ASR pass for unique symbols #2149
Conversation
|
@certik This PR now passes all the integration tests with LLVM and C backend with every possible name-mangling. It just fails for one test and that has something to do with nested structs and External Symbol. Can you please test this PR against your use cases? |
|
Both the test cases in #2129 as well as |
|
I would structure this ASR pass like this:
We can then keep extending the logic in the first pass as needed, and it could be that there is no single way forward, and users will select the behavior on the command line that they need for their application / build system. |
|
Register the pass in https://github.com/lcompilers/lpython/blob/main/src/libasr/gen_pass.py and run the script to generate the |
src/libasr/pass/unique_symbols.cpp
Outdated
| std::map<std::string, ASR::symbol_t*> current_scope; | ||
|
|
||
| UniqueSymbolVisitor(Allocator& al_, | ||
| std::map<ASR::symbol_t*, std::string> sn) : al(al_), sym_to_new_name(sn){} |
There was a problem hiding this comment.
a hashtable might be faster here.
| @@ -1616,7 +1635,7 @@ int main(int argc, char *argv[]) | |||
| app.require_subcommand(0, 1); | |||
| CLI11_PARSE(app, argc, argv); | |||
|
|
|||
| lcompilers_unique_ID = LCompilers::get_unique_ID(); | |||
| lcompilers_unique_ID = separate_compilation ? LCompilers::get_unique_ID(): ""; | |||
There was a problem hiding this comment.
I would not do that here, but rather have lcompilers_unique_ID always have a unique ID. But have this logic of turning it on and off in asr_scopes. If asr_scopes doesn't have access to separate_compilation then let's create another variable lcompilers_unique_ID_separate_compilation and have that variable on and off as needed.
#2129, #2142