How to add additional IO under certain condition to RoCCIO in Chisel - riscv

I am trying to add custom instruction to RISC-V by using ROCC, and my base is the rocket-chip.
Some of the accelerators requires additional IO to be added to the RoCCIO class.
I am trying to understand the conditional instantiation of additional IO's, like in the case of the Floating point unit & page table walker (PTW) ,I am new to Chisel & Scala so it very cryptic to me.
Can someone explain how this code implement conditional instantiation:
class RoCCIO(outer: LazyRoCC)(implicit p: Parameters) extends RoCCCoreIO()(p) {
val ptw = Vec(p(RoccNPTWPorts), new TLBPTWIO)
val fpu_req = Decoupled(new FPInput)
val fpu_resp = Decoupled(new FPResult).flip
}

Your example of RoCC is using dead code elimination (however, you must be careful to tie off the signals you aren't using just to be sure).
You can instead use Option to perform conditional instantiation in your IO bundle declaration:
val a = if (cond) Some(UInt(width=5.W)) else None

This question is probably better suited for the rocket-chip repository so those developers will see it: https://github.com/freechipsproject/rocket-chip/issues
That being said, there are a couple of ways things can be conditionally instantiated in Chisel code.
Inheritance: when a Bundle extends another Bundle it can add additional fields, thus the inclusion of those fields is conditional upon which Bundle is actually being instantiated
Dead Code Elimination: Chisel/FIRRTL has robust Dead Code Elimination so if some ports are unused they will simply be removed. Thus the fields can be conditionally removed in configurations that do not use them. I believe this is the case here (but am unsure).

Related

how to get hierarchy signals with similar name?

suppose in my test bench, i had following signals
top.module0.expect
top.module1.expect
yes, we instantiates module0/1 with same module
now, in a function get_expect_sig(int module_idx) (module_idx could be 0 or 1), i want to get the signal according to module_idx
what i'm currently do is as follow:
if (module_idx == 0) return top.module0.expect;
else if (module_idx == 1) return top.module1.expect;
this solution is pretty ugly.
if there any brilliant method that help me with this? because i got 32 modules, not 2 modules
thanks!
A case statement would be slightly less ugly. Since your design choose to instantiate the modules with unique names instead of an array of instances, there is no way to iterate over them.
Some tools allow you to look up a signal's value using a string name. Then you could format the string using the index value. You'll need to read your tools manual.
If this is testbench code (non-synthesizable), if module 0 and module 1 are instances of the same module, you could bind an interface to it. A bind statement will cause this interface to be instantiated in all instances of the module. You could then put some helper code in this interface to 'register' with a configuration class. In UVM, this could be the config db, but in SystemVerilog you could use a static or singleton class.
Example: https://www.edaplayground.com/x/5ZUG
This uses a few concepts (abstract-concrete polymorphic interfaces) that you may need to read about. But hopefully the example is clear enough.

Can one change/influence JAXB's code generation?

I was wondering whether one can influence the "style" of the code that JAXB generates from XML schema (.xsd) fles. E.g. I would like to:
emit a comment inside newly generated classes, specifically if the class is empty, since that triggers warnings in my environment.
change all setter-methods to return the object instead of "void", so one can do call-chaining like:
X someMethod() {
return new X().setFoo(5).setBar("something");
}
instead of the tedious:
X someMethod() {
X x = new (X);
x.setFoo(5);
x.setBar("something");
return x;
}
Is there some "template" anywhere that JAXB uses and that one could tweak, to achieve such things? Or is that all hard-coded?
M.
There is no template for modifying the generated code easily.
There is, however, a number of plugins. For instance: https://java.net/projects/jaxb2-commons/pages/Fluent-api which is just what you want according to your 2nd bullet.
There are other plugins, e.g. for annotations suppressing warnings - that may help against the 1st bullet.
As an extra, I'd like to mention that not generating Java classes from an XML schema but writing them by hand (plus annotations, of course) is a plausible alternative, provided the XML schema isn't too complex. It may have other advantages besides solving #1 and #2.

