diff --git a/pi_digits_benchmark/README.md b/pi_digits_benchmark/README.md
index 55f8ea1..802438a 100644
--- a/pi_digits_benchmark/README.md
+++ b/pi_digits_benchmark/README.md
@@ -29,5 +29,5 @@ Per-run outputs are stored under `runs/`:
 
 - `pi_digits.txt` — final concatenated digits
 - `workers.csv` — per-worker durations and chunk metadata
-- `results.csv` — per-run summary metrics (includes aggregated RSS/VSZ and snapshot overhead)
+- `results.csv` — per-run summary metrics (includes aggregated RSS/VSZ/PSS and snapshot overhead)
 - `stdout.log` / `stderr.log` — benchmark logs
diff --git a/pi_digits_benchmark/pi_digits_benchmark b/pi_digits_benchmark/pi_digits_benchmark
index d41562f..54178b9 100755
Binary files a/pi_digits_benchmark/pi_digits_benchmark and b/pi_digits_benchmark/pi_digits_benchmark differ
diff --git a/pi_digits_benchmark/pi_digits_benchmark.c b/pi_digits_benchmark/pi_digits_benchmark.c
index 5d10aae..51e437b 100644
--- a/pi_digits_benchmark/pi_digits_benchmark.c
+++ b/pi_digits_benchmark/pi_digits_benchmark.c
@@ -54,6 +54,7 @@ typedef struct {
     long max_vsz_kb;
     long agg_rss_kb;
     long agg_vsz_kb;
+    long agg_pss_kb;
     double mem_snapshot_s;
     size_t output_bytes;
     int exit_code;
@@ -348,17 +349,36 @@ static int read_self_status_kb(long *rss_kb, long *vsz_kb) {
     return 0;
 }
 
+static long read_self_pss_kb(void) {
+    FILE *f = fopen("/proc/self/smaps", "r");
+    if (!f) {
+        return 0;
+    }
+    long total = 0;
+    char line[256];
+    while (fgets(line, sizeof(line), f)) {
+        long v = 0;
+        if (sscanf(line, "Pss: %ld", &v) == 1) {
+            total += v;
+        }
+    }
+    fclose(f);
+    return total;
+}
+
 static int write_worker_stats(const char *path) {
     long rss_kb = 0;
     long vsz_kb = 0;
+    long pss_kb = 0;
     if (read_self_status_kb(&rss_kb, &vsz_kb) != 0) {
         return -1;
     }
+    pss_kb = read_self_pss_kb();
     FILE *f = fopen(path, "w");
     if (!f) {
         return -1;
     }
-    fprintf(f, "%ld,%ld\n", rss_kb, vsz_kb);
+    fprintf(f, "%ld,%ld,%ld\n", rss_kb, vsz_kb, pss_kb);
     fclose(f);
     return 0;
 }
@@ -573,31 +593,18 @@ static void collect_metrics(run_metrics_t *metrics, double start_wall, double en
     read_self_status_kb(&metrics->max_rss_kb, &metrics->max_vsz_kb);
 }
 
-static int read_worker_stats(const char *path, long *rss_kb, long *vsz_kb) {
-    *rss_kb = 0;
-    *vsz_kb = 0;
-    FILE *f = fopen(path, "r");
-    if (!f) {
-        return -1;
-    }
-    if (fscanf(f, "%ld,%ld", rss_kb, vsz_kb) != 2) {
-        fclose(f);
-        return -1;
-    }
-    fclose(f);
-    return 0;
-}
-
 static void aggregate_worker_memory(const bench_config_t *cfg, run_mode_t mode, const worker_info_t *workers, run_metrics_t *metrics) {
     double t0 = now_wall();
     long total_rss = 0;
     long total_vsz = 0;
+    long total_pss = 0;
 
     long self_rss = 0;
     long self_vsz = 0;
     read_self_status_kb(&self_rss, &self_vsz);
     total_rss += self_rss;
     total_vsz += self_vsz;
+    total_pss += read_self_pss_kb();
 
     if (mode == MODE_PROCESS) {
         for (long i = 0; i < cfg->process_count; ++i) {
@@ -606,15 +613,22 @@ static void aggregate_worker_memory(const bench_config_t *cfg, run_mode_t mode,
             }
             long rss_kb = 0;
             long vsz_kb = 0;
-            if (read_worker_stats(workers[i].stats_path, &rss_kb, &vsz_kb) == 0) {
+            long pss_kb = 0;
+            FILE *f = fopen(workers[i].stats_path, "r");
+            if (f && fscanf(f, "%ld,%ld,%ld", &rss_kb, &vsz_kb, &pss_kb) == 3) {
+                fclose(f);
                 total_rss += rss_kb;
                 total_vsz += vsz_kb;
+                total_pss += pss_kb;
+            } else if (f) {
+                fclose(f);
             }
         }
     }
 
     metrics->agg_rss_kb = total_rss;
     metrics->agg_vsz_kb = total_vsz;
+    metrics->agg_pss_kb = total_pss;
     metrics->mem_snapshot_s = now_wall() - t0;
 }
 
