1
0
Fork 0
mirror of https://github.com/eclipse-cdt/cdt synced 2025-04-29 19:45:01 +02:00
cdt/plugins/org.eclipse.dd.doc.dsf/docs/dsf_concurrency_model.html

432 lines
30 KiB
HTML

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="content-type">
<title>DSF Concurrency Model</title>
</head>
<body>
<h2>DSF Concurrency Model</h2>
<h3>
</h3>
<p class="MsoNormal" style="line-height: normal;"><b><span
style="font-size: 12pt; font-family: &quot;Times New Roman&quot;;">Version
1.0<br>
Pawel Piech<br>
&copy; 2006, Wind River Systems.<span style="">&nbsp; </span>Release
under EPL version 1.0.</span></b><b><span
style="font-size: 18pt; font-family: &quot;Times New Roman&quot;;"><o:p></o:p></span></b></p>
<h3>Introduction</h3>
Providing a solution to concurrency problems is the primary design goal
of DSF.&nbsp; To that end DSF imposes a rather draconian
restriction on services that use it: <span style="font-weight: bold;">1)
All service interface methods must be called using a single designated
dispatch thread, unless explicitly stated otherwise, 2) The dispatch
thread should never be used to make a blocking call (a call that waits
on I/O or a call that makes a long-running computation).&nbsp; </span>What
the first restriction effectively means, is that the dispatch thread
becomes a global "lock" that all DSF services in a given session
share with each other, and which controls access to most of services'
shared data.&nbsp; It's important to note that <span
style="font-weight: bold;">multi-threading is still allowed</span>
within individual service implementation. but when crossing the service
interface boundaries, only the dispatch thread can be used.&nbsp; The
second restriction just ensures that the performance of the whole
system is not killed by one service that needs to read a huge file over
the network.&nbsp; Another way of looking at it is that the
service implementations practice co-operative multi-threading using the
single dispatch thread.<br>
<br>
There are a couple of obvious side effects that result from this rule:<br>
<ol>
<li>When executing within the dispatch thread, the state of the
services is guaranteed not to change.&nbsp; This means that
thread-defensive programming techniques, such as making duplicates of
lists before iterating over them, are not necessary.&nbsp; Also it's
possible to implement much more complicated logic which polls the state
of many objects, without the worry about dead-locks.</li>
<li>Whenever a blocking operation needs to be performed, it must be
done using an asynchronous method.&nbsp; By the time the operation is
completed, and the caller regains the dispatch thread, this caller may
need to retest the relevant state of the system, because it could
change completely while the asynchronous operation was executing.</li>
</ol>
<h3>The Mechanics</h3>
<h4><span style="font-family: monospace;">java.util.concurrent.ExecutorService</span><br>
</h4>
DSF builds on the vast array of tools added in Java 5.0's
java.util.concurrent package (see <a
href="http://java.sun.com/j2se/1.5.0/docs/guide/concurrency/index.html">http://java.sun.com/j2se/1.5.0/docs/guide/concurrency/index.html</a>
for details), where the most important is the <a
style="font-family: monospace;"
href="http://java.sun.com/j2se/1.5.0/docs/api/java/util/concurrent/ExecutorService.html">ExecutorService</a>
interface.&nbsp; <span style="font-family: monospace;">ExecutorService
</span>is a formal interface for submitting <span
style="font-family: monospace;">Runnable</span> objects that will be
executed according to executor's rules, which could be to execute the
<span style="font-family: monospace;">Runnable </span>immediately,
within a thread pool, using a display thread,
etc.&nbsp; For DSF, the main rule for executors is that they have
to use a single thread to execute the runnable and that the runnables
be executed in the order that they were submitted.&nbsp; To give the
DSF clients and services a method for checking whether they are
being called on the dispatch thread, we extended the <span
style="font-family: monospace;">ExecutorService
