Links tagged “benchmark”

arxiv.org

published scores almost never say which inference engine produced them.

arxiv.org

writing more rules does not make an agent follow them.

metr.org

An external evaluation of autonomous software capability, complicated by the model attempting to cheat the tasks, and finding no catastrophic misalignment.

llm-fighter.com

A versus fighting game in which models play each other, as a way of comparing them.

itnext.io

Throughput, resource cost and feature comparison of Kubernetes CNI plugins on a 10Gbit/s bare-metal cluster, re-run on Kubernetes 1.14 in April 2019.

pawelmhm.github.io

Benchmarks python-aiohttp making up to a million HTTP requests, with asyncio client code and a comparison against synchronous alternatives.

peterbe.com

Benchmarks several ways to remove duplicates from a Python list, and finds list(dict.fromkeys(seq)) fastest when insertion order must be preserved.

techspot.zzzeek.org

Benchmarked argument that asyncio slows database-bound Python code down, because interpreter overhead — not database I/O — dominates typical CRUD work.

baremetrics.com

Live benchmark dashboard of SaaS metrics — churn, revenue per user, failed payments, pricing — aggregated from hundreds of Baremetrics customers.

users.softlab.ece.ntua.gr

A Python implementation of the Ulam prime spiral, and a benchmark where running the same grid generation under PyPy comes out around fifty times faster than CPython.

nbonvin.wordpress.com

Benchmark of Apache (worker and event), Nginx, Varnish, G-WAN and Cherokee serving a 100-byte static file, followed by a comments thread disputing the methodology.

code.google.com

Project page for Multi-Mechanize, a Python framework that replays python-mechanize scripts across processes and threads to generate load against a site or web service.

tsung.erlang-projects.org

Home page of Tsung, an open-source distributed load-testing tool written in Erlang that stresses HTTP, WebDAV, SOAP, database, LDAP and XMPP servers.