From 20ad11de6c8976956e53c0c27009fa81d97a67c2 Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Thu, 16 Jul 2026 10:40:17 -0500
Subject: [PATCH 01/18] refactor(tests): link internal tests against
 cuopt_static

Prerequisite structural work for enabling -fvisibility=hidden on
libcuopt.so (issue #1213).

- Introduce cuopt_objs OBJECT library so cuopt (shared) and
  cuopt_static (static, BUILD_TESTS only) share one compile step
- Add STATIC_LIB option to ConfigureTest; when set, links cuopt_static
  and omits cuopttestutils to avoid transitive cuopt-shared pull-in
- Route all tests with compiled internal symbol dependencies through
  STATIC_LIB: DUAL_SIMPLEX_TEST, LP_INTERNAL_TEST, PDLP_TEST,
  MPS_PARSER_TEST, most MIP tests, SOCP_TEST, ROUTING_INTERNAL_TEST
- Create ROUTING_INTERNAL_TEST consolidating GES, scross, local-search
  candidate, and top-k tests; register previously unbuilt test files
  (l0_scross_test.cu, local_search_cand_test.cu, load_balancing_test.cu,
  feasibility_jump_tests.cu)
- Add gRPC source files directly to GRPC_CLIENT_TEST and
  GRPC_INTEGRATION_TEST so they do not depend on internal symbols being
  exported from the shared library
- Promote is_cusparse_runtime_mixed_precision_supported() to the public
  C API as cuOptIsCusparseRuntimeMixedPrecisionSupported(); update
  c_api_tests.cpp to use the public function

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/CMakeLists.txt                            | 130 +++++++++++++-----
 .../cuopt/mathematical_optimization/cuopt_c.h |   8 ++
 cpp/src/pdlp/cuopt_c.cpp                      |   8 ++
 cpp/tests/CMakeLists.txt                      |  22 +--
 cpp/tests/dual_simplex/CMakeLists.txt         |   3 +-
 cpp/tests/linear_programming/CMakeLists.txt   |  15 +-
 .../c_api_tests/c_api_tests.cpp               |   7 +-
 .../linear_programming/grpc/CMakeLists.txt    |  10 ++
 cpp/tests/mip/CMakeLists.txt                  |  45 ++++--
 cpp/tests/routing/CMakeLists.txt              |  18 ++-
 cpp/tests/socp/CMakeLists.txt                 |   3 +-
 11 files changed, 186 insertions(+), 83 deletions(-)

diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
index 61f6eb91df..e3f887dbe5 100644
--- a/cpp/CMakeLists.txt
+++ b/cpp/CMakeLists.txt
@@ -512,56 +512,41 @@ if (NOT SKIP_GRPC_BUILD)
             APPEND PROPERTY COMPILE_OPTIONS "-DNDEBUG")
 endif (NOT SKIP_GRPC_BUILD)
 
-add_library(cuopt SHARED
+add_library(cuopt_objs OBJECT
         ${CUOPT_SRC_FILES}
 )
 
-set_target_properties(cuopt
-        PROPERTIES BUILD_RPATH "\$ORIGIN"
-        INSTALL_RPATH "\$ORIGIN"
-        INTERFACE_POSITION_INDEPENDENT_CODE ON
+set_target_properties(cuopt_objs
+        PROPERTIES POSITION_INDEPENDENT_CODE ON
         CXX_SCAN_FOR_MODULES OFF
 )
 
-target_compile_definitions(cuopt
-        PUBLIC "CUOPT_LOG_ACTIVE_LEVEL=RAPIDS_LOGGER_LOG_LEVEL_${LIBCUOPT_LOGGING_LEVEL}"
-        PUBLIC CUSPARSE_ENABLE_EXPERIMENTAL_API
+target_compile_definitions(cuopt_objs
+  PUBLIC "CUOPT_LOG_ACTIVE_LEVEL=RAPIDS_LOGGER_LOG_LEVEL_${LIBCUOPT_LOGGING_LEVEL}"
+  PUBLIC CUSPARSE_ENABLE_EXPERIMENTAL_API
 )
 
-target_compile_options(cuopt
+target_compile_options(cuopt_objs
         PRIVATE "$<$<COMPILE_LANGUAGE:CXX>:${CUOPT_CXX_FLAGS}>"
         "$<$<COMPILE_LANGUAGE:CUDA>:${CUOPT_CUDA_FLAGS}>"
 )
 
-if (WRITE_FATBIN)
-    file(WRITE "${CUOPT_BINARY_DIR}/fatbin.ld"
-            [=[
-  SECTIONS
-  {
-    .nvFatBinSegment : { *(.nvFatBinSegment) }
-    .nv_fatbin : { *(.nv_fatbin) }
-  }
-  ]=])
-    target_link_options(cuopt PRIVATE "${CUOPT_BINARY_DIR}/fatbin.ld")
-endif ()
-
-add_library(cuopt::cuopt ALIAS cuopt)
 # ##################################################################################################
 # - include paths ---------------------------------------------------------------------------------
 message(STATUS "target include directories CUDSS_INCLUDES = ${CUDSS_INCLUDE}")
 
 # Adding Papilo as a system include messes up clang's include resolution if papilo is already installed as a conda package
-target_include_directories(cuopt PRIVATE
+target_include_directories(cuopt_objs PRIVATE
         "${papilo_SOURCE_DIR}/src"
         "${papilo_BINARY_DIR}"
 )
 
-target_include_directories(cuopt SYSTEM PRIVATE
-        "${pslp_SOURCE_DIR}/include"
-        "${dejavu_SOURCE_DIR}"
+target_include_directories(cuopt_objs SYSTEM PRIVATE
+  "${pslp_SOURCE_DIR}/include"
+  "${dejavu_SOURCE_DIR}"
 )
 
-target_include_directories(cuopt
+target_include_directories(cuopt_objs
         PRIVATE
         "${CMAKE_CURRENT_SOURCE_DIR}/../thirdparty"
         "${CMAKE_CURRENT_SOURCE_DIR}/src"
@@ -581,8 +566,8 @@ target_include_directories(cuopt
 )
 
 # Link PSLP by file to avoid export dependency tracking
-target_link_libraries(cuopt PRIVATE $<TARGET_FILE:PSLP>)
-add_dependencies(cuopt PSLP)
+target_link_libraries(cuopt_objs PRIVATE $<TARGET_FILE:PSLP>)
+add_dependencies(cuopt_objs PSLP)
 
 # ##################################################################################################
 # - link libraries --------------------------------------------------------------------------------
@@ -597,7 +582,7 @@ list(PREPEND CUOPT_PRIVATE_CUDA_LIBS CUDA::cublasLt)
 
 # Pass CUDSS_MT_LIB_FILE_NAME as a compile definition
 get_filename_component(CUDSS_MT_LIB_FILE_NAME "${CUDSS_MT_LIB_FILE}" NAME)
-target_compile_definitions(cuopt PRIVATE CUDSS_MT_LIB_FILE_NAME="${CUDSS_MT_LIB_FILE_NAME}")
+target_compile_definitions(cuopt_objs PRIVATE CUDSS_MT_LIB_FILE_NAME="${CUDSS_MT_LIB_FILE_NAME}")
 
 execute_process(
         COMMAND git rev-parse --short HEAD
@@ -616,14 +601,14 @@ configure_file(
 )
 
 # Add the generated include directory
-target_include_directories(cuopt PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/include)
+target_include_directories(cuopt_objs PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/include)
 
 list(JOIN CMAKE_CUDA_ARCHITECTURES "," JOINED_CUDA_ARCHITECTURES)
-target_compile_definitions(cuopt PUBLIC
+target_compile_definitions(cuopt_objs PUBLIC
         CUOPT_CUDA_ARCHITECTURES="${JOINED_CUDA_ARCHITECTURES}"
         CUOPT_CPU_ARCHITECTURE="${CMAKE_SYSTEM_PROCESSOR}")
 
-target_link_libraries(cuopt
+target_link_libraries(cuopt_objs
         PUBLIC
         CUDA::cublas
         CUDA::cusparse
@@ -634,9 +619,6 @@ target_link_libraries(cuopt
         ${CUDSS_LIB_FILE}
         PRIVATE
         ${CUOPT_PRIVATE_CUDA_LIBS}
-        simde::simde
-        OpenMP::OpenMP_CXX
-        OpenMP::OpenMP_CUDA
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:protobuf::libprotobuf>
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:gRPC::grpc++>
 )
@@ -645,9 +627,85 @@ target_link_libraries(cuopt
 # - generate tests --------------------------------------------------------------------------------
 if (BUILD_TESTS)
     include(CTest)
+    add_library(cuopt_static STATIC $<TARGET_OBJECTS:cuopt_objs>)
+    target_link_libraries(cuopt_static
+            PUBLIC
+            CUDA::cublas
+            CUDA::cusparse
+            rmm::rmm
+            rapids_logger::rapids_logger
+            CCCL::CCCL
+            raft::raft
+            ${CUDSS_LIB_FILE}
+            PRIVATE
+            ${CUOPT_PRIVATE_CUDA_LIBS}
+            $<$<BOOL:${CUOPT_ENABLE_GRPC}>:protobuf::libprotobuf>
+            $<$<BOOL:${CUOPT_ENABLE_GRPC}>:gRPC::grpc++>
+    )
+    target_include_directories(cuopt_static
+            PUBLIC
+            "$<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/include>"
+            "$<BUILD_INTERFACE:${CMAKE_CURRENT_BINARY_DIR}/include>"
+            "$<INSTALL_INTERFACE:include>"
+    )
+    target_compile_definitions(cuopt_static PUBLIC
+            CUOPT_CUDA_ARCHITECTURES="${JOINED_CUDA_ARCHITECTURES}"
+            CUOPT_CPU_ARCHITECTURE="${CMAKE_SYSTEM_PROCESSOR}"
+            "CUOPT_LOG_ACTIVE_LEVEL=RAPIDS_LOGGER_LOG_LEVEL_${LIBCUOPT_LOGGING_LEVEL}"
+            CUSPARSE_ENABLE_EXPERIMENTAL_API
+    )
+    target_link_libraries(cuopt_static PRIVATE $<TARGET_FILE:PSLP>)
+    add_dependencies(cuopt_static PSLP)
     add_subdirectory(tests)
 endif (BUILD_TESTS)
 
+add_library(cuopt SHARED $<TARGET_OBJECTS:cuopt_objs>)
+add_library(cuopt::cuopt ALIAS cuopt)
+set_target_properties(cuopt
+        PROPERTIES BUILD_RPATH "\$ORIGIN"
+        INSTALL_RPATH "\$ORIGIN"
+        INTERFACE_POSITION_INDEPENDENT_CODE ON
+        CXX_SCAN_FOR_MODULES OFF
+        LINKER_LANGUAGE CUDA
+)
+# cuopt needs PUBLIC interface for consumers
+target_include_directories(cuopt
+        PUBLIC
+        "$<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/include>"
+        "$<BUILD_INTERFACE:${CMAKE_CURRENT_BINARY_DIR}/include>"
+        INTERFACE
+        "$<INSTALL_INTERFACE:include>"
+)
+target_link_libraries(cuopt
+        PUBLIC
+        CUDA::cublas
+        CUDA::cusparse
+        rmm::rmm
+        rapids_logger::rapids_logger
+        CCCL::CCCL
+        raft::raft
+        ${CUDSS_LIB_FILE}
+        PRIVATE
+        ${CUOPT_PRIVATE_CUDA_LIBS}
+        simde::simde
+        OpenMP::OpenMP_CXX
+        OpenMP::OpenMP_CUDA
+        $<$<BOOL:${CUOPT_ENABLE_GRPC}>:protobuf::libprotobuf>
+        $<$<BOOL:${CUOPT_ENABLE_GRPC}>:gRPC::grpc++>
+)
+
+if (WRITE_FATBIN)
+    file(WRITE "${CUOPT_BINARY_DIR}/fatbin.ld"
+            [=[
+  SECTIONS
+  {
+    .nvFatBinSegment : { *(.nvFatBinSegment) }
+    .nv_fatbin : { *(.nv_fatbin) }
+  }
+  ]=])
+    target_link_options(cuopt PRIVATE "${CUOPT_BINARY_DIR}/fatbin.ld")
+endif ()
+
 # ##################################################################################################
 # - install targets -------------------------------------------------------------------------------
 
diff --git a/cpp/include/cuopt/mathematical_optimization/cuopt_c.h b/cpp/include/cuopt/mathematical_optimization/cuopt_c.h
index 218402a7ef..d66c908f1d 100644
--- a/cpp/include/cuopt/mathematical_optimization/cuopt_c.h
+++ b/cpp/include/cuopt/mathematical_optimization/cuopt_c.h
@@ -1055,6 +1055,14 @@ cuopt_int_t cuOptGetDualObjectiveValue(cuOptSolution solution,
  */
 cuopt_int_t cuOptGetReducedCosts(cuOptSolution solution, cuopt_float_t* reduced_cost_ptr);
 
+/**
+ * @brief Query whether the runtime cuSPARSE version supports mixed-precision SpMV
+ *        (FP32 matrix × FP64 vector). Requires cuSPARSE >= 12.5 at runtime.
+ *
+ * @return 1 if supported, 0 otherwise.
+ */
+cuopt_int_t cuOptIsCusparseRuntimeMixedPrecisionSupported();
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/cpp/src/pdlp/cuopt_c.cpp b/cpp/src/pdlp/cuopt_c.cpp
index 963bd6ada7..d07641cff6 100644
--- a/cpp/src/pdlp/cuopt_c.cpp
+++ b/cpp/src/pdlp/cuopt_c.cpp
@@ -15,6 +15,7 @@
 #include <cuopt/mathematical_optimization/solver_settings.hpp>
 #include <cuopt/utilities/timestamp_utils.hpp>
 #include <pdlp/cuopt_c_internal.hpp>
+#include <pdlp/cusparse_view.hpp>
 #include <utilities/logger.hpp>
 
 #include <cuopt/mathematical_optimization/io/parser.hpp>
@@ -1296,3 +1297,10 @@ cuopt_int_t cuOptGetReducedCosts(cuOptSolution solution, cuopt_float_t* reduced_
     return CUOPT_INVALID_ARGUMENT;
   }
 }
+
+cuopt_int_t cuOptIsCusparseRuntimeMixedPrecisionSupported()
+{
+  return cuopt::mathematical_optimization::pdlp::is_cusparse_runtime_mixed_precision_supported()
+           ? 1
+           : 0;
+}
diff --git a/cpp/tests/CMakeLists.txt b/cpp/tests/CMakeLists.txt
index 27f6c94983..05c7e7a23a 100644
--- a/cpp/tests/CMakeLists.txt
+++ b/cpp/tests/CMakeLists.txt
@@ -37,11 +37,19 @@ endif()
 set(CUOPT_TEST_DIR ${CMAKE_CURRENT_SOURCE_DIR})
 
 # ################################################################ ------------------------------------------------------------------
-# ConfigureTest(NAME source1.cu source2.cu [LABELS label1 label2 ...])
+# ConfigureTest(NAME source1.cu source2.cu [STATIC_LIB] [LABELS label1 label2 ...])
 #
 # LABELS sets CTest labels for selective local test execution via `ctest -L <label>`.
+# STATIC_LIB links cuopt_static instead of cuopt (for tests needing internal symbols).
 function(ConfigureTest CMAKE_TEST_NAME)
-    cmake_parse_arguments(CT "" "" "LABELS" ${ARGN})
+    cmake_parse_arguments(CT "STATIC_LIB" "" "LABELS" ${ARGN})
+    if(CT_STATIC_LIB)
+        set(_cuopt_link_target cuopt_static)
+        set(_extra_libs "")
+    else()
+        set(_cuopt_link_target cuopt)
+        set(_extra_libs cuopttestutils)
+    endif()
     add_executable(${CMAKE_TEST_NAME} ${CT_UNPARSED_ARGUMENTS})
     target_include_directories(${CMAKE_TEST_NAME}
         PRIVATE
@@ -53,13 +61,10 @@ function(ConfigureTest CMAKE_TEST_NAME)
         "${pslp_SOURCE_DIR}/include"
         "${dejavu_SOURCE_DIR}"
     )
-
     target_link_libraries(${CMAKE_TEST_NAME}
        PRIVATE
-        cuopt
-        cuopttestutils
-        OpenMP::OpenMP_CXX
-        OpenMP::OpenMP_CUDA
+        ${_cuopt_link_target}
+        ${_extra_libs}
         GTest::gmock
         GTest::gmock_main
         GTest::gtest
@@ -69,13 +74,10 @@ function(ConfigureTest CMAKE_TEST_NAME)
     if(NOT DEFINED INSTALL_TARGET OR "${INSTALL_TARGET}" STREQUAL "")
       target_link_options(${CMAKE_TEST_NAME} PRIVATE -Wl,--enable-new-dtags)
     endif()
-
     add_test(NAME ${CMAKE_TEST_NAME} COMMAND ${CMAKE_TEST_NAME})
-
     if(CT_LABELS)
         set_tests_properties(${CMAKE_TEST_NAME} PROPERTIES LABELS "${CT_LABELS}")
     endif()
-
     install(
         TARGETS ${CMAKE_TEST_NAME}
         COMPONENT testing
diff --git a/cpp/tests/dual_simplex/CMakeLists.txt b/cpp/tests/dual_simplex/CMakeLists.txt
index 1f6740378f..6a8d2e46d4 100644
--- a/cpp/tests/dual_simplex/CMakeLists.txt
+++ b/cpp/tests/dual_simplex/CMakeLists.txt
@@ -7,4 +7,5 @@ ConfigureTest(DUAL_SIMPLEX_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/solve.cpp
     ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/solve_barrier.cu
     ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/right_looking_ldlt.cpp
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
diff --git a/cpp/tests/linear_programming/CMakeLists.txt b/cpp/tests/linear_programming/CMakeLists.txt
index 6db30755c3..e7bba9c2d1 100644
--- a/cpp/tests/linear_programming/CMakeLists.txt
+++ b/cpp/tests/linear_programming/CMakeLists.txt
@@ -4,22 +4,29 @@
 # cmake-format: on
 
 ConfigureTest(LP_UNIT_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/optimization_problem_test.cu
     ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/solver_settings_test.cu
-    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/presolve_test.cu
     ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/solution_interface_test.cu
     LABELS numopt)
+
+ConfigureTest(LP_INTERNAL_TEST
+    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/optimization_problem_test.cu
+    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/presolve_test.cu
+    LABELS numopt
+    STATIC_LIB)
+
 # ##################################################################################################
 # - Linear programming PDLP tests ----------------------------------------------------------------------
 ConfigureTest(PDLP_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/pdlp_test.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 
 # ##################################################################################################
 # - MPS / LP parser tests --------------------------------------------------------------------------
 ConfigureTest(MPS_PARSER_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/parser_test.cpp
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 
 ConfigureTest(MPS_FAST_PARSER_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/experimental_mps_fast/fast_fp64_parser_test.cpp
diff --git a/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp b/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
index 467af308b7..2778096119 100644
--- a/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
+++ b/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
@@ -21,10 +21,6 @@
 #include <utilities/common_utils.hpp>
 #include <utilities/error.hpp>
 
-namespace cuopt::mathematical_optimization::pdlp {
-bool is_cusparse_runtime_mixed_precision_supported();
-}
-
 #include <gtest/gtest.h>
 
 TEST(c_api, int_size) { EXPECT_EQ(test_int_size(), sizeof(int32_t)); }
@@ -441,12 +437,11 @@ TEST(c_api, pdlp_precision_single)
 
 TEST(c_api, pdlp_precision_mixed)
 {
-  using namespace cuopt::mathematical_optimization::pdlp;
   const std::string& rapidsDatasetRootDir = cuopt::test::get_rapids_dataset_root_dir();
   std::string filename           = rapidsDatasetRootDir + "/linear_programming/afiro_original.mps";
   cuopt_int_t termination_status = -1;
   cuopt_float_t objective;
-  if (!cuopt::mathematical_optimization::pdlp::is_cusparse_runtime_mixed_precision_supported()) {
+  if (!cuOptIsCusparseRuntimeMixedPrecisionSupported()) {
     auto status = test_pdlp_precision_mixed(filename.c_str(), &termination_status, &objective);
     bool solve_returned_error = (status != CUOPT_SUCCESS);
     bool solve_returned_non_optimal =
diff --git a/cpp/tests/linear_programming/grpc/CMakeLists.txt b/cpp/tests/linear_programming/grpc/CMakeLists.txt
index aadbf23df3..4261a7fe60 100644
--- a/cpp/tests/linear_programming/grpc/CMakeLists.txt
+++ b/cpp/tests/linear_programming/grpc/CMakeLists.txt
@@ -9,6 +9,12 @@
 
 add_executable(GRPC_CLIENT_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/grpc_client_test.cpp
+    "${CUOPT_SOURCE_DIR}/src/grpc/grpc_problem_mapper.cpp"
+    "${CUOPT_SOURCE_DIR}/src/grpc/grpc_service_mapper.cpp"
+    "${CUOPT_SOURCE_DIR}/src/grpc/grpc_settings_mapper.cpp"
+    "${CUOPT_SOURCE_DIR}/src/grpc/grpc_solution_mapper.cpp"
+    "${CUOPT_SOURCE_DIR}/src/grpc/client/grpc_client.cpp"
+    "${CUOPT_SOURCE_DIR}/src/grpc/client/grpc_client_env.cpp"
 )
 
 target_include_directories(GRPC_CLIENT_TEST
@@ -16,6 +22,7 @@ target_include_directories(GRPC_CLIENT_TEST
     "${CUOPT_SOURCE_DIR}/include"
     "${CUOPT_SOURCE_DIR}/src/grpc"
     "${CUOPT_SOURCE_DIR}/src/grpc/client"
+    "${CUOPT_SOURCE_DIR}/src/grpc/server"
     "${CUOPT_SOURCE_DIR}/src/grpc/codegen/generated"
     "${CUOPT_TEST_DIR}"
     "${CMAKE_CURRENT_SOURCE_DIR}"  # For grpc_client_test_helper.hpp
@@ -93,6 +100,9 @@ install(
 
 add_executable(GRPC_INTEGRATION_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/grpc_integration_test.cpp
+    "${CUOPT_SOURCE_DIR}/src/grpc/grpc_service_mapper.cpp"
+    "${CUOPT_SOURCE_DIR}/src/grpc/client/grpc_client.cpp"
+    "${CUOPT_SOURCE_DIR}/src/grpc/client/grpc_client_env.cpp"
 )
 
 target_include_directories(GRPC_INTEGRATION_TEST
diff --git a/cpp/tests/mip/CMakeLists.txt b/cpp/tests/mip/CMakeLists.txt
index d533f09c2d..7e20764511 100644
--- a/cpp/tests/mip/CMakeLists.txt
+++ b/cpp/tests/mip/CMakeLists.txt
@@ -7,22 +7,27 @@
 # - MIP tests ----------------------------------------------------------------------
 ConfigureTest(MIP_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/miplib_test.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 ConfigureTest(SEMI_CONTINUOUS_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/semi_continuous_test.cu
     LABELS numopt)
 ConfigureTest(PROBLEM_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/problem_test.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 ConfigureTest(ELIM_VAR_REMAP_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/elim_var_remap_test.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 ConfigureTest(STANDARDIZATION_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/bounds_standardization_test.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 ConfigureTest(MULTI_PROBE_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/multi_probe_test.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 ConfigureTest(INCUMBENT_CALLBACK_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/incumbent_callback_test.cu
     LABELS numopt)
@@ -31,27 +36,37 @@ ConfigureTest(DOC_EXAMPLE_TEST
     LABELS numopt)
 ConfigureTest(CUTS_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/cuts_test.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 ConfigureTest(UNIT_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/unit_test.cu
     ${CMAKE_CURRENT_SOURCE_DIR}/integer_with_real_bounds.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 ConfigureTest(EMPTY_FIXED_PROBLEMS_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/empty_fixed_problems_test.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 ConfigureTest(PRESOLVE_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/presolve_test.cu
-    LABELS numopt)
-# Disable for now
-# ConfigureTest(FEASIBILITY_JUMP_TEST
-#    ${CMAKE_CURRENT_SOURCE_DIR}/feasibility_jump_tests.cu
-# )
+    LABELS numopt
+    STATIC_LIB)
+ConfigureTest(FEASIBILITY_JUMP_TEST
+    ${CMAKE_CURRENT_SOURCE_DIR}/feasibility_jump_tests.cu
+    LABELS numopt
+    STATIC_LIB)
+ConfigureTest(LOAD_BALANCING_TEST
+    ${CMAKE_CURRENT_SOURCE_DIR}/load_balancing_test.cu
+    LABELS numopt
+    STATIC_LIB)
 ConfigureTest(MIP_TERMINATION_STATUS_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/termination_test.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 ConfigureTest(DETERMINISM_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/determinism_test.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)
 ConfigureTest(HEURISTICS_HYPER_PARAMS_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/heuristics_hyper_params_test.cu
     LABELS numopt)
diff --git a/cpp/tests/routing/CMakeLists.txt b/cpp/tests/routing/CMakeLists.txt
index 569f84beb7..d6edc8c6e0 100644
--- a/cpp/tests/routing/CMakeLists.txt
+++ b/cpp/tests/routing/CMakeLists.txt
@@ -5,8 +5,14 @@
 
 ConfigureTest(ROUTING_TEST ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_routing_test.cu
     LABELS routing)
-ConfigureTest(ROUTING_GES_TEST ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_ges_test.cu
-    LABELS routing)
+
+ConfigureTest(ROUTING_INTERNAL_TEST
+    ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_ges_test.cu
+    ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_scross_test.cu
+    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/local_search_cand_test.cu
+    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/top_k.cu
+    LABELS routing
+    STATIC_LIB)
 
 ConfigureTest(VEHICLE_ORDER_TEST ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_vehicle_order_match.cu
     LABELS routing)
@@ -15,17 +21,10 @@ ConfigureTest(VEHICLE_TYPES_TEST ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_vehicle_t
 ConfigureTest(OBJECTIVE_FUNCTION_TEST ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_objective_function_test.cu
     LABELS routing)
 
-# ##################################################################################################
-# - L1 advanced retail tests --------------------------------------------------------------------------
 ConfigureTest(RETAIL_L1TEST ${CMAKE_CURRENT_SOURCE_DIR}/level1/l1_retail_test.cu
     LABELS routing)
-
-# ##################################################################################################
-# - L1  tests for quick regression check --------------------------------------------------------------------------
 ConfigureTest(ROUTING_L1TEST ${CMAKE_CURRENT_SOURCE_DIR}/level1/l1_routing_test.cu
     LABELS routing)
-
-# Disabled: both tests are slow and known to be broken.
 set_tests_properties(RETAIL_L1TEST ROUTING_L1TEST PROPERTIES DISABLED TRUE)
 
 # # - ${CMAKE_CURRENT_SOURCE_DIR} unit tests ----------------------------------------------------------------------------
@@ -41,7 +40,6 @@ ConfigureTest(ROUTING_UNIT_TEST
       ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/heterogenous_fleet.cu
       ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/prize_collection.cu
       ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/objective_function.cu
-      ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/top_k.cu
       ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/batch_tsp.cu
       ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/set_shmem_of_kernel.cu
     LABELS routing)
diff --git a/cpp/tests/socp/CMakeLists.txt b/cpp/tests/socp/CMakeLists.txt
index f380984225..402edee901 100644
--- a/cpp/tests/socp/CMakeLists.txt
+++ b/cpp/tests/socp/CMakeLists.txt
@@ -7,4 +7,5 @@ ConfigureTest(SOCP_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/second_order_cone_kernels.cu
     ${CMAKE_CURRENT_SOURCE_DIR}/solve_barrier_socp.cu
     ${CMAKE_CURRENT_SOURCE_DIR}/general_quadratic_test.cu
-    LABELS numopt)
+    LABELS numopt
+    STATIC_LIB)

From de93a4d9e62412c773da03fa50de610b7fa327fa Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Thu, 16 Jul 2026 11:20:12 -0500
Subject: [PATCH 02/18] refactor(tests): consolidate internal numopt tests into
 single binary

Collapse 19 separate STATIC_LIB test binaries (dual_simplex, LP internal,
PDLP, MPS parser, 11 MIP, SOCP) into a single NUMOPT_INTERNAL_TEST binary
linked against cuopt_static.  Routing internal tests remain in
ROUTING_INTERNAL_TEST.  This reduces total static-linked binary count from
19 to 2, cutting link time and disk usage.

Also fix cuopt_c.cpp to forward-declare
is_cusparse_runtime_mixed_precision_supported() instead of including
cusparse_view.hpp, which pulled CUDA intrinsics into a CXX translation unit.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/src/pdlp/cuopt_c.cpp                    |  5 +-
 cpp/tests/CMakeLists.txt                    |  1 +
 cpp/tests/dual_simplex/CMakeLists.txt       |  7 +--
 cpp/tests/internal/CMakeLists.txt           | 38 ++++++++++++++
 cpp/tests/linear_programming/CMakeLists.txt | 21 +-------
 cpp/tests/mip/CMakeLists.txt                | 56 +--------------------
 cpp/tests/socp/CMakeLists.txt               |  7 +--
 7 files changed, 48 insertions(+), 87 deletions(-)
 create mode 100644 cpp/tests/internal/CMakeLists.txt

diff --git a/cpp/src/pdlp/cuopt_c.cpp b/cpp/src/pdlp/cuopt_c.cpp
index d07641cff6..ed49c9baf9 100644
--- a/cpp/src/pdlp/cuopt_c.cpp
+++ b/cpp/src/pdlp/cuopt_c.cpp
@@ -15,7 +15,6 @@
 #include <cuopt/mathematical_optimization/solver_settings.hpp>
 #include <cuopt/utilities/timestamp_utils.hpp>
 #include <pdlp/cuopt_c_internal.hpp>
-#include <pdlp/cusparse_view.hpp>
 #include <utilities/logger.hpp>
 
 #include <cuopt/mathematical_optimization/io/parser.hpp>
@@ -1298,6 +1297,10 @@ cuopt_int_t cuOptGetReducedCosts(cuOptSolution solution, cuopt_float_t* reduced_
   }
 }
 
+namespace cuopt::mathematical_optimization::pdlp {
+bool is_cusparse_runtime_mixed_precision_supported();
+}
+
 cuopt_int_t cuOptIsCusparseRuntimeMixedPrecisionSupported()
 {
   return cuopt::mathematical_optimization::pdlp::is_cusparse_runtime_mixed_precision_supported()
diff --git a/cpp/tests/CMakeLists.txt b/cpp/tests/CMakeLists.txt
index 05c7e7a23a..b3d001c1c6 100644
--- a/cpp/tests/CMakeLists.txt
+++ b/cpp/tests/CMakeLists.txt
@@ -109,6 +109,7 @@ if(BUILD_TESTS)
   add_subdirectory(linear_programming)
   add_subdirectory(dual_simplex)
   add_subdirectory(socp)
+  add_subdirectory(internal)
   add_subdirectory(qp)
   add_subdirectory(utilities)
 
diff --git a/cpp/tests/dual_simplex/CMakeLists.txt b/cpp/tests/dual_simplex/CMakeLists.txt
index 6a8d2e46d4..8c7cd7005a 100644
--- a/cpp/tests/dual_simplex/CMakeLists.txt
+++ b/cpp/tests/dual_simplex/CMakeLists.txt
@@ -3,9 +3,4 @@
 # SPDX-License-Identifier: Apache-2.0
 # cmake-format: on
 
-ConfigureTest(DUAL_SIMPLEX_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/solve.cpp
-    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/solve_barrier.cu
-    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/right_looking_ldlt.cpp
-    LABELS numopt
-    STATIC_LIB)
+# Internal tests consolidated into NUMOPT_INTERNAL_TEST (cpp/tests/internal/CMakeLists.txt).
diff --git a/cpp/tests/internal/CMakeLists.txt b/cpp/tests/internal/CMakeLists.txt
new file mode 100644
index 0000000000..5e2441ad92
--- /dev/null
+++ b/cpp/tests/internal/CMakeLists.txt
@@ -0,0 +1,38 @@
+# cmake-format: off
+# SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+# cmake-format: on
+
+# Single binary for all numopt tests that require internal symbols (cuopt_static).
+# Add new internal numopt tests here instead of creating separate STATIC_LIB targets.
+ConfigureTest(NUMOPT_INTERNAL_TEST
+    # dual_simplex
+    ${CUOPT_TEST_DIR}/dual_simplex/unit_tests/solve.cpp
+    ${CUOPT_TEST_DIR}/dual_simplex/unit_tests/solve_barrier.cu
+    ${CUOPT_TEST_DIR}/dual_simplex/unit_tests/right_looking_ldlt.cpp
+    # linear_programming
+    ${CUOPT_TEST_DIR}/linear_programming/pdlp_test.cu
+    ${CUOPT_TEST_DIR}/linear_programming/unit_tests/optimization_problem_test.cu
+    ${CUOPT_TEST_DIR}/linear_programming/unit_tests/presolve_test.cu
+    ${CUOPT_TEST_DIR}/linear_programming/parser_test.cpp
+    # mip
+    ${CUOPT_TEST_DIR}/mip/miplib_test.cu
+    ${CUOPT_TEST_DIR}/mip/problem_test.cu
+    ${CUOPT_TEST_DIR}/mip/elim_var_remap_test.cu
+    ${CUOPT_TEST_DIR}/mip/bounds_standardization_test.cu
+    ${CUOPT_TEST_DIR}/mip/multi_probe_test.cu
+    ${CUOPT_TEST_DIR}/mip/cuts_test.cu
+    ${CUOPT_TEST_DIR}/mip/unit_test.cu
+    ${CUOPT_TEST_DIR}/mip/integer_with_real_bounds.cu
+    ${CUOPT_TEST_DIR}/mip/empty_fixed_problems_test.cu
+    ${CUOPT_TEST_DIR}/mip/presolve_test.cu
+    ${CUOPT_TEST_DIR}/mip/feasibility_jump_tests.cu
+    ${CUOPT_TEST_DIR}/mip/load_balancing_test.cu
+    ${CUOPT_TEST_DIR}/mip/termination_test.cu
+    ${CUOPT_TEST_DIR}/mip/determinism_test.cu
+    # socp
+    ${CUOPT_TEST_DIR}/socp/second_order_cone_kernels.cu
+    ${CUOPT_TEST_DIR}/socp/solve_barrier_socp.cu
+    ${CUOPT_TEST_DIR}/socp/general_quadratic_test.cu
+    LABELS numopt
+    STATIC_LIB)
diff --git a/cpp/tests/linear_programming/CMakeLists.txt b/cpp/tests/linear_programming/CMakeLists.txt
index e7bba9c2d1..fc39944609 100644
--- a/cpp/tests/linear_programming/CMakeLists.txt
+++ b/cpp/tests/linear_programming/CMakeLists.txt
@@ -8,26 +8,6 @@ ConfigureTest(LP_UNIT_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/solution_interface_test.cu
     LABELS numopt)
 
-ConfigureTest(LP_INTERNAL_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/optimization_problem_test.cu
-    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/presolve_test.cu
-    LABELS numopt
-    STATIC_LIB)
-
-# ##################################################################################################
-# - Linear programming PDLP tests ----------------------------------------------------------------------
-ConfigureTest(PDLP_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/pdlp_test.cu
-    LABELS numopt
-    STATIC_LIB)
-
-# ##################################################################################################
-# - MPS / LP parser tests --------------------------------------------------------------------------
-ConfigureTest(MPS_PARSER_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/parser_test.cpp
-    LABELS numopt
-    STATIC_LIB)
-
 ConfigureTest(MPS_FAST_PARSER_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/experimental_mps_fast/fast_fp64_parser_test.cpp
     ${CMAKE_CURRENT_SOURCE_DIR}/experimental_mps_fast/fast_parser_edge_test.cpp
@@ -38,6 +18,7 @@ target_include_directories(MPS_FAST_PARSER_TEST
 )
 target_link_libraries(MPS_FAST_PARSER_TEST PRIVATE simde::simde)
 
+
 # ##################################################################################################
 # - C API Tests----------------------------------------------------------------------
 # The C API tests require a separate library to be linked against. So we don't use the ConfigureTest macro.
diff --git a/cpp/tests/mip/CMakeLists.txt b/cpp/tests/mip/CMakeLists.txt
index 7e20764511..91fb87a953 100644
--- a/cpp/tests/mip/CMakeLists.txt
+++ b/cpp/tests/mip/CMakeLists.txt
@@ -4,69 +4,17 @@
 # cmake-format: on
 
 # ##################################################################################################
-# - MIP tests ----------------------------------------------------------------------
-ConfigureTest(MIP_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/miplib_test.cu
-    LABELS numopt
-    STATIC_LIB)
+# - MIP tests that use only public API (no internal symbols) ---------------------------------------
+# Internal MIP tests are consolidated into NUMOPT_INTERNAL_TEST (cpp/tests/internal/CMakeLists.txt).
 ConfigureTest(SEMI_CONTINUOUS_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/semi_continuous_test.cu
     LABELS numopt)
-ConfigureTest(PROBLEM_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/problem_test.cu
-    LABELS numopt
-    STATIC_LIB)
-ConfigureTest(ELIM_VAR_REMAP_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/elim_var_remap_test.cu
-    LABELS numopt
-    STATIC_LIB)
-ConfigureTest(STANDARDIZATION_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/bounds_standardization_test.cu
-    LABELS numopt
-    STATIC_LIB)
-ConfigureTest(MULTI_PROBE_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/multi_probe_test.cu
-    LABELS numopt
-    STATIC_LIB)
 ConfigureTest(INCUMBENT_CALLBACK_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/incumbent_callback_test.cu
     LABELS numopt)
 ConfigureTest(DOC_EXAMPLE_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/doc_example_test.cu
     LABELS numopt)
-ConfigureTest(CUTS_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/cuts_test.cu
-    LABELS numopt
-    STATIC_LIB)
-ConfigureTest(UNIT_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/unit_test.cu
-    ${CMAKE_CURRENT_SOURCE_DIR}/integer_with_real_bounds.cu
-    LABELS numopt
-    STATIC_LIB)
-ConfigureTest(EMPTY_FIXED_PROBLEMS_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/empty_fixed_problems_test.cu
-    LABELS numopt
-    STATIC_LIB)
-ConfigureTest(PRESOLVE_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/presolve_test.cu
-    LABELS numopt
-    STATIC_LIB)
-ConfigureTest(FEASIBILITY_JUMP_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/feasibility_jump_tests.cu
-    LABELS numopt
-    STATIC_LIB)
-ConfigureTest(LOAD_BALANCING_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/load_balancing_test.cu
-    LABELS numopt
-    STATIC_LIB)
-ConfigureTest(MIP_TERMINATION_STATUS_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/termination_test.cu
-    LABELS numopt
-    STATIC_LIB)
-ConfigureTest(DETERMINISM_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/determinism_test.cu
-    LABELS numopt
-    STATIC_LIB)
 ConfigureTest(HEURISTICS_HYPER_PARAMS_TEST
     ${CMAKE_CURRENT_SOURCE_DIR}/heuristics_hyper_params_test.cu
     LABELS numopt)
diff --git a/cpp/tests/socp/CMakeLists.txt b/cpp/tests/socp/CMakeLists.txt
index 402edee901..9490527469 100644
--- a/cpp/tests/socp/CMakeLists.txt
+++ b/cpp/tests/socp/CMakeLists.txt
@@ -3,9 +3,4 @@
 # SPDX-License-Identifier: Apache-2.0
 # cmake-format: on
 
-ConfigureTest(SOCP_TEST
-    ${CMAKE_CURRENT_SOURCE_DIR}/second_order_cone_kernels.cu
-    ${CMAKE_CURRENT_SOURCE_DIR}/solve_barrier_socp.cu
-    ${CMAKE_CURRENT_SOURCE_DIR}/general_quadratic_test.cu
-    LABELS numopt
-    STATIC_LIB)
+# Internal tests consolidated into NUMOPT_INTERNAL_TEST (cpp/tests/internal/CMakeLists.txt).

From e6589c1f4c4e31d7cce6b4fe25868472e67f2e4b Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Thu, 16 Jul 2026 13:13:15 -0500
Subject: [PATCH 03/18] fix(tests): make combined internal test binaries link
 and run

Fix four issues discovered when building NUMOPT_INTERNAL_TEST and
ROUTING_INTERNAL_TEST:

- omp_helpers.hpp: closing namespace brace was outside #ifdef _OPENMP,
  causing "expected a declaration" when compiled without -fopenmp.

- cuopt_static: propagate OpenMP::OpenMP_CXX PUBLIC and -fopenmp for
  CUDA TUs so test binaries that link cuopt_static compile OMP-heavy
  internal headers correctly.

- cuopt shared: restore CUSPARSE_ENABLE_EXPERIMENTAL_API and
  CUOPT_LOG_ACTIVE_LEVEL as PUBLIC compile definitions; they were lost
  when compile definitions moved from cuopt to cuopt_objs (the
  $<TARGET_OBJECTS:...> pattern does not carry INTERFACE properties).

- Combined binaries: add CUOPT_DISABLE_TEST_MAIN guard to
  CUOPT_TEST_PROGRAM_MAIN() in base_fixture.hpp; provide a single
  main.cu entry point per binary; make init_handler static in three
  MIP test files to resolve ODR violations; compile check_constraints.cu
  directly into ROUTING_INTERNAL_TEST for check_route; add missing
  src/ and src/io/ include paths to GRPC_CLIENT_TEST and
  GRPC_INTEGRATION_TEST.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/CMakeLists.txt                            | 14 +++++++++++++
 cpp/tests/internal/CMakeLists.txt             | 10 ++++++++--
 cpp/tests/internal/main.cu                    | 20 +++++++++++++++++++
 .../linear_programming/grpc/CMakeLists.txt    |  3 +++
 cpp/tests/mip/bounds_standardization_test.cu  |  2 +-
 cpp/tests/mip/elim_var_remap_test.cu          |  2 +-
 cpp/tests/mip/multi_probe_test.cu             |  2 +-
 cpp/tests/routing/CMakeLists.txt              |  9 +++++++--
 cpp/tests/utilities/base_fixture.hpp          |  8 ++++++++
 9 files changed, 63 insertions(+), 7 deletions(-)
 create mode 100644 cpp/tests/internal/main.cu

diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
index e3f887dbe5..a0117fb7ba 100644
--- a/cpp/CMakeLists.txt
+++ b/cpp/CMakeLists.txt
@@ -637,11 +637,17 @@ if (BUILD_TESTS)
             CCCL::CCCL
             raft::raft
             ${CUDSS_LIB_FILE}
+            OpenMP::OpenMP_CXX
             PRIVATE
             ${CUOPT_PRIVATE_CUDA_LIBS}
             $<$<BOOL:${CUOPT_ENABLE_GRPC}>:protobuf::libprotobuf>
             $<$<BOOL:${CUOPT_ENABLE_GRPC}>:gRPC::grpc++>
     )
+    # Internal test sources include OMP-dependent headers (omp_helpers.hpp, omp_atomic_t).
+    # Propagate -fopenmp for CUDA TUs since OpenMP::OpenMP_CXX only covers CXX.
+    target_compile_options(cuopt_static INTERFACE
+            "$<$<COMPILE_LANGUAGE:CUDA>:-fopenmp>"
+    )
     target_include_directories(cuopt_static
             PUBLIC
             "$<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/include>"
@@ -693,6 +699,14 @@ target_link_libraries(cuopt
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:protobuf::libprotobuf>
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:gRPC::grpc++>
 )
+# Propagate compile definitions that consumers need when including cuopt headers.
+# These were on cuopt directly before the cuopt_objs refactor; $<TARGET_OBJECTS:...>
+# does not carry INTERFACE properties, so we restore them explicitly.
+target_compile_definitions(cuopt
+        PUBLIC
+        "CUOPT_LOG_ACTIVE_LEVEL=RAPIDS_LOGGER_LOG_LEVEL_${LIBCUOPT_LOGGING_LEVEL}"
+        CUSPARSE_ENABLE_EXPERIMENTAL_API
+)
 
 if (WRITE_FATBIN)
     file(WRITE "${CUOPT_BINARY_DIR}/fatbin.ld"
diff --git a/cpp/tests/internal/CMakeLists.txt b/cpp/tests/internal/CMakeLists.txt
index 5e2441ad92..fb24cae936 100644
--- a/cpp/tests/internal/CMakeLists.txt
+++ b/cpp/tests/internal/CMakeLists.txt
@@ -5,7 +5,12 @@
 
 # Single binary for all numopt tests that require internal symbols (cuopt_static).
 # Add new internal numopt tests here instead of creating separate STATIC_LIB targets.
+#
+# main.cu provides the single main() entry point (RMM-aware, same as CUOPT_TEST_PROGRAM_MAIN).
+# CUOPT_DISABLE_TEST_MAIN suppresses CUOPT_TEST_PROGRAM_MAIN() in individual test files
+# so only main.cu defines main().
 ConfigureTest(NUMOPT_INTERNAL_TEST
+    ${CUOPT_TEST_DIR}/internal/main.cu
     # dual_simplex
     ${CUOPT_TEST_DIR}/dual_simplex/unit_tests/solve.cpp
     ${CUOPT_TEST_DIR}/dual_simplex/unit_tests/solve_barrier.cu
@@ -26,8 +31,6 @@ ConfigureTest(NUMOPT_INTERNAL_TEST
     ${CUOPT_TEST_DIR}/mip/integer_with_real_bounds.cu
     ${CUOPT_TEST_DIR}/mip/empty_fixed_problems_test.cu
     ${CUOPT_TEST_DIR}/mip/presolve_test.cu
-    ${CUOPT_TEST_DIR}/mip/feasibility_jump_tests.cu
-    ${CUOPT_TEST_DIR}/mip/load_balancing_test.cu
     ${CUOPT_TEST_DIR}/mip/termination_test.cu
     ${CUOPT_TEST_DIR}/mip/determinism_test.cu
     # socp
@@ -36,3 +39,6 @@ ConfigureTest(NUMOPT_INTERNAL_TEST
     ${CUOPT_TEST_DIR}/socp/general_quadratic_test.cu
     LABELS numopt
     STATIC_LIB)
+
+# Suppress individual main() definitions — main.cu is the single entry point.
+target_compile_definitions(NUMOPT_INTERNAL_TEST PRIVATE CUOPT_DISABLE_TEST_MAIN)
diff --git a/cpp/tests/internal/main.cu b/cpp/tests/internal/main.cu
new file mode 100644
index 0000000000..9541c143fb
--- /dev/null
+++ b/cpp/tests/internal/main.cu
@@ -0,0 +1,20 @@
+/* clang-format off */
+/* SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+/* SPDX-License-Identifier: Apache-2.0
+/* clang-format on */
+
+// Single main() entry point for combined internal test binaries.
+// Individual test files use CUOPT_TEST_PROGRAM_MAIN() which is suppressed
+// (via CUOPT_DISABLE_TEST_MAIN) when they are compiled into a combined binary.
+#include <utilities/base_fixture.hpp>
+
+// Intentionally NOT suppressed here — this file owns the main() for the binary.
+int main(int argc, char** argv)
+{
+  ::testing::InitGoogleTest(&argc, argv);
+  auto const cmd_opts = parse_test_options(argc, argv);
+  auto const rmm_mode = cmd_opts["rmm_mode"].as<std::string>();
+  auto resource       = cuopt::test::create_memory_resource(rmm_mode);
+  rmm::mr::set_current_device_resource(resource);
+  return RUN_ALL_TESTS();
+}
diff --git a/cpp/tests/linear_programming/grpc/CMakeLists.txt b/cpp/tests/linear_programming/grpc/CMakeLists.txt
index 4261a7fe60..7f0d7a0e59 100644
--- a/cpp/tests/linear_programming/grpc/CMakeLists.txt
+++ b/cpp/tests/linear_programming/grpc/CMakeLists.txt
@@ -20,6 +20,8 @@ add_executable(GRPC_CLIENT_TEST
 target_include_directories(GRPC_CLIENT_TEST
     PRIVATE
     "${CUOPT_SOURCE_DIR}/include"
+    "${CUOPT_SOURCE_DIR}/src"
+    "${CUOPT_SOURCE_DIR}/src/io"
     "${CUOPT_SOURCE_DIR}/src/grpc"
     "${CUOPT_SOURCE_DIR}/src/grpc/client"
     "${CUOPT_SOURCE_DIR}/src/grpc/server"
@@ -108,6 +110,7 @@ add_executable(GRPC_INTEGRATION_TEST
 target_include_directories(GRPC_INTEGRATION_TEST
     PRIVATE
     "${CUOPT_SOURCE_DIR}/include"
+    "${CUOPT_SOURCE_DIR}/src"
     "${CUOPT_SOURCE_DIR}/src/grpc"
     "${CUOPT_SOURCE_DIR}/src/grpc/client"
     "${CUOPT_TEST_DIR}"
diff --git a/cpp/tests/mip/bounds_standardization_test.cu b/cpp/tests/mip/bounds_standardization_test.cu
index 5219366ce6..fffaec4989 100644
--- a/cpp/tests/mip/bounds_standardization_test.cu
+++ b/cpp/tests/mip/bounds_standardization_test.cu
@@ -31,7 +31,7 @@
 
 namespace cuopt::mathematical_optimization::test {
 
-void init_handler(const raft::handle_t* handle_ptr)
+static void init_handler(const raft::handle_t* handle_ptr)
 {
   // Init cuBlas / cuSparse context here to avoid having it during solving time
   RAFT_CUBLAS_TRY(raft::linalg::detail::cublassetpointermode(
diff --git a/cpp/tests/mip/elim_var_remap_test.cu b/cpp/tests/mip/elim_var_remap_test.cu
index eb6f2ae750..1cbc1cc60f 100644
--- a/cpp/tests/mip/elim_var_remap_test.cu
+++ b/cpp/tests/mip/elim_var_remap_test.cu
@@ -34,7 +34,7 @@
 
 namespace cuopt::mathematical_optimization::test {
 
-void init_handler(const raft::handle_t* handle_ptr)
+static void init_handler(const raft::handle_t* handle_ptr)
 {
   // Init cuBlas / cuSparse context here to avoid having it during solving time
   RAFT_CUBLAS_TRY(raft::linalg::detail::cublassetpointermode(
diff --git a/cpp/tests/mip/multi_probe_test.cu b/cpp/tests/mip/multi_probe_test.cu
index d7241f9325..9438bf6183 100644
--- a/cpp/tests/mip/multi_probe_test.cu
+++ b/cpp/tests/mip/multi_probe_test.cu
@@ -33,7 +33,7 @@ namespace cuopt::mathematical_optimization::test {
 
 inline auto make_async() { return rmm::mr::cuda_async_memory_resource(); }
 
-void init_handler(const raft::handle_t* handle_ptr)
+static void init_handler(const raft::handle_t* handle_ptr)
 {
   // Init cuBlas / cuSparse context here to avoid having it during solving time
   RAFT_CUBLAS_TRY(raft::linalg::detail::cublassetpointermode(
diff --git a/cpp/tests/routing/CMakeLists.txt b/cpp/tests/routing/CMakeLists.txt
index d6edc8c6e0..84b8aefa30 100644
--- a/cpp/tests/routing/CMakeLists.txt
+++ b/cpp/tests/routing/CMakeLists.txt
@@ -7,13 +7,18 @@ ConfigureTest(ROUTING_TEST ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_routing_test.cu
     LABELS routing)
 
 ConfigureTest(ROUTING_INTERNAL_TEST
+    ${CUOPT_TEST_DIR}/internal/main.cu
     ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_ges_test.cu
-    ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_scross_test.cu
-    ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/local_search_cand_test.cu
     ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/top_k.cu
+    ${CMAKE_CURRENT_SOURCE_DIR}/utilities/check_constraints.cu
     LABELS routing
     STATIC_LIB)
 
+# Suppress individual main() definitions — main.cu is the single entry point.
+# check_constraints.cu provides check_route used by l0_ges_test.cu directly
+# to avoid linking cuopttestutils (which depends on cuopt shared lib).
+target_compile_definitions(ROUTING_INTERNAL_TEST PRIVATE CUOPT_DISABLE_TEST_MAIN)
+
 ConfigureTest(VEHICLE_ORDER_TEST ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_vehicle_order_match.cu
     LABELS routing)
 ConfigureTest(VEHICLE_TYPES_TEST ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_vehicle_types_test.cu
diff --git a/cpp/tests/utilities/base_fixture.hpp b/cpp/tests/utilities/base_fixture.hpp
index 31d7923dfa..4739ae9926 100644
--- a/cpp/tests/utilities/base_fixture.hpp
+++ b/cpp/tests/utilities/base_fixture.hpp
@@ -111,7 +111,12 @@ inline auto parse_test_options(int argc, char** argv)
  * the original functionality. In addition, this custom `main` function parses
  * the command line to customize test behavior, like the allocation mode used
  * for creating the default memory resource.
+ *
+ * When CUOPT_DISABLE_TEST_MAIN is defined (e.g. when compiling multiple test
+ * files into a single combined binary), this macro expands to nothing so that
+ * only one translation unit provides main().
  */
+#ifndef CUOPT_DISABLE_TEST_MAIN
 #define CUOPT_TEST_PROGRAM_MAIN()                                        \
   int main(int argc, char** argv)                                        \
   {                                                                      \
@@ -122,3 +127,6 @@ inline auto parse_test_options(int argc, char** argv)
     rmm::mr::set_current_device_resource(resource);                      \
     return RUN_ALL_TESTS();                                              \
   }
+#else
+#define CUOPT_TEST_PROGRAM_MAIN()
+#endif

From a59bd8cc3ca2a140be0cf1e6fc606c29bc2c4ce9 Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Thu, 16 Jul 2026 14:55:03 -0500
Subject: [PATCH 04/18] style(tests): remove stale comments from empty
 CMakeLists stubs

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/tests/dual_simplex/CMakeLists.txt | 2 --
 cpp/tests/socp/CMakeLists.txt         | 2 --
 2 files changed, 4 deletions(-)

diff --git a/cpp/tests/dual_simplex/CMakeLists.txt b/cpp/tests/dual_simplex/CMakeLists.txt
index 8c7cd7005a..6fe84e33af 100644
--- a/cpp/tests/dual_simplex/CMakeLists.txt
+++ b/cpp/tests/dual_simplex/CMakeLists.txt
@@ -2,5 +2,3 @@
 # SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 # cmake-format: on
-
-# Internal tests consolidated into NUMOPT_INTERNAL_TEST (cpp/tests/internal/CMakeLists.txt).
diff --git a/cpp/tests/socp/CMakeLists.txt b/cpp/tests/socp/CMakeLists.txt
index 9490527469..73e0084147 100644
--- a/cpp/tests/socp/CMakeLists.txt
+++ b/cpp/tests/socp/CMakeLists.txt
@@ -2,5 +2,3 @@
 # SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 # cmake-format: on
-
-# Internal tests consolidated into NUMOPT_INTERNAL_TEST (cpp/tests/internal/CMakeLists.txt).

From bd2774526e91090f9a6bb005512a3e8bfd267966 Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Thu, 16 Jul 2026 15:07:19 -0500
Subject: [PATCH 05/18] refactor(tests): remove empty dual_simplex and socp
 CMakeLists stubs

All tests from these directories moved to NUMOPT_INTERNAL_TEST.
The subdirectories and add_subdirectory calls serve no purpose now.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/tests/CMakeLists.txt              | 2 --
 cpp/tests/dual_simplex/CMakeLists.txt | 4 ----
 cpp/tests/socp/CMakeLists.txt         | 4 ----
 3 files changed, 10 deletions(-)
 delete mode 100644 cpp/tests/dual_simplex/CMakeLists.txt
 delete mode 100644 cpp/tests/socp/CMakeLists.txt

diff --git a/cpp/tests/CMakeLists.txt b/cpp/tests/CMakeLists.txt
index b3d001c1c6..a5f0cf6e49 100644
--- a/cpp/tests/CMakeLists.txt
+++ b/cpp/tests/CMakeLists.txt
@@ -107,8 +107,6 @@ if(BUILD_TESTS)
     add_subdirectory(mip)
   endif()
   add_subdirectory(linear_programming)
-  add_subdirectory(dual_simplex)
-  add_subdirectory(socp)
   add_subdirectory(internal)
   add_subdirectory(qp)
   add_subdirectory(utilities)
diff --git a/cpp/tests/dual_simplex/CMakeLists.txt b/cpp/tests/dual_simplex/CMakeLists.txt
deleted file mode 100644
index 6fe84e33af..0000000000
--- a/cpp/tests/dual_simplex/CMakeLists.txt
+++ /dev/null
@@ -1,4 +0,0 @@
-# cmake-format: off
-# SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-# SPDX-License-Identifier: Apache-2.0
-# cmake-format: on
diff --git a/cpp/tests/socp/CMakeLists.txt b/cpp/tests/socp/CMakeLists.txt
deleted file mode 100644
index 73e0084147..0000000000
--- a/cpp/tests/socp/CMakeLists.txt
+++ /dev/null
@@ -1,4 +0,0 @@
-# cmake-format: off
-# SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-# SPDX-License-Identifier: Apache-2.0
-# cmake-format: on

From 0e5689a231029e2cc4fa7e94f19d372664c20729 Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Thu, 16 Jul 2026 15:16:59 -0500
Subject: [PATCH 06/18] fix(build): link PSLP directly on cuopt shared target

$<TARGET_OBJECTS:cuopt_objs> does not propagate PRIVATE link
dependencies to consuming targets; PSLP must be linked explicitly
on both cuopt_static and cuopt (shared).

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/CMakeLists.txt | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
index a0117fb7ba..fe4cddcd6e 100644
--- a/cpp/CMakeLists.txt
+++ b/cpp/CMakeLists.txt
@@ -699,6 +699,8 @@ target_link_libraries(cuopt
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:protobuf::libprotobuf>
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:gRPC::grpc++>
 )
+target_link_libraries(cuopt PRIVATE $<TARGET_FILE:PSLP>)
+add_dependencies(cuopt PSLP)
 # Propagate compile definitions that consumers need when including cuopt headers.
 # These were on cuopt directly before the cuopt_objs refactor; $<TARGET_OBJECTS:...>
 # does not carry INTERFACE properties, so we restore them explicitly.

From 90e5d7904c87d325b7516a52fe2d9b5eeb1c7a37 Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Thu, 16 Jul 2026 17:38:47 -0500
Subject: [PATCH 07/18] refactor(tests): delete CUOPT_TEST_PROGRAM_MAIN from
 combined-binary test files

Files compiled into NUMOPT_INTERNAL_TEST and ROUTING_INTERNAL_TEST no
longer need their own main(); remove the call directly rather than
suppressing it via the CUOPT_DISABLE_TEST_MAIN guard. Simplifies
base_fixture.hpp and the CMakeLists files.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/tests/internal/CMakeLists.txt                        | 7 -------
 cpp/tests/internal/main.cu                               | 4 ----
 cpp/tests/linear_programming/pdlp_test.cu                | 2 --
 cpp/tests/linear_programming/unit_tests/presolve_test.cu | 2 --
 cpp/tests/routing/CMakeLists.txt                         | 5 -----
 cpp/tests/routing/level0/l0_ges_test.cu                  | 2 --
 cpp/tests/routing/unit_tests/top_k.cu                    | 4 +---
 cpp/tests/utilities/base_fixture.hpp                     | 7 -------
 8 files changed, 1 insertion(+), 32 deletions(-)

diff --git a/cpp/tests/internal/CMakeLists.txt b/cpp/tests/internal/CMakeLists.txt
index fb24cae936..bb343366a8 100644
--- a/cpp/tests/internal/CMakeLists.txt
+++ b/cpp/tests/internal/CMakeLists.txt
@@ -5,10 +5,6 @@
 
 # Single binary for all numopt tests that require internal symbols (cuopt_static).
 # Add new internal numopt tests here instead of creating separate STATIC_LIB targets.
-#
-# main.cu provides the single main() entry point (RMM-aware, same as CUOPT_TEST_PROGRAM_MAIN).
-# CUOPT_DISABLE_TEST_MAIN suppresses CUOPT_TEST_PROGRAM_MAIN() in individual test files
-# so only main.cu defines main().
 ConfigureTest(NUMOPT_INTERNAL_TEST
     ${CUOPT_TEST_DIR}/internal/main.cu
     # dual_simplex
@@ -39,6 +35,3 @@ ConfigureTest(NUMOPT_INTERNAL_TEST
     ${CUOPT_TEST_DIR}/socp/general_quadratic_test.cu
     LABELS numopt
     STATIC_LIB)
-
-# Suppress individual main() definitions — main.cu is the single entry point.
-target_compile_definitions(NUMOPT_INTERNAL_TEST PRIVATE CUOPT_DISABLE_TEST_MAIN)
diff --git a/cpp/tests/internal/main.cu b/cpp/tests/internal/main.cu
index 9541c143fb..869f2b2bec 100644
--- a/cpp/tests/internal/main.cu
+++ b/cpp/tests/internal/main.cu
@@ -3,12 +3,8 @@
 /* SPDX-License-Identifier: Apache-2.0
 /* clang-format on */
 
-// Single main() entry point for combined internal test binaries.
-// Individual test files use CUOPT_TEST_PROGRAM_MAIN() which is suppressed
-// (via CUOPT_DISABLE_TEST_MAIN) when they are compiled into a combined binary.
 #include <utilities/base_fixture.hpp>
 
-// Intentionally NOT suppressed here — this file owns the main() for the binary.
 int main(int argc, char** argv)
 {
   ::testing::InitGoogleTest(&argc, argv);
diff --git a/cpp/tests/linear_programming/pdlp_test.cu b/cpp/tests/linear_programming/pdlp_test.cu
index 33ac914c35..5a3e1c40fe 100644
--- a/cpp/tests/linear_programming/pdlp_test.cu
+++ b/cpp/tests/linear_programming/pdlp_test.cu
@@ -4765,5 +4765,3 @@ TEST(pdlp_class, batch_with_optimal_size_query)
 }
 
 }  // namespace cuopt::mathematical_optimization::test
-
-CUOPT_TEST_PROGRAM_MAIN()
diff --git a/cpp/tests/linear_programming/unit_tests/presolve_test.cu b/cpp/tests/linear_programming/unit_tests/presolve_test.cu
index 75e0565d67..32bfc12aab 100644
--- a/cpp/tests/linear_programming/unit_tests/presolve_test.cu
+++ b/cpp/tests/linear_programming/unit_tests/presolve_test.cu
@@ -1062,5 +1062,3 @@ TEST(submip_presolve, ex9_fully_reduced)
 }
 
 }  // namespace cuopt::mathematical_optimization::test
-
-CUOPT_TEST_PROGRAM_MAIN()
diff --git a/cpp/tests/routing/CMakeLists.txt b/cpp/tests/routing/CMakeLists.txt
index 84b8aefa30..15cbbaffd1 100644
--- a/cpp/tests/routing/CMakeLists.txt
+++ b/cpp/tests/routing/CMakeLists.txt
@@ -14,11 +14,6 @@ ConfigureTest(ROUTING_INTERNAL_TEST
     LABELS routing
     STATIC_LIB)
 
-# Suppress individual main() definitions — main.cu is the single entry point.
-# check_constraints.cu provides check_route used by l0_ges_test.cu directly
-# to avoid linking cuopttestutils (which depends on cuopt shared lib).
-target_compile_definitions(ROUTING_INTERNAL_TEST PRIVATE CUOPT_DISABLE_TEST_MAIN)
-
 ConfigureTest(VEHICLE_ORDER_TEST ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_vehicle_order_match.cu
     LABELS routing)
 ConfigureTest(VEHICLE_TYPES_TEST ${CMAKE_CURRENT_SOURCE_DIR}/level0/l0_vehicle_types_test.cu
diff --git a/cpp/tests/routing/level0/l0_ges_test.cu b/cpp/tests/routing/level0/l0_ges_test.cu
index b3e72c56e6..47944c478d 100644
--- a/cpp/tests/routing/level0/l0_ges_test.cu
+++ b/cpp/tests/routing/level0/l0_ges_test.cu
@@ -246,5 +246,3 @@ INSTANTIATE_TEST_SUITE_P(level0_ges, double_test_vrp, ::testing::Values(std::mak
 }  // namespace test
 }  // namespace routing
 }  // namespace cuopt
-
-CUOPT_TEST_PROGRAM_MAIN()
diff --git a/cpp/tests/routing/unit_tests/top_k.cu b/cpp/tests/routing/unit_tests/top_k.cu
index e402c1ec43..c6d377a63f 100644
--- a/cpp/tests/routing/unit_tests/top_k.cu
+++ b/cpp/tests/routing/unit_tests/top_k.cu
@@ -1,6 +1,6 @@
 /* clang-format off */
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2024-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 /* clang-format on */
@@ -322,5 +322,3 @@ INSTANTIATE_TEST_SUITE_P(top_k,
 }  // namespace test
 }  // namespace routing
 }  // namespace cuopt
-
-CUOPT_TEST_PROGRAM_MAIN()
diff --git a/cpp/tests/utilities/base_fixture.hpp b/cpp/tests/utilities/base_fixture.hpp
index 4739ae9926..28283ac2d9 100644
--- a/cpp/tests/utilities/base_fixture.hpp
+++ b/cpp/tests/utilities/base_fixture.hpp
@@ -112,11 +112,7 @@ inline auto parse_test_options(int argc, char** argv)
  * the command line to customize test behavior, like the allocation mode used
  * for creating the default memory resource.
  *
- * When CUOPT_DISABLE_TEST_MAIN is defined (e.g. when compiling multiple test
- * files into a single combined binary), this macro expands to nothing so that
- * only one translation unit provides main().
  */
-#ifndef CUOPT_DISABLE_TEST_MAIN
 #define CUOPT_TEST_PROGRAM_MAIN()                                        \
   int main(int argc, char** argv)                                        \
   {                                                                      \
@@ -127,6 +123,3 @@ inline auto parse_test_options(int argc, char** argv)
     rmm::mr::set_current_device_resource(resource);                      \
     return RUN_ALL_TESTS();                                              \
   }
-#else
-#define CUOPT_TEST_PROGRAM_MAIN()
-#endif

From 3e58684399e906f5587d99fbdf62c56a19fa0834 Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Fri, 17 Jul 2026 10:30:16 -0500
Subject: [PATCH 08/18] refactor(tests): replace runtime cuSPARSE check with
 compile-time macro

Remove cuOptIsCusparseRuntimeMixedPrecisionSupported from the public
C API. The c_api_test already includes internal headers; use the
compile-time CUSPARSE_VERSION macro instead, which is accurate for
cuOpt's deployment model (compile-time and runtime toolkit match).

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 .../cuopt/mathematical_optimization/cuopt_c.h  |  8 --------
 cpp/src/pdlp/cuopt_c.cpp                       | 11 -----------
 .../c_api_tests/c_api_tests.cpp                | 18 ++++++++++--------
 3 files changed, 10 insertions(+), 27 deletions(-)

diff --git a/cpp/include/cuopt/mathematical_optimization/cuopt_c.h b/cpp/include/cuopt/mathematical_optimization/cuopt_c.h
index d66c908f1d..218402a7ef 100644
--- a/cpp/include/cuopt/mathematical_optimization/cuopt_c.h
+++ b/cpp/include/cuopt/mathematical_optimization/cuopt_c.h
@@ -1055,14 +1055,6 @@ cuopt_int_t cuOptGetDualObjectiveValue(cuOptSolution solution,
  */
 cuopt_int_t cuOptGetReducedCosts(cuOptSolution solution, cuopt_float_t* reduced_cost_ptr);
 
-/**
- * @brief Query whether the runtime cuSPARSE version supports mixed-precision SpMV
- *        (FP32 matrix × FP64 vector). Requires cuSPARSE >= 12.5 at runtime.
- *
- * @return 1 if supported, 0 otherwise.
- */
-cuopt_int_t cuOptIsCusparseRuntimeMixedPrecisionSupported();
-
 #ifdef __cplusplus
 }
 #endif
diff --git a/cpp/src/pdlp/cuopt_c.cpp b/cpp/src/pdlp/cuopt_c.cpp
index ed49c9baf9..963bd6ada7 100644
--- a/cpp/src/pdlp/cuopt_c.cpp
+++ b/cpp/src/pdlp/cuopt_c.cpp
@@ -1296,14 +1296,3 @@ cuopt_int_t cuOptGetReducedCosts(cuOptSolution solution, cuopt_float_t* reduced_
     return CUOPT_INVALID_ARGUMENT;
   }
 }
-
-namespace cuopt::mathematical_optimization::pdlp {
-bool is_cusparse_runtime_mixed_precision_supported();
-}
-
-cuopt_int_t cuOptIsCusparseRuntimeMixedPrecisionSupported()
-{
-  return cuopt::mathematical_optimization::pdlp::is_cusparse_runtime_mixed_precision_supported()
-           ? 1
-           : 0;
-}
diff --git a/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp b/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
index 2778096119..ca8d700aca 100644
--- a/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
+++ b/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
@@ -17,6 +17,7 @@
 #include <pdlp/cuopt_c_internal.hpp>
 
 #include <cuda_runtime.h>
+#include <cusparse.h>
 
 #include <utilities/common_utils.hpp>
 #include <utilities/error.hpp>
@@ -441,18 +442,19 @@ TEST(c_api, pdlp_precision_mixed)
   std::string filename           = rapidsDatasetRootDir + "/linear_programming/afiro_original.mps";
   cuopt_int_t termination_status = -1;
   cuopt_float_t objective;
-  if (!cuOptIsCusparseRuntimeMixedPrecisionSupported()) {
-    auto status = test_pdlp_precision_mixed(filename.c_str(), &termination_status, &objective);
-    bool solve_returned_error = (status != CUOPT_SUCCESS);
-    bool solve_returned_non_optimal =
-      (status == CUOPT_SUCCESS && termination_status != CUOPT_TERMINATION_STATUS_OPTIMAL);
-    EXPECT_TRUE(solve_returned_error || solve_returned_non_optimal);
-    return;
-  }
+  // Mixed-precision SpMV (FP32 matrix × FP64 vector) requires cuSPARSE >= 12.5.
+#if CUSPARSE_VERSION >= 12500
   EXPECT_EQ(test_pdlp_precision_mixed(filename.c_str(), &termination_status, &objective),
             CUOPT_SUCCESS);
   EXPECT_EQ(termination_status, CUOPT_TERMINATION_STATUS_OPTIMAL);
   EXPECT_NEAR(objective, -464.7531, 1e-1);
+#else
+  auto status = test_pdlp_precision_mixed(filename.c_str(), &termination_status, &objective);
+  bool solve_returned_error = (status != CUOPT_SUCCESS);
+  bool solve_returned_non_optimal =
+    (status == CUOPT_SUCCESS && termination_status != CUOPT_TERMINATION_STATUS_OPTIMAL);
+  EXPECT_TRUE(solve_returned_error || solve_returned_non_optimal);
+#endif
 }
 
 // =============================================================================

From c9476b2fbafda15fd28bff84c8e0325ea91fc453 Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Fri, 17 Jul 2026 10:40:20 -0500
Subject: [PATCH 09/18] fix(tests): require error when mixed precision is
 unsupported
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

When cuSPARSE < 12.5, cuopt_expects throws ValidationError before
the solve begins, so the C API always returns an error — CUOPT_SUCCESS
with non-optimal is impossible. Assert status != CUOPT_SUCCESS directly.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp b/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
index ca8d700aca..5d6f315a5d 100644
--- a/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
+++ b/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
@@ -449,11 +449,10 @@ TEST(c_api, pdlp_precision_mixed)
   EXPECT_EQ(termination_status, CUOPT_TERMINATION_STATUS_OPTIMAL);
   EXPECT_NEAR(objective, -464.7531, 1e-1);
 #else
-  auto status = test_pdlp_precision_mixed(filename.c_str(), &termination_status, &objective);
-  bool solve_returned_error = (status != CUOPT_SUCCESS);
-  bool solve_returned_non_optimal =
-    (status == CUOPT_SUCCESS && termination_status != CUOPT_TERMINATION_STATUS_OPTIMAL);
-  EXPECT_TRUE(solve_returned_error || solve_returned_non_optimal);
+  // cuopt_expects throws ValidationError when mixed precision is requested without support,
+  // so the C API always returns an error code — never CUOPT_SUCCESS.
+  EXPECT_NE(test_pdlp_precision_mixed(filename.c_str(), &termination_status, &objective),
+            CUOPT_SUCCESS);
 #endif
 }
 

From 1f52c5e3da3a463c62f19bf8fd7b639c7faa3857 Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Fri, 17 Jul 2026 14:39:16 -0500
Subject: [PATCH 10/18] fix(tests): fix invalid CSR in problem_test
 create_problem helper

After thrust::exclusive_scan, offsets.back() already equals the total
number of non-zeros (standard CSR sentinel). The previous + 1 produced
one entry beyond the last row, which is invalid CSR and caused
cusparseXcsrsort to produce incorrect results when assertions are
enabled (PR CI uses -a / DEFINE_ASSERT=ON which enables -UNDEBUG).

The resulting out-of-bounds column index triggered the
check_csr_representation assertion in compute_transpose_of_problem,
crashing NUMOPT_INTERNAL_TEST with SIGABRT in PR CI.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/tests/mip/problem_test.cu | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/cpp/tests/mip/problem_test.cu b/cpp/tests/mip/problem_test.cu
index 1504df1ad4..a53d1ee44b 100644
--- a/cpp/tests/mip/problem_test.cu
+++ b/cpp/tests/mip/problem_test.cu
@@ -109,7 +109,7 @@ lp::optimization_problem_t<i_t, f_t> create_problem(raft::handle_t const* h, i_t
   auto offsets = rand_vec<i_t, i_t>(n_cnst + 1, average_vars_per_cnst, average_vars_per_cnst * 5);
   thrust::exclusive_scan(thrust::host, offsets.begin(), offsets.end(), offsets.begin());
 
-  i_t nnz = offsets.back() + 1;
+  i_t nnz = offsets.back();
 
   // a_values - non-zero coefficients
   auto coeff = rand_vec<i_t, f_t>(nnz, 0.2, 10.0);

From 545ce45983ce3c99b839966c947758be17934e2d Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Fri, 17 Jul 2026 16:02:26 -0500
Subject: [PATCH 11/18] fix(test): check runtime cuSPARSE version in
 pdlp_precision_mixed test

The compile-time #if CUSPARSE_VERSION >= 12500 guard only checks the
header version used during build. On systems where the runtime cuSPARSE
library is older than 12.5 (e.g., oldest-deps conda builds), the solver
correctly throws ValidationError via is_cusparse_runtime_mixed_precision_supported(),
but the test unconditionally expected CUOPT_SUCCESS.

Fix by calling cusparseGetProperty() at runtime in the test to mirror
the same check the solver performs, routing the test to the success or
failure assertion accordingly.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 .../c_api_tests/c_api_tests.cpp               | 23 +++++++++++++++----
 1 file changed, 18 insertions(+), 5 deletions(-)

diff --git a/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp b/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
index 5d6f315a5d..e90a912885 100644
--- a/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
+++ b/cpp/tests/linear_programming/c_api_tests/c_api_tests.cpp
@@ -442,12 +442,25 @@ TEST(c_api, pdlp_precision_mixed)
   std::string filename           = rapidsDatasetRootDir + "/linear_programming/afiro_original.mps";
   cuopt_int_t termination_status = -1;
   cuopt_float_t objective;
-  // Mixed-precision SpMV (FP32 matrix × FP64 vector) requires cuSPARSE >= 12.5.
+  // Mixed-precision SpMV (FP32 matrix × FP64 vector) requires cuSPARSE >= 12.5 at BOTH
+  // compile time (header) and runtime (loaded .so). The header version (#if) guards symbol
+  // availability; the runtime check below mirrors is_cusparse_runtime_mixed_precision_supported().
 #if CUSPARSE_VERSION >= 12500
-  EXPECT_EQ(test_pdlp_precision_mixed(filename.c_str(), &termination_status, &objective),
-            CUOPT_SUCCESS);
-  EXPECT_EQ(termination_status, CUOPT_TERMINATION_STATUS_OPTIMAL);
-  EXPECT_NEAR(objective, -464.7531, 1e-1);
+  int cusparseMajor = 0, cusparseMinor = 0;
+  cusparseGetProperty(MAJOR_VERSION, &cusparseMajor);
+  cusparseGetProperty(MINOR_VERSION, &cusparseMinor);
+  bool runtimeSupported = (cusparseMajor > 12) || (cusparseMajor == 12 && cusparseMinor >= 5);
+  if (runtimeSupported) {
+    EXPECT_EQ(test_pdlp_precision_mixed(filename.c_str(), &termination_status, &objective),
+              CUOPT_SUCCESS);
+    EXPECT_EQ(termination_status, CUOPT_TERMINATION_STATUS_OPTIMAL);
+    EXPECT_NEAR(objective, -464.7531, 1e-1);
+  } else {
+    // cuopt_expects throws ValidationError when mixed precision is requested without runtime
+    // support, so the C API always returns an error code — never CUOPT_SUCCESS.
+    EXPECT_NE(test_pdlp_precision_mixed(filename.c_str(), &termination_status, &objective),
+              CUOPT_SUCCESS);
+  }
 #else
   // cuopt_expects throws ValidationError when mixed precision is requested without support,
   // so the C API always returns an error code — never CUOPT_SUCCESS.

From 0effbce469b133d288802c295ffa7b09e7ed18f4 Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Mon, 20 Jul 2026 09:53:12 -0500
Subject: [PATCH 12/18] fix(tests): restore RMM setup in ROUTING_UNIT_TEST

top_k.cu was moved to ROUTING_INTERNAL_TEST because it uses internal
symbols (routing/detail). It was also the only source in ROUTING_UNIT_TEST
that provided CUOPT_TEST_PROGRAM_MAIN(), which sets up the RMM device
memory resource before running tests.

Without RMM setup, the routing solver allocations use the default
cuda_memory_resource, causing SIGSEGV in vehicle_breaks.non_uniform_breaks
(and potentially other tests that stress the allocator).

Fix by adding internal/main.cu to ROUTING_UNIT_TEST sources. The
custom main() it provides takes precedence over GTest::gmock_main's
default main and properly initializes the RMM memory resource.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/tests/routing/CMakeLists.txt | 1 +
 1 file changed, 1 insertion(+)

diff --git a/cpp/tests/routing/CMakeLists.txt b/cpp/tests/routing/CMakeLists.txt
index 15cbbaffd1..c544267797 100644
--- a/cpp/tests/routing/CMakeLists.txt
+++ b/cpp/tests/routing/CMakeLists.txt
@@ -29,6 +29,7 @@ set_tests_properties(RETAIL_L1TEST ROUTING_L1TEST PROPERTIES DISABLED TRUE)
 
 # # - ${CMAKE_CURRENT_SOURCE_DIR} unit tests ----------------------------------------------------------------------------
 ConfigureTest(ROUTING_UNIT_TEST
+      ${CUOPT_TEST_DIR}/internal/main.cu
       ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/vehicle_types.cu
       ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/breaks.cu
       ${CMAKE_CURRENT_SOURCE_DIR}/unit_tests/heterogenous_breaks.cu

From 755ec5d876f7e273e64575b9c4250aa5d02f9e4d Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Mon, 20 Jul 2026 10:42:32 -0500
Subject: [PATCH 13/18] fix(tests): fix malformed SPDX block comment in
 internal/main.cu

The SPDX header used bare `/* text` lines without closing `*/` on each
line, leaving an unclosed C comment. Subsequent `/*` markers triggered
-Werror=comment when internal/main.cu was compiled into new targets
(e.g. ROUTING_UNIT_TEST). Use the standard block-comment style matching
the rest of the test suite.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/tests/internal/main.cu | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/cpp/tests/internal/main.cu b/cpp/tests/internal/main.cu
index 869f2b2bec..f79fbc5201 100644
--- a/cpp/tests/internal/main.cu
+++ b/cpp/tests/internal/main.cu
@@ -1,6 +1,8 @@
 /* clang-format off */
-/* SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-/* SPDX-License-Identifier: Apache-2.0
+/*
+ * SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-License-Identifier: Apache-2.0
+ */
 /* clang-format on */
 
 #include <utilities/base_fixture.hpp>

From 8f23153cdd9048eb875868cbc87e7f70a58f7c1a Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Mon, 20 Jul 2026 14:06:34 -0500
Subject: [PATCH 14/18] fix(build): add OpenMP::OpenMP_CUDA to cuopt_objs
 PRIVATE deps

cuopt_objs is the OBJECT library that compiles all CUDA TUs. Before
the cuopt_objs refactor, cuopt SHARED had OpenMP::OpenMP_CUDA as a
PRIVATE dep, which passed -fopenmp to nvcc and defined _OPENMP for
all .cu files. After the refactor, cuopt_objs only inherited
OpenMP::OpenMP_CXX (via CUOPT_PRIVATE_CUDA_LIBS), which only applies
to CXX compilation. Without -fopenmp, omp_atomic_t<T> is undefined
in CUDA TUs that include omp_helpers.hpp (e.g. via
dual_simplex/solution.hpp), causing build failures.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/CMakeLists.txt | 1 +
 1 file changed, 1 insertion(+)

diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
index fe4cddcd6e..e27a657d63 100644
--- a/cpp/CMakeLists.txt
+++ b/cpp/CMakeLists.txt
@@ -619,6 +619,7 @@ target_link_libraries(cuopt_objs
         ${CUDSS_LIB_FILE}
         PRIVATE
         ${CUOPT_PRIVATE_CUDA_LIBS}
+        OpenMP::OpenMP_CUDA
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:protobuf::libprotobuf>
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:gRPC::grpc++>
 )

From c1598967acc32ff1e346eac8e7b83d41d9846ee9 Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Tue, 21 Jul 2026 10:06:13 -0500
Subject: [PATCH 15/18] fix(build): add simde::simde to cuopt_objs PRIVATE deps

The fast MPS parser sources include simde/x86/avx2.h. Before the
cuopt_objs refactor, simde::simde was a PRIVATE dep of cuopt SHARED
and its include path was available during source compilation. After
the refactor, sources are compiled in cuopt_objs, so simde::simde
must be listed there.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 cpp/CMakeLists.txt | 1 +
 1 file changed, 1 insertion(+)

diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
index e27a657d63..b17b8befec 100644
--- a/cpp/CMakeLists.txt
+++ b/cpp/CMakeLists.txt
@@ -619,6 +619,7 @@ target_link_libraries(cuopt_objs
         ${CUDSS_LIB_FILE}
         PRIVATE
         ${CUOPT_PRIVATE_CUDA_LIBS}
+        simde::simde
         OpenMP::OpenMP_CUDA
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:protobuf::libprotobuf>
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:gRPC::grpc++>

From 3a55206fac0d39db98fd332bb209d09047730e5e Mon Sep 17 00:00:00 2001
From: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Date: Tue, 21 Jul 2026 14:43:10 -0500
Subject: [PATCH 16/18] fix(tests): add missing symmetric QCMATRIX entry in
 fast parser edge test

QcMatrixRowsMatchReferenceBitwise provided X1 X2 -2.5 but not the
required matching X2 X1 -2.5. The classic MPS parser enforces that
each off-diagonal (i,j) entry has an explicit (j,i) counterpart;
verify_fixture_bitwise routes through both parsers and the reference
parser was throwing ValidationError on the asymmetric input.

Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
---
 .../experimental_mps_fast/fast_parser_edge_test.cpp            | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/cpp/tests/linear_programming/experimental_mps_fast/fast_parser_edge_test.cpp b/cpp/tests/linear_programming/experimental_mps_fast/fast_parser_edge_test.cpp
index 69cc29a0d3..48fd77ae35 100644
--- a/cpp/tests/linear_programming/experimental_mps_fast/fast_parser_edge_test.cpp
+++ b/cpp/tests/linear_programming/experimental_mps_fast/fast_parser_edge_test.cpp
@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+// SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 // SPDX-License-Identifier: Apache-2.0
 
 #include "fast_parser.hpp"
@@ -853,6 +853,7 @@ TEST(FastMpsParserEdgeTest, QcMatrixRowsMatchReferenceBitwise)
                          "QCMATRIX   QC1\n"
                          " X1 X1 1.25\n"
                          " X1 X2 -2.5\n"
+                         " X2 X1 -2.5\n"
                          "QCMATRIX   QC2\n"
                          " X2 X2 3.75\n"
                          "ENDATA\n");

From 6610d4448ba797d5b31914da1f7fcd163ad15842 Mon Sep 17 00:00:00 2001
From: Arha Gatram <agatram@nvidia.com>
Date: Thu, 23 Jul 2026 10:03:57 -0700
Subject: [PATCH 17/18] check symbols script

---
 ci/check_symbols.sh                | 84 ++++++++++++++++++++++++++++++
 conda/recipes/libcuopt/recipe.yaml |  4 +-
 2 files changed, 87 insertions(+), 1 deletion(-)
 create mode 100644 ci/check_symbols.sh

diff --git a/ci/check_symbols.sh b/ci/check_symbols.sh
new file mode 100644
index 0000000000..ed17b27213
--- /dev/null
+++ b/ci/check_symbols.sh
@@ -0,0 +1,84 @@
+#!/bin/bash
+# SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+set -eEuo pipefail
+
+echo "checking for symbol visibility issues"
+
+LIBRARY="${1}"
+
+echo ""
+echo "Checking exported symbols in '${LIBRARY}'"
+symbol_file="$(mktemp)"
+match_file="$(mktemp)"
+trap 'rm -f "${symbol_file}" "${match_file}"' EXIT
+
+# Ignore WEAK and UNIQUE symbols since UNIQUE symbols should be exported and
+# WEAK symbols may come from template instantiations.
+# Ignore symbols containing "_error" since these are likely exception types
+# and should be exported.
+
+readelf --dyn-syms --wide "${LIBRARY}" \
+    | awk '$7 != "UND" && $5 != "WEAK" && $5 != "UNIQUE"' \
+    | c++filt --no-params \
+    | awk '$0 !~ /_error/' \
+    > "${symbol_file}"
+
+patterns=(
+    'cub::'
+    'thrust::'
+    'raft::'
+    'rmm::'
+    'cuopt::mathematical_optimization::detail'
+    'cuopt::routing::detail'
+    'cuopt::detail'
+    'grpc::'
+    'google::protobuf'
+    'tbb::'
+    'absl::'
+    'dejavu::'
+    'papilo::'
+    'boost::'
+)
+
+failed=0
+
+for pattern in "${patterns[@]}"; do
+    echo "Checking for '${pattern}' symbols..."
+
+    awk -v pattern="${pattern}" '
+        BEGIN { has_trailing_scope = (substr(pattern, length(pattern) - 1) == "::") }
+        $1 ~ /^[0-9]+:/ {
+            symbol = ""
+            for (i = 8; i <= NF; ++i) {
+                symbol = symbol (i == 8 ? "" : " ") $i
+            }
+
+            sub(/<.*/, "", symbol)
+            sub(/^.*[[:space:]](for|to)[[:space:]]+/, "", symbol)
+
+            if (has_trailing_scope) {
+                matched = (index(symbol, pattern) == 1)
+            } else {
+                matched = (symbol == pattern || index(symbol, pattern "::") == 1)
+            }
+
+            if (matched) { print }
+        }
+    ' "${symbol_file}" > "${match_file}"
+
+    matches=$(awk 'END { print NR }' "${match_file}")
+    if [[ "${matches}" -ne 0 ]]; then
+        sed -n '1,20p' "${match_file}"
+        echo "ERROR: Found exported symbols in ${LIBRARY} matching the pattern ${pattern}."
+        echo "ERROR: Total matching symbols: ${matches}"
+        failed=1
+    fi
+done
+
+if [[ "${failed}" -ne 0 ]]; then
+    exit 1
+fi
+
+echo "No symbol visibility issues found in ${LIBRARY}"
diff --git a/conda/recipes/libcuopt/recipe.yaml b/conda/recipes/libcuopt/recipe.yaml
index 77c957eade..ab89215e13 100644
--- a/conda/recipes/libcuopt/recipe.yaml
+++ b/conda/recipes/libcuopt/recipe.yaml
@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2025-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 schema_version: 1
 
@@ -107,6 +107,7 @@ outputs:
       script:
         content: |
           cmake --install cpp/build
+          ./ci/check_symbols.sh cpp/build/libcuopt.so
       dynamic_linking:
         overlinking_behavior: "error"
       prefix_detection:
@@ -118,6 +119,7 @@ outputs:
       build:
         - cmake ${{ cmake_version }}
         - ${{ stdlib("c") }}
+        - binutils
       host:
         - libboost-devel
         - cuda-version =${{ cuda_version }}

From 2ca4e8076b9635ea42e3954a0639d6a4b1bc0a82 Mon Sep 17 00:00:00 2001
From: Arha Gatram <agatram@nvidia.com>
Date: Thu, 23 Jul 2026 12:09:37 -0700
Subject: [PATCH 18/18] Symbol visibility controls and exports

