Chisel3 Float Type

The repository contains the source code implementing in Chisel3 a floating point representation compliant with ieee754. The objective being to use it in the designing of hardware accelerators or co-processors running alongside traditional CPUs (e.g. Xilinx Zynq & UltraScale, Intel Stratix), only the implementation of 32 bit float (float), 64 bit float (double) and 128 bit float (long double) are considered for the time being.

All the details about how to use the library and its features are available on the project wiki.

Launching tests

C

The C implementations are separated in two different files so one must each time compile both of them and then run them.

cd c/
gcc float.c -o testFloat.out
./testFloat.out
gcc double.c -o testDouble.out
./testDouble.out

Chisel

All the types and operations are included in the same testing process. Consequently, running the following line is sufficient.

cd chisel/
sbt "test:runMain fpu.FPUMain"

Generate the package

The generation of the package can be done as follows:

cd chisel/
sbt "package"
cp target/scala-2.11/chisel3-float-type_2.11-1.0.jar ../
mv ../chisel3-float-type_2.11-1.0.jar ../chisel-float-type.jar

Coming updates

• Make the project an importable library
• Implementing casting methods such as toFloat and others
• Implementation of the division operator
• Implementation of double
• Implementation of long double
• Addition logic optimisation
• Fix the rounding/precision issue