Hello there. I am implementing matrix multiplication and am trying to have my stuff run as fast as possible. But then I noticed something strange. Using Xcode, resultmatrix = prod(matrix1,matrix2); is 10 to 16 times slower (with 100x100 matrixes) than simply doing something like that: for(...) for(...) for(..) multiply_things(); And 20 to 40 times slower than a simple threaded approach. Using visual studio, it's 2 times faster than naive multiplication. This is on a release build, with #define BOOST_UBLAS_NDEBUG 1 #define NDEBUG 1 So what gives?
On 30 Jun 2010, at 16:55, Raphaël Amiot wrote:
Hello there.
I am implementing matrix multiplication and am trying to have my stuff run as fast as possible. But then I noticed something strange. Using Xcode,
resultmatrix = prod(matrix1,matrix2);
is 10 to 16 times slower (with 100x100 matrixes) than simply doing something like that:
for(...) for(...) for(..) multiply_things();
And 20 to 40 times slower than a simple threaded approach.
Using visual studio, it's 2 times faster than naive multiplication.
This is on a release build, with #define BOOST_UBLAS_NDEBUG 1 #define NDEBUG 1
So what gives?_______________________________________________
Could you provide a small complete program which shows your problem? Chris
Code here. You'll need to change the path to boost headers in the includes to make it run.
Part that runs way too slow:
chrono.begin();
for(int i=0; i On 30 Jun 2010, at 16:55, Raphaël Amiot wrote: Hello there. I am implementing matrix multiplication and am trying to have my stuff run as fast as possible.
But then I noticed something strange.
Using Xcode, resultmatrix = prod(matrix1,matrix2); is 10 to 16 times slower (with 100x100 matrixes) than simply doing something like that: for(...)
for(...)
for(..)
multiply_things(); And 20 to 40 times slower than a simple threaded approach. Using visual studio, it's 2 times faster than naive multiplication. This is on a release build, with
#define BOOST_UBLAS_NDEBUG 1
#define NDEBUG 1 So what gives?_______________________________________________ Could you provide a small complete program which shows your problem? Chris _______________________________________________
Boost-users mailing list
Boost-users@lists.boost.org
http://lists.boost.org/mailman/listinfo.cgi/boost-users
participants (2)
-
Christopher Jefferson -
Raphaël Amiot