chadbrewbaker

• Easyperf discussion for 12 September 2021 - Random topic group discussion

• JVM: Thomas Dullien's Prodfiler https://prodfiler.com/ also benchmark against unix tools to see how much slower the JVM code is.

• Perf for JVM analysis https://www.brendangregg.com/perf.html#JIT_Symbols

• Parallel pseudorandom number generators https://dl.acm.org/doi/10.1145/237578.237591

• LLVM Machine Code Analyzer https://llvm.org/docs/CommandGuide/llvm-mca.html

• Selecting between GPU and CPU https://karimali.ca/resources/papers/gpucpuselect.pdf

• System Design Interview 2nd Ed. https://www.amazon.com/System-Design-Interview-insiders-Second/dp/B08CMF2CQF/

• Discrete Event Simulation - A First Course https://www.amazon.com/Discrete-Event-Simulation-Lawrence-M-Leemis/dp/0131429175

• COZ profiler https://github.com/plasma-umass/coz

chadbrewbaker

Guest Thomas Dullien

https://twitter.com/halvarflake

https://optimyze.cloud/

• Ofensive security is understanding large scale legacy systems.
• Whole stack analysis required.
• Tool built for Microsoft patch analysis (bindiff) is useful for seeing compiler changes
• Weird machines, exploitability, and provable unexploitability

chadbrewbaker

Guest: @i_bogosavljevic Ivica Bogosavljevic

• Biweekly performance meetup - diverse domains of expertese.
• Performance challange coming in a few week.
• Free online course rollout.
• Ivica specalizes in embeddeed performance consulting at Johny's Software Lab.
• Audio latency, high frequency trading, gaming all latency sensitive.
• Memory bottlenecks are the most common performance issue.
• The process is very agile, a series of experiments with the client.
• Loop tiling, loop sectioning/merging, and loop interchange are useful transformations.

chadbrewbaker

Host: @dendibakh Guest: @i_bogosavljevic Ivica Bogosavljevic

• Biweekly performance meetup - diverse domains of expertese.
• Performance challange coming in a few week.
• Free online course rollout.

chadbrewbaker

Daniel Lemire https://lemire.me/en/

Experimental vs asymptotic software engineering, Ullman's essay

Emery Berger

John Reghr

Denis' Book

chadbrewbaker

# To display the perf.data header info, please use --header/--header-only options.
#
#
# Total Lost Samples: 0
#
# Samples: 35K of event 'cpu-clock:uhpppH'
# Event count (approx.): 357151511580
#
# Children      Self  Command  Shared Object        Symbol                                                                                                                                                                                                                
# ........  ........  .......  ...................  ......................................................................................................................................................................................................................

chadbrewbaker

name: CICD

on: [push, pull_request]

jobs:

  build:
    name: Build
    runs-on: ${{ matrix.job.os }}
    strategy:

chadbrewbaker

# The individual has been paid wages for insured work during the individual’s base period in an amount at least one and one-quarter times the wages paid to the individual during that quarter of the individual’s base period in which the individual’s wages were highest; provided that the individual has been paid wages for insured work totaling at least three and five-tenths percent of the state wide average annual wage for insured work,computed for the preceding calendar year if the individual’s benefit year begins on or after the first full week in July and computed for the second preceding calendar year if the individual’s benefit year begins before the first full week in July, in that calendar quarter in the individual’s base period in which the individual’s wages were highest, and the individual has been paid wages for insured work totaling at least one-half of the amount of wages required under this paragraph in the calendar quarter of the base period in which the individual’s wages were highest, in a cale

chadbrewbaker

Chrome UTF-8 Decoder

NGINX UTF-8 Javascript

Chrome UTF-8 DFA

Rust UTF-8 Decoder/validator

MUSL mbrtowc

chadbrewbaker

US-East-1 on 12-1-2020

Invocation through web console clicking Test button.

Python after cold start seems to stay around 1ms. Nodejs seems to bounce back to 33 ms cold restart.

The kernel is 4.14.203-112.332.amzn2.x86_64.

Sadly no /dev/shm or /dev/zram. No swapon or mkswap commands to set swapspace.