It's Been Five Years Since The Open64 5.0 Compiler Release

[phoronix]

Phoronix: It's Been Five Years Since The Open64 5.0 Compiler Release

This week marked five years since the release of the Open64 5.0 compiler in what is the latest and likely last-ever release of this once-promising code compiler...

http://www.phoronix.com/scan.php?page=news_item&px=Open64-5.0-5-Year-Compiler

[ssokolow]

At least GCC and LLVM/Clang continue providing competition to one another and keep advancing in-step as open-source, well-supported compilers.

And, for specialized use cases, development continues on the v2 fork of OpenWatcom.

(The v2 fork is basically LibreOffice to the original branch's OpenOffice. According to the EDM2 wiki, it got started when contributors got fed up with the core team unilaterally throwing out their code.)

It may lag behind its better-supported contemporaries in supporting language features and distros don't want to include it because its custom copyleft license goes farther than the GPL about when you must make source available, but they do try and it's nice to have effortless cross-compilation to all supported targets (including old and/or niche stuff like COM files, the DOS4G/W and Win386 extenders, and OS/2) from a single install.

[slacka]

Yes the v2 Open Watcom fork is extremely active.

[torsionbar28]

Seeing as how AMD abandoned their "CMT" architecture CPU's, I don't think it's all that surprising that their compiler hasn't seen active development. Clearly the motivation behind a chipmaker releasing their own compiler, is superior performance on their own chips. History has shown this to be a lost cause on the CMT chips, and their marketshare is now extremely small.

With the upcoming Zen chips using SMT instead, and performance expected to be excellent, there may be renewed interest in this compiler. Or maybe not, since AMD has mentioned that Zen will be optimized to run existing binaries (i.e. those built with intel optimizations), without requiring a re-compile.

[coder111]

I'm not sure how AMD can do that without reporting "GenuineIntel" as CPU ID.

Intel's compiler will disable some optimizations if it detects non-intel CPU, and lots of Windows stuff ships compiled with ICC. I think Intel was even fined for this, but AFAIK they continue doing it.

Or maybe not, since AMD has mentioned that Zen will be optimized to run existing binaries (i.e. those built with intel optimizations), without requiring a re-compile.