The printed total size differs from calculated from .json

[wangnuannuan]

Describe the bug
I want to get the rom and ram size by reading the JSON file, so I use the script https:/zephyrproject-rtos/zephyr/blob/master/scripts/footprint/size_report to parse the elf file.

Take rom size as example, the script will print the total size in the terminal and generate a file rom.json. But when I use the following code to calculate the rom size in the JSON file, the printed total size differs from calculated from JSON file.

from anytree.importer import JsonImporter

importer = JsonImporter()
size = 0
with open("./rom.json", "r") as fp:
    root = importer.read(fp)

    for leave in root.leaves:
        size += leave.size
print(size)

@dcpleung @nashif

[tejlmand]

One reason for this is because the size report itself will take the rom_start, datas and devices sections into consideration, which means the final size reported:

...
├── mpu_config                                                                                      8 0.05%
└── transitions.10586                                                                              12 0.07%
==============================================================================================================
                                                                                                16974

will be larger than the sum of the sizes of the individual symbols.

When you sum the json, it is the sum of the size of each symbol, but the rom_start, datas and devices sections are not having symbols in flash. Furthermore the datas and devices are loaded into ram at startup, so any symbols would be referring to ram addresses.
Everything that takes up space in flash are included in the total sum, but cannot be presented as a symbol in the overview and the json file.

If you compare to objdump, you should find some of the missing size:

$ objdump -h zephyr/zephyr.elf 
Sections:
Idx Name          Size      VMA       LMA       File off  Algn
  0 rom_start     00000100  00000000  00000000  000000c0  2**2
                  CONTENTS, ALLOC, LOAD, CODE
...
  8 datas         00000040  20000000  00004190  00004250  2**2
                  CONTENTS, ALLOC, LOAD, DATA
  9 devices       00000090  20000040  000041d0  00004290  2**2
                  CONTENTS, ALLOC, LOAD, READONLY, DATA

Note, the section list above is not necessarily complete.

[dcpleung]

The sizes of individual items are of the items themselves, while the total size takes into account of alignments and padding. That's why the total size is larger.

And one important thing to remember, not all symbols are itemized. There are symbols in the image (for example, strings) that do not appear in the symbol table, and thus not in the list but counted towards total size.

[galak]

@wangnuannuan any updates on this

[wangnuannuan]

#33112