From b3387d1c30ae7da28b34378fb5d9bc29a213e991 Mon Sep 17 00:00:00 2001 From: Nandor Licker Date: Thu, 9 Apr 2020 19:13:38 +0100 Subject: [PATCH] [ConstExprPreter] Updated constant interpreter documentation Summary: Updated the documentation to better reflect features implemented on the constexpr branch at https://github.com/nandor/llvm-project and extended the TODO list with known missing features Reviewers: rsmith, Bigcheese, dexonsmith, jfb Subscribers: lebedev.ri, cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D75726 --- clang/docs/ConstantInterpreter.rst | 326 +++++++++++++++++++++++++++++-------- 1 file changed, 258 insertions(+), 68 deletions(-) diff --git a/clang/docs/ConstantInterpreter.rst b/clang/docs/ConstantInterpreter.rst index a86161c..67a17c7 100644 --- a/clang/docs/ConstantInterpreter.rst +++ b/clang/docs/ConstantInterpreter.rst @@ -8,129 +8,256 @@ Constant Interpreter Introduction ============ -The constexpr interpreter aims to replace the existing tree evaluator in clang, improving performance on constructs which are executed inefficiently by the evaluator. The interpreter is activated using the following flags: +The constexpr interpreter aims to replace the existing tree evaluator in +clang, improving performance on constructs which are executed inefficiently +by the evaluator. The interpreter is activated using the following flags: -* ``-fexperimental-new-constant-interpreter`` enables the interpreter, emitting an error if an unsupported feature is encountered +* ``-fexperimental-new-constant-interpreter`` enables the interpreter, +emitting an error if an unsupported feature is encountered Bytecode Compilation ==================== -Bytecode compilation is handled in ``ByteCodeStmtGen.h`` for statements and ``ByteCodeExprGen.h`` for expressions. The compiler has two different backends: one to generate bytecode for functions (``ByteCodeEmitter``) and one to directly evaluate expressions as they are compiled, without generating bytecode (``EvalEmitter``). All functions are compiled to bytecode, while toplevel expressions used in constant contexts are directly evaluated since the bytecode would never be reused. This mechanism aims to pave the way towards replacing the evaluator, improving its performance on functions and loops, while being just as fast on single-use toplevel expressions. - -The interpreter relies on stack-based, strongly-typed opcodes. The glue logic between the code generator, along with the enumeration and description of opcodes, can be found in ``Opcodes.td``. The opcodes are implemented as generic template methods in ``Interp.h`` and instantiated with the relevant primitive types by the interpreter loop or by the evaluating emitter. +Bytecode compilation is handled in ``ByteCodeStmtGen.h`` for statements +and ``ByteCodeExprGen.h`` for expressions. The compiler has two different +backends: one to generate bytecode for functions (``ByteCodeEmitter``) and +one to directly evaluate expressions as they are compiled, without +generating bytecode (``EvalEmitter``). All functions are compiled to +bytecode, while toplevel expressions used in constant contexts are directly +evaluated since the bytecode would never be reused. This mechanism aims to +pave the way towards replacing the evaluator, improving its performance on +functions and loops, while being just as fast on single-use toplevel +expressions. + +The interpreter relies on stack-based, strongly-typed opcodes. The glue +logic between the code generator, along with the enumeration and +description of opcodes, can be found in ``Opcodes.td``. The opcodes are +implemented as generic template methods in ``Interp.h`` and instantiated +with the relevant primitive types by the interpreter loop or by the +evaluating emitter. Primitive Types --------------- * ``PT_{U|S}int{8|16|32|64}`` - Signed or unsigned integers of a specific bit width, implemented using the ```Integral``` type. + Signed or unsigned integers of a specific bit width, implemented using + the ```Integral``` type. * ``PT_{U|S}intFP`` - Signed or unsigned integers of an arbitrary, but fixed width used to implement - integral types which are required by the target, but are not supported by the host. - Under the hood, they rely on APValue. The ``Integral`` specialisation for these - types is required by opcodes to share an implementation with fixed integrals. + Signed or unsigned integers of an arbitrary, but fixed width used to + implement integral types which are required by the target, but are not + supported by the host. Under the hood, they rely on APValue. The + ``Integral`` specialisation for these types is required by opcodes to + share an implementation with fixed integrals. * ``PT_Bool`` - Representation for boolean types, essentially a 1-bit unsigned ``Integral``. + Representation for boolean types, essentially a 1-bit unsigned + ``Integral``. * ``PT_RealFP`` - Arbitrary, but fixed precision floating point numbers. Could be specialised in - the future similarly to integers in order