[1/2] config: add Graviton2(arm64) meson configuration
diff mbox series

Message ID 20200909011109.2048-2-vcchunga@amazon.com
State Superseded, archived
Delegated to: Thomas Monjalon
Headers show
Series
  • build: add Graviton2(arm64) config
Related show

Checks

Context Check Description
ci/checkpatch success coding style OK

Commit Message

Vimal Chungath Sept. 9, 2020, 1:11 a.m. UTC
Add meson build configuration for Graviton2 platform
with 64-bit ARM Neoverse N1 cores. This patch makes the
following changes to generic Neoverse N1 config:

1. increase lcore limit to 64
2. increase memory support to 1TB
3. remove +crc from -march as that is default when setting armv8.2

For more information about Graviton2 platform, refer to:
https://aws.amazon.com/ec2/graviton/

Signed-off-by: Vimal Chungath <vcchunga@amazon.com>
---
 config/arm/arm64_graviton2_linux_gcc | 17 +++++++++++++++++
 config/arm/meson.build               | 12 +++++++++++-
 2 files changed, 28 insertions(+), 1 deletion(-)
 create mode 100644 config/arm/arm64_graviton2_linux_gcc

Comments

Honnappa Nagarahalli Sept. 12, 2020, 3:23 a.m. UTC | #1
+Jerin, Hemant, Dharmik

<snip>
Hi Vimal,
	Few comments inline.

> 
> Add meson build configuration for Graviton2 platform with 64-bit ARM
> Neoverse N1 cores. This patch makes the following changes to generic
> Neoverse N1 config:
> 
> 1. increase lcore limit to 64
> 2. increase memory support to 1TB
There will be multiple SoCs with N1 cores. All of them will have the same implementor ID and part number. But, they will have different values for these configurable parameters.
IMO, from usage perspective, we have 2 cases:
1) Ability to build a portable binary that can run on multiple Arm SoCs (for ex: BlueField, thunderx1, thunderx2, N1SDP, Graviton2 etc)
2) Ability to build a binary which would run only on a SoC it was compiled for and provide the most optimized binary for that SoC. But, this may not be portable.

For 1) we have default march.

For 2) we do not have the capability today in meson build (at least, this is my understanding, please correct me if I am wrong). In this case, the user knows the target platform for compilation. IMO, we should add the capability to take the target platform as an input from the user (similar to the make build system) and Graviton2 can be one such target platform.

> 3. remove +crc from -march as that is default when setting armv8.2
> 
> For more information about Graviton2 platform, refer to:
> https://aws.amazon.com/ec2/graviton/
> 
> Signed-off-by: Vimal Chungath <vcchunga@amazon.com>
> ---
>  config/arm/arm64_graviton2_linux_gcc | 17 +++++++++++++++++
>  config/arm/meson.build               | 12 +++++++++++-
>  2 files changed, 28 insertions(+), 1 deletion(-)  create mode 100644
> config/arm/arm64_graviton2_linux_gcc
> 
> diff --git a/config/arm/arm64_graviton2_linux_gcc
> b/config/arm/arm64_graviton2_linux_gcc
> new file mode 100644
> index 000000000..022e06303
> --- /dev/null
> +++ b/config/arm/arm64_graviton2_linux_gcc
> @@ -0,0 +1,17 @@
> +[binaries]
> +c = 'aarch64-linux-gnu-gcc'
> +cpp = 'aarch64-linux-gnu-cpp'
> +ar = 'aarch64-linux-gnu-gcc-ar'
> +strip = 'aarch64-linux-gnu-strip'
> +pkgconfig = 'aarch64-linux-gnu-pkg-config'
> +pcap-config = ''
> +
> +[host_machine]
> +system = 'linux'
> +cpu_family = 'aarch64'
> +cpu = 'armv8-a'
> +endian = 'little'
> +
> +[properties]
> +implementor_id = '0x41'
> +implementor_pn = '0xd0c'
> diff --git a/config/arm/meson.build b/config/arm/meson.build index
> 8728051d5..64e277ebc 100644
> --- a/config/arm/meson.build
> +++ b/config/arm/meson.build
> @@ -86,6 +86,16 @@ flags_octeontx2_extra = [
>  	['RTE_ARM_FEATURE_ATOMICS', true],
>  	['RTE_EAL_IGB_UIO', false],
>  	['RTE_USE_C11_MEM_MODEL', true]]
> +flags_n1generic_extra = [
> +	['RTE_MACHINE', '"neoverse-n1"'],
> +	['RTE_MAX_LCORE', 64],
> +	['RTE_CACHE_LINE_SIZE', 64],
> +	['RTE_ARM_FEATURE_ATOMICS', true],
> +	['RTE_USE_C11_MEM_MODEL', true],
> +	['RTE_MAX_MEM_MB', 1048576],
> +	['RTE_MAX_NUMA_NODES', 1],
> +	['RTE_EAL_NUMA_AWARE_HUGEPAGES', false],
> +	['RTE_LIBRTE_VHOST_NUMA', false]]
> 
>  machine_args_generic = [
>  	['default', ['-march=armv8-a+crc']],
> @@ -97,7 +107,7 @@ machine_args_generic = [
>  	['0xd09', ['-mcpu=cortex-a73']],
>  	['0xd0a', ['-mcpu=cortex-a75']],
>  	['0xd0b', ['-mcpu=cortex-a76']],
> -	['0xd0c', ['-march=armv8.2-a+crc+crypto', '-mcpu=neoverse-n1'],
> flags_n1sdp_extra]]
> +	['0xd0c', ['-march=armv8.2-a+crypto', '-mcpu=neoverse-n1'],
> +flags_n1generic_extra]]
> 
>  machine_args_cavium = [
>  	['default', ['-march=armv8-a+crc+crypto','-mcpu=thunderx']],
> --
> 2.16.6
Vimal Chungath Sept. 12, 2020, 6:24 p.m. UTC | #2
On 9/11/20 8:23 PM, Honnappa Nagarahalli wrote:
>
> +Jerin, Hemant, Dharmik
>
> <snip>
> Hi Vimal,
>         Few comments inline.
>
>>
>> Add meson build configuration for Graviton2 platform with 64-bit ARM
>> Neoverse N1 cores. This patch makes the following changes to generic
>> Neoverse N1 config:
>>
>> 1. increase lcore limit to 64
>> 2. increase memory support to 1TB
> There will be multiple SoCs with N1 cores. All of them will have the same implementor ID and part number. But, they will have different values for these configurable parameters.
> IMO, from usage perspective, we have 2 cases:
> 1) Ability to build a portable binary that can run on multiple Arm SoCs (for ex: BlueField, thunderx1, thunderx2, N1SDP, Graviton2 etc)
> 2) Ability to build a binary which would run only on a SoC it was compiled for and provide the most optimized binary for that SoC. But, this may not be portable.
>
> For 1) we have default march.
>
> For 2) we do not have the capability today in meson build (at least, this is my understanding, please correct me if I am wrong). In this case, the user knows the target platform for compilation. IMO, we should add the capability to take the target platform as an input from the user (similar to the make build system) and Graviton2 can be one such target platform.

My intention was to have parameters that work for both N1SDP and Graviton2
rather than 2). Does the change to RTE_MAX_LCORE and RTE_MAX_MEM_MB make them
incompatible with N1SDP?

I'm not sure if taking target platform from user is the best option here.
Would this be specific to N1 since other platforms like thunderx 
differentiate the flags with part number?

>
>> 3. remove +crc from -march as that is default when setting armv8.2
>>
>> For more information about Graviton2 platform, refer to:
>> https://aws.amazon.com/ec2/graviton/
>>
>> Signed-off-by: Vimal Chungath <vcchunga@amazon.com>
>> ---
>>  config/arm/arm64_graviton2_linux_gcc | 17 +++++++++++++++++
>>  config/arm/meson.build               | 12 +++++++++++-
>>  2 files changed, 28 insertions(+), 1 deletion(-)  create mode 100644
>> config/arm/arm64_graviton2_linux_gcc
>>
>> diff --git a/config/arm/arm64_graviton2_linux_gcc
>> b/config/arm/arm64_graviton2_linux_gcc
>> new file mode 100644
>> index 000000000..022e06303
>> --- /dev/null
>> +++ b/config/arm/arm64_graviton2_linux_gcc
>> @@ -0,0 +1,17 @@
>> +[binaries]
>> +c = 'aarch64-linux-gnu-gcc'
>> +cpp = 'aarch64-linux-gnu-cpp'
>> +ar = 'aarch64-linux-gnu-gcc-ar'
>> +strip = 'aarch64-linux-gnu-strip'
>> +pkgconfig = 'aarch64-linux-gnu-pkg-config'
>> +pcap-config = ''
>> +
>> +[host_machine]
>> +system = 'linux'
>> +cpu_family = 'aarch64'
>> +cpu = 'armv8-a'
>> +endian = 'little'
>> +
>> +[properties]
>> +implementor_id = '0x41'
>> +implementor_pn = '0xd0c'
>> diff --git a/config/arm/meson.build b/config/arm/meson.build index
>> 8728051d5..64e277ebc 100644
>> --- a/config/arm/meson.build
>> +++ b/config/arm/meson.build
>> @@ -86,6 +86,16 @@ flags_octeontx2_extra = [
>>       ['RTE_ARM_FEATURE_ATOMICS', true],
>>       ['RTE_EAL_IGB_UIO', false],
>>       ['RTE_USE_C11_MEM_MODEL', true]]
>> +flags_n1generic_extra = [
>> +     ['RTE_MACHINE', '"neoverse-n1"'],
>> +     ['RTE_MAX_LCORE', 64],
>> +     ['RTE_CACHE_LINE_SIZE', 64],
>> +     ['RTE_ARM_FEATURE_ATOMICS', true],
>> +     ['RTE_USE_C11_MEM_MODEL', true],
>> +     ['RTE_MAX_MEM_MB', 1048576],
>> +     ['RTE_MAX_NUMA_NODES', 1],
>> +     ['RTE_EAL_NUMA_AWARE_HUGEPAGES', false],
>> +     ['RTE_LIBRTE_VHOST_NUMA', false]]
>>
>>  machine_args_generic = [
>>       ['default', ['-march=armv8-a+crc']],
>> @@ -97,7 +107,7 @@ machine_args_generic = [
>>       ['0xd09', ['-mcpu=cortex-a73']],
>>       ['0xd0a', ['-mcpu=cortex-a75']],
>>       ['0xd0b', ['-mcpu=cortex-a76']],
>> -     ['0xd0c', ['-march=armv8.2-a+crc+crypto', '-mcpu=neoverse-n1'],
>> flags_n1sdp_extra]]
>> +     ['0xd0c', ['-march=armv8.2-a+crypto', '-mcpu=neoverse-n1'],
>> +flags_n1generic_extra]]
>>
>>  machine_args_cavium = [
>>       ['default', ['-march=armv8-a+crc+crypto','-mcpu=thunderx']],
>> --
>> 2.16.6
>
Honnappa Nagarahalli Sept. 14, 2020, 4:49 a.m. UTC | #3
<snip>

> 
> On 9/11/20 8:23 PM, Honnappa Nagarahalli wrote:
> >
> > +Jerin, Hemant, Dharmik
> >
> > <snip>
> > Hi Vimal,
> >         Few comments inline.
> >
> >>
> >> Add meson build configuration for Graviton2 platform with 64-bit ARM
> >> Neoverse N1 cores. This patch makes the following changes to generic
> >> Neoverse N1 config:
> >>
> >> 1. increase lcore limit to 64
> >> 2. increase memory support to 1TB
> > There will be multiple SoCs with N1 cores. All of them will have the same
> implementor ID and part number. But, they will have different values for
> these configurable parameters.
> > IMO, from usage perspective, we have 2 cases:
> > 1) Ability to build a portable binary that can run on multiple Arm
> > SoCs (for ex: BlueField, thunderx1, thunderx2, N1SDP, Graviton2 etc)
> > 2) Ability to build a binary which would run only on a SoC it was compiled
> for and provide the most optimized binary for that SoC. But, this may not be
> portable.
> >
> > For 1) we have default march.
> >
> > For 2) we do not have the capability today in meson build (at least, this is
> my understanding, please correct me if I am wrong). In this case, the user
> knows the target platform for compilation. IMO, we should add the capability
> to take the target platform as an input from the user (similar to the make
> build system) and Graviton2 can be one such target platform.
> 
> My intention was to have parameters that work for both N1SDP and
> Graviton2 rather than 2). Does the change to RTE_MAX_LCORE and
> RTE_MAX_MEM_MB make them incompatible with N1SDP?
They are not optimal for N1SDP. In the future these parameters might have to be changed. For ex: if there a N1 based SoC with more than 64 CPU cores.

> 
> I'm not sure if taking target platform from user is the best option here.
> Would this be specific to N1 since other platforms like thunderx differentiate
> the flags with part number?
This issue is specific to Arm CPU cores in general. So, it applies to N1 too.

> 
> >
> >> 3. remove +crc from -march as that is default when setting armv8.2
> >>
> >> For more information about Graviton2 platform, refer to:
> >> https://aws.amazon.com/ec2/graviton/
> >>
> >> Signed-off-by: Vimal Chungath <vcchunga@amazon.com>
> >> ---
> >>  config/arm/arm64_graviton2_linux_gcc | 17 +++++++++++++++++
> >>  config/arm/meson.build               | 12 +++++++++++-
> >>  2 files changed, 28 insertions(+), 1 deletion(-)  create mode 100644
> >> config/arm/arm64_graviton2_linux_gcc
> >>
> >> diff --git a/config/arm/arm64_graviton2_linux_gcc
> >> b/config/arm/arm64_graviton2_linux_gcc
> >> new file mode 100644
> >> index 000000000..022e06303
> >> --- /dev/null
> >> +++ b/config/arm/arm64_graviton2_linux_gcc
> >> @@ -0,0 +1,17 @@
> >> +[binaries]
> >> +c = 'aarch64-linux-gnu-gcc'
> >> +cpp = 'aarch64-linux-gnu-cpp'
> >> +ar = 'aarch64-linux-gnu-gcc-ar'
> >> +strip = 'aarch64-linux-gnu-strip'
> >> +pkgconfig = 'aarch64-linux-gnu-pkg-config'
> >> +pcap-config = ''
> >> +
> >> +[host_machine]
> >> +system = 'linux'
> >> +cpu_family = 'aarch64'
> >> +cpu = 'armv8-a'
> >> +endian = 'little'
> >> +
> >> +[properties]
> >> +implementor_id = '0x41'
> >> +implementor_pn = '0xd0c'
> >> diff --git a/config/arm/meson.build b/config/arm/meson.build index
> >> 8728051d5..64e277ebc 100644
> >> --- a/config/arm/meson.build
> >> +++ b/config/arm/meson.build
> >> @@ -86,6 +86,16 @@ flags_octeontx2_extra = [
> >>       ['RTE_ARM_FEATURE_ATOMICS', true],
> >>       ['RTE_EAL_IGB_UIO', false],
> >>       ['RTE_USE_C11_MEM_MODEL', true]]
> >> +flags_n1generic_extra = [
> >> +     ['RTE_MACHINE', '"neoverse-n1"'],
> >> +     ['RTE_MAX_LCORE', 64],
> >> +     ['RTE_CACHE_LINE_SIZE', 64],
> >> +     ['RTE_ARM_FEATURE_ATOMICS', true],
> >> +     ['RTE_USE_C11_MEM_MODEL', true],
> >> +     ['RTE_MAX_MEM_MB', 1048576],
> >> +     ['RTE_MAX_NUMA_NODES', 1],
> >> +     ['RTE_EAL_NUMA_AWARE_HUGEPAGES', false],
> >> +     ['RTE_LIBRTE_VHOST_NUMA', false]]
> >>
> >>  machine_args_generic = [
> >>       ['default', ['-march=armv8-a+crc']], @@ -97,7 +107,7 @@
> >> machine_args_generic = [
> >>       ['0xd09', ['-mcpu=cortex-a73']],
> >>       ['0xd0a', ['-mcpu=cortex-a75']],
> >>       ['0xd0b', ['-mcpu=cortex-a76']],
> >> -     ['0xd0c', ['-march=armv8.2-a+crc+crypto', '-mcpu=neoverse-n1'],
> >> flags_n1sdp_extra]]
> >> +     ['0xd0c', ['-march=armv8.2-a+crypto', '-mcpu=neoverse-n1'],
> >> +flags_n1generic_extra]]
> >>
> >>  machine_args_cavium = [
> >>       ['default', ['-march=armv8-a+crc+crypto','-mcpu=thunderx']],
> >> --
> >> 2.16.6
> >
Jerin Jacob Sept. 17, 2020, 8:23 a.m. UTC | #4
On Mon, Sep 14, 2020 at 10:19 AM Honnappa Nagarahalli
<Honnappa.Nagarahalli@arm.com> wrote:
>
> <snip>
>
> >
> > On 9/11/20 8:23 PM, Honnappa Nagarahalli wrote:
> > >
> > > +Jerin, Hemant, Dharmik
> > >
> > > <snip>
> > > Hi Vimal,
> > >         Few comments inline.
> > >
> > >>
> > >> Add meson build configuration for Graviton2 platform with 64-bit ARM
> > >> Neoverse N1 cores. This patch makes the following changes to generic
> > >> Neoverse N1 config:
> > >>
> > >> 1. increase lcore limit to 64
> > >> 2. increase memory support to 1TB
> > > There will be multiple SoCs with N1 cores. All of them will have the same
> > implementor ID and part number. But, they will have different values for
> > these configurable parameters.
> > > IMO, from usage perspective, we have 2 cases:
> > > 1) Ability to build a portable binary that can run on multiple Arm
> > > SoCs (for ex: BlueField, thunderx1, thunderx2, N1SDP, Graviton2 etc)
> > > 2) Ability to build a binary which would run only on a SoC it was compiled
> > for and provide the most optimized binary for that SoC. But, this may not be
> > portable.
> > >
> > > For 1) we have default march.
> > >
> > > For 2) we do not have the capability today in meson build (at least, this is
> > my understanding, please correct me if I am wrong). In this case, the user
> > knows the target platform for compilation. IMO, we should add the capability
> > to take the target platform as an input from the user (similar to the make
> > build system) and Graviton2 can be one such target platform.
> >
> > My intention was to have parameters that work for both N1SDP and
> > Graviton2 rather than 2). Does the change to RTE_MAX_LCORE and
> > RTE_MAX_MEM_MB make them incompatible with N1SDP?
> They are not optimal for N1SDP. In the future these parameters might have to be changed. For ex: if there a N1 based SoC with more than 64 CPU cores.


Sorry for the late reply.

Looking at the Bluefield, Graviton2, and upcoming SoCs based on ARM IP,
It is very clear that MIDR value can not be changed by the silicon vendors.
So our existing build scheme of using the MIDR  value-based probe does
not work anymore with ARM IP.
So IMO, We need to change our build scheme. i.e

1) For native build just use -march=native
2) For cross-build, explicitly, mention the target to pick the
configuration values instead of
probing the MIDR value-based scheme.

If we agree, Any volunteers for the update to new scheme?

>
> >
> > I'm not sure if taking target platform from user is the best option here.
> > Would this be specific to N1 since other platforms like thunderx differentiate
> > the flags with part number?
> This issue is specific to Arm CPU cores in general. So, it applies to N1 too.
>
> >
> > >
> > >> 3. remove +crc from -march as that is default when setting armv8.2
> > >>
> > >> For more information about Graviton2 platform, refer to:
> > >> https://aws.amazon.com/ec2/graviton/
> > >>
> > >> Signed-off-by: Vimal Chungath <vcchunga@amazon.com>
> > >> ---
> > >>  config/arm/arm64_graviton2_linux_gcc | 17 +++++++++++++++++
> > >>  config/arm/meson.build               | 12 +++++++++++-
> > >>  2 files changed, 28 insertions(+), 1 deletion(-)  create mode 100644
> > >> config/arm/arm64_graviton2_linux_gcc
> > >>
> > >> diff --git a/config/arm/arm64_graviton2_linux_gcc
> > >> b/config/arm/arm64_graviton2_linux_gcc
> > >> new file mode 100644
> > >> index 000000000..022e06303
> > >> --- /dev/null
> > >> +++ b/config/arm/arm64_graviton2_linux_gcc
> > >> @@ -0,0 +1,17 @@
> > >> +[binaries]
> > >> +c = 'aarch64-linux-gnu-gcc'
> > >> +cpp = 'aarch64-linux-gnu-cpp'
> > >> +ar = 'aarch64-linux-gnu-gcc-ar'
> > >> +strip = 'aarch64-linux-gnu-strip'
> > >> +pkgconfig = 'aarch64-linux-gnu-pkg-config'
> > >> +pcap-config = ''
> > >> +
> > >> +[host_machine]
> > >> +system = 'linux'
> > >> +cpu_family = 'aarch64'
> > >> +cpu = 'armv8-a'
> > >> +endian = 'little'
> > >> +
> > >> +[properties]
> > >> +implementor_id = '0x41'
> > >> +implementor_pn = '0xd0c'
> > >> diff --git a/config/arm/meson.build b/config/arm/meson.build index
> > >> 8728051d5..64e277ebc 100644
> > >> --- a/config/arm/meson.build
> > >> +++ b/config/arm/meson.build
> > >> @@ -86,6 +86,16 @@ flags_octeontx2_extra = [
> > >>       ['RTE_ARM_FEATURE_ATOMICS', true],
> > >>       ['RTE_EAL_IGB_UIO', false],
> > >>       ['RTE_USE_C11_MEM_MODEL', true]]
> > >> +flags_n1generic_extra = [
> > >> +     ['RTE_MACHINE', '"neoverse-n1"'],
> > >> +     ['RTE_MAX_LCORE', 64],
> > >> +     ['RTE_CACHE_LINE_SIZE', 64],
> > >> +     ['RTE_ARM_FEATURE_ATOMICS', true],
> > >> +     ['RTE_USE_C11_MEM_MODEL', true],
> > >> +     ['RTE_MAX_MEM_MB', 1048576],
> > >> +     ['RTE_MAX_NUMA_NODES', 1],
> > >> +     ['RTE_EAL_NUMA_AWARE_HUGEPAGES', false],
> > >> +     ['RTE_LIBRTE_VHOST_NUMA', false]]
> > >>
> > >>  machine_args_generic = [
> > >>       ['default', ['-march=armv8-a+crc']], @@ -97,7 +107,7 @@
> > >> machine_args_generic = [
> > >>       ['0xd09', ['-mcpu=cortex-a73']],
> > >>       ['0xd0a', ['-mcpu=cortex-a75']],
> > >>       ['0xd0b', ['-mcpu=cortex-a76']],
> > >> -     ['0xd0c', ['-march=armv8.2-a+crc+crypto', '-mcpu=neoverse-n1'],
> > >> flags_n1sdp_extra]]
> > >> +     ['0xd0c', ['-march=armv8.2-a+crypto', '-mcpu=neoverse-n1'],
> > >> +flags_n1generic_extra]]
> > >>
> > >>  machine_args_cavium = [
> > >>       ['default', ['-march=armv8-a+crc+crypto','-mcpu=thunderx']],
> > >> --
> > >> 2.16.6
> > >
>
Honnappa Nagarahalli Sept. 17, 2020, 5:11 p.m. UTC | #5
<snip>

> > >
> > > On 9/11/20 8:23 PM, Honnappa Nagarahalli wrote:
> > > >
> > > > +Jerin, Hemant, Dharmik
> > > >
> > > > <snip>
> > > > Hi Vimal,
> > > >         Few comments inline.
> > > >
> > > >>
> > > >> Add meson build configuration for Graviton2 platform with 64-bit
> > > >> ARM Neoverse N1 cores. This patch makes the following changes to
> > > >> generic Neoverse N1 config:
> > > >>
> > > >> 1. increase lcore limit to 64
> > > >> 2. increase memory support to 1TB
> > > > There will be multiple SoCs with N1 cores. All of them will have
> > > > the same
> > > implementor ID and part number. But, they will have different values
> > > for these configurable parameters.
> > > > IMO, from usage perspective, we have 2 cases:
> > > > 1) Ability to build a portable binary that can run on multiple Arm
> > > > SoCs (for ex: BlueField, thunderx1, thunderx2, N1SDP, Graviton2
> > > > etc)
> > > > 2) Ability to build a binary which would run only on a SoC it was
> > > > compiled
> > > for and provide the most optimized binary for that SoC. But, this
> > > may not be portable.
> > > >
> > > > For 1) we have default march.
> > > >
> > > > For 2) we do not have the capability today in meson build (at
> > > > least, this is
> > > my understanding, please correct me if I am wrong). In this case,
> > > the user knows the target platform for compilation. IMO, we should
> > > add the capability to take the target platform as an input from the
> > > user (similar to the make build system) and Graviton2 can be one such
> target platform.
> > >
> > > My intention was to have parameters that work for both N1SDP and
> > > Graviton2 rather than 2). Does the change to RTE_MAX_LCORE and
> > > RTE_MAX_MEM_MB make them incompatible with N1SDP?
> > They are not optimal for N1SDP. In the future these parameters might have
> to be changed. For ex: if there a N1 based SoC with more than 64 CPU cores.
> 
> 
> Sorry for the late reply.
> 
> Looking at the Bluefield, Graviton2, and upcoming SoCs based on ARM IP, It is
> very clear that MIDR value can not be changed by the silicon vendors.
> So our existing build scheme of using the MIDR  value-based probe does not
> work anymore with ARM IP.
> So IMO, We need to change our build scheme. i.e
> 
> 1) For native build just use -march=native
I think we do not need the native build. With the native build, it is not possible to identify the SoC and use the correct configuration parameters.

> 2) For cross-build, explicitly, mention the target to pick the configuration
> values instead of probing the MIDR value-based scheme.
Agree

> 
> If we agree, Any volunteers for the update to new scheme?
Arm will work on this.

> 
> >
> > >
> > > I'm not sure if taking target platform from user is the best option here.
> > > Would this be specific to N1 since other platforms like thunderx
> > > differentiate the flags with part number?
> > This issue is specific to Arm CPU cores in general. So, it applies to N1 too.
> >
> > >
> > > >
> > > >> 3. remove +crc from -march as that is default when setting
> > > >> armv8.2
> > > >>
> > > >> For more information about Graviton2 platform, refer to:
> > > >> https://aws.amazon.com/ec2/graviton/
> > > >>
> > > >> Signed-off-by: Vimal Chungath <vcchunga@amazon.com>
> > > >> ---
> > > >>  config/arm/arm64_graviton2_linux_gcc | 17 +++++++++++++++++
> > > >>  config/arm/meson.build               | 12 +++++++++++-
> > > >>  2 files changed, 28 insertions(+), 1 deletion(-)  create mode
> > > >> 100644 config/arm/arm64_graviton2_linux_gcc
> > > >>
> > > >> diff --git a/config/arm/arm64_graviton2_linux_gcc
> > > >> b/config/arm/arm64_graviton2_linux_gcc
> > > >> new file mode 100644
> > > >> index 000000000..022e06303
> > > >> --- /dev/null
> > > >> +++ b/config/arm/arm64_graviton2_linux_gcc
> > > >> @@ -0,0 +1,17 @@
> > > >> +[binaries]
> > > >> +c = 'aarch64-linux-gnu-gcc'
> > > >> +cpp = 'aarch64-linux-gnu-cpp'
> > > >> +ar = 'aarch64-linux-gnu-gcc-ar'
> > > >> +strip = 'aarch64-linux-gnu-strip'
> > > >> +pkgconfig = 'aarch64-linux-gnu-pkg-config'
> > > >> +pcap-config = ''
> > > >> +
> > > >> +[host_machine]
> > > >> +system = 'linux'
> > > >> +cpu_family = 'aarch64'
> > > >> +cpu = 'armv8-a'
> > > >> +endian = 'little'
> > > >> +
> > > >> +[properties]
> > > >> +implementor_id = '0x41'
> > > >> +implementor_pn = '0xd0c'
> > > >> diff --git a/config/arm/meson.build b/config/arm/meson.build
> > > >> index 8728051d5..64e277ebc 100644
> > > >> --- a/config/arm/meson.build
> > > >> +++ b/config/arm/meson.build
> > > >> @@ -86,6 +86,16 @@ flags_octeontx2_extra = [
> > > >>       ['RTE_ARM_FEATURE_ATOMICS', true],
> > > >>       ['RTE_EAL_IGB_UIO', false],
> > > >>       ['RTE_USE_C11_MEM_MODEL', true]]
> > > >> +flags_n1generic_extra = [
> > > >> +     ['RTE_MACHINE', '"neoverse-n1"'],
> > > >> +     ['RTE_MAX_LCORE', 64],
> > > >> +     ['RTE_CACHE_LINE_SIZE', 64],
> > > >> +     ['RTE_ARM_FEATURE_ATOMICS', true],
> > > >> +     ['RTE_USE_C11_MEM_MODEL', true],
> > > >> +     ['RTE_MAX_MEM_MB', 1048576],
> > > >> +     ['RTE_MAX_NUMA_NODES', 1],
> > > >> +     ['RTE_EAL_NUMA_AWARE_HUGEPAGES', false],
> > > >> +     ['RTE_LIBRTE_VHOST_NUMA', false]]
> > > >>
> > > >>  machine_args_generic = [
> > > >>       ['default', ['-march=armv8-a+crc']], @@ -97,7 +107,7 @@
> > > >> machine_args_generic = [
> > > >>       ['0xd09', ['-mcpu=cortex-a73']],
> > > >>       ['0xd0a', ['-mcpu=cortex-a75']],
> > > >>       ['0xd0b', ['-mcpu=cortex-a76']],
> > > >> -     ['0xd0c', ['-march=armv8.2-a+crc+crypto', '-mcpu=neoverse-n1'],
> > > >> flags_n1sdp_extra]]
> > > >> +     ['0xd0c', ['-march=armv8.2-a+crypto', '-mcpu=neoverse-n1'],
> > > >> +flags_n1generic_extra]]
> > > >>
> > > >>  machine_args_cavium = [
> > > >>       ['default',
> > > >> ['-march=armv8-a+crc+crypto','-mcpu=thunderx']],
> > > >> --
> > > >> 2.16.6
> > > >
> >
Jerin Jacob Sept. 18, 2020, 4:02 a.m. UTC | #6
On Thu, Sep 17, 2020 at 10:41 PM Honnappa Nagarahalli
<Honnappa.Nagarahalli@arm.com> wrote:
>
> <snip>
>
> > > >
> > > > On 9/11/20 8:23 PM, Honnappa Nagarahalli wrote:
> > > > >
> > > > > +Jerin, Hemant, Dharmik
> > > > >
> > > > > <snip>
> > > > > Hi Vimal,
> > > > >         Few comments inline.
> > > > >
> > > > >>
> > > > >> Add meson build configuration for Graviton2 platform with 64-bit
> > > > >> ARM Neoverse N1 cores. This patch makes the following changes to
> > > > >> generic Neoverse N1 config:
> > > > >>
> > > > >> 1. increase lcore limit to 64
> > > > >> 2. increase memory support to 1TB
> > > > > There will be multiple SoCs with N1 cores. All of them will have
> > > > > the same
> > > > implementor ID and part number. But, they will have different values
> > > > for these configurable parameters.
> > > > > IMO, from usage perspective, we have 2 cases:
> > > > > 1) Ability to build a portable binary that can run on multiple Arm
> > > > > SoCs (for ex: BlueField, thunderx1, thunderx2, N1SDP, Graviton2
> > > > > etc)
> > > > > 2) Ability to build a binary which would run only on a SoC it was
> > > > > compiled
> > > > for and provide the most optimized binary for that SoC. But, this
> > > > may not be portable.
> > > > >
> > > > > For 1) we have default march.
> > > > >
> > > > > For 2) we do not have the capability today in meson build (at
> > > > > least, this is
> > > > my understanding, please correct me if I am wrong). In this case,
> > > > the user knows the target platform for compilation. IMO, we should
> > > > add the capability to take the target platform as an input from the
> > > > user (similar to the make build system) and Graviton2 can be one such
> > target platform.
> > > >
> > > > My intention was to have parameters that work for both N1SDP and
> > > > Graviton2 rather than 2). Does the change to RTE_MAX_LCORE and
> > > > RTE_MAX_MEM_MB make them incompatible with N1SDP?
> > > They are not optimal for N1SDP. In the future these parameters might have
> > to be changed. For ex: if there a N1 based SoC with more than 64 CPU cores.
> >
> >
> > Sorry for the late reply.
> >
> > Looking at the Bluefield, Graviton2, and upcoming SoCs based on ARM IP, It is
> > very clear that MIDR value can not be changed by the silicon vendors.
> > So our existing build scheme of using the MIDR  value-based probe does not
> > work anymore with ARM IP.
> > So IMO, We need to change our build scheme. i.e
> >
> > 1) For native build just use -march=native
> I think we do not need the native build. With the native build, it is not possible to identify the SoC and use the correct configuration parameters.

Ack. Both native and cross can use same config.

>
> > 2) For cross-build, explicitly, mention the target to pick the configuration
> > values instead of probing the MIDR value-based scheme.
> Agree
>
> >
> > If we agree, Any volunteers for the update to new scheme?
> Arm will work on this.

Thanks. We wait for the patch for the rework before applying Graviton patches.


>
> >
> > >
> > > >
> > > > I'm not sure if taking target platform from user is the best option here.
> > > > Would this be specific to N1 since other platforms like thunderx
> > > > differentiate the flags with part number?
> > > This issue is specific to Arm CPU cores in general. So, it applies to N1 too.
> > >
> > > >
> > > > >
> > > > >> 3. remove +crc from -march as that is default when setting
> > > > >> armv8.2
> > > > >>
> > > > >> For more information about Graviton2 platform, refer to:
> > > > >> https://aws.amazon.com/ec2/graviton/
> > > > >>
> > > > >> Signed-off-by: Vimal Chungath <vcchunga@amazon.com>
> > > > >> ---
> > > > >>  config/arm/arm64_graviton2_linux_gcc | 17 +++++++++++++++++
> > > > >>  config/arm/meson.build               | 12 +++++++++++-
> > > > >>  2 files changed, 28 insertions(+), 1 deletion(-)  create mode
> > > > >> 100644 config/arm/arm64_graviton2_linux_gcc
> > > > >>
> > > > >> diff --git a/config/arm/arm64_graviton2_linux_gcc
> > > > >> b/config/arm/arm64_graviton2_linux_gcc
> > > > >> new file mode 100644
> > > > >> index 000000000..022e06303
> > > > >> --- /dev/null
> > > > >> +++ b/config/arm/arm64_graviton2_linux_gcc
> > > > >> @@ -0,0 +1,17 @@
> > > > >> +[binaries]
> > > > >> +c = 'aarch64-linux-gnu-gcc'
> > > > >> +cpp = 'aarch64-linux-gnu-cpp'
> > > > >> +ar = 'aarch64-linux-gnu-gcc-ar'
> > > > >> +strip = 'aarch64-linux-gnu-strip'
> > > > >> +pkgconfig = 'aarch64-linux-gnu-pkg-config'
> > > > >> +pcap-config = ''
> > > > >> +
> > > > >> +[host_machine]
> > > > >> +system = 'linux'
> > > > >> +cpu_family = 'aarch64'
> > > > >> +cpu = 'armv8-a'
> > > > >> +endian = 'little'
> > > > >> +
> > > > >> +[properties]
> > > > >> +implementor_id = '0x41'
> > > > >> +implementor_pn = '0xd0c'
> > > > >> diff --git a/config/arm/meson.build b/config/arm/meson.build
> > > > >> index 8728051d5..64e277ebc 100644
> > > > >> --- a/config/arm/meson.build
> > > > >> +++ b/config/arm/meson.build
> > > > >> @@ -86,6 +86,16 @@ flags_octeontx2_extra = [
> > > > >>       ['RTE_ARM_FEATURE_ATOMICS', true],
> > > > >>       ['RTE_EAL_IGB_UIO', false],
> > > > >>       ['RTE_USE_C11_MEM_MODEL', true]]
> > > > >> +flags_n1generic_extra = [
> > > > >> +     ['RTE_MACHINE', '"neoverse-n1"'],
> > > > >> +     ['RTE_MAX_LCORE', 64],
> > > > >> +     ['RTE_CACHE_LINE_SIZE', 64],
> > > > >> +     ['RTE_ARM_FEATURE_ATOMICS', true],
> > > > >> +     ['RTE_USE_C11_MEM_MODEL', true],
> > > > >> +     ['RTE_MAX_MEM_MB', 1048576],
> > > > >> +     ['RTE_MAX_NUMA_NODES', 1],
> > > > >> +     ['RTE_EAL_NUMA_AWARE_HUGEPAGES', false],
> > > > >> +     ['RTE_LIBRTE_VHOST_NUMA', false]]
> > > > >>
> > > > >>  machine_args_generic = [
> > > > >>       ['default', ['-march=armv8-a+crc']], @@ -97,7 +107,7 @@
> > > > >> machine_args_generic = [
> > > > >>       ['0xd09', ['-mcpu=cortex-a73']],
> > > > >>       ['0xd0a', ['-mcpu=cortex-a75']],
> > > > >>       ['0xd0b', ['-mcpu=cortex-a76']],
> > > > >> -     ['0xd0c', ['-march=armv8.2-a+crc+crypto', '-mcpu=neoverse-n1'],
> > > > >> flags_n1sdp_extra]]
> > > > >> +     ['0xd0c', ['-march=armv8.2-a+crypto', '-mcpu=neoverse-n1'],
> > > > >> +flags_n1generic_extra]]
> > > > >>
> > > > >>  machine_args_cavium = [
> > > > >>       ['default',
> > > > >> ['-march=armv8-a+crc+crypto','-mcpu=thunderx']],
> > > > >> --
> > > > >> 2.16.6
> > > > >
> > >
Vimal Chungath Sept. 18, 2020, 5:42 p.m. UTC | #7
On 9/17/20 9:02 PM, Jerin Jacob wrote:
>
> On Thu, Sep 17, 2020 at 10:41 PM Honnappa Nagarahalli
> <Honnappa.Nagarahalli@arm.com> wrote:
>>
>> <snip>
>>
>>>>>
>>>>> On 9/11/20 8:23 PM, Honnappa Nagarahalli wrote:
>>>>>>
>>>>>> +Jerin, Hemant, Dharmik
>>>>>>
>>>>>> <snip>
>>>>>> Hi Vimal,
>>>>>>         Few comments inline.
>>>>>>
>>>>>>>
>>>>>>> Add meson build configuration for Graviton2 platform with 64-bit
>>>>>>> ARM Neoverse N1 cores. This patch makes the following changes to
>>>>>>> generic Neoverse N1 config:
>>>>>>>
>>>>>>> 1. increase lcore limit to 64
>>>>>>> 2. increase memory support to 1TB
>>>>>> There will be multiple SoCs with N1 cores. All of them will have
>>>>>> the same
>>>>> implementor ID and part number. But, they will have different values
>>>>> for these configurable parameters.
>>>>>> IMO, from usage perspective, we have 2 cases:
>>>>>> 1) Ability to build a portable binary that can run on multiple Arm
>>>>>> SoCs (for ex: BlueField, thunderx1, thunderx2, N1SDP, Graviton2
>>>>>> etc)
>>>>>> 2) Ability to build a binary which would run only on a SoC it was
>>>>>> compiled
>>>>> for and provide the most optimized binary for that SoC. But, this
>>>>> may not be portable.
>>>>>>
>>>>>> For 1) we have default march.
>>>>>>
>>>>>> For 2) we do not have the capability today in meson build (at
>>>>>> least, this is
>>>>> my understanding, please correct me if I am wrong). In this case,
>>>>> the user knows the target platform for compilation. IMO, we should
>>>>> add the capability to take the target platform as an input from the
>>>>> user (similar to the make build system) and Graviton2 can be one such
>>> target platform.
>>>>>
>>>>> My intention was to have parameters that work for both N1SDP and
>>>>> Graviton2 rather than 2). Does the change to RTE_MAX_LCORE and
>>>>> RTE_MAX_MEM_MB make them incompatible with N1SDP?
>>>> They are not optimal for N1SDP. In the future these parameters might have
>>> to be changed. For ex: if there a N1 based SoC with more than 64 CPU cores.
>>>
>>>
>>> Sorry for the late reply.
>>>
>>> Looking at the Bluefield, Graviton2, and upcoming SoCs based on ARM IP, It is
>>> very clear that MIDR value can not be changed by the silicon vendors.
>>> So our existing build scheme of using the MIDR  value-based probe does not
>>> work anymore with ARM IP.
>>> So IMO, We need to change our build scheme. i.e
>>>
>>> 1) For native build just use -march=native
>> I think we do not need the native build. With the native build, it is not possible to identify the SoC and use the correct configuration parameters.
>
> Ack. Both native and cross can use same config.
>
>>
>>> 2) For cross-build, explicitly, mention the target to pick the configuration
>>> values instead of probing the MIDR value-based scheme.
>> Agree
>>
>>>
>>> If we agree, Any volunteers for the update to new scheme?
>> Arm will work on this.
>
> Thanks. We wait for the patch for the rework before applying Graviton patches.
>
Ack. I will update Graviton patch once we have the meson build changes.

