Memory Model

The EVM uses a linear memory (i.e. starting at zero and growing), without a defined limit. Allocating memory costs a quadratic amount of gas, which severely limits the amount of memory usable in a single context (i.e. contract call).

In addition, the EVM has a separate memory space per context. Contexts may communicate with each other my copying data to call data and return data buffers.

The FuelVM uses a single share memory block per transaction. Memory is allocated statically with a known upper bound, allowing for straightforward implementation of heap types such as vectors. Memory in the FuelVM is globally readable across contexts, but locally writable. Each context may only write to portions of the stack and the heap which it has ownership over.