# Platform¶

class simtk.openmm.openmm.Platform(*args, **kwargs)

A Platform defines an implementation of all the kernels needed to perform some calculation. More precisely, a Platform object acts as a registry for a set of KernelFactory objects which together implement the kernels. The Platform class, in turn, provides a static registry of all available Platform objects.

To get a Platform object, call

<pre> Platform& platform = Platform::findPlatform(kernelNames); </pre>

passing in the names of all kernels that will be required for the calculation you plan to perform. It will return the fastest available Platform which provides implementations of all the specified kernels. You can then call createKernel() to construct particular kernels as needed.

__init__(*args, **kwargs)

Methods

 __init__(*args, **kwargs) findPlatform((kernelNames) -> Platform) Find a Platform which can be used to perform a calculation. getDefaultPluginsDirectory(...) Get the default directory from which to load plugins. getName((self) -> std::string const &) Get the name of this platform. getNumPlatforms(() -> int) Get the number of Platforms that have been registered. getOpenMMVersion(() -> std::string const &) Get a string containing the version number of the OpenMM library. getPlatform((index) -> Platform) Get a registered Platform by index. getPlatformByName((name) -> Platform) Get the registered Platform with a particular name. getPluginLoadFailures(() -> vectorstring) Get any failures caused during the last call to loadPluginsFromDirectory getPropertyDefaultValue((self, ...) Get the default value of a Platform-specific property. getPropertyNames((self) -> vectorstring) Get the names of all Platform-specific properties this Platform supports. getPropertyValue((self, context, ...) Get the value of a Platform-specific property for a Context. getSpeed((self) -> double) Get an estimate of how fast this Platform class is. linkedContextCreated(self, context, ...) This is called whenever a new Context is created using ContextImpl::createLinkedContext(). loadPluginLibrary(file) Load a dynamic library (DLL) which contains an OpenMM plugin. loadPluginsFromDirectory(...) Load multiple dynamic libraries (DLLs) which contain OpenMM plugins from a single directory. registerPlatform(platform) Register a new Platform. setPropertyDefaultValue(self, property, value) Set the default value of a Platform-specific property. setPropertyValue(self, context, property, value) Set the value of a Platform-specific property for a Context. supportsDoublePrecision((self) -> bool) Get whether this Platform supports double precision arithmetic. supportsKernels((self, kernelNames) -> bool) Determine whether this Platforms provides implementations of a set of kernels.
static registerPlatform(platform)

Register a new Platform.

static getNumPlatforms() → int

Get the number of Platforms that have been registered.

static getPlatform(index) → Platform

Get a registered Platform by index.

static getPluginLoadFailures() → vectorstring

Get any failures caused during the last call to loadPluginsFromDirectory

static getPlatformByName(name) → Platform

Get the registered Platform with a particular name. If no Platform with that name has been registered, this throws an exception.

static findPlatform(kernelNames) → Platform

Find a Platform which can be used to perform a calculation.

Parameters: kernelNames (vector< std::string >) – the names of all kernels which will be needed for the calculation the fastest registered Platform which supports all of the requested kernels. If no Platform exists which supports all of them, this will throw an exception. Platform
static loadPluginLibrary(file)

Load a dynamic library (DLL) which contains an OpenMM plugin. Typically, each Platform is distributed as a separate dynamic library. This method can then be called at runtime to load each available library. Each library should contain an initializer function to register any Platforms and KernelFactories that it contains.

If the file does not exist or cannot be loaded, an exception is thrown.

Parameters: file (string) – the path to the dynamic library file. This is interpreted using the operating system’s rules for loading libraries. Typically it may be either an absolute path or relative to a set of standard locations.
static loadPluginsFromDirectory(directory) → vectorstring

Load multiple dynamic libraries (DLLs) which contain OpenMM plugins from a single directory. This method loops over every file contained in the specified directory and calls loadPluginLibrary() for each one. If an error occurs while trying to load a particular file, that file is simply ignored. You can retrieve a list of all such errors by calling getPluginLoadFailures().

Parameters: directory (string) – the path to the directory containing libraries to load the names of all files which were successfully loaded as libraries vector< std::string >
static getDefaultPluginsDirectory() → std::string const &

Get the default directory from which to load plugins. If the environment variable OPENMM_PLUGIN_DIR is set, this returns its value. Otherwise, it returns a platform specific default location.

Returns: the path to the default plugin directory string
static getOpenMMVersion() → std::string const &

Get a string containing the version number of the OpenMM library.

getName(self) → std::string const &

Get the name of this platform. This should be a unique identifier which can be used to recognized it.

getSpeed(self) → double

Get an estimate of how fast this Platform class is. This need not be precise. It only is expected to return an order or magnitude estimate of the relative performance of different Platform classes. An unoptimized reference implementation should return 1.0, and all other Platforms should return a larger value that is an estimate of how many times faster they are than the reference implementation.

supportsDoublePrecision(self) → bool

Get whether this Platform supports double precision arithmetic. If this returns false, the platform is permitted to represent double precision values internally as single precision.

Deprecated

This method is not well defined, and is too simplistic to describe the actual behavior of some Platforms, such as ones that offer multiple precision modes. It will be removed in a future release.
getPropertyNames(self) → vectorstring

Get the names of all Platform-specific properties this Platform supports.

getPropertyValue(self, context, property) → std::string const &

Get the value of a Platform-specific property for a Context.

Parameters: context (Context) – the Context for which to get the property property (string) – the name of the property to get the value of the property string
setPropertyValue(self, context, property, value)

Set the value of a Platform-specific property for a Context.

Parameters: context (Context) – the Context for which to set the property property (string) – the name of the property to set value (string) – the value to set for the property
getPropertyDefaultValue(self, property) → std::string const &

Get the default value of a Platform-specific property. This is the value that will be used for newly created Contexts.

Parameters: property (string) – the name of the property to get the default value of the property string
setPropertyDefaultValue(self, property, value)

Set the default value of a Platform-specific property. This is the value that will be used for newly created Contexts.

Parameters: property (string) – the name of the property to set value (string) – the value to set for the property
linkedContextCreated(self, context, originalContext)

This is called whenever a new Context is created using ContextImpl::createLinkedContext(). It gives the Platform a chance to initialize the context and store platform-specific data in it.

Parameters: context (ContextImpl) – the newly created context originalContext (ContextImpl) – the original context it is linked to
supportsKernels(self, kernelNames) → bool

Determine whether this Platforms provides implementations of a set of kernels.

Parameters: kernelNames (vector< std::string >) – the names of the kernels of interests true if this Platform provides implementations of all the kernels in the list, false if there are any which it does not support bool
__delattr__

x.__delattr__(‘name’) <==> del x.name

__format__()

default object formatter

__getattribute__

x.__getattribute__(‘name’) <==> x.name

__hash__
__reduce__()

helper for pickle

__reduce_ex__()

helper for pickle

__sizeof__() → int

size of object in memory, in bytes

__str__