Project Highlight: BSDScheme

Posted on

Last year, Phil Eaton started working on BSDScheme, a Scheme interpreter that he ultimately intends to support Scheme R7RS. In college, he had completed two compiler projects in C++ for two different courses. One was a Scheme to Forth compiler, the other an implementation of the Tiger language from Andrew Appel’s ‘Modern Compiler Implementation’ books.

I hadn’t really written a complete interpreter or compiler since then, but I’d been trying to get back into language implementation. I planned to write a Scheme interpreter that was at least bootstrapped from a compiled language so I could go through the traditional steps of lexing, parsing, optimizing, compiling, etc., and not just build a language of the meta-circular interpreter variety. I was spurred to action when my co-worker at Linode, Brian Steffens, wrote bshift, a compiler for a C-like language.

For his new project, he wanted to use something other than C++. Though he knows the language and likes some of the features, he overall finds it a “complicated mess”. So he started on BSDScheme using C, building generic ADTs via macros.

As he worked on the project, he referred to Brian’s bshift for inspiration. As it happens, bshift is implemented in D. Over time, he discovered that he “really liked the power and simplicity of D”. That eventually led him to drop C for D.

It was clear it would save me a ton of time implementing all the same data structures and flows one implements in almost every new C project. The combination of compile-time type checking, GC, generic ADT support, and great C interoperability was appealing.

He’s been developing the project on Mac and FreeBSD, using LDC, the LLVM-based D compiler. In that time, he has found a number of D features beneficial, but two stand out for him above the rest. The first is nested functions.

They’re a good step up from C, where nested functions are not part of the standard and only unofficially supported (in different ways) by different compilers. C++ has lambdas, but that’s not really the same thing. It is a helpful syntactic sugar used in BSDScheme for defining new functions with external context (the names of the parameters to bind).

As for the second, hold on to your seats: it’s the GC.

The existence of a standard GC is a big benefit of D over C and C++. Sure, you could use the Boehm GC, but how that works with threads is up to you to discover. It is not fun to do prototyping in a GC-less language because the amount of boilerplate distracts from the goals. People often say this when they’re referring to Python, Ruby, or Node, but D is not at all comparable for (among) a few reasons: 1) compile-time type-checking, 2) dead-simple C interop, 3) multi-processing support.

Spend some time in the D forums and you’ll often find newcomers from C and C++ who, unlike Phil, have a strong aversion to garbage collection and are actively seeking to avoid it. You’ll also find replies from long-time D coders who started out the same way, but eventually came to embrace the GC completely and learned how to make it work for them. The D GC can certainly be problematic for certain types of software, but it is a net win for others. This point is reiterated frequently in the GC series on this blog, which shows how to get the most out of the GC, profile it, and mitigate its impact if it does become a performance problem.

As Phil learned the language, he identified areas for improvement in the D documentation.

Certainly it is advantageous compared to the C preprocessor that there is not an entirely separate language for doing compile-time macros, but the behavior difference and transition guides are missing or poorly written. A comparison between D templates and C++ templates (in all their complexity) could also be a great source of explanation.

We’re always looking to improve the documentation and make it more friendly to newcomers of all backgrounds. The docs are often written by people who are already well-versed in the language and its ecosystem, making blind spots somewhat inevitable. Anyone in the process of learning D is welcome and encouraged to help improve both the Language and Library docs. In the top right corner of each page are two links: “Report a bug” and “Improve this page”. The first takes you to D’s bug tracker to report a documentation bug, the second allows anyone with a logged-in GitHub account to quickly fork, edit the page online, and submit a pull request.

In addition to the ulitmate goal of supporting Scheme R7RS, Phil plans to add FFI support in order to allow BSDScheme to call D functions directly, as well as support for D threads and an LLVM-based backend.

Overall, he seems satisfied with his decision to move to D for the implementation.

I think D has been a good time investment. It is a very practical language with a lot of the necessary high-level aspects and libraries for modern development. In the future, I plan to dig more into the libraries and ecosystem for backend web systems. Furthermore, unlike with C or C++, so far I’d feel comfortable choosing D for projects where I am not the sole developer. This includes issues ranging from prospective ease of onboarding to long-term performance and maintainability.

A big thanks to Phil for taking the time to contribute to this post (be sure to check out his blog, too). We’re always happy to hear about new projects in D. BSDScheme is released under the three-clause BSD license, so it’s a great place to start for anyone looking for an interesting open-source D project to contribute to or learn from. Have fun!