to improve floating point performance. + Arbitrary, but fixed precision floating point numbers. Could be + specialised in the future similarly to integers in order to improve + floating point performance. * ``PT_Ptr`` - Pointer type, defined in ``"Pointer.h"``. + Pointer type, defined in ``"Pointer.h"``. A pointer can be either null, + reference interpreter-allocated memory (``BlockPointer``) or point to an + address which can be derived, but not accessed (``ExternPointer``). * ``PT_FnPtr`` - Function pointer type, can also be a null function pointer. Defined in ``"Pointer.h"``. + Function pointer type, can also be a null function pointer. Defined + in ``"FnPointer.h"``. * ``PT_MemPtr`` - Member pointer type, can also be a null member pointer. Defined in ``"Pointer.h"`` + Member pointer type, can also be a null member pointer. Defined + in ``"MemberPointer.h"`` + +* ``PT_VoidPtr`` + + Void pointer type, can be used for rount-trip casts. Represented as + the union of all pointers which can be cast to void. + Defined in ``"VoidPointer.h"``. + +* ``PT_ObjCBlockPtr`` + + Pointer type for ObjC blocks. Defined in ``"ObjCBlockPointer.h"``. Composite types --------------- -The interpreter distinguishes two kinds of composite types: arrays and records. Unions are represented as records, except a single field can be marked as active. The contents of inactive fields are kept until they -are reactivated and overwritten. +The interpreter distinguishes two kinds of composite types: arrays and +records (structs and classes). Unions are represented as records, except +at most a single field can be marked as active. The contents of inactive +fields are kept until they are reactivated and overwritten. +Complex numbers (``_Complex``) and vectors +(``__attribute((vector_size(16)))``) are treated as arrays. Bytecode Execution ================== -Bytecode is executed using a stack-based interpreter. The execution context consists of an ``InterpStack``, along with a chain of ``InterpFrame`` objects storing the call frames. Frames are built by call instructions and destroyed by return instructions. They perform one allocation to reserve space for all locals in a single block. These objects store all the required information to emit stack traces whenever evaluation fails. +Bytecode is executed using a stack-based interpreter. The execution +context consists of an ``InterpStack``, along with a chain of +``InterpFrame`` objects storing the call frames. Frames are built by +call instructions and destroyed by return instructions. They perform +one allocation to reserve space for all locals in a single block. +These objects store all the required information to emit stack traces +whenever evaluation fails. Memory Organisation =================== Memory management in the interpreter relies on 3 data structures: ``Block`` -object which store the data and associated inline metadata, ``Pointer`` objects -which refer to or into blocks, and ``Descriptor`` structures which describe -blocks and subobjects nested inside blocks. +objects which store the data and associated inline metadata, ``Pointer`` +objects which refer to or into blocks, and ``Descriptor`` structures which +describe blocks and subobjects nested inside blocks. Blocks ------ -Blocks contain data interleaved with metadata. They are allocated either statically -in the code generator (globals, static members, dummy parameter values etc.) or -dynamically in the interpreter, when creating the frame containing the local variables -of a function. Blocks are associated with a descriptor that characterises the entire -allocation, along with a few additional attributes: - -* ``IsStatic`` indicates whether the block has static duration in the interpreter, i.e. it is not a local in a frame. - -* ``IsExtern`` indicates that the block was created for an extern and the storage cannot be read or written. +Blocks contain data interleaved with metadata. They are allocated either +statically in the code generator (globals, static members, dummy parameter +values etc.) or dynamically in the interpreter, when creating the frame +containing the local variables of a function. Blocks are associated with a +descriptor that characterises the entire allocation, along with a few +additional attributes: + +* ``IsStatic`` indicates whether the block has static duration in the +interpreter, i.e. it is not a local in a frame. + +* ``DeclID`` identifies each global declaration (it is set to an invalid +and irrelevant value for locals) in order to prevent illegal writes and +reads involving globals and temporaries with static storage duration. + +Static blocks are never deallocated, but local ones might be deallocated +even when there are live pointers to them. Pointers are only valid as +long as the blocks they point to are valid, so a block with pointers to +it whose lifetime ends is kept alive until all pointers to it go out of +scope. Since the frame is destroyed on function exit, such blocks are +turned into a ``DeadBlock`` and copied to storage managed by the +interpreter itself, not the frame. Reads and writes to these blocks are +illegal and cause an appropriate diagnostic to be emitted. When the last +pointer goes out of scope, dead blocks are also deallocated. + +The lifetime of blocks is managed through 3 methods stored in the +descriptor of the block: + +* **CtorFn**: initializes the metadata which is store in the block, +alongside actual data. Invokes the default constructors of objects +which are not trivial (``Pointer``, ``RealFP``, etc.) -* ``DeclID`` identifies each global declaration (it is set to an invalid and irrelevant value for locals) in order to prevent illegal writes and reads involving globals and temporaries with static storage duration. - -Static blocks are never deallocated, but local ones might be deallocated even when there are live pointers to them. Pointers are only valid as long as the blocks they point to are valid, so a block with pointers to it whose lifetime ends is kept alive until all pointers to it go out of scope. Since the frame is destroyed on function exit, such blocks are turned into a ``DeadBlock`` and copied to storage managed by the interpreter itself, not the frame. Reads and writes to these blocks are illegal and cause an appropriate diagnostic to be emitted. When the last pointer goes out of scope, dead blocks are also deallocated. - -The lifetime of blocks is managed through 3 methods stored in the descriptor of the block: - -* **CtorFn**: initializes the metadata which is store in the block, alongside actual data. Invokes the default constructors of objects which are not trivial (``Pointer``, ``RealFP``, etc.) * **DtorFn**: invokes the destructors of non-trivial objects. -* **MoveFn**: moves a block to dead storage. -Non-static blocks track all the pointers into them through an intrusive doubly-linked list, this is required in order to adjust all pointers when transforming a block into a dead block. +* **MoveFn**: moves a block to dead storage. -Descriptors ------------ +Non-static blocks track all the pointers into them through an intrusive +doubly-linked list, required to adjust and invalidate all pointers when +transforming a block into a dead block. If the lifetime of an object ends, +all pointers to it are invalidated, emitting the appropriate diagnostics when +dereferenced. -Descriptor are generated at bytecode compilation time and contain information required to determine if a particular memory access is allowed in constexpr. Even though there is a single descriptor object, it encodes information for several kinds of objects: +The interpreter distinguishes 3 different kinds of blocks: * **Primitives** - A block containing a primitive reserved storage only for the primitive. + A block containing a single primitive with no additional metadata. * **Arrays of primitives** - An array of primitives contains a pointer to an ``InitMap`` storage as its first field: the initialisation map is a bit map indicating all elements of the array which were initialised. If the pointer is null, no elements were initialised, while a value of ``(InitMap)-1`` indicates that the object was fully initialised. when all fields are initialised, the map is deallocated and replaced with that token. + An array of primitives contains a pointer to an ``InitMap`` storage as its + first field: the initialisation map is a bit map indicating all elements of + the array which were initialised. If the pointer is null, no elements were + initialised, while a value of ``(InitMap*)-1`` indicates that the object was + fully initialised. When all fields are initialised, the map is deallocated + and replaced with that token. - Array elements are stored sequentially, without padding, after the pointer to the map. + Array elements are stored sequentially, without padding, after the pointer + to the map. * **Arrays of composites and records** - Each element in an array of composites is preceded by an ``InlineDescriptor``. Descriptors and elements are stored sequentially in the block. Records are laid out identically to arrays of composites: each field and base class is preceded by an inline descriptor. The ``InlineDescriptor`` has the following field: + Each element in an array of composites is preceded by an ``InlineDescriptor`` + which stores the attributes specific to the field and not the whole + allocation site. Descriptors and elements are stored sequentially in the + block. + Records are laid out identically to arrays of composites: each field and base + class is preceded by an inline descriptor. The ``InlineDescriptor`` + has the following fields: + + * **Offset**: byte offset into the array or record, used to step back to the + parent array or record. + * **IsConst**: flag indicating if the field is const-qualified. + * **IsInitialized**: flag indicating whether the field or element was + initialized. For non-primitive fields, this is only relevant to determine + the dynamic type of objects during construction. + * **IsBase**: flag indicating whether