Message ID | 1339328813-6269-5-git-send-email-s.martin49@gmail.com |
---|---|
State | Superseded |
Headers | show |
On 06/10/12 13:46, Samuel Martin wrote: > +# PowerPC cpu features > +config BR2_POWERPC_CPU_G3 > + bool > +config BR2_POWERPC_CPU_G4 > + bool > +config BR2_POWERPC_CPU_G5 > + bool I was expecting Altivec here, not the PowerPC CPU generations... These CPU features are relevant mostly to allow libraries to select optimized assembly code, and I don't see much use of G5-specific assembly. And anyway, the -mcpu options for a particular CPU are already passed by buildroot. Regards, Arnout
On 2012-06-12 03:15, Arnout Vandecappelle wrote: > On 06/10/12 13:46, Samuel Martin wrote: >> +# PowerPC cpu features >> +config BR2_POWERPC_CPU_G3 >> + bool >> +config BR2_POWERPC_CPU_G4 >> + bool >> +config BR2_POWERPC_CPU_G5 >> + bool > > I was expecting Altivec here, not the PowerPC CPU generations... > These CPU > features are relevant mostly to allow libraries to select optimized > assembly > code, and I don't see much use of G5-specific assembly. And anyway, > the > -mcpu options for a particular CPU are already passed by buildroot. > > Regards, > Arnout Also the G5 is 64-bit capable so one might expect some issues down the line since we're not ppc64-capable at the moment. Regards.
Hi Arnout, Gustavoz, 2012/6/12 Gustavo Zacarias <gustavo@zacarias.com.ar>: > On 2012-06-12 03:15, Arnout Vandecappelle wrote: > >> On 06/10/12 13:46, Samuel Martin wrote: >>> >>> +# PowerPC cpu features >>> +config BR2_POWERPC_CPU_G3 >>> + bool >>> +config BR2_POWERPC_CPU_G4 >>> + bool >>> +config BR2_POWERPC_CPU_G5 >>> + bool >> >> >> I was expecting Altivec here, not the PowerPC CPU generations... These >> CPU >> features are relevant mostly to allow libraries to select optimized >> assembly >> code, and I don't see much use of G5-specific assembly. And anyway, the >> -mcpu options for a particular CPU are already passed by buildroot. I would expect nothing less ;-) >> >> Regards, >> Arnout > > > Also the G5 is 64-bit capable so one might expect some issues down the line > since we're not ppc64-capable at the moment. > Regards. > So, to sum up and make the relation with the question I asked on the pacth 6/6 of the serie, you suggest I always turn off the ENABLE_POWERPC option in the opencv package. That's it? BTW, I can add an Altivec symbol, and let you (or others) use it in packages that may take benefit from such optimization. Let me know... Regards,
diff --git a/target/Config.in.arch b/target/Config.in.arch index 49c01e5..02aa931 100644 --- a/target/Config.in.arch +++ b/target/Config.in.arch @@ -600,6 +600,14 @@ config BR2_xtensa_core_name default "dc232b" if BR2_xtensa_dc232b # default "s5000" if BR2_xtensa_s5000 +# PowerPC cpu features +config BR2_POWERPC_CPU_G3 + bool +config BR2_POWERPC_CPU_G4 + bool +config BR2_POWERPC_CPU_G5 + bool + choice prompt "Target Architecture Variant" depends on BR2_powerpc @@ -640,12 +648,18 @@ config BR2_powerpc_630 bool "630" config BR2_powerpc_740 bool "740" + select BR2_POWERPC_CPU_G3 config BR2_powerpc_7400 bool "7400" + select BR2_POWERPC_CPU_G3 + select BR2_POWERPC_CPU_G4 config BR2_powerpc_7450 bool "7450" + select BR2_POWERPC_CPU_G3 + select BR2_POWERPC_CPU_G4 config BR2_powerpc_750 bool "750" + select BR2_POWERPC_CPU_G3 config BR2_powerpc_801 bool "801" config BR2_powerpc_821 @@ -656,6 +670,9 @@ config BR2_powerpc_860 bool "860" config BR2_powerpc_970 bool "970" + select BR2_POWERPC_CPU_G3 + select BR2_POWERPC_CPU_G4 + select BR2_POWERPC_CPU_G5 config BR2_powerpc_8540 bool "8540 / e500v1" config BR2_powerpc_8548
Signed-off-by: Samuel Martin <s.martin49@gmail.com>