On 2013-07-04 7:20 PM, Peter FELECAN wrote: > Maybe I misunderstood the issue but openssl is implemented mostly in C > and there is no C++. Consequently, the above described benchmark is > about code generation when the source is C but we are discussing g++ > against SSCC as C++ compilers, wxwidgets being mainly C++.
We are, but the C/C++ bit is really only relevant for the ABI. For binary code performance, both gcc and g++ use the same backend, as explained eg here: http://www.redhat.com/magazine/002dec04/features/gcc/#internal-org-gcc If I remember correctly, some years after that, Fortran was included too. As far as I know, Solaris Studio (and all compilers) work in the same ways, because it's simply easier to maintain a single optimization engine (or rather, to integrate several of them a single time from processor teams for Intel/AMD/SPARC64/UltraSPARC-T design). You can see it for Studio in that the patch that contains binary code fixes is "Compiler Common components", 148906-06. Those specific to C/C++ don't relate to binary optimization. So openssl is indeed useful to compare binary performance. C or C++ does not matter for that particular benchmark. > BTW, is wxwidgets in the class of libraries requiring such a great > optimization, > such as openssl, that it doesn't suffer a little bit of lag on SPARC? I would say, any interactive UI lib needs as much optimization as it can :-) As for the specifics, I don't know how to bench it and if it would be perceptible or not. Any idea how to do that? Laurent _______________________________________________ maintainers mailing list [email protected] https://lists.opencsw.org/mailman/listinfo/maintainers .:: This mailing list's archive is public. ::.
