Merge branch 'stan-dev:develop' into feature/matrix-input-to-mvn

stan-dev · Jul 18, 2021 · 03de55a · 03de55a
2 parents aa929c9 + c87156b
commit 03de55a
Show file tree

Hide file tree

Showing 115 changed files with 4,517 additions and 173 deletions.
diff --git a/Jenkinsfile b/Jenkinsfile
@@ -245,6 +245,11 @@ pipeline {
                                 sh "echo STAN_OPENCL=true>> make/local"
                                 sh "echo OPENCL_PLATFORM_ID=${env.OPENCL_PLATFORM_ID_CPU}>> make/local"
                                 sh "echo OPENCL_DEVICE_ID=${env.OPENCL_DEVICE_ID_CPU}>> make/local"
+                                // skips tests that require specific support in OpenCL
+                                sh 'echo "ifdef NO_CPU_OPENCL_INT64_BASE_ATOMIC" >> make/local'
+                                sh 'echo "CXXFLAGS += -DSTAN_TEST_SKIP_REQUIRING_OPENCL_INT64_BASE_ATOMIC" >> make/local'
+                                sh 'echo "endif" >> make/local'
+
                                 runTests("test/unit/math/opencl", false)
                                 runTests("test/unit/multiple_translation_units_test.cpp")
                             } else {

diff --git a/benchmarks/benchmark.py b/benchmarks/benchmark.py
@@ -419,18 +419,22 @@ def benchmark(
                 value = 0.4
                 if function_name in special_arg_values:
                     if isinstance(special_arg_values[function_name][n], str):
-                        make_arg_function = special_arg_values[function_name][n]
+                        make_arg_function = make_special_arg_values[special_arg_values[function_name][n]]
                     elif isinstance(
                             special_arg_values[function_name][n], numbers.Number
                     ):
                         value = special_arg_values[function_name][n]
-                if scalar == "double":
+                if not is_argument_autodiff or (
+                      not is_argument_scalar and (
+                          opencl == "base" or varmat == "base" or make_arg_function != "make_arg"
+                )):
+                    arg_type_prim = cpp_arg_template.replace("SCALAR", "double");
                     setup += (
                         "  {} {} = stan::test::{}<{}>({}, state.range(0));\n".format(
-                            arg_type,
+                            arg_type_prim,
                             var_name,
                             make_arg_function,
-                            arg_type,
+                            arg_type_prim,
                             value,
                         )
                     )
@@ -440,11 +444,22 @@ def benchmark(
                                 var_name + "_cl", var_name
                             )
                             var_name += "_cl"
+                            if is_argument_autodiff:
+                                var_conversions += (
+                                  "    stan::math::var_value<stan::math::matrix_cl<double>> {}({});\n".format(
+                                    var_name + "_var", var_name)
+                                )
+                                var_name += "_var"
                         elif varmat == "base" and arg_overload == "Rev":
-                            setup += "  auto {} = stan::math::to_var_value({});\n".format(
-                                var_name + "_varmat", var_name
+                            var_conversions += "      stan::math::var_value<{}> {}({});\n".format(
+                                arg_type_prim, var_name + "_varmat", var_name
                             )
                             var_name += "_varmat"
+                        elif is_argument_autodiff: #rev
+                            var_conversions += "    {} {} = {};\n".format(
+                                arg_type, var_name + "_var", var_name
+                            )
+                            var_name += "_var"
                 else:
                     var_conversions += (
                         "    {} {} = stan::test::{}<{}>({}, state.range(0));\n".format(

diff --git a/lib/tbb_2020.3/build/macos.inc b/lib/tbb_2020.3/build/macos.inc
@@ -36,10 +36,14 @@ ifndef arch
      export arch:=ppc32
    endif
  else
-   ifeq ($(shell /usr/sbin/sysctl -n hw.optional.x86_64 2>/dev/null),1)
-     export arch:=intel64
+   ifeq ($(shell /usr/sbin/sysctl -n hw.machine),arm64)
+    export arch:=arm64
    else
-     export arch:=ia32
+    ifeq ($(shell /usr/sbin/sysctl -n hw.optional.x86_64 2>/dev/null),1)
+      export arch:=intel64
+    else
+      export arch:=ia32
+    endif
    endif
  endif
 endif

diff --git a/make/compiler_flags b/make/compiler_flags
@@ -91,7 +91,7 @@ CXX_MINOR := $(word 2,$(subst ., ,$(CXX_VERSION)))
 ##
 
 ifdef STAN_NO_RANGE_CHECKS
-  CXXFLAGS_THREADS ?= -DSTAN_NO_RANGE_CHECKS
+  CXXFLAGS_NO_RANGE_CHECKS ?= -DSTAN_NO_RANGE_CHECKS
 endif
 
 ################################################################################
@@ -306,8 +306,8 @@ ifdef STAN_MPI
   CXXFLAGS_MPI ?= -Wno-delete-non-virtual-dtor
 endif
 
-CXXFLAGS += $(CXXFLAGS_LANG) $(CXXFLAGS_OS) $(CXXFLAGS_WARNINGS) $(CXXFLAGS_BOOST) $(CXXFLAGS_EIGEN) $(CXXFLAGS_OPENCL) $(CXXFLAGS_MPI) $(CXXFLAGS_THREADS) $(CXXFLAGS_TBB) $(CXXFLAGS_FLTO) $(CXXFLAGS_OPTIM) -O$(O) $(INC)
-CPPFLAGS += $(CPPFLAGS_LANG) $(CPPFLAGS_OS) $(CPPFLAGS_WARNINGS) $(CPPFLAGS_BOOST) $(CPPFLAGS_EIGEN) $(CPPFLAGS_OPENCL) $(CPPFLAGS_TBB) $(CPPFLAGS_MPI) $(CPPFLAGS_TBB) $(CPPFLAGS_FLTO) $(CPPFLAGS_OPTIM)
+CXXFLAGS += $(CXXFLAGS_LANG) $(CXXFLAGS_OS) $(CXXFLAGS_WARNINGS) $(CXXFLAGS_BOOST) $(CXXFLAGS_EIGEN) $(CXXFLAGS_OPENCL) $(CXXFLAGS_MPI) $(CXXFLAGS_THREADS) $(CXXFLAGS_TBB) $(CXXFLAGS_FLTO) $(CXXFLAGS_OPTIM) $(CXXFLAGS_NO_RANGE_CHECKS) -O$(O) $(INC)
+CPPFLAGS += $(CPPFLAGS_LANG) $(CPPFLAGS_OS) $(CPPFLAGS_WARNINGS) $(CPPFLAGS_BOOST) $(CPPFLAGS_EIGEN) $(CPPFLAGS_OPENCL) $(CPPFLAGS_TBB) $(CPPFLAGS_MPI) $(CPPFLAGS_TBB) $(CPPFLAGS_FLTO) $(CPPFLAGS_OPTIM) $(CXXFLAGS_NO_RANGE_CHECKS)
 LDFLAGS += $(LDFLAGS_LANG) $(LDFLAGS_OS) $(LDFLAGS_WARNINGS) $(LDFLAGS_BOOST) $(LDFLAGS_EIGEN) $(LDFLAGS_OPENCL) $(LDFLAGS_MPI) $(LDFLAGS_TBB) $(LDFLAGS_FLTO) $(LDFLAGS_OPTIM)
 LDLIBS += $(LDLIBS_LANG) $(LDLIBS_OS) $(LDLIBS_WARNINGS) $(LDLIBS_BOOST) $(LDLIBS_EIGEN) $(LDLIBS_OPENCL) $(LDLIBS_MPI) $(LDLIBS_TBB)
 

diff --git a/make/tests b/make/tests
@@ -100,7 +100,7 @@ else
 endif
 
 %.hpp-test : %.hpp test/dummy.cpp
-	$(COMPILE.cpp) $(CXXFLAGS) -O0 -include $^ -o $(DEV_NULL)
+	$(COMPILE.cpp) $(CXXFLAGS) -O0 -include $^ -o $(DEV_NULL) -Wunused-local-typedefs
 
 test/dummy.cpp:
 	@mkdir -p test

diff --git a/stan/math/fwd/fun/mdivide_left.hpp b/stan/math/fwd/fun/mdivide_left.hpp
@@ -60,7 +60,6 @@ template <typename T1, typename T2,
 inline Eigen::Matrix<value_type_t<T2>, T1::RowsAtCompileTime,
                      T2::ColsAtCompileTime>
 mdivide_left(const T1& A, const T2& b) {
-  using T = typename value_type_t<T2>::Scalar;
   constexpr int S1 = T1::RowsAtCompileTime;
   constexpr int C2 = T2::ColsAtCompileTime;
 

diff --git a/stan/math/fwd/fun/mdivide_left_tri_low.hpp b/stan/math/fwd/fun/mdivide_left_tri_low.hpp
@@ -55,7 +55,6 @@ template <typename T1, typename T2, require_eigen_t<T1>* = nullptr,
 inline Eigen::Matrix<value_type_t<T2>, T1::RowsAtCompileTime,
                      T2::ColsAtCompileTime>
 mdivide_left_tri_low(const T1& A, const T2& b) {
-  using T = typename value_type_t<T2>::Scalar;
   constexpr int S1 = T1::RowsAtCompileTime;
   constexpr int C2 = T2::ColsAtCompileTime;
 

diff --git a/stan/math/opencl/copy.hpp b/stan/math/opencl/copy.hpp
@@ -55,8 +55,7 @@ inline matrix_cl<scalar_type_t<T>> to_matrix_cl(T&& src) {
  * @param src source matrix on the OpenCL device
  * @return Eigen matrix with a copy of the data in the source matrix
  */
-template <typename T_ret, typename T,
-          require_eigen_vt<std::is_arithmetic, T_ret>* = nullptr,
+template <typename T_ret, typename T, require_eigen_t<T_ret>* = nullptr,
           require_matrix_cl_t<T>* = nullptr,
           require_st_same<T_ret, T>* = nullptr>
 inline auto from_matrix_cl(const T& src) {