Re: [OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
On Mon, Jul 31, 2017 at 12:31:58PM -0500, Mark Hatle wrote: > On 7/31/17 12:16 PM, Armin Kuster wrote: > > Signed-off-by: Armin Kuster> > --- > > meta/conf/machine/include/arm/arch-armv8.inc | 25 + > > 1 file changed, 25 insertions(+) > > > > diff --git a/meta/conf/machine/include/arm/arch-armv8.inc > > b/meta/conf/machine/include/arm/arch-armv8.inc > > index 5e832fa..dc1ba5e 100644 > > --- a/meta/conf/machine/include/arm/arch-armv8.inc > > +++ b/meta/conf/machine/include/arm/arch-armv8.inc > > @@ -1 +1,26 @@ > > +DEFAULTTUNE ?= "armv8-a" > > do we want the '-a'? The other arm (7) are of the format armv7a (no '-'). On the other hand, gcc went with armv7-a instead of armv7a... So, maybe having '-a' is actually more aligned with others? > > + > > +TUNEVALID[armv8-a] = "Enable instructions for ARMv8-a" > > +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', ' > > -march=armv8-a', '', d)}" > > + > > +TUNEVALID[crypto] = "Enable crypto instructions for ARMv8-a" > > +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'crypto', '+crypto', > > '', d)}" > > + > > +MACHINEOVERRIDES =. "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', > > 'armv8a:', '' ,d)}" > > + > > require conf/machine/include/arm/arch-arm64.inc > > + > > +AVAILTUNES += "armv8-a armv8-a-crypto" > > +ARMPKGARCH_tune-armv8-a ?= "armv8-a" > > + > > +# Enable Crypto extension. This also enables Advanced SIMD and > > floating-point instructions > > +ARMPKGARCH_tune-armv8-a-crypto ?= "armv8-a" > > + > > +TUNE_FEATURES_tune-armv8-a = "aarch64 armv8-a" > > +TUNE_FEATURES_tune-armv8-a-crypto = "${TUNE_FEATURES_tune-armv8-a} crypto" > > + > > +PACKAGE_EXTRA_ARCHS_tune-armv8-a = "aarch64 armv8-a" > > +PACKAGE_EXTRA_ARCHS_tune-armv8-a-crypto = > > "${PACKAGE_EXTRA_ARCHS_tune-armv8-a} armv8-a-crypto" > > + > > +BASE_LIB_tune-armv8-a = "${BASE_LIB_tune-aarch64}" > > +BASE_LIB_tune-armv8-a-crypto = "${BASE_LIB_tune-aarch64}" > > > > -- > ___ > Openembedded-core mailing list > Openembedded-core@lists.openembedded.org > http://lists.openembedded.org/mailman/listinfo/openembedded-core -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
On 8/3/17 10:56 AM, akuster808 wrote: > > > On 08/01/2017 08:46 AM, Khem Raj wrote: >> On Tue, Aug 1, 2017 at 8:24 AM, Mark Hatlewrote: >>> On 8/1/17 10:20 AM, Khem Raj wrote: On Tue, Aug 1, 2017 at 8:17 AM, akuster808 wrote: > > On 07/31/2017 12:04 PM, Khem Raj wrote: >> >> On 7/31/17 10:51 AM, Mark Hatle wrote: >>> On 7/31/17 12:40 PM, akuster808 wrote: On 07/31/2017 10:31 AM, Mark Hatle wrote: > On 7/31/17 12:16 PM, Armin Kuster wrote: >> Signed-off-by: Armin Kuster >> --- >>meta/conf/machine/include/arm/arch-armv8.inc | 25 >> + >>1 file changed, 25 insertions(+) >> >> diff --git a/meta/conf/machine/include/arm/arch-armv8.inc >> b/meta/conf/machine/include/arm/arch-armv8.inc >> index 5e832fa..dc1ba5e 100644 >> --- a/meta/conf/machine/include/arm/arch-armv8.inc >> +++ b/meta/conf/machine/include/arm/arch-armv8.inc >> @@ -1 +1,26 @@ >> +DEFAULTTUNE ?= "armv8-a" > do we want the '-a'? The other arm (7) are of the format armv7a (no > '-'). works for me either way. While we are at it. How would we want ‘armv8.1-a’, ‘armv8.2-a’, ‘armv8.3-a' formated as? >>> My preference is to drop the '-'. As for the '.', I'm not sure.. not >>> something >>> we've run across before. >>> >>> We could just drop it (the '.'), but it really depends on if armv81a >>> would >>> confuse someone (familiar with arm) or not. >> I would suggest to also sync with other distros and ensure that we dont >> do something different. Its very costly later. Since applications get >> ported to most common combination > Sync what part? naming of file? ( sorry lost regarding your comment) naming convention for arch e.g. >>> I'd say Debian/Ubuntu -- Fedora/RH would be the two primary sources I'd >>> look at. >>> Between debian package names and RPM package names, it should give a clue >>> if >>> there is any community standard forming for those names. >>> >>> (May still be too premature for those environments for the armv8.X-a >>> naming... >>> and I've not seen it discussed on any other lists either.) >> yeah, it will be good to not go through the pain of community doing >> something else later >> like *-*-gnueabihf triplet was invented and to date we keep fixing >> packages for OE >> to account for gnueabihf can be gnueabi + -mfloat-abi=hard but >> packages just assume > > I think I now understand the issue. Any interest in having additional > tune files warehoused somewhere so folks don't have to reinvent the > wheel? Something like "meta-extra-tunes" ? > > I prefer fostering contributions for additional tune files that get > vetted by the community rather than the multitude of unseen > implementations I have found sitting around. I wouldn't be against it. The original thought was that new tunes (and related configurations) would be housed in the BSP until submitted to the YP. But I'm not sure in hindsight that is very robust.. Having a general meta-* layer that can just collect arch settings would be useful to help coordinate people, but keep out in-progress stuff from the base. The only issue is someone will have to further coordinate acceptance into the extra layer, and help push them into the oe-core proper when the tune quality is at a correct level --or-- the tune becomes prevalent in the world. (I just don't want it to become a dumping ground and it never ends up in the core... Similarly moving older tunes out of the core might some day be useful as well.) Key of course is this should include tunes, as well as things like siteinfo.bbclass and insane.bbclass or other extensions. (via: PACKAGEQA_EXTRA_MACHDEFFUNCS and similar.) While the toolchains, arch recipe patches, etc should be elsewhere... --Mark > regards, > armin > >> >> >>> --Mark > -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
On 08/01/2017 08:46 AM, Khem Raj wrote: On Tue, Aug 1, 2017 at 8:24 AM, Mark Hatlewrote: On 8/1/17 10:20 AM, Khem Raj wrote: On Tue, Aug 1, 2017 at 8:17 AM, akuster808 wrote: On 07/31/2017 12:04 PM, Khem Raj wrote: On 7/31/17 10:51 AM, Mark Hatle wrote: On 7/31/17 12:40 PM, akuster808 wrote: On 07/31/2017 10:31 AM, Mark Hatle wrote: On 7/31/17 12:16 PM, Armin Kuster wrote: Signed-off-by: Armin Kuster --- meta/conf/machine/include/arm/arch-armv8.inc | 25 + 1 file changed, 25 insertions(+) diff --git a/meta/conf/machine/include/arm/arch-armv8.inc b/meta/conf/machine/include/arm/arch-armv8.inc index 5e832fa..dc1ba5e 100644 --- a/meta/conf/machine/include/arm/arch-armv8.inc +++ b/meta/conf/machine/include/arm/arch-armv8.inc @@ -1 +1,26 @@ +DEFAULTTUNE ?