Merge branch 'main' into conda-limited-api-abi3

Author: ramakrishnap-nv

.github/CODEOWNERS

@@ -2,8 +2,8 @@
 *                  @nvidia/cuopt-infra-codeowners
 
 #cpp code owners
-cpp/               @nvidia/cuopt-engine-codeowners
-benchmarks/        @nvidia/cuopt-engine-codeowners
+cpp/               @nvidia/cuopt-solver-codeowners
+benchmarks/        @nvidia/cuopt-solver-codeowners
 
 #python code owners
 python/            @nvidia/cuopt-infra-codeowners

.github/workflows/pr.yaml

@@ -115,6 +115,7 @@ jobs:
           - '!ci/release/**'
           - '!ci/run_*.pytests.sh'
           - '!ci/run_ctests.sh'
+          - '!ci/run-vale.sh'
           - '!ci/test_cpp*.sh'
           - '!ci/test_notebooks.sh'
           - '!ci/test_python.sh'
@@ -175,6 +176,7 @@ jobs:
           - '!ci/check_style.sh'
           - '!ci/docker/**'
           - '!ci/release/**'
+          - '!ci/run-vale.sh'
           - '!ci/test_python.sh'
           - '!ci/test_self_hosted_service.sh'
           - '!ci/test_wheel*.sh'
@@ -245,6 +247,7 @@ jobs:
           - '!ci/check_style.sh'
           - '!ci/docker/**'
           - '!ci/release/**'
+          - '!ci/run-vale.sh'
           - '!ci/test_self_hosted_service.sh'
           - '!ci/test_wheel*.sh'
           - '!ci/thirdparty-testing/**'
@@ -315,6 +318,7 @@ jobs:
           - '!ci/docker/**'
           - '!ci/release/**'
           - '!ci/run_ctests.sh'
+          - '!ci/run-vale.sh'
           - '!ci/test_python.sh'
           - '!ci/thirdparty-testing/**'
           - '!ci/utils/sync_skills_version.sh'

.pre-commit-config.yaml

@@ -137,8 +137,14 @@ repos:
         files: ^(VERSION|\.claude-plugin/marketplace\.json|\.cursor-plugin/plugin\.json|gemini-extension\.json)$
       - id: vale
         name: Vale docs prose lint
-        entry: vale
+        entry: ci/run-vale.sh
+        verbose: true
         language: system
+        # 'pass_filenames: false' prevents pre-commit from invoking 'vale' once per file,
+        # (instead passes all files at once to 1 invocation)
+        pass_filenames: false
+        args: [--output=line, --minAlertLevel=warning, docs/cuopt/source]
+        # needs to be kept in sync with vale.ini
         files: ^docs/cuopt/source/.*\.(rst|md)$
       - id: validate-skills
         name: Validate agent skills

build.sh

@@ -14,7 +14,7 @@ ARGS=$*
 REPODIR=$(cd "$(dirname "$0")"; pwd)
 LIBCUOPT_BUILD_DIR=${LIBCUOPT_BUILD_DIR:=${REPODIR}/cpp/build}
 
-VALIDARGS="clean codegen libcuopt cuopt_grpc_server cuopt cuopt_server cuopt_sh_client docs deb -a -b -g -fsanitize -tsan -msan -v -l= --verbose-pdlp --build-lp-only  --no-fetch-rapids --skip-c-python-adapters --skip-tests-build --skip-routing-build --skip-grpc-build --skip-fatbin-write --host-lineinfo [--cmake-args=\\\"<args>\\\"] [--cache-tool=<tool>] --install --allgpuarch --ci-only-arch --show_depr_warn -h --help"
+VALIDARGS="clean codegen libcuopt cuopt_grpc_server cuopt cuopt_server cuopt_sh_client docs deb -a -b -g -fsanitize -tsan -msan -v -l= --verbose-pdlp --build-lp-only  --no-fetch-rapids --skip-c-python-adapters --skip-tests-build --skip-routing-build --skip-grpc-build --skip-fatbin-write --host-lineinfo --split-compile [--cmake-args=\\\"<args>\\\"] [--cache-tool=<tool>] --install --allgpuarch --ci-only-arch --show_depr_warn -h --help"
 HELP="$0 [<target> ...] [<flag> ...]
  where <target> is:
    clean            - remove all existing build artifacts and configuration (start over)
@@ -45,6 +45,7 @@ HELP="$0 [<target> ...] [<flag> ...]
    --skip-grpc-build    - skip building gRPC and protobuf components (auto-enabled with -tsan)
    --skip-fatbin-write      - skip the fatbin write
    --host-lineinfo           - build with debug line information for host code
+   --split-compile           - opt in to nvcc split compilation; builds may be nondeterministic
    --cache-tool=<tool> - pass the build cache tool (eg: ccache, sccache, distcc) that will be used
                       to speedup the build process.
    --cmake-args=\\\"<args>\\\"   - pass arbitrary list of CMake configuration options (escape all quotes in argument)
@@ -260,6 +261,11 @@ fi
 if hasArg --host-lineinfo; then
     HOST_LINEINFO=1
 fi
+if hasArg --split-compile; then
+    # nvcc split compilation can produce nondeterministic cuOpt builds, so keep it opt-in.
+    echo "WARNING: nvcc split compilation may produce nondeterministic cuOpt builds."
+    export NVCC_PREPEND_FLAGS="${NVCC_PREPEND_FLAGS:+${NVCC_PREPEND_FLAGS} }--split-compile=0"
+fi
 
 function contains_string {
     local search_string="$1"

ci/build_wheel_cuopt.sh

@@ -41,6 +41,7 @@ EXCLUDE_ARGS=(
   --exclude "libcusolver.so.*"
   --exclude "libcusparse.so.*"
   --exclude "libcuopt.so"
+  --exclude "libnvJitLink.so*"
   --exclude "librapids_logger.so"
   --exclude "librmm.so"
 )

ci/build_wheel_libcuopt.sh

@@ -103,7 +103,7 @@ EXCLUDE_ARGS=(
   --exclude "libcusolver.so.*"
   --exclude "libcusparse.so.*"
   --exclude "libnccl.so.*"
-  --exclude "libnvJitLink*"
+  --exclude "libnvJitLink.so*"
   --exclude "librapids_logger.so"
   --exclude "librmm.so"
   # OpenSSL 3 is intentionally NOT bundled. Resolving libssl.so.3 / libcrypto.so.3

ci/run-vale.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+
+# SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+set -euo pipefail
+
+if ! command -v vale &>/dev/null; then
+    if [[ "${CI:-}" != "true" ]]; then
+        echo "WARNING: vale not found — skipping prose lint. Install 'vale' to run this check locally." >&2
+        exit 0
+    fi
+    echo "ERROR: vale not found and CI=true." >&2
+    exit 1
+fi
+
+exec vale "$@"

cpp/CMakeLists.txt

@@ -190,7 +190,7 @@ endif ()
 # Add jobserver flags for parallel compilation if PARALLEL_LEVEL is set
 if (PARALLEL_LEVEL AND NOT "${PARALLEL_LEVEL}" STREQUAL "")
     message(STATUS "Enabling nvcc parallel compilation support")
-    list(APPEND CUOPT_CUDA_FLAGS --threads=0 --split-compile=0)
+    list(APPEND CUOPT_CUDA_FLAGS --threads=0)
     if (USE_NVCC_JOBSERVER AND CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 13.0)
         message(STATUS "Enabling nvcc jobserver support (NVCC >= 13.0)")
         list(APPEND CUOPT_CUDA_FLAGS --jobserver)
@@ -295,7 +295,7 @@ FetchContent_MakeAvailable(papilo)
 FetchContent_Declare(
         pslp
         GIT_REPOSITORY "https://github.com/dance858/PSLP.git"
-        GIT_TAG "v0.0.10"
+        GIT_TAG "v0.0.11"
         GIT_PROGRESS TRUE
         EXCLUDE_FROM_ALL
         SYSTEM
@@ -495,12 +495,6 @@ set(CUOPT_SRC_FILES)
 set(MPS_FAST_SRC_FILES)
 add_subdirectory(src)
 