DMD 2.078.0 Has Been Released

Posted on

Another major release of DMD, this time 2.078.0, has been packaged and delivered in time for the new year. See the full changelog at and download the compiler for your platform either from the main download page or the 2.078.0 release directory.

This release brings a number of quality-of-life improvements, fixing some minor annoyances and inconsistencies, three of which are targeted at smoothing out the experience of programming in D without DRuntime.

C runtime construction and destruction

D has included static constructors and destructors, both as aggregate type members and at module level, for most of its existence. The former are called in lexical order as DRuntime goes through its initialization routine, and the latter are called in reverse lexical order as the runtime shuts down. But when programming in an environment without DRuntime, such as when using the -betterC compiler switch, or using a stubbed-out runtime, static construction and destruction are not available.

DMD 2.078.0 brings static module construction and destruction to those environments in the form of two new pragmas, pragma(crt_constructor) and pragma(crt_destructor) respectively. The former causes any function to which it’s applied to be executed before the C main, and the latter after the C main, as in this example:


// Compile with:    dmd crun1.d
// Alternatively:   dmd -betterC crun1.d

import core.stdc.stdio;

// Each of the following functions should have
// C linkage (cdecl).

void init()

void fini()

void main()
    puts("C main");

The compiler requires that any function annotated with the new pragmas be declared with the extern(C) linkage attribute. In this example, though it isn’t required, main is also declared as extern(C). The colon syntax on line 8 applies the attribute to every function that follows, up to the end of the module or until a new linkage attribute appears.

In a normal D program, the C main is the entry point for DRuntime and is generated by the compiler. When the C runtime calls the C main, the D runtime does its initialization, which includes starting up the GC, executing static constructors, gathering command-line arguments into a string array, and calling the application’s main function, a.k.a. D main.

When a D module’s main is annotated with extern(C), it essentially replaces DRuntime’s implementation, as the compiler will never generate a C main function for the runtime in that case. If -betterC is not supplied on the command line, or an alternative implementation is not provided, DRuntime itself is still available and can be manually initialized and terminated.

The example above is intended to clearly show that the crt_constructor pragma will cause init to execute before the C main and the crt_destructor causes fini to run after. This introduces new options for scenarios where DRuntime is unavailable. However, take away the extern(C) from main and the same execution order will print to the command line:


// Compile with:    dmd crun2.d

import core.stdc.stdio;

extern(C) void init()

extern(C) void fini()

void main()
    import std.stdio : writeln;
    writeln("D main");

The difference is that the C main now belongs to DRuntime and our main is the D main. The execution order is: init, C main, D main, fini. This means init is effectively called before DRuntime is initialized and fini after it terminates. Because this example uses the DRuntime function writeln, it can’t be compiled with -betterC.

You may discover that writeln works if you import it at the top of the module and substitute it for puts in the example. However, always remember that even though DRuntime may be available, it’s not in a valid state when a crt_constructor and a crt_destructor are executed.

RAII for -betterC

One of the limitations in -betterC mode has been the absence of RAII. In normal D code, struct destructors are executed when an instance goes out of scope. This has always depended on DRuntime, and since the runtime isn’t available in -betterC mode, neither are struct destructors. With DMD 2.078.0, the are in the preceding sentence becomes were.


// Compile with:    dmd -betterC destruct.d

import core.stdc.stdio : puts;

struct DestroyMe
        puts("Destruction complete.");

extern(C) void main()
    DestroyMe d;

Interestingly, this is implemented in terms of try..finally, so a side-effect is that -betterC mode now supports try and finally blocks:


// Compile with:    dmd -betterC cleanup1.d

import core.stdc.stdlib,

extern(C) void main()
    int* ints;
        // acquire resources here
        ints = cast(int*)malloc(int.sizeof * 10);
        // release resources here

Since D’s scope(exit) feature is also implemented in terms of try..finally, this is now possible in -betterC mode also:


// Compile with: dmd -betterC cleanup2.d

import core.stdc.stdlib,

extern(C) void main()
    auto ints1 = cast(int*)malloc(int.sizeof * 10);
        puts("Freeing ints1!");

    auto ints2 = cast(int*)malloc(int.sizeof * 10);
        puts("Freeing ints2!");

Note that exceptions are not implemented for -betterC mode, so there’s no catch, scope(success), or scope(failure).

Optional ModuleInfo

One of the seemingly obscure features dependent upon DRuntime is the ModuleInfo type. It’s a type that works quietly behind the scenes as one of the enabling mechanisms of reflection and most D programmers will likely never hear of it. That is, unless they start trying to stub out their own minimal runtime. That’s when linker errors start cropping up complaining about the missing ModuleInfo type, since the compiler will have generated an instance of it for each module in the program.

