13 stories tagged floating-point, newest first.
102 points by pwg | 33 comments
Ken Shirriff blog post examining the Intel 8087 coprocessor's microcode and circuitry for floating-point addition and subtraction, including special cases.
39 points by dverlaeckt80 | 6 comments
Blog post on using superoptimization to find an optimal instruction sequence for a floating-point emulator function returning a float's lowest set bit.
243 points by pngmangi | 21 comments
Blog post examining the x86 FMA instruction vfmaddrnd231pd, which performs fused multiply-add with per-instruction rounding control.
47 points by zX41ZdbW | 5 comments
Blog post on purplesyringa explaining how to implement vectorized count-leading-zeros and count-trailing-zeros using floating-point tricks.
64 points by signa11 | 82 comments
Blog post explaining why standard C offers no portable way to check for errors in string-to-float conversion functions like strtod, following up on math_errhandling.
58 points by ibobev | 12 comments
Blog post arguing that float arithmetic should be handled by reacting to results (NaN/inf) rather than the preventive checks used for integer overflow in C and Rust.
52 points by ibobev | 30 comments
A blog post exploring when adding decimal numbers as IEEE double-precision floats produces rounding errors and when the results come out exact.
98 points by pwg | 11 comments
Ken Shirriff's blog post reverse-engineering the Intel 8087 coprocessor's FPTAN tangent instruction, which combines CORDIC with polynomial approximation.
409 points by rafaelcosta | 194 comments
Blog post describing how an iPhone 16 Pro Max produces garbage output when running LLMs with Apple's MLX framework, and investigating the cause.
374 points by SomaticPirate | 98 comments
An interactive web tool that shows the bit pattern, sign, exponent, and significand of half, bfloat, float, and double floating-point values.
354 points by Tomte | 138 comments
Ken Shirriff's blog post reverse-engineering the Pentium's ×3 multiplier circuit, used in its base-8 floating-point multiplication, and explaining how it works.
1683 points by pie_flavor | 407 comments
Blog post retelling a tweet thread on why building an accurate calculator app is hard, covering precision issues like (10^100)+1-(10^100) returning 0.
362 points by gslin | 110 comments
Ken Shirriff's blog post examining the Pentium die to show where the 1994 floating-point division bug lies in the silicon and why it occurred.