Framework Analysis

This section evaluates how the internal stages of the benchmarking framework—specifically mapping, routing, and translation—contribute to the overall overhead and effectiveness of QEC codes.

1. Mapping and Routing Overhead

We show that mapping and routing stages add 134.095% more two-qubit gates on average due to SWAP overhead. In practical applications, this corresponds to just over four extra two-qubit gates for every original one, which greatly amplifies two-qubit errors.

Mapping and Routing Overhead

Figure 1: Comparison of gate overhead introduced by different mapping and routing strategies.

Takeaway #6: Mapping and routing add significant gate overhead, highlighting the need for strategies tailored specifically to QEC codes to preserve fault tolerance.

2. Translation Overhead

Our investigation shows that even the most optimized translation into native gate sets introduces significant overhead, with an average of 3.166 additional two-qubit gates per original gate.

Translation Overhead

Figure 2: Gate count increase across various translation SDKs (Qiskit, TKET, BQSKit).

Takeaway #7: Compared to unoptimized circuits, careful compilation and translation are crucial to minimize overhead and limit error accumulation.