@@ -635,13 +649,13 @@ static void write_results_csv(const char *results_csv, const char *run_id, const
         kernel[sizeof(kernel) - 1] = '\0';
     }
 
-    fprintf(csv, "run_id,timestamp,host,kernel,mode,worker_count,pi_digits,chunk_size,exit_code,wall_s,cpu_user_s,cpu_sys_s,cpu_pct,max_rss_kb,max_vsz_kb,agg_rss_kb,agg_vsz_kb,mem_snapshot_s,final_output_bytes\n");
+    fprintf(csv, "run_id,timestamp,host,kernel,mode,worker_count,pi_digits,chunk_size,exit_code,wall_s,cpu_user_s,cpu_sys_s,cpu_pct,max_rss_kb,max_vsz_kb,agg_rss_kb,agg_vsz_kb,agg_pss_kb,mem_snapshot_s,final_output_bytes\n");
     long workers = (strcmp(mode, "thread") == 0) ? cfg->thread_count : cfg->process_count;
-    fprintf(csv, "%s,%s,%s,%s,%s,%ld,%ld,%ld,%d,%.6f,%.6f,%.6f,%.2f,%ld,%ld,%ld,%ld,%.6f,%zu\n",
+    fprintf(csv, "%s,%s,%s,%s,%s,%ld,%ld,%ld,%d,%.6f,%.6f,%.6f,%.2f,%ld,%ld,%ld,%ld,%ld,%.6f,%zu\n",
         run_id, ts, host, kernel, mode, workers, cfg->pi_digits, cfg->chunk_size,
         metrics->exit_code, metrics->wall_s, metrics->cpu_user_s, metrics->cpu_sys_s,
         metrics->cpu_pct, metrics->max_rss_kb, metrics->max_vsz_kb,
-        metrics->agg_rss_kb, metrics->agg_vsz_kb, metrics->mem_snapshot_s, metrics->output_bytes);
+        metrics->agg_rss_kb, metrics->agg_vsz_kb, metrics->agg_pss_kb, metrics->mem_snapshot_s, metrics->output_bytes);
     fclose(csv);
 }
 
@@ -741,18 +755,17 @@ int main(int argc, char **argv) {
         metrics.exit_code = 1;
     }
 
+    double t_calc = now_wall();
+    collect_metrics(&metrics, t0, t_calc);
+    aggregate_worker_memory(&cfg, mode, workers, &metrics);
+
     int agg_rc = concatenate_outputs(&cfg, tmp_dir, final_path, worker_count);
     if (agg_rc != 0) {
         metrics.exit_code = 1;
     }
 
-    aggregate_worker_memory(&cfg, mode, workers, &metrics);
-
     remove_dir_tree(tmp_dir);
     unlink(pid_file);
-
-    double t1 = now_wall();
-    collect_metrics(&metrics, t0, t1);
     metrics.output_bytes = file_size(final_path);
 
     write_results_csv(results_csv, run_id, mode_name, &cfg, &metrics);
diff --git a/pi_digits_benchmark/run_pi_digits_benchmark.sh b/pi_digits_benchmark/run_pi_digits_benchmark.sh
index 1c47527..ccc4017 100755
--- a/pi_digits_benchmark/run_pi_digits_benchmark.sh
+++ b/pi_digits_benchmark/run_pi_digits_benchmark.sh
@@ -20,7 +20,7 @@ RUN_ID=$(date +"%Y%m%d_%H%M%S")
 
 RESULTS_CSV="$DIR/results.csv"
 if [ ! -f "$RESULTS_CSV" ]; then
-  echo "run_id,timestamp,host,kernel,mode,worker_count,pi_digits,chunk_size,exit_code,wall_s,cpu_user_s,cpu_sys_s,cpu_pct,max_rss_kb,max_vsz_kb,agg_rss_kb,agg_vsz_kb,mem_snapshot_s,final_output_bytes" > "$RESULTS_CSV"
+  echo "run_id,timestamp,host,kernel,mode,worker_count,pi_digits,chunk_size,exit_code,wall_s,cpu_user_s,cpu_sys_s,cpu_pct,max_rss_kb,max_vsz_kb,agg_rss_kb,agg_vsz_kb,agg_pss_kb,mem_snapshot_s,final_output_bytes" > "$RESULTS_CSV"
 fi
 
 run_mode() {