5 ms·

The comment was referring to CPU architectures.

• maaku 11y ago
  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?
  • MereInterest 11y ago
    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.
    • maaku 11y ago
      Ah ok. More backends.
      • groovy2shoes 11y ago
        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 http://marc.info/?l=openbsd-misc&m=137530560232232&w=2