EIP198 - Bigint modexp precompiled contract

[gumb0]

#3615
ethereum/EIPs#198

• Contract code implementation
• Calculating the cost
• Optimize code implementation if boost::multiprecision::powm is too slow

[codecov-io]

Codecov Report

Merging #4028 into develop will increase coverage by 0.33%.
The diff coverage is 92.64%.

@@             Coverage Diff             @@
##           develop    #4028      +/-   ##
===========================================
+ Coverage     65.1%   65.43%   +0.33%     
===========================================
  Files          308      309       +1     
  Lines        22552    22670     +118     
===========================================
+ Hits         14683    14835     +152     
+ Misses        7869     7835      -34

Impacted Files	Coverage Δ
libethashseal/genesis/metropolisTest.cpp	100% <ø> (ø)	⬆️
libethcore/ChainOperationParams.h	100% <100%> (ø)	⬆️
libethcore/ChainOperationParams.cpp	100% <100%> (ø)	⬆️
test/unittests/libethcore/PrecompiledTest.cpp	100% <100%> (ø)
libethereum/Account.cpp	80.55% <55.55%> (-1.8%)	⬇️
libethcore/Precompiled.h	66.66% <66.66%> (-8.34%)	⬇️
libethcore/Precompiled.cpp	94.11% <96.29%> (+2.45%)	⬆️
libp2p/Host.cpp	72.36% <0%> (-1.06%)	⬇️
libp2p/Common.cpp	52% <0%> (-0.8%)	⬇️
... and 3 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 83b2679...aaf29c7. Read the comment docs.

[gumb0]

Looks like the boost::multiprecision::powm is not slower that the manual implemetation, since we don't know the modulo length and can't get rid of intermediate modulo operations (like we do for the EXP opcode implementation)

[gcolvin]

Looks good to me.

[gcolvin]

I don't recall the history here, but why didn't we just add a MODEXP opcode?

[chfast]

This is actually a good question, @gcolvin. Should we go to EIPs with it to ask for clarification when a features go as opcodes and when as precompiled contracts...

[chriseth]

@gcolvin for this particular function, the reason to choose a precompiled contract over an opcode was because the arguments can be larger than 32 bits and it would be weird to have such an opcode fetch from memory. Furthermore, I think the idea is that as much as possible should go to precompiled contracts instead of opcodes, and precompiled contracts have to be pure functions.