= "armv8-a" do we want the '-a'? The other arm (7) are of the format armv7a (no '-'). works for me either way. While we are at it. How would we want ‘armv8.1-a’, ‘armv8.2-a’, ‘armv8.3-a' formated as? My preference is to drop the '-'. As for the '.', I'm not sure.. not something we've run across before. We could just drop it (the '.'), but it really depends on if armv81a would confuse someone (familiar with arm) or not. I would suggest to also sync with other distros and ensure that we dont do something different. Its very costly later. Since applications get ported to most common combination Sync what part? naming of file? ( sorry lost regarding your comment) naming convention for arch e.g. I'd say Debian/Ubuntu -- Fedora/RH would be the two primary sources I'd look at. Between debian package names and RPM package names, it should give a clue if there is any community standard forming for those names. (May still be too premature for those environments for the armv8.X-a naming... and I've not seen it discussed on any other lists either.) yeah, it will be good to not go through the pain of community doing something else later like *-*-gnueabihf triplet was invented and to date we keep fixing packages for OE to account for gnueabihf can be gnueabi + -mfloat-abi=hard but packages just assume I think I now understand the issue. Any interest in having additional tune files warehoused somewhere so folks don't have to reinvent the wheel? Something like "meta-extra-tunes" ? I prefer fostering contributions for additional tune files that get vetted by the community rather than the multitude of unseen implementations I have found sitting around. regards, armin --Mark -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
On Tue, Aug 1, 2017 at 8:24 AM, Mark Hatlewrote: > On 8/1/17 10:20 AM, Khem Raj wrote: >> On Tue, Aug 1, 2017 at 8:17 AM, akuster808 wrote: >>> >>> >>> On 07/31/2017 12:04 PM, Khem Raj wrote: On 7/31/17 10:51 AM, Mark Hatle wrote: > > On 7/31/17 12:40 PM, akuster808 wrote: >> >> >> On 07/31/2017 10:31 AM, Mark Hatle wrote: >>> >>> On 7/31/17 12:16 PM, Armin Kuster wrote: Signed-off-by: Armin Kuster --- meta/conf/machine/include/arm/arch-armv8.inc | 25 + 1 file changed, 25 insertions(+) diff --git a/meta/conf/machine/include/arm/arch-armv8.inc b/meta/conf/machine/include/arm/arch-armv8.inc index 5e832fa..dc1ba5e 100644 --- a/meta/conf/machine/include/arm/arch-armv8.inc +++ b/meta/conf/machine/include/arm/arch-armv8.inc @@ -1 +1,26 @@ +DEFAULTTUNE ?= "armv8-a" >>> >>> do we want the '-a'? The other arm (7) are of the format armv7a (no >>> '-'). >> >> works for me either way. >> >> While we are at it. How would we want ‘armv8.1-a’, ‘armv8.2-a’, >> ‘armv8.3-a' >> formated as? > > My preference is to drop the '-'. As for the '.', I'm not sure.. not > something > we've run across before. > > We could just drop it (the '.'), but it really depends on if armv81a > would > confuse someone (familiar with arm) or not. I would suggest to also sync with other distros and ensure that we dont do something different. Its very costly later. Since applications get ported to most common combination >>> >>> Sync what part? naming of file? ( sorry lost regarding your comment) >> >> naming convention for arch e.g. >> > > I'd say Debian/Ubuntu -- Fedora/RH would be the two primary sources I'd look > at. > Between debian package names and RPM package names, it should give a clue if > there is any community standard forming for those names. > > (May still be too premature for those environments for the armv8.X-a naming... > and I've not seen it discussed on any other lists either.) yeah, it will be good to not go through the pain of community doing something else later like *-*-gnueabihf triplet was invented and to date we keep fixing packages for OE to account for gnueabihf can be gnueabi + -mfloat-abi=hard but packages just assume > > --Mark -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
On 8/1/17 10:20 AM, Khem Raj wrote: > On Tue, Aug 1, 2017 at 8:17 AM, akuster808wrote: >> >> >> On 07/31/2017 12:04 PM, Khem Raj wrote: >>> >>> >>> On 7/31/17 10:51 AM, Mark Hatle wrote: On 7/31/17 12:40 PM, akuster808 wrote: > > > On 07/31/2017 10:31 AM, Mark Hatle wrote: >> >> On 7/31/17 12:16 PM, Armin Kuster wrote: >>> >>> Signed-off-by: Armin Kuster >>> --- >>> meta/conf/machine/include/arm/arch-armv8.inc | 25 >>> + >>> 1 file changed, 25 insertions(+) >>> >>> diff --git a/meta/conf/machine/include/arm/arch-armv8.inc >>> b/meta/conf/machine/include/arm/arch-armv8.inc >>> index 5e832fa..dc1ba5e 100644 >>> --- a/meta/conf/machine/include/arm/arch-armv8.inc >>> +++ b/meta/conf/machine/include/arm/arch-armv8.inc >>> @@ -1 +1,26 @@ >>> +DEFAULTTUNE ?= "armv8-a" >> >> do we want the '-a'? The other arm (7) are of the format armv7a (no >> '-'). > > works for me either way. > > While we are at it. How would we want ‘armv8.1-a’, ‘armv8.2-a’, > ‘armv8.3-a' > formated as? My preference is to drop the '-'. As for the '.', I'm not sure.. not something we've run across before. We could just drop it (the '.'), but it really depends on if armv81a would confuse someone (familiar with arm) or not. >>> >>> I would suggest to also sync with other distros and ensure that we dont >>> do something different. Its very costly later. Since applications get >>> ported to most common combination >> >> Sync what part? naming of file? ( sorry lost regarding your comment) > > naming convention for arch e.g. > I'd say Debian/Ubuntu -- Fedora/RH would be the two primary sources I'd look at. Between debian package names and RPM package names, it should give a clue if there is any community standard forming for those names. (May still be too premature for those environments for the armv8.X-a naming... and I've not seen it discussed on any other lists either.) --Mark -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
On Tue, Aug 1, 2017 at 8:17 AM, akuster808wrote: > > > On 07/31/2017 12:04 PM, Khem Raj wrote: >> >> >> On 7/31/17 10:51 AM, Mark Hatle wrote: >>> >>> On 7/31/17 12:40 PM, akuster808 wrote: On 07/31/2017 10:31 AM, Mark Hatle wrote: > > On 7/31/17 12:16 PM, Armin Kuster wrote: >> >> Signed-off-by: Armin Kuster >> --- >> meta/conf/machine/include/arm/arch-armv8.inc | 25 >> + >> 1 file changed, 25 insertions(+) >> >> diff --git a/meta/conf/machine/include/arm/arch-armv8.inc >> b/meta/conf/machine/include/arm/arch-armv8.inc >> index 5e832fa..dc1ba5e 100644 >> --- a/meta/conf/machine/include/arm/arch-armv8.inc >> +++ b/meta/conf/machine/include/arm/arch-armv8.inc >> @@ -1 +1,26 @@ >> +DEFAULTTUNE ?