Hi, I've been working on further improving the perf profile view in LNT that James added recently. I've got a series of patches ready that reconstruct the control flow graph from the perf profiles, which helps a lot in quickly finding the codegen differences that cause performance deltas. I've used the D3 javascript library (https://d3js.org/) to implement the visualization. My question is if there is any objection to add D3 to LNT? I think D3 will allow us to produce nicer and/or more informative charts and visualizations in the future. I've attached an example screenshot of the control flow graph visualization that I plan to upstream if there is no objection to add D3 to LNT. Thanks, Kristof [cid:4BFAFE45-D153-4064-9E07-8E01CC5D13C0] -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160421/714c1481/attachment-0001.html> -------------- next part -------------- A non-text attachment was scrubbed... Name: Screen Shot 2016-04-21 at 15.08.26.png Type: image/png Size: 326528 bytes Desc: Screen Shot 2016-04-21 at 15.08.26.png URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160421/714c1481/attachment-0001.png>
I love D3. This is a great idea. Maybe we can finally kill flot! We are going to have to start using the html data attributes more; but that is a good thing IMO. D3 would greatly simply some of our CSS/templates too. I am regularly surprised how hard it is to make a cell turn colors in our current jinja macro setup. Is it okay to have D3 checked into the LNT repo, or should we use a CDN link?> On Apr 21, 2016, at 6:23 AM, Kristof Beyls via llvm-dev <llvm-dev at lists.llvm.org> wrote: > > Hi, > > I've been working on further improving the perf profile view in LNT that James added recently. > I've got a series of patches ready that reconstruct the control flow graph from the perf profiles, which helps a lot in quickly finding the codegen differences that cause performance deltas. > I've used the D3 javascript library (https://d3js.org/ <https://d3js.org/>) to implement the visualization. > > My question is if there is any objection to add D3 to LNT? > > I think D3 will allow us to produce nicer and/or more informative charts and visualizations in the future. > I've attached an example screenshot of the control flow graph visualization that I plan to upstream if there is no objection to add D3 to LNT. > > Thanks, > > Kristof > > <Screen Shot 2016-04-21 at 15.08.26.png> > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160421/e0d96d3f/attachment.html>
> On Apr 21, 2016, at 6:23 AM, Kristof Beyls via llvm-dev <llvm-dev at lists.llvm.org> wrote: > > Hi, > > I've been working on further improving the perf profile view in LNT that James added recently. > I've got a series of patches ready that reconstruct the control flow graph from the perf profiles, which helps a lot in quickly finding the codegen differences that cause performance deltas. > I've used the D3 javascript library (https://d3js.org/ <https://d3js.org/>) to implement the visualization. > > My question is if there is any objection to add D3 to LNT? > > I think D3 will allow us to produce nicer and/or more informative charts and visualizations in the future. > I've attached an example screenshot of the control flow graph visualization that I plan to upstream if there is no objection to add D3 to LNT.Beautiful! I love it :) Thanks Kristof. -- Mehdi> > Thanks, > > Kristof > > <Screen Shot 2016-04-21 at 15.08.26.png> > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160421/cdcb31a4/attachment.html>
> On 21 Apr 2016, at 17:15, Chris Matthews <chris.matthews at apple.com> wrote: > > I love D3. This is a great idea. Maybe we can finally kill flot! > > We are going to have to start using the html data attributes more; but that is a good thing IMO. > > D3 would greatly simply some of our CSS/templates too. I am regularly surprised how hard it is to make a cell turn colors in our current jinja macro setup. > > Is it okay to have D3 checked into the LNT repo, or should we use a CDN link?Not using a CDN link would mean you can work offline with an LNT server installed on your laptop, for example. Otherwise, I'm as happy with using a CDN link instead of importing it into the LNT repo. I can't think of another reason not to use a CDN link. I don't really have a strong preference either way. Do you have a preference? Thanks, Kristof
Possibly Parallel Threads
- [RFC] Using D3 in LNT?
- Proper attribution in Authors@R for the d3.js library by Mike Bostock
- convert R plots into annotated web-graphics
- [LLVMdev] [LNT] Does LNT's mouse wheel zooming work well anywhere?
- Extending TableGen's 'foreach' to work with 'multiclass' and 'defm'