Loop unrolling

From Wikipedia, the free encyclopedia
  (Redirected from Loop unwinding)
Jump to: navigation, search

Loop unrolling, also known as loop unwinding, is a loop transformation technique that attempts to optimize a program's execution speed at the expense of its binary size (space-time tradeoff). The transformation can be undertaken manually by the programmer or by an optimizing compiler.

The goal of loop unwinding is to increase a program's speed by reducing (or eliminating) instructions that control the loop, such as pointer arithmetic and "end of loop" tests on each iteration;[1] reducing branch penalties; as well as "hiding latencies, in particular, the delay in reading data from memory".[2] To eliminate this overhead, loops can be re-written as a repeated sequence of similar independent statements.[3]

Advantages[edit]

The overhead in "tight" loops often consists of instructions to increment a pointer or index to the next element in an array (pointer arithmetic), as well as "end of loop" tests. If an optimizing compiler or assembler is able to pre-calculate offsets to each individually referenced array variable, these can be built into the machine code instructions directly, therefore requiring no additional arithmetic operations at run time (note that in the example given below this is not the case).

  • Significant gains can be realized if the reduction in executed instructions compensates for any performance reduction caused by any increase in the size of the program.
  • branch penalty is minimised.[4]
  • If the statements in the loop are independent of each other (i.e. where statements that occur earlier in the loop do not affect statements that follow them), the statements can potentially be executed in parallel.
  • Can be implemented dynamically if the number of array elements is unknown at compile time (as in Duff's device)

Optimizing compilers will sometimes perform the unrolling automatically, or upon request.

Disadvantages[edit]

  • Increased program code size, which can be undesirable, particularly for embedded applications.
    • Can also cause an increase in instruction cache misses, which may adversely affect performance.
  • Unless performed transparently by an optimizing compiler, the code may become less readable.
  • If the code in the body of the loop involves function calls, it may not be possible to combine unrolling with inlining, since the increase in code size might be excessive. Thus there can be a trade-off between the two optimizations.
  • Possible increased register usage in a single iteration to store temporary variables[dubious ], which may reduce performance, though much will depend on possible optimizations.[5]
  • Apart from very small and simple codes, unrolled loops that contain branches are even slower than recursions [6]

Static/manual loop unrolling[edit]

Manual (or static) loop unrolling involves the programmer analyzing the loop and interpreting the iterations into a sequence of instructions which will reduce the loop overhead. This is in contrast to dynamic unrolling which is accomplished by the compiler.

A simple manual example in C[edit]

A procedure in a computer program is to delete 100 items from a collection. This is normally accomplished by means of a for-loop which calls the function delete(item_number). If this part of the program is to be optimized, and the overhead of the loop requires significant resources compared to those for the delete(x) loop, unwinding can be used to speed it up.

Normal loop After loop unrolling
 int x;
 for (x = 0; x < 100; x++)
 {
     delete(x);
 }
 int x; 
 for (x = 0; x < 100; x+=5)
 {
     delete(x);
     delete(x+1);
     delete(x+2);
     delete(x+3);
     delete(x+4);
 }

As a result of this modification, the new program has to make only 20 iterations, instead of 100. Afterwards, only 20% of the jumps and conditional branches need to be taken, and represents, over many iterations, a potentially significant decrease in the loop administration overhead. To produce the optimal benefit, no variables should be specified in the unrolled code that require pointer arithmetic. This usually requires "base plus offset" addressing, rather than indexed referencing.

On the other hand, this manual loop unrolling expands the source code size from 3 lines to 7, that have to be produced, checked, and debugged, and the compiler may have to allocate more registers to store variables in the expanded loop iteration[dubious ]. In addition, the loop control variables and number of operations inside the unrolled loop structure have to be chosen carefully so that the result is indeed the same as in the original code (assuming this is a later optimization on already working code). For example, consider the implications if the iteration count were not divisible by 5. The manual amendments required also become somewhat more complicated if the test conditions are variables. See also Duff's device.

