Conversation
| // gas ssaCFGOptimized: 809663 | ||
| // gas ssaCFGOptimized code: 570000 | ||
| // gas ssaCFGOptimized: 809711 | ||
| // gas ssaCFGOptimized code: 570600 |
There was a problem hiding this comment.
This is a small regression, but it is the only one on external contracts.
There was a problem hiding this comment.
regression by 0.0059% for runtime and by 0.105% for deploy.. i think it's fine..
| // gas legacy: 4234020 | ||
| // gas legacyOptimized: 2318668 | ||
| // gas ssaCFGOptimized: 1882355 | ||
| // gas ssaCFGOptimized: 1841235 |
There was a problem hiding this comment.
This is actually a 2% improvement.
| // SWAP3| * | lit2 v2 v0 | ||
| // POP| v0 | lit2 v2 | ||
| // DUP3| v0 | lit2 v2 v0 |
There was a problem hiding this comment.
This is a bit unfortunate, I would like to fix this somehow.
clonker
left a comment
clonker
left a comment
There was a problem hiding this comment.
nice, the regression seems tiny in comparison to what we gain (less complexity in the algo and even one gas win)
|
I would actually like to get back to this one to have another look, but for now, I think we can park it. |
This piece of logic was swapping top to the tail region if case the top is not in its right place and it wants to be in tail anyway. I don't have an explanation why this is useful, but it seems to improve code generation on most external benchmarks.
blishko
force-pushed
the
ssa-remove-more-shuffler-logic
branch
from
929b4e2 to
93cd9f7
Compare
|
This pull request is stale because it has been open for 14 days with no activity. |
github-actions
Bot
added
the
stale
blishko
removed
the
stale
2 participants
This piece of logic was swapping top to the tail region if case the top is not in its right place and it wants to be in tail anyway.
I don't have an explanation why this is useful, but it seems to improve code generation on most external benchmarks.