---
 cpp/CMakeLists.txt                            |  6 ++
 cpp/include/cuopt/error.hpp                   |  5 +-
 cpp/include/cuopt/export.hpp                  | 14 ++++
 cpp/include/cuopt/grpc/cython_grpc_client.hpp |  7 +-
 cpp/include/cuopt/grpc/grpc_client_env.hpp    |  7 +-
 .../backend_selection.hpp                     |  8 ++-
 .../cpu_optimization_problem.hpp              |  7 +-
 .../cpu_optimization_problem_solution.hpp     |  7 +-
 .../cpu_pdlp_warm_start_data.hpp              |  8 ++-
 .../cuopt/mathematical_optimization/cuopt_c.h |  9 +++
 .../io/data_model_view.hpp                    |  9 ++-
 .../io/mps_data_model.hpp                     | 10 ++-
 .../io/mps_writer.hpp                         |  9 ++-
 .../mathematical_optimization/io/parser.hpp   |  9 ++-
 .../io/utilities/cython_parser.hpp            |  5 +-
 .../mathematical_optimization/io/writer.hpp   |  9 ++-
 .../mip/solver_settings.hpp                   |  7 +-
 .../mip/solver_solution.hpp                   |  7 +-
 .../optimization_problem.hpp                  |  7 +-
 .../optimization_problem_solution.hpp         |  7 +-
 .../pdlp/pdlp_warm_start_data.hpp             |  8 ++-
 .../pdlp/solver_settings.hpp                  |  7 +-
 .../pdlp/solver_solution.hpp                  |  7 +-
 .../cuopt/mathematical_optimization/solve.hpp |  7 +-
 .../solve_remote.hpp                          |  7 +-
 .../solver_settings.hpp                       |  7 +-
 .../utilities/cython_solve.hpp                |  5 +-
 .../utilities/cython_types.hpp                |  5 +-
 cpp/include/cuopt/routing/assignment.hpp      |  7 +-
 cpp/include/cuopt/routing/cython/cython.hpp   |  5 +-
 cpp/include/cuopt/routing/data_model_view.hpp |  5 +-
 .../cuopt/routing/routing_structures.hpp      |  4 +-
 cpp/include/cuopt/routing/solve.hpp           |  7 +-
 cpp/include/cuopt/routing/solver_settings.hpp |  7 +-
 cpp/src/grpc/client/solve_remote.cpp          |  5 +-
 cpp/src/grpc/grpc_problem_mapper.cpp          | 50 ++++++++-------
 cpp/src/grpc/grpc_settings_mapper.cpp         | 42 +++++++-----
 cpp/src/grpc/grpc_solution_mapper.cpp         | 64 ++++++++++---------
 cpp/src/io/data_model_view.cpp                |  5 +-
 cpp/src/io/lp_parser.cpp                      | 11 ++--
 cpp/src/io/mps_data_model.cpp                 |  9 +--
 cpp/src/io/mps_writer.cpp                     |  5 +-
 cpp/src/io/parser.cpp                         | 23 ++++---
 cpp/src/io/writer.cpp                         |  9 +--
 cpp/src/math_optimization/solution_reader.hpp |  9 ++-
 cpp/src/math_optimization/solver_settings.cu  | 47 +++++++++-----
 cpp/src/mip_heuristics/solve.cu               |  9 +--
 cpp/src/mip_heuristics/solver_settings.cu     |  5 +-
 cpp/src/mip_heuristics/solver_solution.cu     |  5 +-
 cpp/src/pdlp/cpu_optimization_problem.cpp     |  5 +-
 cpp/src/pdlp/cpu_pdlp_warm_start_data.cu      |  9 +--
 cpp/src/pdlp/optimization_problem.cu          |  7 +-
 cpp/src/pdlp/pdlp_warm_start_data.cu          |  5 +-
 cpp/src/pdlp/solution_conversion.cu           |  9 +--
 cpp/src/pdlp/solve.cu                         | 14 ++--
 cpp/src/pdlp/solver_settings.cu               |  5 +-
 cpp/src/pdlp/solver_solution.cu               |  5 +-
 cpp/src/routing/assignment.cu                 |  5 +-
 cpp/src/routing/data_model_view.cu            |  3 +-
 .../distance_engine/waypoint_matrix.cpp       |  5 +-
 cpp/src/routing/solve.cu                      |  7 +-
 cpp/src/routing/solver_settings.cu            |  5 +-
 cpp/src/routing/utilities/cython.cu           |  3 +-
 cpp/src/utilities/logger.hpp                  | 12 ++--
 64 files changed, 415 insertions(+), 227 deletions(-)
 create mode 100644 cpp/include/cuopt/export.hpp

diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
index b17b8befec..9b0ff0a735 100644
--- a/cpp/CMakeLists.txt
+++ b/cpp/CMakeLists.txt
@@ -510,6 +510,8 @@ if (NOT SKIP_GRPC_BUILD)
     # at runtime with "undefined symbol: absl::…::Mutex::Dtor".
     set_property(SOURCE ${GRPC_INFRA_FILES} DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}
             APPEND PROPERTY COMPILE_OPTIONS "-DNDEBUG")
+    set_property(SOURCE ${PROTO_SRCS} ${GRPC_PROTO_SRCS} ${GRPC_SERVICE_SRCS} ${DATA_PROTO_SRCS} DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
+            APPEND PROPERTY COMPILE_OPTIONS "$<$<COMPILE_LANGUAGE:CXX>:-fvisibility=default>")
 endif (NOT SKIP_GRPC_BUILD)
 
 add_library(cuopt_objs OBJECT
@@ -518,6 +520,9 @@ add_library(cuopt_objs OBJECT
 
 set_target_properties(cuopt_objs
         PROPERTIES POSITION_INDEPENDENT_CODE ON
+        CXX_VISIBILITY_PRESET hidden
+        CUDA_VISIBILITY_PRESET hidden
+        VISIBILITY_INLINES_HIDDEN ON
         CXX_SCAN_FOR_MODULES OFF
 )
 
@@ -620,6 +625,7 @@ target_link_libraries(cuopt_objs
         PRIVATE
         ${CUOPT_PRIVATE_CUDA_LIBS}
         simde::simde
+        OpenMP::OpenMP_CXX
         OpenMP::OpenMP_CUDA
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:protobuf::libprotobuf>
         $<$<BOOL:${CUOPT_ENABLE_GRPC}>:gRPC::grpc++>
diff --git a/cpp/include/cuopt/error.hpp b/cpp/include/cuopt/error.hpp
index 95179c5ec4..40be060c5b 100644
--- a/cpp/include/cuopt/error.hpp
+++ b/cpp/include/cuopt/error.hpp
@@ -6,13 +6,14 @@
 /* clang-format on */
 #pragma once
 
+#include <cuopt/export.hpp>
 #include "cuopt/mathematical_optimization/constants.h"
 
 #include <stdarg.h>
 
 #include <raft/core/error.hpp>
 
-namespace cuopt {
+namespace CUOPT_EXPORT cuopt {
 
 /**
  * @brief Indicates different type of exceptions which cuOpt might throw
@@ -168,4 +169,4 @@ void execute_cuopt_fail(Args... args)
   throw cuopt::logic_error(msg, error_type_t::RuntimeError);
 }
 
-}  // namespace cuopt
+}  // namespace CUOPT_EXPORT cuopt
diff --git a/cpp/include/cuopt/export.hpp b/cpp/include/cuopt/export.hpp
new file mode 100644
index 0000000000..5f7209c5f5
--- /dev/null
+++ b/cpp/include/cuopt/export.hpp
@@ -0,0 +1,14 @@
+/* clang-format off */
+/*
+ * SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-License-Identifier: Apache-2.0
+ */
+/* clang-format on */
+
+#pragma once
+
+#if defined(__GNUC__) || defined(__clang__)
+#define CUOPT_EXPORT __attribute__((visibility("default")))
+#else
+#define CUOPT_EXPORT
+#endif
diff --git a/cpp/include/cuopt/grpc/cython_grpc_client.hpp b/cpp/include/cuopt/grpc/cython_grpc_client.hpp
index 593014cedc..3f6db62269 100644
--- a/cpp/include/cuopt/grpc/cython_grpc_client.hpp
+++ b/cpp/include/cuopt/grpc/cython_grpc_client.hpp
@@ -5,6 +5,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/utilities/cython_solve.hpp>
 
 #include <cstddef>
@@ -22,7 +23,8 @@ class data_model_view_t;
 }  // namespace io
 }  // namespace cuopt::mathematical_optimization
 
-namespace cuopt::cython {
+namespace cuopt {
+namespace CUOPT_EXPORT cython {
 
 /** Mirrors cuopt::mathematical_optimization::job_status_t for the Python bindings. */
 enum class grpc_job_status_t : int {
@@ -166,4 +168,5 @@ class grpc_python_client_t {
   std::unique_ptr<impl_t> impl_;
 };
 
-}  // namespace cuopt::cython
+}  // namespace CUOPT_EXPORT cython
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/grpc/grpc_client_env.hpp b/cpp/include/cuopt/grpc/grpc_client_env.hpp
index dab3dbe531..d8f88b7273 100644
--- a/cpp/include/cuopt/grpc/grpc_client_env.hpp
+++ b/cpp/include/cuopt/grpc/grpc_client_env.hpp
@@ -5,9 +5,11 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include "grpc_client.hpp"
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 /** How TLS is chosen when building a grpc_client_config_t. */
 enum class grpc_tls_mode_t {
@@ -49,4 +51,5 @@ grpc_client_config_t make_grpc_client_config(const std::string& host,
                                              grpc_tls_mode_t tls_mode,
                                              const grpc_explicit_tls_t* explicit_tls = nullptr);
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/backend_selection.hpp b/cpp/include/cuopt/mathematical_optimization/backend_selection.hpp
index f36b1d2bff..8c6bfeb73f 100644
--- a/cpp/include/cuopt/mathematical_optimization/backend_selection.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/backend_selection.hpp
@@ -7,7 +7,10 @@
 
 #pragma once
 
-namespace cuopt::mathematical_optimization {
+#include <cuopt/export.hpp>
+
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 /**
  * @brief Enum for execution mode (local vs remote solve)
@@ -51,4 +54,5 @@ execution_mode_t get_execution_mode();
  */
 memory_backend_t get_memory_backend_type();
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/cpu_optimization_problem.hpp b/cpp/include/cuopt/mathematical_optimization/cpu_optimization_problem.hpp
index e4ebcc35b8..28aa91a82f 100644
--- a/cpp/include/cuopt/mathematical_optimization/cpu_optimization_problem.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/cpu_optimization_problem.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/optimization_problem_interface.hpp>
 
 #include <raft/core/handle.hpp>
@@ -18,7 +19,8 @@
 #include <string>
 #include <vector>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 namespace io {
 template <typename i_t, typename f_t>
@@ -238,4 +240,5 @@ class cpu_optimization_problem_t : public optimization_problem_interface_t<i_t,
   std::vector<std::string> row_names_{};
 };
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/cpu_optimization_problem_solution.hpp b/cpp/include/cuopt/mathematical_optimization/cpu_optimization_problem_solution.hpp
index b4e5edec44..5e04e3bbd4 100644
--- a/cpp/include/cuopt/mathematical_optimization/cpu_optimization_problem_solution.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/cpu_optimization_problem_solution.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/cpu_pdlp_warm_start_data.hpp>
 #include <cuopt/mathematical_optimization/mip/solver_solution.hpp>
 #include <cuopt/mathematical_optimization/mip/solver_stats.hpp>
@@ -18,7 +19,8 @@
 
 #include <vector>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 /**
  * @brief CPU-backed LP solution (uses std::vector instead of rmm::device_uvector)
@@ -389,4 +391,5 @@ class cpu_mip_solution_t : public mip_solution_interface_t<i_t, f_t> {
   i_t num_simplex_iterations_;
 };
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/cpu_pdlp_warm_start_data.hpp b/cpp/include/cuopt/mathematical_optimization/cpu_pdlp_warm_start_data.hpp
index 9cf4740c96..1a76da0fa3 100644
--- a/cpp/include/cuopt/mathematical_optimization/cpu_pdlp_warm_start_data.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/cpu_pdlp_warm_start_data.hpp
@@ -7,10 +7,13 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/pdlp/pdlp_warm_start_data.hpp>
+
 #include <vector>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 // CPU version of pdlp_warm_start_data_t using std::vector for remote execution
 template <typename i_t, typename f_t>
@@ -118,4 +121,5 @@ template <typename i_t, typename f_t>
 pdlp_warm_start_data_t<i_t, f_t> convert_to_gpu_warmstart(
   const cpu_pdlp_warm_start_data_t<i_t, f_t>& cpu_data, rmm::cuda_stream_view stream);
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/cuopt_c.h b/cpp/include/cuopt/mathematical_optimization/cuopt_c.h
index 218402a7ef..7efa160924 100644
--- a/cpp/include/cuopt/mathematical_optimization/cuopt_c.h
+++ b/cpp/include/cuopt/mathematical_optimization/cuopt_c.h
@@ -9,6 +9,7 @@
 #define CUOPT_C_API_H
 
 #include <cuopt/mathematical_optimization/constants.h>
+#include <cuopt/export.hpp>
 
 #include <stdint.h>
 
@@ -17,6 +18,10 @@
 extern "C" {
 #endif
 
+#if defined(__GNUC__) || defined(__clang__)
+#pragma GCC visibility push(default)
+#endif
+
 /**
  * @brief A ``cuOptOptimizationProblem`` object contains a representation of
  * an LP, MIP, QP, or QCQP. It is created by ``cuOptCreateProblem``,
@@ -1055,6 +1060,10 @@ cuopt_int_t cuOptGetDualObjectiveValue(cuOptSolution solution,
  */
 cuopt_int_t cuOptGetReducedCosts(cuOptSolution solution, cuopt_float_t* reduced_cost_ptr);
 
+#if defined(__GNUC__) || defined(__clang__)
+#pragma GCC visibility pop
+#endif
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/cpp/include/cuopt/mathematical_optimization/io/data_model_view.hpp b/cpp/include/cuopt/mathematical_optimization/io/data_model_view.hpp
index 7cf4511eb9..41e1a98920 100644
--- a/cpp/include/cuopt/mathematical_optimization/io/data_model_view.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/io/data_model_view.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/io/mps_data_model.hpp>
 
 #include <cstdint>
@@ -15,7 +16,9 @@
 #include <type_traits>
 #include <vector>
 
-namespace cuopt::mathematical_optimization::io {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
+namespace io {
 
 /**
  * @brief A representation of a linear programming (LP) optimization problem
@@ -482,4 +485,6 @@ class data_model_view_t {
   std::vector<typename mps_data_model_t<i_t, f_t>::quadratic_constraint_t> quadratic_constraints_;
 };  // class data_model_view_t
 
-}  // namespace cuopt::mathematical_optimization::io
+}  // namespace io
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/io/mps_data_model.hpp b/cpp/include/cuopt/mathematical_optimization/io/mps_data_model.hpp
index 4ed1d7244f..7557327c29 100644
--- a/cpp/include/cuopt/mathematical_optimization/io/mps_data_model.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/io/mps_data_model.hpp
@@ -7,13 +7,17 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
+
 #include <cstdint>
 #include <span>
 #include <string>
 #include <type_traits>
 #include <vector>
 
-namespace cuopt::mathematical_optimization::io {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
+namespace io {
 
 /**
  * @brief A representation of a linear programming (LP) optimization problem
@@ -401,4 +405,6 @@ template <typename i_t, typename f_t>
 void canonicalize_quadratic_constraints(
   std::vector<typename mps_data_model_t<i_t, f_t>::quadratic_constraint_t>& constraints);
 
-}  // namespace cuopt::mathematical_optimization::io
+}  // namespace io
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/io/mps_writer.hpp b/cpp/include/cuopt/mathematical_optimization/io/mps_writer.hpp
index 6184ce56b6..9de061bbc9 100644
--- a/cpp/include/cuopt/mathematical_optimization/io/mps_writer.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/io/mps_writer.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/io/data_model_view.hpp>
 #include <cuopt/mathematical_optimization/io/mps_data_model.hpp>
 
@@ -18,7 +19,9 @@
 #include <unordered_set>
 #include <vector>
 
-namespace cuopt::mathematical_optimization::io {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
+namespace io {
 
 /**
  * @brief Main writer class for MPS files
@@ -60,4 +63,6 @@ class mps_writer_t {
   static data_model_view_t<i_t, f_t> create_view(const mps_data_model_t<i_t, f_t>& model);
 };  // class mps_writer_t
 
-}  // namespace cuopt::mathematical_optimization::io
+}  // namespace io
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/io/parser.hpp b/cpp/include/cuopt/mathematical_optimization/io/parser.hpp
index e3944c97e2..210cba0adc 100644
--- a/cpp/include/cuopt/mathematical_optimization/io/parser.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/io/parser.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/io/mps_data_model.hpp>
 
 #include <algorithm>
@@ -16,7 +17,9 @@
 #include <string>
 #include <string_view>
 
-namespace cuopt::mathematical_optimization::io {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
+namespace io {
 
 /**
  * @brief Selects which MPS reader implementation should be used by dispatching entry points.
@@ -201,4 +204,6 @@ inline mps_data_model_t<i_t, f_t> read(const std::string& path, bool fixed_mps_f
   return read<i_t, f_t>(path, mps_reader_type_t::default_reader, fixed_mps_format);
 }
 
-}  // namespace cuopt::mathematical_optimization::io
+}  // namespace io
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/io/utilities/cython_parser.hpp b/cpp/include/cuopt/mathematical_optimization/io/utilities/cython_parser.hpp
index a7863199df..e561325456 100644
--- a/cpp/include/cuopt/mathematical_optimization/io/utilities/cython_parser.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/io/utilities/cython_parser.hpp
@@ -7,12 +7,13 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/io/mps_data_model.hpp>
 
 #include <memory>
 
 namespace cuopt {
-namespace cython {
+namespace CUOPT_EXPORT cython {
 
 std::unique_ptr<cuopt::mathematical_optimization::io::mps_data_model_t<int, double>> call_read(
   const std::string& file_path, bool fixed_mps_format);
@@ -20,5 +21,5 @@ std::unique_ptr<cuopt::mathematical_optimization::io::mps_data_model_t<int, doub
 std::unique_ptr<cuopt::mathematical_optimization::io::mps_data_model_t<int, double>> call_parse_mps(
   const std::string& mps_file_path, bool fixed_mps_format);
 
-}  // namespace cython
+}  // namespace CUOPT_EXPORT cython
 }  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/io/writer.hpp b/cpp/include/cuopt/mathematical_optimization/io/writer.hpp
index 032865220e..f1a9514d5a 100644
--- a/cpp/include/cuopt/mathematical_optimization/io/writer.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/io/writer.hpp
@@ -7,9 +7,12 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/io/data_model_view.hpp>
 
-namespace cuopt::mathematical_optimization::io {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
+namespace io {
 
 /**
  * @brief Writes the problem to an MPS formatted file
@@ -23,4 +26,6 @@ namespace cuopt::mathematical_optimization::io {
 template <typename i_t, typename f_t>
 void write_mps(const data_model_view_t<i_t, f_t>& problem, const std::string& mps_file_path);
 
-}  // namespace cuopt::mathematical_optimization::io
+}  // namespace io
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/mip/solver_settings.hpp b/cpp/include/cuopt/mathematical_optimization/mip/solver_settings.hpp
index 8ddcdbbb8a..8bb676cb94 100644
--- a/cpp/include/cuopt/mathematical_optimization/mip/solver_settings.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/mip/solver_settings.hpp
@@ -11,6 +11,7 @@
 #include <vector>
 
 #include <cuopt/mathematical_optimization/constants.h>
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/mip/diving_hyper_params.hpp>
 #include <cuopt/mathematical_optimization/mip/heuristics_hyper_params.hpp>
 #include <cuopt/mathematical_optimization/mip/submip_hyper_params.hpp>
@@ -22,7 +23,8 @@
 
 #include <vector>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 struct benchmark_info_t {
   double last_improvement_of_best_feasible    = 0;
@@ -231,4 +233,5 @@ struct mip_solver_settings_accessor {
   }
 };
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/mip/solver_solution.hpp b/cpp/include/cuopt/mathematical_optimization/mip/solver_solution.hpp
index 19a8e5e531..1ad58b9e10 100644
--- a/cpp/include/cuopt/mathematical_optimization/mip/solver_solution.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/mip/solver_solution.hpp
@@ -9,6 +9,7 @@
 
 #include <cuopt/mathematical_optimization/constants.h>
 #include <cuopt/error.hpp>
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/mip/solver_stats.hpp>
 #include <cuopt/mathematical_optimization/utilities/internals.hpp>
 
@@ -21,7 +22,8 @@
 #include <string>
 #include <vector>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 enum class mip_termination_status_t : int8_t {
   NoTermination         = CUOPT_TERMINATION_STATUS_NO_TERMINATION,
@@ -92,4 +94,5 @@ class mip_solution_t : public base_solution_t {
   std::vector<rmm::device_uvector<f_t>> solution_pool_;
 };
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/optimization_problem.hpp b/cpp/include/cuopt/mathematical_optimization/optimization_problem.hpp
index 5c755281ca..bdfc2ffbd4 100644
--- a/cpp/include/cuopt/mathematical_optimization/optimization_problem.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/optimization_problem.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/optimization_problem_interface.hpp>
 #include <cuopt/mathematical_optimization/utilities/internals.hpp>
 
@@ -20,7 +21,8 @@
 #include <string>
 #include <vector>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 // Forward declarations
 template <typename i_t, typename f_t>
@@ -425,4 +427,5 @@ class optimization_problem_t : public optimization_problem_interface_t<i_t, f_t>
   std::vector<std::string> row_names_{};
 };
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/optimization_problem_solution.hpp b/cpp/include/cuopt/mathematical_optimization/optimization_problem_solution.hpp
index 822bd54de3..577d5727ec 100644
--- a/cpp/include/cuopt/mathematical_optimization/optimization_problem_solution.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/optimization_problem_solution.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/cpu_optimization_problem_solution.hpp>
 #include <cuopt/mathematical_optimization/mip/solver_solution.hpp>
 #include <cuopt/mathematical_optimization/optimization_problem_solution_interface.hpp>
@@ -16,7 +17,8 @@
 #include <raft/core/copy.hpp>
 #include <rmm/cuda_stream_view.hpp>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 /**
  * @brief GPU-backed LP solution (wraps optimization_problem_solution_t)
@@ -476,4 +478,5 @@ class gpu_mip_solution_t : public mip_solution_interface_t<i_t, f_t> {
   mip_solution_t<i_t, f_t> solution_;
 };
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/pdlp/pdlp_warm_start_data.hpp b/cpp/include/cuopt/mathematical_optimization/pdlp/pdlp_warm_start_data.hpp
index 07f54672f2..52a800c3c2 100644
--- a/cpp/include/cuopt/mathematical_optimization/pdlp/pdlp_warm_start_data.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/pdlp/pdlp_warm_start_data.hpp
@@ -7,11 +7,14 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
+
 #include <rmm/device_uvector.hpp>
 
 #include <span>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 template <typename i_t, typename f_t>
 struct pdlp_warm_start_data_view_t;
@@ -99,4 +102,5 @@ struct pdlp_warm_start_data_view_t {
   i_t iterations_since_last_restart_{-1};
 };
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/pdlp/solver_settings.hpp b/cpp/include/cuopt/mathematical_optimization/pdlp/solver_settings.hpp
index 96f548ec32..c6558369ab 100644
--- a/cpp/include/cuopt/mathematical_optimization/pdlp/solver_settings.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/pdlp/solver_settings.hpp
@@ -8,6 +8,7 @@
 #pragma once
 
 #include <cuopt/mathematical_optimization/constants.h>
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/cpu_pdlp_warm_start_data.hpp>
 #include <cuopt/mathematical_optimization/pdlp/pdlp_hyper_params.cuh>
 #include <cuopt/mathematical_optimization/pdlp/pdlp_warm_start_data.hpp>
@@ -21,7 +22,8 @@
 
 #include <cuda/std/span>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 // Forward declare solver_settings_t for friend class
 template <typename i_t, typename f_t>
@@ -351,4 +353,5 @@ class pdlp_solver_settings_t {
   friend class solver_settings_t<i_t, f_t>;
 };
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/pdlp/solver_solution.hpp b/cpp/include/cuopt/mathematical_optimization/pdlp/solver_solution.hpp
index 5bacff1101..be48ea4baf 100644
--- a/cpp/include/cuopt/mathematical_optimization/pdlp/solver_solution.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/pdlp/solver_solution.hpp
@@ -9,6 +9,7 @@
 
 #include <cuopt/mathematical_optimization/constants.h>
 #include <cuopt/error.hpp>
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/pdlp/pdlp_warm_start_data.hpp>
 #include <cuopt/mathematical_optimization/pdlp/solver_settings.hpp>
 #include <cuopt/mathematical_optimization/utilities/internals.hpp>
@@ -22,7 +23,8 @@
 #include <string>
 #include <vector>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 // Possible reasons for terminating
 enum class pdlp_termination_status_t : int8_t {
@@ -310,4 +312,5 @@ class optimization_problem_solution_t : public base_solution_t {
   /** error struct */
   cuopt::logic_error error_status_;
 };
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/solve.hpp b/cpp/include/cuopt/mathematical_optimization/solve.hpp
index 961fcd29b6..abc30aa0d7 100644
--- a/cpp/include/cuopt/mathematical_optimization/solve.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/solve.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/cpu_optimization_problem.hpp>
 #include <cuopt/mathematical_optimization/mip/solver_settings.hpp>
 #include <cuopt/mathematical_optimization/mip/solver_solution.hpp>
@@ -23,7 +24,8 @@
 #include <string>
 #include <vector>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 /**
  * @brief Linear programming solve function.
@@ -213,4 +215,5 @@ std::unique_ptr<mip_solution_interface_t<i_t, f_t>> solve_mip(
 
