You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
sha2_blocks is a test case which is resource intensive and generates a big circuit. However in term of feature it is doing the same as sha2_byte, except that it has a much bigger input (3, 56 and 112 bytes input vs one).
Having sha2_blocks as part of the CI is slowing down CI a lot which is not good, but it cannot help to find more issues, except, incidentally, performance issues when the CI hits its timeout and fail.
We should have a dedicated way to detect performance issues, and not rely and something that was not designed for this.
Happy Case
Remove sha2_blocks from the CI
Alternatives Considered
No response
Additional Context
The overflow checks PR #2713 fail CI because of this test case. It turns out it is because sha256 is only using wrapping operations, and their stdlib implementation is not performant. Creating the circuit takes 8 minutes with stdlib wrapping. The resulting circuit is 20% bigger which seems OK.
We should analyse why stdlib wrapping takes more time and propose a builtin alternative.
Would you like to submit a PR for this Issue?
No
Support Needs
No response
The text was updated successfully, but these errors were encountered:
Problem
sha2_blocks is a test case which is resource intensive and generates a big circuit. However in term of feature it is doing the same as sha2_byte, except that it has a much bigger input (3, 56 and 112 bytes input vs one).
Having sha2_blocks as part of the CI is slowing down CI a lot which is not good, but it cannot help to find more issues, except, incidentally, performance issues when the CI hits its timeout and fail.
We should have a dedicated way to detect performance issues, and not rely and something that was not designed for this.
Happy Case
Remove sha2_blocks from the CI
Alternatives Considered
No response
Additional Context
The overflow checks PR #2713 fail CI because of this test case. It turns out it is because sha256 is only using wrapping operations, and their stdlib implementation is not performant. Creating the circuit takes 8 minutes with stdlib wrapping. The resulting circuit is 20% bigger which seems OK.
We should analyse why stdlib wrapping takes more time and propose a builtin alternative.
Would you like to submit a PR for this Issue?
No
Support Needs
No response
The text was updated successfully, but these errors were encountered: