LLVM 5.0-RC1 Up For Testing

[phoronix]

Phoronix: LLVM 5.0-RC1 Up For Testing

Following the LLVM 5 branching earlier this week, release manager Hans Wennborg has now tagged the first release candidate...

http://www.phoronix.com/scan.php?page=news_item&px=LLVM-5.0-RC1-Testing

[sdack]

Hello,

some dimwit is spreading rumours about Phoronix being too stupid to apply compiler flags properly:

https://kristerw.blogspot.com/2017/07/phoronix-scimark-benchmarking-results.html

Please reassure your readers that Phoronix knows what they're doing, and thanks in advance.

[FireBurn]

If you do a CFLAGS="-O2 -O3" then -O3 takes precedence, CFLAGS="-O3 -O2" it's -O2 that takes precedence, it's a feature I sometimes take advantage of updating the flags for individual packages.

[sdack]

If you do a CFLAGS="-O2 -O3" then -O3 takes precedence, CFLAGS="-O3 -O2" it's -O2 that takes precedence, it's a feature I sometimes take advantage of updating the flags for individual packages.

Of course this is a well known behaviour not only found with compilers but many command-line tools. While it's possible for Phoronix to make mistakes do I have my doubts they'd be making one as obvious as this one. Anyhow, it's either a faux pas by Phoronix or an elaborate troll by a blogger.

[Ansla]

There actually was a problem with the PTS test, and it was fixed. It's not about being stupid or not, every piece of software has bugs, including the PTS.

http://openbenchmarking.org/test/pts/scimark2

 Revision History
pts/scimark2-1.3.0 [25 Jul 2017 09:13:45 UTC]
- Remove accidentally left in -O1 from build string