lightgbm: restore the workflow and use static OpenMP schedules

.github/workflows/build-lightgbm.yml

@@ -29,115 +29,7 @@ env:
   MANYLINUX_RISCV64_IMAGE: quay.io/pypa/manylinux_2_39_riscv64
 
 jobs:
-  omp_probe:
-    name: Probe libgomp and glibc malloc on riscv64
-    runs-on: ubuntu-24.04-riscv
-    timeout-minutes: 180
-    steps:
-      - name: Write the probe
-        run: |
-          cat > probe.c <<'PROBE_EOF'
-          /* Minimal riscv64 probe: OpenMP work-share vs glibc malloc under threads. */
-          #define _GNU_SOURCE
-          #include <omp.h>
-          #include <pthread.h>
-          #include <stdio.h>
-          #include <stdlib.h>
-          #include <string.h>
-
-          static long alloc_body(int i) {
-            char *p = malloc(64 + (i & 255));
-            p[0] = (char)i;
-            long v = p[0];
-            free(p);
-            return v;
-          }
-
-          static long calc_body(int i) { return (long)i * 3 % 7; }
-
-          static int nthreads, nrounds, n;
-
-          static void *pth(void *arg) {
-            long s = 0;
-            for (int r = 0; r < nrounds; r++)
-              for (int i = 0; i < n; i++) s += alloc_body(i);
-            return (void *)s;
-          }
-
-          int main(int argc, char **argv) {
-            const char *kind = argc > 1 ? argv[1] : "guided-alloc";
-            nrounds = argc > 2 ? atoi(argv[2]) : 100000;
-            n = argc > 3 ? atoi(argv[3]) : 300;
-            nthreads = omp_get_max_threads();
-            long total = 0;
-            printf("kind=%s threads=%d rounds=%d n=%d\n", kind, nthreads, nrounds, n);
-            fflush(stdout);
-
-            if (!strcmp(kind, "pthread-alloc")) {
-              pthread_t t[64];
-              if (nthreads > 64) nthreads = 64;
-              for (int i = 0; i < nthreads; i++) pthread_create(&t[i], NULL, pth, NULL);
-              for (int i = 0; i < nthreads; i++) { void *r; pthread_join(t[i], &r); total += (long)r; }
-              printf("PROBE_OK %s %ld\n", kind, total);
-              return 0;
-            }
-
-            for (int r = 0; r < nrounds; r++) {
-              long s = 0;
-              if (!strcmp(kind, "guided-alloc")) {
-          #pragma omp parallel for num_threads(nthreads) schedule(guided) reduction(+ : s)
-                for (int i = 0; i < n; i++) s += alloc_body(i);
-              } else if (!strcmp(kind, "guided-calc")) {
-          #pragma omp parallel for num_threads(nthreads) schedule(guided) reduction(+ : s)
-                for (int i = 0; i < n; i++) s += calc_body(i);
-              } else if (!strcmp(kind, "static-alloc")) {
-          #pragma omp parallel for num_threads(nthreads) schedule(static) reduction(+ : s)
-                for (int i = 0; i < n; i++) s += alloc_body(i);
-              } else if (!strcmp(kind, "dynamic-alloc")) {
-          #pragma omp parallel for num_threads(nthreads) schedule(dynamic) reduction(+ : s)
-                for (int i = 0; i < n; i++) s += alloc_body(i);
-              } else {
-                fprintf(stderr, "unknown kind\n");
-                return 2;
-              }
-              total += s;
-            }
-            printf("PROBE_OK %s %ld\n", kind, total);
-            return 0;
-          }
-          PROBE_EOF
-      - name: Bare runner
-        run: |
-          set +e
-          gcc --version | head -1
-          nproc; head -2 /proc/meminfo
-          grep -E 'processor|isa|uarch|mmu|model|hart' /proc/cpuinfo | head -12
-          gcc -O2 -fopenmp -pthread -o probe probe.c
-          for kind in guided-calc static-alloc dynamic-alloc guided-alloc pthread-alloc; do
-            for i in 1 2 3; do
-              timeout 1200 ./probe "$kind" 100000 300 > out.log 2>&1
-              echo "### bare $kind attempt $i rc=$? $(tail -1 out.log)"
-            done
-          done
-          echo "### BARE DONE"
-      - name: Manylinux image
-        run: |
-          set +e
-          podman run -t --rm --network=host -v "${PWD}:/work" -w /work \
-            "${MANYLINUX_RISCV64_IMAGE}" bash -c '
-              gcc --version | head -1
-              gcc -O2 -fopenmp -pthread -o /tmp/probe /work/probe.c
-              for kind in guided-calc static-alloc dynamic-alloc guided-alloc pthread-alloc; do
-                for i in 1 2 3; do
-                  timeout 1200 /tmp/probe "$kind" 100000 300 > /tmp/out.log 2>&1
-                  echo "### container $kind attempt $i rc=$? $(tail -1 /tmp/out.log)"
-                done
-              done
-              echo "### CONTAINER DONE"
-            '
-          echo "### STEP DONE"
   python_sdist:
-    if: github.run_id == '0'  # probe-only run
     name: Build lightgbm ${{ inputs.version || '4.7.0' }} sdist
     runs-on: ubuntu-latest
     outputs: