Hi Andrei,
I have summarized the results with LDC on Mac OSX in a previous
posts:
orignal: 760ms
* change v = [x,x,x] to v[0]=x... in this(float) constructor -
540ms (220ms improvment)
* remove this(float[]) constructor and replace it with
this(float,float,float) - 410ms (130ms improvment)
* "final" to all class methods - 350 (60ms improvment)
* hand unroll overloaded vector arithmetic operators, 260ms (90ms
improvment)
However I retested on a windows 7 machine with GDC compiler and
the results were very different.
orignal: 545ms
* the first 2 optimizations which helped the most on OSX with LDC
has almost zero effect
* hand unroll overloaded vector arithmetic operators - 280ms
(265ms improvment)
* "final" to all class methods - 200ms (80ms improvment)
On Saturday, 1 June 2013 at 13:36:44 UTC, Andrei Alexandrescu
wrote:
Would be great to collect a summary of the changes and their
benefits.
Thanks finalpatch and everyone else for this work.
Andrei
