[ANN] macstl 0.2.2 -- multiply high, class redesign, faster than autovectorization

[g***@pixelglow.com]

Hi All,

I'm pleased to announce the release of macstl 0.2.2, which features a
multipy-high implementation and a redesign of the expression template
and iterators to make it easier to extend.

http://www.pixelglow.com/macstl/

Here's a list of changes:

- Fixed error when linking more than 1 object file, template function
should be inline [KHe].
- Added mulhi function for vec and valarray [MSh].
- Improved valarray expression template and iterator design.

macstl features a portable SIMD (single instruction multiple data)
toolkit featuring fast transcendental and integer division functions,
complex number arithmetic and cross-platform programming, all in an
easy-to-use syntax. It produces code that is 3x to 16x faster than
hand-coded scalar loops, and is actually faster than the
autovectorization technique.

http://www.pixelglow.com/stories/macstl-intel-autovectorization/

If you're Down Under in the month of April, get to see macstl
demonstrated live and in your face. I'll be touring with the Apple
Australia/New Zealand WWDC 2005 Preview roadshow, where you also get to
see lots of super secret Tiger tech.

http://www.apple.com.au/seminars/wwdc05/


Cheers, Glen Low


---
pixelglow software | simply brilliant stuff
www.pixelglow.com

[Stephen Chu]

Post by g***@pixelglow.com
Hi All,
I'm pleased to announce the release of macstl 0.2.2, which features a
multipy-high implementation and a redesign of the expression template
and iterators to make it easier to extend.
http://www.pixelglow.com/macstl/
- Fixed error when linking more than 1 object file, template function
should be inline [KHe].
- Added mulhi function for vec and valarray [MSh].
- Improved valarray expression template and iterator design.
macstl features a portable SIMD (single instruction multiple data)
toolkit featuring fast transcendental and integer division functions,
complex number arithmetic and cross-platform programming, all in an
easy-to-use syntax. It produces code that is 3x to 16x faster than
hand-coded scalar loops, and is actually faster than the
autovectorization technique.
http://www.pixelglow.com/stories/macstl-intel-autovectorization/
If you're Down Under in the month of April, get to see macstl
demonstrated live and in your face. I'll be touring with the Apple
Australia/New Zealand WWDC 2005 Preview roadshow, where you also get to
see lots of super secret Tiger tech.
http://www.apple.com.au/seminars/wwdc05/
Cheers, Glen Low

Looks like good stuffs. One question though. Can it be used with
CodeWarrior CFM projects?

Thanks.

Stephen

[g***@pixelglow.com]

Post by Stephen Chu
Looks like good stuffs. One question though. Can it be used with
CodeWarrior CFM projects?

Stephen:

macstl works a treat with CodeWarrior Mach-O on the Mac, other
platforms and combinations might require a bit more work. CFM should be
alright since I try not to use OS X libraries directly if I can help
it, only the memory mapping stuff might be specific to OS X. If you're
interested in making it work for Codewarrior CFM, email me directly at
macstl at pixelglow dot com, or you can discuss it at our dev list at

http://www.pixelglow.com/lists/listinfo/macstl-dev

Some developers there are currently working on a Yellow Dog Linux port
for PowerPC for example.

Cheers,
Glen Low, Pixelglow Software
www.pixelglow.com