Changes between Version 2 and Version 3 of Commentary/Libraries/Integer

Timestamp:

09/25/11 06:29:50 (21 months ago)

Author:

igloo

Comment:

--

Commentary/Libraries/Integer

@@ -3 +3 @@
 = GHC Commentary: Libraries/Integer =
 
-TODO
+GHC is set up to allow different implementations of the `Integer` type to be chosen at build time.
 
 == Selecting an Integer implementation  ==
 
-You can select which implementation of Integer is used by defining `INTEGER_LIBRARY` in `mk/build.mk`. The code is in `libraries/$(INTEGER_LIBRARY)`.
+You can select which implementation of Integer is used by defining `INTEGER_LIBRARY` in `mk/build.mk`. This tells the build system to build the library in `libraries/$(INTEGER_LIBRARY)`, and the `cIntegerLibrary` and `cIntegerLibraryType` values in `Config.hs` are defined accordingly.
 
 The default value is `integer-gmp`, which uses the [http://gmplib.org/ GNU Multiple Precision Arithmetic Library (GMP)] to define the Integer type and its operations.
@@ -15 +15 @@
 == The Integer interface  ==
 
-TODO
+All Integer implementations should export the same set of types and functions from `GHC.Integer`. These exports are used by the `base` package However, all of these types and functions must actually be defined in `GHC.Integer.Type`, so that GHC knows where to find them.
 
 == How Integer is handled inside GHC  ==
 
-TODO
+Integers are represented using the `HsInteger` constructor of `HsLit` for the early phases of compilation (e.g. type checking), but for later stages, once we use the `Core` representation, they are converted to the `LitInteger` constructor of the `Literal` type by `mkIntegerExpr`. While `Integer`s aren't "machine literals" like the other `Literal` constructors, it is more convenient when writing rules to pretend that they are literals rather than having to understand their real core representation.
+
+All of the types and functions in the `Integer` interface have built-in names, e.g. `plusIntegerName`, defined in [[GhcFile(compiler/prelude/PrelNames.lhs)]] and included in `basicKnownKeyNames`. This allows us to match on all of the functions in `builtinIntegerRules` in [[GhcFile(compiler/prelude/PrelRules.lhs)]], so we can constant-fold Integer expressions.
+
+We keep the `LitInteger` representation as late as possible; in particular, it's important that this representation is used in unfoldings in interface files, so that constant folding can happen on expressions that get inlined. We only convert it to a proper core representation of Integer in [[GhcFile(compiler/coreSyn/CorePrep.lhs)]], where we normally use the Id for `mkInteger` (which we are carrying around in the `LitInteger` constructor) to build an expression like `mkInteger True [123, 456]` (where the `Bool` represents the sign, and the list of `Int`s are 31 bit chunks of the absolute value from lowest to highest).