diff options
Diffstat (limited to 'doc')
| -rw-r--r-- | doc/style.md | 276 |
1 files changed, 276 insertions, 0 deletions
diff --git a/doc/style.md b/doc/style.md new file mode 100644 index 00000000..8529dfa1 --- /dev/null +++ b/doc/style.md @@ -0,0 +1,276 @@ +# Chisel Style Guide + +The Chisel style guide reflects the [Google Java style +guide](http://https://google.github.io/styleguide/javaguide.html) and the [General Public Scala style +guide](http://docs.scala-lang.org/style/). Specific rules below are to clarify +the style used for the chisel3 repo and repos related to Chisel (Firrtl). + +**Goal:** Readability and consistency are the main purposes of the style guide. +Writing your code so someone else (or yourself) can grok it later is important +to code health and quality. + +## Filenames +The source file name consists of the case-sensitive name of the top-level class +it contains, plus ".scala". + +## Packages + +Package definitions must contain the full path to the package from scala. If +you create a subpackage, it should go in a subdirectory. + + package directory.name.to.get.you.to.your.source + +As in Scala, packages follow the [Java package naming convention](https://google.github.io/styleguide/javaguide.html#s5.2.1-package-names). +Note that these guidelines call for all lowercase, no underscores. + +```scala +// Do this +package hardware.chips.topsecret.masterplan + +// Not this +package hardware.chips.veryObvious.bad_style +``` + +We also suggest you do not use chisel3 as a package, and especially do not use it +as the final (innermost) package. + +```scala +// Don't do this +package hardware.chips.newchip.superfastcomponent.chisel3 + +// This will lead to instantiating package members like so: +val module = Module(new chisel3.FastModule) + +// Which collides with the chisel namespace +import chisel3._ +``` + +## Imports +Avoid wildcard ( ._ ) imports, with the exception of chisel3._ +All other imports must call out used methods. +import chisel3._ must be first and separated from remaining imports with extra +blank line. + +**Reason:** This makes it clear where methods are defined. + +Remaining imports must be listed alphabetically. + +```scala +import chisel3._ + +import the.other.thing.that.i.reference.inline +import the.other.things.that.i.reference.{ClassOne, ClassTwo} + + +val myInline = inline.MakeAnInline() +val myClassOne = new ClassOne +``` + +## Tests +Test classes are named starting with the name of the class they are testing, and +ending with "Test". +Test files must start with the name of the class you are testing and end with +"Test.scala". +Test files should reside in a subdirectory called "tests". +The tests package should be composed of the package class you are testing. + +```scala +package class.under.test.class +package tests +``` + +## Comments +We use scaladoc to automatically generate documentation from the source code. + +```scala +/** Multiple lines of ScalaDoc text are written here, + * wrapped normally... + */ +public int method(String p1) { ... } +``` + +... or in this single-line example: + +```scala +/** An especially short bit of Javadoc. */ +``` + +Write documentation as if the person reading it knows more about Scala and +Chisel than you. If you find comments in the code consider breaking them up +into seperate methods. + +## Module Classes and Instances + +Modules can take different forms in Chisel. First, in the verilog sense where +you instance the module and then hook it up. In this case Module(new MyMod()) is +returning a reference to the module. + +```scala +val myMod = Module(new MyMod()) +myMod.io <> hookUp +``` + +Second, in a more programmatic inline style with factory methods. In this case +Queue is actually returning the part of the IO bundle representing the queue's +output. The factory method takes the input IO to the queue and an optional param +for depth. + +```scala +val queueOut = Queue(queueIn, depth=10) +``` + +The latter can be used for composing multiple functions into a single