Count instructions where timing cannot tell 3% apart (110)

016's second stage passed its criterion: six interleaved pairs against
ee21aec gave TPC-C (RAM, W=1) -0.6% at the median and the bank's
queries 0.0% to +2.2%, no run interfered. The ranks query was slower
all six times, and chasing it showed that on this Pi timing cannot
separate 1-3%: two servers asked in turns disagree with themselves
(ee21aec to e02bfc9 +1.6%, e02bfc9 to 481642f -1.2%, ee21aec to 481642f
+2.8%), and a query whose instructions grew 0.6% ran 3.5% faster.

Instructions do not move so: pcount.c counts the server's user-mode
instructions with perf_event_open (no perf here), and psample.c samples
where they go. ranks had grown 0.4% in instructions (the rest was where
code and data landed), per_branch 3.6% and tables 20 instructions a row,
which 07047c3 halved. The six pairs of 07047c3 were interfered (another
session's build, load 4-12) and are kept apart, unused.

1 files changed, 13 insertions, 0 deletions

diff --git a/research/experiments/025-i64-cost/ab-sessions.sh b/research/experiments/025-i64-cost/ab-sessions.sh
new file mode 100755
index 0000000..b4b5b1d
--- /dev/null
+++ b/research/experiments/025-i64-cost/ab-sessions.sh
@@ -0,0 +1,13 @@
+#!/bin/zsh
+# ab-sessions.sh DIR A B SESSIONS: ab.py on fresh processes each session,
+# pure-serve-A and pure-serve-B from DIR, which starts first alternating;
+# each session waits for a load average below 2.
+here=${0:A:h}
+dir=$1 a=$2 b=$3 n=$4
+for i in $(seq 1 $n); do
+  until [ "$(cut -d' ' -f1 /proc/loadavg | cut -d. -f1)" -lt 2 ]; do sleep 5; done
+  if (( i % 2 )); then first=$a second=$b; else first=$b second=$a; fi
+  echo "== session $i first=$first $(cut -d' ' -f1-3 /proc/loadavg)"
+  python3 $here/ab.py $dir/pure-serve-$first $dir/pure-serve-$second 6 30 2>&1
+done
+echo "== end"