LLVM Coding Standards¶

Introduction
Languages, Libraries, and Standards
Mechanical Source Issues
- Source Code Formatting
- Language and Compiler Issues
Style Issues
See Also

Introduction ¶

This document describes coding standards that are used in the LLVM project. Although no coding standards should be regarded as absolute requirements to be followed in all instances, coding standards are particularly important for large-scale code bases that follow a library-based design (like LLVM).

While this document may provide guidance for some mechanical formatting issues, whitespace, or other “microscopic details”, these are not fixed standards. Always follow the golden rule:

If you are extending, enhancing, or bug fixing already implemented code, use the style that is already being used so that the source is uniform and easy to follow.

Note that some code bases (e.g. libc++) have special reasons to deviate from the coding standards. For example, in the case of libc++, this is because the naming and other conventions are dictated by the C++ standard.

There are some conventions that are not uniformly followed in the code base (e.g. the naming convention). This is because they are relatively new, and a lot of code was written before they were put in place. Our long term goal is for the entire codebase to follow the convention, but we explicitly do not want patches that do large-scale reformatting of existing code. On the other hand, it is reasonable to rename the methods of a class if you’re about to change it in some other way. Please commit such changes separately to make code review easier.

The ultimate goal of these guidelines is to increase the readability and maintainability of our common source base.

Languages, Libraries, and Standards ¶

Most source code in LLVM and other LLVM projects using these coding standards is C++ code. There are some places where C code is used either due to environment restrictions, historical restrictions, or due to third-party source code imported into the tree. Generally, our preference is for standards conforming, modern, and portable C++ code as the implementation language of choice.

C++ Standard Versions ¶

Unless otherwise documented, LLVM subprojects are written using standard C++14 code and avoid unnecessary vendor-specific extensions.

Nevertheless, we restrict ourselves to features which are available in the major toolchains supported as host compilers (see Getting Started with the LLVM System page, section Software).

Each toolchain provides a good reference for what it accepts:

C++ Standard Library ¶

Use the C++ standard library facilities whenever they are available for a particular task. LLVM and related projects emphasize and rely on the standard library facilities as much as possible.

We avoid some standard facilities, like the I/O streams, and instead use LLVM’s streams library (raw_ostream). More detailed information on these subjects is available in the LLVM Programmer’s Manual.

LLVM support libraries (for example, ADT) implement functionality missing in the standard library. Such libraries are usually implemented in the llvm namespace and follow the expected standard interface, when there is one.

Guidelines for Go code ¶

Any code written in the Go programming language is not subject to the formatting rules below. Instead, we adopt the formatting rules enforced by the gofmt tool.

Go code should strive to be idiomatic. Two good sets of guidelines for what this means are Effective Go and Go Code Review Comments.

Mechanical Source Issues ¶

Source Code Formatting ¶

Commenting ¶

Comments are important for readability and maintainability. When writing comments, write them as English prose, using proper capitalization, punctuation, etc. Aim to describe what the code is trying to do and why, not how it does it at a micro level. Here are a few important things to document:

File Headers ¶

Every source file should have a header on it that describes the basic purpose of the file. The standard header looks like this:

//===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
///
/// \file
/// This file contains the declaration of the Instruction class, which is the
/// base class for all of the VM instructions.
///
//===----------------------------------------------------------------------===//

A few things to note about this particular format: The “-*- C++ -*-” string on the first line is there to tell Emacs that the source file is a C++ file, not a C file (Emacs assumes .h files are C files by default).

Note

This tag is not necessary in .cpp files. The name of the file is also on the first line, along with a very short description of the purpose of the file.

The next section in the file is a concise note that defines the license that the file is released under. This makes it perfectly clear what terms the source code can be distributed under and should not be modified in any way.

The main body is a Doxygen comment (identified by the /// comment marker instead of the usual //) describing the purpose of the file. The first sentence (or a passage beginning with \brief) is used as an abstract. Any additional information should be separated by a blank line. If an algorithm is based on a paper or is described in another source, provide a reference.

Class overviews ¶

Classes are a fundamental part of an object-oriented design. As such, a class definition should have a comment block that explains what the class is used for and how it works. Every non-trivial class is expected to have a doxygen comment block.

Method information ¶

Methods and global functions should also be documented. A quick note about what it does and a description of the edge cases is all that is necessary here. The reader should be able to understand how to use interfaces without reading the code itself.

Good things to talk about here are what happens when something unexpected happens, for instance, does the method return null?

Comment Formatting ¶

In general, prefer C++-style comments (// for normal comments, /// for doxygen documentation comments). There are a few cases when it is useful to use C-style (/* */) comments however:

When writing C code to be compatible with C89.
When writing a header file that may be #included by a C source file.
When writing a source file that is used by a tool that only accepts C-style comments.
When documenting the significance of constants used as actual parameters in a call. This is most helpful for bool parameters, or passing 0 or nullptr. The comment should contain the parameter name, which ought to be meaningful. For example, it’s not clear what the parameter means in this call:
```
Object.emitName(nullptr);
```
An in-line C-style comment makes the intent obvious:
```
Object.emitName(/*Prefix=*/nullptr);
```

Commenting out large blocks of code is discouraged, but if you really have to do this (for documentation purposes or as a suggestion for debug printing), use #if 0 and #endif. These nest properly and are better behaved in general than C style comments.

