|
Boost-Commit : |
Subject: [Boost-commit] svn:boost r64556 - sandbox/SOC/2010/quasi_random/boost/random
From: jvd_at_[hidden]
Date: 2010-08-02 18:05:37
Author: qrng
Date: 2010-08-02 18:05:37 EDT (Mon, 02 Aug 2010)
New Revision: 64556
URL: http://svn.boost.org/trac/boost/changeset/64556
Log:
integer_pow optimization.
Text files modified:
sandbox/SOC/2010/quasi_random/boost/random/faure.hpp | 24 +++++++++++++++++++-----
1 files changed, 19 insertions(+), 5 deletions(-)
Modified: sandbox/SOC/2010/quasi_random/boost/random/faure.hpp
==============================================================================
--- sandbox/SOC/2010/quasi_random/boost/random/faure.hpp (original)
+++ sandbox/SOC/2010/quasi_random/boost/random/faure.hpp 2010-08-02 18:05:37 EDT (Mon, 02 Aug 2010)
@@ -72,12 +72,26 @@
return ilog;
}
-inline std::size_t integer_pow(std::size_t arg, std::size_t p)
+// Implements exponentiation by squaring, for p > ~4 this is computationally
+// more efficient than naïvely multiplying the base with itself repeatedly.
+// In erroneous situations, e.g., integer_pow(0, 0) the function returns 1
+// and does not report the error. This is the intended behavior.
+inline std::size_t integer_pow(std::size_t base, std::size_t p)
{
- std::size_t ipow = 1;
- while( p-- )
- ipow *= arg;
- return ipow;
+ std::size_t result = 1;
+ for( ; p != 0; p >>= 1, base *= base )
+ {
+ // A typical way to implement the multiplication
+ // would be something like this:
+ // if( p & 1 )
+ // result *= base;
+ // Apart from being simple this method, however,
+ // does not have a lot of other advantages, it is not, especially,
+ // friendly to CPU branch prediction routines.
+ std::size_t k = (p & 1); // k in {0,1}
+ result *= (k * base + !k);
+ }
+ return result;
}
// Computes a table of binomial coefficients modulo qs.
Boost-Commit list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk