http://embeddedgurus.com/stack-overflow/2009/06/division-of-integers-by-constants/ Thoughts on embedded systems by Nigel Jones Wed, 28 Jul 2010 00:59:13 +0000 http://wordpress.org/?v=2.9.1 hourly 1 http://embeddedgurus.com/stack-overflow/2009/06/division-of-integers-by-constants/comment-page-1/#comment-234 victorh Wed, 17 Jun 2009 19:44:33 +0000 http://www.gfcdev.org/test-stack/2009/06/05/division-of-integers-by-constants/#comment-234 Hello Nigel!In order to give you the exact figures, I have to undust my code. But I can tell you from the top of my head that the reciprocal multiplication is more than 10 times faster than the naive ones (i.e. sprintf and built-in division by 10)All the approaches were implemented in C, using HI-TECH PICC standard (not PRO) compiler.By the way, just yesterday Embedded.com published a nice article about this very same topic. Check it out.http://www.embedded.com/217900224?printable=true Hello Nigel!In order to give you the exact figures, I have to undust my code. But I can tell you from the top of my head that the reciprocal multiplication is more than 10 times faster than the naive ones (i.e. sprintf and built-in division by 10)All the approaches were implemented in C, using HI-TECH PICC standard (not PRO) compiler.By the way, just yesterday Embedded.com published a nice article about this very same topic. Check it out.http://www.embedded.com/217900224?printable=true

]]> http://embeddedgurus.com/stack-overflow/2009/06/division-of-integers-by-constants/comment-page-1/#comment-233 Nigel Jones Wed, 10 Jun 2009 11:36:04 +0000 http://www.gfcdev.org/test-stack/2009/06/05/division-of-integers-by-constants/#comment-233 A fellow fan! Did you implement it in assembly language or C? Also what was the magnitude of the variation between the various methods? I ask because I'm intrigued to see how the method ports to other architectures. A fellow fan! Did you implement it in assembly language or C? Also what was the magnitude of the variation between the various methods? I ask because I'm intrigued to see how the method ports to other architectures.

]]> http://embeddedgurus.com/stack-overflow/2009/06/division-of-integers-by-constants/comment-page-1/#comment-232 victorh Tue, 09 Jun 2009 20:11:35 +0000 http://www.gfcdev.org/test-stack/2009/06/05/division-of-integers-by-constants/#comment-232 Some time ago, I needed to implement an efficient and effective way (in terms of program size, RAM usage and speed) to convert an uint32_t to a BCD on a PIC16F690. After trying and finetuning such different approaches as the compiler built-in division, sprintf, <a href="http://forum.htsoft.com/all/showthreaded.php/Cat/0/Number/15168/" rel="nofollow">successive subtractions</a>, and some <a href="http://www.microchipc.com/sourcecode/PIC_Hi-Tech_C_decimal_output_routines.zip" rel="nofollow">faster</a>-but-<a href="http://forum.htsoft.com/all/showthreaded.php/Cat/0/Number/13307/" rel="nofollow">obscure</a> algorithms, a reciprocal multiplication implemented with <a href="http://www.cs.uiowa.edu/~jones/bcd/divide.html#nomul" rel="nofollow">shifts and additions</a> resulted in a non-convoluted method, and the fastest one I benchmarked. Truly, Douglas Jones' paper was a lifesaver for me. Really good stuff. Some time ago, I needed to implement an efficient and effective way (in terms of program size, RAM usage and speed) to convert an uint32_t to a BCD on a PIC16F690. After trying and finetuning such different approaches as the compiler built-in division, sprintf, successive subtractions, and some faster-but-obscure algorithms, a reciprocal multiplication implemented with shifts and additions resulted in a non-convoluted method, and the fastest one I benchmarked. Truly, Douglas Jones' paper was a lifesaver for me. Really good stuff.

]]>