Yossi Kreinin - blog archive

• Unit tests mark territory more than squash bugs
• Doing everyone else's job
• LLMs struggle with the shell, too
• All means are fair except solving the problem
• "Enabling" C threads in a Python / Wasm environment
• LLMs aren't world models
• 0+0 > 0: C++ thread-local storage performance
• Profiling in production with function call traces
• Advantages of incompetent management
• Profiling with Ctrl-C
• A 100x speedup with unsafe Python
• The state of AI for hand-drawn animation inbetweening
• Managers have no human rights
• refix: fast, debuggable, reproducible builds
• Don't ask if a monorepo is good for you – ask if you're good enough for a monorepo
• Patents: how and why to get them
• Things want to work, not punish errors
• Fun won't get it done
• The habitat of hardware bugs
• Evil tip: avoid "easy" things
• Love thy coworker; thy work, not necessarily
• The overblown frequency vs cost efficiency trade-off
• People can read their manager's mind
• Stock options: a balanced approach
• A manager's pitfall: striving to "add value"
• The C++ FQA is on GitHub
• Fun with UB in C: returning uninitialized floats
• Accidentally quadratic: rm -rf??
• "Information asymmetry" cuts both ways
• A Sokoban levels design programming contest
• Capital vs labor: who risks more?
• Do company names actually matter?
• A better future
• Things from Python I'd miss in Go
• How to make a heap profiler
• Why bad scientific code beats code following "best practices"
• Working simultaneously vs waiting simultaneously
• Can your static type system handle linear algebra?
• A simple way to "get more people to code"
• Very funny, gdb. Ve-ery funny.
• Delayed printf for real-time logging
• Coroutines in one page of C
• Do call yourself a programmer, and other career advice
• How FPGAs work, and why you'll buy one
• The bright side of dark silicon
• Parallelism and concurrency need different tools
• 10x more selective
• checkedthreads: bug-free shared memory parallelism
• Amdahl's law in reverse: the wimpy core advantage
• Is program speed less important than X?
• Efficiency is fundamentally at odds with elegance
• How profilers lie: the cases of gprof and KCachegrind
• It's "locking" if it's blocking
• C++ template fuckwittery
• Why custom allocators/pools are hard
• "Value", the irksome euphemism
• Will OpenCL help displace GPGPU? Parallella, P2012, ...
• Do you really want to be making this much money when you're 50?
• C as an intermediate language
• Error codes vs exceptions: critical code vs typical code
• Aren't side effects fundamental in complexity analysis?
• What "Worse is Better vs The Right Thing" is really about
• "It's done in hardware so it's cheap"
• Work on unimportant problems
• Hardware macroarchitecture vs microarchitecture
• Which of those would you like me to write?
• Passing shell script arguments to a subprocess
• Why programming isn't for everyone
• Compensation, rationality and the project/person fit
• Cycles, memory, fuel and parking
• Could SOPA give us back a decentralized Internet?
• Coding standards: is consistency prettier than freedom?
• Graham & Coase: when big companies are a good idea
• Engineers vs managers: economics vs business
• SIMD < SIMT < SMT: parallelism in NVIDIA GPUs
• An unusual hardware architecture: APA (Associative Processing Array)
• Machine code monkey patching
• Making data races manifest themselves
• The Iron Fist Coding Standard
• My history with Forth & stack machines
• The Internet age/reputation paradox
• If a tree falls in a forest, it kills Schrödinger's cat
• Applied mathematics in business consulting
• Lack of wealth through lack of empathy
• API users & API wrappers
• Digital asses in the computing industry
• The Virtue of a Manager
• Getting the call stack without a frame pointer
• What makes cover-up preferable to error handling
• The C++ Sucks Series: petrifying functions
• Coding standards: having more errors in code than code
• The nomadic programmer
• Humans and compilers need each other: the VLIW SIMD case
• Pearls of wisdom
• The C++ Sucks Series: the quest for the entry point
• The internal free market
• Consistency: how to defeat the purpose of IEEE floating point
• I want a struct linker
• The cardinal programming jokes
• I love globals, or Google Core Dump
• Extreme Programming Explained
• OO C is passable
• Redundancy vs dependencies: which is worse?
• I can't believe I'm praising Tcl
• Python: teaching kids and biting bits don't mix
• Side effects or not, aliasing kills you
• Optimal processor size
• IHateCamelCase
• Code, data and interactive programming
• The Algorithmic Virtual Machine
• "High-level CPU": follow-up
• The "high-level CPU" challenge
• Everybody agrees with yosefk
• Fun at the Turing tar pit
• Interrupt? Let the Bastard handle it!
• Why don't we have a word for it?
• Teeth marks at the rear end
• AI problems
• A writer of the lame kind
• Low-level is easy
• Blogging is hard