Doxygen Use in Documentation Comments ¶

Use the \file command to turn the standard file header into a file-level comment.

Include descriptive paragraphs for all public interfaces (public classes, member and non-member functions). Avoid restating the information that can be inferred from the API name. The first sentence (or a paragraph beginning with \brief) is used as an abstract. Try to use a single sentence as the \brief adds visual clutter. Put detailed discussion into separate paragraphs.

To refer to parameter names inside a paragraph, use the \p name command. Don’t use the \arg name command since it starts a new paragraph that contains documentation for the parameter.

Wrap non-inline code examples in \code ... \endcode.

To document a function parameter, start a new paragraph with the \param name command. If the parameter is used as an out or an in/out parameter, use the \param [out] name or \param [in,out] name command, respectively.

To describe function return value, start a new paragraph with the \returns command.

A minimal documentation comment:

/// Sets the xyzzy property to \p Baz.
void setXyzzy(bool Baz);

A documentation comment that uses all Doxygen features in a preferred way:

/// Does foo and bar.
///
/// Does not do foo the usual way if \p Baz is true.
///
/// Typical usage:
/// \code
///   fooBar(false, "quux", Res);
/// \endcode
///
/// \param Quux kind of foo to do.
/// \param [out] Result filled with bar sequence on foo success.
///
/// \returns true on success.
bool fooBar(bool Baz, StringRef Quux, std::vector<int> &Result);

Don’t duplicate the documentation comment in the header file and in the implementation file. Put the documentation comments for public APIs into the header file. Documentation comments for private APIs can go to the implementation file. In any case, implementation files can include additional comments (not necessarily in Doxygen markup) to explain implementation details as needed.

Don’t duplicate function or class name at the beginning of the comment. For humans it is obvious which function or class is being documented; automatic documentation processing tools are smart enough to bind the comment to the correct declaration.

Avoid:

// Example.h:

// example - Does something important.
void example();

// Example.cpp:

// example - Does something important.
void example() { ... }

Preferred:

// Example.h:

/// Does something important.
void example();

// Example.cpp:

/// Builds a B-tree in order to do foo.  See paper by...
void example() { ... }

`#include` Style ¶

Immediately after the header file comment (and include guards if working on a header file), the minimal list of #includes required by the file should be listed. We prefer these #includes to be listed in this order:

Main Module Header
Local/Private Headers
LLVM project/subproject headers (clang/..., lldb/..., llvm/..., etc)
System #includes

and each category should be sorted lexicographically by the full path.

The Main Module Header file applies to .cpp files which implement an interface defined by a .h file. This #include should always be included first regardless of where it lives on the file system. By including a header file first in the .cpp files that implement the interfaces, we ensure that the header does not have any hidden dependencies which are not explicitly #included in the header, but should be. It is also a form of documentation in the .cpp file to indicate where the interfaces it implements are defined.

LLVM project and subproject headers should be grouped from most specific to least specific, for the same reasons described above. For example, LLDB depends on both clang and LLVM, and clang depends on LLVM. So an LLDB source file should include lldb headers first, followed by clang headers, followed by llvm headers, to reduce the possibility (for example) of an LLDB header accidentally picking up a missing include due to the previous inclusion of that header in the main source file or some earlier header file. clang should similarly include its own headers before including llvm headers. This rule applies to all LLVM subprojects.

Source Code Width ¶

Write your code to fit within 80 columns.

There must be some limit to the width of the code in order to allow developers to have multiple files side-by-side in windows on a modest display. If you are going to pick a width limit, it is somewhat arbitrary but you might as well pick something standard. Going with 90 columns (for example) instead of 80 columns wouldn’t add any significant value and would be detrimental to printing out code. Also many other projects have standardized on 80 columns, so some people have already configured their editors for it (vs something else, like 90 columns).

Whitespace ¶

In all cases, prefer spaces to tabs in source files. People have different preferred indentation levels, and different styles of indentation that they like; this is fine. What isn’t fine is that different editors/viewers expand tabs out to different tab stops. This can cause your code to look completely unreadable, and it is not worth dealing with.

As always, follow the Golden Rule above: follow the style of existing code if you are modifying and extending it.

Do not add trailing whitespace. Some common editors will automatically remove trailing whitespace when saving a file which causes unrelated changes to appear in diffs and commits.

Format Lambdas Like Blocks Of Code ¶

When formatting a multi-line lambda, format it like a block of code. If there is only one multi-line lambda in a statement, and there are no expressions lexically after it in the statement, drop the indent to the standard two space indent for a block of code, as if it were an if-block opened by the preceding part of the statement:

std::sort(foo.begin(), foo.end(), [&](Foo a, Foo b) -> bool {
  if (a.blah < b.blah)
    return true;
  if (a.baz < b.baz)
    return true;
  return a.bam < b.bam;
});

To take best advantage of this formatting, if you are designing an API which accepts a continuation or single callable argument (be it a function object, or a std::function), it should be the last argument if at all possible.

