Enable link-time optimization for nrf targets #2031
Conversation
Testing performed: installed freshly built .uf2 on a Particle Xenon. Checked that circuitpython still starts. Checked that the size of all .uf2 files for nrf builds are plausible. Aside from memory savings, the performance of Python code (pystone) increased by about +14%. However, this adds about 12-16 seconds to each nrf build. Timings & Sizes (build system: i5-3320M, -j5 parallelism on 4 threads): Before: $ make -j5 BOARD=particle_xenon 765004 bytes free in flash out of 1048576 bytes ( 1024.0 kb ). 232076 bytes free in ram for stack out of 245760 bytes ( 240.0 kb ). 68.54user 11.83system 0:34.34elapsed 234%CPU pystones before: 570 After: $ make -j5 BOARD=particle_xenon 804284 bytes free in flash out of 1048576 bytes ( 1024.0 kb ). 232072 bytes free in ram for stack out of 245760 bytes ( 240.0 kb ). 71.06user 11.77system 0:46.91elapsed 176%CPU pystones after: 650 Timings on travis: Before: Build feather_nrf52840_express for pl took 55.79s and succeeded Build feather_nrf52840_express for zh_Latn_pinyin took 3.18s and succeeded After: Build feather_nrf52840_express for pl took 62.72s and succeeded Build feather_nrf52840_express for zh_Latn_pinyin took 19.10s Closes: adafruit#1396
|
Travis overall times changed from "4 hrs 11 min 5 sec" (https://travis-ci.com/adafruit/circuitpython/builds/121819759) to "4 hrs 38 min 26 sec" (https://travis-ci.com/adafruit/circuitpython/builds/121828350). In addition, because job # 1 is now badly imbalanced, the total wait time increased from 53 min 51 sec to 1 hr 21 min 13 sec. |
|
I'll leave it up to @dhalbert |
There was a problem hiding this comment.
I tested this and it works with the RGB BLE example (server and client: two boards). I am seeing about 42KB of extra space.
It's possible the -flto-partition=none is the key, but I don't know. If it stops working again we can always turn it off again.
Thanks for being persistent about experimenting about this.
|
I will look at the build times and re-balance the Travis jobs after this finishes, and then submit a PR for that. @tannewt is working on GitHub actions to use for CI instead of Travis, but until that is done we could increase the number of jobs to 6 or 7 for now. |
The good: It gets back about 60kB of flash space and increases performance (pystone) by about +14%
The bad: It adds about 12-16 seconds to each target built
The ugly: This is exactly the same thing tried for #1396 by others, and it's not clear why it seems to be working for me today.
See the commit message for further info.