flamenco, runtime: implement static_instruction_limit

[topointon-jump]

Implement solana-foundation/solana-improvement-documents#160

• Implementation
• Unit tests
• Ledger

Also drive-by fix to make fd_executor_verify_transaction take a const bank.

[github-actions]

Performance Measurements ⏳

Suite	Baseline	New	Change
backtest mainnet-368528500-perf per slot	0.050832 s	0.050828 s	-0.008% ✅
backtest mainnet-368528500-perf snapshot load	1.641 s	1.639 s	-0.122% ✅
backtest mainnet-368528500-perf total elapsed	50.832055 s	50.828497 s	-0.007% ✅
firedancer mem usage with mainnet.toml	1005.23 GiB	1005.23 GiB	0.000% ✅

[github-actions]

Performance Measurements ⏳

Suite	Baseline	New	Change
backtest mainnet-368528500-perf per slot	0.050994 s	0.051115 s	0.237% ✅
backtest mainnet-368528500-perf snapshot load	1.625 s	1.635 s	0.615% ✅
backtest mainnet-368528500-perf total elapsed	50.994451 s	51.115122 s	0.237% ✅
firedancer mem usage with mainnet.toml	1005.23 GiB	1005.23 GiB	0.000% ✅

[github-actions]

Performance Measurements ⏳

Suite	Baseline	New	Change
backtest mainnet-368528500-perf per slot	0.050902 s	0.050771 s	-0.257% ✅
backtest mainnet-368528500-perf snapshot load	1.639 s	1.628 s	-0.671% ✅
backtest mainnet-368528500-perf total elapsed	50.901701 s	50.770682 s	-0.257% ✅
firedancer mem usage with mainnet.toml	1005.23 GiB	1005.23 GiB	0.000% ✅

[github-actions]

Performance Measurements ⏳

Suite	Baseline	New	Change
backtest mainnet-368528500-perf per slot	0.050535 s	0.050789 s	0.503% ✅
backtest mainnet-368528500-perf snapshot load	1.637 s	1.633 s	-0.244% ✅
backtest mainnet-368528500-perf total elapsed	50.534575 s	50.789403 s	0.504% ✅
firedancer mem usage with mainnet.toml	1005.23 GiB	1005.23 GiB	0.000% ✅

[topointon-jump]

This ledger passes through an epoch boundary in which static_instruction_limit is activated, and then tries to execute a transaction with 64 instructions which succeeds. Note that the script to create this ledger also sent transactions with 65 instructions, but because these were filtered out by Agave's sanitisation code (which is what this feature changes) they don't appear in the ledger.

[github-actions]

Performance Measurements ⏳

Suite	Baseline	New	Change
backtest mainnet-368528500-perf per slot	0.050698 s	0.050597 s	-0.199% ✅
backtest mainnet-368528500-perf snapshot load	1.636 s	1.611 s	-1.528% ✅
backtest mainnet-368528500-perf total elapsed	50.697835 s	50.59691 s	-0.199% ✅
firedancer mem usage with mainnet.toml	1005.23 GiB	1005.23 GiB	0.000% ✅

[github-actions]

Performance Measurements ⏳

Suite	Baseline	New	Change
backtest mainnet-368528500-perf per slot	0.051036 s	0.050746 s	-0.568% ✅
backtest mainnet-368528500-perf snapshot load	1.698 s	1.678 s	-1.178% ✅
backtest mainnet-368528500-perf total elapsed	51.035906 s	50.746093 s	-0.568% ✅
firedancer mem usage with mainnet.toml	1005.23 GiB	1005.23 GiB	0.000% ✅

[github-actions]

Performance Measurements ⏳

Suite	Baseline	New	Change
backtest mainnet-368528500-perf per slot	0.051478 s	0.050634 s	-1.640% ✅
backtest mainnet-368528500-perf snapshot load	1.749 s	1.661 s	-5.031% ✅
backtest mainnet-368528500-perf total elapsed	51.477847 s	50.633591 s	-1.640% ✅
firedancer mem usage with mainnet.toml	1005.23 GiB	1005.23 GiB	0.000% ✅

[ibhatt-jumptrading]

this doesn't fully cover the case where a transaction is worst case sized for > 64 instructions because it will violate limits in the transaction parser (fd_txn_parse). The correct thing to do here imo is to change txn parse to always fail if there are >64 top level instructions.

[topointon-jump]

^Discussed offline, will pr in a cleanup pr in after the feature flag is activated everywhere (~3 weeks)