Editing Cilk (section)

===Task parallelism: spawn and sync===
{{See also|Fork–join model}}

Cilk's main addition to C are two keywords that together allow writing task-parallel programs.

* The {{mono|spawn}} keyword, when preceding a function call ({{mono|spawn f(x)}}), indicates that the function call ({{mono|f(x)}}) can safely run in parallel with the statements following it in the calling function. Note that the scheduler is not ''obligated'' to run this procedure in parallel; the keyword merely alerts the scheduler that it can do so.
* A {{mono|sync}} statement indicates that execution of the current function cannot proceed until all previously spawned function calls have completed. This is an example of a [[barrier (computer science)|barrier]] method.

(In Cilk Plus, the keywords are spelled {{mono|_Cilk_spawn}} and {{mono|_Cilk_sync}}, or {{mono|cilk_spawn}} and {{mono|cilk_sync}} if the Cilk Plus headers are included.)

Below is a [[recursion|recursive]] implementation of the [[Fibonacci number|Fibonacci]] function in Cilk, with parallel recursive calls, which demonstrates the {{mono|spawn}}, and {{mono|sync}} keywords. The original Cilk required any function using these to be annotated with the {{mono|cilk}} keyword, which is gone as of Cilk Plus. (Cilk program code is not numbered; the numbers have been added only to make the discussion easier to follow.)

<syntaxhighlight lang="c" line highlight="8,9,11">
cilk int fib(int n) {
    if (n < 2) {
        return n;
    }
    else {
       int x, y;

       x = spawn fib(n - 1);
       y = spawn fib(n - 2);

       sync;

       return x + y;
    }
}
</syntaxhighlight>

If this code was executed by a ''single'' processor to determine the value of {{mono|fib(2)}}, that processor would create a [[Stack frame|frame]] for {{mono|fib(2)}}, and execute lines 1 through 5. On line 6, it would create spaces in the frame to hold the values of {{mono|x}} and {{mono|y}}. On line 8, the processor would have to suspend the current frame, create a new frame to execute the procedure {{mono|fib(1)}}, execute the code of that frame until reaching a return statement, and then resume the {{mono|fib(2)}} frame with the value of fib(1) placed into {{mono|fib(2)}}<nowiki>'s</nowiki> {{mono|x}} variable. On the next line, it would need to suspend again to execute {{mono|fib(0)}} and place the result in {{mono|fib(2)}}<nowiki>'s</nowiki> {{mono|y}} variable.

When the code is executed on a ''multiprocessor'' machine, however, execution proceeds differently. One processor starts the execution of {{mono|fib(2)}}; when it reaches line 8, however, the {{mono|spawn}} keyword modifying the call to {{mono|fib(n-1)}} tells the processor that it can safely give the job to a second processor: this second processor can create a frame for {{mono|fib(1)}}, execute its code, and store its result in {{mono|fib(2)}}<nowiki>'s</nowiki> frame when it finishes; the first processor continues executing the code of {{mono|fib(2)}} at the same time. A processor is not obligated to assign a spawned procedure elsewhere; if the machine only has two processors and the second is still busy on {{mono|fib(1)}} when the processor executing {{mono|fib(2)}} gets to the procedure call, the first processor will suspend {{mono|fib(2)}} and execute {{mono|fib(0)}} itself, as it would if it were the only processor. Of course, if another processor is available, then it will be called into service, and all three processors would be executing separate frames simultaneously.

(The preceding description is not entirely accurate. Even though the common terminology for discussing Cilk refers to processors making the decision to spawn off work to other processors, it is actually the scheduler which assigns procedures to processors for execution, using a policy called ''work-stealing'', described later.)

If the processor executing {{mono|fib(2)}} were to execute line 13 before both of the other processors had completed their frames, it would generate an incorrect result or an error; {{mono|fib(2)}} would be trying to add the values stored in {{mono|x}} and {{mono|y}}, but one or both of those values would be missing. This is the purpose of the {{mono|sync}} keyword, which we see in line 11: it tells the processor executing a frame that it must suspend its own execution until all the procedure calls it has spawned off have returned. When {{mono|fib(2)}} is allowed to proceed past the {{mono|sync}} statement in line 11, it can only be because {{mono|fib(1)}} and {{mono|fib(0)}} have completed and placed their results in {{mono|x}} and {{mono|y}}, making it safe to perform calculations on those results.

The code example above uses the syntax of Cilk-5. The original Cilk (Cilk-1) used a rather different syntax that required programming in an explicit [[continuation-passing style]], and the Fibonacci examples looks as follows:<ref>{{cite conference |url=http://supertech.csail.mit.edu/papers/PPoPP95.pdf |title=Cilk: An Efficient Multithreaded Runtime System |first1=Robert D. |last1=Blumofe |first2=Christopher F. |last2=Joerg |first3=Bradley C. |last3=Kuszmaul |first4=Charles E. |last4=Leiserson |first5=Keith H. |last5=Randall |first6=Yuli |last6=Zhou |conference=Proc. ACM SIGPLAN [[Symposium on Principles and Practice of Parallel Programming|Symp. Principles and Practice of Parallel Programming]] |pages=207–216 |year=1995}}</ref>

<syntaxhighlight lang="c">
thread fib(cont int k, int n)
{
    if (n < 2) {
        send_argument(k, n);
    }
    else {
        cont int x, y;
        spawn_next sum(k, ?x, ?y);
        spawn fib(x, n - 1);
        spawn fib(y, n - 2);
    }
}

thread sum(cont int k, int x, int y)
{
     send_argument(k, x + y);
}
</syntaxhighlight>

Inside {{mono|fib}}'s recursive case, the {{mono|spawn_next}} keyword indicates the creation of a ''successor'' thread (as opposed to the ''child'' threads created by {{mono|spawn}}), which executes the {{mono|sum}} subroutine after waiting for the ''continuation variables'' {{mono|x}} and {{mono|y}} to be filled in by the recursive calls. The base case and {{mono|sum}} use a {{mono|send_argument(k, n)}} operation to set their continuation variable {{mono|k}} to the value of {{mono|n}}, effectively "returning" the value to the successor thread.