A gas-efficient Solidity compiler for @ethereum, powered by LLVM.
Ethereum
Joined May 2025
- Tweets72
- Following9
- Followers527
- Likes61
- Compilation times: already faster
- Gas efficiency: ~13% on average, more to come
- Stack-too-deep: fixed
Job is not finished
Maybe unpopular opinion, I think Solidity as-is is mostly fine, and the highest priority thing to fix is _not_ new syntax / language features but the complier:
- make compilation 100x faster
- make gas optimization 100x better
- solving stack too deep
- making the compiler easier to extend & experiment so that EVM devs can propose new EIPs that improve the EVM
All for generics & better syntax, just doesn't feel like high priority. I fear a python2 to python3 moment.
What’s missing now is tooling.
solx already works with @HardhatHQ and Foundry, but @etherscan don’t support solx yet.
Once infra catches up, “stack too deep” will finally be history.
That’s the key difference:
In solc, “stack too deep” remains a compiler limitation.
In solx, it’s solved at compile time — the only challenge left is ecosystem support.
That retry sounds expensive, but solx parallelizes tasks from the IR stage onward.
Even with the extra pass, it’s still ~5× faster than solc –via-ir.
solc, by contrast, can’t parallelize — every stage runs sequentially, so doubling the passes would double compile time.
solx takes a different route:
1️⃣ It tries stackification first.
2️⃣ If that fails, it measures the minimal spill needed.
3️⃣ It recompiles, moves exactly those vars, and succeeds.
So compilation always completes.
solc --via-ir fixes stack overflows in some cases — when stackification doesn’t fail.
It reduces stack pressure by spilling variables to memory early, but if that still doesn’t work, compilation fails.
teapaw.org/Designer_Misbah/status…
Stack too deep is still half-solved I think. Some projects are so complex that adjusting your local variables is not enough due to dependencies.
solx compiler does compile that kind of code. which is great. But how do you get the code verified on Etherscan since they don't support solx yet?
@etherscan feature request. Will you start supporting solx soon?
Solx retweeted
⚡️Smaller, faster, smarter.
solx 0.1.2 by @solx_compiler, cuts contract size, lowers deployment costs, and now integrates seamlessly with Hardhat v3. Big win for devs.
🤝 Tooling
Integration with @HardhatHQ v3 is now first-class.
Developers can point Hardhat directly to solx, and the compiler outputs all required artifacts.
This makes solx easier to adopt in existing project workflows.
🧪 Test coverage extended
The suite now includes 24 real-world projects (Uniswap V4, Aave V3, Solady, Beefy, Lens, and more) and over 2,700 contracts.
Every PR is validated against this set; regressions are not accepted.
🧹 Codegen improvements
Dead Store Elimination removes memory/storage writes that can never be observed.
Conditional jumps compile into fewer, simpler instructions.
Together, these reduce bytecode size and runtime overhead.
📦 Automatic 24 kB handling
If solx detects your contract will exceed the 24 kB cap, it progressively swaps gas-optimal code for size-optimal code—starting outside loops and moving inward—stopping as soon as it fits.
📉 Binary size
- Median contract size: 13% smaller vs 0.1.1
- ~22% smaller than solc 0.8.30 –via-ir
- Across 2,700+ real contracts, solx produces the smaller binary in most cases—and when it does, the margin is significant.
Full details 👉 zksync.mirror.xyz/rt7yjujPtQ…
solx 0.1.0, 0.1.1, 0.1.2 and solc 0.8.30 are now available to try online: godbolt.org/z/Kv64jqK7T
If you’re comfortable reading EVM assembly, explore compiler efficiency and spot optimizations. Notice something missing in solx? Report it github.com/matter-labs/solx/…
👉 Try the demo:
github.com/abinavpp/solx-mli…
And stay tuned — this is just the beginning.