Abe

Thanks for the conversation and/or explanation, Vladimir. In that case, perhaps the spec. should be modified with wording to the effect of “all operations in P4 are atomic unless annotated...

@jafingerhut, thanks for your detailed and thoughtful reply. Given that reply, I humbly re-instate my request/suggestion: "atomic" should be first class, **not** an annotation. IMO & TTBOMK, annotations are by...

@apinski-cavium I am **not** proposing/suggesting _"Doing things atomicly by default"_ at _all_. What I _am_ proposing/suggesting, in a nutshell: - **get rid of the '@'** in "@atomic" - make it...

@vgurevich: re _"the current language spec allows this to happen. I simply stated that no current implementation works this way"_: if the spec. allows for random interleavings when "@atomic" is...

@jafingerhut: I respectfully — but **_strongly_** — disagree with “_a P4 compiler should_ [perhaps] _automatically infer critical sections_”. IMO & AFAIK, automatic detection of where critical sections begin and end...

@jafingerhut ... > ``` > bit tmp2_0; > bit tmp3_0; > bit tmp; > action ct_table_hit(bit p1, bit p2) { > @atomic { > drop_packet(); > tmp = h1_0.get_hash((tuple_0){f0 =...

TLDR: I _strongly_ agree with most/all of what Mihai recently posted. On Thu, Apr 6, 2023 at 2:40 PM Mihai Budiu ***@***.***> wrote: > > I would think that this...

@jnfoster Doesn`t specifying an arch. implicitly specify a target?… e.g. “t2na” ⇒ “tofino2”.

@apinski-cavium I｀d rather not make this [a] macro[s] if we don｀t have to. AFAIK, the only valid use-case for macros is e.g. > #if TOFINO2 > > > > #elseif...