Testing required: visited link performance
I've just enabled NetSurf's visited link handling. This allows links that have been visited to be rendered differently to unvisited links. It was originally implemented several years ago, but disabled since it had a detrimental effect on performance. Since then we've made many improvements to the browser, so performance may be more acceptable. However, most of the developers use fast hardware now, so performance issues are harder to spot. Please could people test builds #1177 and #1178 on a variety of pages and us know how page load times vary? I am particularly interested to hear from people using old hardware such as RiscPCs, Iyonixes, Ataris, etc. We also need to know the size of your URL file. On RISC OS, if you shift double click on the !NetSurf application directory and run OpenChoices, your URL file should be in the directory that opens. So a helpful report could take the form: System: Iyonix URL file size: 246K Site: http://www.bbc.co.uk/news/ Build #1177: 8.6s Build #1178: 9.0s Site: http://slashdot.org/ Build #1177: 5.4s Build #1178: 5.7s You can test whatever pages you like. Pages with more links on them require more searching of the browsing history. Also, the more you use NetSurf, the bigger the history to search (and the bigger the URL file). Note that the absolute page load times can vary anyway, due to network issues or other activity the computer is doing, so average timings over several runs are best. Cheers, -- Michael Drake (tlsa) http://www.netsurf-browser.org/
Re: Testing required: visited link performance
On Fri, 17 May 2013 13:06:07 +0100, Michael Drake wrote: It was originally implemented several years ago, but disabled since it had a detrimental effect on performance. IIRC it took minutes to display even the Google homepage when this was initially enabled. I'm happy to report it is not having any perceivable impact on performance at all now. System: SAM440EP (600MHz PowerPC) URL file size: 50K Site: http://www.google.co.uk Build #1177: 0.6s Build #1178: 0.6s Site: http://news.google.co.uk Build #1177: 6.6s Build #1178: 6.8s Site: http://www.digitalspy.co.uk Build #1177: 5.1s Build #1178: 5.2s Site: http://www.netsurf-browser.org Build #1177: 2.2s Build #1178: 2.1s The visited links don't display differently on Google because of the annoying redirect links they use these days, which change on every reload. It is visibly working on the NetSurf homepage however - except for the Wakefield show link, where I think the initial URL isn't being remembered by NetSurf, as the link redirects immediately to http://www.wakefieldshow.org.uk/index.php Chris
Re: Testing required: visited link performance
In article out-51966c4e.md-1.4.17.chris.yo...@unsatisfactorysoftware.co.uk, Chris Young chris.yo...@unsatisfactorysoftware.co.uk wrote: except for the Wakefield show link, where I think the initial URL isn't being remembered by NetSurf, as the link redirects Yep, it doesn't know about redirects. Supporting them probably requires quite a big change. Thanks for the feedback. -- Michael Drake (tlsa) http://www.netsurf-browser.org/
Re: Testing required: visited link performance
On Friday 17 May 2013 13:06:07 Michael Drake wrote: Please could people test builds #1177 and #1178 on a variety of pages and us know how page load times vary? I am particularly interested to hear from people using old hardware such as RiscPCs, Iyonixes, Ataris, etc. I quickly tested the Atari m5475 builds and didn't notice any difference in performance between the two. So I didn't test any further :-) Good work! Cheers, JFL -- Jean-François Lemaire
Re: Testing required: visited link performance
Hope this helps. System: Iyonix URL file size: 66k Site: http://www.bbc.co.uk/news/ Build #1177: 13.0s Build #1178: 13.7s Build #1179: 22.6s Site: http://slashdot.org/ Build #1177: 8.5s Build #1178: 8.4s Build #1179: 25.0s
Re: Testing required: visited link performance
In article 534d633ccb...@timil.com, Tim Hill t...@timil.com wrote: Site: http://www.bbc.co.uk/news/ Build #1177: 13.0s Build #1178: 13.7s Build #1179: 22.6s Site: http://slashdot.org/ Build #1177: 8.5s Build #1178: 8.4s Build #1179: 25.0s The #1179 results are odd. Was that with JavaScript on, or with the computer busy doing something else? -- Michael Drake (tlsa) http://www.netsurf-browser.org/
Re: Testing required: visited link performance
In article 534d692c69t...@netsurf-browser.org, Michael Drake t...@netsurf-browser.org wrote: In article 534d633ccb...@timil.com, Tim Hill t...@timil.com wrote: Site: http://www.bbc.co.uk/news/ Build #1177: 13.0s Build #1178: 13.7s Build #1179: 22.6s Site: http://slashdot.org/ Build #1177: 8.5s Build #1178: 8.4s Build #1179: 25.0s The #1179 results are odd. Was that with JavaScript on, or with the computer busy doing something else? On. On the one hand I didn't think anything else was going at the time but on the other I can't repeat those times. Now 10s and 13.2s. Timings are representative of several attempts in all cases.