= "armv8-a" > > do we want the '-a'? The other arm (7) are of the format armv7a (no > '-'). works for me either way. While we are at it. How would we want ‘armv8.1-a’, ‘armv8.2-a’, ‘armv8.3-a' formated as? >>> >>> My preference is to drop the '-'. As for the '.', I'm not sure.. not >>> something >>> we've run across before. >>> >>> We could just drop it (the '.'), but it really depends on if armv81a >>> would >>> confuse someone (familiar with arm) or not. >> >> I would suggest to also sync with other distros and ensure that we dont >> do something different. Its very costly later. Since applications get >> ported to most common combination > > Sync what part? naming of file? ( sorry lost regarding your comment) naming convention for arch e.g. -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
On 07/31/2017 12:04 PM, Khem Raj wrote: On 7/31/17 10:51 AM, Mark Hatle wrote: On 7/31/17 12:40 PM, akuster808 wrote: On 07/31/2017 10:31 AM, Mark Hatle wrote: On 7/31/17 12:16 PM, Armin Kuster wrote: Signed-off-by: Armin Kuster--- meta/conf/machine/include/arm/arch-armv8.inc | 25 + 1 file changed, 25 insertions(+) diff --git a/meta/conf/machine/include/arm/arch-armv8.inc b/meta/conf/machine/include/arm/arch-armv8.inc index 5e832fa..dc1ba5e 100644 --- a/meta/conf/machine/include/arm/arch-armv8.inc +++ b/meta/conf/machine/include/arm/arch-armv8.inc @@ -1 +1,26 @@ +DEFAULTTUNE ?= "armv8-a" do we want the '-a'? The other arm (7) are of the format armv7a (no '-'). works for me either way. While we are at it. How would we want ‘armv8.1-a’, ‘armv8.2-a’, ‘armv8.3-a' formated as? My preference is to drop the '-'. As for the '.', I'm not sure.. not something we've run across before. We could just drop it (the '.'), but it really depends on if armv81a would confuse someone (familiar with arm) or not. I would suggest to also sync with other distros and ensure that we dont do something different. Its very costly later. Since applications get ported to most common combination Sync what part? naming of file? ( sorry lost regarding your comment) - armin --Mark - armin --Mark + +TUNEVALID[armv8-a] = "Enable instructions for ARMv8-a" +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', ' -march=armv8-a', '', d)}" + +TUNEVALID[crypto] = "Enable crypto instructions for ARMv8-a" +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'crypto', '+crypto', '', d)}" + +MACHINEOVERRIDES =. "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', 'armv8a:', '' ,d)}" + require conf/machine/include/arm/arch-arm64.inc + +AVAILTUNES += "armv8-a armv8-a-crypto" +ARMPKGARCH_tune-armv8-a ?= "armv8-a" + +# Enable Crypto extension. This also enables Advanced SIMD and floating-point instructions +ARMPKGARCH_tune-armv8-a-crypto ?= "armv8-a" + +TUNE_FEATURES_tune-armv8-a = "aarch64 armv8-a" +TUNE_FEATURES_tune-armv8-a-crypto = "${TUNE_FEATURES_tune-armv8-a} crypto" + +PACKAGE_EXTRA_ARCHS_tune-armv8-a = "aarch64 armv8-a" +PACKAGE_EXTRA_ARCHS_tune-armv8-a-crypto = "${PACKAGE_EXTRA_ARCHS_tune-armv8-a} armv8-a-crypto" + +BASE_LIB_tune-armv8-a = "${BASE_LIB_tune-aarch64}" +BASE_LIB_tune-armv8-a-crypto = "${BASE_LIB_tune-aarch64}" -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
