Re: [lmi] Compiling takes longer with gcc-4.9.2

[Vadim Zeitlin]

On Tue, 05 Jan 2016 03:19:45 +0000 Greg Chicares <address@hidden> wrote:

GC> Thus, whether (MinGW-w64 native) beats (MinGW-w64 Cygwin cross compiler)
GC> seems to depend on...well, who knows--maybe the hardware in your two
GC> different hosts. But neither MinGW-w64 choice is universally faster.

 Yes, and this is quite surprising. I could understand that the relative
amount of overhead added by Cygwin is more or less important depending on
the hardware, but I just can't figure out how can it beat the native
compiler.

GC> Your table above actually lists two different versions of MinGW-w64
GC> (last digit '1' or '2').

 This is because Cygwin version is 4.9.2 while Debian version is 4.9.1 and
I wanted to ensure that the binaries created by the native compiler used
inside the VM and Debian cross-compiler were compatible because I thought
to use the libraries created by the former with the binaries generated by
the latter at some moment. But then I decided it was a bad idea and that if
we want to cross-compile, we should cross-compile everything instead of
relying on doing part of the work inside the VM and part of it outside it.

 Also, 4.9.1 and 4.9.2 have the same ABI, so it shouldn't matter which one
is used in any case. And I don't think (although I didn't measure it,
should I?) there is any difference in build times between them neither, at
least I don't see anything that would seem to be related to the compiler
performance in the list of fixes in 4.9.2 at

https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2

 Regards,
VZ