 // Remote execution functions are declared in solve_remote.hpp (included above)
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/solve_remote.hpp b/cpp/include/cuopt/mathematical_optimization/solve_remote.hpp
index 3636195660..bdb19f9c9d 100644
--- a/cpp/include/cuopt/mathematical_optimization/solve_remote.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/solve_remote.hpp
@@ -7,12 +7,14 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 // Include the solution interface definitions so unique_ptr can properly delete them
 #include <cuopt/mathematical_optimization/optimization_problem_solution_interface.hpp>
 
 #include <memory>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 // Forward declarations (only declaration needed, not definition)
 template <typename i_t, typename f_t>
@@ -44,4 +46,5 @@ std::unique_ptr<mip_solution_interface_t<i_t, f_t>> solve_mip_remote(
   cpu_optimization_problem_t<i_t, f_t> const& cpu_problem,
   mip_solver_settings_t<i_t, f_t> const& settings);
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/solver_settings.hpp b/cpp/include/cuopt/mathematical_optimization/solver_settings.hpp
index df373b209d..6b47805702 100644
--- a/cpp/include/cuopt/mathematical_optimization/solver_settings.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/solver_settings.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/pdlp/pdlp_warm_start_data.hpp>
 
 #include <raft/core/device_span.hpp>
@@ -22,7 +23,8 @@
 #include <optional>
 #include <vector>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+namespace CUOPT_EXPORT mathematical_optimization {
 
 template <typename i_t, typename f_t>
 class solver_settings_t {
@@ -109,4 +111,5 @@ class solver_settings_t {
   std::vector<parameter_info_t<std::string>> string_parameters;
 };
 
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/utilities/cython_solve.hpp b/cpp/include/cuopt/mathematical_optimization/utilities/cython_solve.hpp
index 79ef572d9c..f84119a8dc 100644
--- a/cpp/include/cuopt/mathematical_optimization/utilities/cython_solve.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/utilities/cython_solve.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/optimization_problem_interface.hpp>
 #include <cuopt/mathematical_optimization/optimization_problem_solution_interface.hpp>
 #include <cuopt/mathematical_optimization/solver_settings.hpp>
@@ -20,7 +21,7 @@
 #include <vector>
 
 namespace cuopt {
-namespace cython {
+namespace CUOPT_EXPORT cython {
 
 // Type definitions moved to cython_types.hpp to avoid circular dependencies
 // The types linear_programming_ret_t and mip_ret_t are defined in cython_types.hpp.
@@ -65,5 +66,5 @@ std::pair<std::vector<std::unique_ptr<solver_ret_t>>, double> call_batch_solve(
   std::vector<cuopt::mathematical_optimization::io::data_model_view_t<int, double>*>,
   mathematical_optimization::solver_settings_t<int, double>*);
 
-}  // namespace cython
+}  // namespace CUOPT_EXPORT cython
 }  // namespace cuopt
diff --git a/cpp/include/cuopt/mathematical_optimization/utilities/cython_types.hpp b/cpp/include/cuopt/mathematical_optimization/utilities/cython_types.hpp
index 966018bd9f..69d6f91604 100644
--- a/cpp/include/cuopt/mathematical_optimization/utilities/cython_types.hpp
+++ b/cpp/include/cuopt/mathematical_optimization/utilities/cython_types.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/mip/solver_solution.hpp>
 #include <cuopt/mathematical_optimization/pdlp/solver_solution.hpp>
 #include <cuopt/mathematical_optimization/utilities/internals.hpp>
@@ -19,7 +20,7 @@
 #include <vector>
 
 namespace cuopt {
-namespace cython {
+namespace CUOPT_EXPORT cython {
 
 using gpu_buffer = std::unique_ptr<rmm::device_buffer>;
 using cpu_buffer = std::vector<double>;
@@ -112,5 +113,5 @@ struct mip_ret_t {
   bool is_gpu() const { return std::holds_alternative<gpu_buffer>(solution_); }
 };
 
-}  // namespace cython
+}  // namespace CUOPT_EXPORT cython
 }  // namespace cuopt
diff --git a/cpp/include/cuopt/routing/assignment.hpp b/cpp/include/cuopt/routing/assignment.hpp
index f6fcc3d7b4..b138382d6d 100644
--- a/cpp/include/cuopt/routing/assignment.hpp
+++ b/cpp/include/cuopt/routing/assignment.hpp
@@ -1,6 +1,6 @@
 /* clang-format off */
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2021-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 /* clang-format on */
@@ -8,6 +8,7 @@
 #pragma once
 
 #include <cuopt/error.hpp>
+#include <cuopt/export.hpp>
 #include <cuopt/routing/routing_structures.hpp>
 #include <fstream>
 #include <rmm/cuda_stream_view.hpp>
@@ -16,7 +17,7 @@
 #include <vector>
 
 namespace cuopt {
-namespace routing {
+namespace CUOPT_EXPORT routing {
 
 /*! Routing assignment default strings */
 class solution_string_t {
@@ -264,5 +265,5 @@ struct host_assignment_t {
   std::vector<i_t> accepted{};
 };
 
-}  // namespace routing
+}  // namespace CUOPT_EXPORT routing
 }  // namespace cuopt
diff --git a/cpp/include/cuopt/routing/cython/cython.hpp b/cpp/include/cuopt/routing/cython/cython.hpp
index c1ac4e1dd6..76b0de7d2e 100644
--- a/cpp/include/cuopt/routing/cython/cython.hpp
+++ b/cpp/include/cuopt/routing/cython/cython.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/routing/assignment.hpp>
 #include <cuopt/routing/cython/generator.hpp>
 #include <cuopt/routing/data_model_view.hpp>
@@ -19,7 +20,7 @@
 #include <vector>
 
 namespace cuopt {
-namespace cython {
+namespace CUOPT_EXPORT cython {
 
 template <typename i_t, typename f_t>
 void populate_dataset_params(routing::generator::dataset_params_t<i_t, f_t>& params,
@@ -91,5 +92,5 @@ std::vector<std::unique_ptr<vehicle_routing_ret_t>> call_batch_solve(
 std::unique_ptr<dataset_ret_t> call_generate_dataset(
   raft::handle_t& handle, routing::generator::dataset_params_t<int, float> const& params);
 
-}  // namespace cython
+}  // namespace CUOPT_EXPORT cython
 }  // namespace cuopt
diff --git a/cpp/include/cuopt/routing/data_model_view.hpp b/cpp/include/cuopt/routing/data_model_view.hpp
index fa7aec73ec..b025469144 100644
--- a/cpp/include/cuopt/routing/data_model_view.hpp
+++ b/cpp/include/cuopt/routing/data_model_view.hpp
@@ -7,6 +7,7 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/routing/routing_structures.hpp>
 
 #include <raft/core/device_span.hpp>
@@ -14,7 +15,7 @@
 
 #include <unordered_set>
 namespace cuopt {
-namespace routing {
+namespace CUOPT_EXPORT routing {
 
 /**
  * @brief A container of vehicle routing solver input
@@ -663,5 +664,5 @@ class data_model_view_t {
   raft::device_span<i_t const> initial_sol_offsets_{};
   std::map<i_t, std::vector<detail::vehicle_break_t<i_t>>> vehicle_breaks_{};
 };
-}  // namespace routing
+}  // namespace CUOPT_EXPORT routing
 }  // namespace cuopt
diff --git a/cpp/include/cuopt/routing/routing_structures.hpp b/cpp/include/cuopt/routing/routing_structures.hpp
index 807ef6d880..64bcc22ce5 100644
--- a/cpp/include/cuopt/routing/routing_structures.hpp
+++ b/cpp/include/cuopt/routing/routing_structures.hpp
@@ -1,12 +1,14 @@
 /* clang-format off */
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2021-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 /* clang-format on */
 
 #pragma once
 
+#include <cuopt/export.hpp>
+
 #include <raft/core/device_span.hpp>
 
 #include <cstdint>
diff --git a/cpp/include/cuopt/routing/solve.hpp b/cpp/include/cuopt/routing/solve.hpp
index 82b36dd483..41a8cf932a 100644
--- a/cpp/include/cuopt/routing/solve.hpp
+++ b/cpp/include/cuopt/routing/solve.hpp
@@ -1,17 +1,18 @@
 /* clang-format off */
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 /* clang-format on */
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/routing/assignment.hpp>
 #include <cuopt/routing/data_model_view.hpp>
 #include <cuopt/routing/solver_settings.hpp>
 namespace cuopt {
-namespace routing {
+namespace CUOPT_EXPORT routing {
 
 /**
  * @brief Routing solve function
@@ -26,5 +27,5 @@ template <typename i_t, typename f_t>
 assignment_t<i_t> solve(
   data_model_view_t<i_t, f_t> const& data_model,
   solver_settings_t<i_t, f_t> const& settings = solver_settings_t<i_t, f_t>{});
-}  // namespace routing
+}  // namespace CUOPT_EXPORT routing
 }  // namespace cuopt
diff --git a/cpp/include/cuopt/routing/solver_settings.hpp b/cpp/include/cuopt/routing/solver_settings.hpp
index 36f301276d..3aae7ff0ef 100644
--- a/cpp/include/cuopt/routing/solver_settings.hpp
+++ b/cpp/include/cuopt/routing/solver_settings.hpp
@@ -1,19 +1,20 @@
 /* clang-format off */
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 /* clang-format on */
 
 #pragma once
 
+#include <cuopt/export.hpp>
 #include <cuopt/routing/routing_structures.hpp>
 #include <fstream>
 #include <limits>
 #include <ostream>
 
 namespace cuopt {
-namespace routing {
+namespace CUOPT_EXPORT routing {
 
 template <typename i_t, typename f_t>
 class solver_settings_t {
@@ -94,5 +95,5 @@ class solver_settings_t {
   std::string best_result_file_name_;
 };
 
-}  // namespace routing
+}  // namespace CUOPT_EXPORT routing
 }  // namespace cuopt
diff --git a/cpp/src/grpc/client/solve_remote.cpp b/cpp/src/grpc/client/solve_remote.cpp
index e8fbdf7139..eabea39e05 100644
--- a/cpp/src/grpc/client/solve_remote.cpp
+++ b/cpp/src/grpc/client/solve_remote.cpp
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/grpc/grpc_client_env.hpp>
 #include <cuopt/mathematical_optimization/cpu_optimization_problem.hpp>
 #include <cuopt/mathematical_optimization/cpu_optimization_problem_solution.hpp>
@@ -218,10 +219,10 @@ std::unique_ptr<mip_solution_interface_t<i_t, f_t>> solve_mip_remote(
 }
 
 // Explicit template instantiations for remote execution stubs
-template std::unique_ptr<lp_solution_interface_t<int, double>> solve_lp_remote(
+template CUOPT_EXPORT std::unique_ptr<lp_solution_interface_t<int, double>> solve_lp_remote(
   cpu_optimization_problem_t<int, double> const&, pdlp_solver_settings_t<int, double> const&);
 
-template std::unique_ptr<mip_solution_interface_t<int, double>> solve_mip_remote(
+template CUOPT_EXPORT std::unique_ptr<mip_solution_interface_t<int, double>> solve_mip_remote(
   cpu_optimization_problem_t<int, double> const&, mip_solver_settings_t<int, double> const&);
 
 }  // namespace cuopt::mathematical_optimization
diff --git a/cpp/src/grpc/grpc_problem_mapper.cpp b/cpp/src/grpc/grpc_problem_mapper.cpp
index c9698b7976..e330bca6f6 100644
--- a/cpp/src/grpc/grpc_problem_mapper.cpp
+++ b/cpp/src/grpc/grpc_problem_mapper.cpp
@@ -5,6 +5,8 @@
 
 #include "grpc_problem_mapper.hpp"
 
+#include <cuopt/export.hpp>
+
 #include <cuopt/mathematical_optimization/constants.h>
 #include <cuopt_remote.pb.h>
 #include <cuopt_remote_service.pb.h>
@@ -214,60 +216,64 @@ std::vector<cuopt::remote::SendArrayChunkRequest> build_array_chunk_requests(
 
 // Explicit template instantiations
 #if CUOPT_INSTANTIATE_FLOAT
-template void map_problem_to_proto(const cpu_optimization_problem_t<int32_t, float>& cpu_problem,
-                                   cuopt::remote::OptimizationProblem* pb_problem);
-template void map_proto_to_problem(const cuopt::remote::OptimizationProblem& pb_problem,
-                                   cpu_optimization_problem_t<int32_t, float>& cpu_problem);
-template size_t estimate_problem_proto_size(
-  const cpu_optimization_problem_t<int32_t, float>& cpu_problem);
-template void populate_chunked_header_lp(
+template CUOPT_EXPORT void map_problem_to_proto(
+  const cpu_optimization_problem_t<int32_t, float>& cpu_problem,
+  cuopt::remote::OptimizationProblem* pb_problem);
+template CUOPT_EXPORT void map_proto_to_problem(
+  const cuopt::remote::OptimizationProblem& pb_problem,
+  cpu_optimization_problem_t<int32_t, float>& cpu_problem);
+template CUOPT_EXPORT size_t
+estimate_problem_proto_size(const cpu_optimization_problem_t<int32_t, float>& cpu_problem);
+template CUOPT_EXPORT void populate_chunked_header_lp(
   const cpu_optimization_problem_t<int32_t, float>& cpu_problem,
   const pdlp_solver_settings_t<int32_t, float>& settings,
   cuopt::remote::ChunkedProblemHeader* header);
-template void populate_chunked_header_mip(
+template CUOPT_EXPORT void populate_chunked_header_mip(
   const cpu_optimization_problem_t<int32_t, float>& cpu_problem,
   const mip_solver_settings_t<int32_t, float>& settings,
   bool enable_incumbents,
   cuopt::remote::ChunkedProblemHeader* header);
-template void map_chunked_header_to_problem(
+template CUOPT_EXPORT void map_chunked_header_to_problem(
   const cuopt::remote::ChunkedProblemHeader& header,
   cpu_optimization_problem_t<int32_t, float>& cpu_problem);
-template void map_chunked_arrays_to_problem(
+template CUOPT_EXPORT void map_chunked_arrays_to_problem(
   const cuopt::remote::ChunkedProblemHeader& header,
   const std::map<int32_t, std::vector<uint8_t>>& arrays,
   const std::map<container_array_key_t, std::vector<uint8_t>>& container_arrays,
   cpu_optimization_problem_t<int32_t, float>& cpu_problem);
-template std::vector<cuopt::remote::SendArrayChunkRequest> build_array_chunk_requests(
+template CUOPT_EXPORT std::vector<cuopt::remote::SendArrayChunkRequest> build_array_chunk_requests(
   const cpu_optimization_problem_t<int32_t, float>& problem,
   const std::string& upload_id,
   int64_t chunk_size_bytes);
 #endif
 
 #if CUOPT_INSTANTIATE_DOUBLE
-template void map_problem_to_proto(const cpu_optimization_problem_t<int32_t, double>& cpu_problem,
-                                   cuopt::remote::OptimizationProblem* pb_problem);
-template void map_proto_to_problem(const cuopt::remote::OptimizationProblem& pb_problem,
-                                   cpu_optimization_problem_t<int32_t, double>& cpu_problem);
-template size_t estimate_problem_proto_size(
-  const cpu_optimization_problem_t<int32_t, double>& cpu_problem);
-template void populate_chunked_header_lp(
+template CUOPT_EXPORT void map_problem_to_proto(
+  const cpu_optimization_problem_t<int32_t, double>& cpu_problem,
+  cuopt::remote::OptimizationProblem* pb_problem);
+template CUOPT_EXPORT void map_proto_to_problem(
+  const cuopt::remote::OptimizationProblem& pb_problem,
+  cpu_optimization_problem_t<int32_t, double>& cpu_problem);
+template CUOPT_EXPORT size_t
+estimate_problem_proto_size(const cpu_optimization_problem_t<int32_t, double>& cpu_problem);
+template CUOPT_EXPORT void populate_chunked_header_lp(
   const cpu_optimization_problem_t<int32_t, double>& cpu_problem,
   const pdlp_solver_settings_t<int32_t, double>& settings,
   cuopt::remote::ChunkedProblemHeader* header);
-template void populate_chunked_header_mip(
+template CUOPT_EXPORT void populate_chunked_header_mip(
   const cpu_optimization_problem_t<int32_t, double>& cpu_problem,
   const mip_solver_settings_t<int32_t, double>& settings,
   bool enable_incumbents,
   cuopt::remote::ChunkedProblemHeader* header);
-template void map_chunked_header_to_problem(
+template CUOPT_EXPORT void map_chunked_header_to_problem(
   const cuopt::remote::ChunkedProblemHeader& header,
   cpu_optimization_problem_t<int32_t, double>& cpu_problem);
-template void map_chunked_arrays_to_problem(
+template CUOPT_EXPORT void map_chunked_arrays_to_problem(
   const cuopt::remote::ChunkedProblemHeader& header,
   const std::map<int32_t, std::vector<uint8_t>>& arrays,
   const std::map<container_array_key_t, std::vector<uint8_t>>& container_arrays,
   cpu_optimization_problem_t<int32_t, double>& cpu_problem);
-template std::vector<cuopt::remote::SendArrayChunkRequest> build_array_chunk_requests(
+template CUOPT_EXPORT std::vector<cuopt::remote::SendArrayChunkRequest> build_array_chunk_requests(
   const cpu_optimization_problem_t<int32_t, double>& problem,
   const std::string& upload_id,
   int64_t chunk_size_bytes);
diff --git a/cpp/src/grpc/grpc_settings_mapper.cpp b/cpp/src/grpc/grpc_settings_mapper.cpp
index 34926f89e2..5c076c5e3d 100644
--- a/cpp/src/grpc/grpc_settings_mapper.cpp
+++ b/cpp/src/grpc/grpc_settings_mapper.cpp
@@ -5,6 +5,8 @@
 
 #include "grpc_settings_mapper.hpp"
 
+#include <cuopt/export.hpp>
+
 #include <cuopt/mathematical_optimization/constants.h>
 #include <cuopt_remote.pb.h>
 #include <cuopt/mathematical_optimization/mip/solver_settings.hpp>
@@ -92,25 +94,33 @@ void map_proto_to_mip_settings(const cuopt::remote::MIPSolverSettings& pb_settin
 
 // Explicit template instantiations
 #if CUOPT_INSTANTIATE_FLOAT
-template void map_pdlp_settings_to_proto(const pdlp_solver_settings_t<int32_t, float>& settings,
-                                         cuopt::remote::PDLPSolverSettings* pb_settings);
-template void map_proto_to_pdlp_settings(const cuopt::remote::PDLPSolverSettings& pb_settings,
-                                         pdlp_solver_settings_t<int32_t, float>& settings);
-template void map_mip_settings_to_proto(const mip_solver_settings_t<int32_t, float>& settings,
-                                        cuopt::remote::MIPSolverSettings* pb_settings);
-template void map_proto_to_mip_settings(const cuopt::remote::MIPSolverSettings& pb_settings,
-                                        mip_solver_settings_t<int32_t, float>& settings);
+template CUOPT_EXPORT void map_pdlp_settings_to_proto(
+  const pdlp_solver_settings_t<int32_t, float>& settings,
+  cuopt::remote::PDLPSolverSettings* pb_settings);
+template CUOPT_EXPORT void map_proto_to_pdlp_settings(
+  const cuopt::remote::PDLPSolverSettings& pb_settings,
+  pdlp_solver_settings_t<int32_t, float>& settings);
+template CUOPT_EXPORT void map_mip_settings_to_proto(
+  const mip_solver_settings_t<int32_t, float>& settings,
+  cuopt::remote::MIPSolverSettings* pb_settings);
+template CUOPT_EXPORT void map_proto_to_mip_settings(
+  const cuopt::remote::MIPSolverSettings& pb_settings,
+  mip_solver_settings_t<int32_t, float>& settings);
 #endif
 
 #if CUOPT_INSTANTIATE_DOUBLE
-template void map_pdlp_settings_to_proto(const pdlp_solver_settings_t<int32_t, double>& settings,
-                                         cuopt::remote::PDLPSolverSettings* pb_settings);
-template void map_proto_to_pdlp_settings(const cuopt::remote::PDLPSolverSettings& pb_settings,
-                                         pdlp_solver_settings_t<int32_t, double>& settings);
-template void map_mip_settings_to_proto(const mip_solver_settings_t<int32_t, double>& settings,
-                                        cuopt::remote::MIPSolverSettings* pb_settings);
-template void map_proto_to_mip_settings(const cuopt::remote::MIPSolverSettings& pb_settings,
-                                        mip_solver_settings_t<int32_t, double>& settings);
+template CUOPT_EXPORT void map_pdlp_settings_to_proto(
+  const pdlp_solver_settings_t<int32_t, double>& settings,
+  cuopt::remote::PDLPSolverSettings* pb_settings);
+template CUOPT_EXPORT void map_proto_to_pdlp_settings(
+  const cuopt::remote::PDLPSolverSettings& pb_settings,
+  pdlp_solver_settings_t<int32_t, double>& settings);
+template CUOPT_EXPORT void map_mip_settings_to_proto(
+  const mip_solver_settings_t<int32_t, double>& settings,
+  cuopt::remote::MIPSolverSettings* pb_settings);
+template CUOPT_EXPORT void map_proto_to_mip_settings(
+  const cuopt::remote::MIPSolverSettings& pb_settings,
+  mip_solver_settings_t<int32_t, double>& settings);
 #endif
 
 }  // namespace cuopt::mathematical_optimization
diff --git a/cpp/src/grpc/grpc_solution_mapper.cpp b/cpp/src/grpc/grpc_solution_mapper.cpp
index 5a823986d6..7b71614628 100644
--- a/cpp/src/grpc/grpc_solution_mapper.cpp
+++ b/cpp/src/grpc/grpc_solution_mapper.cpp
@@ -5,6 +5,8 @@
 
 #include "grpc_solution_mapper.hpp"
 
+#include <cuopt/export.hpp>
+
 #include <cuopt/mathematical_optimization/constants.h>
 #include <cuopt_remote.pb.h>
 #include <cuopt_remote_service.pb.h>
@@ -193,66 +195,66 @@ void build_mip_solution_proto(const cuopt::remote::ChunkedResultHeader& header,
 
 // Explicit template instantiations
 #if CUOPT_INSTANTIATE_FLOAT
-template void map_lp_solution_to_proto(const cpu_lp_solution_t<int32_t, float>& solution,
-                                       cuopt::remote::LPSolution* pb_solution);
-template cpu_lp_solution_t<int32_t, float> map_proto_to_lp_solution(
+template CUOPT_EXPORT void map_lp_solution_to_proto(
+  const cpu_lp_solution_t<int32_t, float>& solution, cuopt::remote::LPSolution* pb_solution);
+template CUOPT_EXPORT cpu_lp_solution_t<int32_t, float> map_proto_to_lp_solution(
   const cuopt::remote::LPSolution& pb_solution);
-template void map_mip_solution_to_proto(const cpu_mip_solution_t<int32_t, float>& solution,
-                                        cuopt::remote::MIPSolution* pb_solution);
-template cpu_mip_solution_t<int32_t, float> map_proto_to_mip_solution(
+template CUOPT_EXPORT void map_mip_solution_to_proto(
+  const cpu_mip_solution_t<int32_t, float>& solution, cuopt::remote::MIPSolution* pb_solution);
+template CUOPT_EXPORT cpu_mip_solution_t<int32_t, float> map_proto_to_mip_solution(
   const cuopt::remote::MIPSolution& pb_solution);
-template void populate_chunked_result_header_lp(const cpu_lp_solution_t<int32_t, float>& solution,
-                                                cuopt::remote::ChunkedResultHeader* header);
-template void populate_chunked_result_header_mip(const cpu_mip_solution_t<int32_t, float>& solution,
-                                                 cuopt::remote::ChunkedResultHeader* header);
-template std::map<int32_t, std::vector<uint8_t>> collect_lp_solution_arrays(
+template CUOPT_EXPORT void populate_chunked_result_header_lp(
+  const cpu_lp_solution_t<int32_t, float>& solution, cuopt::remote::ChunkedResultHeader* header);
+template CUOPT_EXPORT void populate_chunked_result_header_mip(
+  const cpu_mip_solution_t<int32_t, float>& solution, cuopt::remote::ChunkedResultHeader* header);
+template CUOPT_EXPORT std::map<int32_t, std::vector<uint8_t>> collect_lp_solution_arrays(
   const cpu_lp_solution_t<int32_t, float>& solution);
-template std::map<int32_t, std::vector<uint8_t>> collect_mip_solution_arrays(
+template CUOPT_EXPORT std::map<int32_t, std::vector<uint8_t>> collect_mip_solution_arrays(
   const cpu_mip_solution_t<int32_t, float>& solution);
-template cpu_lp_solution_t<int32_t, float> chunked_result_to_lp_solution(
+template CUOPT_EXPORT cpu_lp_solution_t<int32_t, float> chunked_result_to_lp_solution(
   const cuopt::remote::ChunkedResultHeader& header,
   const std::map<int32_t, std::vector<uint8_t>>& arrays);
-template cpu_mip_solution_t<int32_t, float> chunked_result_to_mip_solution(
+template CUOPT_EXPORT cpu_mip_solution_t<int32_t, float> chunked_result_to_mip_solution(
   const cuopt::remote::ChunkedResultHeader& header,
   const std::map<int32_t, std::vector<uint8_t>>& arrays);
-template void build_lp_solution_proto<int32_t, float>(
+template CUOPT_EXPORT void build_lp_solution_proto<int32_t, float>(
   const cuopt::remote::ChunkedResultHeader& header,
   const std::map<int32_t, std::vector<uint8_t>>& arrays,
   cuopt::remote::LPSolution* proto);
-template void build_mip_solution_proto<int32_t, float>(
+template CUOPT_EXPORT void build_mip_solution_proto<int32_t, float>(
   const cuopt::remote::ChunkedResultHeader& header,
   const std::map<int32_t, std::vector<uint8_t>>& arrays,
   cuopt::remote::MIPSolution* proto);
 #endif
 
 #if CUOPT_INSTANTIATE_DOUBLE
-template void map_lp_solution_to_proto(const cpu_lp_solution_t<int32_t, double>& solution,
-                                       cuopt::remote::LPSolution* pb_solution);
-template cpu_lp_solution_t<int32_t, double> map_proto_to_lp_solution(
+template CUOPT_EXPORT void map_lp_solution_to_proto(
+  const cpu_lp_solution_t<int32_t, double>& solution, cuopt::remote::LPSolution* pb_solution);
+template CUOPT_EXPORT cpu_lp_solution_t<int32_t, double> map_proto_to_lp_solution(
   const cuopt::remote::LPSolution& pb_solution);
-template void map_mip_solution_to_proto(const cpu_mip_solution_t<int32_t, double>& solution,
-                                        cuopt::remote::MIPSolution* pb_solution);
-template cpu_mip_solution_t<int32_t, double> map_proto_to_mip_solution(
+template CUOPT_EXPORT void map_mip_solution_to_proto(
+  const cpu_mip_solution_t<int32_t, double>& solution, cuopt::remote::MIPSolution* pb_solution);
+template CUOPT_EXPORT cpu_mip_solution_t<int32_t, double> map_proto_to_mip_solution(
   const cuopt::remote::MIPSolution& pb_solution);
-template void populate_chunked_result_header_lp(const cpu_lp_solution_t<int32_t, double>& solution,
-                                                cuopt::remote::ChunkedResultHeader* header);
-template void populate_chunked_result_header_mip(
+template CUOPT_EXPORT void populate_chunked_result_header_lp(
+  const cpu_lp_solution_t<int32_t, double>& solution, cuopt::remote::ChunkedResultHeader* header);
+template CUOPT_EXPORT void populate_chunked_result_header_mip(
   const cpu_mip_solution_t<int32_t, double>& solution, cuopt::remote::ChunkedResultHeader* header);
-template std::map<int32_t, std::vector<uint8_t>> collect_lp_solution_arrays(
+template CUOPT_EXPORT std::map<int32_t, std::vector<uint8_t>> collect_lp_solution_arrays(
   const cpu_lp_solution_t<int32_t, double>& solution);
-template std::map<int32_t, std::vector<uint8_t>> collect_mip_solution_arrays(
+template CUOPT_EXPORT std::map<int32_t, std::vector<uint8_t>> collect_mip_solution_arrays(
   const cpu_mip_solution_t<int32_t, double>& solution);
-template cpu_lp_solution_t<int32_t, double> chunked_result_to_lp_solution(
+template CUOPT_EXPORT cpu_lp_solution_t<int32_t, double> chunked_result_to_lp_solution(
   const cuopt::remote::ChunkedResultHeader& header,
   const std::map<int32_t, std::vector<uint8_t>>& arrays);
-template cpu_mip_solution_t<int32_t, double> chunked_result_to_mip_solution(
+template CUOPT_EXPORT cpu_mip_solution_t<int32_t, double> chunked_result_to_mip_solution(
   const cuopt::remote::ChunkedResultHeader& header,
   const std::map<int32_t, std::vector<uint8_t>>& arrays);
-template void build_lp_solution_proto<int32_t, double>(
+template CUOPT_EXPORT void build_lp_solution_proto<int32_t, double>(
   const cuopt::remote::ChunkedResultHeader& header,
   const std::map<int32_t, std::vector<uint8_t>>& arrays,
   cuopt::remote::LPSolution* proto);
-template void build_mip_solution_proto<int32_t, double>(
+template CUOPT_EXPORT void build_mip_solution_proto<int32_t, double>(
   const cuopt::remote::ChunkedResultHeader& header,
   const std::map<int32_t, std::vector<uint8_t>>& arrays,
   cuopt::remote::MIPSolution* proto);
diff --git a/cpp/src/io/data_model_view.cpp b/cpp/src/io/data_model_view.cpp
index 132d87b866..df1efcd52f 100644
--- a/cpp/src/io/data_model_view.cpp
+++ b/cpp/src/io/data_model_view.cpp
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/io/data_model_view.hpp>
 #include <utilities/error.hpp>
 
@@ -377,8 +378,8 @@ data_model_view_t<i_t, f_t>::get_quadratic_constraints() const noexcept
 }
 
 // NOTE: Explicitly instantiate all types here in order to avoid linker error
-template class data_model_view_t<int, float>;
+template class CUOPT_EXPORT data_model_view_t<int, float>;
 
-template class data_model_view_t<int, double>;
+template class CUOPT_EXPORT data_model_view_t<int, double>;
 
 }  // namespace cuopt::mathematical_optimization::io
diff --git a/cpp/src/io/lp_parser.cpp b/cpp/src/io/lp_parser.cpp
index cf9d8d6538..5de0565801 100644
--- a/cpp/src/io/lp_parser.cpp
+++ b/cpp/src/io/lp_parser.cpp
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/io/parser.hpp>
 
 #include <file_to_string.hpp>
@@ -1526,9 +1527,11 @@ mps_data_model_t<i_t, f_t> read_lp_from_string(std::string_view lp_contents)
   return problem;
 }
 
-template mps_data_model_t<int, float> read_lp<int, float>(const std::string&);
-template mps_data_model_t<int, double> read_lp<int, double>(const std::string&);
-template mps_data_model_t<int, float> read_lp_from_string<int, float>(std::string_view);
-template mps_data_model_t<int, double> read_lp_from_string<int, double>(std::string_view);
+template CUOPT_EXPORT mps_data_model_t<int, float> read_lp<int, float>(const std::string&);
+template CUOPT_EXPORT mps_data_model_t<int, double> read_lp<int, double>(const std::string&);
+template CUOPT_EXPORT mps_data_model_t<int, float> read_lp_from_string<int, float>(
+  std::string_view);
+template CUOPT_EXPORT mps_data_model_t<int, double> read_lp_from_string<int, double>(
+  std::string_view);
 
 }  // namespace cuopt::mathematical_optimization::io
diff --git a/cpp/src/io/mps_data_model.cpp b/cpp/src/io/mps_data_model.cpp
index 941f16a96a..f3444fe12f 100644
--- a/cpp/src/io/mps_data_model.cpp
+++ b/cpp/src/io/mps_data_model.cpp
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/io/mps_data_model.hpp>
 #include <utilities/error.hpp>
 
@@ -469,13 +470,13 @@ void canonicalize_quadratic_constraints(
 }
 
 // NOTE: Explicitly instantiate all types here in order to avoid linker error
-template class mps_data_model_t<int, float>;
+template class CUOPT_EXPORT mps_data_model_t<int, float>;
 
-template class mps_data_model_t<int, double>;
+template class CUOPT_EXPORT mps_data_model_t<int, double>;
 
-template void canonicalize_quadratic_constraints<int, float>(
+template CUOPT_EXPORT void canonicalize_quadratic_constraints<int, float>(
   std::vector<mps_data_model_t<int, float>::quadratic_constraint_t>&);
-template void canonicalize_quadratic_constraints<int, double>(
+template CUOPT_EXPORT void canonicalize_quadratic_constraints<int, double>(
   std::vector<mps_data_model_t<int, double>::quadratic_constraint_t>&);
 //  TODO current raft to cusparse wrappers only support int64_t
 //  can be CUSPARSE_INDEX_16U, CUSPARSE_INDEX_32I, CUSPARSE_INDEX_64I
diff --git a/cpp/src/io/mps_writer.cpp b/cpp/src/io/mps_writer.cpp
index d269d6ec8a..31a0bbc25c 100644
--- a/cpp/src/io/mps_writer.cpp
+++ b/cpp/src/io/mps_writer.cpp
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/io/mps_writer.hpp>
 
 #include <cuopt/mathematical_optimization/io/data_model_view.hpp>
@@ -530,7 +531,7 @@ void mps_writer_t<i_t, f_t>::write(const std::string& mps_file_path)
   mps_file.close();
 }
 
-template class mps_writer_t<int, float>;
-template class mps_writer_t<int, double>;
+template class CUOPT_EXPORT mps_writer_t<int, float>;
+template class CUOPT_EXPORT mps_writer_t<int, double>;
 
 }  // namespace cuopt::mathematical_optimization::io
diff --git a/cpp/src/io/parser.cpp b/cpp/src/io/parser.cpp
index 0a626bd26c..eb574fd38a 100644
--- a/cpp/src/io/parser.cpp
+++ b/cpp/src/io/parser.cpp
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/io/parser.hpp>
 
 #include <experimental_mps_fast/fast_parser.hpp>
@@ -33,12 +34,14 @@ mps_data_model_t<i_t, f_t> read_mps_from_string(std::string_view mps_contents,
   return problem;
 }
 
-template mps_data_model_t<int, float> read_mps(const std::string& mps_file, bool fixed_mps_format);
-template mps_data_model_t<int, double> read_mps(const std::string& mps_file, bool fixed_mps_format);
-template mps_data_model_t<int, float> read_mps_from_string(std::string_view mps_contents,
-                                                           bool fixed_mps_format);
-template mps_data_model_t<int, double> read_mps_from_string(std::string_view mps_contents,
+template CUOPT_EXPORT mps_data_model_t<int, float> read_mps(const std::string& mps_file,
                                                             bool fixed_mps_format);
+template CUOPT_EXPORT mps_data_model_t<int, double> read_mps(const std::string& mps_file,
+                                                             bool fixed_mps_format);
+template CUOPT_EXPORT mps_data_model_t<int, float> read_mps_from_string(
+  std::string_view mps_contents, bool fixed_mps_format);
+template CUOPT_EXPORT mps_data_model_t<int, double> read_mps_from_string(
+  std::string_view mps_contents, bool fixed_mps_format);
 
 template <typename i_t, typename f_t>
 mps_data_model_t<i_t, f_t> read_mps_fast_experimental(const std::string& mps_file_path)
@@ -47,11 +50,13 @@ mps_data_model_t<i_t, f_t> read_mps_fast_experimental(const std::string& mps_fil
   return detail::parse_mps_fast_file<i_t, f_t>(mps_file_path);
 }
 
-template mps_data_model_t<int, float> read_mps_fast_experimental(const std::string& mps_file_path);
-template mps_data_model_t<int, double> read_mps_fast_experimental(const std::string& mps_file_path);
-template mps_data_model_t<int64_t, float> read_mps_fast_experimental(
+template CUOPT_EXPORT mps_data_model_t<int, float> read_mps_fast_experimental(
   const std::string& mps_file_path);
-template mps_data_model_t<int64_t, double> read_mps_fast_experimental(
+template CUOPT_EXPORT mps_data_model_t<int, double> read_mps_fast_experimental(
+  const std::string& mps_file_path);
+template CUOPT_EXPORT mps_data_model_t<int64_t, float> read_mps_fast_experimental(
+  const std::string& mps_file_path);
+template CUOPT_EXPORT mps_data_model_t<int64_t, double> read_mps_fast_experimental(
   const std::string& mps_file_path);
 
 }  // namespace cuopt::mathematical_optimization::io
diff --git a/cpp/src/io/writer.cpp b/cpp/src/io/writer.cpp
index c67a1aac4b..eeb11b7dc0 100644
--- a/cpp/src/io/writer.cpp
+++ b/cpp/src/io/writer.cpp
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/io/writer.hpp>
 
 #include <cuopt/mathematical_optimization/io/mps_writer.hpp>
@@ -18,9 +19,9 @@ void write_mps(const data_model_view_t<i_t, f_t>& problem, const std::string& mp
   writer.write(mps_file_path);
 }
 
-template void write_mps<int, float>(const data_model_view_t<int, float>& problem,
-                                    const std::string& mps_file_path);
-template void write_mps<int, double>(const data_model_view_t<int, double>& problem,
-                                     const std::string& mps_file_path);
+template CUOPT_EXPORT void write_mps<int, float>(const data_model_view_t<int, float>& problem,
+                                                 const std::string& mps_file_path);
+template CUOPT_EXPORT void write_mps<int, double>(const data_model_view_t<int, double>& problem,
+                                                  const std::string& mps_file_path);
 
 }  // namespace cuopt::mathematical_optimization::io
diff --git a/cpp/src/math_optimization/solution_reader.hpp b/cpp/src/math_optimization/solution_reader.hpp
index dd0976028a..18d065dc9b 100644
--- a/cpp/src/math_optimization/solution_reader.hpp
+++ b/cpp/src/math_optimization/solution_reader.hpp
@@ -7,10 +7,14 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
+
 #include <string>
 #include <vector>
 
-namespace cuopt::mathematical_optimization {
+namespace cuopt {
+// Ideally internal symbol should not be exported, but cuopt_cli uses it
+namespace CUOPT_EXPORT mathematical_optimization {
 
 /**
  * @brief Reads a solution file and returns the values of specified variables
@@ -24,4 +28,5 @@ class solution_reader_t {
   static std::vector<double> get_variable_values_from_sol_file(
     const std::string& sol_file_path, const std::vector<std::string>& variable_names);
 };
-}  // namespace cuopt::mathematical_optimization
+}  // namespace CUOPT_EXPORT mathematical_optimization
+}  // namespace cuopt
diff --git a/cpp/src/math_optimization/solver_settings.cu b/cpp/src/math_optimization/solver_settings.cu
index 9193112d71..be24b4c283 100644
--- a/cpp/src/math_optimization/solver_settings.cu
+++ b/cpp/src/math_optimization/solver_settings.cu
@@ -6,6 +6,7 @@
 /* clang-format on */
 
 #include <cuopt/error.hpp>
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/solver_settings.hpp>
 #include <mip_heuristics/mip_constants.hpp>
 #include <utilities/logger.hpp>
@@ -664,25 +665,39 @@ bool solver_settings_t<i_t, f_t>::dump_parameters_to_file(const std::string& pat
 }
 
 #if MIP_INSTANTIATE_FLOAT
-template class solver_settings_t<int, float>;
-template void solver_settings_t<int, float>::set_parameter(const std::string& name, int value);
-template void solver_settings_t<int, float>::set_parameter(const std::string& name, float value);
-template void solver_settings_t<int, float>::set_parameter(const std::string& name, bool value);
-template int solver_settings_t<int, float>::get_parameter(const std::string& name) const;
-template float solver_settings_t<int, float>::get_parameter(const std::string& name) const;
-template bool solver_settings_t<int, float>::get_parameter(const std::string& name) const;
-template std::string solver_settings_t<int, float>::get_parameter(const std::string& name) const;
+template class CUOPT_EXPORT solver_settings_t<int, float>;
+template CUOPT_EXPORT void solver_settings_t<int, float>::set_parameter(const std::string& name,
+                                                                        int value);
+template CUOPT_EXPORT void solver_settings_t<int, float>::set_parameter(const std::string& name,
+                                                                        float value);
+template CUOPT_EXPORT void solver_settings_t<int, float>::set_parameter(const std::string& name,
+                                                                        bool value);
+template CUOPT_EXPORT int solver_settings_t<int, float>::get_parameter(
+  const std::string& name) const;
+template CUOPT_EXPORT float solver_settings_t<int, float>::get_parameter(
+  const std::string& name) const;
+template CUOPT_EXPORT bool solver_settings_t<int, float>::get_parameter(
+  const std::string& name) const;
+template CUOPT_EXPORT std::string solver_settings_t<int, float>::get_parameter(
+  const std::string& name) const;
 #endif
 
 #if MIP_INSTANTIATE_DOUBLE
-template class solver_settings_t<int, double>;
-template void solver_settings_t<int, double>::set_parameter(const std::string& name, int value);
-template void solver_settings_t<int, double>::set_parameter(const std::string& name, double value);
-template void solver_settings_t<int, double>::set_parameter(const std::string& name, bool value);
-template int solver_settings_t<int, double>::get_parameter(const std::string& name) const;
-template double solver_settings_t<int, double>::get_parameter(const std::string& name) const;
-template bool solver_settings_t<int, double>::get_parameter(const std::string& name) const;
-template std::string solver_settings_t<int, double>::get_parameter(const std::string& name) const;
+template class CUOPT_EXPORT solver_settings_t<int, double>;
+template CUOPT_EXPORT void solver_settings_t<int, double>::set_parameter(const std::string& name,
+                                                                         int value);
+template CUOPT_EXPORT void solver_settings_t<int, double>::set_parameter(const std::string& name,
+                                                                         double value);
+template CUOPT_EXPORT void solver_settings_t<int, double>::set_parameter(const std::string& name,
+                                                                         bool value);
+template CUOPT_EXPORT int solver_settings_t<int, double>::get_parameter(
+  const std::string& name) const;
+template CUOPT_EXPORT double solver_settings_t<int, double>::get_parameter(
+  const std::string& name) const;
+template CUOPT_EXPORT bool solver_settings_t<int, double>::get_parameter(
+  const std::string& name) const;
+template CUOPT_EXPORT std::string solver_settings_t<int, double>::get_parameter(
+  const std::string& name) const;
 #endif
 
 }  // namespace cuopt::mathematical_optimization
diff --git a/cpp/src/mip_heuristics/solve.cu b/cpp/src/mip_heuristics/solve.cu
index 59b582188f..27242e7f0c 100644
--- a/cpp/src/mip_heuristics/solve.cu
+++ b/cpp/src/mip_heuristics/solve.cu
@@ -6,6 +6,7 @@
 /* clang-format on */
 
 #include <cuopt/error.hpp>
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/solve_remote.hpp>
 
 #include <linear_algebra/sort_csr.cuh>
@@ -945,19 +946,19 @@ std::unique_ptr<mip_solution_interface_t<i_t, f_t>> solve_mip(
 }
 
 #define INSTANTIATE(F_TYPE)                                                                    \
-  template mip_solution_t<int, F_TYPE> solve_mip(                                              \
+  template CUOPT_EXPORT mip_solution_t<int, F_TYPE> solve_mip(                                 \
     optimization_problem_t<int, F_TYPE>& op_problem,                                           \
     mip_solver_settings_t<int, F_TYPE> const& settings);                                       \
                                                                                                \
-  template mip_solution_t<int, F_TYPE> solve_mip(                                              \
+  template CUOPT_EXPORT mip_solution_t<int, F_TYPE> solve_mip(                                 \
     raft::handle_t const* handle_ptr,                                                          \
     const cuopt::mathematical_optimization::io::mps_data_model_t<int, F_TYPE>& mps_data_model, \
     mip_solver_settings_t<int, F_TYPE> const& settings);                                       \
                                                                                                \
-  template std::unique_ptr<mip_solution_interface_t<int, F_TYPE>> solve_mip(                   \
+  template CUOPT_EXPORT std::unique_ptr<mip_solution_interface_t<int, F_TYPE>> solve_mip(      \
     cpu_optimization_problem_t<int, F_TYPE>&, mip_solver_settings_t<int, F_TYPE> const&);      \
                                                                                                \
-  template std::unique_ptr<mip_solution_interface_t<int, F_TYPE>> solve_mip(                   \
+  template CUOPT_EXPORT std::unique_ptr<mip_solution_interface_t<int, F_TYPE>> solve_mip(      \
     optimization_problem_interface_t<int, F_TYPE>*, mip_solver_settings_t<int, F_TYPE> const&);
 
 #if MIP_INSTANTIATE_FLOAT
diff --git a/cpp/src/mip_heuristics/solver_settings.cu b/cpp/src/mip_heuristics/solver_settings.cu
index 645b8c8b6e..8b454c949b 100644
--- a/cpp/src/mip_heuristics/solver_settings.cu
+++ b/cpp/src/mip_heuristics/solver_settings.cu
@@ -6,6 +6,7 @@
 /* clang-format on */
 
