# EEST Benchmark Analysis

https://github.com/ethereum/execution-spec-tests/releases/tag/benchmark%40v0.0.7

For each test targetting an opcode/precompile, look at its `opcode_counts` within each test and see which opcodes actually dominate.

Key patterns:
- Simple binary ops (EQ, SUB, MUL, AND, OR, XOR, SHL): dominated by DUP2 + the target opcode in near-equal counts (~59M each at 7 tests), reflecting tight benchmark loops
- Shift-based ops (SAR, SHR): SAR/SHR leads, with SHL as supporting infrastructure
- Call-type ops (CALL, CALLCODE, DELEGATECALL, STATICCALL): heavy MSIZE/MSTORE/PUSH32/XOR scaffolding (~134M) with the call itself at ~23M
- Precompiles (SHA2-256, identity): PUSH0/PUSH1 dominate setup, with STATICCALL as the call mechanism
- CALLDATACOPY: highest total volume test suite (434 tests, 1.7B PUSH1 ops)
- SELFDESTRUCT: uniquely diverse opcode profile with EXTCODESIZE, SUB, ADD in top 10

## REVERT

Example where target opcode is 8th most frequently called

**Number of tests:** 35

### Aggregate Opcode Frequency (all tests combined)

| Rank | Opcode | Total Count |
|------|--------|-------------|
| 1 | PUSH0 | 16,943,210 |
| 2 | PUSH1 | 10,765,996 |
| 3 | POP | 5,647,599 |
| 4 | PUSH20 | 5,647,576 |
| 5 | CALLDATASIZE | 5,647,558 |
| 6 | GAS | 5,647,436 |
| 7 | STATICCALL | 5,647,436 |
| 8 | REVERT | 5,647,298 |
| 9 | PUSH2 | 3,797,104 |
| 10 | CODECOPY | 1,087,140 |
| 11 | JUMPDEST | 7,153 |
| 12 | JUMP | 6,473 |
| 13 | PUSH3 | 651 |
| 14 | JUMPI | 490 |
| 15 | SSTORE | 385 |

example test

<details><summary><code>test_return_revert[fork_Osaka-blockchain_test-1KiB of non-zero data-opcode_REVERT-benchmark-gas-value_100M]</code></summary>

| Opcode | Count |
|--------|-------|
| PUSH1 | 916,473 |
| PUSH0 | 916,140 |
| PUSH2 | 610,744 |
| CALLDATASIZE | 305,376 |
| POP | 305,375 |
| PUSH20 | 305,374 |
| GAS | 305,370 |
| STATICCALL | 305,370 |
| CODECOPY | 305,369 |
| REVERT | 305,364 |
| JUMPDEST | 370 |
| JUMP | 348 |