- Why DecoupledIO instead of ReadyValidIO?
- Why do I have to wrap module instantiations in
Module(...)? - Why Chisel?
There are multiple kinds of Ready/Valid interfaces that impose varying restrictions on the producers and consumers. Chisel currently provides the following:
- DecoupledIO - No guarantees
- IrrevocableIO - Producer promises to not change the value of 'bits' after a cycle where 'valid' is high and 'ready' is low. Additionally, once 'valid' is raised it will never be lowered until after 'ready' has also been raised.
In short: Limitations of Scala
Chisel Modules are written by defining a Scala class and implementing its constructor. As elaboration runs, Chisel constructs a hardware AST from these Modules. The compiler needs hooks to run before and after the actual construction of the Module object. In Scala, superclasses are fully initialized before subclasses, so by extending Module, Chisel has the ability to run some initialization code before the user's Module is constructed. However, there is no such hook to run after the Module object is initialized. By wrapping Module instantiations in the Module object's apply method (ie. Module(...)), Chisel is able to perform post-initialization actions. There is a proposed solution, so eventually this requirement will be lifted, but for now, wrap those Modules!
Borrowed from [Chisel Introduction](Chisel Introduction)
We were motivated to develop a new hardware language by years of struggle with existing hardware description languages in our research projects and hardware design courses. Verilog and VHDL were developed as hardware simulation languages, and only later did they become a basis for hardware synthesis. Much of the semantics of these languages are not appropriate for hardware synthesis and, in fact, many constructs are simply not synthesizable. Other constructs are non-intuitive in how they map to hardware implementations, or their use can accidently lead to highly inefficient hardware structures. While it is possible to use a subset of these languages and still get acceptable results, they nonetheless present a cluttered and confusing specification model, particularly in an instructional setting.
However, our strongest motivation for developing a new hardware language is our desire to change the way that electronic system design takes place. We believe that it is important to not only teach students how to design circuits, but also to teach them how to design circuit generators ---programs that automatically generate designs from a high-level set of design parameters and constraints. Through circuit generators, we hope to leverage the hard work of design experts and raise the level of design abstraction for everyone. To express flexible and scalable circuit construction, circuit generators must employ sophisticated programming techniques to make decisions concerning how to best customize their output circuits according to high-level parameter values and constraints. While Verilog and VHDL include some primitive constructs for programmatic circuit generation, they lack the powerful facilities present in modern programming languages, such as object-oriented programming, type inference, support for functional programming, and reflection.
Instead of building a new hardware design language from scratch, we chose to embed hardware construction primitives within an existing language. We picked Scala not only because it includes the programming features we feel are important for building circuit generators, but because it was specifically developed as a base for domain-specific languages.