line. + +```scala +val queueOut = Queue( + Arbitrate.byRoundRobin( + Queue(a), // depth assumed to be 1 + Queue(b, depth=3), + Queue(c, depth=4) + ), + depth=10 +) +``` + +## Naming Conventions + +Chisel follows the [Scala Naming Conventions](http://docs.scala-lang.org/style/naming-conventions.html). +In general, Chisel code should use CamelCase for naming (ie. the first letter +of each word is capitalized except sometimes the first word). + +### Why CamelCase instead of Snake\_Case? + +The compiler inserts underscores when splitting Chisel/FIRRTL aggregate types +into Verilog types. The compiler uses underscores to preserve the original +structure of the data in the resulting Verilog. Because of the special meaning +of underscores in Chisel-generated Verilog, their use in naming is **strongly** +discouraged. + +Consider the following Chisel code: + +```scala +val msg = Wire(new Bundle { + val valid = Bool() + val addr = UInt(32) + val data = UInt(64) +}) +val msg_rec = Wire(Bool()) +``` + +Which compiles to the Verilog: + +```verilog +wire msg_valid; +wire [31:0] msg_addr; +wire [63:0] msg_data; +wire msg_rec; +``` + +The Verilog maintains the structure of the original aggregate wire `msg`. +However, because we named another variable `msg_rec`, it appears in the Verilog +as if `msg` had 4 fields instead of its actual 3! If we instead follow the +lowerCamelCase for values naming convention, the resulting Verilog makes more +sense: + +```scala +val msg = Wire(new Bundle { + val valid = Bool() + val addr = UInt(32) + val data = UInt(64) +}) +val msgRec = Wire(Bool()) +``` + +And its resulting Verilog: + +```verilog +wire msg_valid; +wire [31:0] msg_addr; +wire [63:0] msg_data; +wire msgRec; +``` + +Much better. + +### Modules and Bundles (Classes, Traits, and Objects) + +Modules are Scala classes and thus use UpperCamelCase. + +```scala +class ModuleNamingExample extends Module { + ... +} +``` + +Similarly, other classes (Chisel & Scala) should be UpperCamelCase as well. + +```scala +trait UsefulScalaUtilities { + def isEven(n: Int): Boolean = (n % 2) == 0 + def isOdd(n: Int): Boolean = !isEven(n) +} + +class MyCustomBundle extends Bundle { + ... +} +// Companion object to MyCustomBundle +object MyCustomBundle { + ... +} + +``` + +### Values and Methods + +Values and methods should use lowerCamelCase. (Unless the value is a constant.) + +```scala +val mySuperReg = Reg(init = 0.asUInt(32)) +def myImportantMethod(a: UInt): Bool = a < 23.asUInt +``` + +### Constants + +Unlike the Google Java style, constants use UpperCamelCase, which is in line +with the official [Scala Naming +Conventions](https://docs.scala-lang.org/style/naming-conventions.html). +Constants are final fields (val or object) whose contents are deeply immutable +and belong to a package object or an object. Examples: + +```scala +// Constants +object Constants { + val Number = 5 + val Names = "Ed" :: "Ann" :: Nil + val Ages = Map("Ed" -> 35, "Ann" -> 32) +} + +// Not constants +class NonConstantsInClass { + val inClass: String = "in-class" +} + +object nonConstantsInObject { + var varString = "var-string" + val mutableCollection: scala.collection.mutable.Set[String] + val mutableElements = Set(mutable) +} +``` + +### UpperCamelCase vs. lowerCamelCase + +There is more than one reasonable way to covert English prose into camel case. +We follow the convention defined in the [Google Java style +guide](https://google.github.io/styleguide/javaguide.html#s5.3-camel-case). The +potentially non-obvious rule being to treat acronymns as words for the purpose +of camel case. + +Note that the casing of the original words is almost entirely disregarded. +Example: + +Prose form | UpperCamelCase | lowerCamelCase | Incorrect +:------------- | :------------- | :------------- | :------------ +find GCD | FindGcd | findGcd | ~~findGCD~~ +state for FSM | StateForFsm | stateForFsm | ~~stateForFSM~~ +mock dut | MockDut | mockDut | ~~MockDUT~~ +FIFO Generator | FifoGenerator | fifoGenerator | ~~FIFOGenerator~~ |
