Recommend -Os.

Changes

@@ -11,6 +11,8 @@ The contributors that suggested a given feature are shown in []. Thanks!
 
 ***   Fix genblk naming with directly nested generate blocks, #2176. [Alexander Grobman]
 
+****  Use gcc -Os in examples instead of -O2 for better average performance.
+
 ****  Fix undeclared VL_SHIFTR_WWQ, #2114. [Alex Solomatnikov]
 
 

bin/verilator

@@ -1964,19 +1964,18 @@ OPT, OPT_FAST, or OPT_SLOW lib/verilated.mk.  Or, use the -CFLAGS and/or
 the compiler or linker.  Or, just for one run, pass them on the command
 line to make:
 
-    make OPT_FAST="-O2 -fno-stack-protector" -f Vour.mk Vour__ALL.a
+    make OPT_FAST="-Os -fno-stack-protector" -f Vour.mk Vour__ALL.a
 
 OPT_FAST specifies optimizations for those programs that are part of the
 fast path, mostly code that is executed every cycle.  OPT_SLOW specifies
 optimizations for slow-path files (plus tracing), which execute only
 rarely, yet take a long time to compile with optimization on.  OPT
 specifies overall optimization and affects all compiles, including those
-OPT_FAST and OPT_SLOW control.  For best results, use OPT="-O2", and link
+OPT_FAST and OPT_SLOW control.  For best results, use OPT="-Os", 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 "-O3" may help, but gcc compile times may be excessive under O3 on
-even medium sized designs.  Alternatively, some larger designs report
-better performance using "-Os".
+such as "-O2" or "-O3" may help, but gcc compile times may be excessive
+under O3 on even medium sized designs.
 
 Unfortunately, using the optimizer with SystemC files can result in
 compiles taking several minutes.  (The SystemC libraries have many little

examples/cmake_tracing_c/CMakeLists.txt

@@ -34,5 +34,5 @@ add_executable(example ../make_tracing_c/sim_main.cpp)
 # Add the Verilated circuit to the target
 verilate(example COVERAGE TRACE
   INCLUDE_DIRS "../make_tracing_c"
-  VERILATOR_ARGS -f ../make_tracing_c/input.vc -O2 -x-assign 0
+  VERILATOR_ARGS -f ../make_tracing_c/input.vc -Os -x-assign 0
   SOURCES ../make_tracing_c/top.v)

examples/cmake_tracing_sc/CMakeLists.txt

@@ -41,7 +41,7 @@ add_executable(example ../make_tracing_sc/sc_main.cpp)
 # Add the Verilated circuit to the target
 verilate(example SYSTEMC COVERAGE TRACE
   INCLUDE_DIRS "../make_tracing_sc"
-  VERILATOR_ARGS -f ../make_tracing_sc/input.vc -O2 -x-assign 0
+  VERILATOR_ARGS -f ../make_tracing_sc/input.vc -Os -x-assign 0
   SOURCES ../make_tracing_sc/top.v)
 
 verilator_link_systemc(example)

examples/make_protect_lib/Makefile

@@ -34,7 +34,7 @@ VERILATOR_FLAGS =
 # Generate C++
 VERILATOR_FLAGS += -cc
 # Optimize
-VERILATOR_FLAGS += -O2 -x-assign 0
+VERILATOR_FLAGS += -Os -x-assign 0
 # Warn abount lint issues; may not want this on less solid designs
 VERILATOR_FLAGS += -Wall
 # Make waveforms

examples/make_tracing_c/Makefile

@@ -38,7 +38,7 @@ VERILATOR_FLAGS += -cc --exe
 # Generate makefile dependencies (not shown as complicates the Makefile)
 #VERILATOR_FLAGS += -MMD
 # Optimize
-VERILATOR_FLAGS += -O2 -x-assign 0
+VERILATOR_FLAGS += -Os -x-assign 0
 # Warn abount lint issues; may not want this on less solid designs
 VERILATOR_FLAGS += -Wall
 # Make waveforms

examples/make_tracing_c/Makefile_obj

@@ -38,7 +38,7 @@ endif
 # SystemC takes minutes to optimize, thus it is off by default.
 OPT_SLOW =
 # Fast path optimizations.  Most time is spent in these classes.
-OPT_FAST = -O2 -fstrict-aliasing
+OPT_FAST = -Os -fstrict-aliasing
 #OPT_FAST = -O
 #OPT_FAST =
 

examples/make_tracing_sc/Makefile

@@ -38,7 +38,7 @@ VERILATOR_FLAGS += -sc --exe
 # Generate makefile dependencies (not shown as complicates the Makefile)
 #VERILATOR_FLAGS += -MMD
 # Optimize
-VERILATOR_FLAGS += -O2 -x-assign 0
+VERILATOR_FLAGS += -Os -x-assign 0
 # Warn abount lint issues; may not want this on less solid designs
 VERILATOR_FLAGS += -Wall
 # Make waveforms

examples/make_tracing_sc/Makefile_obj

@@ -46,7 +46,7 @@ endif
 # SystemC takes minutes to optimize, thus it is off by default.
 OPT_SLOW =
 # Fast path optimizations.  Most time is spent in these classes.
-OPT_FAST = -O2 -fstrict-aliasing
+OPT_FAST = -Os -fstrict-aliasing
 #OPT_FAST = -O
 #OPT_FAST =
 

include/verilated.mk.in

@@ -86,7 +86,7 @@ LDLIBS   += $(VM_USER_LDLIBS)
 # SystemC takes minutes to optimize, thus it is off by default.
 #OPT_SLOW =
 # Fast path optimizations.  Most time is spent in these classes.
-#OPT_FAST = -O2 -fstrict-aliasing
+#OPT_FAST = -Os -fstrict-aliasing
 #OPT_FAST = -O
 #OPT_FAST =
 

test_regress/driver.pl

@@ -1048,7 +1048,7 @@ sub compile {
                                 "-DTEST_VERBOSE=\"".($self->{verbose} ? 1 : 0)."\"",
                                 "-DTEST_SYSTEMC=\"" .($self->sc ? 1 : 0). "\"",
                                 "-DCMAKE_PREFIX_PATH=\"".(($ENV{SYSTEMC_INCLUDE}||$ENV{SYSTEMC}||'')."/..\""),
-                                "-DTEST_OPT_FAST=\"" . ($param{benchmark}?"-O2":"") . "\"",
+                                "-DTEST_OPT_FAST=\"" . ($param{benchmark} ? "-Os" : "") . "\"",
                                 "-DTEST_VERILATION=\"" . $::Opt_Verilation . "\"",
                         ]);
             return 1 if $self->errors || $self->skips || $self->unsupporteds;
@@ -1066,7 +1066,7 @@ sub compile {
                                 "TEST_OBJ_DIR=$self->{obj_dir}",
                                 "CPPFLAGS_DRIVER=-D".uc($self->{name}),
                                 ($self->{verbose} ? "CPPFLAGS_DRIVER2=-DTEST_VERBOSE=1":""),
-                                ($param{benchmark}?"OPT_FAST=-O2":""),
+                                ($param{benchmark} ? "OPT_FAST=-Os" : ""),
                                 "$self->{VM_PREFIX}",  # bypass default rule, as we don't need archive
                                 ($param{make_flags}||""),
                         ]);