On Tuesday 21 Feb 2017 21:26:20 Walter Dnes wrote: > On Tue, Feb 21, 2017 at 10:19:16PM +0000, Mick wrote > > > Perhaps I do not understand ... why should the chrooted system need > > to use different flags? > > See > https://gcc.gnu.org/onlinedocs/gcc-4.9.4/gcc/i386-and-x86-64-Options.html#i > 386-and-x86-64-Options > > The desktop is "-march=ivybridge" and the netbook is "-march=bonnell". > Neither of them can run the other's "-march=native" code. "ivybridge" > does not have MOVBE, while "bonnell" does not have SSE4.1, SSE4.2, > POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C. gcc will gladly > build for whatever Intel cpu you tell it to. I can build "bonnell" code > on the "ivybridge". But I can't run it on the "ivybridge" machine. > > As a compromise, I suppose I could declare the chroot "-march=core2". > "core2" is "bonnell" minus MOVBE, so both the netbook and the desktop > could run that code, with the netbook getting some, but not all, of the > possible optimization. > > I have a "hot backup" to my desktop, which has a "silvermont" cpu. > That's a newer Atom cpu, and it can run "bonnell" code, no problem. But > the "ivybridge" machine is not that old, and I prefer to keep my > machines until they start dying.
I probably misunderstood your intent. I thought you have 3 PCs and want to build binary packages for the oldest/slowest PC only, within a chroot on the fastest PC. The 2 faster PCs build their own packages and do not need a chroot build environment. -- Regards, Mick
signature.asc
Description: This is a digitally signed message part.
