHits, bestMoveChanges;
   int selDepth, nmpMinPly;
-  Color nmpColor;
   Value bestValue, optimism[COLOR_NB];
 
   Position rootPos;
   StateInfo rootState;
   Search::RootMoves rootMoves;
-  Depth rootDepth, completedDepth, previousDepth;
+  Depth rootDepth, completedDepth;
   Value rootDelta;
   CounterMoveHistory counterMoves;
   ButterflyHistory mainHistory;
@@ -101,13 +99,13 @@ struct MainThread : public Thread {
 /// parking and, most importantly, launching a thread. All the access to threads
 /// is done through this class.
 
-struct ThreadPool : public std::vector<Thread*> {
+struct ThreadPool {
 
   void start_thinking(Position&, StateListPtr&, const Search::LimitsType&, bool = false);
   void clear();
   void set(size_t);
 
-  MainThread* main()        const { return static_cast<MainThread*>(front()); }
+  MainThread* main()        const { return static_cast<MainThread*>(threads.front()); }
   uint64_t nodes_searched() const { return accumulate(&Thread::nodes); }
   uint64_t tb_hits()        const { return accumulate(&Thread::tbHits); }
   Thread* get_best_thread() const;
@@ -116,13 +114,21 @@ struct ThreadPool : public std::vector<Thread*> {
 
   std::atomic_bool stop, increaseDepth;
 
+  auto cbegin() const noexcept { return threads.cbegin(); }
+  auto begin() noexcept { return threads.begin(); }
+  auto end() noexcept { return threads.end(); }
+  auto cend() const noexcept { return threads.cend(); }
+  auto size() const noexcept { return threads.size(); }
+  auto empty() const noexcept { return threads.empty(); }
+
 private:
   StateListPtr setupStates;
+  std::vector<Thread*> threads;
 
   uint64_t accumulate(std::atomic<uint64_t> Thread::* member) const {
 
     uint64_t sum = 0;
-    for (Thread* th : *this)
+    for (Thread* th : threads)
         sum += (th->*member).load(std::memory_order_relaxed);
     return sum;
   }

src/thread_win32_osx.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -41,7 +41,7 @@ void* start_routine(void* ptr)
    P* p = reinterpret_cast<P*>(ptr);
    (p->first->*(p->second))(); // Call member function pointer
    delete p;
-   return NULL;
+   return nullptr;
 }
 
 class NativeThread {
@@ -56,7 +56,7 @@ public:
     pthread_attr_setstacksize(attr, TH_STACK_SIZE);
     pthread_create(&thread, attr, start_routine<T>, new P(obj, fun));
   }
-  void join() { pthread_join(thread, NULL); }
+  void join() { pthread_join(thread, nullptr); }
 };
 
 } // namespace Stockfish
@@ -65,7 +65,7 @@ public:
 
 namespace Stockfish {
 
-typedef std::thread NativeThread;
+using NativeThread = std::thread;
 
 } // namespace Stockfish
 

src/timeman.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -36,6 +36,12 @@ TimeManagement Time; // Our global time management object
 
 void TimeManagement::init(Search::LimitsType& limits, Color us, int ply) {
 
+  // if we have no time, no need to initialize TM, except for the start time,
+  // which is used by movetime.
+  startTime = limits.startTime;
+  if (limits.time[us] == 0)
+      return;
+
   TimePoint moveOverhead    = TimePoint(Options["Move Overhead"]);
   TimePoint slowMover       = TimePoint(Options["Slow Mover"]);
   TimePoint npmsec          = TimePoint(Options["nodestime"]);
@@ -59,8 +65,6 @@ void TimeManagement::init(Search::LimitsType& limits, Color us, int ply) {
       limits.npmsec = npmsec;
   }
 
-  startTime = limits.startTime;
-
   // Maximum move horizon of 50 moves
   int mtg = limits.movestogo ? std::min(limits.movestogo, 50) : 50;
 

src/timeman.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/tt.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -87,7 +87,7 @@ void TranspositionTable::clear() {
 
   std::vector<std::thread> threads;
 
-  for (size_t idx = 0; idx < Options["Threads"]; ++idx)
+  for (size_t idx = 0; idx < size_t(Options["Threads"]); ++idx)
   {
       threads.emplace_back([this, idx]() {
 
@@ -98,7 +98,7 @@ void TranspositionTable::clear() {
           // Each thread will zero its part of the hash table
           const size_t stride = size_t(clusterCount / Options["Threads"]),
                        start  = size_t(stride * idx),
-                       len    = idx != Options["Threads"] - 1 ?
+                       len    = idx != size_t(Options["Threads"]) - 1 ?
                                 stride : clusterCount - start;
 
           std::memset(&table[start], 0, len * sizeof(Cluster));

src/tt.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/tune.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by

src/tune.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -26,8 +26,8 @@
 
 namespace Stockfish {
 
-typedef std::pair<int, int> Range; // Option's min-max values
+using Range = std::pair<int, int>; // Option's min-max values
-typedef Range (RangeFun) (int);
+using RangeFun = Range (int);
 
 // Default Range function, to calculate Option's min-max values
 inline Range default_range(int v) {
@@ -75,7 +75,7 @@ struct SetRange {
 
 class Tune {
 
-  typedef void (PostUpdate) (); // Post-update function
+  using PostUpdate = void (); // Post-update function
 
   Tune() { read_results(); }
   Tune(const Tune&) = delete;

src/types.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -103,8 +103,8 @@ constexpr bool Is64Bit = true;
 constexpr bool Is64Bit = false;
 #endif
 
-typedef uint64_t Key;
+using Key = uint64_t;
-typedef uint64_t Bitboard;
+using Bitboard = uint64_t;
 
 constexpr int MAX_MOVES = 256;
 constexpr int MAX_PLY   = 246;
@@ -186,6 +186,9 @@ enum Value : int {
   VALUE_MATE_IN_MAX_PLY  =  VALUE_MATE - MAX_PLY,
   VALUE_MATED_IN_MAX_PLY = -VALUE_MATE_IN_MAX_PLY,
 
+  // In the code, we make the assumption that these values
+  // are such that non_pawn_material() can be used to uniquely
+  // identify the material on the board.
   PawnValueMg   = 126,   PawnValueEg   = 208,
   KnightValueMg = 781,   KnightValueEg = 854,
   BishopValueMg = 825,   BishopValueEg = 915,
@@ -215,7 +218,7 @@ constexpr Value PieceValue[PHASE_NB][PIECE_NB] = {
     VALUE_ZERO, PawnValueEg, KnightValueEg, BishopValueEg, RookValueEg, QueenValueEg, VALUE_ZERO, VALUE_ZERO }
 };
 
-typedef int Depth;
+using Depth = int;
 
 enum : int {
   DEPTH_QS_CHECKS     =  0,
@@ -413,6 +416,10 @@ inline Color color_of(Piece pc) {
   return Color(pc >> 3);
 }
 
+constexpr bool is_ok(Move m) {
+  return m != MOVE_NONE && m != MOVE_NULL;
+}
+
 constexpr bool is_ok(Square s) {
   return s >= SQ_A1 && s <= SQ_H8;
 }
@@ -442,10 +449,12 @@ constexpr Direction pawn_push(Color c) {
 }
 
 constexpr Square from_sq(Move m) {
+  assert(is_ok(m));
   return Square((m >> 6) & 0x3F);
 }
 
 constexpr Square to_sq(Move m) {
+  assert(is_ok(m));
   return Square(m & 0x3F);
 }
 
@@ -470,10 +479,6 @@ constexpr Move make(Square from, Square to, PieceType pt = KNIGHT) {
   return Move(T + ((pt - KNIGHT) << 12) + (from << 6) + to);
 }
 
-constexpr bool is_ok(Move m) {
-  return from_sq(m) != to_sq(m); // Catch MOVE_NULL and MOVE_NONE
-}
-
 /// Based on a congruential pseudo random number generator
 constexpr Key make_key(uint64_t seed) {
   return seed * 6364136223846793005ULL + 1442695040888963407ULL;

src/uci.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -22,6 +22,7 @@
 #include <sstream>
 #include <string>
 
+#include "benchmark.h"
 #include "evaluate.h"
 #include "movegen.h"
 #include "position.h"
@@ -31,13 +32,12 @@
 #include "tt.h"
 #include "uci.h"
 #include "syzygy/tbprobe.h"
+#include "nnue/evaluate_nnue.h"
 
 using namespace std;
 
 namespace Stockfish {
 
-extern vector<string> setup_bench(const Position&, istream&);
-
 namespace {
 
   // FEN string for the initial position in standard chess
@@ -161,7 +161,7 @@ namespace {
     uint64_t num, nodes = 0, cnt = 1;
 
     vector<string> list = setup_bench(pos, args);
-    num = count_if(list.begin(), list.end(), [](string s) { return s.find("go ") == 0 || s.find("eval") == 0; });
+    num = count_if(list.begin(), list.end(), [](const string& s) { return s.find("go ") == 0 || s.find("eval") == 0; });
 
     TimePoint elapsed = now();
 
@@ -207,8 +207,8 @@ namespace {
      // The coefficients of a third-order polynomial fit is based on the fishtest data
      // for two parameters that need to transform eval to the argument of a logistic
      // function.
-     constexpr double as[] = {  -0.58270499,    2.68512549,   15.24638015,  344.49745382};
+     constexpr double as[] = {   0.38036525,   -2.82015070,   23.17882135,  307.36768407};
-     constexpr double bs[] = {  -2.65734562,   15.96509799,  -20.69040836,   73.61029937 };
+     constexpr double bs[] = {  -2.29434733,   13.27689788,  -14.26828904,   63.45318330 };
 
      // Enforce that NormalizeToPawnValue corresponds to a 50% win rate at ply 64
      static_assert(UCI::NormalizeToPawnValue == int(as[0] + as[1] + as[2] + as[3]));
@@ -315,8 +315,13 @@ string UCI::value(Value v) {
 
   stringstream ss;
 
-  if (abs(v) < VALUE_MATE_IN_MAX_PLY)
+  if (abs(v) < VALUE_TB_WIN_IN_MAX_PLY)
       ss << "cp " << v * 100 / NormalizeToPawnValue;
+  else if (abs(v) < VALUE_MATE_IN_MAX_PLY)
+  {
+      const int ply = VALUE_MATE_IN_MAX_PLY - 1 - std::abs(v);  // recompute ss->ply
+      ss << "cp " << (v > 0 ? 20000 - ply : -20000 + ply);
+  }
   else
       ss << "mate " << (v > 0 ? VALUE_MATE - v + 1 : -VALUE_MATE - v) / 2;
 
@@ -354,15 +359,15 @@ std::string UCI::square(Square s) {
 
 string UCI::move(Move m, bool chess960) {
 
-  Square from = from_sq(m);
-  Square to = to_sq(m);
-
   if (m == MOVE_NONE)
       return "(none)";
 
   if (m == MOVE_NULL)
       return "0000";
 
+  Square from = from_sq(m);
+  Square to = to_sq(m);
+
   if (type_of(m) == CASTLING && !chess960)
       to = make_square(to > from ? FILE_G : FILE_C, rank_of(from));
 

src/uci.h

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -35,7 +35,7 @@ namespace UCI {
 // the win_rate_model() such that Stockfish outputs an advantage of
 // "100 centipawns" for a position if the engine has a 50% probability to win
 // from this position in selfplay at fishtest LTC time control.
-const int NormalizeToPawnValue = 361;
+const int NormalizeToPawnValue = 328;
 
 class Option;
 
@@ -45,12 +45,12 @@ struct CaseInsensitiveLess {
 };
 
 /// The options container is defined as a std::map
-typedef std::map<std::string, Option, CaseInsensitiveLess> OptionsMap;
+using OptionsMap = std::map<std::string, Option, CaseInsensitiveLess>;
 