If there are multiple multi-line lambdas in a statement, or additional parameters after the lambda, indent the block two spaces from the indent of the []:

dyn_switch(V->stripPointerCasts(),
           [] (PHINode *PN) {
             // process phis...
           },
           [] (SelectInst *SI) {
             // process selects...
           },
           [] (LoadInst *LI) {
             // process loads...
           },
           [] (AllocaInst *AI) {
             // process allocas...
           });

Braced Initializer Lists ¶

Starting from C++11, there are significantly more uses of braced lists to perform initialization. For example, they can be used to construct aggregate temporaries in expressions. They now have a natural way of ending up nested within each other and within function calls in order to build up aggregates (such as option structs) from local variables.

The historically common formatting of braced initialization of aggregate variables does not mix cleanly with deep nesting, general expression contexts, function arguments, and lambdas. We suggest new code use a simple rule for formatting braced initialization lists: act as-if the braces were parentheses in a function call. The formatting rules exactly match those already well understood for formatting nested function calls. Examples:

foo({a, b, c}, {1, 2, 3});

llvm::Constant *Mask[] = {
    llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 0),
    llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 1),
    llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 2)};

This formatting scheme also makes it particularly easy to get predictable, consistent, and automatic formatting with tools like Clang Format.

Language and Compiler Issues ¶

Treat Compiler Warnings Like Errors ¶

Compiler warnings are often useful and help improve the code. Those that are not useful, can be often suppressed with a small code change. For example, an assignment in the if condition is often a typo:

if (V = getValue()) {
  ...
}

Several compilers will print a warning for the code above. It can be suppressed by adding parentheses:

if ((V = getValue())) {
  ...
}

Write Portable Code ¶

In almost all cases, it is possible to write completely portable code. When you need to rely on non-portable code, put it behind a well-defined and well-documented interface.

Do not use RTTI or Exceptions ¶

In an effort to reduce code and executable size, LLVM does not use exceptions or RTTI (runtime type information, for example, dynamic_cast<>).

That said, LLVM does make extensive use of a hand-rolled form of RTTI that use templates like isa<>, cast<>, and dyn_cast<>. This form of RTTI is opt-in and can be added to any class.

Do not use Static Constructors ¶

Static constructors and destructors (e.g., global variables whose types have a constructor or destructor) should not be added to the code base, and should be removed wherever possible.

Globals in different source files are initialized in arbitrary order <https://yosefk.com/c++fqa/ctors.html#fqa-10.12>, making the code more difficult to reason about.

Static constructors have negative impact on launch time of programs that use LLVM as a library. We would really like for there to be zero cost for linking in an additional LLVM target or other library into an application, but static constructors undermine this goal.

Use of `class` and `struct` Keywords ¶

In C++, the class and struct keywords can be used almost interchangeably. The only difference is when they are used to declare a class: class makes all members private by default while struct makes all members public by default.

All declarations and definitions of a given class or struct must use the same keyword. For example:

// Avoid if `Example` is defined as a struct.
class Example;

// OK.
struct Example;

struct Example { ... };

struct should be used when all members are declared public.

// Avoid using `struct` here, use `class` instead.
struct Foo {
private:
  int Data;
public:
  Foo() : Data(0) { }
  int getData() const { return Data; }
  void setData(int D) { Data = D; }
};

// OK to use `struct`: all members are public.
struct Bar {
  int Data;
  Bar() : Data(0) { }
};

Do not use Braced Initializer Lists to Call a Constructor ¶

Starting from C++11 there is a “generalized initialization syntax” which allows calling constructors using braced initializer lists. Do not use these to call constructors with non-trivial logic or if you care that you’re calling some particular constructor. Those should look like function calls using parentheses rather than like aggregate initialization. Similarly, if you need to explicitly name the type and call its constructor to create a temporary, don’t use a braced initializer list. Instead, use a braced initializer list (without any type for temporaries) when doing aggregate initialization or something notionally equivalent. Examples:

class Foo {
public:
  // Construct a Foo by reading data from the disk in the whizbang format, ...
  Foo(std::string filename);

  // Construct a Foo by looking up the Nth element of some global data ...
  Foo(int N);

  // ...
};

// The Foo constructor call is reading a file, don't use braces to call it.
std::fill(foo.begin(), foo.end(), Foo("name"));

// The pair is being constructed like an aggregate, use braces.
bar_map.insert({my_key, my_value});

If you use a braced initializer list when initializing a variable, use an equals before the open curly brace:

int data[] = {0, 1, 2, 3};

Use `auto` Type Deduction to Make Code More Readable ¶

Some are advocating a policy of “almost always auto” in C++11, however LLVM uses a more moderate stance. Use auto if and only if it makes the code more readable or easier to maintain. Don’t “almost always” use auto, but do use auto with initializers like cast<Foo>(...) or other places where the type is already obvious from the context. Another time when auto works well for these purposes is when the type would have been abstracted away anyways, often behind a container’s typedef such as std::vector<T>::iterator.

