Design Example: Floating-Point Converter
Binary floating-point arithmetic is commonly used to represent real numbers on computers because it can represent a wide range of numbers. When implemented in FPGAs, however, floating-point can use considerable logic resources. Therefore, designers use fixed-point because of its better performance and lesser resource utilization. Essentially, a floating-point number is converted to a fixed-point before an operation. The result is then converted back to a floating-point for compatibility with other systems.
Efinix provides example design that demonstrates the conversion between floating-point and fixed-point numbers. The design targets the Titanium Ti60 F225 Development Board and Trion® T120 BGA576 Development Board. The example design includes Efinity Virtual I/O Debugger core for customizing and monitoring the design.
- Floating-point to fixed-point conversion
- Fixed-point to floating-point conversion
- Binary floating-point complies with IEEE 754 standard
- Supports floating-point of single-precision (32-bit) and double-precision (64-bit)
- Supports all floating-point numbers
- Supports a fixed-point width of 3 to 64 bits
- Optional asynchronous reset and clock enable ports for both float-to-fixed and fixed-to-float converters
- Optional output ports indicating overflow, underflow, not-a-number (NaN), infinity, zero, and denormalized number in float-to-fixed converter
- Fully-pipelined design with six pipeline stages
- RTL design in Verilog HDL
- Round to nearest, ties away from zero rounding rule
This content is locked and requires you to login before viewing more.