-# nvcc 13.0.3 ICE (signal 11) compiling sliding_window.cu with 7 GPU architectures;
-# --split-compile breaks the codegen into per-arch sub-jobs to avoid the crash
-set_source_files_properties(
-        ${CMAKE_CURRENT_SOURCE_DIR}/src/routing/local_search/sliding_window.cu
-        PROPERTIES COMPILE_OPTIONS "--split-compile=0")
-
 if (HOST_LINEINFO)
     set_source_files_properties(${CUOPT_SRC_FILES} DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR} PROPERTIES COMPILE_OPTIONS "-g1")
 endif ()

cpp/include/cuopt/mathematical_optimization/constants.h

@@ -20,68 +20,69 @@
 #define CUOPT_INSTANTIATE_INT64  0
 
 /* @brief LP/MIP parameter string constants */
-#define CUOPT_ABSOLUTE_DUAL_TOLERANCE              "absolute_dual_tolerance"
-#define CUOPT_RELATIVE_DUAL_TOLERANCE              "relative_dual_tolerance"
-#define CUOPT_ABSOLUTE_PRIMAL_TOLERANCE            "absolute_primal_tolerance"
-#define CUOPT_RELATIVE_PRIMAL_TOLERANCE            "relative_primal_tolerance"
-#define CUOPT_ABSOLUTE_GAP_TOLERANCE               "absolute_gap_tolerance"
-#define CUOPT_RELATIVE_GAP_TOLERANCE               "relative_gap_tolerance"
-#define CUOPT_INFEASIBILITY_DETECTION              "infeasibility_detection"
-#define CUOPT_STRICT_INFEASIBILITY                 "strict_infeasibility"
-#define CUOPT_PRIMAL_INFEASIBLE_TOLERANCE          "primal_infeasible_tolerance"
-#define CUOPT_DUAL_INFEASIBLE_TOLERANCE            "dual_infeasible_tolerance"
-#define CUOPT_ITERATION_LIMIT                      "iteration_limit"
-#define CUOPT_TIME_LIMIT                           "time_limit"
-#define CUOPT_WORK_LIMIT                           "work_limit"
-#define CUOPT_NODE_LIMIT                           "node_limit"
-#define CUOPT_PDLP_SOLVER_MODE                     "pdlp_solver_mode"
-#define CUOPT_METHOD                               "method"
-#define CUOPT_PER_CONSTRAINT_RESIDUAL              "per_constraint_residual"
-#define CUOPT_SAVE_BEST_PRIMAL_SO_FAR              "save_best_primal_so_far"
-#define CUOPT_FIRST_PRIMAL_FEASIBLE                "first_primal_feasible"
-#define CUOPT_LOG_FILE                             "log_file"
-#define CUOPT_LOG_TO_CONSOLE                       "log_to_console"
-#define CUOPT_CROSSOVER                            "crossover"
-#define CUOPT_FOLDING                              "folding"
-#define CUOPT_AUGMENTED                            "augmented"
-#define CUOPT_DUALIZE                              "dualize"
-#define CUOPT_ORDERING                             "ordering"
-#define CUOPT_BARRIER_DUAL_INITIAL_POINT           "barrier_dual_initial_point"
-#define CUOPT_POSTSOLVE_INFO                       "postsolve_info"
-#define CUOPT_BARRIER_ITERATIVE_REFINEMENT         "barrier_iterative_refinement"
-#define CUOPT_BARRIER_STEP_SCALE                   "barrier_step_scale"
-#define CUOPT_ELIMINATE_DENSE_COLUMNS              "eliminate_dense_columns"
-#define CUOPT_CUDSS_DETERMINISTIC                  "cudss_deterministic"
-#define CUOPT_PRESOLVE                             "presolve"
-#define CUOPT_MIP_PROBING                          "mip_probing"
-#define CUOPT_DUAL_POSTSOLVE                       "dual_postsolve"
-#define CUOPT_MIP_DETERMINISM_MODE                 "mip_determinism_mode"
-#define CUOPT_MIP_ABSOLUTE_TOLERANCE               "mip_absolute_tolerance"
-#define CUOPT_MIP_RELATIVE_TOLERANCE               "mip_relative_tolerance"
-#define CUOPT_MIP_INTEGRALITY_TOLERANCE            "mip_integrality_tolerance"
-#define CUOPT_MIP_ABSOLUTE_GAP                     "mip_absolute_gap"
-#define CUOPT_MIP_RELATIVE_GAP                     "mip_relative_gap"
-#define CUOPT_MIP_HEURISTICS_ONLY                  "mip_heuristics_only"
-#define CUOPT_MIP_SCALING                          "mip_scaling"
-#define CUOPT_MIP_PRESOLVE                         "mip_presolve"
-#define CUOPT_MIP_SYMMETRY                         "mip_symmetry"
-#define CUOPT_MIP_RELIABILITY_BRANCHING            "mip_reliability_branching"
-#define CUOPT_MIP_CUT_PASSES                       "mip_cut_passes"
-#define CUOPT_MIP_MIXED_INTEGER_ROUNDING_CUTS      "mip_mixed_integer_rounding_cuts"
-#define CUOPT_MIP_MIXED_INTEGER_GOMORY_CUTS        "mip_mixed_integer_gomory_cuts"
-#define CUOPT_MIP_KNAPSACK_CUTS                    "mip_knapsack_cuts"
-#define CUOPT_MIP_FLOW_COVER_CUTS                  "mip_flow_cover_cuts"
-#define CUOPT_MIP_IMPLIED_BOUND_CUTS               "mip_implied_bound_cuts"
-#define CUOPT_MIP_CLIQUE_CUTS                      "mip_clique_cuts"
-#define CUOPT_MIP_ZERO_HALF_CUTS                   "mip_zero_half_cuts"
-#define CUOPT_MIP_STRONG_CHVATAL_GOMORY_CUTS       "mip_strong_chvatal_gomory_cuts"
-#define CUOPT_MIP_REDUCED_COST_STRENGTHENING       "mip_reduced_cost_strengthening"
-#define CUOPT_MIP_RINS                             "mip_rins"
-#define CUOPT_MIP_OBJECTIVE_STEP                   "mip_objective_step"
-#define CUOPT_MIP_CUT_CHANGE_THRESHOLD             "mip_cut_change_threshold"
-#define CUOPT_MIP_CUT_MIN_ORTHOGONALITY            "mip_cut_min_orthogonality"
-#define CUOPT_MIP_BATCH_PDLP_STRONG_BRANCHING      "mip_batch_pdlp_strong_branching"
-#define CUOPT_MIP_BATCH_PDLP_RELIABILITY_BRANCHING "mip_batch_pdlp_reliability_branching"
+#define CUOPT_ABSOLUTE_DUAL_TOLERANCE               "absolute_dual_tolerance"
+#define CUOPT_RELATIVE_DUAL_TOLERANCE               "relative_dual_tolerance"
+#define CUOPT_ABSOLUTE_PRIMAL_TOLERANCE             "absolute_primal_tolerance"
+#define CUOPT_RELATIVE_PRIMAL_TOLERANCE             "relative_primal_tolerance"
+#define CUOPT_ABSOLUTE_GAP_TOLERANCE                "absolute_gap_tolerance"
+#define CUOPT_RELATIVE_GAP_TOLERANCE                "relative_gap_tolerance"
+#define CUOPT_INFEASIBILITY_DETECTION               "infeasibility_detection"
+#define CUOPT_STRICT_INFEASIBILITY                  "strict_infeasibility"
+#define CUOPT_PRIMAL_INFEASIBLE_TOLERANCE           "primal_infeasible_tolerance"
+#define CUOPT_DUAL_INFEASIBLE_TOLERANCE             "dual_infeasible_tolerance"
+#define CUOPT_ITERATION_LIMIT                       "iteration_limit"
+#define CUOPT_TIME_LIMIT                            "time_limit"
+#define CUOPT_WORK_LIMIT                            "work_limit"
+#define CUOPT_NODE_LIMIT                            "node_limit"
+#define CUOPT_PDLP_SOLVER_MODE                      "pdlp_solver_mode"
+#define CUOPT_METHOD                                "method"
+#define CUOPT_PER_CONSTRAINT_RESIDUAL               "per_constraint_residual"
+#define CUOPT_SAVE_BEST_PRIMAL_SO_FAR               "save_best_primal_so_far"
+#define CUOPT_FIRST_PRIMAL_FEASIBLE                 "first_primal_feasible"
+#define CUOPT_LOG_FILE                              "log_file"
+#define CUOPT_LOG_TO_CONSOLE                        "log_to_console"
+#define CUOPT_CROSSOVER                             "crossover"
+#define CUOPT_FOLDING                               "folding"
+#define CUOPT_AUGMENTED                             "augmented"
+#define CUOPT_DUALIZE                               "dualize"
