topic Which version of IPP are you in Intel® Integrated Performance Primitives

Inplace Normalize?

Ockham_s_Razor — Wed, 19 Sep 2012 18:54:29 GMT

The documentation for Add, Multiply states that to use inplace just use the same pointers.

For ippsNormalize_xxx there is not that comment.

Can I call the ippsNormalize_xxx functions with the same pointers.

Thank you.

-jam

PS: I could not find any way to search this forum.

Which version of IPP are you

Chuck_De_Sylva — Thu, 20 Sep 2012 03:21:41 GMT

Which version of IPP are you using?

I'm using 7.0 right now,

Ockham_s_Razor — Thu, 20 Sep 2012 17:54:24 GMT

I'm using 7.0 right now, moving to 7.1 shortly

only out of place mode is

Gennady_F_Intel — Sun, 30 Sep 2012 05:06:48 GMT

only out of place mode is valid in that function.

I did a little experiment. .

John_Mertus — Tue, 02 Oct 2012 18:55:52 GMT

I did a little experiment. . I found that on a i7 2.8 GHz processor (Dell Precsion 1500) the normalize took about 54.1 milliseconds for an input of 50 million floats. I used Add in place followed by amultiple in place and found it averaged 54.2 milliseconds for 50 million floats. So there was little difference. That said, it would be nice, and apparently easly, to have an inplace normalize.