Frida 17.3.2 Released ∞
release
Fresh bits are ready! This release focuses on squeezing even more performance out of our Fruity backend:
- fruity: Batch datagram delivery to make cross-thread hand-off more deterministic and reduce context-switch overhead.
- ncm: Rework host→device scheduling. We now keep a rolling window of OUT transfers and refill as soon as any URB completes, keeping the bulk pipe busy and bumping HS throughput from ~29 MB/s to ~34 MB/s in our tests.
- ncm: Switch to a fixed-slot NDP layout, turning the O(k²) “shrink until it fits” packer into an O(k) one. On a 256 MiB transfer this drops layout time from ~1.1 s to the noise floor.
Enjoy, and let us know how it works for you!