Class core.thread.fiber.Fiber
This class provides a cooperative concurrency mechanism integrated with the threading and garbage collection functionality. Calling a fiber may be considered a blocking operation that returns when the fiber yields (via Fiber.yield()). Execution occurs within the context of the calling thread so synchronization is not necessary to guarantee memory visibility so long as the same thread calls the fiber each time. Please note that there is no requirement that a fiber be bound to one specific thread. Rather, fibers may be freely passed between threads so long as they are not currently executing. Like threads, a new fiber thread may be created using either derivation or composition, as in the following example.
class Fiber
: core .thread .fiber .base .FiberBase;
Constructors
Name | Description |
---|---|
this
(fn, sz, guardPageSize)
|
Initializes a fiber object which is associated with a static D function. |
this
(dg, sz, guardPageSize)
|
Initializes a fiber object which is associated with a dynamic D function. |
Methods
Name | Description |
---|---|
getThis
()
|
Provides a reference to the calling fiber or null if no fiber is currently active. |
allocStack
(sz, guardPageSize)
|
Warning
Status registers are not saved by the current implementations. This means floating point exception status bits (overflow, divide by 0), rounding mode and similar stuff is set per-thread, not per Fiber!
Warning
On ARM FPU registers are not saved if druntime was compiled as ARM_SoftFloat. If such a build is used on a ARM_SoftFP system which actually has got a FPU and other libraries are using the FPU registers (other code is compiled as ARM_SoftFP) this can cause problems. Druntime must be compiled as ARM_SoftFP in this case.
Authors
Sean Kelly, Walter Bright, Alex Rønne Petersen, Martin Nowak
License
Distributed under the Boost Software License 1.0. (See accompanying file LICENSE)