Early complexity[edit]

In the simple case, the loop control is merely an administrative overhead that arranges the productive statements. The loop itself contributes nothing to the results desired, merely saving the programmer the tedium of replicating the code a hundred times which could have been done by a pre-processor generating the replications, or a text editor. Similarly, if-statements and other flow control statements could be replaced by code replication, except that code bloat can be the result. Computer programs easily track the combinations, but programmers find this repetition boring and make mistakes. Consider:

Normal loop After loop unrolling
for i := 1:8 do
    if i mod 2 = 0 then do_evenstuff(i) 
                   else do_oddstuff(i);
    next i;
do_oddstuff(1); do_evenstuff(2);
do_oddstuff(3); do_evenstuff(4);
do_oddstuff(5); do_evenstuff(6);
do_oddstuff(7); do_evenstuff(8);

But of course, the code performed need not be the invocation of a procedure, and this next example involves the index variable in computation:

Normal loop After loop unrolling
x(1) := 1;
For i := 2:9 do
    x(i) := x(i - 1)*i;
    print i,x(i);
    next i;
x(1) := 1;
x(2) := x(1)*2; print 2,x(2);
x(3) := x(2)*3; print 3,x(3);
x(4) := x(3)*4; print 4,x(4);
...etc.

which, if compiled, might produce a lot of code (print statements being notorious) but further optimization is possible. This example make reference only to x(i) and x(i - 1) in the loop (the latter only to develop the new value x(i)) therefore, given that there is no later reference to the array x developed here, its usages could be replaced by a simple variable. Such a change would however mean a simple variable whose value is changed whereas if staying with the array, the compiler's analysis might note that the array's values are constant, each derived from a previous constant, and therefore carries forward the constant values so that the code becomes

print 2,2;
print 3,6;
print 4,24;
...etc.

It would be quite a surprise if the compiler were to recognise x(n) = Factorial(n).

In general, the content of a loop might be large, involving intricate array indexing. These cases are probably best left to optimizing compilers to unroll. Replicating innermost loops might allow many possible optimisations yet yield only a small gain unless n is large.

Unrolling WHILE loops[edit]

A pseudocode WHILE loop - similar to the following -

Normal loop After loop unrolling Unrolled & "tweaked" loop
  WHILE (condition) DO
         action
  ENDWHILE
.
.
.
.
.
.
  WHILE (condition) DO
         action
         IF NOT(condition) THEN GOTO break
         action
         IF NOT(condition) THEN GOTO break
         action
 ENDWHILE
 LABEL break:
.
 IF (condition) THEN
  REPEAT {
         action
         IF NOT(condition) THEN GOTO break
         action
         IF NOT(condition) THEN GOTO break
          action
         } WHILE (condition)
 LABEL break:

Unrolling is faster because the ENDWHILE (that will be compiled to a jump to the start of the loop) will be executed 66% less often.

Even better, the "tweaked" pseudocode example, that may be performed automatically by some optimizing compilers, eliminating unconditional jumps altogether.

Dynamic unrolling[edit]

Since the benefits of loop unrolling are frequently dependent on the size of an array—which may often not be known until run time—JIT compilers (for example) can determine whether to invoke a "standard" loop sequence or instead generate a (relatively short) sequence of individual instructions for each element. This flexibility is one of the advantages of just-in-time techniques versus static or manual optimization in the context of loop unrolling. In this situation, it is often with relatively small values of n where the savings are still useful—requiring quite small (if any) overall increase in program size (that might be included just once, as part of a standard library).

