PPCSymbolDB: Improve symbol map column detection #8932
Conversation
|
|
||
| // Two columns format: | ||
| // 80004000 zz_80004000_ | ||
| if (!(oss >> word) || word.length() != 8 || !is_hex_str(word)) |
There was a problem hiding this comment.
Did you consider isxdigit coupled with std::all_of ? Not sure if there are any advantages to that.
There was a problem hiding this comment.
I did and I need to wrap isxdigit(int) around a lambda with an unsigned cast to avoid an assert error (char might be signed and negative) or use the locale version of it. However, I don't know if isxdigit is faster or not.
| // 80004000 zz_80004000_ | ||
| if (!(oss >> word) || word.length() != 8 || !is_hex_str(word)) | ||
| continue; | ||
| column_count = 2; |
There was a problem hiding this comment.
Would it be faster to tokenize the string by spaces to determine the column count, then later validate whether they are hex or not?
There was a problem hiding this comment.
I don't think it will be faster since we don't need to tokenize the whole line if the first word isn't a hexstring. If I understand correctly it will also require a container like a vector to store the string tokens and then process them which might be slower.
|
@jordan-woyak Good catch, done. Rebased on latest master and I renamed |
|
Just a simple rebase. I've tested it against the map files I have and didn't find any regression. |
This PR fixes https://dolp.in/i11995.
Ready to be review & merged.