Optimize crypto

[Ashuaidehao]

Description

Cahced crypto object could optimize verify signature performance。

Fixes # (no related)

Type of change

• Optimize

How Has This Been Tested?

Syncing blocks speed up about 25%.

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• Any dependent changes have been merged and published in downstream modules

[vncoelho]

Looks good and a simple change.
Can you check @superboyiii ?

[Jim8y]

This is already tested by @superboyiii .

[roman-khimov]

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]

Use FIFOCache instead of regular dictionary?

[Ashuaidehao]

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.
Or we can change it to time-out cache?

[roman-khimov]

with a large limit

To me it's still very, very similar to ECPointCache, so I'd use the same limit by default. Most of ECPoint uses is for verification anyway, so hitting one cache but not hitting the other is suboptimal to me.

[vncoelho]

@superboyiii ,can you check the performance difference when syncing?

[vncoelho]

If it is 25% as described we can merge now and later modify with other suggestions.

[Jim8y]

@superboyiii ,can you check the performance difference when syncing?

Its 25% - 30% performance increase from the benchmark. But its for syncing the chain.

[vncoelho]

@superboyiii ,can you check the performance difference when syncing?

Its 25% - 30% performance increase from the benchmark. But its for syncing the chain.

Very good,did u also replicate this result?

[shargon]

I found a bug in ECPoint comparation, it doesn't take care about the curve, now it should work

[roman-khimov]

Tests failing, can't be merged.

[shargon]

Tests failing, can't be merged.

The test checked that the bug was fine xD

obsolete

[superboyiii]

@superboyiii ,can you check the performance difference when syncing?

Yes, it's almost 30% speed up.