Metaprogramming: adding equals(Object o) and hashCode() to a library class

I have a library of domain objects which need to be used in the project, however we've found a couple of the classes haven't got an equals or hashCode method implemented.
I'm looking for the simplest (and Grooviest) way to add those methods. Obviously I could create a subclass which only adds the methods, but this would be confusing for developers used to the library and would mean we'd have to refactor existing code.
It is not possible to get the source changed (currently).
If I could edit the class I would just use the #EqualsAndHashCode annotation to carry out an AST transformation (at compile time?), but I can't find a way to instruct the compiler to carry out the transformation on a class which I can't directly annotate.
I'm currently trying to work up an example using the ExpandoMetaClass, so I'd do something like:
MySuperClass.metaClass.hashCode = { ->
// Add dynamic hashCode calculation bits here
}
MySuperClass.metaClass.equals = { ->
// Add dynamic hashCode calculation bits here
}
I don't really want to hand-code the hashCode/equals methods for each class, so I'm looking for something dyamic (like #EqualsAndHashCode) which will work with this.
Am I on the right track? Is there a groovier way?
AST Transforms are only applied at compile time, so you'll get no help from the likes of #EqualsAndHashCode. MetaClass hacks are going to be your only option. That said, there are more-elegant ways to impose MetaClass behavior.
Shameless Self Plug I did a talk about this kind of stuff last year at SpringOne 2GX: http://www.infoq.com/presentations/groovy-app-architecture
In short, you might find benefit in creating extensions (unless you're in Grails) - http://mrhaki.blogspot.com/2013/01/groovy-goodness-adding-extra-methods.html, or by explicitly adding mixins - http://groovy.codehaus.org/Runtime+mixins ... But in general, these are just cleaner ways to do the exact same thing you're already doing.

Ignore certain TypeScript compile errors?

I am wondering if there is a way to ignore certain TypeScript errors upon compilation?
I basically have the same issues most people with large projects have around using the this keyword, and I don't want to put all my classes methods into the constructor.
So I have got an example like so:
TypeScript Example
Which seems to create perfectly valid JS and allows me to get around the this keyword issue, however as you can see in the example the typescript compiler tells me that I cannot compile that code as the keyword this is not valid within that scope. However I don't see why it is an error as it produces okay code.
So is there a way to tell it to ignore certain errors? I am sure given time there will be a nice way to manage the this keyword, but currently I find it pretty dire.
== Edit ==
(Do not read unless you care about context of this question and partial rant)
Just to add some context to all this to show that I'm not just some nut-job (I am sure a lot of you will still think I am) and that I have some good reasons why I want to be able to allow these errors to go through.
Here are some previous questions I have made which highlight some major problems (imo) with TypeScript current this implementation.
Using lawnchair with Typescript
Issue with child scoping of this in Typescript
https://typescript.codeplex.com/discussions/429350 (And some comments I make down the bottom)
The underlying problem I have is that I need to guarantee that all logic is within a consistent scope, I need to be able to access things within knockout, jQuery etc and the local instance of a class. I used to do this with the var self = this; within the class declaration in JavaScript and worked great. As mentioned in some of these previous questions I cannot do that now, so the only way I can guarantee the scope is to use lambda methods, and the only way I can define one of these as a method within a class is within the constructor, and this part is HEAVILY down to personal preference, but I find it horrific that people seem to think that using that syntax is classed as a recommended pattern and not just a work around.
I know TypeScript is in alpha phase and a lot will change, and I HOPE so much that we get some nicer way to deal with this but currently I either make everything a huge mess just to get typescript working (and this is within Hundreds of files which I'm migrating over to TypeScript ) or I just make the call that I know better than the compiler in this case (VERY DANGEROUS I KNOW) so I can keep my code nice and hopefully when a better pattern comes out for handling this I can migrate it then.
Also just on a side note I know a lot of people are loving the fact that TypeScript is embracing and trying to stay as close to the new JavaScript features and known syntax as possible which is great, but typescript is NOT the next version of JavaScript so I don't see a problem with adding some syntactic sugar to the language as people who want to use the latest and greatest official JavaScript implementation can still do so.
The author's specific issue with this seems to be solved but the question is posed about ignoring errors, and for those who end up here looking how to ignore errors:
If properly fixing the error or using more decent workarounds like already suggested here are not an option, as of TypeScript 2.6 (released on Oct 31, 2017), now there is a way to ignore all errors from a specific line using // #ts-ignore comments before the target line.
The mendtioned documentation is succinct enough, but to recap:
// #ts-ignore
const s : string = false
disables error reporting for this line.
However, this should only be used as a last resort when fixing the error or using hacks like (x as any) is much more trouble than losing all type checking for a line.
As for specifying certain errors, the current (mid-2018) state is discussed here, in Design Meeting Notes (2/16/2018) and further comments, which is basically
"no conclusion yet"
and strong opposition to introducing this fine tuning.
I think your question as posed is an XY problem. What you're going for is how can I ensure that some of my class methods are guaranteed to have a correct this context?
For that problem, I would propose this solution:
class LambdaMethods {
constructor(private message: string) {
this.DoSomething = this.DoSomething.bind(this);
}
public DoSomething() {
alert(this.message);
}
}
This has several benefits.
First, you're being explicit about what's going on. Most programmers are probably not going to understand the subtle semantics about what the difference between the member and method syntax are in terms of codegen.
Second, it makes it very clear, from looking at the constructor, which methods are going to have a guaranteed this context. Critically, from a performance, perspective, you don't want to write all your methods this way, just the ones that absolutely need it.
Finally, it preserves the OOP semantics of the class. You'll actually be able to use super.DoSomething from a derived class implementation of DoSomething.
I'm sure you're aware of the standard form of defining a function without the arrow notation. There's another TypeScript expression that generates the exact same code but without the compile error:
class LambdaMethods {
private message: string;
public DoSomething: () => void;
constructor(message: string) {
this.message = message;
this.DoSomething = () => { alert(this.message); };
}
}
So why is this legal and the other one isn't? Well according to the spec: an arrow function expression preserves the this of its enclosing context. So it preserves the meaning of this from the scope it was declared. But declaring a function at the class level this doesn't actually have a meaning.
Here's an example that's wrong for the exact same reason that might be more clear:
class LambdaMethods {
private message: string;
constructor(message: string) {
this.message = message;
}
var a = this.message; // can't do this
}
The way that initializer works by being combined with the constructor is an implementation detail that can't be relied upon. It could change.
I am sure given time there will be a nice way to manage the this keyword, but currently I find it pretty dire.
One of the high-level goals (that I love) in TypeScript is to extend the JavaScript language and work with it, not fight it. How this operates is tricky but worth learning.

