Howdy

<snip>
>
>         Looks great; we use only real-cores to calculate on[1]; so I'd
> mention
> 2 cores - rather than 4 threads. As such we got ~2x as fast with 2 cores
> ;-)
>
>         Do you have a 4 core machine nearby - or better a 24 one ?
>

Thought I did. My test machine for this is described as:
Processor: AMD A10-5800k apu with radeon hd graphics x4
Memory: 7 GiB
Graphics: AMD Aruba
OS; Ubuntu 18.04.1 LTS
OS type: 64 bit
Gnome: 3.28.2
Disk: 308 GB

The files used came from a source patch referring to benchmarking OpenCL
functionality and includes 5 spreadsheets (xls all). Included is a python
script to run them as a batch and to write out recalc times.
So, if I understand correctly OpenCL doesn't mean much on this Linux
desktop, but I will be able to try, a little later, on a Windows 10 machine
also so I took that this would however still show the multi-corps threading
support.

I went back and reran the batch with as little else (easily possible)
running on the test machine. Performance increase varied based on the file,
ranging from ~76% - 30%
Stock History
6.0.7 : 13.9 sec
6.2.0 : 8.71 sec
                        62.68%
37.32% performance increase
Stocks Price Time Correlation
6.0.7 : 21.38 sec
6.2.0 : 9.57 sec
                        44.79%
55.31% performance increase

Energy Market
6.0.7 : 28.44 sec
6.2.0 : 12.43 sec
                        43.74%
58.36% performance increase

Ground Water Daily
6.0.7 :  92.52 sec
6.2.0 : 32.38 sec
                        35.00%
65.00% performance increase
Gross Profit Supermarkets
6.0.7 : 66.95 sec
6.2.0 : 20.15 sec
                        30.09%
69.91% performance increase

Building Design
6.0.7 : 50.14 sec
6.2.0 : 12.06 sec
                        24.05%
75.95% performance increase

The files each contain a write up as to what the data represents and what
statistical functions are being exercised.

More (on other follow up email) on these as this we go.

Best wishes,

Drew


>
>         Lovely video though; I love the raw motorbike-with-hole-in-exhaust
> ricer feel =)
>
>         Good stuff,
>
>                 Michael.
>
> [1] - hyper-threading is not our friend in the case of some heavy usage
> and can make it slower ;-)
> --
> michael.me...@collabora.com <><, GM Collabora Productivity
> Hangout: mejme...@gmail.com, Skype: mmeeks
> (M) +44 7795 666 147 - timezone usually UK / Europe
>

-- 
To unsubscribe e-mail to: marketing+unsubscr...@global.libreoffice.org
Problems? https://www.libreoffice.org/get-help/mailing-lists/how-to-unsubscribe/
Posting guidelines + more: https://wiki.documentfoundation.org/Netiquette
List archive: https://listarchives.libreoffice.org/global/marketing/
Privacy Policy: https://www.documentfoundation.org/privacy

Reply via email to