+#define CUOPT_ORDERING                              "ordering"
+#define CUOPT_BARRIER_DUAL_INITIAL_POINT            "barrier_dual_initial_point"
+#define CUOPT_POSTSOLVE_INFO                        "postsolve_info"
+#define CUOPT_BARRIER_PRESOLVE_BOUND_FREE_VARIABLES "barrier_presolve_bound_free_variables"
+#define CUOPT_BARRIER_ITERATIVE_REFINEMENT          "barrier_iterative_refinement"
+#define CUOPT_BARRIER_STEP_SCALE                    "barrier_step_scale"
+#define CUOPT_ELIMINATE_DENSE_COLUMNS               "eliminate_dense_columns"
+#define CUOPT_CUDSS_DETERMINISTIC                   "cudss_deterministic"
+#define CUOPT_PRESOLVE                              "presolve"
+#define CUOPT_MIP_PROBING                           "mip_probing"
+#define CUOPT_DUAL_POSTSOLVE                        "dual_postsolve"
+#define CUOPT_MIP_DETERMINISM_MODE                  "mip_determinism_mode"
+#define CUOPT_MIP_ABSOLUTE_TOLERANCE                "mip_absolute_tolerance"
+#define CUOPT_MIP_RELATIVE_TOLERANCE                "mip_relative_tolerance"
+#define CUOPT_MIP_INTEGRALITY_TOLERANCE             "mip_integrality_tolerance"
+#define CUOPT_MIP_ABSOLUTE_GAP                      "mip_absolute_gap"
+#define CUOPT_MIP_RELATIVE_GAP                      "mip_relative_gap"
+#define CUOPT_MIP_HEURISTICS_ONLY                   "mip_heuristics_only"
+#define CUOPT_MIP_SCALING                           "mip_scaling"
+#define CUOPT_MIP_PRESOLVE                          "mip_presolve"
+#define CUOPT_MIP_SYMMETRY                          "mip_symmetry"
+#define CUOPT_MIP_RELIABILITY_BRANCHING             "mip_reliability_branching"
+#define CUOPT_MIP_CUT_PASSES                        "mip_cut_passes"
+#define CUOPT_MIP_MIXED_INTEGER_ROUNDING_CUTS       "mip_mixed_integer_rounding_cuts"
+#define CUOPT_MIP_MIXED_INTEGER_GOMORY_CUTS         "mip_mixed_integer_gomory_cuts"
+#define CUOPT_MIP_KNAPSACK_CUTS                     "mip_knapsack_cuts"
+#define CUOPT_MIP_FLOW_COVER_CUTS                   "mip_flow_cover_cuts"
+#define CUOPT_MIP_IMPLIED_BOUND_CUTS                "mip_implied_bound_cuts"
+#define CUOPT_MIP_CLIQUE_CUTS                       "mip_clique_cuts"
+#define CUOPT_MIP_ZERO_HALF_CUTS                    "mip_zero_half_cuts"
+#define CUOPT_MIP_STRONG_CHVATAL_GOMORY_CUTS        "mip_strong_chvatal_gomory_cuts"
+#define CUOPT_MIP_REDUCED_COST_STRENGTHENING        "mip_reduced_cost_strengthening"
+#define CUOPT_MIP_RINS                              "mip_rins"
+#define CUOPT_MIP_OBJECTIVE_STEP                    "mip_objective_step"
+#define CUOPT_MIP_CUT_CHANGE_THRESHOLD              "mip_cut_change_threshold"
+#define CUOPT_MIP_CUT_MIN_ORTHOGONALITY             "mip_cut_min_orthogonality"
+#define CUOPT_MIP_BATCH_PDLP_STRONG_BRANCHING       "mip_batch_pdlp_strong_branching"
+#define CUOPT_MIP_BATCH_PDLP_RELIABILITY_BRANCHING  "mip_batch_pdlp_reliability_branching"
 #define CUOPT_MIP_STRONG_BRANCHING_SIMPLEX_ITERATION_LIMIT \
   "mip_strong_branching_simplex_iteration_limit"
 

cpp/include/cuopt/mathematical_optimization/pdlp/solver_settings.hpp

@@ -296,6 +296,7 @@ class pdlp_solver_settings_t {
   i_t ordering{-1};
   i_t barrier_dual_initial_point{-1};
   i_t postsolve_info{-1};
+  i_t barrier_presolve_bound_free_variables{-1};  // -1 automatic, 0 disabled, 1 enabled
   // Ruiz equilibration for QCQP (barrier) scaling: -1 automatic (row/column
   // imbalance heuristic), 0 disabled, 1 enabled. Distinct from PDLP's own Ruiz
   // scaling in pdlp_hyper_params_t.