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

hwcap supporting architectures?



Hi,

 I'm repackaging mesa, and I'd like to add support for hwcaps, but I'm
 having trouble finding documentation for it.  I know that on my PC the
 search path looks like this:

    /lib/tls/i686/mmx/cmov
    /lib/tls/i686/mmx
    /lib/tls/i686/cmov
    /lib/tls/i686
    /lib/tls/mmx/cmov
    /lib/tls/mmx
    /lib/tls/cmov
    /lib/tls
    /lib/i686/mmx/cmov
    /lib/i686/mmx
    /lib/i686/cmov
    /lib/i686
    /lib/mmx/cmov
    /lib/mmx
    /lib/cmov
    /lib
    /usr/lib/tls/i686/mmx/cmov
    /usr/lib/tls/i686/mmx
    /usr/lib/tls/i686/cmov
    /usr/lib/tls/i686
    /usr/lib/tls/mmx/cmov
    /usr/lib/tls/mmx
    /usr/lib/tls/cmov
    /usr/lib/tls
    /usr/lib/i686/mmx/cmov
    /usr/lib/i686/mmx
    /usr/lib/i686/cmov
    /usr/lib/i686
    /usr/lib/mmx/cmov
    /usr/lib/mmx
    /usr/lib/cmov
    /usr/lib

 Since the optimized i386 library does use MMX (it does perform runtime
 checking) and is compiled with -mcpu=i686 [0] (gcc doesn't seem to be
 emitting CMOVs though) I think the proper place is
 /usr/lib/i686/mmx/cmov (I have no intention of babysitting the
 generation of cmovs with each difference compiler release, therefore
 mmx/cmov instead of just mmx)

 My understanding is that this is also significant on sparc (-mcpu=v9)
 and that this belongs in /usr/lib/v9.  Is this right?

 Mesa upstream uses -mcpu=ev5 -mieee on alpha.  Is that ok?  Where does
 this belong into? /usr/lib/ev5?

 It also uses -mcpu=603 on powerpc.  From my understanding this is a lot
 hairier than other architectures since there's a whole load more
 subarchitectures which are potentially incompatible with each other.

 TIA,

    Marcelo
 
 [0] before someone asks, yes, this does make a difference, I didn't
     test extensively, but with a couple of applications I saw a
     difference of 5% in framerate; at obscene 300 fps 5% doesn't make a
     difference, but at 20 fps, it's one fps more which from a
     perceptual POV is significant.



Reply to: