Duff's device
|
|
In computer science, Duff's Device is an optimized implementation of a serial copy that uses a technique widely applied in assembly language for loop unwinding. Its discovery is credited to Tom Duff in November of 1983, who at the time was working for Lucasfilm. It is perhaps the most dramatic use of case label fall-through in the C programming language to date. Note that Duff does not claim credit for discovering the concept of loop unrolling, just this particular expression of it in C.
While optimizing a serial copy, Duff realized that an unrolled version of his loop could be implemented by interlacing the structures of a switch and a loop:
int n = (count + 7) / 8; /* count > 0 assumed */
switch (count % 8)
{
case 0: do { *to = *from++;
case 7: *to = *from++;
case 6: *to = *from++;
case 5: *to = *from++;
case 4: *to = *from++;
case 3: *to = *from++;
case 2: *to = *from++;
case 1: *to = *from++;
} while (--n > 0);
}
Based on a widely-used algorithm used by assemblers for minimizing the number of tests and branches during a copy, it appears out of place when implemented in C. The device is indeed perfectly valid, legal C, and many compilers will optimize the switch into a jump table just as would be done in an assembler implementation. C's default fall-through in case statements has long been its most controversial single feature; Duff observed that "This code forms some sort of argument in that debate, but I'm not sure whether it's for or against."
- This article was originally based on material from the Free On-line Dictionary of Computing, which is licensed under the GFDL.
Additional note: The original Device was made for copying to a register. To actually copy memory from one location to another, you must add an auto-increment to every reference to to, like so:
*to++ = *from++;
This modified form of the Device appears as a "what does this code do?" exercise in Bjarne Stroustrup's book The C++ Programming Language, presumably because novice programmers cannot be expected to know about memory-mapped output registers. However, it is not particularly useful in this form since the standard C library already supplies a memory-copying function which is likely to be even better optimized.
Books
- Bjarne Stroustrup: The C++ Programming Language, Third Edition, Addison-Wesley, ISBN 0-201-88954-4
External links
- Description and original mail by Duff at Lysator (http://www.lysator.liu.se/c/duffs-device.html)
- Article at FOLDOC (http://foldoc.doc.ic.ac.uk/foldoc/foldoc.cgi?Duff%27s+device)
- Article at the Jargon File (http://www.catb.org/~esr/jargon/html/D/Duffs-device.html)
- Google copy of original USENET post (http://groups-beta.google.com/group/net.lang.c/msg/66008138e07aa94c)