</span>interface as such:<br>
<pre>public interface DsfExecutor extends ScheduledExecutorService<br>{<br> /**<br> * Checks if the thread that this method is called in is the same as the<br> * executor's dispatch thread.<br> * @return true if in DSF executor's dispatch thread<br> */<br> public boolean isInExecutorThread();<br>}<br></pre>
<h4><a
href="http://java.sun.com/j2se/1.5.0/docs/api/java/util/concurrent/Future.html"><span
style="font-family: monospace;">java.lang.concurrent.Future</span></a>
vs <a
href="http://dsdp.eclipse.org/help/latest/topic/org.eclipse.dd.dsf.doc/reference/api/org/eclipse/dd/dsf/concurrent/Done.html"><span
style="font-family: monospace;">org.eclipse.dd.dsf.concurrent.Done</span></a></h4>
The <span style="font-family: monospace;">Done </span>object
encapsulates the return value of an asynchronous call in DSF.&nbsp; It
is actually merely a <span style="font-family: monospace;">Runnable </span>with
an attached <span style="font-family: monospace;">org.eclipse.core.runtime.IStatus</span>
object , but it can be extended by the services or clients to hold
whatever additional data is needed.&nbsp;&nbsp; Typical pattern in how
the <span style="font-family: monospace;">Done </span>object is used,
is as follows:<br>
<pre>Service:<br> public class Service {<br> void asyncMethod(Done done) {<br> new Job() {<br> public void run() {<br> // perform calculation <br> ... <br> done.setStatus(new Status(IStatus.ERROR, ...));<br> fExecutor.execute(done);<br> }<br> }.schedule();<br> }<br> }<br><br>Client:<br> ...<br> Service service = new Service();<br> final String clientData = "xyz";<br> ...<br> service.asynMethod(new Done() {<br> public void run() {<br> if (getStatus().isOK()) {<br> // Handle return data<br> ...<br> } else {<br> // Handle error<br> ...<br> }<br> }<br> }<br></pre>
The service performs the asynchronous operation a background thread,
but
it can still submit the <span style="font-family: monospace;">Done </span>runnable
with the executor.&nbsp; In other words, the <span
style="font-family: monospace;">Done</span> and other runnables can be
submitted from any thread, but will always execute in the single
dispatch thread.&nbsp; Also if the implementation of the <span
style="font-family: monospace;">asyncMethod()</span> is non-blocking,
it does not need to start a job, it could just perform the operation in
the dispatch thread.&nbsp; On the client side, care has to be taken to
save appropriate state before the asynchronous method is called,
because by the time the <span style="font-family: monospace;">Done </span>is
executed, the client state may change.<br>
<br>
The <span style="font-family: monospace;">java.lang.concurrent</span>
package
doesn't already have a <span style="font-family: monospace;">Done</span>,
because the generic concurrent
package is geared more towards large thread pools, where clients submit
tasks to be run in a style similar to Eclipse's Jobs, rather than using
the single dispatch thread model of DSF.&nbsp; To this end<span
style="font-family: monospace;">,</span> the
concurrent package does have an equivalent object, <a
style="font-family: monospace;"
href="http://java.sun.com/j2se/1.5.0/docs/api/java/util/concurrent/Future.html">Future</a>.&nbsp;
<span style="font-family: monospace;">Future </span>has methods that
allows the client to call the <span style="font-family: monospace;">get()</span>
method, and block while waiting for a result, and for this reason it
cannot
be used from the dispatch thread.&nbsp; But it can be used, in a
limited way, by clients which are running on background thread that
still
need to retrieve data from <span style="text-decoration: underline;">synchronous</span>
DSF methods.&nbsp; In this case the code might look like the
following:<br>
<pre>Service:<br> public class Service {<br> int syncMethod() {<br> // perform calculation<br> ...<br> return result;<br> }<br> }<br><br>Client:<br> ...<br> DsfExecutor executor = new DsfExecutor();<br> final Service service = new Service(executor);<br> Future&lt;Integer&gt; future = executor.submit(new Callable&lt;Integer&gt;() {<br> Integer call() {<br> return service.syncMethod();<br> }<br> });<br> int result = future.get();<br></pre>
