From 4e115d4b69d3978abb5220e33bcba6e316489502 Mon Sep 17 00:00:00 2001
From: Wilson Snyder <wsnyder@wsnyder.org>
Date: Mon, 3 Jun 2019 19:13:03 -0400
Subject: [PATCH] Fix performance when mulithreaded on 1 CPU, bug1455.

---
 Changes                                |  2 ++
 include/verilated_threads.h            | 15 +++++++++++---
 test_regress/t/t_bench_mux4k_onecpu.pl | 28 ++++++++++++++++++++++++++
 3 files changed, 42 insertions(+), 3 deletions(-)
 create mode 100755 test_regress/t/t_bench_mux4k_onecpu.pl

diff --git a/Changes b/Changes
index 92f1d8734..a423786f9 100644
--- a/Changes
+++ b/Changes
@@ -28,6 +28,8 @@ The contributors that suggested a given feature are shown in []. Thanks!
 
 ****  Fix invalid XML output due to special chars, bug1444. [Kanad Kanhere]
 
+****  Fix performance when mulithreaded on 1 CPU, bug1455. [Stefan Wallentowitz]
+
 
 * Verilator 4.014 2019-05-08
 
diff --git a/include/verilated_threads.h b/include/verilated_threads.h
index bd5b3d5f3..363a0ce35 100644
--- a/include/verilated_threads.h
+++ b/include/verilated_threads.h
@@ -52,6 +52,10 @@ public:
 
     // METHODS
     static vluint64_t yields() { return s_yields; }
+    static void yieldThread() {
+        ++s_yields;  // Statistics
+        std::this_thread::yield();
+    }
 
     // Block until notify() has occurred, then return.
     // If notify() has already occurred, return immediately.
@@ -63,11 +67,10 @@ public:
         unsigned ct = 0;
         while (VL_UNLIKELY(!notified())) {
             VL_CPU_RELAX();
-            ct++;
+            ++ct;
             if (VL_UNLIKELY(ct > VL_LOCK_SPINS)) {
                 ct = 0;
-                ++s_yields;  // Statistics
-                std::this_thread::yield();
+                yieldThread();
             }
         }
     }
@@ -149,8 +152,14 @@ public:
         return m_upstreamDepsDone.load(std::memory_order_acquire) == target;
     }
     inline void waitUntilUpstreamDone(bool evenCycle) const {
+        unsigned ct = 0;
         while (VL_UNLIKELY(!areUpstreamDepsDone(evenCycle))) {
             VL_CPU_RELAX();
+            ++ct;
+            if (VL_UNLIKELY(ct > VL_LOCK_SPINS)) {
+                ct = 0;
+                VlNotification::yieldThread();
+            }
         }
     }
 };
diff --git a/test_regress/t/t_bench_mux4k_onecpu.pl b/test_regress/t/t_bench_mux4k_onecpu.pl
new file mode 100755
index 000000000..e15715eb5
--- /dev/null
+++ b/test_regress/t/t_bench_mux4k_onecpu.pl
@@ -0,0 +1,28 @@
+#!/usr/bin/perl
+if (!$::Driver) { use FindBin; exec("$FindBin::Bin/bootstrap.pl", @ARGV, $0); die; }
+# DESCRIPTION: Verilator: Verilog Test driver/expect definition
+#
+# Copyright 2003 by Wilson Snyder. This program is free software; you can
+# redistribute it and/or modify it under the terms of either the GNU
+# Lesser General Public License Version 3 or the Perl Artistic License
+# Version 2.0.
+
+scenarios(vltmt => 1);
+
+top_filename("t/t_bench_mux4k.v");
+
+compile(
+    v_flags2 => ["--stats"],
+    );
+
+if (`numactl --show` !~ /cpu/) {
+    skip("No numactl available");
+} else {
+    execute(
+        run_env => 'numactl -m 0 -C 0,0,0,0,0,0,0,0',
+        check_finished => 1,
+        );
+    ok(1);
+}
+
+1;