bench: RESULTS with a table for each machine, HTTPS with streams and envoy

Every table is now for one machine, ARM or Intel, with no averages of
the two, the copying proxies and those using splice() in one table:
bold the best of each group and * the best of the row for the HTTP proxy
and the relay, only the best of the row for CONNECT and SOCKS. The
tables for 13 October replace those of 11-12 October, whose columns for
splice by direction and 16 KB buffers are gone with them.

The HTTPS section is measured again with today's devel, with the TLS
streams and envoy. report.py takes the TLS streams and 1 MB answers for
bulk data.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Vladimir DubrovinandClaude Opus 5.5 committed 2026-10-13 13:00:21 +03:00
1 parent 18d3ff101a
commit ef0c98e4c1
4 files changed
+762 -463

No files matched your search

+4 -3
View File
@@ -36,7 +36,8 @@ import csv
import statistics
STREAM = ("stream1", "stream1-down", "stream16", "connect-stream1-down", "connect-stream16",
"socks-stream1-down", "socks-stream16", "relay-stream1-down", "relay-stream16")
"socks-stream1-down", "socks-stream16", "relay-stream1-down", "relay-stream16",
"tls-term-stream1-down", "tls-term-stream16", "tls-init-stream1-down", "tls-init-stream16")
# connect-ka-1m etc. are bulk too when present
@@ -117,7 +118,7 @@ def load(paths):
# In the combined table, the bulk scenarios in Gbit/s, the others in
# thousands of requests per second.
BULK = STREAM + ("http-ka-1m",)
BULK = STREAM + ("http-ka-1m", "tls-term-ka-1m", "tls-init-ka-1m")
def combined(machines, labels, splice, prefix=None, split=False, overall_only=False):
@@ -288,7 +289,7 @@ def main():
sel = by.get((s, p), [])
if p == "direct":
return "–"
if s in STREAM or s in ("http-ka-1m", "http-ka-50k-2000", "http-ka-2k-50k-2000"):
if s in BULK or s in ("http-ka-50k-2000", "http-ka-2k-50k-2000"):
return fmt(median(sel, "cpu_s_per_gb"), 2) + " s/GB"
return fmt(median(sel, "cpu_us_per_req"), 0) + " µs/req"