I recently ran Linux perf (more precisely perf record -F 99 -g -- my_app) on a large application - peak memory ~350Gbytes, about 3 days runtime.
My problem now is that the resulting perf.data (41Gbytes) seems almost impossible to read. I launched perf report over a day ago and the yellow progress bar is only on about 60%, and perf is using 0.467t VIRT and 0.411t RES according to top. The machine [RHEL 7.4 amd64] has 512Gbytes RAM and 256Gbytes swap. Fingers crossed that it finishes before being OOMed. I also tried perf script, but I had to give up when the output filled the disk mount (~550Gbytes). I'll try to find somewhere with more space, but considering that I was hoping to produce a flame graph, I'm not very confident that perl will be able to cope with the next 2 stages.
Obviously I can try to modify the parameters of my application to try to shrink the time and memory.
Any suggestions for options that would make perf report run more quickly and use less memory?