Elimination of unused sections
Elimination of unused sections is the most significant optimization on image size that is performed by the linker.
Unused section elimination:
-
Removes unreachable code and data from the final image.
-
Is suppressed in cases that might result in the removal of all sections.
To control this optimization use the --remove
,
--no_remove
, --first
, --last
, and
--keep
linker options.
Unused section elimination requires an entry point. Therefore, if there is no entry point
specified for an image, use the --entry
linker option to specify an entry
point and permit unused section elimination to work, if it is enabled.
Note
By default, unused section elimination is disabled if you are building DLLs with
--dll
, or shared libraries with --shared
. Therefore,
you must explicitly include --remove
to re-enable unused section
elimination.
Use the --info unused
linker option to instruct the linker to generate a
list of the unused sections that it eliminates.
An input section is retained in the final image when:
It contains an entry point.
It is referred to, directly or indirectly, by a non-weak reference from an input section containing an entry point.
It is specified as the first or last input section by the
--first
or--last
option (or a scatter-loading equivalent).It is marked as unremovable by the
--keep
option.
Note
Compilers usually collect functions and data together and emit one section for each category. The linker can only eliminate a section if it is entirely unused.
You can mark a function or variable in source code with the __attribute__((used))
attribute. This causes armcc
to generate the symbol __tagsym$$used
for each of the functions and variables. A section containing a
definition of __tagsym$$used
is not removed by unused section
elimination.
You can also use the --split_sections
compiler command-line option to
instruct the compiler to generate one ELF section for each function in the source
file.