Numeric Data Types in the Wild; or: Let’s Look at Last Night’s Dinner Receipt!
summary
A shipped POS printout shows specific failures: a brisket line prints an absurdly precise float — effectively 1200× precision beyond any measurable reality — side dishes sometimes render exact rationals but then suppress “1” so totals break coherence, and a fractional share appears as one slash seven sometimes and as 0.428571... other times. The product works as a receipt printer but produces incorrect, illegible, and incoherent numeric results caused by mixed numeric types and ad hoc control logic.
AI-written summary · about
related talks
02mWhen Computers Can't Math - Floating Point Rounding Error Explained
51mFloating-Point Arithmetic on Deterministic Systems
56mHow to Quickly Convert Floating-Point Numbers in C++
48mFuture-Proofing Numerical Accuracy with Formal Specifications
38mStrength in Numbers: Unums and the Quest for Reliable Arithmetic
01h 00mFloating-Point ＜charconv＞: Making Your Code 10x Faster With C++17's Final Boss
22mCrossing the chasm between quality and performance
08mYou Won’t Believe This One Weird CPU Instruction!