WIP core perf optimizations#5236

Draft

dmkozh wants to merge 103 commits intostellar:masterfrom

dmkozh:budget_test2

Commits on Apr 10, 2026

Commits on Apr 13, 2026

Commits on Apr 15, 2026

add tracy support to bench matrix
dmkozh
committed
Switch SHA256 from libsodium (pure C) to OpenSSL (SHA-NI hardware accel)

authored and
dmkozh
committed
openssl SHA256 -~20ms for buckets? faster then the previous baseline
dmkozh
committed
Parallelize InMemoryIndex construction with bucket put loop (saves ~25ms/ledger)

authored and
dmkozh
committed
Bench for parallel in-memory index - ~-25ms
dmkozh
committed
perf: overlap per-thread commit with parallel execution (+13.6% TPS)

Garand Tyson
authored and
dmkozh
committed
overlap commit with execution bench - seems like regression
dmkozh
committed
Revert "perf: overlap per-thread commit with parallel execution (+13.6% TPS)"
dmkozh
committed
Benchmark to confirm revert
dmkozh
committed
perf: eliminate per-tx child LTX in fee processing (+19.2% TPS)

Garand Tyson
authored and
dmkozh
committed
Bench for no child ltx. over several runs seems like it's neutral to ~-10ms. The change is kind of sketchy, but there may be something to it's concept.
dmkozh
committed
perf: eliminate child LTX in refundSorobanFee

authored and
dmkozh
committed
perf: skip child LTX in removeAccountSigner via peek

authored and
dmkozh
committed
bench for more child LTX skips - seemingly no improvement
dmkozh
committed
Revert "perf: eliminate child LTX in refundSorobanFee"
dmkozh
committed
Revert "perf: skip child LTX in removeAccountSigner via peek"
dmkozh
committed
Revert "Use createWithoutLoading/updateWithoutLoading in parallel commit"
dmkozh
committed
perf: track entry existence in ParallelApplyEntry to skip SHA256 lookups

Garand Tyson
authored and
dmkozh
committed
perf: move semantics in commitChangesToLedgerTxn to avoid XDR copies

authored and
dmkozh
committed

Commits on Apr 16, 2026

Commits on Apr 17, 2026

Commits on Apr 20, 2026

Commits on Apr 21, 2026