News for hackers. Stuff that matters.
Re:Memory Leaks Detection Paper Co-Authored by Netflix CEO R... by fnordpiglet on Sunday December 04, 2022 @08:05PM (#33860042)
Pure products were amazing. I always found the way Quantify represented profiling to be incredibly helpful back in the ancient days. I know there’s more advanced visualizations but I always found the logarithmically scaled directed edges in quantify to speak to me better than flame charts and other related techniques, I think because it represented the full execution graph rather than individual transactional paths. Anyway huge respect for Reed as an engineer and have always been impressed with both Netflix and the quality of the performance, debugging, and other efforts that have come out from them. He clearly values excellent engineering.
Re:Memory Leaks Detection Paper Co-Authored by Netflix CEO R... by iot_devs on Monday December 05, 2022 @04:35AM (#33863372)
Do you have a picture of these "logarithmically scaled directed edges"?

I can imagine what you mean but a visual aid would be super useful.

Re:Memory Leaks Detection Paper Co-Authored by Netflix CEO R... by fnordpiglet on Monday December 05, 2022 @12:02PM (#33868228)
I kagi’ed around and I can’t, sorry. The software is 30 some odd years old and was acquired by IBM and sublimated as all ibm acquisitions do. But my memory of it could well be approximated by using dot to create nodes for execution points with edges to calls that are rendered with a weight proportional to the log of the execution time. The nice thing is it allowed for recursive calls to be rendered as cycles, and there are a lot of non recursive cycles I found. It also did a reasonably good job of clustering so highly popular calls fanned into a sub graph but with each graph calling in clustered independently.
Re:Memory Leaks Detection Paper Co-Authored by Netflix CEO R... by iot_devs on Monday December 05, 2022 @06:13PM (#33873304)
Thanks a lot!

I appreciated the description anyway!