turbo-persistence: make batch_get callback-oriented
batch_get now reports each key index to a callback, possibly concurrently, instead of returning an ordered Vec. The backend decodes TaskStorage inside the callback, so decoding runs in parallel with the reads for large batches, and writes the results into its restore entries directly. This replaces the separate parallel decode step from #100453 (decode_task_storage_batch, PARALLEL_DECODE_THRESHOLD, batch_lookup_raw and the batch_lookup bench). The batch lookup round-trip test now covers the callback API on both sides of batch_get's parallel cutoff, including a malformed value. <!-- NEXT_JS_LLM --> Co-authored-by:Luke Sandberg <210140+lukesandberg@users.noreply.github.com> Co-authored-by:
Tobias Koppers <1365881+sokra@users.noreply.github.com>
Showing
This diff is collapsed.
This diff is collapsed.
This diff is collapsed.
Please
register
or
sign in
to comment