|
|
Blitz Devel : |
From: Suresh kumar (mdsuresh__remove_this__123213_till__13_here_at_[hidden])
Date: 2002-11-02 04:49:23
If Todd, or Julian Cummings already sent me a email about this, please
resend it to mdsuresh_at_[hidden] I have lost access to my
yahoo account.
please remove REMOVEME from my email address.
Anyways guys, i have forked blitz a little and made a lot of controversial
changes. Btw, these are far from official.
You can find these changes at
http://mdsuresh.tripod.com/blitz-0.6.x.zip
Btw, i have only got it compiling. [with g++ 2.95, Intel, Cygwin B19,
WinXP ] I still have a lot of bugs to hunt down and fix. In a few days, i
will get them working. Here's the pilot compilability tester .
http://mdsuresh.tripod.com/func.cpp
Anyways, here are some important changes i made
First thing is Range
Range(0, 5,2) is considered a valid range and wont invoke an assertion.
Range(0,5,2) = 0,2,4
Range(0,-1,1) is a null range and wont invoke an assertion
IF Array<int,2> A(4,5)
A(Range(0,-1,1), 2) is a null slice
Slicing [ partial ordering]
Suppose Array<int,3> A(3,4,5)
and u do A( Range(2,3) , 1) --> this is same as A( Range(2,3) , 1,
Range:all())
and its rank is 2
I added some classes being a matlab phenatic.
lhs_ and rhs_ and the relavent function lhs and rhs
for example, SVDs can return multiple values using rhs() .
int a; float b; char c;
lhs( a, b, c) = rhs( int(1) , float(0.5) , '1')
would set a,b,c to their appropriate values.
Morever, suppose
Array<float, 1> b;
lhs( a, b, c) = rhs( int(1) , Array<float,1>(10) , '1')
would be same as
a = 1; b= Array<float,1>(10); c = '1';
whereas
lhs( a, ref(b), c) = rhs( int(1) , Array<float,1>(10) , '1')
would be
a = 1; b.reference(Array<float,1>(10)); c = '1';
Then, i changed list initialzers for 2-d and 1-d matrices
Array<int, 2> a(4,4);
Array<int,1> b(3);
Array<int,2> c(2,4);
b = 3;
c = 5;
a = b, 1,
c,
10,9,8,7;
would set a equal to
a = 3,3,3,1
5,5,5,5
5,5,5,5
10,9,8,7
And i added some algorithms to blitz. I have named them Vectorized Template
eXtenstions, or otherwise _VTx . As the name says, it will exploit
parallelism wherever possible. And most functions are rip off of the
fortran -90 version of Numerical Recipies.
Anyways, look over. And something really really bugs u, please email me at
mdsuresh_at_REMOVEMErediffmail.com. (Remember to remove REMOVEME). Mostly,
it's still in prerelease mode. Your opnions are very important now, since
it's the right time to shape the library.
cheers
Suresh
_______________________________________________
Blitz-dev mailing list
Blitz-dev_at_[hidden]
http://www.oonumerics.org/mailman/listinfo.cgi/blitz-dev