Fuel, fees and deposits
Fuel
Every operation costs fuel. A transaction reserves a maximum (max_fuel); if execution needs more, the call fails and everything is reverted. Prices are calibrated so that about 20 ns of CPU on a 2-vCPU server correspond to one unit of fuel, which keeps a full block (50 000 000 fuel on mainnet by default) at about one second in the worst case.
| Operation | Fuel |
|---|---|
| Statement · expression | 2 · 1 |
| Per 32 bytes of values read, copied or combined | 1 |
| Function call | 20 |
| Storage read | 250 |
| Storage write or delete | 400 + 4 per byte of key and value |
send · emit · destroy | 300 · 100 + 1 per byte · 1 000 |
sha256, blake3 | 60 + 20 per 64 bytes |
verify_ed25519 | 3 500 + 1 per 64 bytes |
ring_verify | 5 000 + 10 000 per key |
| Deploy: compiling | 5 per byte of source |
| Invoke: loading the contract | 100 + 1 per 100 bytes of compiled code |
Changes in language version 2
| Operation | Version 1 | Version 2 |
|---|---|---|
| Copying a value (reading a local, a constant) | 1 per 32 bytes | 1 per 32 bytes + 4 per heap allocation (each text, bytes or list inside) |
| Decoding a value read from storage | included in the read | + 1 per 32 bytes + 6 per allocation |
xs[i] and len(xs) on a local list | copies the whole list | reads without copying |
| Call to another contract | — | 700 + 1 per 32 bytes of arguments + loading (100 + 1 per 100 bytes of code) |
mul_div, isqrt, pow | — | 30 |
code_hash, is_contract, is_final | — | 250 |
| Enum transition check | — | 1 per checked level |
Version 1 sources compiled as version 2 produce the same results, storage and events; their fuel can differ only where they copy lists or large values. The reasons are explained in Security.
Fees
The minimum fee of a transaction is
fee = (base_fee + ⌈bytes × fee_per_kb ÷ 1000⌉ + ⌈max_fuel × fee_per_kfuel ÷ 1000⌉) × congestionwith The Coin's default parameters base_fee = 1 000, fee_per_kb = 10 000 and fee_per_kfuel = 1 000 motes. The congestion multiplier starts at 1× and adjusts with block usage; the part above 1× is burned. Wallet priorities low | normal | high | urgent pay 1×, 1.25×, ≥ 2× and ≥ 4× the minimum. Parameters are set by governance, so check the network's current values before relying on exact numbers.
- You pay for the fuel you reserve. The wallet reserves the measured fuel × 1.3 + 5 000.
- A failed transaction pays its fee; all of its effects are reverted.
- Views called through the API are free and do not create transactions.
tccl run and the playground print an estimate with the default parameters, for example:
estimated on-chain cost (default parameters): fee 9247 motes for max_fuel 6497 and ~175 bytes · storage deposit +0 motesStorage deposits
Contract state is backed by a refundable deposit:
- The size of a contract is its compiled code plus every storage entry (key + value).
- Required deposit = ⌈size ÷ 1 000⌉ ×
storage_deposit_per_kb(default 100 000 motes, 0.001 TCN per started kB). - A deploy or call that makes the state grow pays the missing deposit, up to the transaction's
max_deposit(wallet default 1 TCN). - A call that makes the state shrink receives
deposit × freed ÷ old sizeback. Storing a default value deletes an entry, so clearing data is rewarded. destroy(to)pays the whole deposit and the balance toto.
The simulator shows deposits as estimates and does not deduct them from the fictitious balances.
Measured performance
TCCL is a tree-walking interpreter written in Rust. Rust gives memory safety and predictable performance; it does not run contracts at native speed. What matters for the network is that fuel follows CPU time.
Measured with cargo test -p tccl --release --test perf -- --ignored --nocapture on an Intel Xeon E5-1620 v2 @ 3.70 GHz (the simulator's in-memory storage is faster than a node's disk, so storage rows are optimistic):
| Workload | Language | ns per fuel | Peak heap | 50 M-fuel block |
|---|---|---|---|---|
| Arithmetic loop | 1 · 2 | 16.4 · 17.6 | < 0.1 MB | 0.8 s · 0.9 s |
| Map writes | 1 · 2 | 1.4 · 1.2 | 0.4 MB | 0.1 s |
| BLAKE3 chain | 1 · 2 | 5.0 · 4.7 | < 0.1 MB | 0.2 s |
| Building records | 2 | 13.8 | < 0.1 MB | 0.7 s |
| Copying a list of 4 000 texts (adversarial) | 1 · 2 | 2 109 · 12.2 | 0.3 MB | 105 s · 0.6 s |
| Decoding a stored list (adversarial) | 1 · 2 | 1 656 · 21.1 | 0.4 MB | 83 s · 1.1 s |
| Holding large local lists (adversarial) | 1 · 2 | 10.1 · 6.4 | 65.6 MB · 16.8 MB (limit) | 0.5 s · 0.3 s |
Your numbers will differ; run tccl bench for a quick check on your machine.