This commit enables a mixed bench, to improve CI and allow for PGO (profile-build...

Joint work gvreuls / vondele

* Download the default NNUE net in AppVeyor
* Download net in travis CI `make net`
* Adjust tests to cover more archs, speedup instrumented testing
* Introduce 'mixed' bench as default, with further options:

classical, NNUE, mixed.

mixed (default) and NNUE require the default net to be present,
which can be obtained with

```
make net
```

Further examples (first is equivalent to `./stockfish bench`):

```
./stockfish bench 16 1 13 default depth mixed
./stockfish bench 16 1 13 default depth classical
./stockfish bench 16 1 13 default depth NNUE
```

The net is now downloaded automatically if needed for `profile-build`
(usual `build` works fine without net present)

PGO gives a nice speedup on fishtest:

passed STC:
LLR: 2.93 (-2.94,2.94) {-0.50,1.50}
Total: 3360 W: 469 L: 343 D: 2548
Ptnml(0-2): 20, 246, 1030, 356, 28
https://tests.stockfishchess.org/tests/view/5f31b5499081672066537569

passed LTC:
LLR: 2.97 (-2.94,2.94) {0.25,1.75}
Total: 8824 W: 609 L: 502 D: 7713
Ptnml(0-2): 8, 430, 3438, 519, 17
https://tests.stockfishchess.org/tests/view/5f31c87b908167206653757c

closes https://github.com/official-stockfish/Stockfish/pull/2931

fixes https://github.com/official-stockfish/Stockfish/issues/2907

requires fishtest updates before commit

Bench: 4290577

diff --git a/.travis.yml b/.travis.yml
index d563a1e1992dd875b336e2e55f416c529eaaa5bc..0dd380479cca2c8ebd418df9ade97553b80b6b90 100644 (file)
--- a/.travis.yml
+++ b/.travis.yml
@@ -43,6 +43,9 @@ before_script:
   - cd src
 
 script:
+  # Download net
+  - make net
+
   # Obtain bench reference from git log
   - git log HEAD | grep "\b[Bb]ench[ :]\+[0-9]\{7\}" | head -n 1 | sed "s/[^0-9]*\([0-9]*\).*/\1/g" > git_sig
   - export benchref=$(cat git_sig)
@@ -55,14 +58,26 @@ script:
   #
   # Verify bench number against various builds
   - export CXXFLAGS="-Werror -D_GLIBCXX_DEBUG"
-  - make clean && make -j2 ARCH=x86-64 optimize=no debug=yes build && ../tests/signature.sh $benchref
+  - make clean && make -j2 ARCH=x86-64-modern optimize=no debug=yes build && ../tests/signature.sh $benchref
+  - export CXXFLAGS="-Werror"
+  - make clean && make -j2 ARCH=x86-64-modern build && ../tests/signature.sh $benchref
+  - make clean && make -j2 ARCH=x86-64-ssse3 build && ../tests/signature.sh $benchref
+  - make clean && make -j2 ARCH=x86-64-sse3-popcnt build && ../tests/signature.sh $benchref
+  - make clean && make -j2 ARCH=x86-64 build && ../tests/signature.sh $benchref
   - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then make clean && make -j2 ARCH=x86-32 optimize=no debug=yes build && ../tests/signature.sh $benchref; fi
   - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then make clean && make -j2 ARCH=x86-32 build && ../tests/signature.sh $benchref; fi
+  - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then make clean && make -j2 ARCH=x86-32-old build && ../tests/signature.sh $benchref; fi
+  - if [[ "$TRAVIS_OS_NAME" == "linux" && "$COMP" == "gcc" ]]; then make clean && make -j2 ARCH=x86-64-modern profile-build && ../tests/signature.sh $benchref; fi
+
+  # compile only for some more advanced architectures (might not run in travis)
+  - make clean && make -j2 ARCH=x86-64-avx2 build
+  - make clean && make -j2 ARCH=x86-64-bmi2 build
+  # needs gcc 10 to compile
+  - if [[ "$COMPILER" != "g++-8" ]]; then make clean && make -j2 ARCH=x86-64-avx512 build; fi
 
   #
   # Check perft and reproducible search
-  - export CXXFLAGS="-Werror"
-  - make clean && make -j2 ARCH=x86-64 build
+  - make clean && make -j2 ARCH=x86-64-modern build
   - ../tests/perft.sh
   - ../tests/reprosearch.sh
 
@@ -70,11 +85,11 @@ script:
   # Valgrind
   #
   - export CXXFLAGS="-O1 -fno-inline"
-  - if [ -x "$(command -v valgrind )" ]; then make clean && make -j2 ARCH=x86-64 debug=yes optimize=no build > /dev/null && ../tests/instrumented.sh --valgrind; fi
+  - if [ -x "$(command -v valgrind )" ]; then make clean && make -j2 ARCH=x86-64-modern debug=yes optimize=no build > /dev/null && ../tests/instrumented.sh --valgrind; fi
   - if [ -x "$(command -v valgrind )" ]; then ../tests/instrumented.sh --valgrind-thread; fi
 
   #
   # Sanitizer
   #
-  - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then make clean && make -j2 ARCH=x86-64 sanitize=undefined optimize=no debug=yes build > /dev/null && ../tests/instrumented.sh --sanitizer-undefined; fi
-  - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then make clean && make -j2 ARCH=x86-64 sanitize=thread    optimize=no debug=yes build > /dev/null && ../tests/instrumented.sh --sanitizer-thread; fi
+  - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then make clean && make -j2 ARCH=x86-64-modern sanitize=undefined optimize=no debug=yes build > /dev/null && ../tests/instrumented.sh --sanitizer-undefined; fi
+  - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then make clean && make -j2 ARCH=x86-64-modern sanitize=thread    optimize=no debug=yes build > /dev/null && ../tests/instrumented.sh --sanitizer-thread; fi

diff --git a/appveyor.yml b/appveyor.yml
index d356ba2f101f566238cc25f3237c29e4f71a62d8..a3732a23f4a8882ab8becf85265cac52371b6890 100644 (file)
--- a/appveyor.yml
+++ b/appveyor.yml
@@ -61,6 +61,20 @@ before_build:
 
 build_script:
   - cmake --build . --config %CONFIGURATION% -- /verbosity:minimal
+  - ps: |
+      # Download default NNUE net from fishtest
+      $nnuenet = Get-Content -Path src\ucioption.cpp | Select-String -CaseSensitive -Pattern "Option" | Select-String -CaseSensitive -Pattern "nn-[a-z0-9]{12}.nnue"
+      $dummy = $nnuenet -match "(?<nnuenet>nn-[a-z0-9]{12}.nnue)"
+      $nnuenet = $Matches.nnuenet
+      Write-Host "Default net:" $nnuenet
+      $nnuedownloadurl = "https://tests.stockfishchess.org/api/nn/$nnuenet"
+      $nnuefilepath = "src\${env:CONFIGURATION}\$nnuenet"
+      if (Test-Path -Path $nnuefilepath) {
+            Write-Host "Already available."
+      } else {
+            Write-Host "Downloading $nnuedownloadurl to $nnuefilepath"
+            Invoke-WebRequest -Uri $nnuedownloadurl -OutFile $nnuefilepath
+      }
 
 before_test:
   - cd src/%CONFIGURATION%

diff --git a/src/Makefile b/src/Makefile
index e34fbf615e1924f23ff26c8fdd39a3b456d1ab05..c00b60b5a56abfc6737689a1e2c2604719853f5a 100644 (file)
--- a/src/Makefile
+++ b/src/Makefile
@@ -569,7 +569,7 @@ help:
 build: config-sanity
        $(MAKE) ARCH=$(ARCH) COMP=$(COMP) all
 
-profile-build: config-sanity objclean profileclean
+profile-build: config-sanity objclean profileclean net
        @echo ""
        @echo "Step 1/4. Building instrumented executable ..."
        $(MAKE) ARCH=$(ARCH) COMP=$(COMP) $(profile_make)

diff --git a/src/benchmark.cpp b/src/benchmark.cpp
index 6041d64287da6353d275a195172bb075b12eee12..806e98401a466b93ba4204e6c0dd3a966c2068b8 100644 (file)
--- a/src/benchmark.cpp
+++ b/src/benchmark.cpp
@@ -95,8 +95,9 @@ const vector<string> Defaults = {
 /// setup_bench() builds a list of UCI commands to be run by bench. There
 /// are five parameters: TT size in MB, number of search threads that
 /// should be used, the limit value spent for each position, a file name
-/// where to look for positions in FEN format and the type of the limit:
-/// depth, perft, nodes and movetime (in millisecs).
+/// where to look for positions in FEN format, the type of the limit:
+/// depth, perft, nodes and movetime (in millisecs), and evaluation type
+/// mixed (default), classical, NNUE.
 ///
 /// bench -> search default positions up to depth 13
 /// bench 64 1 15 -> search default positions up to depth 15 (TT = 64MB)
@@ -115,6 +116,7 @@ vector<string> setup_bench(const Position& current, istream& is) {
   string limit     = (is >> token) ? token : "13";
   string fenFile   = (is >> token) ? token : "default";
   string limitType = (is >> token) ? token : "depth";
+  string evalType  = (is >> token) ? token : "mixed";
 
   go = limitType == "eval" ? "eval" : "go " + limitType + " " + limit;
 
@@ -146,13 +148,20 @@ vector<string> setup_bench(const Position& current, istream& is) {
   list.emplace_back("setoption name Hash value " + ttSize);
   list.emplace_back("ucinewgame");
 
+  size_t posCounter = 0;
+
   for (const string& fen : fens)
       if (fen.find("setoption") != string::npos)
           list.emplace_back(fen);
       else
       {
+          if (evalType == "classical" || (evalType == "mixed" && posCounter % 2 == 0))
+              list.emplace_back("setoption name Use NNUE value false");
+          else if (evalType == "NNUE" || (evalType == "mixed" && posCounter % 2 != 0))
+              list.emplace_back("setoption name Use NNUE value true");
           list.emplace_back("position fen " + fen);
           list.emplace_back(go);
+          ++posCounter;
       }
 
   return list;

diff --git a/tests/instrumented.sh b/tests/instrumented.sh
index ae6d5c4b905f2e79464a1a36c45ddc72a323a184..03ded74abfd023011c8ee901716173354b37235f 100755 (executable)
--- a/tests/instrumented.sh
+++ b/tests/instrumented.sh
@@ -70,7 +70,7 @@ for args in "eval" \
             "go depth 10" \
             "go movetime 1000" \
             "go wtime 8000 btime 8000 winc 500 binc 500" \
-            "bench 128 $threads 10 default depth"
+            "bench 128 $threads 8 default depth"
 do
 
    echo "$prefix $exeprefix ./stockfish $args $postfix"
@@ -80,7 +80,7 @@ done
 
 # more general testing, following an uci protocol exchange
 cat << EOF > game.exp
- set timeout 10
+ set timeout 240
  spawn $exeprefix ./stockfish
 
  send "uci\n"
@@ -98,7 +98,7 @@ cat << EOF > game.exp
  expect "bestmove"
 
  send "position fen 5rk1/1K4p1/8/8/3B4/8/8/8 b - - 0 1\n"
- send "go depth 30\n"
+ send "go depth 20\n"
  expect "bestmove"
 
  send "quit\n"
@@ -121,7 +121,7 @@ cat << EOF > syzygy.exp
  send "uci\n"
  send "setoption name SyzygyPath value ../tests/syzygy/\n"
  expect "info string Found 35 tablebases" {} timeout {exit 1}
- send "bench 128 1 10 default depth\n"
+ send "bench 128 1 8 default depth\n"
  send "quit\n"
  expect eof