DMD 2.078.0 changes things up. The compiler is aware of the presence of the runtime implementation at compile time, so it can see whether or not the current implementation provides a ModuleInfo declaration. If it does, instances will be generated as appropriate. If it doesn’t, then the instances won’t be generated. This makes it just that much easier to stub out your own runtime, which is something you’d want to do if you were, say, writing a kernel in D.

Other notable changes

New users of DMD on Windows will now have an easier time getting a 64-bit environment set up. It’s still necessary to install the Microsoft build tools, but now DMD will detect the installation of either the Microsoft Build Tools package or Visual Studio at runtime when either -m64 or -m32mscoff is specified on the command line. Previously, configuration was handled automatically only by the installer; manual installs had to be configured manually.

DRuntime has been enhanced to allow more fine-grained control over unit tests. Of particular note is the --DRT-testmode flag which can be passed to any D executable. With the argument "run-main", the current default, any unit tests present will be run and then main will execute if they all pass; with "test-or-main", the planned default beginning with DMD 2.080.0, any unit tests present will run and the program will exit with a summary of the results, otherwise main will execute; with "test-only", main will not be executed, but test results will still be summarized if present.

Onward into 2018

This is the first DMD release of 2018. We can’t wait to see what the next 12 months bring for the D programming language community. From everyone at the D Language Foundation, we hope you have a very Happy New Year!

D’s Newfangled Name Mangling

Posted on

Rainer Schuetze is the creator and maintainer of Visual D, the D plugin for Visual Studio. Recently, he implemented a new name mangling algorithm for the D frontend, which was released in DMD 2.077.0. In this post, he explains why it was needed and what it does.

What is symbol name mangling?

D embraces the separate compilation model that compiles D source code to object files and uses a linker to bind the object files to an executable binary file. This allows the reuse of precompiled object files and libraries, speeding up the build process. As the linker is usually one that’s also used for other languages with the same compilation model, e.g. C/C++ or Fortran, mixing object files from different languages is straightforward.

In an object file, a symbol name is assigned to each function or global variable, both when it is defined and when it is used via a call or access. The linker uses these names to connect references to definitions of the same name with only very bare knowledge about the symbol. For example, the symbol for this C function declaration,

extern(C) const(char)* find(int ch, const(char)* str);

does not tell the linker anything about function arguments or return type, as the C language uses the plain function name find as the symbol name (some platforms prepend a _ to the symbol). If you later change the order of the arguments to

extern(C) const(char)* find(const(char)* str, int ch);

but fail to update and recompile all source files that use the new declarartion, the linker will happily bind the resulting object files. In that case, the program is likely to crash, since a character passed to the function will be interpreted as a string pointer and vice versa.

D and C++ avoid this problem by adding more information to the symbol name, i.e. they encode into a symbol name the scope in which the symbol is defined, the function argument types, and the return type. Even if the linker does not interpret this information, linking fails with an undefined symbol error if the definitions used to build the object files don’t match. For example, the D function declaration

module test;
extern(D) const(char)* find(int ch, const(char)* str);

has a symbol name _D4test4findFiPxaZPxa, where _D is a prefix to identify the symbol as being generated from a D source symbol, 4test4find encodes the “fully qualified name” find in module test, and FiPxaZPxa describes the function type with an integer argument (designated by i) and the C-style string pointer type Pxa by just concatenating the encodings for argument types. Z terminates the function argument list and is followed by the encoding for the return type, again Pxa for a C-style string pointer. In contrast,

extern(D) const(char)* find(const(char)* str, int ch);

is encoded as _D4test4findFPxaiZPxa, making it a different symbol with the argument types reversed. The encoding ensures a normalized representation of types and scopes while also providing shorter symbols than minimal source code. This encoding is called “name mangling”.

Ed: Note that extern(C) and extern(D) are linkage attributes. When a function is declared in D without an explicit linkage attribute, extern(D) is the default.

In D, some function attributes are also mangled into the symbol name, e.g. @safe, nothrow, pure and @nogc. In theory, mangling could also cover parameter names, user defined attributes, or even contracts, but that is currently considered excessive.

Please note that even though name mangling can detect some mismatches in the binary interface of functions (i.e. how arguments are passed in registers or on the stack), it won’t catch every error; for example, structs, classes and other user defined types are mangled by name only, so that a change to their definition will still pass unnoticed by the linker.

