diff --git a/Changes b/Changes
index 889f493b0..b1b2bc8f6 100644
--- a/Changes
+++ b/Changes
@@ -33,6 +33,8 @@ The contributors that suggested a given feature are shown in []. Thanks!
       medium and large designs with default options. User makefiles may
       require changes.
 
+****  The run-time library is now compiled with -Os by default. (#2369, #2373)
+
 
 * Verilator 4.034 2020-05-03
 
diff --git a/bin/verilator b/bin/verilator
index e3de112cc..3700ea166 100755
--- a/bin/verilator
+++ b/bin/verilator
@@ -2103,13 +2103,21 @@ OPT_FAST specifies optimizations for those parts of the program that are on the
 fast path. This is mostly code that is executed every cycle.  OPT_SLOW
 specifies optimizations for slow-path files, which execute only rarely, yet
 take a long time to compile with optimization on.  OPT_SLOW is ignored if
-VM_PARALLEL_BUILDS is not 1, in which case all code is compliled with OPT_FAST.
+VM_PARALLEL_BUILDS is not 1, in which case all code is compiled with OPT_FAST.
 See also the C<--output-split> option.  OPT specifies overall optimization and
 affects all compiles, including those OPT_FAST and OPT_SLOW control.  For best
 results, use OPT="-Os -march=native", and link with "-static".  Nearly the same
 results can be had with much better compile times with OPT_FAST="-O1
 -fstrict-aliasing".  Higher optimization such as "-O2" or "-O3" may help, but
 gcc compile times may be excessive under O3 on even medium sized designs.
+There is a third variable, OPT_GLOBAL, which applies to common code in the
+run-time library used by verilated models. This is set to "-Os" by default
+and there should rarely be a need to change it. As the run-time library is
+small in comparison to a lot of verilated models, disabling optimization on
+the run-time library should not have a serious effect on overall compilation
+time, but can have highly detrimental effect on run-time performance,
+especially with tracing. The OPT variable also applies to files that are
+controlled by OPT_GLOBAL.
 
 Unfortunately, using the optimizer with SystemC files can result in
 compiles taking several minutes.  (The SystemC libraries have many little
@@ -2670,7 +2678,7 @@ Verilator and add the generated C++ sources to the target specified.
     verilate(target SOURCES source ... [TOP_MODULE top] [PREFIX name]
              [TRACE] [TRACE_FST] [SYSTEMC] [COVERAGE]
              [INCLUDE_DIRS dir ...] [OPT_SLOW ...] [OPT_FAST ...]
-             [DIRECTORY dir] [VERILATOR_ARGS ...])
+             [OPT_GLOBAL ..] [DIRECTORY dir] [VERILATOR_ARGS ...])
 
 Lowercase and ... should be replaced with arguments, the uppercase parts
 delimit the arguments and can be passed in any order, or left out entirely
@@ -2761,6 +2769,11 @@ optimization level to improve compile times with large designs.
 
 Optional. Set compiler flags for the fast path.
 
+=item OPT_GLOBAL
+
+Optional. Set compiler flags for the common run-time library used by verilated
+models.
+
 =item DIRECTORY
 
 Optional. Set the verilator output directory. It is preferable to use the
diff --git a/include/verilated.mk.in b/include/verilated.mk.in
index 9c583b083..ac8ebfb69 100644
--- a/include/verilated.mk.in
+++ b/include/verilated.mk.in
@@ -92,6 +92,11 @@ LDLIBS   += $(VM_USER_LDLIBS)
 #OPT_FAST = -Os -fstrict-aliasing
 #OPT_FAST = -O
 #OPT_FAST =
+# Optimization applied to the common run-time library used by verilated models.
+# For compatibility this is called OPT_GLOBAL even though it only applies to
+# files in the run-time library. Normally there should be no need for the user
+# to change this.
+OPT_GLOBAL = -Os
 
 #######################################################################
 ##### SystemC builds
@@ -170,6 +175,9 @@ VK_SLOW_OBJS = $(addsuffix .o, $(VM_SLOW))
 
 VK_USER_OBJS   = $(addsuffix .o, $(VM_USER_CLASSES))
 
+# Note VM_GLOBAL_FAST and VM_GLOBAL_SLOW holds the files required from the
+# run-time library. In practice everything is actually in VM_GLOBAL_FAST,
+# but keeping the distinction for compatibility for now.
 VK_GLOBAL_OBJS = $(addsuffix .o, $(VM_GLOBAL_FAST) $(VM_GLOBAL_SLOW))
 
 ifneq ($(VM_PARALLEL_BUILDS),1)
@@ -197,15 +205,16 @@ $(VM_PREFIX)__ALL.a: $(VK_OBJS)
 ### Compile rules
 
 ifneq ($(VM_DEFAULT_RULES),0)
-$(VM_PREFIX)__ALL.o: $(VM_PREFIX)__ALL.cpp
-	$(OBJCACHE) $(CXX) $(CXXFLAGS) $(CPPFLAGS) $(OPT_FAST) -c -o $@ $<
-
-# Anything not in $(VK_SLOW_OBJS), including verilated.o use this rule
+# Anything not in $(VK_SLOW_OBJS) or $(VK_GLOBAL_OBJS), including verilated.o
+# and user files passed on the Verilator command line use this rule.
 %.o: %.cpp
 	$(OBJCACHE) $(CXX) $(CXXFLAGS) $(CPPFLAGS) $(OPT_FAST) -c -o $@ $<
 
 $(VK_SLOW_OBJS): %.o: %.cpp
 	$(OBJCACHE) $(CXX) $(CXXFLAGS) $(CPPFLAGS) $(OPT_SLOW) -c -o $@ $<
+
+$(VK_GLOBAL_OBJS): %.o: %.cpp
+	$(OBJCACHE) $(CXX) $(CXXFLAGS) $(CPPFLAGS) $(OPT_GLOBAL) -c -o $@ $<
 endif
 
 #Default rule embedded in make:
diff --git a/test_regress/CMakeLists.txt b/test_regress/CMakeLists.txt
index e07568e89..0cb22c4b7 100644
--- a/test_regress/CMakeLists.txt
+++ b/test_regress/CMakeLists.txt
@@ -13,7 +13,8 @@
 ######################################################################
 
 cmake_minimum_required(VERSION 3.8)
-set(TEST_REQUIRED_VARS NAME CSOURCES OPT_FAST VERILATOR_ROOT VERILATOR_ARGS
+set(TEST_REQUIRED_VARS NAME CSOURCES OPT_FAST OPT_GLOBAL
+                            VERILATOR_ROOT VERILATOR_ARGS
                             VERILATOR_SOURCES SYSTEMC VERBOSE VERILATION)
 foreach(var ${TEST_REQUIRED_VARS})
   if (NOT DEFINED TEST_${var})
@@ -74,6 +75,9 @@ endif()
 if(TEST_OPT_FAST)
   list(APPEND verilate_ARGS OPT_FAST ${TEST_OPT_FAST})
 endif()
+if(TEST_OPT_GLOBAL)
+  list(APPEND verilate_ARGS OPT_GLOBAL ${TEST_OPT_GLOBAL})
+endif()
 if(TEST_THREADS)
   list(APPEND verilate_ARGS THREADS ${TEST_THREADS})
 endif()
diff --git a/test_regress/driver.pl b/test_regress/driver.pl
index 570a9ecaa..a8465552f 100755
--- a/test_regress/driver.pl
+++ b/test_regress/driver.pl
@@ -1112,6 +1112,7 @@ sub compile {
                                 "-DTEST_SYSTEMC=\"" .($self->sc ? 1 : 0). "\"",
                                 "-DCMAKE_PREFIX_PATH=\"".(($ENV{SYSTEMC_INCLUDE}||$ENV{SYSTEMC}||'')."/..\""),
                                 "-DTEST_OPT_FAST=\"" . ($param{benchmark} ? "-Os" : "") . "\"",
+                                "-DTEST_OPT_GLOBAL=\"" . ($param{benchmark} ? "-Os" : "-O0") . "\"",
                                 "-DTEST_VERILATION=\"" . $::Opt_Verilation . "\"",
                         ]);
             return 1 if $self->errors || $self->skips || $self->unsupporteds;
@@ -1130,6 +1131,7 @@ sub compile {
                                 "CPPFLAGS_DRIVER=-D".uc($self->{name}),
                                 ($self->{verbose} ? "CPPFLAGS_DRIVER2=-DTEST_VERBOSE=1":""),
                                 ($param{benchmark} ? "OPT_FAST=-Os" : ""),
+                                ($param{benchmark} ? "" : "OPT_GLOBAL=-O0"),
                                 "$self->{VM_PREFIX}",  # bypass default rule, as we don't need archive
                                 ($param{make_flags}||""),
                         ]);
diff --git a/test_regress/t/t_flag_csplit.pl b/test_regress/t/t_flag_csplit.pl
index 8ca9c6f43..1e126525c 100755
--- a/test_regress/t/t_flag_csplit.pl
+++ b/test_regress/t/t_flag_csplit.pl
@@ -38,6 +38,7 @@ while (1) {
               ($opt_verbose ? "CPPFLAGS_DRIVER2=-DTEST_VERBOSE=1":""),
               "OPT_FAST=-O2",
               "OPT_SLOW=-O0",
+              "OPT_GLOBAL=-Os",
               ($param{make_flags}||""),
         ]);
 
