refactor(vm): optimize opcode energy cost calculation with BigInteger#6613
Conversation
Use BigInteger for memory size computation to improve precision and add unit tests. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
| // witness array zero, amount array non-zero | ||
| Program program = mockProgram(0, 0, 64, 1, 0); | ||
| long cost = EnergyCost.getVoteWitnessCost3(program); | ||
| // witnessMemNeeded = 0 (size is zero) |
There was a problem hiding this comment.
The unit tests in VoteWitnessCost3Test are well-structured — especially appreciate the overflow boundary coverage in testLargeArrayLengthOverflow and testLargeOffsetOverflow, and the consistency check against cost2 in testConsistentWithCost2ForSmallValues. One minor note: the comment in testZeroLengthOneArray states witnessMemNeeded = 0, but cost3 computes size = 0 * 32 + 32 = 32, which is non-zero — worth clarifying if this is an intentional behavior change from cost2.
There was a problem hiding this comment.
Good catch! The comment was indeed inaccurate — in cost3, even when arrayLength is 0, the size is computed as 0 * 32 + 32 = 32 (accounting for the dynamic array length prefix), so witnessMemNeeded is 32, not 0. This is an intentional behavior difference from cost2, where a zero-length DataWord after mul stays zero and memNeeded returns 0. Fixed the comment in the latest push. Thanks for the review!
kuny0707
requested review from
CodeNinjaEvan and
Federico2014
and removed request for
CodeNinjaEvan
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
5 participants
Summary
BigIntegerinstead ofDataWordfor memory size computation in energy cost calculation to improve precisionallowTvmOsakaproposal config inVMConfigfor future activation control