The mangled name of a symbol is also available during compilation using the .mangleof property. This used to be exploited to provide type reflection of the symbol at compile time. This should no longer be necessary due to the introduction of new __traits that make this information accessible faster and more convenient, for example,



__traits(getFunctionAttributes, symbol);

Thus, usage of .mangleof is not recommended except for debugging purposes.

When reversing the mangling process in the “demangler”, all the encoded information is kept to make it available to the user, but that does not always yield correct D syntax. The first definition above demangles as

const(char)* test.find(int, const(char)*)

i.e. the module name test is added to the function name.

Template symbols

The two definitions of find shown above can coexist in D and C++, so name mangling is not only a way to detect errors at link time but also a necessity to represent overloads. It should at least contain enough information to distinguish different overloads of the same scoped identifier.

This becomes even more obvious when considering templates that usually instantiate different functions or variable definitions for each argument type. In D, the template instantiation information is added to the qualified name of a symbol.

Consider expression templates, a common example of meta programming used for delayed evaluation of expressions:

module expr;
struct Mul(X,Y)
    X x;
    Y y;
struct Add(X,Y)
    X x;
    Y y;

auto mul(X,Y)(X x, Y y) { return Mul!(X,Y)(x, y); }
auto add(X,Y)(X x, Y y) { return Add!(X,Y)(x, y); }

A function template is lowered by the compiler to an eponymous template:

template mul(X, Y)
    auto mul(X x, Y y) { return Mul!(X,Y)(x, y); }

The template name is part of the qualified function name, expr.mul!(X,Y).mul, and the auto return type is inferred to be Mul!(X,Y). This causes the symbol to reference the types X and Y three times. The demangled mangled name of an instantiation with types double and float of this template is

expr.Mul!(double,float) expr.mul!(double,float).mul(double,float)

The mangling process of DMD before version 2.077 walks the abstract syntax tree of the declaration and emits the mangled representation of the types whenever it is hit. Now consider stacking operations, e.g.

auto square(X)(X x) { return mul(x, x); }

auto len = square("var");
pragma(msg, len.square.mangleof);
// S4expr66__T3MulTS4expr16__T3MulTAyaTAyaZ3MulTS4expr16__T3MulTAyaTAyaZ3MulZ3Mul

pragma(msg, typeof(len).mangleof.length);
pragma(msg, len.square.mangleof.length);
pragma(msg, len.square.square.mangleof.length);
pragma(msg, len.square.square.square.mangleof.length);
pragma(msg, len.square.square.square.square.mangleof.length);
pragma(msg, len.square.square.square.square.square.mangleof.length);
pragma(msg, len.square.square.square.square.square.square.mangleof.length);

With DMD 2.076 or earlier, this displays 28u, 78u, 179u, 381u, 785u, 1594u, 3212u, showing exponential growth of the mangled symbol name length even though the expression in the source code just grows linearly. This happens because types like Mul!(Mul!(string, string), Mul!(string, string)) are combined and the mangling repeats their full representation every time they are referenced.

Create a chain of 12 calls to square above and the symbol length increases to 207,114. Even worse, the resulting object file for COFF/64-bit is larger than 15 MB and the time to compile increases from 0.1 seconds to about 1 minute. Most of that time is spent generating code for functions only used at compile time.

Voldemort types returned from template functions can be similar, as they carry the function signature including the template arguments as part of the type name. These can also show a dramatic increase in build times without generating as much code as in the example.

Symbol compression to the rescue

In early 2016, a couple of attempts were made to shorten these long symbols:

  • cut off symbol names if they exceed a given threshold, but append a checksum of the full symbol instead. This was already done with an MD5 hash when emitting symbols for the DigitalMars C compiler tool chain as the OMF object file format does not allow symbols longer than 255 characters. The downside to this is that these symbols can no longer be demangled, so that symbols in linker messages cannot be translated back into human digestible names.
  • apply binary compression to the symbol name. Standard techniques use part of the full symbol name as a dictionary to encode repetitions within the name. This is usually done by encoding a position-length pair using characters outside the normal identifier set. Again, this is already in use when DMD tries to fit symbols into the OMF limit of 255 characters (before applying the MD5 hash trick), but it also has shown some disadvantages: when using characters above the ASCII range, this interferes with UTF8 encoded characters that are also allowed as symbol characters in the D language. It can also break linker output as the console might misinterpret it as a locale-specific character encoding. Avoiding this by applying a binary to ASCII conversion like base64 to the symbols would obfuscate the actual symbol name even more.
  • extend the mangling grammar by allowing references to entities already encoded. This is similar to binary compression, but does not need to encode match length as the entities have terminators embedded into the grammar. The most prominent entities are types. This is the road C++ has taken, as it is also affected by the issues described here. In C++, name mangling is not standardised by the language, but by the compiler or the platform. GNU g++ uses Itanium C++ ABI mangling, which does a pretty good job with C++ code similar to the example above or in the Voldemort issue. Even though Microsoft’s Visual C++ can encode recurring types as well, it still generates very long names because it limits the encoding to the first 10 types of an argument list.

The first attempts at applying the latter scheme to the mangling of D symbols showed disappointing results. As it turned out, these implementations missed a subtle detail of the mangler in the DMD front end at that time; it reused cached representations of mangled type names to combine them to more complex types. This fails to find repetitions of the types from which a cached type name was built.

This is where I stepped in to create a proof-of-concept version of the mangling without these omissions. Early results were promising, so I looked for more opportunities to reduce symbol length:

  • with fully qualified names always containing the package and module names of a symbol, identifiers tend to appear often in a mangled name.
  • qualified names are likely to come from the same module or package, so it would be nice to encode them as a single entity.

The unit tests of the Phobos runtime library are benchmark candidates, as they contain a lot of symbols for template-heavy code. At the given time there were 127,172 symbols found in the map file of the Windows build. These were the results of the different manglings:

back references max length average length
none 416133 369
types 2095 157
types+identifiers 1263 128
types+identifiers+qualified names 1114 117

(This has been measured with the implementation at the time, which is not exactly the same as the final mangling; it used special characters for the different back reference types, but this turned out not to be a good idea. The D mangling is supposed to be the same on all platforms, but these characters will have a special meaning to the linker on one of them.)

It’s rather simple in DMD to determine the identity of identifiers and types, as the latter are merged according to their mangling anyway. Qualified names and their associated symbols turned out to introduce a number of complications, though. Namely, mangleFunc in core.demangle allows building a mangled name of a function from a fully qualified name given as a string function argument and a type specified as a template argument. Implementing this for run-time usage requires copying the full mangling machinery and introspection capabilities of the compiler, which is unrealistic. Considering the limited benefit shown by the above Phobos statistics, the idea of encoding qualified names was dropped.

Here are some details about the new mangling:

  • Back references are now encoded by the character Q, followed by the relative position of the original appearance of the same identifier or type. These positions are encoded with respect to base 26, with the last digit encoded by a lowercase letter and the other digits encoded by an uppercase letter. That way, most back references are 2 or 3 characters long, 4 in extreme cases. Using a different encoding for the last digit allows determining the end of a number without looking at the next character. This helps to avoid ambiguities. (The Itanium C++ ABI mangling uses base 36 encoding by combining numbers and letters, but need a termination character _.)
  • Counting encodable entities as in the C++ mangling would result in slightly shorter mangled names, but needs the mangler to keep a dynamic list of respective positions. The current demangler is designed not to allocate as long as the supplied output buffer is large enough.
  • Relative positions are chosen instead of absolute positions to allow prepending the _D prefix without having to re-encode the symbol. Some platforms also prepend an additional underscore, for which the relative positions are agnostic.
  • The mangling grammar sometimes allows types and identifiers at the same position, so a demangler needs to distinguish between the two even if given by a back reference. That’s why a lookup to the referenced position is necessary to continue demangling; an identifier always starts with a number, while a type always starts with a letter.
  • Using Q for back references grabs the last free letter used to encode types, but there is at least one type defined in the mangling grammar that is not supposed to appear in a mangling anyway (namely TypeIdent), so it can be resurrected if the necessity appears.

For example, the expression template type shown above now mangles as

pragma(msg, len.square.mangleof);
// S4expr__T3MulTSQo__TQlTAyaTQeZQvTQtZQBb
//                ^^   ^^     ^^ ^^ ^^ ^^^ decode to:
//                |    |      |  |  |  |
//                |    |      |  |  |  +- 3Mul
//                |    |      |  |  +---- S4expr__T3MulTAyaTAyaZ3Mul
//                |    |      |  +------- 3Mul
//                |    |      +---------- Aya
//                |    +----------------- 3Mul
//                +---------------------- 4expr

with a length of 39 instead of 78 without back references. The resulting sizes are 23, 39, 57, 76, 95, 114, 133 showing linear growth. The chain of 12 calls to square shrinks from 207,114 characters to 247, i.e. by a factor of more than 800.