Assembly language programmers (including optimizing compiler writers) are also able to benefit from the technique of dynamic loop unrolling, using a method similar to that used for efficient branch tables. Here the advantage is greatest where the maximum offset of any referenced field in a particular array is less than the maximum offset that can be specified in a machine instruction (which will be flagged by the assembler if exceeded). The example below is for IBM/360 or Z/Architecture assemblers and assumes a field of 100 bytes (at offset zero) is to be copied from array FROM to array TO—both having element lengths of 256 bytes with 50 entries

Assembler example (IBM/360 or Z/Architecture)[edit]

For an x86 example, see External Links.
* initialize all the registers to point to arrays etc  (R14 is return address)
         LM    R15,R2,INIT                       load R15= '16', R0=number in array, R1--> 'FROM array', R2--> 'TO array'
LOOP     EQU   *
         SR    R15,R0                            get 16 minus the number in the array
         BNP   ALL                               if n > 16 need to do all of the sequence, then repeat
* (if # entries = zero, R15 will now still be 16, so all the MVC's will be bypassed)
* calculate an offset (from start of MVC sequence) for unconditional branch to 'unwound' MVC loop
         MH    R15,=AL2(ILEN)                    multiply by length of (MVC..) instruction (=6 in this example)
         B     ALL(R15)                          indexed branch instruction (to MVC with drop through)
* Assignment (move) table - (first entry has maximum allowable offset with single register = X'F00' in this example)
ALL      MVC   15*256(100,R2),15*256(R1)         * move 100 bytes of 16th entry  from array 1 to array 2 (with drop through)
ILEN     EQU   *-ALL                                         length of (MVC...) instruction sequence; in this case =6
         MVC   14*256(100,R2),14*256(R1)         *
         MVC   13*256(100,R2),13*256(R1)         *
         MVC   12*256(100,R2),12*256(R1)         * All 16 of these 'move character' instructions use base plus offset addressing
         MVC   11*256(100,R2),11*256(R1)         * and each to/from offset decreases by the length of one array element (256).
         MVC   10*256(100,R2),10*256(R1)         * This avoids pointer arithmetic being required for each element up to a 
         MVC   09*256(100,R2),09*256(R1)         * maximum permissible offset within the instruction of X'FFF'. The instructions 
         MVC   08*256(100,R2),08*256(R1)         * are in order of decreasing offset, so the first element in the set is moved
         MVC   07*256(100,R2),07*256(R1)         * last.
         MVC   06*256(100,R2),06*256(R1)         *
         MVC   05*256(100,R2),05*256(R1)         *
         MVC   04*256(100,R2),04*256(R1)         *
         MVC   03*256(100,R2),03*256(R1)         *
         MVC   02*256(100,R2),02*256(R1)         *
         MVC   01*256(100,R2),01*256(R1)         move 100 bytes of 2nd entry
         MVC   00*256(100,R2),00*256(R1)         move 100 bytes of 1st entry
*
         S     R0,MAXM1                          reduce Count = remaining entries to process
         BNPR  R14                               ... no more, so return to address in R14
         AH    R1,=AL2(16*256)                   increment 'FROM' register pointer beyond first set
         AH    R2,=AL2(16*256)                   increment 'TO'   register pointer beyond first set
         L     R15,MAXM1                         re-load (maximum MVC's) in R15 (destroyed by calculation earlier)
         B     LOOP                              go and execute loop again
*
* ----- Define static Constants and variables (These could be passed as parameters) ---------------------------------  *
INIT     DS    0A                                4 addresses (pointers) to be pre-loaded with a 'LM' instruction
MAXM1    DC    A(16)                             maximum MVC's 
N        DC    A(50)                             number of actual entries in array (a variable, set elsewhere)
         DC    A(FROM)                           address of start of array 1 ("pointer")
         DC    A(TO)                             address of start of array 2 ("pointer")
* ----- Define static Arrays (These could be dynamically acquired) --------------------------------------------------  *
FROM     DS    50CL256                          array of (max) 50 entries of 256 bytes each
TO       DS    50CL256                          array of (max) 50 entries of 256 bytes each

In this example, approximately 202 instructions would be required with a 'conventional' loop (50 iterations) whereas the above dynamic code would require only about 89 instructions (or a saving of approximately 56%). If the array had consisted of only 2 entries, it would still execute in approximately the same time as the original unwound loop. The increase in code size is only about 108 bytes - even if there are thousands of entries in the array. Similar techniques can of course be used where multiple instructions are involved, as long as the combined instruction length is adjusted accordingly. For example, in this same example, if it is required to clear the rest of each array entry to nulls immediately after the 100 byte field copied, an additional clear instruction, 'XC xx*256+100(156,R1),xx*256+100(R2)', can be added immediately after every MVC in the sequence (where 'xx' matches the value in the MVC above it).

