Cache calls to compute_index in the pipeline emitter.
This avoids repeatedly emitting the same index computation. PiperOrigin-RevId: 971387035
P
Peter Hawkins committed
4a50b9b3af75b5a753411da972885931fc7cc7d7
Parent: d5f2afb
Committed by jax authors <google-ml-automation@google.com>
on 8/26/2026, 6:09:20 PM