GPU optimization feedback (I hope ;-)

[Drew Jensen]

Howdy,

Thought I'd pass on a twitter exchange, to the list here, between myself
and Wikisteff regarding another attempt to benchmark some of the changes to
Calc recently.

Wikisteff: "Oh, yeah!
I wrote a 25th order Runge-Kutta integrator in Excel without using VB as
well."

Wikisteff: "Well, it *was* part of a head-to-head test to see how GPU
parallel evaluation of floating-point code worked in practice on
@LibreOffice  Calc.
@MSExcel  was the control. ;)"

Me: "I would love to see results from a GPU parallel evaluation of
floating-point code comparison between Calc and Excel. Is that available
somewhere?  thanks."

Wikisteff: "I didn't publish my results, but they amounted to "if the
calculation is simple, like a uniform linear transformation or a matrix
multiplication, high speedup. If the calculation is nonroutine, like a
nonlinear transform interwoven with large matrix products, no speedup."

Anyway, I haven't replied on twitter yet, will in a minute or two and thank
him for his time and that piece of feedback.

Personally I don't know if any of that is useful as feedback, hope so.

Best wishes,

Drew
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/libreoffice/attachments/20191209/267760c6/attachment.html>