Wilson Snyder
Wilson Snyder
Sounds like there's no Verilator change needed here in the end - Unless I hear otherwise I'll close next week.
When you say test, you mean your tests, not Verilator's test_regress? What is the size of the sources? If
Also compare to the .o sizes. So for one large test what's the comparison of sizes of the .cpp, .gch and .o?
And note if you have e.g. -ggdb it will be much larger - avoid that ;)
What are the GCH sizes using clang?
./configure CXX=clang++ then rebuild verilator. I don't know if cocotb needs something else.
Another thing to try, edit verilated.mk: ``` $(VK_OBJS_FAST): %.o: %.cpp $(VK_PCH_H).fast.gch $(OBJCACHE) $(CXX) $(OPT_FAST) $(CXXFLAGS) $(CPPFLAGS) $(VK_PCH_I_FAST) -c -o $@ $< $(VK_OBJS_SLOW): %.o: %.cpp $(VK_PCH_H).slow.gch $(OBJCACHE) $(CXX) $(OPT_SLOW) $(CXXFLAGS) $(CPPFLAGS)...
Can you repeat with a "ccache -C" between each? I would expect if only the main.cpp file changes (not verilator generated .cpp's) and if the .gch is kept around the...
Thanks, does this pass on other simulators? IEEE isn't clear if not initializing everything (or using default) is legal or not.
Confirmed this is illegal. Verilator does have a check for wrong size (`"Assignment pattern with too many elements"`) so not sure why it isn't triggering. Might you be able to...