Similarly, C++14 adds generic lambda expressions where parameter types can be auto. Use these where you would have used a template.

Beware unnecessary copies with `auto`¶

The convenience of auto makes it easy to forget that its default behavior is a copy. Particularly in range-based for loops, careless copies are expensive.

Use auto & for values and auto * for pointers unless you need to make a copy.

// Typically there's no reason to copy.
for (const auto &Val : Container) { observe(Val); }
for (auto &Val : Container) { Val.change(); }

// Remove the reference if you really want a new copy.
for (auto Val : Container) { Val.change(); saveSomewhere(Val); }

// Copy pointers, but make it clear that they're pointers.
for (const auto *Ptr : Container) { observe(*Ptr); }
for (auto *Ptr : Container) { Ptr->change(); }

Beware of non-determinism due to ordering of pointers ¶

In general, there is no relative ordering among pointers. As a result, when unordered containers like sets and maps are used with pointer keys the iteration order is undefined. Hence, iterating such containers may result in non-deterministic code generation. While the generated code might work correctly, non-determinism can make it harder to reproduce bugs and debug the compiler.

In case an ordered result is expected, remember to sort an unordered container before iteration. Or use ordered containers like vector/MapVector/SetVector if you want to iterate pointer keys.

Beware of non-deterministic sorting order of equal elements ¶

std::sort uses a non-stable sorting algorithm in which the order of equal elements is not guaranteed to be preserved. Thus using std::sort for a container having equal elements may result in non-determinstic behavior. To uncover such instances of non-determinism, LLVM has introduced a new llvm::sort wrapper function. For an EXPENSIVE_CHECKS build this will randomly shuffle the container before sorting. Default to using llvm::sort instead of std::sort.

Style Issues ¶

The High-Level Issues ¶

Self-contained Headers ¶

Header files should be self-contained (compile on their own) and end in .h. Non-header files that are meant for inclusion should end in .inc and be used sparingly.

All header files should be self-contained. Users and refactoring tools should not have to adhere to special conditions to include the header. Specifically, a header should have header guards and include all other headers it needs.

There are rare cases where a file designed to be included is not self-contained. These are typically intended to be included at unusual locations, such as the middle of another file. They might not use header guards, and might not include their prerequisites. Name such files with the .inc extension. Use sparingly, and prefer self-contained headers when possible.

In general, a header should be implemented by one or more .cpp files. Each of these .cpp files should include the header that defines their interface first. This ensures that all of the dependences of the header have been properly added to the header itself, and are not implicit. System headers should be included after user headers for a translation unit.

Library Layering ¶

A directory of header files (for example include/llvm/Foo) defines a library (Foo). Dependencies between libraries are defined by the LLVMBuild.txt file in their implementation (lib/Foo). One library (both its headers and implementation) should only use things from the libraries listed in its dependencies.

Some of this constraint can be enforced by classic Unix linkers (Mac & Windows linkers, as well as lld, do not enforce this constraint). A Unix linker searches left to right through the libraries specified on its command line and never revisits a library. In this way, no circular dependencies between libraries can exist.

This doesn’t fully enforce all inter-library dependencies, and importantly doesn’t enforce header file circular dependencies created by inline functions. A good way to answer the “is this layered correctly” would be to consider whether a Unix linker would succeed at linking the program if all inline functions were defined out-of-line. (& for all valid orderings of dependencies - since linking resolution is linear, it’s possible that some implicit dependencies can sneak through: A depends on B and C, so valid orderings are “C B A” or “B C A”, in both cases the explicit dependencies come before their use. But in the first case, B could still link successfully if it implicitly depended on C, or the opposite in the second case)

`#include` as Little as Possible ¶

#include hurts compile time performance. Don’t do it unless you have to, especially in header files.

But wait! Sometimes you need to have the definition of a class to use it, or to inherit from it. In these cases go ahead and #include that header file. Be aware however that there are many cases where you don’t need to have the full definition of a class. If you are using a pointer or reference to a class, you don’t need the header file. If you are simply returning a class instance from a prototyped function or method, you don’t need it. In fact, for most cases, you simply don’t need the definition of a class. And not #includeing speeds up compilation.

It is easy to try to go too overboard on this recommendation, however. You must include all of the header files that you are using — you can include them either directly or indirectly through another header file. To make sure that you don’t accidentally forget to include a header file in your module header, make sure to include your module header first in the implementation file (as mentioned above). This way there won’t be any hidden dependencies that you’ll find out about later.

Keep “Internal” Headers Private ¶

Many modules have a complex implementation that causes them to use more than one implementation (.cpp) file. It is often tempting to put the internal communication interface (helper classes, extra functions, etc) in the public module header file. Don’t do this!

If you really need to do something like this, put a private header file in the same directory as the source files, and include it locally. This ensures that your private interface remains private and undisturbed by outsiders.

