[ML] Fix test parallelism for low-core CI machines and add diagnostics

The ceil(nproc/3) formula was too conservative on 4-core machines
(macOS CI Orka VMs), yielding -j 2 which serialised test suites into
5 waves. On <=4 cores, CTest internal parallelism is modest enough
that using all cores avoids unnecessary serialisation.

Also adds diagnostic logging of CPU count and parallelism settings
to build.gradle, docker_entrypoint.sh, and run-tests-individually.cmake
to make CI performance easier to analyse.

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: edsavage

build.gradle

@@ -116,7 +116,9 @@ project.ext.numCpus = Runtime.runtime.availableProcessors()
 // Each suite spawns ctest --parallel <all_cpus> internally, so running too
 // many suites simultaneously causes resource contention. These values were
 // determined empirically (see PR #2900).
-project.ext.testParallel = isWindows ? 2 : Math.max(1, (int) Math.ceil(numCpus / 3.0))
+// On low-core machines (<=4, e.g. macOS CI Orka VMs), use all cores since
+// CTest internal parallelism is modest enough that contention is minimal.
+project.ext.testParallel = isWindows ? 2 : (numCpus <= 4 ? numCpus : Math.max(2, (int) Math.ceil(numCpus / 3.0)))
 project.ext.makeEnvironment = [ 'CPP_CROSS_COMPILE': cppCrossCompile,
                                 'VERSION_QUALIFIER': versionQualifier,
                                 'SNAPSHOT': (isSnapshot ? 'yes' : 'no'),
@@ -232,6 +234,9 @@ task test(type: Exec) {
   workingDir "${projectDir}"
   dependsOn 'strip'
   description = 'Run C++ tests'
+  doFirst {
+    logger.lifecycle("Test parallelism: numCpus=${numCpus}, testParallel=${testParallel} (cmake --build -j ${testParallel})")
+  }
 }
 
 check {

cmake/run-tests-individually.cmake

@@ -188,8 +188,8 @@ if(NOT "${_current_batch}" STREQUAL "")
 endif()
 
 list(LENGTH _batches _num_batches)
-message(STATUS "Running ${_num_tests} test(s) in ${_num_batches} batch(es), "
-  "max ${MAX_PROCS} parallel process(es)")
+message(STATUS "${TEST_SUITE}: ${_num_tests} test(s) in ${_num_batches} batch(es), "
+  "MAX_ARGS=${MAX_ARGS}, MAX_PROCS=${MAX_PROCS} (${_num_cpus} logical CPUs)")
 
 # ---------------------------------------------------------------------------
 # Generate a per-batch runner script invoked by each CTest test entry

dev-tools/docker/docker_entrypoint.sh

@@ -68,7 +68,15 @@ if [ "x$1" = "x--test" ] ; then
     echo passed > build/test_status.txt
     # Each test suite spawns ctest --parallel <nproc> internally, so limit
     # the number of suites running concurrently to avoid resource contention.
-    TEST_PARALLEL=$(( ($(nproc) + 2) / 3 ))
+    # On low-core machines (<=4), use all cores since CTest internal
+    # parallelism is modest enough that contention is minimal.
+    NCPUS=$(nproc)
+    if [ "$NCPUS" -le 4 ]; then
+        TEST_PARALLEL=$NCPUS
+    else
+        TEST_PARALLEL=$(( (NCPUS + 2) / 3 ))
+    fi
+    echo "Test parallelism: nproc=${NCPUS}, TEST_PARALLEL=${TEST_PARALLEL} (cmake --build -j ${TEST_PARALLEL})"
     cmake --build cmake-build-docker ${CMAKE_VERBOSE} -j ${TEST_PARALLEL} -t test_individually || echo failed > build/test_status.txt
 fi