[llvm-dev] difference with autotools, cmake and ninja building methods

[David Chisnall via llvm-dev]

On 1 Dec 2015, at 18:01, David Blaikie <dblaikie at gmail.com> wrote:
> 
> If you plan to run the regression tests with that build, you'll probably lose the time you gain by linking - the startup of all the llvm tools for all those little tests adds up. A "make check-all" of clang/llvm, at least on my machine (32 core, 32GB of RAM, running of an SSD) is slower (even if you just touch one cc file - causing maximal linking, minimal compiling, etc) with a shared library build, unfortunately.

Are you using bfd ld?  On the machine I use for testing (32 cores, 256GB RAM, spinning rust disks but with a 512GB SSD used for L2ARC / ZIL), it takes longer to do the final link of a debug build of clang than it takes to run the entire test suite.  With ld64 or gold, the link is much, much faster (the RAM requirements are also a lot lower) - the final link of a debug build with ld64 is only a few seconds.

David