Note

It’s okay to put extra implementation methods in a public class itself. Just make them private (or protected) and all is well.

Use Early Exits and `continue` to Simplify Code ¶

When reading code, keep in mind how much state and how many previous decisions have to be remembered by the reader to understand a block of code. Aim to reduce indentation where possible when it doesn’t make it more difficult to understand the code. One great way to do this is by making use of early exits and the continue keyword in long loops. Consider this code that does not use an early exit:

Value *doSomething(Instruction *I) {
  if (!I->isTerminator() &&
      I->hasOneUse() && doOtherThing(I)) {
    ... some long code ....
  }

  return 0;
}

This code has several problems if the body of the 'if' is large. When you’re looking at the top of the function, it isn’t immediately clear that this only does interesting things with non-terminator instructions, and only applies to things with the other predicates. Second, it is relatively difficult to describe (in comments) why these predicates are important because the if statement makes it difficult to lay out the comments. Third, when you’re deep within the body of the code, it is indented an extra level. Finally, when reading the top of the function, it isn’t clear what the result is if the predicate isn’t true; you have to read to the end of the function to know that it returns null.

It is much preferred to format the code like this:

Value *doSomething(Instruction *I) {
  // Terminators never need 'something' done to them because ...
  if (I->isTerminator())
    return 0;

  // We conservatively avoid transforming instructions with multiple uses
  // because goats like cheese.
  if (!I->hasOneUse())
    return 0;

  // This is really just here for example.
  if (!doOtherThing(I))
    return 0;

  ... some long code ....
}

This fixes these problems. A similar problem frequently happens in for loops. A silly example is something like this:

for (Instruction &I : BB) {
  if (auto *BO = dyn_cast<BinaryOperator>(&I)) {
    Value *LHS = BO->getOperand(0);
    Value *RHS = BO->getOperand(1);
    if (LHS != RHS) {
      ...
    }
  }
}

When you have very, very small loops, this sort of structure is fine. But if it exceeds more than 10-15 lines, it becomes difficult for people to read and understand at a glance. The problem with this sort of code is that it gets very nested very quickly. Meaning that the reader of the code has to keep a lot of context in their brain to remember what is going immediately on in the loop, because they don’t know if/when the if conditions will have elses etc. It is strongly preferred to structure the loop like this:

for (Instruction &I : BB) {
  auto *BO = dyn_cast<BinaryOperator>(&I);
  if (!BO) continue;

  Value *LHS = BO->getOperand(0);
  Value *RHS = BO->getOperand(1);
  if (LHS == RHS) continue;

  ...
}

This has all the benefits of using early exits for functions: it reduces nesting of the loop, it makes it easier to describe why the conditions are true, and it makes it obvious to the reader that there is no else coming up that they have to push context into their brain for. If a loop is large, this can be a big understandability win.

Don’t use `else` after a `return`¶

For similar reasons as above (reduction of indentation and easier reading), please do not use 'else' or 'else if' after something that interrupts control flow — like return, break, continue, goto, etc. For example:

case 'J': {
  if (Signed) {
    Type = Context.getsigjmp_bufType();
    if (Type.isNull()) {
      Error = ASTContext::GE_Missing_sigjmp_buf;
      return QualType();
    } else {
      break; // Unnecessary.
    }
  } else {
    Type = Context.getjmp_bufType();
    if (Type.isNull()) {
      Error = ASTContext::GE_Missing_jmp_buf;
      return QualType();
    } else {
      break; // Unnecessary.
    }
  }
}

It is better to write it like this:

case 'J':
  if (Signed) {
    Type = Context.getsigjmp_bufType();
    if (Type.isNull()) {
      Error = ASTContext::GE_Missing_sigjmp_buf;
      return QualType();
    }
  } else {
    Type = Context.getjmp_bufType();
    if (Type.isNull()) {
      Error = ASTContext::GE_Missing_jmp_buf;
      return QualType();
    }
  }
  break;

Or better yet (in this case) as:

case 'J':
  if (Signed)
    Type = Context.getsigjmp_bufType();
  else
    Type = Context.getjmp_bufType();

  if (Type.isNull()) {
    Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
                     ASTContext::GE_Missing_jmp_buf;
    return QualType();
  }
  break;

The idea is to reduce indentation and the amount of code you have to keep track of when reading the code.

Turn Predicate Loops into Predicate Functions ¶

It is very common to write small loops that just compute a boolean value. There are a number of ways that people commonly write these, but an example of this sort of thing is:

bool FoundFoo = false;
for (unsigned I = 0, E = BarList.size(); I != E; ++I)
  if (BarList[I]->isFoo()) {
    FoundFoo = true;
    break;
  }

if (FoundFoo) {
  ...
}

Instead of this sort of loop, we prefer to use a predicate function (which may be static) that uses early exits:

/// \returns true if the specified list has an element that is a foo.
static bool containsFoo(const std::vector<Bar*> &List) {
  for (unsigned I = 0, E = List.size(); I != E; ++I)
    if (List[I]->isFoo())
      return true;
  return false;
}
...

if (containsFoo(BarList)) {
  ...
}

There are many reasons for doing this: it reduces indentation and factors out code which can often be shared by other code that checks for the same predicate. More importantly, it forces you to pick a name for the function, and forces you to write a comment for it. In this silly example, this doesn’t add much value. However, if the condition is complex, this can make it a lot easier for the reader to understand the code that queries for this predicate. Instead of being faced with the in-line details of how we check to see if the BarList contains a foo, we can trust the function name and continue reading with better locality.

