Optimization Tools¶
Footprint and Memory Usage¶
The build system offers 3 targets to view and analyse RAM, ROM and stack usage in generated images. The tools run on the final image and give information about size of symbols and code being used in both RAM and ROM. Additionally, with features available through the compiler, we can also generate worst-case stack usage analysis:
Tools that are available as build system targets:
Build Target: puncover¶
This target uses a 3rd party tools called puncover which can be found here. When this target is built, it will launch a local web server which will allow you to open a web client and browse the files and view their ROM, RAM and stack usage. Before you can use this target, you will have to install the puncover python module:
pip3 install git+https://github.com/HBehrens/puncover --user
Then:
Using west:
west build -b reel_board samples/hello_world
west build -t puncover
Using CMake and ninja:
# Use cmake to configure a Ninja-based buildsystem:
cmake -B build -GNinja -DBOARD=reel_board samples/hello_world
# Now run ninja on the generated build system:
ninja -C build puncover
To view worst-case stack usage analysis, build this with the
CONFIG_STACK_USAGE enabled.
Using west:
west build -b reel_board samples/hello_world -- -DCONFIG_STACK_USAGE=y
west build -t puncover
Using CMake and ninja:
# Use cmake to configure a Ninja-based buildsystem:
cmake -B build -GNinja -DBOARD=reel_board -DCONFIG_STACK_USAGE=y samples/hello_world
# Now run ninja on the generated build system:
ninja -C build puncover
Build Target: ram_report¶
List all compiled objects and their RAM usage in a tabular form with bytes per symbol and the percentage it uses. The data is grouped based on the file system location of the object in the tree and the file containing the symbol.
Use the ram_report target with your board:
Using west:
west build -b reel_board samples/hello_world
west build -t ram_report
Using CMake and ninja:
# Use cmake to configure a Ninja-based buildsystem:
cmake -B build -GNinja -DBOARD=reel_board samples/hello_world
# Now run ninja on the generated build system:
ninja -C build ram_report
which will generate something similar to the output below:
Path Size %
==============================================================================================================
...
...
SystemCoreClock 4 0.08%
_kernel 48 0.99%
_sw_isr_table 384 7.94%
cli.10544 16 0.33%
gpio_initialized.9765 1 0.02%
on.10543 4 0.08%
poll_out_lock.9764 4 0.08%
z_idle_threads 128 2.65%
z_interrupt_stacks 2048 42.36%
z_main_thread 128 2.65%
arch 1 0.02%
arm 1 0.02%
core 1 0.02%
aarch32 1 0.02%
cortex_m 1 0.02%
mpu 1 0.02%
arm_mpu.c 1 0.02%
static_regions_num 1 0.02%
drivers 536 11.09%
clock_control 100 2.07%
nrf_power_clock.c 100 2.07%
__device_clock_nrf 16 0.33%
data 80 1.65%
hfclk_users 4 0.08%
...
...
Build Target: rom_report¶
List all compiled objects and their ROM usage in a tabular form with bytes per symbol and the percentage it uses. The data is grouped based on the file system location of the object in the tree and the file containing the symbol.
Use the rom_report to get the ROM report:
Using west:
west build -b reel_board samples/hello_world
west build -t rom_report
Using CMake and ninja:
# Use cmake to configure a Ninja-based buildsystem:
cmake -B build -GNinja -DBOARD=reel_board samples/hello_world
# Now run ninja on the generated build system:
ninja -C build rom_report
which will generate something similar to the output below:
Path Size %
==============================================================================================================
...
...
CSWTCH.5 4 0.02%
SystemCoreClock 4 0.02%
__aeabi_idiv0 2 0.01%
__udivmoddi4 702 3.37%
_sw_isr_table 384 1.85%
delay_machine_code.9114 6 0.03%
levels.8826 20 0.10%
mpu_config 8 0.04%
transitions.10558 12 0.06%
arch 1194 5.74%
arm 1194 5.74%
core 1194 5.74%
aarch32 1194 5.74%
cortex_m 852 4.09%
fault.c 400 1.92%
bus_fault.isra.0 60 0.29%
mem_manage_fault.isra.0 56 0.27%
usage_fault.isra.0 36 0.17%
z_arm_fault 232 1.11%
z_arm_fault_init 16 0.08%
irq_init.c 24 0.12%
z_arm_interrupt_init 24 0.12%
mpu 352 1.69%
arm_core_mpu.c 56 0.27%
z_arm_configure_static_mpu_regions 56 0.27%
arm_mpu.c 296 1.42%
__init_sys_init_arm_mpu_init0 8 0.04%
arm_core_mpu_configure_static_mpu_regions 20 0.10%
arm_core_mpu_disable 16 0.08%
arm_core_mpu_enable 20 0.10%
arm_mpu_init 92 0.44%
mpu_configure_regions 140 0.67%
thread_abort.c 76 0.37%
z_impl_k_thread_abort
76 0.37%
...
...
Data Structures¶
Build Target: pahole¶
Poke-a-hole (pahole) is an object-file analysis tool to find the size of the data structures, and the holes caused due to aligning the data elements to the word-size of the CPU by the compiler.
Poke-a-hole (pahole) must be installed prior to using this target. It can be obtained from https://git.kernel.org/pub/scm/devel/pahole/pahole.git and is available in the dwarves package in both fedora and ubuntu:
sudo apt-get install dwarves
or in fedora:
sudo dnf install dwarves
Using west:
west build -b reel_board samples/hello_world
west build -t pahole
Using CMake and ninja:
# Use cmake to configure a Ninja-based buildsystem:
cmake -B build -GNinja -DBOARD=reel_board samples/hello_world
# Now run ninja on the generated build system:
ninja -C build pahole
After running this target, pahole will output the results to the console:
/* Used at: zephyr/isr_tables.c */
/* <80> ../include/sw_isr_table.h:30 */
struct _isr_table_entry {
void * arg; /* 0 4 */
void (*isr)(void *); /* 4 4 */
/* size: 8, cachelines: 1, members: 2 */
/* last cacheline: 8 bytes */
};
/* Used at: zephyr/isr_tables.c */
/* <eb> ../include/arch/arm/aarch32/cortex_m/mpu/arm_mpu_v7m.h:134 */
struct arm_mpu_region_attr {
uint32_t rasr; /* 0 4 */
/* size: 4, cachelines: 1, members: 1 */
/* last cacheline: 4 bytes */
};
/* Used at: zephyr/isr_tables.c */
/* <112> ../include/arch/arm/aarch32/cortex_m/mpu/arm_mpu.h:24 */
struct arm_mpu_region {
uint32_t base; /* 0 4 */
const char * name; /* 4 4 */
arm_mpu_region_attr_t attr; /* 8 4 */
/* size: 12, cachelines: 1, members: 3 */
/* last cacheline: 12 bytes */
};
...
...