Peak mem difference in Simulator with loaded FIT playable file vs no load?

In have noticed in 4.1.6 that my data field App have 256 bytes less free memory when I load and playback a FIT file vs if I do not load a FIT file. I know its 256 bytes exactly of difference since I checked with:

System.println(System.getSystemStats().freeMemory);

This difference has not been there in the past. And there should be no dynamic difference in my data field App. So, I analyzed it using CTRL+M and got, when not loading a FIT file but still pressed the start btn:

If I play back a FIT file I get:

BUT - the stack is exactly the same:

And, I have peak memory detection enabled and shown.

huh?

Hr. Mortensen over 2 years ago

I have analyzed further:

7 months ago, I tested free memory in simulator vs real device difference, see:

https://forums.garmin.com/developer/connect-iq/f/discussion/293163/free-memory-in-simulator-vs-real-device-difference-which-to-trust

I back then concluded I could allocate 800 bytes more on my fr645 in real life vs the simulator before my app would crash with “out of memory”. If I added 900, it would crash back then on my fr645 device. I have now repeated this experiment with 4.1.7. And it now first crashes if I add more than 800+256 bytes.

So, in my conclusion, something is wrong with the free memory estimation (or at least different vs past) in latest simulator when loading FIT file for playback.

Any comments or suggestions? Otherwise, I’ll file this as a defect in simulator free memory estimation.

_psx_ over 2 years ago in reply to Hr. Mortensen

run app in sim in release mode and you will see almost good memory consumption

on condition there is no bug connected with loading fit file

unfortunately each new sdk leaves less free memory for apps

Hr. Mortensen over 2 years ago in reply to _psx_

Thanks!

Q: "run app in sim in release mode" - Is it that I should use -r --release when I build the PRG?

Doing so does indeed reduce memory usage to something that is closer to release life!

If I build for uploading to App store, what is then downside of the -r switch?

Hr. Mortensen over 2 years ago in reply to Hr. Mortensen

Btw: loading a FIT file still adds the same extra memory

_psx_ over 2 years ago in reply to Hr. Mortensen

building iq is default with -r but you have check if build for prg/device

you can see in memory viewer e.g. building with -r removing all string for object to symbols and maybe something else

_psx_ over 2 years ago in reply to Hr. Mortensen

maybe on device there is also in memory a few samples in memory - probably you don't compare the same

try to load very big fit file and you will see it has impact on memory

Hr. Mortensen over 2 years ago in reply to _psx_

Tried using both a really small FIT file and one from a long run. No difference.

I'll file this as a defect - let's see how Garmin views this one. Thanks for the help - I learned that -r can be useful in memory simulations :-)

Hr. Mortensen over 2 years ago in reply to Hr. Mortensen

Filed:

forums.garmin.com/.../peak-mem-difference-in-new-simulator-with-loading-fit-playable-file-vs-when-not-loading-a-fit-file

_psx_ over 2 years ago in reply to Hr. Mortensen

So it can be "overhead" of sim but 256 bytes can be used for e.g. cache also on device (I don't think system saves data every second due to battery).

You are walking on a very thin line. :) I think it's possible to reduce memory (code/data segment). Have you tried -O3 as you use 4.1.6?

jim_m_58 over 2 years ago in reply to Hr. Mortensen

Personally, I'd build an app based on the sim. The SDK may have a fix that hasn't been rolled out to the devices yet where the VM may need a few more bytes. The VM already takes about 4k of application memory, and 256b could be about the smallest increment they could use.

This has happened in the past, but it's rare.

This is also why I never come close to the memory limit in my apps. I always leave some free space. I have learned the hard way!