Implementing mangleFunc mentioned above for the mangling with back referencing identifiers still is not obvious; while the fully qualified name is not supposed to contain any types (e.g. as a struct template argument) identifiers in the mangled name can appear again in the function type. This was solved by extending the demangler to use “Design by Introspection” (DbI) (as coined by Andrei Alexandrescu):

  • make the Demangle struct a template that parameterizes on a struct that supplies a couple of hooks
    struct NoHooks {}  // supports: static bool parseLName(ref Demangle); ...
    private struct Demangle(Hooks = NoHooks)
    Hooks hooks;
        // ...
        void parseLName()
            static if(__traits(hasMember, Hooks, "parseLName"))
                if (hooks.parseLName(this))
                // normal decode...
  • create a hook that replaces a reoccurring identifier with the appropriate back reference
    struct RemangleHooks
        char[] result;
        size_t[const(char)[]] idpos;
        // ...
        bool parseLName(ref Demangler!RemangleHooks d)
            // flush input so far to result[]
            if (d.front == 'Q')
                // re-encode back reference...
            else if (auto ppos = currentIdentifier in idpos)
                // encode back reference to identifier at *ppos
                idpos[currentIdentifier] = currentPos;
            return true;
  • combine the qualified name and the type as before (core.demangle is still capable of decoding it) and run it through the hooked demangler
    char[] mangleFunc(FuncType)(const(char)[] qualifiedName)
        const(char)mangledQualifiedName = encodeLNames(qualifiedName);
        const(char)mangled = mangledQualifiedName ~ FuncType.mangleof;
        auto d = Demangle!RemangleHooks(mangled, null);
        d.mute = true; // no demangled output
        return d.hooks.result;

Is the new mangling sound?

The back references encoded into the mangling extend the existing mangling. Unfortunately, the latter had ambiguities reported to the D issue tracking system, with more of these likely yet to be uncovered. The demangler in core.demangle rejected about 3% of the unmodified symbols from the Phobos unit tests, while 15% were demangled only partially.

It’s tough to verify the soundness of an addition to an already complex and fragile definition, as a change to the mangling would need an update to the tooling (demangler, debuggers). Anyway, it was a good opportunity to get rid of these, too.

So scrutiny of the existing definition was required. To do this mechanically, the mangling specification from the web site was converted into a grammar digestible by the bison parser generator. Bison can create LALR(1) parser tables, which basically means that, while scanning a mangled symbol, looking at a character and its successor is enough to determine whether the character adds to a previous entity or starts a new one. When conflicts are reported when processing a grammar, they might be resolvable with a larger context, but they can also hint at actual problems or undesirable complexity. Adding pseudo-tokens representing handcrafted parser control flow can avoid these conflicts.

This gist shows a grammar for the D mangling scheme without the back references. It still has a couple of conflicts when run through Bison, one of which was determined to be an actual ambiguity in the definition. Adding back references to
the grammar doesn’t add any conflicts.

In addition, core.demangle was fixed to work for all symbols but those exposing the known ambiguities.


Some of the implementations in std.traits used the mangling of a symbol to introspect compile-time properties, for example, to determine the linkage. This was done using a simplified demangler. With the introduction of back references, these
didn’t work any more except for simple symbol names. Using a solution as for core.mangleFunc is feasible, but can slow down compilation considerably as the demangling needs to be executed via CTFE. Fortunately, new __traits have been added which cover all information that can be found in the mangling.

While most users will not notice any changes to their programs other than smaller object and executable file sizes, the new mangling can be a breaking change to external tools like the linker or a debugger. These will continue to work, but until they are updated, be prepared to eventually see the new mangled names for a little while instead of the demangled ones.

Thanks to Mike Parker, Walter Bright and Steven Schveighoffer for review.

DMD 2.077.0 Released

Posted on

The D Language Foundation is happy to announce DMD 2.077.0. This latest release of the reference compiler for the D programming language is available from the Downloads page. Among the usual slate of bug and regression fixes, this release brings a couple of particulary beneficial enhancements that will have an immediate impact on some existing projects.

Cutting symbol bloat

Thanks to Rainer Schütze, the compiler now produces significantly smaller mangled names in situations where they had begun to get out of control, particularly in the case of IFTI (Implicit Function Template Instantiation) where Voldemort types are involved. That may call for a bit of a detour here.

The types that shall not be named

Voldemort types are perhaps one of D’s more interesting features. They look like this:

auto getHeWhoShallNotBeNamed() 
    struct NoName 
        void castSpell() 
            import std.stdio : writeln;
    return NoName();

void main() 
    auto voldemort = getHeWhoShallNotBeNamed();

Here we have an auto function, a function for which the return type is inferred, returning an instance of a type declared inside the function. It’s possible to access public members on the instance even though its type can never be named outside of the function where it was declared. Coupled with type inference in variable declarations, it’s possible to store the returned instance and reuse it. This serves as an extra level of encapsulation where it’s desired.