The biggest drawback to using <span style="font-family: monospace;">Future
</span>with DSF services, is that it does not work with
asynchronous methods.&nbsp; This is because the <a
href="http://java.sun.com/j2se/1.5.0/docs/api/java/util/concurrent/Callable.html#call%28%29"><span
style="font-family: monospace;">Callable.call()</span></a>
implementation
has to return a value within a single dispatch cycle.&nbsp; To get
around this, DSF has an additional object called <span
style="font-family: monospace;">DsfQuery</span>, which works like a <span
style="font-family: monospace;">Future </span>combined with a <span
style="font-family: monospace;">Callable</span>, but allows the
implementation to make multiple dispatches before setting the return
value to the client.&nbsp; The <span style="font-family: monospace;">DsfQuery<span
style="font-family: monospace;"> object works as follows:<br>
</span></span>
<ol>
<li>Client creates the query object with its own implementation of <span
style="font-family: monospace;">DsfQuery.execute()</span>.<br>
</li>
<li>Client calls the <span style="font-family: monospace;">DsfQuery.get()</span>
method on non-dispatch thread, and blocks.</li>
<li>The query is queued with the executor, and eventually the <span
style="font-family: monospace;">DsfQuery.execute()</span> method is
called on the dispatch thread.</li>
<li>The query <span style="font-family: monospace;">DsfQuery.execute()</span>
calls synchronous and asynchronous methods that are needed to do its
job.</li>
<li>The query code calls <span style="font-family: monospace;">DsfQuery.done()</span>
method with the result.</li>
<li>The <span style="font-family: monospace;">DsfQuery.get()</span>
method un-blocks and returns the result to the client.<br>
</li>
</ol>
<h3><a
href="http://dsdp.eclipse.org/help/latest/topic/org.eclipse.dd.dsf.doc/reference/api/org/eclipse/dd/dsf/examples/concurrent/package-summary.html">Slow
Data Provider Example</a></h3>
The point of DSF concurrency can be most easily explained through
a practical example.&nbsp; Suppose there is a viewer which needs to
show data that originates from a remote "provider".&nbsp; There is a
considerable delay in transmitting the data to and from the provider,
and some delay in processing the data.&nbsp; The viewer is a
lazy-loading table, which means that it request information only about
items that are visible on the screen, and as the table is scrolled, new
requests for data are generated.&nbsp; The diagram below illustrates
the
logical relationship between components:<br>
<br>
.<img alt="" title="Slow Data Provider Diagram"
src="dsf_concurrency_model-1.png" style="width: 636px; height: 128px;"><br>
<p>In detail, these components look like this:<span
style="text-decoration: underline;"></span></p>
<p><span style="text-decoration: underline;"></span></p>
<span style="text-decoration: underline;">Table Viewer</span><br>
<p>The table viewer is the standard
<span style="font-family: monospace;">org.eclipse.jface.viewers.TableViewer</span>,
created with <span style="font-family: monospace;">SWT.VIRTUAL</span>
flag.&nbsp; It has an associated content
provider, SlowDataProviderContentProvider) which handles all the
interactions with the data provider.&nbsp; The lazy content provider
operates in a very simple cycle:</p>
<ol>
<li>Table viewer tells content provider that the input has changed by
calling <span style="font-family: monospace;">IContentProvider.inputChanged()</span>.&nbsp;
This means that the content provider has to query initial state of the
data.</li>
<li>Next the content provider tells the viewer how many elements
there are, by calling <span style="font-family: monospace;">TableViewer.setItemCount()</span>.</li>
<li>At this point, the table resizes, and it requests data values for
items that are visible.&nbsp; So for each visible item it calls: <span
style="font-family: monospace;">ILazyContentProvider.updateElement()</span>.</li>
<li>After calculating the value, the content provider tells the table
what the value is, by calling <span style="font-family: monospace;">TableViewer.replace().</span></li>
<li>If the data ever changes, the content provider tells the table to
rerequest the data, by calling <span style="font-family: monospace;">TableViewer.clear()</span>.</li>
</ol>
Table viewer operates in the
SWT display thread, which means that the content provider must switch
from the display thread to the DSF dispatch thread, whenever it is
called by the table viewer, as in the example below:<br>
<pre> public void updateElement(final int index) {<br> assert fTableViewer != null;<br> if (fDataProvider == null) return;<br><br> fDataProvider.getExecutor().execute(<br> new Runnable() { public void run() {<br> // Must check again, in case disposed while redispatching.<br> if (fDataProvider == null) return;<br> <br> queryItemData(index);<br> }});<br> }<br></pre>
Likewise, when the content provider calls the table viewer, it also has
to switch back into the display thread as in following example, when
the content provider receives an event from the data provider, that an
item value has changed.<br>
<pre> public void dataChanged(final Set&lt;Integer&gt; indexes) {<br> // Check for dispose.<br> if (fDataProvider == null) return;<br><br> // Clear changed items in table viewer.<br> if (fTableViewer != null) {<br> final TableViewer tableViewer = fTableViewer;<br> tableViewer.getTable().getDisplay().asyncExec(<br> new Runnable() { public void run() {<br> // Check again if table wasn't disposed when <br> // switching to the display thread.<br> if (tableViewer.getTable().isDisposed()) return; // disposed<br> for (Integer index : indexes) {<br> tableViewer.clear(index);<br> }<br> }});<br> }<br> }<br></pre>
All of this switching back and forth between threads makes the code
look a lot more complicated than it really is, and it takes some
getting used to, but this is the price to be paid for multi-threading.
Whether the participants use semaphores or the dispatch thread, the
logic is equally complicated, and we believe that using a single
dispatch thread, makes the synchronization very explicit and thus less
error-prone.<br>
<p><span style="text-decoration: underline;">Data Provider Service</span></p>
<p>The data provider service interface, <span
style="font-family: monospace;">DataProvider</span>, is very similar
to that of the lazy content provider.&nbsp; It has methods to: </p>
<ul>
<li>get item count</li>
<li>get a value for given item</li>
<li>register as listener for changes in data count and data values</li>
</ul>
But this is a DSF interface, and all methods must be called on the
service's dispatch thread.&nbsp; For this reason, the <span
style="font-family: monospace;">DataProvider </span>interface returns
an instance of <span style="font-family: monospace;">DsfExecutor</span>,
which must be used with the interface.<br>
<p><span style="text-decoration: underline;">Slow Data Provider</span></p>
<p>The data provider is actually implemented as a thread which is an
inner class of <span style="font-family: monospace;">SlowDataProvider</span>
service.&nbsp; The provider thread
communicates with the service by reading Request objects from a shared
queue, and by posting Runnable objects directly to the <span
style="font-family: monospace;">DsfExecutor</span> but
with a simulated transmission delay.&nbsp; Separately, an additional
flag is also used to control the shutdown of the provider thread.</p>
To simulate a real back end, the data provider randomly invalidates a
set of items and notifies the listeners to update themselves.&nbsp; It
also periodically invalidates the whole table and forces the clients to
requery all items.<br>
<h4>Data and Control Flow<br>
</h4>
This can be described in following steps:<br>
<ol>
<li>The table viewer requests data for an item at a given index (<span
style="font-family: monospace;">SlowDataProviderContentProvider.updateElement</span>).<br>
</li>
<li>The table viewer's content provider executes a <span
style="font-family: monospace;">Runnable </span>in the DSF
dispatch thread and calls the data provider interface (<span
style="font-family: monospace;">SlowDataProviderContentProvider.queryItemData</span>).</li>
<li>Data provider service creates a Request object, and files it in a
queue (<span style="font-family: monospace;">SlowDataProvider.getItem</span>).</li>
<li>Data provider thread de-queues the Request object and acts on it,
calculating the value (<span style="font-family: monospace;">ProviderThread.processItemRequest</span>).</li>
<li>Data provider thread schedules the calculation result to be
posted with DSF executor (<span style="font-family: monospace;">SlowDataProvider.java:185</span>).</li>
<li>The Done callback sets the result data in the table viewer (<span
style="font-family: monospace;">SlowDataProviderContentProvider.java:167</span>).<br>
</li>
</ol>
<h4>Running the example and full sources</h4>
This example is implemented in the <span
style="font-family: monospace;">org.eclipse.dd.dsf.examples</span>
plugin, in the <span style="font-family: monospace;">org.eclipse.dd.dsf.examples.concurrent</span>
package.&nbsp; <br>
<br>
To run the example:<br>
<ol>
<li>Build the test plugin (along with the <span
style="font-family: monospace;">org.eclipse.dsdp.DSF plugin</span>)
and launch the PDE.&nbsp; <br>
</li>
<li>Make sure to add the <span style="font-style: italic;">DSF
Tests</span> action set to your current perspective.</li>
<li>From the main menu, select <span style="font-style: italic;">DSF
Tests -&gt; Slow Data Provider</span>.</li>
<li>A dialog will open and after a delay it will populate with data.</li>
<li>Scroll and resize dialog and observe the update behavior.</li>
</ol>
<h4>Initial Notes<br>
</h4>
This example is supposed to be representative of a typical embedded
debugger design problem.&nbsp; Embedded debuggers are often slow in
retrieving and processing data, and can sometimes be accessed through a
relatively slow data channel, such as serial port or JTAG
connection.&nbsp; But as such, this basic example presents a couple
of major usability problems<br>
<ol>
<li>The data provider service interface mirrors the table's content
provider interface, in that it has a method to retrieve a single piece
of data at a time.&nbsp; The result of this is visible to the user as
lines of data are filled in one-by-one in the table.&nbsp; However,
most debugger back ends are in fact capable of retrieving data in
batches and are much more efficient at it than retrieving data items
one-by-one.</li>
<li>When scrolling quickly through the table, the requests are
generated by the table viewer for items which are quickly scrolled out
of view, but the service still queues them up and calculates them in
the order they were received.&nbsp; As a result, it takes a very long
time for the table to be populated with data at the location where the
user is looking.&nbsp; <br>
</li>
</ol>
These two problems are very common in creating UI for embedded
debugging, and there are common patterns which can be used to solve
these problems in DSF services. <br>
<h3>Coalescing</h3>
Coalescing many single-item requests into fewer multi-item requests is
the surest way to improve performance in communication with a remote
debugger, although it's not necessarily the simplest.&nbsp; There are
two basic patterns in which coalescing is achieved:<br>
<ol>
<li>The back end provides an interface for retrieving data in large
chunks.&nbsp; So when the service implementation receives a request for
a single item, it retrieves a whole chunk of data, returns the single
item, and stores the rest of the data in a local cache.</li>
<li>The back end providers an interface for retrieving data in
variable size chunks.&nbsp; When the service implementation receives a
request for a single item, it buffers the request, and waits for other
requests to come in.&nbsp; After a delay, the service clears the buffer
and submits a request for the combined items to the data provider.</li>
</ol>
In practice, a combination of the two patterns is needed, but for
purpose of an example, we implemented the second pattern in the
"Input-Coalescing Slow Data Provider" (<span
style="font-family: monospace;">InputCoalescingSlowDataProvider.java</span>).&nbsp;
<br>
<p><span style="text-decoration: underline;">Input Buffer</span></p>
<p>The main feature of this pattern is a buffer for holding the
requests before sending them to the data provider.&nbsp; In this
example the user requests are buffered in two arrays: <span
style="font-family: monospace;">fGetItemIndexesBuffer</span> and <span
style="font-family: monospace;">fGetItemDonesBuffer</span>.&nbsp; The
<span style="font-family: monospace;">DataProvider.getItem()</span>
implementation is changed as follows:</p>
<pre> public void getItem(final int index, final GetDataDone&lt;String&gt; done) {<br> // Schedule a buffer-servicing call, if one is needed.<br> if (fGetItemIndexesBuffer.isEmpty()) {<br> fExecutor.schedule(<br> new Runnable() { public void run() {<br> fileBufferedRequests();<br> }},<br> COALESCING_DELAY_TIME, <br> TimeUnit.MILLISECONDS);<br> }<br> <br> // Add the call data to the buffer. <br> // Note: it doesn't matter that the items were added to the buffer <br> // after the buffer-servicing request was scheduled. This is because<br> // the buffers are guaranteed not to be modified until this dispatch<br> // cycle is over.<br> fGetItemIndexesBuffer.add(index);<br> fGetItemDonesBuffer.add(done);<br> } <br><br></pre>
And method that services the buffer looks like this:<br>
<pre> public void fileBufferedRequests() { <br> // Remove a number of getItem() calls from the buffer, and combine them<br> // into a request.<br> int numToCoalesce = Math.min(fGetItemIndexesBuffer.size(), COALESCING_COUNT_LIMIT);<br> final ItemRequest request = new ItemRequest(new Integer[numToCoalesce], new GetDataDone[numToCoalesce]); <br> for (int i = 0; i &lt; numToCoalesce; i++) {<br> request.fIndexes[i] = fGetItemIndexesBuffer.remove(0);<br> request.fDones[i] = fGetItemDonesBuffer.remove(0);<br> }<br><br> // Queue the coalesced request, with the appropriate transmission delay.<br> fQueue.add(request);<br> <br> // If there are still calls left in the buffer, execute another <br> // buffer-servicing call, but without any delay.<br> if (!fGetItemIndexesBuffer.isEmpty()) {<br> fExecutor.execute(new Runnable() { public void run() {<br> fileBufferedRequests();<br> }});<br> }<br> }<br></pre>
The most interesting feature of this implementation is the fact that
there are no semaphores anywhere to control access to the input
buffers.&nbsp; Even though the buffers are serviced with a delay and
multiple clients can call the <span style="font-family: monospace;">getItem()</span>
method, the use of a single
dispatch thread prevents any race conditions that could corrupt the
buffer data.&nbsp; In real-world implementations, the buffers and
caches that need to be used are far more sophisticated with much more
complicated logic, and this is where managing access to them using the
dispatch thread is ever more important.<br>
<h3>Cancellability</h3>
<p><span style="text-decoration: underline;">Table Viewer</span></p>
<p><span style="text-decoration: underline;"></span></p>
Unlike coalescing, which can be implemented entirely within the
service, cancellability requires that the client be modified as well
to take advantage of this capability.&nbsp; For the table viewer
content provider, this means that additional features have to be
added.&nbsp; In <span style="font-family: monospace;">CancellingSlowDataProviderContentProvider.java</span>
<span style="font-family: monospace;">ILazyContentProvider.updateElement()</span>
was changes as follows:<br>
<pre> public void updateElement(final int index) {<br> assert fTableViewer != null;<br> if (fDataProvider == null) return;<br> <br> // Calculate the visible index range.<br> final int topIdx = fTableViewer.getTable().getTopIndex();<br> final int botIdx = topIdx + getVisibleItemCount(topIdx);<br> <br> fCancelCallsPending.incrementAndGet();<br> fDataProvider.getExecutor().execute(<br> new Runnable() { public void run() {<br> // Must check again, in case disposed while redispatching.<br> if (fDataProvider == null || fTableViewer.getTable().isDisposed()) return;<br> if (index &gt;= topIdx &amp;&amp; index &lt;= botIdx) {<br> queryItemData(index);<br> }<br> cancelStaleRequests(topIdx, botIdx);<br> }});<br> }<br></pre>
Now the client keeps track of the requests it made to the service in <span
style="font-family: monospace;">fItemDataDones</span>, and above, <span
style="font-family: monospace;">cancelStaleRequests()</span> iterates
through all the outstanding requests and cancels the ones that are no
longer in the visible range.<br>
<p><span style="text-decoration: underline;">Data Provider Service<span
style="text-decoration: underline;"></span></span></p>
<p><span style="text-decoration: underline;"><span
style="text-decoration: underline;"></span></span></p>
<p>The data provider implementation
(<span style="font-family: monospace;">CancellableInputCoalescingSlowDataProvider.java</span>),
builds on top of the
coalescing data provider.&nbsp; To make the canceling feature useful,
the data provider service has to limit the size of the request
queue.&nbsp; This is because in this example which simulates
communication with a target and once requests are filed into the
request
queue, they cannot be canceled, just like a client can't cancel
request once it sends them over a socket.&nbsp; So instead, if a flood
of <span style="font-family: monospace;">getItem()</span>
calls comes in, the service has to hold most of them in the coalescing
buffer in case the client decides to cancel them.&nbsp; Therefore the
<span style="font-family: monospace;">fileBufferedRequests()</span>
method includes a simple check before servicing
the buffer, and if the request queue is full, the buffer servicing call
is delayed.</p>
<pre> if (fQueue.size() &gt;= REQUEST_QUEUE_SIZE_LIMIT) {<br> if (fGetItemIndexesBuffer.isEmpty()) {<br> fExecutor.schedule(<br> new Runnable() { public void run() {<br> fileBufferedRequests();<br> }},<br> REQUEST_BUFFER_FULL_RETRY_DELAY, <br> TimeUnit.MILLISECONDS);<br> }<br> return;<br> } <br></pre>
Beyond this change, the only other significant change is that before
the requests are queued, they are checked for cancellation.<br>
<h3>Final Notes<br>
</h3>
The example given here is fairly simplistic, and chances are that the
same example could be implemented using semaphores and free threading
with perhaps fewer lines of code.&nbsp; But what we have found is that
as the problem gets bigger, the amount of
features in the data provider increases, the state of the
communication protocol gets more complicated, and the number of modules
needed in the service layer increases, using free threading and
semaphores does not safely scale.&nbsp; Using a dispatch thread for
synchronization certainly doesn't make the inherent problems of the
system less complicated, but it does help eliminate the race conditions
and deadlocks from the overall system.<br>
<p>Coalescing and Cancellability are both optimizations.&nbsp; Neither
of these optimizations affected the original interface of the service,
and one of them only needed a service-side modification.&nbsp; But as
with all optimizations, it is often better to first make sure that the
whole system is working correctly and then add optimizations where they
can make the biggest difference in user experience.&nbsp; </p>
<p>The above examples of optimizations can take many forms, and as
mentioned with coalescing, caching data that is retrieved from the data
provider is the most common form of data coalescing.&nbsp; For
cancellation, many services in DSF build on top of other services,
which means that even a low-level service can cause a higher
level service to retrieve data, while another event might cause it to
cancel those requests.&nbsp; The perfect example of this is a Variables
service, which is responsible for calculating the value of expressions
shown in the Variables view.&nbsp; The Variables service reacts to the
Run Control service, which issues a suspended event and then requests a
set of variables to be evaluated by the debugger back end.&nbsp; But as
soon as a resumed event is issued by Run Control, the Variables service
needs to cancel&nbsp; the pending evaluation requests.<br>
</p>
<br>
<br>
</body>
</html>