Function Pointers

Summary

This proposal provides language constructs that expose IL opcodes that cannot currently be accessed efficiently, or at all, in C# today: ldftn and calli. These IL opcodes can be important in high performance code and developers need an efficient way to access them.

Motivation

The motivations and background for this feature are described in the following issue (as is a potential implementation of the feature):

https://github.com/dotnet/csharplang/issues/191

This is an alternate design proposal to compiler intrinsics

Detailed Design

Function pointers

The language will allow for the declaration of function pointers using the delegate* syntax. The full syntax is described in detail in the next section but it is meant to resemble the syntax used by Func and Action type declarations.

unsafe class Example {
    void Example(Action<int> a, delegate*<int, void> f) {
        a(42);
        f(42);
    }
}

These types are represented using the function pointer type as outlined in ECMA-335. This means invocation of a delegate* will use calli where invocation of a delegate will use callvirt on the Invoke method. Syntactically though invocation is identical for both constructs.

The ECMA-335 definition of method pointers includes the calling convention as part of the type signature (section 7.1). The default calling convention will be managed. Unmanaged calling conventions can by specified by putting an unmanaged keyword afer the delegate* syntax, which will use the runtime platform default. Specific unmanaged conventions can then be specified in brackets to the unmanaged keyword by specifying any type starting with CallConv in the System.Runtime.CompilerServices namespace, leaving off the CallConv prefix. These types must come from the program's core library, and the set of valid combinations is platform-dependent.

//This method has a managed calling convention. This is the same as leaving the managed keyword off.
delegate* managed<int, int>;

// This method will be invoked using whatever the default unmanaged calling convention on the runtime
// platform is. This is platform and architecture dependent and is determined by the CLR at runtime.
delegate* unmanaged<int, int>;

// This method will be invoked using the cdecl calling convention
// Cdecl maps to System.Runtime.CompilerServices.CallConvCdecl
delegate* unmanaged[Cdecl] <int, int>;

// This method will be invoked using the stdcall calling convention, and suppresses GC transition
// Stdcall maps to System.Runtime.CompilerServices.CallConvStdcall
// SuppressGCTransition maps to System.Runtime.CompilerServices.CallConvSuppressGCTransition
delegate* unmanaged[Stdcall, SuppressGCTransition] <int, int>;

Conversions between delegate* types is done based on their signature including the calling convention.

unsafe class Example {
    void Conversions() {
        delegate*<int, int, int> p1 = ...;
        delegate* managed<int, int, int> p2 = ...;
        delegate* unmanaged<int, int, int> p3 = ...;

        p1 = p2; // okay p1 and p2 have compatible signatures
        Console.WriteLine(p2 == p1); // True
        p2 = p3; // error: calling conventions are incompatible
    }
}

A delegate* type is a pointer type which means it has all of the capabilities and restrictions of a standard pointer type:

  • Only valid in an unsafe context.
  • Methods which contain a delegate* parameter or return type can only be called from an unsafe context.
  • Cannot be converted to object.
  • Cannot be used as a generic argument.
  • Can implicitly convert delegate* to void*.
  • Can explicitly convert from void* to delegate*.

Restrictions:

  • Custom attributes cannot be applied to a delegate* or any of its elements.
  • A delegate* parameter cannot be marked as params
  • A delegate* type has all of the restrictions of a normal pointer type.
  • Pointer arithmetic cannot be performed directly on function pointer types.

Function pointer syntax

The full function pointer syntax is represented by the following grammar:

pointer_type
    : ...
    | funcptr_type
    ;

funcptr_type
    : 'delegate' '*' calling_convention_specifier? '<' funcptr_parameter_list funcptr_return_type '>'
    ;

calling_convention_specifier
    : 'managed'
    | 'unmanaged' ('[' unmanaged_calling_convention ']')?
    ;

unmanaged_calling_convention
    : 'Cdecl'
    | 'Stdcall'
    | 'Thiscall'
    | 'Fastcall'
    | identifier (',' identifier)*
    ;

funptr_parameter_list
    : (funcptr_parameter ',')*
    ;

funcptr_parameter
    : funcptr_parameter_modifier? type
    ;

funcptr_return_type
    : funcptr_return_modifier? return_type
    ;

funcptr_parameter_modifier
    : 'ref'
    | 'out'
    | 'in'
    ;

funcptr_return_modifier
    : 'ref'
    | 'ref readonly'
    ;

