Beyond Floating Point: Next-Generation Computer Arithmetic
Bluestein
16 points
5 comments
November 16, 2026
Related Discussions
Found 5 related stories in 197.7ms across 13,334 title embeddings via pgvector HNSW
- The Intel 8087 chip: floating-point addition and subtraction pwg · 97 pts · October 18, 2026 · 58% similar
- Microcode in Intel's 8087 floating-point chip: the scale instruction pwg · 99 pts · September 12, 2026 · 55% similar
- z486: A 486-Class Pipelined FPGA CPU with Integrated Floating-Point matt_d · 19 pts · August 15, 2026 · 52% similar
- Breaking the 1.58-bit Barrier for Ternary LLMs matt_d · 177 pts · September 16, 2026 · 50% similar
- Processing in Memory: DRAM Is About to Do Math bhouston · 12 pts · August 26, 2026 · 49% similar
Discussion Highlights (3 comments)
0xcoffee
Somewhat related, Anger Fog has a document about improving the IEEE 754 standard: https://www.agner.org/optimize/nan_propagation.pdf
AlotOfReading
Gustafson's arguments haven't held up well to further development, in my opinion. As originally designed they aren't faster, they don't "solve" numerical analysis, and a lot of the issues people have with floats don't go away. They get a few things right like eliminating NaNs, but the reality is just so underwhelming compared to Gustafson's grandiose claims.
Panzerschrek
For any bit width, they have a larger dynamic range, higher accuracy Isn't it a contradiction? One can't increase range without sacrificing accuracy.