You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The current sigverify limits incoming batches to 5k packets. This constrains things to ~125k TPS (~40ms per iteration) on typical HW. Current sigverify implementation can be scaled to ~200k TPS just by increasing max batch size.
Port AVX-512 implementation from Firedancer, should be ~2x performance increase per core through sigverify
We can also avoid doing some amount of sigverify by feeding some minimum PF from banking stage → QUIC fetch and filtering out transactions of low value before wasting time sigverifying and filtering them. We need the txview #2255 changes to be able to quickly and efficiently parse transactions and extract the PF.
The current sigverify limits incoming batches to 5k packets. This constrains things to ~125k TPS (~40ms per iteration) on typical HW. Current sigverify implementation can be scaled to ~200k TPS just by increasing max batch size.
Port AVX-512 implementation from Firedancer, should be ~2x performance increase per core through sigverify
We can also avoid doing some amount of sigverify by feeding some minimum PF from banking stage → QUIC fetch and filtering out transactions of low value before wasting time sigverifying and filtering them. We need the txview #2255 changes to be able to quickly and efficiently parse transactions and extract the PF.
The text was updated successfully, but these errors were encountered: