QuBLAS is pronounced as [tɕʰy.blɑ:s]
This library provides simulation capabilities for fixed-point arithmetic operations in C++, tailored for behavioral modeling and accuracy testing of algorithms in a hardware context.
QuBLAS is designed to mimic the behavior of fixed-point arithmetic as closely as possible to what one would expect when running an algorithm on actual hardware such as ASICs or FPGAs. The library supports various quantization modes and overflow handling schemes, allowing users to simulate and explore different numerical and structural scenarios for their designs.