| 12 Feb 2025 |
 motiejus | Relevant quote:
Fedora and Ubuntu have now returned frame pointers, which is great news. People should start running these releases in 2024 and will find that CPU flame graphs make more sense, Off-CPU flame graphs work for the first time, and other new things become possible. It's also a win for continuous profilers, as they don't need to convince their customers to make OS changes to get profiles to fully work.
As you can see, this debugging session would benefit from those greatly. :)
| 07:49:27 |
| motiejus | * Relevant quote:
Fedora and Ubuntu have now returned frame pointers, which is great news. People should start running these releases in 2024 and will find that CPU flame graphs make more sense, Off-CPU flame graphs work for the first time, and other new things become possible. It's also a win for continuous profilers, as they don't need to convince their customers to make OS changes to get profiles to fully work.
As you can see, this debugging session would benefit from those greatly.
| 08:02:37 |
| motiejus | Uploaded a bunch of flame graphs, probably including worker nodes (if I got the perf command right). | 08:26:07 |
| motiejus | * Uploaded a bunch of flame graphs, probably including worker threads (if I got the perf command right). | 08:26:25 |
 K900 | In reply to@motiejus:jakstys.lt
Unrelated: can I start lobbying a CFLAG -fno-omit-frame-pointer, at least on x86_64-linux? So I don't need to do such partial hacks. Some background reading by Brendan Gregg. We tend to use debuginfo for this instead | 08:57:01 |
| motiejus | Can you elaborate? | 09:07:19 |
| K900 | We have debuginfo for most hot path packages | 10:00:19 |
| K900 | And you can use that with perf to get proper unwinding without frame pointers | 10:00:33 |
| K900 | (at the cost of some post-processing time) | 10:00:41 |
| motiejus | Great! I see Python is in the list that has separate debuginfo pkgs; I'll poke around. | 15:56:01 |
| motiejus | back to my argument: although the same can be achieved with dbginfo for select packages and is hugely helpful for gdb, but in profiling use case, we don't need to bear the cost of separate dbginfo -- the size and binary is almost the same with frame pointers. :) | 15:56:58 |
| motiejus | * back to the original argument about frame pointers: although the same can be achieved with dbginfo for select packages and is hugely helpful for gdb, but in profiling use case, we don't need to bear the cost of separate dbginfo -- the size and binary is almost the same with frame pointers. :) | 16:01:15 |
| motiejus | * back to the original argument about frame pointers: yes, the same can be achieved with dbginfo for select packages and is hugely helpful for gdb. However, in profiling use case, we don't need to bear the cost of separate dbginfo -- the size and binary is almost the same with frame pointers. :)e | 16:01:33 |
| motiejus | * back to the original argument about frame pointers: yes, the same can be achieved with dbginfo for select packages and is hugely helpful for gdb. However, in profiling use case, we don't need to bear the cost of separate dbginfo -- the size and binary is almost the same with frame pointers. :) | 16:01:39 |
| motiejus | e.g. in my case, this flag is not available on OpenMP, of which I'd like to have stack frames of in the flamegraph | 16:03:22 |
|  Kira changed their display name from kira to Kira. | 19:59:42 |
 @hexa:lossy.network | 02:00.0 System peripheral: Global Unichip Corp. Coral Edge TPU
| 20:15:00 |
| @hexa:lossy.network | [ 20.102983] apex 0000:02:00.0: Apex performance not throttled due to temperature
| 20:15:50 |
| @hexa:lossy.network | oh thx | 20:15:52 |
| @hexa:lossy.network | frigate[8635]: ValueError: Failed to load delegate from libedgetpu.so.1.0
| 20:52:43 |
| @hexa:lossy.network | hrm | 20:52:46 |
| @hexa:lossy.network | device must be pci not pcie apparently | 20:59:15 |
| @hexa:lossy.network | so, it's up and not killing the n100 yet | 20:59:27 |
| @hexa:lossy.network | the gpu is 10% busy with video things | 20:59:57 |
| @hexa:lossy.network | * the gpu vpu is 10% busy with video things | 21:00:07 |
| @hexa:lossy.network | for two cameras at 720p | 21:00:18 |
| @hexa:lossy.network | ok, my config is really simple at this point | 21:22:33 |
| @hexa:lossy.network | cameras:
Foo:
detect:
enabled: true
ffmpeg:
inputs:
- input_args: preset-rtsp-restream
path: rtsp://127.0.0.1:8554/g3-flex?mp4
roles:
- record
motion:
mask:
- 0,0,223,0,224,27,0,27
Bar:
detect:
enabled: false
ffmpeg:
inputs:
- input_args: preset-rtsp-restream
path: rtsp://127.0.0.1:8554/g3?mp4
roles: []
motion:
mask:
- 0,0,223,0,224,27,0,27
database:
path: /var/lib/frigate/frigate.db
detectors:
coral:
device: pci
enabled: true
type: edgetpu
ffmpeg:
hwaccel_args: preset-vaapi
mqtt:
enabled: false
host: localhost
record:
enabled: false
retain:
days: 2
mode: all
| 21:22:47 |
| @hexa:lossy.network | video encode/decode using vaapi with intel quicksync video | 21:23:22 |
| @hexa:lossy.network | detection with a coral edgetpu pcie device | 21:23:32 |