 /// The Option class implements each option as specified by the UCI protocol
 class Option {
 
-  typedef void (*OnChange)(const Option&);
+  using OnChange = void (*)(const Option&);
 
 public:
   Option(OnChange = nullptr);
@@ -61,7 +61,7 @@ public:
 
   Option& operator=(const std::string&);
   void operator<<(const Option&);
-  operator double() const;
+  operator int() const;
   operator std::string() const;
   bool operator==(const char*) const;
 

src/ucioption.cpp

@@ -1,6 +1,6 @@
 /*
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
-  Copyright (C) 2004-2022 The Stockfish developers (see AUTHORS file)
+  Copyright (C) 2004-2023 The Stockfish developers (see AUTHORS file)
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -38,13 +38,13 @@ UCI::OptionsMap Options; // Global object
 namespace UCI {
 
 /// 'On change' actions, triggered by an option's value change
-void on_clear_hash(const Option&) { Search::clear(); }
+static void on_clear_hash(const Option&) { Search::clear(); }
-void on_hash_size(const Option& o) { TT.resize(size_t(o)); }
+static void on_hash_size(const Option& o) { TT.resize(size_t(o)); }
-void on_logger(const Option& o) { start_logger(o); }
+static void on_logger(const Option& o) { start_logger(o); }
-void on_threads(const Option& o) { Threads.set(size_t(o)); }
+static void on_threads(const Option& o) { Threads.set(size_t(o)); }
-void on_tb_path(const Option& o) { Tablebases::init(o); }
+static void on_tb_path(const Option& o) { Tablebases::init(o); }
-void on_use_NNUE(const Option& ) { Eval::NNUE::init(); }
+static void on_use_NNUE(const Option&) { Eval::NNUE::init(); }
-void on_eval_file(const Option& ) { Eval::NNUE::init(); }
+static void on_eval_file(const Option&) { Eval::NNUE::init(); }
 
 /// Our case insensitive less() function as required by UCI protocol
 bool CaseInsensitiveLess::operator() (const string& s1, const string& s2) const {
@@ -73,7 +73,7 @@ void init(OptionsMap& o) {
   o["UCI_Chess960"]          << Option(false);
   o["UCI_AnalyseMode"]       << Option(false);
   o["UCI_LimitStrength"]     << Option(false);
-  o["UCI_Elo"]               << Option(1350, 1350, 2850);
+  o["UCI_Elo"]               << Option(1320, 1320, 3190);
   o["UCI_ShowWDL"]           << Option(false);
   o["SyzygyPath"]            << Option("<empty>", on_tb_path);
   o["SyzygyProbeDepth"]      << Option(1, 1, 100);
@@ -128,9 +128,9 @@ Option::Option(double v, int minv, int maxv, OnChange f) : type("spin"), min(min
 Option::Option(const char* v, const char* cur, OnChange f) : type("combo"), min(0), max(0), on_change(f)
 { defaultValue = v; currentValue = cur; }
 
-Option::operator double() const {
+Option::operator int() const {
   assert(type == "check" || type == "spin");
-  return (type == "spin" ? stof(currentValue) : currentValue == "true");
+  return (type == "spin" ? std::stoi(currentValue) : currentValue == "true");
 }
 
 Option::operator std::string() const {

tests/instrumented.sh

@@ -70,7 +70,8 @@ for args in "eval" \
             "go depth 10" \
             "go movetime 1000" \
             "go wtime 8000 btime 8000 winc 500 binc 500" \
-            "bench 128 $threads 8 default depth"
+            "bench 128 $threads 8 default depth" \
+            "export_net verify.nnue"
 do
 
    echo "$prefix $exeprefix ./stockfish $args $postfix"
@@ -78,6 +79,11 @@ do
 
 done
 
+# verify the generated net equals the base net
+network=`./stockfish uci | grep 'option name EvalFile type string default' | awk '{print $NF}'`
+echo "Comparing $network to the written verify.nnue"
+diff $network verify.nnue
+
 # more general testing, following an uci protocol exchange
 cat << EOF > game.exp
  set timeout 240