The Low-Level Issues ¶

Name Types, Functions, Variables, and Enumerators Properly ¶

Poorly-chosen names can mislead the reader and cause bugs. We cannot stress enough how important it is to use descriptive names. Pick names that match the semantics and role of the underlying entities, within reason. Avoid abbreviations unless they are well known. After picking a good name, make sure to use consistent capitalization for the name, as inconsistency requires clients to either memorize the APIs or to look it up to find the exact spelling.

In general, names should be in camel case (e.g. TextFileReader and isLValue()). Different kinds of declarations have different rules:

Type names (including classes, structs, enums, typedefs, etc) should be nouns and start with an upper-case letter (e.g. TextFileReader).
Variable names should be nouns (as they represent state). The name should be camel case, and start with an upper case letter (e.g. Leader or Boats).
Function names should be verb phrases (as they represent actions), and command-like function should be imperative. The name should be camel case, and start with a lower case letter (e.g. openFile() or isFoo()).
Enum declarations (e.g. enum Foo {...}) are types, so they should follow the naming conventions for types. A common use for enums is as a discriminator for a union, or an indicator of a subclass. When an enum is used for something like this, it should have a Kind suffix (e.g. ValueKind).
Enumerators (e.g. enum { Foo, Bar }) and public member variables should start with an upper-case letter, just like types. Unless the enumerators are defined in their own small namespace or inside a class, enumerators should have a prefix corresponding to the enum declaration name. For example, enum ValueKind { ... }; may contain enumerators like VK_Argument, VK_BasicBlock, etc. Enumerators that are just convenience constants are exempt from the requirement for a prefix. For instance:
```
enum {
  MaxSize = 42,
  Density = 12
};
```

As an exception, classes that mimic STL classes can have member names in STL’s style of lower-case words separated by underscores (e.g. begin(), push_back(), and empty()). Classes that provide multiple iterators should add a singular prefix to begin() and end() (e.g. global_begin() and use_begin()).

Here are some examples:

class VehicleMaker {
  ...
  Factory<Tire> F;            // Avoid: a non-descriptive abbreviation.
  Factory<Tire> Factory;      // Better: more descriptive.
  Factory<Tire> TireFactory;  // Even better: if VehicleMaker has more than one
                              // kind of factories.
};

Vehicle makeVehicle(VehicleType Type) {
  VehicleMaker M;                         // Might be OK if scope is small.
  Tire Tmp1 = M.makeTire();               // Avoid: 'Tmp1' provides no information.
  Light Headlight = M.makeLight("head");  // Good: descriptive.
  ...
}

Assert Liberally ¶

Use the “assert” macro to its fullest. Check all of your preconditions and assumptions, you never know when a bug (not necessarily even yours) might be caught early by an assertion, which reduces debugging time dramatically. The “<cassert>” header file is probably already included by the header files you are using, so it doesn’t cost anything to use it.

To further assist with debugging, make sure to put some kind of error message in the assertion statement, which is printed if the assertion is tripped. This helps the poor debugger make sense of why an assertion is being made and enforced, and hopefully what to do about it. Here is one complete example:

inline Value *getOperand(unsigned I) {
  assert(I < Operands.size() && "getOperand() out of range!");
  return Operands[I];
}

Here are more examples:

assert(Ty->isPointerType() && "Can't allocate a non-pointer type!");

assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");

assert(idx < getNumSuccessors() && "Successor # out of range!");

assert(V1.getType() == V2.getType() && "Constant types must be identical!");

assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");

You get the idea.

In the past, asserts were used to indicate a piece of code that should not be reached. These were typically of the form:

assert(0 && "Invalid radix for integer literal");

This has a few issues, the main one being that some compilers might not understand the assertion, or warn about a missing return in builds where assertions are compiled out.

Today, we have something much better: llvm_unreachable:

llvm_unreachable("Invalid radix for integer literal");

When assertions are enabled, this will print the message if it’s ever reached and then exit the program. When assertions are disabled (i.e. in release builds), llvm_unreachable becomes a hint to compilers to skip generating code for this branch. If the compiler does not support this, it will fall back to the “abort” implementation.

Neither assertions or llvm_unreachable will abort the program on a release build. If the error condition can be triggered by user input then the recoverable error mechanism described in LLVM Programmer’s Manual should be used instead. In cases where this is not practical, report_fatal_error may be used.

Another issue is that values used only by assertions will produce an “unused value” warning when assertions are disabled. For example, this code will warn:

unsigned Size = V.size();
assert(Size > 42 && "Vector smaller than it should be");

bool NewToSet = Myset.insert(Value);
assert(NewToSet && "The value shouldn't be in the set yet");

These are two interesting different cases. In the first case, the call to V.size() is only useful for the assert, and we don’t want it executed when assertions are disabled. Code like this should move the call into the assert itself. In the second case, the side effects of the call must happen whether the assert is enabled or not. In this case, the value should be cast to void to disable the warning. To be specific, it is preferred to write the code like this:

assert(V.size() > 42 && "Vector smaller than it should be");

bool NewToSet = Myset.insert(Value); (void)NewToSet;
assert(NewToSet && "The value shouldn't be in the set yet");

Do Not Use `using namespace std`¶

In LLVM, we prefer to explicitly prefix all identifiers from the standard namespace with an “std::” prefix, rather than rely on “using namespace std;”.

In header files, adding a 'using namespace XXX' directive pollutes the namespace of any source file that #includes the header, creating maintenance issues.

In implementation files (e.g. .cpp files), the rule is more of a stylistic rule, but is still important. Basically, using explicit namespace prefixes makes the code clearer, because it is immediately obvious what facilities are being used and where they are coming from. And more portable, because namespace clashes cannot occur between LLVM code and other namespaces. The portability rule is important because different standard library implementations expose different symbols (potentially ones they shouldn’t), and future revisions to the C++ standard will add more symbols to the std namespace. As such, we never use 'using namespace std;' in LLVM.

The exception to the general rule (i.e. it’s not an exception for the std namespace) is for implementation files. For example, all of the code in the LLVM project implements code that lives in the ‘llvm’ namespace. As such, it is ok, and actually clearer, for the .cpp files to have a 'using namespace llvm;' directive at the top, after the #includes. This reduces indentation in the body of the file for source editors that indent based on braces, and keeps the conceptual context cleaner. The general form of this rule is that any .cpp file that implements code in any namespace may use that namespace (and its parents’), but should not use any others.

Provide a Virtual Method Anchor for Classes in Headers ¶

If a class is defined in a header file and has a vtable (either it has virtual methods or it derives from classes with virtual methods), it must always have at least one out-of-line virtual method in the class. Without this, the compiler will copy the vtable and RTTI into every .o file that #includes the header, bloating .o file sizes and increasing link times.

Don’t use default labels in fully covered switches over enumerations ¶

-Wswitch warns if a switch, without a default label, over an enumeration does not cover every enumeration value. If you write a default label on a fully covered switch over an enumeration then the -Wswitch warning won’t fire when new elements are added to that enumeration. To help avoid adding these kinds of defaults, Clang has the warning -Wcovered-switch-default which is off by default but turned on when building LLVM with a version of Clang that supports the warning.

A knock-on effect of this stylistic requirement is that when building LLVM with GCC you may get warnings related to “control may reach end of non-void function” if you return from each case of a covered switch-over-enum because GCC assumes that the enum expression may take any representable value, not just those of individual enumerators. To suppress this warning, use llvm_unreachable after the switch.

Use range-based `for` loops wherever possible ¶

The introduction of range-based for loops in C++11 means that explicit manipulation of iterators is rarely necessary. We use range-based for loops wherever possible for all newly added code. For example:

BasicBlock *BB = ...
for (Instruction &I : *BB)
  ... use I ...

Don’t evaluate `end()` every time through a loop ¶

In cases where range-based for loops can’t be used and it is necessary to write an explicit iterator-based loop, pay close attention to whether end() is re-evaluted on each loop iteration. One common mistake is to write a loop in this style:

BasicBlock *BB = ...
for (auto I = BB->begin(); I != BB->end(); ++I)
  ... use I ...

The problem with this construct is that it evaluates “BB->end()” every time through the loop. Instead of writing the loop like this, we strongly prefer loops to be written so that they evaluate it once before the loop starts. A convenient way to do this is like so:

BasicBlock *BB = ...
for (auto I = BB->begin(), E = BB->end(); I != E; ++I)
  ... use I ...

The observant may quickly point out that these two loops may have different semantics: if the container (a basic block in this case) is being mutated, then “BB->end()” may change its value every time through the loop and the second loop may not in fact be correct. If you actually do depend on this behavior, please write the loop in the first form and add a comment indicating that you did it intentionally.

Why do we prefer the second form (when correct)? Writing the loop in the first form has two problems. First it may be less efficient than evaluating it at the start of the loop. In this case, the cost is probably minor — a few extra loads every time through the loop. However, if the base expression is more complex, then the cost can rise quickly. I’ve seen loops where the end expression was actually something like: “SomeMap[X]->end()” and map lookups really aren’t cheap. By writing it in the second form consistently, you eliminate the issue entirely and don’t even have to think about it.

The second (even bigger) issue is that writing the loop in the first form hints to the reader that the loop is mutating the container (a fact that a comment would handily confirm!). If you write the loop in the second form, it is immediately obvious without even looking at the body of the loop that the container isn’t being modified, which makes it easier to read the code and understand what it does.

