diff options
Diffstat (limited to 'docs/tutorials/013/page07.html')
| -rw-r--r-- | docs/tutorials/013/page07.html | 41 |
1 files changed, 21 insertions, 20 deletions
diff --git a/docs/tutorials/013/page07.html b/docs/tutorials/013/page07.html index 509e326adf9..d2e22b2bdde 100644 --- a/docs/tutorials/013/page07.html +++ b/docs/tutorials/013/page07.html @@ -1,3 +1,4 @@ +<!-- $Id$ --> <HTML> <HEAD> <META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1"> @@ -16,23 +17,23 @@ <P> I've been trying to justify the chain of tasks by talking about a Work object that implements a state machine. The idea is that your -Work object has to perform a series of discrete steps to complete it's +Work object has to perform a series of discrete steps to complete it's function. Traditionally, all of those steps would take place in one thread of execution. That thread would probably be one from a Task thread pool. <P> -Suppose, however, that some of those steps spend a lot of time waiting -for disk IO. You could find that all of your thread-pool threads +Suppose, however, that some of those steps spend a lot of time waiting +for disk IO. You could find that all of your thread-pool threads are just sitting there waiting for the disk. You might then be tempted to increase the thread pool size to get more work through. -However, if some of the stages are memory intensive, you could run out +However, if some of the stages are memory intensive, you could run out of memory if all of the workers get to that state at the same time. <P> One solution might be to have different thread pools for each state. Each pool could have it's size tuned appropriately for the work that -would be done there. That's where the chain of Tasks comes in. - In this tutorial's implementation I've taken the -easy route and set all of the thread pools to the same size but a more +would be done there. That's where the chain of Tasks comes in. + In this tutorial's implementation I've taken the +easy route and set all of the thread pools to the same size but a more realistic solution would be to set each thread pool in the chain to a specific size as needed by that state of operation. <P> @@ -123,7 +124,7 @@ protected: <font color=blue>#include</font> "<font color=green>work.h</font>" <font color=red>/* - Initialize the state to zero + Initialize the state to zero */</font> <font color=#008888>Unit_Of_Work::Unit_Of_Work</font> (void) : state_ (0) @@ -137,7 +138,7 @@ protected: } <font color=red>/* - Display our instance value + Display our instance value */</font> void <font color=#008888>Unit_Of_Work::who_am_i</font> (void) { @@ -145,7 +146,7 @@ void <font color=#008888>Unit_Of_Work::who_am_i</font> (void) } <font color=red>/* - Dispay our type name + Dispay our type name */</font> void <font color=#008888>Unit_Of_Work::what_am_i</font> (void) { @@ -153,7 +154,7 @@ void <font color=#008888>Unit_Of_Work::what_am_i</font> (void) } <font color=red>/* - Return failure. You should always derive from Unit_Of_Work... + Return failure. You should always derive from Unit_Of_Work... */</font> int <font color=#008888>Unit_Of_Work::process</font> (void) { @@ -161,7 +162,7 @@ int <font color=#008888>Unit_Of_Work::process</font> (void) } <font color=red>/* - ditto + ditto */</font> int <font color=#008888>Unit_Of_Work::fini</font> (void) { @@ -169,7 +170,7 @@ int <font color=#008888>Unit_Of_Work::fini</font> (void) } <font color=red>/* - Default constructor has no "<font color=green>message number</font>" + Default constructor has no "<font color=green>message number</font>" */</font> <font color=#008888>Work::Work</font> (void) :message_ (-1) @@ -179,7 +180,7 @@ int <font color=#008888>Unit_Of_Work::fini</font> (void) <font color=red>/* The useful constructor remembers which message it is and will tell you if - you ask. + you ask. */</font> <font color=#008888>Work::Work</font> (int message) : message_ (message) @@ -193,7 +194,7 @@ int <font color=#008888>Unit_Of_Work::fini</font> (void) } <font color=red>/* - This objects type name is different from the baseclass + This objects type name is different from the baseclass */</font> void <font color=#008888>Work::what_am_i</font> (void) { @@ -202,7 +203,7 @@ void <font color=#008888>Work::what_am_i</font> (void) <font color=red>/* A very simple state machine that just walks through three stages. If it is - called more than that, it will tell you not to bother. + called more than that, it will tell you not to bother. */</font> int <font color=#008888>Work::process</font> (void) { @@ -227,8 +228,8 @@ int <font color=#008888>Work::process</font> (void) <font color=red>/* If you don't have enough subtasks in the chain then the state machine won't - progress to the end. The fini() hook will allow us to recover from that by - executing the remaining states in the final task of the chain. + progress to the end. The fini() hook will allow us to recover from that by + executing the remaining states in the final task of the chain. */</font> int <font color=#008888>Work::fini</font> (void) { @@ -247,9 +248,9 @@ int <font color=#008888>Work::fini</font> (void) <P> And that is that. For a more complex machine that may want to "jump states" you would have to set some "state information" (sorry, bad -choice of terminology again) so that process() could decide what to do +choice of terminology again) so that process() could decide what to do at each call. You might also modify Task::svc() so that it will -respect the return value of process() and do something useful with the +respect the return value of process() and do something useful with the information. <P> <P><HR WIDTH="100%"> |