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
There is no gas bump feature that could control that bump the block gas limit. We only need to support this in perf-devnet-3, as block gas limit in Jochemnet starts at 1G. We also need to investigate into the recent proposal that gives CL full control over the block gas limit, instead of EL. => PR feat(test-fill): align fill-stateful with gas-benchmarks implementation #2923
Batch PRC call for _fetch_receipts helper in ClientBackend object, this was significantly blocking gas-benchmarks payload generation in the past, for test_sstore_invariatns only, it took ~2 hr to complete.
There is currently a max batch transaction size, which is 750 txs. When benchmark generate transactions more than this amount (e.g., ether transfer, contract creation benchmark), the framework would build multiple block with max batch tx size, instead of building one single block.
Post verification being ignored, but we still need such mechanism for benchmarking (e.g., total gas used comparison, final receipt check and more.).
Our current approach use debug_setHead at the end of the benchmark, so that the next benchmark begins at start_block (snapshot_block + some global initialization). Based on Jochem's comment, we need to investigate whether it works as expected. In gas-benchmarks, we use FCU to control the rollback. => PR feat(test-fill): align fill-stateful with gas-benchmarks implementation #2923
Add necessary metadata information for fixture, including the snapshot block, gas bump block number, sender pool size and more.
Tracking suggested features and refactor for PR #2637 .
fill-statefulwithgas-benchmarksimplementation #2923fill-statefulwithgas-benchmarksimplementation #2923_fetch_receiptshelper in ClientBackend object, this was significantly blocking gas-benchmarks payload generation in the past, fortest_sstore_invariatnsonly, it took ~2 hr to complete.debug_setHeadat the end of the benchmark, so that the next benchmark begins atstart_block(snapshot_block+ some global initialization). Based on Jochem's comment, we need to investigate whether it works as expected. In gas-benchmarks, we use FCU to control the rollback. => PR feat(test-fill): alignfill-statefulwithgas-benchmarksimplementation #2923