the record is a base class. In that + case, the offset can be used to identify the derived class. + * **IsActive**: indicates if the field is the active field of a union. + * **IsMutable**: indicates if the field is marked as mutable. + +Inline descriptors are filled in by the `CtorFn` of blocks, which leaves storage +in an uninitialised, but valid state. - * **Offset**: byte offset into the array or record, used to step back to the parent array or record. - * **IsConst**: flag indicating if the field is const-qualified. - * **IsInitialized**: flag indicating whether the field or element was initialized. For non-primitive fields, this is only relevant for base classes. - * **IsBase**: flag indicating whether the record is a base class. In that case, the offset can be used to identify the derived class. - * **IsActive**: indicates if the field is the active field of a union. - * **IsMutable**: indicates if the field is marked as mutable. +Descriptors +----------- -Inline descriptors are filled in by the `CtorFn` of blocks, which leaves storage in an uninitialised, but valid state. +Descriptors are generated at bytecode compilation time and contain information +required to determine if a particular memory access is allowed in constexpr. +They also carry all the information required to emit a diagnostic involving +a memory access, such as the declaration which originates the block. +Currently there is a single kind of descriptor encoding information for all +block types. Pointers -------- -Pointers track a ``Pointee``, the block to which they point or ``nullptr`` for null pointers, along with a ``Base`` and an ``Offset``. The base identifies the innermost field, while the offset points to an array element relative to the base (including one-past-end pointers). Most subobject the pointer points to in block, while the offset identifies the array element the pointer points to. These two fields allow all pointers to be uniquely identified and disambiguated. +Pointers, implemented in ``Pointer.h`` are represented as a tagged union. +Some of these may not yet be available in upstream ``clang``. + + * **BlockPointer**: used to reference memory allocated and managed by the + interpreter, being the only pointer kind which allows dereferencing in the + interpreter + * **ExternPointer**: points to memory which can be addressed, but not read by + the interpreter. It is equivalent to APValue, tracking a declaration and a path + of fields and indices into that allocation. + * **TargetPointer**: represents a target address derived from a base address + through pointer arithmetic, such as ``((int *)0x100)[20]``. Null pointers are + target pointers with a zero offset. + * **TypeInfoPointer**: tracks information for the opaque type returned by + ``typeid`` + * **InvalidPointer**: is dummy pointer created by an invalid operation which + allows the interpreter to continue execution. Does not allow pointer + arithmetic or dereferencing. + +Besides the previously mentioned union, a number of other pointer-like types +have their own type: + + * **ObjCBlockPointer** tracks Objective-C blocks + * **FnPointer** tracks functions and lazily caches their compiled version + * **MemberPointer** tracks C++ object members + +Void pointers, which can be built by casting any of the aforementioned +pointers, are implemented as a union of all pointer types. The ``BitCast`` +opcode is reponsible for performing all legal conversions between these +types and primitive integers. + +BlockPointer +~~~~~~~~~~~~ + +Block pointers track a ``Pointee``, the block to which they point, along +with a ``Base`` and an ``Offset``. The base identifies the innermost field, +while the offset points to an array element relative to the base (including +one-past-end pointers). The offset identifies the array element or field +which is referenced, while the base points to the outer object or array which +contains the field. These two fields allow all pointers to be uniquely +identified, disambiguated and characterised. As an example, consider the following structure: @@ -149,9 +276,14 @@ As an example, consider the following structure: }; constexpr A a; -On the target, ``&a`` and ``&a.b.x`` are equal. So are ``&a.c[0]`` and ``&a.c[0].a``. In the interpreter, all these pointers must be distinguished since the are all allowed to address distinct range of memory. +On the target, ``&a`` and ``&a.b.x`` are equal. So are ``&a.c[0]`` and +``&a.c[0].a``. In the interpreter, all these pointers must be +distinguished since the are all allowed to address distinct range of +memory. -In the interpreter, the object would require 240 bytes of storage and would have its field interleaved with metadata. The pointers which can be derived to the object are illustrated in the following diagram: +In the interpreter, the object would require 240 bytes of storage and +would have its field interleaved with metadata. The pointers which can +be derived to the object are illustrated in the following diagram: :: @@ -164,9 +296,63 @@ In the interpreter, the object would require 240 bytes of storage and would have a |&a.b.x &a.y &a.c |&a.c[0].a |&a.c[1].a | &a.b &a.c[0] &a.c[1] &a.z -The ``Base`` offset of all pointers points to the start of a field or an array and is preceded by an inline descriptor (unless ``Base == 0``, pointing to the root). All the relevant attributes can be read from either the inline descriptor or the descriptor of the block. +The ``Base`` offset of all pointers points to the start of a field or +an array and is preceded by an inline descriptor (unless ``Base`` is +zero, pointing to the root). All the relevant attributes can be read +from either the inline descriptor or the descriptor of the block. + + +Array elements are identified by the ``Offset`` field of pointers, +pointing to past the inline descriptors for composites and before +the actual data in the case of primitive arrays. The ``Offset`` +points to the offset where primitives can be read from. As an example, +``a.c + 1`` would have the same base as ``a.c`` since it is an element +of ``a.c``, but its offset would point to ``&a.c[1]``. The +array-to-pointer decay operation adjusts a pointer to an array (where +the offset is equal to the base) to a pointer to the first element. + +ExternPointer +~~~~~~~~~~~~~ + +Extern pointers can be derived, pointing into symbols which are not +readable from constexpr. An external pointer consists of a base +declaration, along with a path designating a subobject, similar to +the ``LValuePath`` of an APValue. Extern pointers can be converted +to block pointers if the underlying variable is defined after the +pointer is created, as is the case in the following example: + +.. code-block:: c + + extern const int a; + constexpr const int *p = &a; + const int a = 5; + static_assert(*p == 5, "x"); + +TargetPointer +~~~~~~~~~~~~~ + +While null pointer arithmetic or integer-to-pointer conversion is +banned in constexpr, some expressions on target offsets must be folded, +replicating the behaviour of the ``offsetof`` builtin. Target pointers +are characterised by 3 offsets: a field offset, an array offset and a +base offset, along with a descriptor specifying the type the pointer is +supposed to refer to. Array indexing adjusts the array offset, while the +field offset is adjusted when a pointer to a member is created. Casting +an integer to a pointer sets the value of the base offset. As a special +case, null pointers are target pointers with all offets set to 0. + +TypeInfoPointer +~~~~~~~~~~~~~~~ + +``TypeInfoPointer`` tracks two types: the type assigned to +``std::type_info`` and the type which was passed to ``typeinfo``. + +InvalidPointer +~~~~~~~~~~~~~~ -Array elements are identified by the ``Offset`` field of pointers, pointing to past the inline descriptors for composites and before the actual data in the case of primitive arrays. The ``Offset`` points to the offset where primitives can be read from. As an example, ``a.c + 1`` would have the same base as ``a.c`` since it is an element of ``a.c``, but its offset would point to ``&a.c[1]``. The ``*`` operation narrows the scope of the pointer, adjusting the base to ``&a.c[1]``. The reverse operator, ``&``, expands the scope of ``&a.c[1]``, turning it into ``a.c + 1``. When a one-past-end pointer is narrowed, its offset is set to ``-1`` to indicate that it is an invalid value (expanding returns the past-the-end pointer). As a special case, narrowing ``&a.c`` results in ``&a.c[0]``. The `narrow` and `expand` methods can be used to follow the chain of equivalent pointers. +Such pointers are built by operations which cannot generate valid +pointers, allowing the interpreter to continue execution after emitting +a warning. Inspecting such a pointer stops execution. TODO ==== @@ -174,20 +360,24 @@ TODO Missing Language Features ------------------------- -* Definition of externs must override previous declaration * Changing the active field of unions -* Union copy constructors -* ``typeid`` * ``volatile`` * ``__builtin_constant_p`` -* ``std::initializer_list`` -* lambdas -* range-based for loops -* ``vector_size`` * ``dynamic_cast`` +* ``new`` and ``delete`` +* Fixed Point numbers and arithmetic on Complex numbers +* Several builtin methods, including string operations and +``__builtin_bit_cast`` +* Continue-after-failure: a form of exception handling at the bytecode +level should be implemented to allow execution to resume. As an example, +argument evaluation should resume after the computation of an argument fails. +* Pointer-to-Integer conversions +* Lazy descriptors: the interpreter creates a ``Record`` and ``Descriptor`` +when it encounters a type: ones which are not yet defined should be lazily +created when required Known Bugs ---------- -* Pointer comparison for equality needs to narrow/expand pointers -* If execution fails, memory storing APInts and APFloats is leaked when the stack is cleared +* If execution fails, memory storing APInts and APFloats is leaked when the +stack is cleared -- 2.7.4