Suggested by @Rumata888: lagrange_last in ECCVM is currently set based on dyadic size. It could be much more efficient to set it based on actual trace usage (as we do in Ultra/MegaHonk). This is only viable if it becomes a witness since otherwise the VK depends on transaction stack depth. Would require new relations establishing that ops are "empty" beyond the corresponding "final" idx.
Suggested by @Rumata888:
lagrange_lastin ECCVM is currently set based on dyadic size. It could be much more efficient to set it based on actual trace usage (as we do in Ultra/MegaHonk). This is only viable if it becomes a witness since otherwise the VK depends on transaction stack depth. Would require new relations establishing that ops are "empty" beyond the corresponding "final" idx.