If no calling_convention_specifier is provided, the default is managed. The precise metadata encoding of the calling_convention_specifier and what identifiers are valid in the unmanaged_calling_convention is covered in Metadata Representation of Calling Conventions.

delegate int Func1(string s);
delegate Func1 Func2(Func1 f);

// Function pointer equivalent without calling convention
delegate*<string, int>;
delegate*<delegate*<string, int>, delegate*<string, int>>;

// Function pointer equivalent with calling convention
delegate* managed<string, int>;
delegate*<delegate* managed<string, int>, delegate*<string, int>>;

Function pointer conversions

In an unsafe context, the set of available implicit conversions (Implicit conversions) is extended to include the following implicit pointer conversions:

  • Existing conversions
  • From funcptr_type F0 to another funcptr_type F1, provided all of the following are true:
    • F0 and F1 have the same number of parameters, and each parameter D0n in F0 has the same ref, out, or in modifiers as the corresponding parameter D1n in F1.
    • For each value parameter (a parameter with no ref, out, or in modifier), an identity conversion, implicit reference conversion, or implicit pointer conversion exists from the parameter type in F0 to the corresponding parameter type in F1.
    • For each ref, out, or in parameter, the parameter type in F0 is the same as the corresponding parameter type in F1.
    • If the return type is by value (no ref or ref readonly), an identity, implicit reference, or implicit pointer conversion exists from the return type of F1 to the return type of F0.
    • If the return type is by reference (ref or ref readonly), the return type and ref modifiers of F1 are the same as the return type and ref modifiers of F0.
    • The calling convention of F0 is the same as the calling convention of F1.

Allow address-of to target methods

Method groups will now be allowed as arguments to an address-of expression. The type of such an expression will be a delegate* which has the equivalent signature of the target method and a managed calling convention:

unsafe class Util {
    public static void Log() { }

    void Use() {
        delegate*<void> ptr1 = &Util.Log;

        // Error: type "delegate*<void>" not compatible with "delegate*<int>";
        delegate*<int> ptr2 = &Util.Log;
   }
}

In an unsafe context, a method M is compatible with a function pointer type F if all of the following are true:

  • M and F have the same number of parameters, and each parameter in M has the same ref, out, or in modifiers as the corresponding parameter in F.
  • For each value parameter (a parameter with no ref, out, or in modifier), an identity conversion, implicit reference conversion, or implicit pointer conversion exists from the parameter type in M to the corresponding parameter type in F.
  • For each ref, out, or in parameter, the parameter type in M is the same as the corresponding parameter type in F.
  • If the return type is by value (no ref or ref readonly), an identity, implicit reference, or implicit pointer conversion exists from the return type of F to the return type of M.
  • If the return type is by reference (ref or ref readonly), the return type and ref modifiers of F are the same as the return type and ref modifiers of M.
  • The calling convention of M is the same as the calling convention of F. This includes both the calling convention bit, as well as any calling convention flags specified in the unmanaged identifier.
  • M is a static method.

In an unsafe context, an implicit conversion exists from an address-of expression whose target is a method group E to a compatible function pointer type F if E contains at least one method that is applicable in its normal form to an argument list constructed by use of the parameter types and modifiers of F, as described in the following.

  • A single method M is selected corresponding to a method invocation of the form E(A) with the following modifications:
    • The arguments list A is a list of expressions, each classified as a variable and with the type and modifier (ref, out, or in) of the corresponding funcptr_parameter_list of F.
    • The candidate methods are only those methods that are applicable in their normal form, not those applicable in their expanded form.
    • The candidate methods are only those methods that are static.
  • If the algorithm of overload resolution produces an error, then a compile-time error occurs. Otherwise, the algorithm produces a single best method M having the same number of parameters as F and the conversion is considered to exist.
  • The selected method M must be compatible (as defined above) with the function pointer type F. Otherwise, a compile-time error occurs.
  • The result of the conversion is a function pointer of type F.

This means developers can depend on overload resolution rules to work in conjunction with the address-of operator:

unsafe class Util {
    public static void Log() { }
    public static void Log(string p1) { }
    public static void Log(int i) { };

    void Use() {
        delegate*<void> a1 = &Log; // Log()
        delegate*<int, void> a2 = &Log; // Log(int i)

        // Error: ambiguous conversion from method group Log to "void*"
        void* v = &Log;
    }

The address-of operator will be implemented using the ldftn instruction.

Restrictions of this feature:

  • Only applies to methods marked as static.
  • Non-static local functions cannot be used in &. The implementation details of these methods are deliberately not specified by the language. This includes whether they are static vs. instance or exactly what signature they are emitted with.

Operators on Function Pointer Types

The section in unsafe code on operators is modified as such:

In an unsafe context, several constructs are available for operating on all _pointer_type_s that are not _funcptr_type_s:

In an unsafe context, several constructs are available for operating on all _funcptr_type_s:

Additionally, we modify all the sections in Pointers in expressions to forbid function pointer types, except Pointer comparison and The sizeof operator.

Better function member

The better function member specification will be changed to include the following line:

A delegate* is more specific than void*

This means that it is possible to overload on void* and a delegate* and still sensibly use the address-of operator.

Metadata representation of in, out, and ref readonly parameters and return types

Function pointer signatures have no parameter flags location, so we must encode whether parameters and the return type are in, out, or ref readonly by using modreqs.

in

We reuse System.Runtime.InteropServices.InAttribute, applied as a modreq to the ref specifier on a parameter or return type, to mean the following:

  • If applied to a parameter ref specifier, this parameter is treated as in.
  • If applied to the return type ref specifier, the return type is treated as ref readonly.

out

We use System.Runtime.InteropServices.OutAttribute, applied as a modreq to the ref specifier on a parameter type, to mean that the parameter is an out parameter.

Errors

  • It is an error to apply OutAttribute as a modreq to a return type.
  • It is an error to apply both InAttribute and OutAttribute as a modreq to a parameter type.
  • If either are specified via modopt, they are ignored.

Metadata Representation of Calling Conventions

Calling conventions are encoded in a method signature in metadata by a combination of the CallKind flag in the signature and zero or more modopts at the start of the signature. ECMA-335 currently declares the following elements in the CallKind flag:

CallKind
   : default
   | unmanaged cdecl
   | unmanaged fastcall
   | unmanaged thiscall
   | unmanaged stdcall
   | varargs
   ;

Of these, function pointers in C# will support all but varargs.

In addition, the runtime (and eventually 335) will be updated to include a new CallKind on new platforms. This does not have a formal name currently, but this document will use unmanaged ext as a placeholder to stand for the new extensible calling convention format. With no modopts, unmanaged ext is the platform default calling convention, unmanaged without the square brackets.

Mapping the calling_convention_specifier to a CallKind

A calling_convention_specifier that is omitted, or specified as managed, maps to the default CallKind. This is default CallKind of any method not attributed with UnmanagedCallersOnly.

C# recognizes 4 special identifiers that map to specific existing unmanaged CallKinds from ECMA 335. In order for this mapping to occur, these identifiers must be specified on their own, with no other identifiers, and this requirement is encoded into the spec for unmanaged_calling_conventions. These identifiers are Cdecl, Thiscall, Stdcall, and Fastcall, which correspond to unmanaged cdecl, unmanaged thiscall, unmanaged stdcall, and unmanaged fastcall, respectively. If more than one identifer is specified, or the single identifier is not of the specially recognized identifiers, we perform special name lookup on the identifier with the following rules:

  • We prepend the identifier with the string CallConv
  • We look only at types defined in the System.Runtime.CompilerServices namespace.
  • We look only at types defined in the core library of the application, which is the library that defines System.Object and has no dependencies.
  • We look only at public types.

