I have a locally stored maildir with 3.7 millions e-mails spread in ~50 folders. aerc spends ~15 seconds at startup with a spinning [...] before it shows any e-mail.
Is there something that can be done to improve the startup time?
The strace -c output looks like this:
% time seconds usecs/call calls errors syscall
------ ----------- ----------- --------- --------- ------------------
76,17 58,883267 9050 6506 421 futex
6,63 5,126167 28478 180 1 read
6,58 5,084948 2867 1773 2 epoll_pwait
6,23 4,818495 326 14776 nanosleep
3,86 2,982285 43 68400 getdents64
0,20 0,158011 38 4098 tgkill
0,12 0,092559 38 2390 madvise
0,06 0,048912 11 4098 getpid
0,06 0,043334 10 4098 10 rt_sigreturn
0,02 0,017345 6 2520 sched_yield
0,01 0,010641 9 1093 110 newfstatat
0,01 0,006921 89 77 mprotect
0,01 0,005638 16 339 mmap
0,01 0,004878 21 231 write
0,01 0,004832 29 164 rt_sigprocmask
0,00 0,002931 117 25 clone3
0,00 0,002803 2 1350 fcntl
0,00 0,001915 5 366 1 openat
0,00 0,001754 4 366 close
0,00 0,001637 31 52 sigaltstack
0,00 0,000954 36 26 set_robust_list
0,00 0,000900 34 26 gettid
0,00 0,000743 35 21 munmap
0,00 0,000690 26 26 rseq
0,00 0,000659 1 341 333 epoll_ctl
0,00 0,000446 3 114 rt_sigaction
0,00 0,000147 4 31 fstat
0,00 0,000093 46 2 socket
0,00 0,000079 26 3 2 unlinkat
0,00 0,000066 66 1 bind
0,00 0,000055 18 3 inotify_add_watch
0,00 0,000047 47 1 1 connect
0,00 0,000046 46 1 getrlimit
0,00 0,000037 37 1 1 accept4
0,00 0,000022 3 7 ioctl
0,00 0,000021 10 2 statfs
0,00 0,000018 18 1 pipe2
0,00 0,000016 16 1 setrlimit
0,00 0,000014 4 3 brk
0,00 0,000009 9 1 setsockopt
0,00 0,000008 4 2 sched_getaffinity
0,00 0,000008 8 1 epoll_create1
0,00 0,000007 3 2 prlimit64
0,00 0,000006 3 2 1 access
0,00 0,000005 5 1 arch_prctl
0,00 0,000005 5 1 set_tid_address
0,00 0,000004 4 1 getrandom
0,00 0,000003 3 1 listen
0,00 0,000002 2 1 getsockname
0,00 0,000000 0 2 pread64
0,00 0,000000 0 1 execve
0,00 0,000000 0 1 inotify_init1
0,00 0,000000 0 1 faccessat2
------ ----------- ----------- --------- --------- ------------------
100,00 77,304383 680 113531 883 total
BTW, thanks for your great talk at FOSDEM 2025, I was in the room!
Hi there,
Could you capture the following and paste the results here?
perf record --call-graph=dwarf aerc # :quit aerc as soon as you can see emails listed perf report -g --stdio | head -n200
Sorry for the delay, here is the output of "perf report" as requested. I have no idea if this output is useful/relevant, let me know.
# To display the perf.data header info, please use --header/--header-only options. # # # Total Lost Samples: 0 # # Samples: 3K of event 'cpu_atom/cycles/Pu' # Event count (approx.): 1797132652 # # Children Self Command Shared Object Symbol # ........ ........ ....... .................... ............................................................................................................................................................................. # 75.62% 0.00% aerc [unknown] [.] 0x01ffffffffffff7f | ---0x1ffffffffffff7f | |--74.65%--runtime.systemstack.abi0 | | | |--71.80%--runtime.gcBgMarkWorker.func2 | | | | | |--37.68%--runtime.gcDrainMarkWorkerIdle (inlined) | | | runtime.gcDrain | | | | | | | --37.53%--runtime.scanobject | | | | | | | |--19.22%--runtime.greyobject | | | | | | | | | |--10.02%--runtime.markBits.setMarked (inlined) | | | | | | | | | |--5.98%--runtime.(*mspan).markBitsForIndex (inlined) | | | | | runtime.(*gcBits).bitp (inlined) | | | | | | | | | | | --5.37%--runtime.(*gcBits).bytep (inlined) | | | | | | | | | --0.63%--runtime.pageIndexOf (inlined) | | | | | | | |--9.95%--runtime.findObject | | | | | | | | | |--4.40%--runtime.(*mSpanStateBox).get (inlined) | | | | | | | | | | | --4.26%--runtime/internal/atomic.(*Uint8).Load (inlined) | | | | | | | | | |--2.25%--runtime.spanOf (inlined) | | | | | | | | | --0.65%--runtime.(*mspan).base (inlined) | | | | | | | |--1.04%--runtime.typePointers.nextFast (inlined) | | | | | | | --0.72%--runtime.typePointers.next | | | | | --34.12%--runtime.gcDrainMarkWorkerDedicated (inlined) | | runtime.gcDrain | | | | | |--33.12%--runtime.scanobject | | | | | | | |--16.59%--runtime.greyobject | | | | | | | | | |--8.65%--runtime.markBits.setMarked (inlined) | | | | | | | | | |--4.58%--runtime.(*mspan).markBitsForIndex (inlined) | | | | | | | | | | | --4.55%--runtime.(*gcBits).bitp (inlined) | | | | | | | | | | | --4.12%--runtime.(*gcBits).bytep (inlined) | | | | | | | | | --0.67%--runtime.pageIndexOf (inlined) | | | | | | | |--7.05%--runtime.findObject | | | | | | | | | |--3.02%--runtime.(*mSpanStateBox).get (inlined) | | | | | | | | | | | --2.98%--runtime/internal/atomic.(*Uint8).Load (inlined) | | | | | | | | | --1.66%--runtime.spanOf (inlined) | | | | | | | |--1.33%--runtime.typePointers.next | | | | | | | --1.02%--runtime.typePointers.nextFast (inlined) | | | | | --0.68%--runtime.markroot | | | |--1.26%--runtime.(*sweepLocked).sweep.(*mheap).freeSpan.func3 | | | | | --1.07%--runtime.(*mheap).freeSpanLocked | | | | | --0.66%--runtime.(*pageAlloc).free | | | --0.81%--runtime.gcAssistAlloc.func1 | runtime.gcAssistAlloc1 | runtime.gcDrainN | | | --0.80%--runtime.scanobject | --0.93%--runtime.mcall runtime.park_m runtime.schedule | --0.89%--runtime.findRunnable 74.65% 0.03% aerc aerc [.] runtime.systemstack.abi0 | --74.63%--runtime.systemstack.abi0 | |--71.77%--runtime.gcBgMarkWorker.func2 | | | |--37.68%--runtime.gcDrainMarkWorkerIdle (inlined) | | runtime.gcDrain | | | | | --37.53%--runtime.scanobject | | | | | |--19.22%--runtime.greyobject | | | | | | | |--10.02%--runtime.markBits.setMarked (inlined) | | | | | | | |--5.98%--runtime.(*mspan).markBitsForIndex (inlined) | | | | runtime.(*gcBits).bitp (inlined) | | | | | | | | | --5.37%--runtime.(*gcBits).bytep (inlined) | | | | | | | --0.63%--runtime.pageIndexOf (inlined) | | | | | |--9.95%--runtime.findObject | | | | | | | |--4.40%--runtime.(*mSpanStateBox).get (inlined) | | | | | | | | | --4.26%--runtime/internal/atomic.(*Uint8).Load (inlined) | | | | | | | |--2.25%--runtime.spanOf (inlined) | | | | | | | --0.65%--runtime.(*mspan).base (inlined) | | | | | |--1.04%--runtime.typePointers.nextFast (inlined) | | | | | --0.72%--runtime.typePointers.next | | | --34.09%--runtime.gcDrainMarkWorkerDedicated (inlined) | runtime.gcDrain | | | |--33.12%--runtime.scanobject | | | | | |--16.59%--runtime.greyobject | | | | | | | |--8.65%--runtime.markBits.setMarked (inlined) | | | | | | | |--4.58%--runtime.(*mspan).markBitsForIndex (inlined) | | | | | | | | | --4.55%--runtime.(*gcBits).bitp (inlined) | | | | | | | | | --4.12%--runtime.(*gcBits).bytep (inlined) | | | | | | | --0.67%--runtime.pageIndexOf (inlined) | | | | | |--7.05%--runtime.findObject | | | | | | | |--3.02%--runtime.(*mSpanStateBox).get (inlined) | | | | | | | | | --2.98%--runtime/internal/atomic.(*Uint8).Load (inlined) | | | | | | | --1.66%--runtime.spanOf (inlined) | | | | | |--1.33%--runtime.typePointers.next | | | | | --1.02%--runtime.typePointers.nextFast (inlined) | | | --0.68%--runtime.markroot | |--1.26%--runtime.(*sweepLocked).sweep.(*mheap).freeSpan.func3 | | | --1.07%--runtime.(*mheap).freeSpanLocked | | | --0.66%--runtime.(*pageAlloc).free | --0.81%--runtime.gcAssistAlloc.func1 runtime.gcAssistAlloc1 runtime.gcDrainN | --0.80%--runtime.scanobject 72.10% 16.13% aerc aerc [.] runtime.scanobject | |--56.30%--runtime.scanobject | | | |--36.09%--runtime.greyobject | | | | | |--18.85%--runtime.markBits.setMarked (inlined) | | | | | |--10.65%--runtime.(*mspan).markBitsForIndex (inlined) | | | | | | | --10.62%--runtime.(*gcBits).bitp (inlined) | | | | | | | --9.58%--runtime.(*gcBits).bytep (inlined) | | | | | |--1.33%--runtime.pageIndexOf (inlined) | | | | | |--0.77%--runtime.markBits.isMarked (inlined) | | | | | --0.57%--runtime.(*mspan).base (inlined) | | | |--17.18%--runtime.findObject | | | | | |--7.42%--runtime.(*mSpanStateBox).get (inlined)
Hmm, I thought 200 lines would be enough to get meaningful information...
If you still have the perf.data file, could you run these two commands without truncating the end:
perf report -g --stdioperf report --stdio
Output of first command: https://paste.ack.tf/847681 Output of second command: https://paste.ack.tf/6c88b8
Note that these captures have been done with the latest master of aerc, while the previous capture was done with whatever aerc version was available in Fedora 40 (which most likely wasn't the latest)
The pastes have expired unfortunately. If this was using IMAP, I think that https://lists.sr.ht/~rjarry/aerc-devel/patches/57911 could help.
It wasn't using IMAP, but a local store, aka maildir://.
ACK, thanks, my bad. Regardless, it'd be great if you could refresh the pastes. Thanks!
Simon Martin referenced this ticket in commit 455cd52.