> Always a pleasure to count the number of clicks needed before you can establish that the only platforms are Mac and Windows. You're just attacking him because he's using a different compiler suite? If you know of problems with the Intel compiler suite, how about you let us know what they are instead of passively aggressively pointing to a single recent problem which from the bug report I'm not sure is the fault of Intel's compiler not being conformant or not functioning with a recent feature, or GCC accepting a non-conformant structure, and whether being cross-compiler portable in that way is even a goal of the Firefox team. Why so critical? He's making a bold claim, but you're not even responding to that. It is neither old nor outdated, it is not a "rebuild" and it does not use obsolete technologies or have security holes. adding TLS 1.1/1.2 support a while back, by offering OCSP-stapling, by keeping a close eye on encryption and the browser's security by continuing to port or re-implement security fixes that apply to Pale Moon as a browser. It's a true fork now and has been employing rapid development (as opposed to rapid release) to solidify this independent direction with its own focus and attempt at keeping the browser sane, lean, and offering users choice and stability.Īt the same time, Pale Moon's focus on security and evolving networking standards has added features and kept pace with those developments in other browsers, by e.g. But we've come a very long way since then with an increasing amount of different code being carried over each time it was re-based on later Firefox code. It was a rebuild with minor changes in the Firefox 4.0 era, yes. Pale Moon has been on a divergent path with its own code for a long time already. Rumor: "Pale Moon is just a rebranded rebuild of an old Firefox version" Note that the use cases of the compilers are very different, so it's probably reasonable for IBM/Intel/etc to have these as defaults and GCC/LLVM to have different defaults, but it doesn't change what you would want to compare if you want to compare generated code performance. People who compare compiler benchmarks rarely actually do this. So if you really want to compare performance numbers between compilers, you have to normalize them to something first, be it "go crazy and do what you want to my program" or "follow the standard" The default at O2 and above is Qstrict and Qnostrict_induction (IE "change the semantics of my program"). See for example, IBM's qstrict and qstrict_induction. A lot of compilers take "liberties" with the standard by default in the name of performance. Assuming you compare apples to apples (IE put them both in the same precision modes, tell them both to actually follow the standard, etc), it's not appreciably faster for anything except really specialized kernels.
0 Comments
Leave a Reply. |