Conversation
vncoelho
left a comment
There was a problem hiding this comment.
Looks good and a simple change.
Can you check @superboyiii ?
|
This is already tested by @superboyiii . |
roman-khimov
left a comment
There was a problem hiding this comment.
It looks similar to ECPointCache, but at another level. Not sure why this can be a problem for C# (it takes time to calculate Y, here you already have it), but if it works (in a size-limited fashion of course) and gives measurable difference, then OK.
shargon
left a comment
There was a problem hiding this comment.
Use FIFOCache instead of regular dictionary?
We can change it to FIFOCache with a large limit. If the limit was small, the performance maybe more lower than no cache. |
To me it's still very, very similar to |
vncoelho
left a comment
There was a problem hiding this comment.
@superboyiii ,can you check the performance difference when syncing?
|
If it is 25% as described we can merge now and later modify with other suggestions. |
Its 25% - 30% performance increase from the benchmark. But its for syncing the chain. |
Very good,did u also replicate this result? |
|
I found a bug in |
roman-khimov
left a comment
There was a problem hiding this comment.
Tests failing, can't be merged.
The test checked that the bug was fine xD |
Yes, it's almost 30% speed up. |
Description
Cahced crypto object could optimize verify signature performance。
Fixes # (no related)
Type of change
How Has This Been Tested?
Syncing blocks speed up about 25%.
Checklist: