Subject: Rare project developers
List archive
- From: Alexander Gall <>
- To: <>
- Subject: Re: [rare-dev] Compilation failures for Tofino2
- Date: Fri, 3 Jun 2022 11:37:21 +0200
On Fri, 3 Jun 2022 11:04:40 +0200, Frédéric LOUI <>
said:
>> The forwarder alreay gets the platform passed in via --platform, so
>> that should be enough.
> OK !
>> We can simply use the __TARGET_TOFINO__ symbol to further distinguish
>> the different targets within the two pipe configurations, so no
>> additional flags are needed.
> FYI in 9.9.0 docs I found:
> - TOFINO1: 64Q, 32Q, 32D, 20D, 18D
> - TOFINO2: 128Q (A0, B0 & B1), 128QM, 64Q, 64D
According to the "Tofino2 Device Family Overview" document:
4-pipe: 128Q, 128QM, 128QMH, 64Q
3-pipe: 80T, 80TM
2-pipe: 64D
Maybe one of the Todino2 documents that have appeared recently on the
Intel site also contains info about the assignment of CPU and
recirculation ports.
Maybe it would be better to use the Chip IDs above instead of the
{DUAL,QUAD}_PIPE flags. This would be easy to do (just a replacement
in
https://bitbucket.software.geant.org/projects/RARE/repos/rare-nix/browse/rare/platforms.nix).
--
Alex
>> Le 3 juin 2022 à 10:37, Alexander Gall <> a écrit :
>>
>> On Fri, 3 Jun 2022 10:04:14 +0200, Frédéric LOUI
>> <> said:
>>
>>> OK, please let me know if you want me to add this --tofino2 trigger to
>>> bf_forwarder.py.
>>> Question:
>>
>>> We have 3 alternatives here:
>>> 1- add a manual trigger « --asic=tofino2 »
>>> 2- make bf_forwarder.py to interrogate bf_switchd via GRPC (granted the
>>> fact that CPU port is identified)
>>> 3- Or check ONIE machines value for AS9516-32D
>>
>> The forwarder alreay gets the platform passed in via --platform, so
>> that should be enough.
>>
>>> My question is that will you maintain a machine file in your build
>>> process ? If yes, it can be safe to use -3-
>>
>> The packaging uses per-platform flags for the compiler and
>> bf_forwarder (currently not used).
>>
>> The port speed stuff can be inferred from the platform identifier by
>> the forwarder, so we should already be good there.
>>
>> The CPU port must be fixed at compile time. Currently, we're using the
>> flags DUAL_PIPE and QUAD_PIPE to distinguish the two flavors:
>>
>> #if defined DUAL_PIPE
>>
>> #define CPU_PORT 192
>> #define RECIR_PORT 68
>>
>> #elif defined QUAD_PIPE
>>
>> #define CPU_PORT 320
>> #define RECIR_PORT 68
>>
>> #else
>>
>> /* Fall-through for the Tofino software emulation (Tofino model). It
>> doesn't emulate the CPU PCIe port, use the CPU Eth port instead. */
>>
>> #define CPU_PORT 64
>> #define RECIR_PORT 68
>>
>> #endif
>>
>> We can simply use the __TARGET_TOFINO__ symbol to further distinguish
>> the different targets within the two pipe configurations, so no
>> additional flags are needed.
>>
>> --
>> Alex
>>
>>
>>> Your opinion ?
>>> Frederic
>>
>>>> Le 3 juin 2022 à 07:46, Alexander Gall <> a écrit :
>>>>
>>>> We'll have to
>>>> at least adapt the forwarder too, at least for the new supported
>>>> interface speeds.
- Re: [rare-dev] Compilation failures for Tofino2, (continued)
- Re: [rare-dev] Compilation failures for Tofino2, mc36, 06/02/2022
- Re: [rare-dev] Compilation failures for Tofino2, Alexander Gall, 06/02/2022
- Re: [rare-dev] Compilation failures for Tofino2, mc36, 06/02/2022
- Re: [rare-dev] Compilation failures for Tofino2, mc36, 06/02/2022
- Re: [rare-dev] Compilation failures for Tofino2, Alexander Gall, 06/03/2022
- Re: [rare-dev] Compilation failures for Tofino2, mc36, 06/03/2022
- Re: [rare-dev] Compilation failures for Tofino2, Alexander Gall, 06/03/2022
- Re: [rare-dev] Compilation failures for Tofino2, Frédéric LOUI, 06/03/2022
- Re: [rare-dev] Compilation failures for Tofino2, Alexander Gall, 06/03/2022
- Re: [rare-dev] Compilation failures for Tofino2, Frédéric LOUI, 06/03/2022
- Re: [rare-dev] Compilation failures for Tofino2, Alexander Gall, 06/03/2022
- Re: [rare-dev] Compilation failures for Tofino2, Frédéric LOUI, 06/03/2022
- Re: [rare-dev] Compilation failures for Tofino2, mc36, 06/02/2022
- Re: [rare-dev] Compilation failures for Tofino2, Frédéric LOUI, 06/02/2022
Archive powered by MHonArc 2.6.19.