#!/bin/sh # OB-1 stage 2, step 3: the leg's full run matrix. # # res-* fully resident on this leg's CPU configuration -- the byte-identity # reference the leased runs are compared against (stage 1's # committed baseline is a different backend; see run-ob1.sh header) # lease-* leased, K in {16, 8, 4}, both acceptance corpora # *-a/*-b an A/A pair (same run twice) to show the run itself is stable # *-fadv the same K=16 prose run with the OS page cache dropped for every # leased range, so the cold tier is genuinely cold NVMe set -e SH=/mnt/f/f32/openbob-wt/research-2/research/ob1/run-ob1.sh AC_PROSE=/mnt/f/f32/stage/research/ob1/AC-PROSE.txt AC_CODE=/mnt/f/f32/stage/research/ob1/AC-CODE.txt sh "$SH" res-prose-a "$AC_PROSE" 0 0 sh "$SH" res-prose-b "$AC_PROSE" 0 0 sh "$SH" res-code-a "$AC_CODE" 0 0 sh "$SH" lease-k16-prose-a "$AC_PROSE" 16 0 sh "$SH" lease-k16-prose-b "$AC_PROSE" 16 0 sh "$SH" lease-k16-code "$AC_CODE" 16 0 sh "$SH" lease-k8-prose "$AC_PROSE" 8 0 sh "$SH" lease-k8-code "$AC_CODE" 8 0 sh "$SH" lease-k4-prose "$AC_PROSE" 4 0 sh "$SH" lease-k4-code "$AC_CODE" 4 0 sh "$SH" lease-k16-prose-fadv "$AC_PROSE" 16 1