P4 2016 draft: nonsensical cube expressions? about p4-spec

[vgurevich]

Hi Ben,

Formally speaking, these expressions are OK from the mathematical point of
view. What is true is that in this proposed syntax there might be several
ways how a particular set can be written. So, in the first case (a set of
all 8-bit numbers with the 3 least significant bits cleared) there are 32
ways to write it out in the proposed syntax, but only one way (8w0 &&& 8w7)
would be considered "canonical" and, frankly, it is in the eye of the
beholder. For example, some people might argue that 8w0xF8 &&& 8w7 (all
"don't care bits are set to 1") is better, depending on which TCAM design
technology they've grown up with.

Rather than making them illegal or undefined, I'd simply recommend the
compiler to issue a warning instead.

Thanks,
Vladimir

On Tue, Aug 16, 2016 at 3:33 PM, Ben Pfaff [email protected] wrote:

Section 10.11.3 doesn't say anything about nonsense cube expressions, that
is, ones where bits are set in the value but not in the mask, such as 8w8
&&& 8w7. According to the definition provided, such an expression is
equivalent to 8w0 &&& 8w7, but it might be worth making the meaning
implementation-defined or undefined since it doesn't make much sense to
have a 1-bit in the value but a 0-bit in the mask.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#41, or mute the thread
https://github.com/notifications/unsubscribe-auth/AMPb88KdXuxL4TOS4H6GzVUNguOcjKZcks5qgjqxgaJpZM4Jl6wu
.

[blp]

I did not realize that some might consider "all-1 don't-cares" canonical. One learns something new every day.

The core of my point was to recommend that the specification note that there's more than one way to specify equivalent matching contraints, so I'm happy with anything that moves in that direction.

[chkim4142]

Can we close this?

[mihaibudiu]

I hope so, but I was waiting for @blp to check that the spec fixes this issue.