>
>>
>>>
>>>>
>>>>>
>>>>> I'm not sure if taking target platform from user is the best option here.
>>>>> Would this be specific to N1 since other platforms like thunderx
>>>>> differentiate the flags with part number?
>>>> This issue is specific to Arm CPU cores in general. So, it applies to N1 too.
>>>>
>>>>>
>>>>>>
>>>>>>> 3. remove +crc from -march as that is default when setting
>>>>>>> armv8.2
>>>>>>>
>>>>>>> For more information about Graviton2 platform, refer to:
>>>>>>> https://aws.amazon.com/ec2/graviton/
>>>>>>>
>>>>>>> Signed-off-by: Vimal Chungath <vcchunga@amazon.com>
>>>>>>> ---
>>>>>>>  config/arm/arm64_graviton2_linux_gcc | 17 +++++++++++++++++
>>>>>>>  config/arm/meson.build               | 12 +++++++++++-
>>>>>>>  2 files changed, 28 insertions(+), 1 deletion(-)  create mode
>>>>>>> 100644 config/arm/arm64_graviton2_linux_gcc
>>>>>>>
>>>>>>> diff --git a/config/arm/arm64_graviton2_linux_gcc
>>>>>>> b/config/arm/arm64_graviton2_linux_gcc
>>>>>>> new file mode 100644
>>>>>>> index 000000000..022e06303
>>>>>>> --- /dev/null
>>>>>>> +++ b/config/arm/arm64_graviton2_linux_gcc
>>>>>>> @@ -0,0 +1,17 @@
>>>>>>> +[binaries]
>>>>>>> +c = 'aarch64-linux-gnu-gcc'
>>>>>>> +cpp = 'aarch64-linux-gnu-cpp'
>>>>>>> +ar = 'aarch64-linux-gnu-gcc-ar'
>>>>>>> +strip = 'aarch64-linux-gnu-strip'
>>>>>>> +pkgconfig = 'aarch64-linux-gnu-pkg-config'
>>>>>>> +pcap-config = ''
>>>>>>> +
>>>>>>> +[host_machine]
>>>>>>> +system = 'linux'
>>>>>>> +cpu_family = 'aarch64'
>>>>>>> +cpu = 'armv8-a'
>>>>>>> +endian = 'little'
>>>>>>> +
>>>>>>> +[properties]
>>>>>>> +implementor_id = '0x41'
>>>>>>> +implementor_pn = '0xd0c'
>>>>>>> diff --git a/config/arm/meson.build b/config/arm/meson.build
>>>>>>> index 8728051d5..64e277ebc 100644
>>>>>>> --- a/config/arm/meson.build
>>>>>>> +++ b/config/arm/meson.build
>>>>>>> @@ -86,6 +86,16 @@ flags_octeontx2_extra = [
>>>>>>>       ['RTE_ARM_FEATURE_ATOMICS', true],
>>>>>>>       ['RTE_EAL_IGB_UIO', false],
>>>>>>>       ['RTE_USE_C11_MEM_MODEL', true]]
>>>>>>> +flags_n1generic_extra = [
>>>>>>> +     ['RTE_MACHINE', '"neoverse-n1"'],
>>>>>>> +     ['RTE_MAX_LCORE', 64],
>>>>>>> +     ['RTE_CACHE_LINE_SIZE', 64],
>>>>>>> +     ['RTE_ARM_FEATURE_ATOMICS', true],
>>>>>>> +     ['RTE_USE_C11_MEM_MODEL', true],
>>>>>>> +     ['RTE_MAX_MEM_MB', 1048576],
>>>>>>> +     ['RTE_MAX_NUMA_NODES', 1],
>>>>>>> +     ['RTE_EAL_NUMA_AWARE_HUGEPAGES', false],
>>>>>>> +     ['RTE_LIBRTE_VHOST_NUMA', false]]
>>>>>>>
>>>>>>>  machine_args_generic = [
>>>>>>>       ['default', ['-march=armv8-a+crc']], @@ -97,7 +107,7 @@
>>>>>>> machine_args_generic = [
>>>>>>>       ['0xd09', ['-mcpu=cortex-a73']],
>>>>>>>       ['0xd0a', ['-mcpu=cortex-a75']],
>>>>>>>       ['0xd0b', ['-mcpu=cortex-a76']],
>>>>>>> -     ['0xd0c', ['-march=armv8.2-a+crc+crypto', '-mcpu=neoverse-n1'],
>>>>>>> flags_n1sdp_extra]]
>>>>>>> +     ['0xd0c', ['-march=armv8.2-a+crypto', '-mcpu=neoverse-n1'],
>>>>>>> +flags_n1generic_extra]]
>>>>>>>
>>>>>>>  machine_args_cavium = [
>>>>>>>       ['default',
>>>>>>> ['-march=armv8-a+crc+crypto','-mcpu=thunderx']],
>>>>>>> --
>>>>>>> 2.16.6
>>>>>>
>>>>
Ali Saidi Sept. 19, 2020, 7:06 p.m. UTC | #8
On 9/17/20 9:02 PM, Jerin Jacob wrote:
>
> On Thu, Sep 17, 2020 at 10:41 PM Honnappa Nagarahalli
> <Honnappa.Nagarahalli@arm.com> wrote:
>>
>> <snip>
>>
>>>>>
>>>>> On 9/11/20 8:23 PM, Honnappa Nagarahalli wrote:
>>>>>>
>>>>>> +Jerin, Hemant, Dharmik
>>>>>>
>>>>>> <snip>
>>>>>> Hi Vimal,
>>>>>>         Few comments inline.
>>>>>>
>>>>>>>
>>>>>>> Add meson build configuration for Graviton2 platform with 64-bit
>>>>>>> ARM Neoverse N1 cores. This patch makes the following changes to
>>>>>>> generic Neoverse N1 config:
>>>>>>>
>>>>>>> 1. increase lcore limit to 64
>>>>>>> 2. increase memory support to 1TB
>>>>>> There will be multiple SoCs with N1 cores. All of them will have
>>>>>> the same
>>>>> implementor ID and part number. But, they will have different values
>>>>> for these configurable parameters.
>>>>>> IMO, from usage perspective, we have 2 cases:
>>>>>> 1) Ability to build a portable binary that can run on multiple Arm
>>>>>> SoCs (for ex: BlueField, thunderx1, thunderx2, N1SDP, Graviton2
>>>>>> etc)
>>>>>> 2) Ability to build a binary which would run only on a SoC it was
>>>>>> compiled
>>>>> for and provide the most optimized binary for that SoC. But, this
>>>>> may not be portable.
>>>>>>
>>>>>> For 1) we have default march.
>>>>>>
>>>>>> For 2) we do not have the capability today in meson build (at
>>>>>> least, this is
>>>>> my understanding, please correct me if I am wrong). In this case,
>>>>> the user knows the target platform for compilation. IMO, we should
>>>>> add the capability to take the target platform as an input from the
>>>>> user (similar to the make build system) and Graviton2 can be one such
>>> target platform.
>>>>>
>>>>> My intention was to have parameters that work for both N1SDP and
>>>>> Graviton2 rather than 2). Does the change to RTE_MAX_LCORE and
>>>>> RTE_MAX_MEM_MB make them incompatible with N1SDP?
>>>> They are not optimal for N1SDP. In the future these parameters might have
>>> to be changed. For ex: if there a N1 based SoC with more than 64 CPU cores.
>>>
>>>
>>> Sorry for the late reply.
>>>
>>> Looking at the Bluefield, Graviton2, and upcoming SoCs based on ARM IP, It is
>>> very clear that MIDR value can not be changed by the silicon vendors.
>>> So our existing build scheme of using the MIDR  value-based probe does not
>>> work anymore with ARM IP.
>>> So IMO, We need to change our build scheme. i.e
>>>
>>> 1) For native build just use -march=native
>> I think we do not need the native build. With the native build, it is not possible to identify the SoC and use the correct configuration parameters.

For native binaries the answer should be either use -march=native or map the MIDR to
the compiler options. The MIDR defines the core and thus the core features and which
architecture features the compiler supports. It seems like the only sticking point here
is RTE_MAX_MEM_MB and RTE_MAX_LCORE. Is the slight increase in memory size by having a
reasonable max for the various SoCs given a single core MIDR (80 seems to be the high
number for N1 currently), reason to add complexity and require the user to explicitly
pass the system they're compiling for?

Thanks,

Ali
Honnappa Nagarahalli Sept. 20, 2020, 12:41 a.m. UTC | #9
<snip>

> On 9/18/20, 12:42 PM, "Vimal Chungath" <vcchunga@amazon.com> wrote:
> 
>     On 9/17/20 9:02 PM, Jerin Jacob wrote:
>     >
>     > On Thu, Sep 17, 2020 at 10:41 PM Honnappa Nagarahalli
>     > <Honnappa.Nagarahalli@arm.com> wrote:
>     >>
>     >> <snip>
>     >>
>     >>>>>
>     >>>>> On 9/11/20 8:23 PM, Honnappa Nagarahalli wrote:
>     >>>>>>
>     >>>>>> +Jerin, Hemant, Dharmik
>     >>>>>>
>     >>>>>> <snip>
>     >>>>>> Hi Vimal,
>     >>>>>>         Few comments inline.
>     >>>>>>
>     >>>>>>>
>     >>>>>>> Add meson build configuration for Graviton2 platform with 64-bit
>     >>>>>>> ARM Neoverse N1 cores. This patch makes the following changes
> to
>     >>>>>>> generic Neoverse N1 config:
>     >>>>>>>
>     >>>>>>> 1. increase lcore limit to 64
>     >>>>>>> 2. increase memory support to 1TB
>     >>>>>> There will be multiple SoCs with N1 cores. All of them will have
>     >>>>>> the same
>     >>>>> implementor ID and part number. But, they will have different values
>     >>>>> for these configurable parameters.
>     >>>>>> IMO, from usage perspective, we have 2 cases:
>     >>>>>> 1) Ability to build a portable binary that can run on multiple Arm
>     >>>>>> SoCs (for ex: BlueField, thunderx1, thunderx2, N1SDP, Graviton2
>     >>>>>> etc)
>     >>>>>> 2) Ability to build a binary which would run only on a SoC it was
>     >>>>>> compiled
>     >>>>> for and provide the most optimized binary for that SoC. But, this
>     >>>>> may not be portable.
> For native binaries the answer should be either use -march=native or map the
> MIDR to the compiler options. The MIDR defines the core and thus the core
> features and which architecture features the compiler supports. It seems like
The MIDR identifies the core. It does not identify the SoC. We will use the option of taking the target SoC from the user and use a config to get the correct compiler options.

> the only sticking point here is RTE_MAX_MEM_MB and RTE_MAX_LCORE. Is
> the slight increase in memory size by having a reasonable max for the various
> SoCs given a single core MIDR (80 seems to be the high number for N1
> currently), reason to add complexity and require the user to explicitly pass
> the system they're compiling for?
The largest I know is 128 cores [1]. But, the SoCs could have as little as 4 cores. The new method provides complete flexibility in controlling more parameters than what can be done currently. IMO, the solution is not complex and is inline with the way the make build system works currently.

[1] https://www.hpcwire.com/off-the-wire/ampere-announces-altra-max-arm-processor-with-128-cores/

> 
> 
> Ali
> 
>

Patch
diff mbox series

diff --git a/config/arm/arm64_graviton2_linux_gcc b/config/arm/arm64_graviton2_linux_gcc
new file mode 100644
index 000000000..022e06303
--- /dev/null
+++ b/config/arm/arm64_graviton2_linux_gcc
@@ -0,0 +1,17 @@ 
+[binaries]
+c = 'aarch64-linux-gnu-gcc'
+cpp = 'aarch64-linux-gnu-cpp'
+ar = 'aarch64-linux-gnu-gcc-ar'
+strip = 'aarch64-linux-gnu-strip'
+pkgconfig = 'aarch64-linux-gnu-pkg-config'
+pcap-config = ''
+
+[host_machine]
+system = 'linux'
+cpu_family = 'aarch64'
+cpu = 'armv8-a'
+endian = 'little'
+
+[properties]
+implementor_id = '0x41'
+implementor_pn = '0xd0c'
diff --git a/config/arm/meson.build b/config/arm/meson.build
index 8728051d5..64e277ebc 100644
--- a/config/arm/meson.build
+++ b/config/arm/meson.build
@@ -86,6 +86,16 @@  flags_octeontx2_extra = [
 	['RTE_ARM_FEATURE_ATOMICS', true],
 	['RTE_EAL_IGB_UIO', false],
 	['RTE_USE_C11_MEM_MODEL', true]]
+flags_n1generic_extra = [
+	['RTE_MACHINE', '"neoverse-n1"'],
+	['RTE_MAX_LCORE', 64],
+	['RTE_CACHE_LINE_SIZE', 64],
+	['RTE_ARM_FEATURE_ATOMICS', true],
+	['RTE_USE_C11_MEM_MODEL', true],
+	['RTE_MAX_MEM_MB', 1048576],
+	['RTE_MAX_NUMA_NODES', 1],
+	['RTE_EAL_NUMA_AWARE_HUGEPAGES', false],
+	['RTE_LIBRTE_VHOST_NUMA', false]]
 
 machine_args_generic = [
 	['default', ['-march=armv8-a+crc']],
@@ -97,7 +107,7 @@  machine_args_generic = [
 	['0xd09', ['-mcpu=cortex-a73']],
 	['0xd0a', ['-mcpu=cortex-a75']],
 	['0xd0b', ['-mcpu=cortex-a76']],
-	['0xd0c', ['-march=armv8.2-a+crc+crypto', '-mcpu=neoverse-n1'], flags_n1sdp_extra]]
+	['0xd0c', ['-march=armv8.2-a+crypto', '-mcpu=neoverse-n1'], flags_n1generic_extra]]
 
 machine_args_cavium = [
 	['default', ['-march=armv8-a+crc+crypto','-mcpu=thunderx']],