While the second form of the loop is a few extra keystrokes, we do strongly prefer it.

`#include <iostream>` is Forbidden ¶

The use of #include <iostream> in library files is hereby forbidden, because many common implementations transparently inject a static constructor into every translation unit that includes it.

Note that using the other stream headers (<sstream> for example) is not problematic in this regard — just <iostream>. However, raw_ostream provides various APIs that are better performing for almost every use than std::ostream style APIs.

Note

New code should always use raw_ostream for writing, or the llvm::MemoryBuffer API for reading files.

Use `raw_ostream`¶

LLVM includes a lightweight, simple, and efficient stream implementation in llvm/Support/raw_ostream.h, which provides all of the common features of std::ostream. All new code should use raw_ostream instead of ostream.

Unlike std::ostream, raw_ostream is not a template and can be forward declared as class raw_ostream. Public headers should generally not include the raw_ostream header, but use forward declarations and constant references to raw_ostream instances.

Avoid `std::endl`¶

The std::endl modifier, when used with iostreams outputs a newline to the output stream specified. In addition to doing this, however, it also flushes the output stream. In other words, these are equivalent:

std::cout << std::endl;
std::cout << '\n' << std::flush;

Most of the time, you probably have no reason to flush the output stream, so it’s better to use a literal '\n'.

Don’t use `inline` when defining a function in a class definition ¶

A member function defined in a class definition is implicitly inline, so don’t put the inline keyword in this case.

Don’t:

class Foo {
public:
  inline void bar() {
    // ...
  }
};

Do:

class Foo {
public:
  void bar() {
    // ...
  }
};

Microscopic Details ¶

This section describes preferred low-level formatting guidelines along with reasoning on why we prefer them.

Spaces Before Parentheses ¶

Put a space before an open parenthesis only in control flow statements, but not in normal function call expressions and function-like macros. For example:

if (X) ...
for (I = 0; I != 100; ++I) ...
while (LLVMRocks) ...

somefunc(42);
assert(3 != 4 && "laws of math are failing me");

A = foo(42, 92) + bar(X);

The reason for doing this is not completely arbitrary. This style makes control flow operators stand out more, and makes expressions flow better.

Prefer Preincrement ¶

Hard fast rule: Preincrement (++X) may be no slower than postincrement (X++) and could very well be a lot faster than it. Use preincrementation whenever possible.

The semantics of postincrement include making a copy of the value being incremented, returning it, and then preincrementing the “work value”. For primitive types, this isn’t a big deal. But for iterators, it can be a huge issue (for example, some iterators contains stack and set objects in them… copying an iterator could invoke the copy ctor’s of these as well). In general, get in the habit of always using preincrement, and you won’t have a problem.

Namespace Indentation ¶

In general, we strive to reduce indentation wherever possible. This is useful because we want code to fit into 80 columns without excessive wrapping, but also because it makes it easier to understand the code. To facilitate this and avoid some insanely deep nesting on occasion, don’t indent namespaces. If it helps readability, feel free to add a comment indicating what namespace is being closed by a }. For example:

namespace llvm {
namespace knowledge {

/// This class represents things that Smith can have an intimate
/// understanding of and contains the data associated with it.
class Grokable {
...
public:
  explicit Grokable() { ... }
  virtual ~Grokable() = 0;

  ...

};

} // end namespace knowledge
} // end namespace llvm

Feel free to skip the closing comment when the namespace being closed is obvious for any reason. For example, the outer-most namespace in a header file is rarely a source of confusion. But namespaces both anonymous and named in source files that are being closed half way through the file probably could use clarification.

Anonymous Namespaces ¶

After talking about namespaces in general, you may be wondering about anonymous namespaces in particular. Anonymous namespaces are a great language feature that tells the C++ compiler that the contents of the namespace are only visible within the current translation unit, allowing more aggressive optimization and eliminating the possibility of symbol name collisions. Anonymous namespaces are to C++ as “static” is to C functions and global variables. While “static” is available in C++, anonymous namespaces are more general: they can make entire classes private to a file.

The problem with anonymous namespaces is that they naturally want to encourage indentation of their body, and they reduce locality of reference: if you see a random function definition in a C++ file, it is easy to see if it is marked static, but seeing if it is in an anonymous namespace requires scanning a big chunk of the file.

Because of this, we have a simple guideline: make anonymous namespaces as small as possible, and only use them for class declarations. For example:

namespace {
class StringSort {
...
public:
  StringSort(...)
  bool operator<(const char *RHS) const;
};
} // end anonymous namespace

static void runHelper() {
  ...
}

bool StringSort::operator<(const char *RHS) const {
  ...
}

Avoid putting declarations other than classes into anonymous namespaces:

namespace {

// ... many declarations ...

void runHelper() {
  ...
}

// ... many declarations ...

} // end anonymous namespace

When you are looking at “runHelper” in the middle of a large C++ file, you have no immediate way to tell if this function is local to the file. In contrast, when the function is marked static, you don’t need to cross-reference faraway places in the file to tell that the function is local.