If lookup succeeds on all of the identifiers specified in an unmanaged_calling_convention, we encode the CallKind as unmanaged ext, and encode each of the resolved types in the set of modopts at the beginning of the function pointer signature. As a note, these rules mean that users cannot prefix these identifiers with CallConv, as that will result in looking up CallConvCallConvVectorCall.

When interpreting metadata, we first look at the CallKind. If it is anything other than unmanaged ext, we ignore all modopts on the return type for the purposes of determining the calling convention, and use only the CallKind. If the CallKind is unmanaged ext, we look at the modopts at the start of the function pointer type, taking the union of all types that meet the following requirements:

  • The is defined in the core library, which is the library that references no other libraries and defines System.Object.
  • The type is defined in the System.Runtime.CompilerServices namespace.
  • The type starts with the prefix CallConv.
  • The type is public.

These represent the types that must be found when performing lookup on the identifiers in an unmanaged_calling_convention when defining a function pointer type in source.

It is an error to attempt to use a function pointer with a CallKind of unmanaged ext if the target runtime does not support the feature. This will be determined by looking for the presence of the System.Runtime.CompilerServices.RuntimeFeature.UnmanagedCallKind constant. If this constant is present, the runtime is considered to support the feature.

System.Runtime.InteropServices.UnmanagedCallersOnlyAttribute

System.Runtime.InteropServices.UnmanagedCallersOnlyAttribute is an attribute used by the CLR to indicate that a method should be called with a specific calling convention. Because of this, we introduce the following support for working with the attribute:

  • It is an error to directly call a method annotated with this attribute from C#. Users must obtain a function pointer to the method and then invoke that pointer.
  • It is an error to apply the attribute to anything other than an ordinary static method or ordinary static local function. The C# compiler will mark any non-static or static non-ordinary methods imported from metadata with this attribute as unsupported by the language.
  • It is an error for a method marked with the attribute to have a parameter or return type that is not an unmanaged_type.
  • It is an error for a method marked with the attribute to have type parameters, even if those type parameters are constrained to unmanaged.
  • It is an error for a method in a generic type to be marked with the attribute.
  • It is an error to convert a method marked with the attribute to a delegate type.
  • It is an error to specify any types for UnmanagedCallersOnly.CallConvs that do not meet the requirements for calling convention modopts in metadata.

When determining the calling convention of a method marked with a valid UnmanagedCallersOnly attribute, the compiler performs the following checks on the types specified in the CallConvs property to determine the effective CallKind and modopts that should be used to determine the calling convention:

  • If no types are specified, the CallKind is treated as unmanaged ext, with no calling convention modopts at the start of the function pointer type.
  • If there is one type specified, and that type is named CallConvCdecl, CallConvThiscall, CallConvStdcall, or CallConvFastcall, the CallKind is treated as unmanaged cdecl, unmanaged thiscall, unmanaged stdcall, or unmanaged fastcall, respectively, with no calling convention modopts at the start of the function pointer type.
  • If multiple types are specified or the single type is not named one of the specially called out types above, the CallKind is treated as unmanaged ext, with the union of the types specified treated as modopts at the start of the function pointer type.

The compiler then looks at this effective CallKind and modopt collection and uses normal metadata rules to determine the final calling convention of the function pointer type.

Open Questions

Detecting runtime support for unmanaged ext

https://github.com/dotnet/runtime/issues/38135 tracks adding this flag. Depending on the feedback from review, we will either use the property specified in the issue, or use the presence of UnmanagedCallersOnlyAttribute as the flag that determines whether the runtimes supports unmanaged ext.

Considerations

Allow instance methods

The proposal could be extended to support instance methods by taking advantage of the EXPLICITTHIS CLI calling convention (named instance in C# code). This form of CLI function pointers puts the this parameter as an explicit first parameter of the function pointer syntax.

unsafe class Instance {
    void Use() {
        delegate* instance<Instance, string> f = &ToString;
        f(this);
    }
}

This is sound but adds some complication to the proposal. Particularly because function pointers which differed by the calling convention instance and managed would be incompatible even though both cases are used to invoke managed methods with the same C# signature. Also in every case considered where this would be valuable to have there was a simple work around: use a static local function.

unsafe class Instance {
    void Use() {
        static string toString(Instance i) = i.ToString();
        delgate*<Instance, string> f = &toString;
        f(this);
    }
}

Don't require unsafe at declaration

Instead of requiring unsafe at every use of a delegate*, only require it at the point where a method group is converted to a delegate*. This is where the core safety issues come into play (knowing that the containing assembly cannot be unloaded while the value is alive). Requiring unsafe on the other locations can be seen as excessive.

This is how the design was originally intended. But the resulting language rules felt very awkward. It's impossible to hide the fact that this is a pointer value and it kept peeking through even without the unsafe keyword. For example the conversion to object can't be allowed, it can't be a member of a class, etc ... The C# design is to require unsafe for all pointer uses and hence this design follows that.

Developers will still be capable of presenting a safe wrapper on top of delegate* values the same way that they do for normal pointer types today. Consider:

unsafe struct Action {
    delegate*<void> _ptr;

    Action(delegate*<void> ptr) => _ptr = ptr;
    public void Invoke() => _ptr();
}

Using delegates

Instead of using a new syntax element, delegate*, simply use existing delegate types with a * following the type:

Func<object, object, bool>* ptr = &object.ReferenceEquals;

Handling calling convention can be done by annotating the delegate types with an attribute that specifies a CallingConvention value. The lack of an attribute would signify the managed calling convention.

Encoding this in IL is problematic. The underlying value needs to be represented as a pointer yet it also must:

  1. Have a unique type to allow for overloads with different function pointer types.
  2. Be equivalent for OHI purposes across assembly boundaries.

The last point is particularly problematic. This mean that every assembly which uses Func<int>* must encode an equivalent type in metadata even though Func<int>* is defined in an assembly though don't control. Additionally any other type which is defined with the name System.Func<T> in an assembly that is not mscorlib must be different than the version defined in mscorlib.

One option that was explored was emitting such a pointer as mod_req(Func<int>) void*. This doesn't work though as a mod_req cannot bind to a TypeSpec and hence cannot target generic instantiations.

Named function pointers

The function pointer syntax can be cumbersome, particularly in complex cases like nested function pointers. Rather than have developers type out the signature every time the language could allow for named declarations of function pointers as is done with delegate.

func* void Action();

unsafe class NamedExample {
    void M(Action a) {
        a();
    }
}

Part of the problem here is the underlying CLI primitive doesn't have names hence this would be purely a C# invention and require a bit of metadata work to enable. That is doable but is a significant about of work. It essentially requires C# to have a companion to the type def table purely for these names.

Also when the arguments for named function pointers were examined we found they could apply equally well to a number of other scenarios. For example it would be just as convenient to declare named tuples to reduce the need to type out the full signature in all cases.

(int x, int y) Point;

class NamedTupleExample {
    void M(Point p) {
        Console.WriteLine(p.x);
    }
}

After discussion we decided to not allow named declaration of delegate* types. If we find there is significant need for this based on customer usage feedback then we will investigate a naming solution that works for function pointers, tuples, generics, etc ... This is likely to be similar in form to other suggestions like full typedef support in the language.

Future Considerations

static delegates

This refers to the proposal to allow for the declaration of delegate types which can only refer to static members. The advantage being that such delegate instances can be allocation free and better in performance sensitive scenarios.

If the function pointer feature is implemented the static delegate proposal will likely be closed out. The proposed advantage of that feature is the allocation free nature. However recent investigations have found that is not possible to achieve due to assembly unloading. There must be a strong handle from the static delegate to the method it refers to in order to keep the assembly from being unloaded out from under it.

To maintain every static delegate instance would be required to allocate a new handle which runs counter to the goals of the proposal. There were some designs where the allocation could be amortized to a single allocation per call-site but that was a bit complex and didn't seem worth the trade off.

That means developers essentially have to decide between the following trade offs:

  1. Safety in the face of assembly unloading: this requires allocations and hence delegate is already a sufficient option.
  2. No safety in face of assembly unloading: use a delegate*. This can be wrapped in a struct to allow usage outside an unsafe context in the rest of the code.