 #include <cuopt/error.hpp>
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/mip/solver_settings.hpp>
 #include <mip_heuristics/mip_constants.hpp>
 #include <raft/util/cudart_utils.hpp>
@@ -48,11 +49,11 @@ mip_solver_settings_t<i_t, f_t>::get_tolerances() const noexcept
 
 // Explicit template instantiations for common types
 #if MIP_INSTANTIATE_FLOAT
-template class mip_solver_settings_t<int, float>;
+template class CUOPT_EXPORT mip_solver_settings_t<int, float>;
 #endif
 
 #if MIP_INSTANTIATE_DOUBLE
-template class mip_solver_settings_t<int, double>;
+template class CUOPT_EXPORT mip_solver_settings_t<int, double>;
 #endif
 
 }  // namespace cuopt::mathematical_optimization
diff --git a/cpp/src/mip_heuristics/solver_solution.cu b/cpp/src/mip_heuristics/solver_solution.cu
index 5b59e9a0fa..1997d684dc 100644
--- a/cpp/src/mip_heuristics/solver_solution.cu
+++ b/cpp/src/mip_heuristics/solver_solution.cu
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/mip/solver_solution.hpp>
 #include <mip_heuristics/mip_constants.hpp>
 #include <utilities/logger.hpp>
@@ -274,10 +275,10 @@ void mip_solution_t<i_t, f_t>::log_detailed_summary() const
 }
 
 #if MIP_INSTANTIATE_FLOAT || PDLP_INSTANTIATE_FLOAT
-template class mip_solution_t<int, float>;
+template class CUOPT_EXPORT mip_solution_t<int, float>;
 #endif
 
 #if MIP_INSTANTIATE_DOUBLE
-template class mip_solution_t<int, double>;
+template class CUOPT_EXPORT mip_solution_t<int, double>;
 #endif
 }  // namespace cuopt::mathematical_optimization
diff --git a/cpp/src/pdlp/cpu_optimization_problem.cpp b/cpp/src/pdlp/cpu_optimization_problem.cpp
index 725424bf6b..4b970eb6ec 100644
--- a/cpp/src/pdlp/cpu_optimization_problem.cpp
+++ b/cpp/src/pdlp/cpu_optimization_problem.cpp
@@ -6,6 +6,7 @@
 /* clang-format on */
 
 #include <cuopt/error.hpp>
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/cpu_optimization_problem.hpp>
 #include <cuopt/mathematical_optimization/csr_matrix_utils.hpp>
 #include <cuopt/mathematical_optimization/io/mps_data_model.hpp>
@@ -1187,10 +1188,10 @@ void cpu_optimization_problem_t<i_t, f_t>::adopt_from_mps_data_model(
 // ==============================================================================
 
 #if MIP_INSTANTIATE_FLOAT
-template class cpu_optimization_problem_t<int32_t, float>;
+template class CUOPT_EXPORT cpu_optimization_problem_t<int32_t, float>;
 #endif
 #if MIP_INSTANTIATE_DOUBLE
-template class cpu_optimization_problem_t<int32_t, double>;
+template class CUOPT_EXPORT cpu_optimization_problem_t<int32_t, double>;
 #endif
 
 }  // namespace cuopt::mathematical_optimization
diff --git a/cpp/src/pdlp/cpu_pdlp_warm_start_data.cu b/cpp/src/pdlp/cpu_pdlp_warm_start_data.cu
index 94c8ed5466..f2af28ba51 100644
--- a/cpp/src/pdlp/cpu_pdlp_warm_start_data.cu
+++ b/cpp/src/pdlp/cpu_pdlp_warm_start_data.cu
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/cpu_pdlp_warm_start_data.hpp>
 #include <cuopt/mathematical_optimization/pdlp/pdlp_warm_start_data.hpp>
 #include <mip_heuristics/mip_constants.hpp>
@@ -109,17 +110,17 @@ pdlp_warm_start_data_t<i_t, f_t> convert_to_gpu_warmstart(
 }
 
 #if MIP_INSTANTIATE_DOUBLE
-template cpu_pdlp_warm_start_data_t<int, double> convert_to_cpu_warmstart(
+template CUOPT_EXPORT cpu_pdlp_warm_start_data_t<int, double> convert_to_cpu_warmstart(
   const pdlp_warm_start_data_t<int, double>&, rmm::cuda_stream_view);
-template pdlp_warm_start_data_t<int, double> convert_to_gpu_warmstart(
+template CUOPT_EXPORT pdlp_warm_start_data_t<int, double> convert_to_gpu_warmstart(
   const cpu_pdlp_warm_start_data_t<int, double>&, rmm::cuda_stream_view);
 #endif
 
 #if MIP_INSTANTIATE_FLOAT || PDLP_INSTANTIATE_FLOAT
-template cpu_pdlp_warm_start_data_t<int, float> convert_to_cpu_warmstart(
+template CUOPT_EXPORT cpu_pdlp_warm_start_data_t<int, float> convert_to_cpu_warmstart(
   const pdlp_warm_start_data_t<int, float>&, rmm::cuda_stream_view);
 
-template pdlp_warm_start_data_t<int, float> convert_to_gpu_warmstart(
+template CUOPT_EXPORT pdlp_warm_start_data_t<int, float> convert_to_gpu_warmstart(
   const cpu_pdlp_warm_start_data_t<int, float>&, rmm::cuda_stream_view);
 #endif
 
diff --git a/cpp/src/pdlp/optimization_problem.cu b/cpp/src/pdlp/optimization_problem.cu
index 85d3ef4df4..95457e2556 100644
--- a/cpp/src/pdlp/optimization_problem.cu
+++ b/cpp/src/pdlp/optimization_problem.cu
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/cpu_optimization_problem.hpp>
 #include <cuopt/mathematical_optimization/optimization_problem.hpp>
 #include <cuopt/mathematical_optimization/optimization_problem_utils.hpp>
@@ -1632,14 +1633,14 @@ optimization_problem_t<i_t, other_f_t> optimization_problem_t<i_t, f_t>::convert
 // ==============================================================================
 // Explicit template instantiations matching MIP constants
 #if MIP_INSTANTIATE_FLOAT || PDLP_INSTANTIATE_FLOAT
-template class optimization_problem_t<int32_t, float>;
+template class CUOPT_EXPORT optimization_problem_t<int32_t, float>;
 #endif
 #if MIP_INSTANTIATE_DOUBLE
-template class optimization_problem_t<int32_t, double>;
+template class CUOPT_EXPORT optimization_problem_t<int32_t, double>;
 #endif
 
 #if PDLP_INSTANTIATE_FLOAT || MIP_INSTANTIATE_FLOAT
-template optimization_problem_t<int32_t, float>
+template CUOPT_EXPORT optimization_problem_t<int32_t, float>
   optimization_problem_t<int32_t, double>::convert_to_other_prec<float>(
     rmm::cuda_stream_view) const;
 #endif
diff --git a/cpp/src/pdlp/pdlp_warm_start_data.cu b/cpp/src/pdlp/pdlp_warm_start_data.cu
index 6039710423..a214f1a165 100644
--- a/cpp/src/pdlp/pdlp_warm_start_data.cu
+++ b/cpp/src/pdlp/pdlp_warm_start_data.cu
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/pdlp/pdlp_warm_start_data.hpp>
 #include <cuopt/mathematical_optimization/solver_settings.hpp>
 
@@ -179,10 +180,10 @@ void pdlp_warm_start_data_t<i_t, f_t>::check_sizes()
 }
 
 #if MIP_INSTANTIATE_FLOAT || PDLP_INSTANTIATE_FLOAT
-template class pdlp_warm_start_data_t<int, float>;
+template class CUOPT_EXPORT pdlp_warm_start_data_t<int, float>;
 #endif
 
 #if MIP_INSTANTIATE_DOUBLE
-template class pdlp_warm_start_data_t<int, double>;
+template class CUOPT_EXPORT pdlp_warm_start_data_t<int, double>;
 #endif
 }  // namespace cuopt::mathematical_optimization
diff --git a/cpp/src/pdlp/solution_conversion.cu b/cpp/src/pdlp/solution_conversion.cu
index 4e7336ea93..8293629e6f 100644
--- a/cpp/src/pdlp/solution_conversion.cu
+++ b/cpp/src/pdlp/solution_conversion.cu
@@ -10,6 +10,7 @@
  * @brief Implementations of conversion methods from solution classes to Cython ret structs
  */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/cpu_optimization_problem_solution.hpp>
 #include <cuopt/mathematical_optimization/optimization_problem_solution.hpp>
 #include <cuopt/mathematical_optimization/utilities/cython_solve.hpp>
@@ -215,11 +216,11 @@ cuopt::cython::mip_ret_t cpu_mip_solution_t<i_t, f_t>::to_cpu_mip_ret_t()
 }
 
 // Explicit template instantiations
-template cuopt::cython::linear_programming_ret_t
+template CUOPT_EXPORT cuopt::cython::linear_programming_ret_t
 gpu_lp_solution_t<int, double>::to_linear_programming_ret_t();
-template cuopt::cython::mip_ret_t gpu_mip_solution_t<int, double>::to_mip_ret_t();
-template cuopt::cython::linear_programming_ret_t
+template CUOPT_EXPORT cuopt::cython::mip_ret_t gpu_mip_solution_t<int, double>::to_mip_ret_t();
+template CUOPT_EXPORT cuopt::cython::linear_programming_ret_t
 cpu_lp_solution_t<int, double>::to_cpu_linear_programming_ret_t();
-template cuopt::cython::mip_ret_t cpu_mip_solution_t<int, double>::to_cpu_mip_ret_t();
+template CUOPT_EXPORT cuopt::cython::mip_ret_t cpu_mip_solution_t<int, double>::to_cpu_mip_ret_t();
 
 }  // namespace cuopt::mathematical_optimization
diff --git a/cpp/src/pdlp/solve.cu b/cpp/src/pdlp/solve.cu
index 25b427fa9a..c68707c7c0 100644
--- a/cpp/src/pdlp/solve.cu
+++ b/cpp/src/pdlp/solve.cu
@@ -6,6 +6,7 @@
 /* clang-format on */
 
 #include <cuopt/error.hpp>
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/solve_remote.hpp>
 #include <pdlp/cusparse_view.hpp>
 #include <pdlp/optimal_batch_size_handler/optimal_batch_size_handler.hpp>
@@ -2306,28 +2307,28 @@ std::unique_ptr<lp_solution_interface_t<i_t, f_t>> solve_lp(
 }
 
 #define INSTANTIATE(F_TYPE)                                                                      \
-  template optimization_problem_solution_t<int, F_TYPE> solve_lp(                                \
+  template CUOPT_EXPORT optimization_problem_solution_t<int, F_TYPE> solve_lp(                   \
     optimization_problem_t<int, F_TYPE>& op_problem,                                             \
     pdlp_solver_settings_t<int, F_TYPE> const& settings,                                         \
     bool problem_checking,                                                                       \
     bool use_pdlp_solver_mode,                                                                   \
     bool is_batch_mode);                                                                         \
                                                                                                  \
-  template optimization_problem_solution_t<int, F_TYPE> solve_lp(                                \
+  template CUOPT_EXPORT optimization_problem_solution_t<int, F_TYPE> solve_lp(                   \
     raft::handle_t const* handle_ptr,                                                            \
     const cuopt::mathematical_optimization::io::mps_data_model_t<int, F_TYPE>& mps_data_model,   \
     pdlp_solver_settings_t<int, F_TYPE> const& settings,                                         \
     bool problem_checking,                                                                       \
     bool use_pdlp_solver_mode);                                                                  \
                                                                                                  \
-  template std::unique_ptr<lp_solution_interface_t<int, F_TYPE>> solve_lp(                       \
+  template CUOPT_EXPORT std::unique_ptr<lp_solution_interface_t<int, F_TYPE>> solve_lp(          \
     cpu_optimization_problem_t<int, F_TYPE>&,                                                    \
     pdlp_solver_settings_t<int, F_TYPE> const&,                                                  \
     bool,                                                                                        \
     bool,                                                                                        \
     bool);                                                                                       \
                                                                                                  \
-  template std::unique_ptr<lp_solution_interface_t<int, F_TYPE>> solve_lp(                       \
+  template CUOPT_EXPORT std::unique_ptr<lp_solution_interface_t<int, F_TYPE>> solve_lp(          \
     optimization_problem_interface_t<int, F_TYPE>*,                                              \
     pdlp_solver_settings_t<int, F_TYPE> const&,                                                  \
     bool,                                                                                        \
@@ -2340,7 +2341,7 @@ std::unique_ptr<lp_solution_interface_t<i_t, f_t>> solve_lp(
     const timer_t& timer,                                                                        \
     bool is_batch_mode);                                                                         \
                                                                                                  \
-  template optimization_problem_solution_t<int, F_TYPE> batch_pdlp_solve(                        \
+  template CUOPT_EXPORT optimization_problem_solution_t<int, F_TYPE> batch_pdlp_solve(           \
     raft::handle_t const* handle_ptr,                                                            \
     const cuopt::mathematical_optimization::io::mps_data_model_t<int, F_TYPE>& mps_data_model,   \
     const std::vector<int>& fractional,                                                          \
@@ -2356,7 +2357,8 @@ std::unique_ptr<lp_solution_interface_t<i_t, f_t>> solve_lp(
                                              bool per_climber_constraint_bounds,                 \
                                              bool collect_solutions);                            \
                                                                                                  \
-  template optimization_problem_t<int, F_TYPE> mps_data_model_to_optimization_problem(           \
+  template CUOPT_EXPORT optimization_problem_t<int, F_TYPE>                                      \
+  mps_data_model_to_optimization_problem(                                                        \
     raft::handle_t const* handle_ptr,                                                            \
     const cuopt::mathematical_optimization::io::mps_data_model_t<int, F_TYPE>& data_model);      \
   template void set_pdlp_solver_mode(pdlp_solver_settings_t<int, F_TYPE>& settings);
diff --git a/cpp/src/pdlp/solver_settings.cu b/cpp/src/pdlp/solver_settings.cu
index 3b8e34ce5b..33d8f1a64b 100644
--- a/cpp/src/pdlp/solver_settings.cu
+++ b/cpp/src/pdlp/solver_settings.cu
@@ -6,6 +6,7 @@
 /* clang-format on */
 
 #include <cuopt/error.hpp>
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/pdlp/pdlp_warm_start_data.hpp>
 #include <cuopt/mathematical_optimization/pdlp/solver_settings.hpp>
 #include <math_optimization/solution_writer.hpp>
@@ -415,11 +416,11 @@ pdlp_solver_settings_t<i_t, f_t>::get_pdlp_warm_start_data_view() const noexcept
 }
 
 #if MIP_INSTANTIATE_FLOAT || PDLP_INSTANTIATE_FLOAT
-template class pdlp_solver_settings_t<int, float>;
+template class CUOPT_EXPORT pdlp_solver_settings_t<int, float>;
 #endif
 
 #if MIP_INSTANTIATE_DOUBLE
-template class pdlp_solver_settings_t<int, double>;
+template class CUOPT_EXPORT pdlp_solver_settings_t<int, double>;
 #endif
 
 }  // namespace cuopt::mathematical_optimization
diff --git a/cpp/src/pdlp/solver_solution.cu b/cpp/src/pdlp/solver_solution.cu
index 0da8649dc4..08e5ee00a8 100644
--- a/cpp/src/pdlp/solver_solution.cu
+++ b/cpp/src/pdlp/solver_solution.cu
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/mathematical_optimization/pdlp/solver_solution.hpp>
 
 #include <math_optimization/solution_writer.hpp>
@@ -453,10 +454,10 @@ void optimization_problem_solution_t<i_t, f_t>::write_to_sol_file(
 }
 
 #if MIP_INSTANTIATE_FLOAT || PDLP_INSTANTIATE_FLOAT
-template class optimization_problem_solution_t<int, float>;
+template class CUOPT_EXPORT optimization_problem_solution_t<int, float>;
 #endif
 
 #if MIP_INSTANTIATE_DOUBLE
-template class optimization_problem_solution_t<int, double>;
+template class CUOPT_EXPORT optimization_problem_solution_t<int, double>;
 #endif
 }  // namespace cuopt::mathematical_optimization
diff --git a/cpp/src/routing/assignment.cu b/cpp/src/routing/assignment.cu
index 4636fa7359..be40bda183 100644
--- a/cpp/src/routing/assignment.cu
+++ b/cpp/src/routing/assignment.cu
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/routing/assignment.hpp>
 #include <raft/util/cudart_utils.hpp>
 #include <utilities/copy_helpers.hpp>
@@ -272,8 +273,8 @@ void host_assignment_t<i_t>::print() const noexcept
   }
 }
 
-template class assignment_t<int>;
-template class host_assignment_t<int>;
+template class CUOPT_EXPORT assignment_t<int>;
+template class CUOPT_EXPORT host_assignment_t<int>;
 
 }  // namespace routing
 }  // namespace cuopt
diff --git a/cpp/src/routing/data_model_view.cu b/cpp/src/routing/data_model_view.cu
index f5865cdad6..392c7a10fb 100644
--- a/cpp/src/routing/data_model_view.cu
+++ b/cpp/src/routing/data_model_view.cu
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/routing/data_model_view.hpp>
 #include <limits>
 #include <ostream>
@@ -743,6 +744,6 @@ raft::handle_t const* data_model_view_t<i_t, f_t>::get_handle_ptr() const noexce
   return handle_ptr_;
 }
 
-template class data_model_view_t<int, float>;
+template class CUOPT_EXPORT data_model_view_t<int, float>;
 }  // namespace routing
 }  // namespace cuopt
diff --git a/cpp/src/routing/distance_engine/waypoint_matrix.cpp b/cpp/src/routing/distance_engine/waypoint_matrix.cpp
index f31d1130d3..030c8790ea 100644
--- a/cpp/src/routing/distance_engine/waypoint_matrix.cpp
+++ b/cpp/src/routing/distance_engine/waypoint_matrix.cpp
@@ -1,11 +1,12 @@
 /* clang-format off */
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 /* clang-format on */
 
 #include <cuopt/error.hpp>
+#include <cuopt/export.hpp>
 #include <cuopt/routing/distance_engine/waypoint_matrix.hpp>
 
 #include <routing/utilities/check_input.hpp>
@@ -408,7 +409,7 @@ void waypoint_matrix_t<i_t, f_t>::compute_shortest_path_costs(f_t* d_custom_matr
   stream_view_.synchronize();
 }
 
-template class waypoint_matrix_t<int, float>;
+template class CUOPT_EXPORT waypoint_matrix_t<int, float>;
 
 }  // namespace distance_engine
 }  // namespace cuopt
diff --git a/cpp/src/routing/solve.cu b/cpp/src/routing/solve.cu
index abac614633..a7caf88ad9 100644
--- a/cpp/src/routing/solve.cu
+++ b/cpp/src/routing/solve.cu
@@ -1,10 +1,11 @@
 /* clang-format off */
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/routing/solve.hpp>
 #include <routing/solver.hpp>
 #include <utilities/logger.hpp>
@@ -29,7 +30,7 @@ assignment_t<i_t> solve(data_model_view_t<i_t, f_t> const& data_model,
   }
 }
 
-template assignment_t<int> solve(data_model_view_t<int, float> const& data_model,
-                                 solver_settings_t<int, float> const& settings);
+template CUOPT_EXPORT assignment_t<int> solve(data_model_view_t<int, float> const& data_model,
+                                              solver_settings_t<int, float> const& settings);
 }  // namespace routing
 }  // namespace cuopt
diff --git a/cpp/src/routing/solver_settings.cu b/cpp/src/routing/solver_settings.cu
index a9025cc08f..6267f39698 100644
--- a/cpp/src/routing/solver_settings.cu
+++ b/cpp/src/routing/solver_settings.cu
@@ -1,11 +1,12 @@
 /* clang-format off */
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 /* clang-format on */
 
 #include <cuopt/error.hpp>
+#include <cuopt/export.hpp>
 #include <cuopt/routing/solver_settings.hpp>
 
 namespace cuopt {
@@ -62,6 +63,6 @@ std::tuple<i_t, bool, std::string> solver_settings_t<i_t, f_t>::get_dump_best_re
   return std::make_tuple(dump_interval_, dump_best_results_, best_result_file_name_);
 }
 
-template class solver_settings_t<int, float>;
+template class CUOPT_EXPORT solver_settings_t<int, float>;
 }  // namespace routing
 }  // namespace cuopt
diff --git a/cpp/src/routing/utilities/cython.cu b/cpp/src/routing/utilities/cython.cu
index cf9e8ab73b..5a0b9bf6b2 100644
--- a/cpp/src/routing/utilities/cython.cu
+++ b/cpp/src/routing/utilities/cython.cu
@@ -5,6 +5,7 @@
  */
 /* clang-format on */
 
+#include <cuopt/export.hpp>
 #include <cuopt/routing/cython/cython.hpp>
 #include <cuopt/routing/solve.hpp>
 #include <raft/core/handle.hpp>
@@ -187,7 +188,7 @@ std::unique_ptr<dataset_ret_t> call_generate_dataset(
   return std::make_unique<dataset_ret_t>(std::move(gen_ret));
 }
 
-template void populate_dataset_params<int, float>(
+template CUOPT_EXPORT void populate_dataset_params<int, float>(
   routing::generator::dataset_params_t<int, float>& params,
   int n_locations,
   bool asymmetric,
diff --git a/cpp/src/utilities/logger.hpp b/cpp/src/utilities/logger.hpp
index f3a5bf6f2f..2f9053b05f 100644
--- a/cpp/src/utilities/logger.hpp
+++ b/cpp/src/utilities/logger.hpp
@@ -7,6 +7,8 @@
 
 #pragma once
 
+#include <cuopt/export.hpp>
+
 #include <cuopt/logger_macros.hpp>
 
 #include <rapids_logger/logger.hpp>
@@ -20,7 +22,7 @@
 #include <type_traits>
 #include <vector>
 
-namespace cuopt {
+namespace CUOPT_EXPORT cuopt {
 
 /**
  * @brief Get the default logger.
@@ -45,7 +47,9 @@ class init_logger_t {
   init_logger_t(std::string log_file, bool log_to_console);
 };
 
-namespace detail {
+}  // namespace CUOPT_EXPORT cuopt
+
+namespace cuopt::detail {
 
 // Returns true for the first N calls sharing this counter.
 template <auto N>
@@ -69,9 +73,7 @@ inline bool log_every_n_should_emit(std::atomic<uint64_t>& counter)
   return counter.fetch_add(1, std::memory_order_relaxed) % (uint64_t)N == 0;
 }
 
-}  // namespace detail
-
-}  // namespace cuopt
+}  // namespace cuopt::detail
 
 // Rate-limited logging built on the generated CUOPT_LOG_<level> macros. `level` is one of
 // TRACE/DEBUG/INFO/WARN/ERROR/CRITICAL; `n` must be a positive compile-time constant. Each