Weight #166
Weight #166
Conversation
Primitives are still used in the specification of jets, but can no longer appear in Dags for serialization. Indeed, attemps to serialize them prior to this commit would fail with a pattern matching error. This removes them from the Dag data structure entirely.
Set the outputSize and inputSize from the type information.
They can be inferred from the type information.
Refactor computeEvalTCOBound into analyseBounds and allowed it to be called outside of evalTCOExpression. This will facilitate testing of the bounds computation.
roconnor-blockstream
force-pushed
the
weight
branch
from
7c67950 to
56f15be
Compare
|
@apoelstra only the C code needs to be reviewed. I've shuffled around the commits quite a bit during development, so I do recommend giving it a pass through CI before looking at it too deeply. |
|
utACK 56f15be. I have a mild preference for the "fix Cost.hs" commit to come first, before any computed values are introduced, avoiding committing then reverting bad numbers. But this is fine. Will run CI when my box is free. |
|
Yeah regarding your "fix Cost.hs" preference. Your choices are between being able to check that the regression test works, by swapping to adjacent commits, or doing your proposal and having to swap far apart commits, regenerate the failing data and dealing with the merge conflict that would arise. I'm okay with either approach if you have a preference. I don't see a way of getting the best of both solutions, other than maybe bizarre branches and merges. |
roconnor-blockstream
force-pushed
the
weight
branch
from
56f15be to
793c7b1
Compare
|
@roconnor-blockstream ah, yeah, good point. You do actually need to include the bad data for the test to still compile when it's swapped. Ok, we'll keep things as is. |
| free(bound); | ||
| return true; | ||
| return (maxCells <= *cellsBound) ? SIMPLICITY_ERR_EXEC_MEMORY | ||
| : (maxCost <= *costBound) ? SIMPLICITY_ERR_EXEC_BUDGET |
There was a problem hiding this comment.
I don't know how I missed it, but these inequalities were strict before and now they are non-strict, which is incorrect. I'll prepare a PR to correct this.
The written specification of "the bound on memory allocation requirements exceed the allowed limits" is just too vague.
This PR does a couple of related things: