Re: calc: faster sums ...

2015-10-30 Thread Noel Grandin
For cross-platform bonus points, there is also libraries like: http://www.yeppp.info/ (There are various other ones, but this is the only one I found that shows signs of recent activity) Disclaimer: http://www.peralex.com/disclaimer.html ___

Re: calc: faster sums ...

2015-10-30 Thread Michael Meeks
On Thu, 2015-10-29 at 14:26 -0400, Kohei Yoshida wrote: > So, if I'm not mistaken of your intent, you can just do > > const double* p = ::numeric_block::at(*rNode.data, nOffset); > > to have direct access to the double array at specified offset, without needing > to do a weird hack like that.

calc: faster sums ...

2015-10-29 Thread Michael Meeks
Hi Kohei, I'd love some input (if you have a minute) on the attached. The punch-line is, that if we want to do really fast arithmetic, we start to need to do some odd things; while I suspect that this piece of unrolling can be done with the iterator - the next step I'm poking at (SSE3

Re: calc: faster sums ...

2015-10-29 Thread Kohei Yoshida
Hi Michael, > On October 29, 2015 at 10:21 AM Michael Meeks > wrote: > > > Hi Kohei, > > I'd love some input (if you have a minute) on the attached. The > punch-line is, that if we want to do really fast arithmetic, we start to > need to do some odd things;

Re: calc: faster sums ...

2015-10-29 Thread Tomaž Vajngerl
Hi, On Thu, Oct 29, 2015 at 3:21 PM, Michael Meeks wrote: > Hi Kohei, > > I'd love some input (if you have a minute) on the attached. The > punch-line is, that if we want to do really fast arithmetic, we start to > need to do some odd things; while I suspect