On 7/31/17 10:51 AM, Mark Hatle wrote: > On 7/31/17 12:40 PM, akuster808 wrote: >> >> >> On 07/31/2017 10:31 AM, Mark Hatle wrote: >>> On 7/31/17 12:16 PM, Armin Kuster wrote: Signed-off-by: Armin Kuster--- meta/conf/machine/include/arm/arch-armv8.inc | 25 + 1 file changed, 25 insertions(+) diff --git a/meta/conf/machine/include/arm/arch-armv8.inc b/meta/conf/machine/include/arm/arch-armv8.inc index 5e832fa..dc1ba5e 100644 --- a/meta/conf/machine/include/arm/arch-armv8.inc +++ b/meta/conf/machine/include/arm/arch-armv8.inc @@ -1 +1,26 @@ +DEFAULTTUNE ?= "armv8-a" >>> do we want the '-a'? The other arm (7) are of the format armv7a (no '-'). >> works for me either way. >> >> While we are at it. How would we want ‘armv8.1-a’, ‘armv8.2-a’, ‘armv8.3-a' >> formated as? > > My preference is to drop the '-'. As for the '.', I'm not sure.. not > something > we've run across before. > > We could just drop it (the '.'), but it really depends on if armv81a would > confuse someone (familiar with arm) or not. I would suggest to also sync with other distros and ensure that we dont do something different. Its very costly later. Since applications get ported to most common combination > > --Mark > >> - armin >>> >>> --Mark >>> + +TUNEVALID[armv8-a] = "Enable instructions for ARMv8-a" +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', ' -march=armv8-a', '', d)}" + +TUNEVALID[crypto] = "Enable crypto instructions for ARMv8-a" +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'crypto', '+crypto', '', d)}" + +MACHINEOVERRIDES =. "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', 'armv8a:', '' ,d)}" + require conf/machine/include/arm/arch-arm64.inc + +AVAILTUNES += "armv8-a armv8-a-crypto" +ARMPKGARCH_tune-armv8-a ?= "armv8-a" + +# Enable Crypto extension. This also enables Advanced SIMD and floating-point instructions +ARMPKGARCH_tune-armv8-a-crypto ?= "armv8-a" + +TUNE_FEATURES_tune-armv8-a = "aarch64 armv8-a" +TUNE_FEATURES_tune-armv8-a-crypto = "${TUNE_FEATURES_tune-armv8-a} crypto" + +PACKAGE_EXTRA_ARCHS_tune-armv8-a = "aarch64 armv8-a" +PACKAGE_EXTRA_ARCHS_tune-armv8-a-crypto = "${PACKAGE_EXTRA_ARCHS_tune-armv8-a} armv8-a-crypto" + +BASE_LIB_tune-armv8-a = "${BASE_LIB_tune-aarch64}" +BASE_LIB_tune-armv8-a-crypto = "${BASE_LIB_tune-aarch64}" >> > signature.asc Description: OpenPGP digital signature -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
On 7/31/17 12:40 PM, akuster808 wrote: > > > On 07/31/2017 10:31 AM, Mark Hatle wrote: >> On 7/31/17 12:16 PM, Armin Kuster wrote: >>> Signed-off-by: Armin Kuster>>> --- >>> meta/conf/machine/include/arm/arch-armv8.inc | 25 + >>> 1 file changed, 25 insertions(+) >>> >>> diff --git a/meta/conf/machine/include/arm/arch-armv8.inc >>> b/meta/conf/machine/include/arm/arch-armv8.inc >>> index 5e832fa..dc1ba5e 100644 >>> --- a/meta/conf/machine/include/arm/arch-armv8.inc >>> +++ b/meta/conf/machine/include/arm/arch-armv8.inc >>> @@ -1 +1,26 @@ >>> +DEFAULTTUNE ?= "armv8-a" >> do we want the '-a'? The other arm (7) are of the format armv7a (no '-'). > works for me either way. > > While we are at it. How would we want ‘armv8.1-a’, ‘armv8.2-a’, ‘armv8.3-a' > formated as? My preference is to drop the '-'. As for the '.', I'm not sure.. not something we've run across before. We could just drop it (the '.'), but it really depends on if armv81a would confuse someone (familiar with arm) or not. --Mark > - armin >> >> --Mark >> >>> + >>> +TUNEVALID[armv8-a] = "Enable instructions for ARMv8-a" >>> +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', ' >>> -march=armv8-a', '', d)}" >>> + >>> +TUNEVALID[crypto] = "Enable crypto instructions for ARMv8-a" >>> +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'crypto', '+crypto', >>> '', d)}" >>> + >>> +MACHINEOVERRIDES =. "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', >>> 'armv8a:', '' ,d)}" >>> + >>> require conf/machine/include/arm/arch-arm64.inc >>> + >>> +AVAILTUNES += "armv8-a armv8-a-crypto" >>> +ARMPKGARCH_tune-armv8-a ?= "armv8-a" >>> + >>> +# Enable Crypto extension. This also enables Advanced SIMD and >>> floating-point instructions >>> +ARMPKGARCH_tune-armv8-a-crypto ?= "armv8-a" >>> + >>> +TUNE_FEATURES_tune-armv8-a = "aarch64 armv8-a" >>> +TUNE_FEATURES_tune-armv8-a-crypto = "${TUNE_FEATURES_tune-armv8-a} crypto" >>> + >>> +PACKAGE_EXTRA_ARCHS_tune-armv8-a = "aarch64 armv8-a" >>> +PACKAGE_EXTRA_ARCHS_tune-armv8-a-crypto = >>> "${PACKAGE_EXTRA_ARCHS_tune-armv8-a} armv8-a-crypto" >>> + >>> +BASE_LIB_tune-armv8-a = "${BASE_LIB_tune-aarch64}" >>> +BASE_LIB_tune-armv8-a-crypto = "${BASE_LIB_tune-aarch64}" >>> > -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
On 07/31/2017 10:31 AM, Mark Hatle wrote: On 7/31/17 12:16 PM, Armin Kuster wrote: Signed-off-by: Armin Kuster--- meta/conf/machine/include/arm/arch-armv8.inc | 25 + 1 file changed, 25 insertions(+) diff --git a/meta/conf/machine/include/arm/arch-armv8.inc b/meta/conf/machine/include/arm/arch-armv8.inc index 5e832fa..dc1ba5e 100644 --- a/meta/conf/machine/include/arm/arch-armv8.inc +++ b/meta/conf/machine/include/arm/arch-armv8.inc @@ -1 +1,26 @@ +DEFAULTTUNE ?= "armv8-a" do we want the '-a'? The other arm (7) are of the format armv7a (no '-'). works for me either way. While we are at it. How would we want ‘armv8.1-a’, ‘armv8.2-a’, ‘armv8.3-a' formated as? - armin --Mark + +TUNEVALID[armv8-a] = "Enable instructions for ARMv8-a" +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', ' -march=armv8-a', '', d)}" + +TUNEVALID[crypto] = "Enable crypto instructions for ARMv8-a" +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'crypto', '+crypto', '', d)}" + +MACHINEOVERRIDES =. "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', 'armv8a:', '' ,d)}" + require conf/machine/include/arm/arch-arm64.inc + +AVAILTUNES += "armv8-a armv8-a-crypto" +ARMPKGARCH_tune-armv8-a ?= "armv8-a" + +# Enable Crypto extension. This also enables Advanced SIMD and floating-point instructions +ARMPKGARCH_tune-armv8-a-crypto ?= "armv8-a" + +TUNE_FEATURES_tune-armv8-a = "aarch64 armv8-a" +TUNE_FEATURES_tune-armv8-a-crypto = "${TUNE_FEATURES_tune-armv8-a} crypto" + +PACKAGE_EXTRA_ARCHS_tune-armv8-a = "aarch64 armv8-a" +PACKAGE_EXTRA_ARCHS_tune-armv8-a-crypto = "${PACKAGE_EXTRA_ARCHS_tune-armv8-a} armv8-a-crypto" + +BASE_LIB_tune-armv8-a = "${BASE_LIB_tune-aarch64}" +BASE_LIB_tune-armv8-a-crypto = "${BASE_LIB_tune-aarch64}" -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
Re: [OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
On 7/31/17 12:16 PM, Armin Kuster wrote: > Signed-off-by: Armin Kuster> --- > meta/conf/machine/include/arm/arch-armv8.inc | 25 + > 1 file changed, 25 insertions(+) > > diff --git a/meta/conf/machine/include/arm/arch-armv8.inc > b/meta/conf/machine/include/arm/arch-armv8.inc > index 5e832fa..dc1ba5e 100644 > --- a/meta/conf/machine/include/arm/arch-armv8.inc > +++ b/meta/conf/machine/include/arm/arch-armv8.inc > @@ -1 +1,26 @@ > +DEFAULTTUNE ?= "armv8-a" do we want the '-a'? The other arm (7) are of the format armv7a (no '-'). --Mark > + > +TUNEVALID[armv8-a] = "Enable instructions for ARMv8-a" > +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', ' > -march=armv8-a', '', d)}" > + > +TUNEVALID[crypto] = "Enable crypto instructions for ARMv8-a" > +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'crypto', '+crypto', > '', d)}" > + > +MACHINEOVERRIDES =. "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', > 'armv8a:', '' ,d)}" > + > require conf/machine/include/arm/arch-arm64.inc > + > +AVAILTUNES += "armv8-a armv8-a-crypto" > +ARMPKGARCH_tune-armv8-a ?= "armv8-a" > + > +# Enable Crypto extension. This also enables Advanced SIMD and > floating-point instructions > +ARMPKGARCH_tune-armv8-a-crypto ?= "armv8-a" > + > +TUNE_FEATURES_tune-armv8-a = "aarch64 armv8-a" > +TUNE_FEATURES_tune-armv8-a-crypto = "${TUNE_FEATURES_tune-armv8-a} crypto" > + > +PACKAGE_EXTRA_ARCHS_tune-armv8-a = "aarch64 armv8-a" > +PACKAGE_EXTRA_ARCHS_tune-armv8-a-crypto = > "${PACKAGE_EXTRA_ARCHS_tune-armv8-a} armv8-a-crypto" > + > +BASE_LIB_tune-armv8-a = "${BASE_LIB_tune-aarch64}" > +BASE_LIB_tune-armv8-a-crypto = "${BASE_LIB_tune-aarch64}" > -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core
[OE-core] [PATCH 1/2] armv8: update to use armv8-a tune
Signed-off-by: Armin Kuster--- meta/conf/machine/include/arm/arch-armv8.inc | 25 + 1 file changed, 25 insertions(+) diff --git a/meta/conf/machine/include/arm/arch-armv8.inc b/meta/conf/machine/include/arm/arch-armv8.inc index 5e832fa..dc1ba5e 100644 --- a/meta/conf/machine/include/arm/arch-armv8.inc +++ b/meta/conf/machine/include/arm/arch-armv8.inc @@ -1 +1,26 @@ +DEFAULTTUNE ?= "armv8-a" + +TUNEVALID[armv8-a] = "Enable instructions for ARMv8-a" +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', ' -march=armv8-a', '', d)}" + +TUNEVALID[crypto] = "Enable crypto instructions for ARMv8-a" +TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'crypto', '+crypto', '', d)}" + +MACHINEOVERRIDES =. "${@bb.utils.contains('TUNE_FEATURES', 'armv8-a', 'armv8a:', '' ,d)}" + require conf/machine/include/arm/arch-arm64.inc + +AVAILTUNES += "armv8-a armv8-a-crypto" +ARMPKGARCH_tune-armv8-a ?= "armv8-a" + +# Enable Crypto extension. This also enables Advanced SIMD and floating-point instructions +ARMPKGARCH_tune-armv8-a-crypto ?= "armv8-a" + +TUNE_FEATURES_tune-armv8-a = "aarch64 armv8-a" +TUNE_FEATURES_tune-armv8-a-crypto = "${TUNE_FEATURES_tune-armv8-a} crypto" + +PACKAGE_EXTRA_ARCHS_tune-armv8-a = "aarch64 armv8-a" +PACKAGE_EXTRA_ARCHS_tune-armv8-a-crypto = "${PACKAGE_EXTRA_ARCHS_tune-armv8-a} armv8-a-crypto" + +BASE_LIB_tune-armv8-a = "${BASE_LIB_tune-aarch64}" +BASE_LIB_tune-armv8-a-crypto = "${BASE_LIB_tune-aarch64}" -- 2.5.0 -- ___ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core