In D, for any given API, as far as the world outside of a module is concerned, module private is the lowest level of encapsulation.

module foobar;

private struct Foo
    int x;

struct Bar 
    private int y;
    int z;

Here, the type Foo is module private. Bar is shown here for completeness, as those new to D are often surprised to learn that private members of an aggregate type are also module private (D’s equivalent of the C++ friend relationship). There is no keyword that indicates a lower level of encapsulation.

Sometimes you just may not want Foo to be visible to the entire module. While it’s true that anyone making a breaking change to Foo’s interface also has access to the parts of the module that break (which is the rationale behind module-private members), there are times when you may not want the entire module to have access to Foo at all. Voldemort types fill that role of hiding details not just from the world, but from the rest of the module.

The evil side of Voldemort types

One unforeseen consequence of Voldemort types that was first reported in mid–2016 was that, when used in templated functions, they caused a serious explosion in the size of the mangled function names (in some cases up to 1 MB!), making for some massive object files. There was a good bit of forum discussion on how to trim them down, with a number of ideas tossed around. Ultimately, Rainer Schütze took it on. His persistence has resulted in shorter mangled names all around, but the wins are particularly impressive when it comes to IFTI and Voldemort types. (Rainer is also the maintainer of Visual D, the D programming language plugin for Visual Studio)

D’s name-mangling scheme is detailed in the ABI documentation. The description of the new enhancement is in the section titled ‘Back references’.

Improved vectorization

D has long supported array operations such as element-wise addtion, multiplication, etc. For example:

int[] arr1 = [0, 1, 2];
int[] arr2 = [3, 4, 5];
int[3] arr3 = arr1[] + arr2[];
assert(arr3 == [3, 5, 7]);

In some cases, such operations could be vectorized. The reason it was some cases and not all cases is because dedicated assembly routines were used to achieve the vectorization and they weren’t implemented for every case.

With 2.077.0, that’s no longer true. Vectorization is now templated so that all array operations benefit. Any codebase out there using array operations that were not previously vectorized can expect a sizable performance increase for those operations thanks to the increased throughput (though whether an application benefits overall is of course context-dependent). How the benefit is received depends on the compiler being used. From the changelog:

For GDC/LDC the implementation relies on auto-vectorization, for DMD the implementation performs the vectorization itself. Support for vector operations with DMD is determined statically (-mcpu=native, -mcpu=avx2) to avoid binary bloat and the small test overhead. DMD enables SSE2 for 64-bit targets by default.

Note that the changelog initially showed -march instead of -mcpu in the quoted lines, and the updated version had not yet been posted when this announcement was published.

DMD’s implementation is implemented in terms of core.simd, which is also part of DRuntime’s public API.

The changelog also notes that there’s a potential for division performed on float arrays in existing code to see a performance decrease in exchange for an increase in precision.

The implementation no longer weakens floating point divisions (e.g. ary[] / scalar) to multiplication (ary[] * (1.0 / scalar)) as that may reduce precision. To preserve the higher performance of float multiplication when loss of precision is acceptable, use either -ffast-math with GDC/LDC or manually rewrite your code to multiply by (1.0 / scalar) for DMD.

Other assorted treats

Just the other day, someone asked in the forums if DMD supports reproducible builds. As of 2.077.0, the answer is affirmative. DMD now ensures that compilation is deterministic such that given the same source code and the same compiler version, the binaries produced will be identical. If this is important to you, be sure not to use any of the non-determistic lexer tokens (__DATE__, __TIME__, and __TIMESTAMP__) in your code.

DMD’s -betterC command line option gets some more love in this release. When it’s enabled, DRuntime is not available. Library authors can now use the predefined version D_BetterC to determine when that is the case so that, where it’s feasible, they can more conveniently support applications with and without the runtime. Also, the option’s behavior is now documented, so it’s no longer necessary to go to the forums or parse through search results to figure out what is and isn’t actually supported in BetterC mode.

The entire changelog is, as always, available at

DCompute: Running D on the GPU

Posted on

Nicholas Wilson is a student at Murdoch University, studying for his BEng (Hons)/BSc in Industrial Computer Systems (Hons) and Instrumentation & Control/ Molecular Biology & Genetics and Biomedical Science. He just finished his thesis on low-cost defect detection of solar cells by electroluminescence imaging, which gives him time to work on DCompute and write about it for the D Blog.He plays the piano, ice skates, and has spent 7 years putting D to use on number bashing, automation, and anything else that he could make a computer do for him.

