[cfe-dev] [LLVMdev] RFC: Upcoming Build System Changes

David A. Greene greened at obbligato.org
Tue Nov 1 17:05:46 PDT 2011


Joerg Sonnenberger <joerg at britannica.bec.de> writes:

> On Tue, Nov 01, 2011 at 06:27:35PM -0500, David A. Greene wrote:
>> Here is actual data comparing an empty LLVM build done recursively (the
>> LLVM build) and non-recursively (the Cray build).
>> 
>> See this?
>> 
>> 0inputs+0outputs (0major+671804minor)pagefaults 0swaps
>> 
>> vs. this?
>> 
>> 0inputs+0outputs (0major+184605minor)pagefaults 0swaps
>> 
>> That's I/O.
>
> Sorry to disappoint you, but no, this isn't I/O. You have just shown
> that a LLVM build touches more mmap'd pages.

It probably is I/O.  All those directories and Makefiles have to go
somewhere.  :)

I'll admit I don't know for sure without more detailed analysis, but it
seems like a reasonable hypothesis to me.

In any event, who cares?  The non-recursive build is much faster, and
that's the bottom line in this sub-thread, right?

It's a bit silly that we're carrying on this much about this when the
bigger issue is clarity and maintenance of the build system.  Daniel's
approach is far from perfect (and I think he'd be the first to admit
it), but it seems better to me than any other specific proposal given so
far.

                           -Dave



More information about the cfe-dev mailing list