It is, of course, perfectly possible to generate the above code 'inline' using a single assembler macro statement, specifying just 4 or 5 operands (or alternatively, make it into a library subroutine, accessed by a simple call, passing a list of parameters), making the optimization readily accessible to inexperienced programmers.

C example[edit]

The following example demonstrates dynamic loop unrolling for a simple program written in C. Unlike the assembler example above, pointer/index arithmetic is still generated by the compiler in this example because a variable (i) is still used to address the array element. Full optimization is only possible if absolute indexes are used in the replacement statements.

#include<stdio.h>
 
#define TOGETHER (8)
 
int main(void)
{ 
 int i = 0; 
 int entries = 50;                                 /* total number to process     */
 int repeat;                                       /* number of times for while.. */
 int left = 0;                                     /* remainder (process later)   */ 
 
 /* If the number of elements is not be divisible by BLOCKSIZE,                   */ 
 /* get repeat times required to do most processing in the while loop             */
 
 repeat = (entries / TOGETHER);                    /* number of times to repeat   */
 left   = (entries % TOGETHER);                    /* calculate remainder         */
 
 /* Unroll the loop in 'bunches' of 8                                             */ 
 while (repeat--) 
  { 
    printf("process(%d)\n", i    );
    printf("process(%d)\n", i + 1); 
    printf("process(%d)\n", i + 2); 
    printf("process(%d)\n", i + 3); 
    printf("process(%d)\n", i + 4); 
    printf("process(%d)\n", i + 5); 
    printf("process(%d)\n", i + 6); 
    printf("process(%d)\n", i + 7);
 
    /* update the index by amount processed in one go                            */ 
    i += TOGETHER; 
  }
 
 /* Use a switch statement to process remaining by jumping to the case label     */ 
 /* at the label that will then drop through to complete the set                 */ 
 switch (left) 
  {
     case 7 : printf("process(%d)\n", i + 6);      /* process and rely on drop through */
     case 6 : printf("process(%d)\n", i + 5); 
     case 5 : printf("process(%d)\n", i + 4);  
     case 4 : printf("process(%d)\n", i + 3);  
     case 3 : printf("process(%d)\n", i + 2); 
     case 2 : printf("process(%d)\n", i + 1);      /* two left                                      */
     case 1 : printf("process(%d)\n", i);          /* just one left to process                      */ 
     case 0 : ;                                    /* none left                                     */
  } 
}

See also[edit]

References[edit]

  1. ^ Ullman, Jeffrey D.; Aho, Alfred V. (1977). Principles of compiler design. Reading, Mass: Addison-Wesley Pub. Co. pp. 471–2. ISBN 0-201-10073-8. 
  2. ^ Petersen, W.P., Arbenz, P. (2004). Introduction to Parallel Computing. Oxford University Press. p. 10. 
  3. ^ Nicolau, Alexandru (1985). "Loop Quantization: Unwinding for Fine-Grain Parallelism Exploitation". Dept. of Computer Science Technical Report. Ithaca, NY: Cornell University. OCLC 14638257. 
  4. ^ Fog, Agner (2012-02-29). "Optimizing subroutines in assembly language". Copenhagen University College of Engineering. p. 100. Retrieved 2012-09-22. "12.11 Loop unrolling" 
  5. ^ Sarkar, Vivek (2001). "Optimized Unrolling of Nested Loops". International Journal of Parallel Programming 29 (5): 545–581. doi:10.1023/A:1012246031671. 
  6. ^ Adam Horvath "Code unwinding - performance is far away"

Further reading[edit]

  • Kennedy, Ken; Allen, Randy (2001). Optimizing Compilers for Modern Architectures: A Dependence-based Approach. Morgan Kaufmann. ISBN 1-55860-286-0. 

External links[edit]