DCompute is a framework and compiler extension to support writing native kernels for OpenCL and CUDA in D to utilize GPUs and other accelerators for computationally intensive code. Its compute API drivers automate the interactions between user code and the tedious and error prone APIs with the goal of enabling the rapid development of high performance D libraries and applications.


This is the second article on DCompute. In the previous article, we looked at the development of DCompute and some trivial examples. While we were able to successfully build kernels, there was no way to run them short of using them with an existing framework or doing everything yourself. This is no longer the case. As of v0.1.0, DCompute now comes with native wrappers for both OpenCL and CUDA, enabling kernel dispatch as easily as CUDA.

In order to run a kernel we need to pass it off to the appropriate compute API, either CUDA or OpenCL. While these APIs both try to achieve similar things they are different enough that to squeeze that last bit of performance out of them you need to treat each API separately. But there is sufficient overlap that we can make the interface reasonably consistent between the two. The C bindings to these APIs, however, are very low level and trying to use them is very tedious and extremely prone to error (yay void*).
In addition to the tedium and error proneness, you have to redundantly specify a lot of information, which further compounds the problem. Fortunately this is D and we can remove a lot of the redundancy through introspection and code generation.

The drivers wrap the C API, providing a clean and consistent interface that’s easy to use. While the documentation is a little sparse at the moment, the source code is for the most part straightforward (if you’re familiar with the C APIs, looking where a function is used is a good place to start). There is the occasional piece of magic to achieve a sane API.

Taming the beasts

OpenCL’s clGet*Info functions are the way to access properties of the class hidden behind the void*. A typical call looks like

cl_foo* foo = ...; 
cl_int refCount;
clGetFooInfo(foo, CL_FOO_REFERENCE_COUNT, refCount.sizeof, &refCount,null);

And that’s not even one for which you have to call, to figure out how much memory you need to allocate, then call again with the allocated buffer (and $DEITY help you if you want to get a cl_program’s binaries).

Using D, I have been able to turn that into this:

struct Foo
    void* raw;
    static struct Info
        @(0x1234) int referenceCount;
    mixin(generateGetInfo!(Info, clGetFooInfo));

Foo foo  = ...;
int refCount = foo.referenceCount;

All the magic is in generateGetInfo to generate a property for each member in Foo.Info, enabling much better scalability and bonus documentation.

CUDA also has properties exposed in a similar manner, however they are not essential (unlike OpenCL) for getting things done so their development has been deferred.

Launching a kernel is a large point of pain when dealing with the C API of both OpenCL and (only marginally less horrible) CUDA, due to the complete lack of type safety and having to use the & operator into a void* far too much. In DCompute this incantation simply becomes

Event e = q.enqueue!(saxpy)([N])(b_res, alpha, b_x, b_y, N);

for OpenCL (1D with N work items), and

q.enqueue!(saxpy)([N, 1, 1], [1 ,1 ,1])(b_res, alpha, b_x, b_y, N);

for CUDA (equivalent to saxpy<<<N,1,0,q>>>(b_res,alpha,b_x,b_y, N);)

Where q is a queue, N is the length of buffers (b_res, b_x & b_y) and saxpy (single-precision a x plus y) is the kernel in this example. A full example may be found here, along with the magic that drives the OpenCL and CUDA enqueue functions.

The future of DCompute

While DCompute is functional, there is still much to do. The drivers still need some polish and user testing, and I need to set up continuous integration. A driver that unifies the different compute APIs is also in the works so that we can be even more cross-platform than the industry cross-platform standard.

Being able to convert SPIR-V into SPIR would enable targeting cl_khr_spir-capable 1.x and 2.0 CL implementations, dramatically increasing the number of devices that can run D kernel code (there’s nothing stopping you using the OpenCL driver for other kernels though).

On the compiler side of things, supporting OpenCL image and CUDA texture & surface operations in LDC would increase the applicability of the kernels that could be written.
I currently maintain a forward-ported fork of Khronos’s SPIR-V LLVM to generate SPIR-V from LLVM IR. I plan to use IWOCL to coordinate efforts to merge it into the LLVM trunk, and in doing so, remove the need for some of the hacks in place to deal with the oddities of the SPIR-V backend.

Using DCompute in your projects

If you want to use DCompute, you’ll need a recent LDC built against LLVM with the NVPTX (for CUDA) and/or SPIRV (for OpenCL 2.1+) targets enabled and should add "dcompute": "~>0.1.0" to your dub.json. LDC 1.4+ releases have NVPTX enabled. If you want to target OpenCL, you’ll need to build LDC yourself against my fork of LLVM.