[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: RH XFree86 spec file question



Thanks for the info!  This is good stuff to learn - I appreciate you posting
it.
James

------------------------------------------------------------------------------
James Fowler
Graduate Student Political Science & Latin America
Arizona State University
jamesvi@asu.edu        					ICQ: 22088266
-----------------------------------------------------------------------------


On Thu, 2 Dec 1999 Jay.Estabrook@compaq.com wrote:

> 
> >>> Craig P Prescott said:
> > 
> > Was building XFree86 on my EV45 (aka 21064A) AlphaStation 250 last
> > night, and I was wondering why everything was getting compiled with
> > "-Wa,-m21164a".  Then saw this in the spec file:
> > 
> > %ifarch alpha
> > make World -C xc CDEBUGFLAGS="$RPM_OPT_FLAGS -Wa,-m21164a"
> > # we are having problems with the compiler on alpha.
> > make -C xc/programs/xterm CDEBUGFLAGS="-Wa,-m21164a"
> > 
> > Anyway, this just seemed kinda weird, so I thought I'd ask about it
> > (I'm sure I'll learn something).  Anybody know what is up with this?
> > Just seems like a bad idea to tell the assembler (in my case) the
> > wrong chip...
> 
> First off, the -Wa,-m21164a is a flag for GAS, the assembler. It says to
> allow instructions that are supported on the 21164A, aka EV56 CPU, or later
> ones, to be processed if found in the assembly code the compiler
> generates. Without this flag, the assembler will give an error if it finds
> any instructions that are not supported on the base EV4/EV5 CPUs, which
> would include the byte/word ones. And the compiler is not being told to
> generate those instructions; that requires the -mcpu= flag, IIRC.
> 
> Secondly, those byte/word ones are *required* in order for certain
> operations, that the Xserver needs to do, to work correctly on the EV6
> CPU. They appear only in "__asm__" constructs, and they will only be
> executed if the Xserver is actually running on an EV6 machine, so there's no
> problem on older ones. They are there because we want a single Xserver
> binary to work on all the Alpha CPUs, much along the same lines as the
> "generic" kernels that are now available.
> 
> --Jay++
> 
> ---------------------------------------------------------------
> Jay A Estabrook                       Alpha Engineering - LINUX
> Compaq Computer Corp. - MRO1-2/K20    (508) 467-2080
> 200 Forest Street, Marlboro MA 01752  Jay.Estabrook@compaq.com
> ---------------------------------------------------------------
> 
> -- 
> To unsubscribe: send e-mail to axp-list-request@redhat.com with
> 'unsubscribe' as the subject.  Do not send it to axp-list@redhat.com
> 
> 



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index] []