Rare project developers

[Alexander Gall <>]

On Thu, 15 Dec 2022 13:55:20 +0100, mc36 <> said:

> well, so i don't have a tofino2 at hand so all i can tell you is that
> the "no dead code elimination" was tested on the modem... according
> to your finding, that is you had a model-only workaround, i'm afraid
> to say a thing but worth a try... in the model, all the tests passed
> after adding this hack to my tofino2 vm...

For all profiles?  Which SDE version did you use? Did you run the
optimizer?

I'm unable to optimize some of the profiles with
"--no-dead-code-elimination":

vpls
geant-testbed
renater-peering-l3
nop-mchome

Those all fail with "no phv record" for some of the fields in ig_md
for all combinations of parameters the optimizer is trying.  That's
with SDE 9.7.4. With 9.11.1 it gets even worse, e.g.

running 
/nix/store/vyhk6qd7fg9254dh5dry3hbfdfk6fvfa-bf-sde-model-9.11.1/bin/bf-p4c 
-I. -I../p4src/ -Xp4c="--no-dead-code-elimination"  -b tofino2 -a t2na 
-DDUAL_PIPE optimizer821280.p4...

optimizer821280.tofino2/pipe/optimizer821280.bfa:4231: error: No phv record 
ig_md.bridge_src
optimizer821280.tofino2/pipe/optimizer821280.bfa:4239: error: No phv record 
ig_md.bridge_src
Assembler BUG
terminate called without an active exception

--
Alex

> br,
> cs

> On 12/15/22 13:25, Alexander Gall wrote:
>> Hi Fr  d  ric
>> 
>> Thanks for helping to debug the issue with the dropped packets on
>> Tofino2. Do I understand correctly that disabling dead-code
>> eliminiation when compiling for Tofino2 is a workaround for this
>> problem until Intel fixes the bug (P4C-5045) in the compiler?
>> 
>> If so, I will add the --no-dead-code-elimination flag to the build
>> procedure for Tofino2-based platforms in the rare-nix repo.
>> 
>> My previous workaround (commit 09e14ee4 in the rare repo) only worked
>> on the model but not on the ASIC. I notice that it has already been
>> removed by the combination of commits 5b0e759 and 3d6d736. That's fine
>> but I'm not sure if it was intentional (I'm still unhappy with these
>> automated commits...).
>>