@@ -111,13 +112,15 @@ sub check_gcc_flags {
     while (defined (my $line = $fh->getline)) {
         chomp $line;
         print ":log: $line\n" if $Self->{verbose};
-        if ($line =~ /\.cpp/) {
+        if ($line =~ /$Self->{VM_PREFIX}\S*\.cpp/) {
             my $filetype = ($line =~ /Slow|Syms/) ? "slow":"fast";
             my $opt = ($line !~ /-O2/) ? "slow":"fast";
             print "$filetype, $opt, $line\n" if $Self->{verbose};
             if ($filetype ne $opt) {
                 error("${filetype} file compiled as if was ${opt}: $line");
             }
+        } elsif ($line =~ /\.cpp/ and $line !~ /-Os/) {
+            error("library file not compiled with OPT_GLOBAL: $line");
         }
     }
 }
diff --git a/verilator-config.cmake.in b/verilator-config.cmake.in
index 50edfdcd9..b13b32301 100644
--- a/verilator-config.cmake.in
+++ b/verilator-config.cmake.in
@@ -116,7 +116,7 @@ define_property(TARGET
 function(verilate TARGET)
   cmake_parse_arguments(VERILATE "COVERAGE;TRACE;TRACE_FST;SYSTEMC"
                                  "PREFIX;TOP_MODULE;THREADS;DIRECTORY"
-                                 "SOURCES;VERILATOR_ARGS;INCLUDE_DIRS;OPT_SLOW;OPT_FAST"
+                                 "SOURCES;VERILATOR_ARGS;INCLUDE_DIRS;OPT_SLOW;OPT_FAST;OPT_GLOBAL"
                                  ${ARGN})
   if (NOT VERILATE_SOURCES)
     message(FATAL_ERROR "Need at least one source")
@@ -290,6 +290,11 @@ function(verilate TARGET)
       set_property(SOURCE "${VFAST}" APPEND_STRING PROPERTY COMPILE_FLAGS " ${OPT_FAST}")
     endforeach()
   endforeach()
+  foreach(VGLOBAL ${${VERILATE_PREFIX}_GLOBAL})
+    foreach(OPT_GLOBAL ${VERILATE_OPT_GLOBAL} ${${VERILATE_PREFIX}_USER_CFLAGS})
+      set_property(SOURCE "${VGLOBAL}" APPEND_STRING PROPERTY COMPILE_FLAGS " ${OPT_GLOBAL}")
+    endforeach()
+  endforeach()
 
   target_include_directories(${TARGET} PUBLIC "${VERILATOR_ROOT}/include"
                                                "${VERILATOR_ROOT}/include/vltstd")