journal 2026-05-05

P86 8-entry TLB shell perf

P86 doubled the unified TLB from four entries to eight and reran the P84 BusyBox shell workload.

Result:

metricP84P86delta
post-load cycles239,533,716223,777,049-6.58%
CPI2.66152.5615-3.76%
fetch walks2,263,0381,117,037-50.64%
load walks2,267,672973,288-57.08%
store walks601,266199,592-66.80%
memory handshakes39,642,30133,111,189-16.48%

Shell phase notes: ls improved 7.68%, cat improved 44.69%, and the 8x ash loop improved 29.02%. The /tmp file phase got 20.44% slower in this run, so the next round should avoid over-reading one aggregate number.

Status: PASS for RTL simulation and shell benchmark. LibreLane hardening is NOT RUN.