EVM Charts 2024: What's hot? What's not? by Dominic Bruetsch | Devcon SEA

[Music] hello everyone good morning welcome to this presentation we're going to see some statistics about e evm especially we're going to compare some data um we collected data from 100,000 blocks before last Devcon and 100,000 blocks before thison Defcon and we're going to see what changed I'm Dominic Bridge one of the lead Auditors and co-founders at chain security so let's Dive Right In but first um I have a question to you what do you think is the most popular op code used in the evm it's push one it's mainly used to push small on the stack next is push two um it's for larger constant and this one is especially used to push jump targets onto the stack um f uh third in row is Jump test this marks jump test in evm code and you might wonder why we don't see the jump instruction itself being used heavily and this is due to in the evm we have two instructions to jump jump and jump I for conditional jump and both together roughly add up to the same number as jump test overall between the data sets of Defcon before the B before the Defcon bota and before this Devcon we see no change in the most popular op codes moving on to the least popular op codes um least popular op code is call code this is basically a helpful variant of delegate call and it's used really few times Only log zero just emits log events with zero topics and invalid op codes they basically simply consume all the gas of the current context when they are called now prior to Devcon 7 this changed slightly create made it onto the list it became less popular invalid is still on the list this one is slightly interesting as this is one of the op code that is dominated by one single contract the invalid op codes are almost exclusively used by the usdt contract and they have an assert in a CMA flary they use which uh produces invalid op code if triggered now especially for the least popular op code it may be dis interesting to discuss some op codes that do not appear on the list particularly self-destruct since the Cancun hard for this op code has basically become Obsolete and Let's ignore some special conoces here it now simply transfers all the EO of the account onwards investigating this we were surprised that the op code is still used for in some cases and yeah um even more interestingly we saw that it's still there are still gas tokens deploying new contracts we don't know why this is still done uh we think it's due to Legacy code still being triggered by some contracts let's have a quick look at the new OP codes introduced since last Devcon keep in mind that all contracts deployed before um these op code were introduced um won't use them so all execution you see here must be from Recently deployed contracts push zero is already used for 0.7 of all op code we've seen so this new OP code is already used more than for example call data load or Mo um you might wonder why t- store is used more than T load because in a typical re-entrance pattern you have t- store tload t- store now what about pre- compiles prior to Def conon 6 EC recover was by far the most heavily used precompile it was mainly used by open C cport contract for signature verification keep in in mind that we only have 11 million evm transaction in our data set so 2 million users is quite a lot for sh 2256 the heaviest user was the E2 deposit contract and identity which is mainly used in Viper for memory operations now before Defcon 7 we see way less easy recover operation and the far heaviest users now are permit 2 and warhole contract for sh 2256 the heaviest users are the if2 deposit contract and igen layer and we saw that Viper popularity rose and we see this reflected in the identity op code yeah the least popular op code interesting here it's one single contract that does all the Blake two calls transaction complexity the numbers of call per transaction went down while the memory consump decreased yeah so super quick summary P zero is super hot easy recover is US less and there are more calls per transaction and we've prepared a blog post which is not fully ready yet but will be published soon it will have full description of the data more plots more results thank you thank you thank you Dominic for the short presentation insightful one of course um do you have questions for Dominic oh okay let me [Music] try hi yes yeah um it's interesting to see how create was uh one of the least op codes do you have a reason or Theses as to why that is the case I'm pretty sure because create two became more common and then the statistics are split right right yeah okay cool thanks next question okay um what can we do with all these new findings I think it's uh important for the core deaths and the people discussing the EIP stuff um especially for the op codes um it's really good for them to see right that push to a new addition there is introduced is really used heavily and has an impact and they may can have take look at the statistic and see what isn't used so heavily but of course it's hard to deprecate something right because we still have contracts relying on them thank you last question