Merge branch 'master' into cleanup-predict

.github/workflows/r_tests.yml

@@ -46,8 +46,8 @@ jobs:
         uses: actions/cache@v4
         with:
           path: ${{ env.R_LIBS_USER }}
-          key: ${{ runner.os }}-r-${{ matrix.r }}-7-${{ hashFiles('R-package/DESCRIPTION') }}
-          restore-keys: ${{ runner.os }}-r-${{ matrix.r }}-7-${{ hashFiles('R-package/DESCRIPTION') }}
+          key: ${{ runner.os }}-r-${{ matrix.r }}-8-${{ hashFiles('R-package/DESCRIPTION') }}
+          restore-keys: ${{ runner.os }}-r-${{ matrix.r }}-8-${{ hashFiles('R-package/DESCRIPTION') }}
       - uses: actions/setup-python@v5
         with:
           python-version: "3.10"

CMakeLists.txt

@@ -497,7 +497,8 @@ if(GOOGLE_TEST)
   configure_file(
     ${xgboost_SOURCE_DIR}/tests/cli/machine.conf.in
     ${xgboost_BINARY_DIR}/tests/cli/machine.conf
-    @ONLY)
+    @ONLY
+    NEWLINE_STYLE UNIX)
   if(BUILD_DEPRECATED_CLI)
     add_test(
       NAME TestXGBoostCLI

R-package/R/xgb.train.R

@@ -459,10 +459,10 @@ xgb.train <- function(params = xgb.params(), data, nrounds, evals = list(),
 #' @param seed Random number seed. If not specified, will take a random seed through R's own RNG engine.
 #' @param booster (default= `"gbtree"`)
 #' Which booster to use. Can be `"gbtree"`, `"gblinear"` or `"dart"`; `"gbtree"` and `"dart"` use tree based models while `"gblinear"` uses linear functions.
-#' @param eta,learning_rate (two aliases for the same parameter) (default=0.3)
+#' @param eta,learning_rate (two aliases for the same parameter)
 #' Step size shrinkage used in update to prevent overfitting. After each boosting step, we can directly get the weights of new features, and `eta` shrinks the feature weights to make the boosting process more conservative.
-#'
-#' range: \eqn{[0,1]}
+#' - range: \eqn{[0,1]}
+#' - default value: 0.3 for tree-based boosters, 0.5 for linear booster.
 #'
 #' Note: should only pass one of `eta` or `learning_rate`. Both refer to the same parameter and there's thus no difference between one or the other.
 #' @param gamma,min_split_loss (two aliases for the same parameter) (for Tree Booster) (default=0, alias: `gamma`)

R-package/configure.win

@@ -0,0 +1,21 @@
+R_EXE="${R_HOME}/bin${R_ARCH_BIN}/R.exe"
+CXX=`"${R_EXE}" CMD config CXX`
+
+cat > test.cpp <<EOL
+#include <xmmintrin.h>
+int main() {
+  char data = 0;
+  const char* address = &data;
+  _mm_prefetch(address, _MM_HINT_NTA);
+  return 0;
+}
+EOL
+
+XGBOOST_MM_PREFETCH_PRESENT=""
+${CXX} -o test test.cpp 2>/dev/null && ./test && XGBOOST_MM_PREFETCH_PRESENT="-DXGBOOST_MM_PREFETCH_PRESENT=1"
+rm -f ./test
+rm -f ./test.cpp
+
+sed \
+    -e "s/@XGBOOST_MM_PREFETCH_PRESENT@/$XGBOOST_MM_PREFETCH_PRESENT/" \
+    < src/Makevars.win.in > src/Makevars.win

R-package/src/Makevars.win → R-package/src/Makevars.win.in

@@ -22,6 +22,8 @@ PKG_CPPFLAGS = \
     -I$(PKGROOT)/include \
     -I$(PKGROOT)/dmlc-core/include \
     -I$(PKGROOT) \
+    -DXGBOOST_BUILTIN_PREFETCH_PRESENT=1 \
+    @XGBOOST_MM_PREFETCH_PRESENT@ \
     $(XGB_RFLAGS)
 
 PKG_CXXFLAGS = \

cmake/Version.cmake

@@ -2,5 +2,7 @@ function(write_version)
   message(STATUS "xgboost VERSION: ${xgboost_VERSION}")
   configure_file(
     ${xgboost_SOURCE_DIR}/cmake/version_config.h.in
-    ${xgboost_SOURCE_DIR}/include/xgboost/version_config.h @ONLY)
+    ${xgboost_SOURCE_DIR}/include/xgboost/version_config.h
+    @ONLY
+    NEWLINE_STYLE UNIX)
 endfunction()

doc/parameter.rst

@@ -348,6 +348,11 @@ Parameters for Linear Booster (``booster=gblinear``)
 
   - L1 regularization term on weights. Increasing this value will make model more conservative. Normalised to number of training examples.
 
+* ``eta`` [default=0.5, alias: ``learning_rate``]
+
+  - Step size shrinkage used in update to prevent overfitting. After each boosting step, we can directly get the weights of new features, and ``eta`` shrinks the feature weights to make the boosting process more conservative.
+  - range: [0,1]
+
 * ``updater`` [default= ``shotgun``]
 
   - Choice of algorithm to fit linear model

doc/tutorials/model.rst

@@ -146,7 +146,7 @@ It remains to ask: which tree do we want at each step?  A natural thing is to ad
 
 .. math::
 
-  \text{obj}^{(t)} & = \sum_{i=1}^n l(y_i, \hat{y}_i^{(t)}) + \sum_{i=1}^t\omega(f_i) \\
+  \text{obj}^{(t)} & = \sum_{i=1}^n l(y_i, \hat{y}_i^{(t)}) + \sum_{k=1}^t\omega(f_k) \\
             & = \sum_{i=1}^n l(y_i, \hat{y}_i^{(t-1)} + f_t(x_i)) + \omega(f_t) + \mathrm{constant}
 
 If we consider using mean squared error (MSE) as our loss function, the objective becomes

ops/script/test_r_package.py

@@ -63,7 +63,7 @@ def pkgroot(path: str) -> None:
     if os.path.exists(osxmakef):
         os.remove(osxmakef)
     pkgroot("Makevars.in")
-    pkgroot("Makevars.win")
+    pkgroot("Makevars.win.in")
     # misc
     rwsp = Path("R-package") / "remove_warning_suppression_pragma.sh"
     if system() != "Windows":