Bigint constants

[jochem-brouwer]

This PR:

• Caches most (not all) constants of BigInts monorepo-wide
• Caches ALL constants in EVM
• The monorepo-wide constants are now exported from util

Note: please review #3034 first (and merge in master if OK, or merge this on top if OK and then merge into master)

[codecov]

Codecov Report

Merging #3050 (53b3b6a) into master (aaa5de6) will decrease coverage by 0.09%.
The diff coverage is 79.61%.

Additional details and impacted files

Flag	Coverage Δ
block	88.78% <88.57%> (+0.05%)	⬆️
blockchain	92.61% <95.23%> (+0.02%)	⬆️
client	87.46% <88.02%> (-0.27%)	⬇️
common	98.19% <100.00%> (+<0.01%)	⬆️
ethash	∅ <ø> (∅)
evm	71.87% <59.50%> (+0.20%)	⬆️
rlp	∅ <ø> (∅)
statemanager	90.15% <ø> (ø)
trie	90.34% <ø> (+0.03%)	⬆️
tx	96.36% <100.00%> (+0.01%)	⬆️
util	86.97% <94.87%> (+0.09%)	⬆️
vm	78.12% <62.50%> (-0.16%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

[holgerd77]

Great stuff, thanks for taking this one step further! 🙏

Looks generally great and good to merge from my side (won’t do myself since I opened the initial PR, would be great if someone else could take upon that). Some tests still failing though.

[holgerd77]

Updated both the base branch and this branch and resolved a merge conflict, since I wanted to get the latest DB-layout (storage trie prefix) changes in to be able to run the profiler on this and compare.

[holgerd77]

Puh puh puh.

It is so hard to get reliable deterministic results, I know realized that having two different CLI tabs (one with the PR and one with master and switching between them) might/seems also to influence performance since the OS might assign different ressources (?) to the specific tab or something. 🤔

I've now switched to staying in one tab and getting the different full paths with the Linux pwd command and then running cd FULL_PATH_PR && npm run client:start ... and cd FULL_PATH_MASTER && npm run client:start ...(having a copied repository there) in alternation, that seems to produce the most "honest" results to me.

Just "live-writing" down the results here, I am not sure yet if they show some clear trend:

Running npm run client:start -- --sync=none --vmProfileBlocks --executeBlocks=2005567-2005590 in the way above and then again looking at the last block and noting here the BpS value (already totally fluent in our new terminology 😂):

PR : 11.32. 10.16. 9.613. 14.971. 15.277. 9.521. 15.092. 10.277
Master : 9.651. 10.23. 10.07. 10.463. 9.971. 16.087. 15.641. 10.241

For some reasons GitHub sometimes thinks to auto-insert a dot after these numbers makes sense, I'll leave it there.

Hmm. So these results are so utterly diverse, I guess one can't just read anything out of this.

I'll run the same command as above but compare the on one block (number) lower (so 2005589):

PR. : 32.583. 32.516. 31.984. 32.448. 30.251. 31.406. 32.788. 31.789
Master : 30.077. 31.789. 29.79. 30.251. 31.66. 31.596. 32.448. 31.853

Ok. Better.

Interesting that some blocks are so diverse.

Hmm. I guess one can't completely say for sure nevertheless if this improves things or not out of this results. Slight tendency eventually that it does.

So my conclusion would be somewhat: safe enough to say that this doesn't worsen things and likely enough that this somewhat improve things that we can merge. 🙂

[acolytec3]

LGTM. Going ahead and approving pending CI (though not sure if all will pass).

[holgerd77]

LGTM