HardwareHacker Newsabout 1 hour ago
IEEE VHDL Standard Float Package Flawed
15 min read
Testing reveals the IEEE VHDL standard float package calculates square roots incorrectly for 735 of 10,000 float32 calls and 1,278 of 10,000 float64 calls. The source code on opensource.ieee.org contains logic errors causing wrong results or simulator stops. This flaw affects 859,771 total float32 calls and 857,234 float64 calls across various operations like add, multiply, and divide.
Level
Hype check
Tap to vote and see what everyone thinks.