Re:GCC 6.1 Released
by Esau on Wednesday April 27, 2016 @11:32AM
(#11581295)
The one thing I hear mentioned is that GCC supports more processors than Clang does. I believe this is the reason that NetBSD and OpenBSD still use GCC.
[ Parent
| on HN ]
What do you mean by support more processors?
The comment was referring to CPU architectures.
I understand that. Still, what was meant? C/C++ don't have mechanisms for automatic parallelization, or if there are obscure extensions they are not commonly used. Compilation is usually a single-processor task, with linear speedup obtained by doing multiple compilation units in parallel (make -jN). So in what way can a compiler be "better" for multi-processor / multi-core CPU architectures?
Re:GCC 6.1 Released
by MereInterest on Wednesday April 27, 2016 @10:24PM
(#11586207)
I believe that in this context, "more processors" meant "more different models of processors". That is, there exist processors for which gcc can produce binaries and clang cannot produce binaries.
Ah ok. More backends.
Re:GCC 6.1 Released
by groovy2shoes on Friday April 29, 2016 @04:34PM
(#11598573)
Well, support for more target architectures, really. I can easily imagine cases where a compiler could have multiple backends for the same target arch, so the number of backends could be misleading in that regard.
For a summary from the OpenBSD perspective, see here: http://marc.info/?l=openbsd-misc&m=137530560232232&w=2