diff --git a/src/gpu.cu b/src/gpu.cu
index 7d00c6b..6fbb239 100644
--- a/src/gpu.cu
+++ b/src/gpu.cu
@@ -1866,7 +1866,7 @@ void GpuThread::run() {
         uint64_t scaled_total_inputs = stage.total_inputs * stage.inputs_multiplier;
         auto [scaled_input_speed, input_speed_unit] = scale_si(scaled_total_inputs / stage.total_time);
         auto [scaled_output_speed, output_speed_unit] = scale_si(stage.total_outputs / stage.total_time);
-        std::printf("%-20s - %9.3f ms | %7.3f %% | %12" PRIu64 " -> %12" PRIu64
+        std::printf("%-20s - %9.3f ms | %7.3f %% | %16" PRIu64 " -> %12" PRIu64
                     " | 1 in %11.3f | %7.3f %cips | %7.3f %cops\n",
                     stage.name.c_str(), stage.total_time * 1e3,
                     stage.total_time / host_total_time * 100.0,
@@ -1882,7 +1882,7 @@ void GpuThread::run() {
       auto [scaled_input_speed, input_speed_unit] = scale_si(total_inputs / host_total_time);
       auto [scaled_output_speed, output_speed_unit] = scale_si(total_outputs / host_total_time);
       std::printf(
-          "total                - %9.3f ms | %7.3f %% | %12" PRIu64
+          "total                - %9.3f ms | %7.3f %% | %16" PRIu64
           " -> %12" PRIu64 " |                  | %7.3f %cips | %7.3f %cops\n",
           host_total_time * 1e3, kernel_total_time / host_total_time * 100.0,
           total_inputs, total_outputs, scaled_input_speed, input_speed_unit,
diff --git a/src/main.cpp b/src/main.cpp
index bf6603c..449b2d9 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -14,12 +14,19 @@
 #include <cstring>
 #include <cinttypes>
 #include <cstdio>
+#include <csignal>
 #include <chrono>
 #include <optional>
 #include <charconv>
 #include <algorithm>
 #include <random>
 
+static std::atomic_bool running{true};
+
+static void signal_handler(int) {
+    running.store(false, std::memory_order_relaxed);
+}
+
 #ifdef NO_GPU
 constexpr bool no_gpu = true;
 #else
@@ -244,7 +251,10 @@ int main_inner(int argc, char **argv) {
     }
 #endif
 
-    for (size_t i = 0;; i++) {
+    std::signal(SIGINT, signal_handler);
+    std::signal(SIGTERM, signal_handler);
+
+    for (size_t i = 0; running.load(std::memory_order_relaxed); i++) {
         if (threads != 0) {
             std::lock_guard lock(cpu_outputs.mutex);
             while (!cpu_outputs.queue.empty()) {
@@ -264,10 +274,20 @@ int main_inner(int argc, char **argv) {
         std::this_thread::sleep_for(std::chrono::seconds(1));
     }
 
+    std::printf("\nShutting down...\n");
+
 #ifndef NO_GPU
     for (auto &thread : gpu_threads) {
         (*thread).stop();
     }
+    std::printf("Waiting for GPU batches to finish...\n");
+    for (auto &thread : gpu_threads) {
+        (*thread).join();
+    }
+    {
+        uint64_t total_seeds_checked = seed_range.pos.load(std::memory_order_relaxed) - start_seed;
+        std::printf("Start seed: %" PRIi64 ", Total seeds checked: %" PRIu64 "\n", start_seed, total_seeds_checked);
+    }
 #endif
 #ifndef NO_CPU
     for (auto &thread : cpu_threads) {
@@ -283,11 +303,6 @@ int main_inner(int argc, char **argv) {
     }
 #endif
 
-#ifndef NO_GPU
-    for (auto &thread : gpu_threads) {
-        (*thread).join();
-    }
-#endif
 #ifndef NO_CPU
     for (auto &thread : cpu_threads) {
         (*thread).join();