Concurrency Namespace (C++ AMP)
Provides classes and functions that accelerate the execution of C++ code on data-parallel hardware. For more information, see C++ AMP Overview
namespace Concurrency;
Members
Namespaces
Name |
Description |
|---|---|
Provides functions that support D3D interoperability. Enables seamless use of D3D resources for compute in AMP code and the use of resources created in AMP in D3D code, without creating redundant intermediate copies. You can use C++ AMP to incrementally accelerate the compute-intensive sections of your DirectX applications and use the D3D API on data produced from AMP computations. |
|
Functions in the fast_math namespace are not C99-compliant. Only single-precision versions of each function are provided. These functions use the DirectX intrinsic functions, which are faster than the corresponding functions in the precise_math namespace and do not require extended double-precision support on the accelerator, but they are less accurate. There are two versions of each function for source-level compatibility with C99 code; both versions take and return single-precision values. |
|
Provides types and functions that are designed for graphics programming. |
|
Functions in the precise_math namespace are C99 compliant. Both single-precision and double-precision versions of each function are included. These functions—this includes the single-precision functions—require extended double-precision support on the accelerator. |
Classes
Name |
Description |
|---|---|
Represents an abstraction of a physical DP-optimized compute node. |
|
Represents a virtual device abstraction on a C++ AMP data-parallel accelerator. |
|
The exception that is thrown when an underlying DirectX call fails due to the Windows timeout-detection-and-recovery mechanism. |
|
A data aggregate on an accelerator_view in the grid domain. It is a collection of variables, one for each element in a grid domain. Each variable holds a value that corresponds to some C++ type. |
|
Represents a view into the data in an array<T,N>. |
|
Represents a future that corresponds to a C++ AMP asynchronous operation. |
|
Represents a vector of N integer values that specify the bounds of an N-dimensional space that has an origin of 0. The values in the coordinate vector are ordered from most significant to least significant. For example, in Cartesian 3-dimensional space, the extent vector (7,5,3) represents a space in which the z coordinate ranges from 0 to 7, the y coordinate ranges from 0 to 5, and the x coordinate ranges from 0 to 3. |
|
Defines an N-dimensional index point. |
|
The exception that's thrown when the runtime can't start a kernel by using the compute domain specified at the parallel_for_each call site. |
|
The exception that is thrown when a method fails because of a lack of system or device memory. |
|
The base type for exceptions in the C++ AMP library. |
|
A capability class that is only creatable by the system and is passed to a tiled parallel_for_each lambda as part of the tiled_index parameter. It provides one method, wait(), whose purpose is to synchronize execution of threads that are running in the thread group (tile). |
|
A tiled_extent object is an extent object of one to three dimensions that subdivides the extent space into one-dimensional, two-dimensional, or three-dimensional tiles. |
|
Provides an index into a tiled_grid object. This class has properties to access element relative to the local tile origin and relative to the global origin. |
|
The exception that is thrown when an uninitialized object is used. |
|
The exception that is thrown when an unsupported feature is used. |
Enumerations
Name |
Description |
|---|---|
Specifies the data access type. |
|
Specifies the queuing modes that are supported on the accelerator. |
Operators
Operator |
Description |
|---|---|
Determines whether the specified data structures are equal. |
|
Determines whether the specified data structures are unequal. |
|
Computes the component-wise sum of the specified arguments. |
|
Computes the component-wise difference between the specified arguments. |
|
Computes the component-wise product of the specified arguments. |
|
Computes the component-wise quotient of the specified arguments. |
|
Computes the modulus of the first specified argument by the second specified argument. |
Functions
Name |
Description |
|---|---|
Blocks execution of all threads in a tile until all memory accesses have been completed. |
|
Uninitializes the C++ AMP runtime. |
|
Overloaded. If the value stored at the specified location compares equal to the first specified value, then the second specified value is stored in the same location as an atomic operation. |
|
Overloaded. Sets the value stored at the specified location to the specified value as an atomic operation. |
|
Overloaded. Sets the value stored at the specified location to the sum of that value and a specified value as an atomic operation. |
|
Overloaded. Sets the value stored at the specified location to the bitwise and of that value and a specified value as an atomic operation. |
|
Overloaded. Decrements the value stored at the specified location and stores the result in the same location as an atomic operation. |
|
Overloaded. Increments the value stored at the specified location and stores the result in the same location as an atomic operation. |
|
Overloaded. Sets the value stored at the specified location to the larger of that value and a specified value as an atomic operation. |
|
Overloaded. Sets the value stored at the specified location to the smaller of that value and a specified value as an atomic operation. |
|
Overloaded. Sets the value stored at the specified location to the bitwise or of that value and a specified value as an atomic operation. |
|
Overloaded. Sets the value stored at the specified location to the difference of that value and a specified value as an atomic operation. |
|
Overloaded. Sets the value stored at the specified location to the bitwise xor of that value and a specified value as an atomic operation. |
|
Copies a C++ AMP object. All synchronous data transfer requirements are met. Data can't be copied when code is running code on an accelerator. The general form of this function is copy(src, dest). |
|
Copies a C++ AMP object and returns completion_future that can be waited on. Data can't be copied when code is running on an accelerator. The general form of this function is copy(src, dest). |
|
Aborts the execution of a function that has the restrict(amp) restriction clause. |
|
Prints a formatted string to the Visual Studio Output window and raises a runtime_exception exception that has the same formatting string. |
|
Prints a formatted string to the Visual Studio Output window. It is called from a function that has the restrict(amp) restriction clause. |
|
Blocks execution of all threads in a tile until all global memory accesses have been completed. |
|
Runs a function across the compute domain. |
|
Blocks execution of all threads in a tile until tile_static memory accesses have been completed. |
Constants
Name |
Description |
|---|---|
The maximum number of buffers allowed by DirectX. |
|
Stores the maximum length of the module name. This value must be the same on the compiler and the runtime. |
Requirements
Header: amp.h