On Tuesday, February 8, 2022 at 2:16:31 PM UTC-5 Jeffrey Walton wrote:
> Hi Everyone, > > strciphr.cpp has a couple of classes that gave us trouble in the past. The > class functions were CFB_CipherTemplate<BASE>::ProcessData and > AdditiveCipherTemplate<S>::ProcessData. The classes encrypt or decrypt a > byte string: > > ProcessData(byte* outString, const byte* inString, size_t length); > > The problem arose because of in-place encryption or decryption (lots of > hand waiving): > > std::string str = "Super secret information"; > ... > ProcessData(&str[0], &str[0], str.size()); > > Clang and GCC started removing writes to outString due to alias violation. > Or I think it was alias violations. > > Initially we cut-in a separate buffer for the classes and then performed a > memcpy. That crushed performance per GH #1010 on some platforms. > > We recently removed the separate buffer and memcpy to restore performance. > We are now testing a volatile pointer to tame the optimizer. Testing is Ok > on the platforms that caused us troubles in the past. > > If you have some time, please test the "strciphr" branch of Crypto++ and > report any troubles you encounter. Here are the steps: > > $ git clone https://github.com/weidai11/cryptopp cryptopp-strciphr > $ cd cryptopp-strciphr > $ git checkout strciphr > > $ make -j 4 && ./cryptest.exe v && ./cryptest.exe tv all > > Please report any failures. > I'm going to merge this into master for more testing. We will probably be releasing Crypto++ 8.7 next weekend. Testing is welcomed. Jeff -- You received this message because you are subscribed to the Google Groups "Crypto++ Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to cryptopp-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/cryptopp-users/3951d12a-bca6-46ab-a49c-1528ba5a8b53n%40googlegroups.com.
