Stefan Seefeld

@mloskot I think we should establish http://boostorg.github.io/gil to be the canonical project homepage.

Just adding a few thoughts here: * yes, adding benchmarks is a great idea, I agree to put them into a `benchmarks` subdirectory. * the choice of benchmark framework should...

Fair enough. My main point wasn't to promote my home-grown solution, but to suggest that we clarify what we want to benchmark and compare, and what the parameter space is,...

Keeping an eye on compilation times for Boost.GIL is of course important, but using an actual benchmark framework for that seems overkill to me. @sdebionne yes, image size and pixel...

@sdebionne I'm not quite sure I understand what you are asking about. Being able to evaluate operations lazily can be a useful optimization technique, in particular to chain / fuse...

So if I understand correctly, the `boost::crc_32_type.process_bytes()` function above computes ("accumulates") different numbers depending on whether you compile a `release` or `debug` variant. That sounds like a bug in either...

How does the checksum accumulator work ? I assume it's iterating over some sequence (of bytes ?). So it ought to be possible to pre-process the data, such that the...

Wow, excellent detective work, Dr. Watson ! :-) What does the last screenshot above actually show ? What are the hex numbers between the lines starting with "Checksum" ? (There...

OK, I see, that makes sense ! Unless you think you could fix the issue quickly, I suggest you add that test to GIL's `test/` (or perhaps we should create...

Has the error only ever been observed with MSVC ? I'm trying to figure out whether it would be worth for me to have a look into it, on Linux...