should it be allowed to change the method signature in a non statically typed language

Hypothetic and academic question.
pseudo-code:
<pre><code>
class Book{
read(theReader)
}
class BookWithMemory extends Book {
read(theReader, aTimestamp = null)
}
</pre></code>
Assuming:
an interface (if supported) would prohibit it
default value for parameters are supported
Notes:
PHP triggers an strict standards error for this.
I'm not surprised that PHP strict mode complains about such an override. It's very easy for a similar situation to arise unintentionally in which part of a class hierarchy was edited to use a new signature and a one or a few classes have fallen out of sync.
To avoid the ambiguity, name the new method something different (for this example, maybe readAt?), and override read to call readAt in the new class. This makes the intent plain to the interpreter as well as anyone reading the code.
The actual behavior in such a case is language-dependent -- more specifically, it depends on how much of the signature makes up the method selector, and how parameters are passed.
If the name alone is the selector (as in PHP or Perl), then it's down to how the language handles mismatched method parameter lists. If default arguments are processed at the call site based on the static type of the receiver instead of at the callee's entry point, when called through a base class reference you'd end up with an undefined argument value instead of your specified default, similarly to what would happen if there was no default specified.
If the number of parameters (with or without their types) are part of the method selector (as in Erlang or E), as is common in dynamic languages that run on JVM or CLR, you have two different methods. Create a new overload taking additional arguments, and override the base method with one that calls the new overload with default argument values.
If I am reading the question correctly, this question seems very language specific (as in it is not applicable to all dynamic languages), as I know you can do this in ruby.
class Book
def read(book)
puts book
end
end
class BookWithMemory < Book
def read(book,aTimeStamp = nil)
super book
puts aTimeStamp
end
end
I am not sure about dynamic languages besides ruby. This seems like a pretty subjective question as well, as at least two languages were designed on either side of the issue (method overloading vs not: ruby vs php).

Resources