interpreters/python: Optmize Python for size

[tmedicci]

Summary

Reduce general firmware size by optimizing Python-related sources.

Impact

Impact on user: Yes. Enable reducing firmware size when CONFIG_INTERPRETERS_CPYTHON is enabled.

Impact on build: Yes. Reduce firmware size by ~6% (tested on ESP32-P4)

Impact on hardware: No.

Impact on documentation: No.

Impact on security: No.

Impact on compatibility: No.

Testing

Build esp32p4-function-ev-board:python defconfig before and after applying this patch.

Building

make -j distclean && ./tools/configure.sh -S esp32p4-function-ev-board:python && make -s -j$(nproc)

Running

After flashing the device, just run python on NSH to ensure that it's being properly initialized.

Results

Comparison between the firmwares built before and after applying this patch:

Before

Memory region         Used Size  Region Size  %age Used
   tcm_idram_seg:         136 B         8 KB      1.66%
        irom_seg:     3159284 B        64 MB      4.71%
        sram_seg:      374776 B     978880 B     38.29%
        drom_seg:    16062776 B        64 MB     23.94%
rev3_mspi_workaround_seg:           0 B          0 B
      lp_ram_seg:         104 B      32744 B      0.32%
 lp_reserved_seg:          24 B         24 B    100.00%
  extern_ram_seg:          2 MB        64 MB      3.12%
CP: nuttx.hex

And the final firmware size:

-rw-r--r-- 1 tiago tiago 16454900 jun 15 11:32 nuttx.bin

After

Memory region         Used Size  Region Size  %age Used
   tcm_idram_seg:         136 B         8 KB      1.66%
        irom_seg:     2192720 B        64 MB      3.27%
        sram_seg:      374776 B     978880 B     38.29%
        drom_seg:    15034520 B        64 MB     22.40%
rev3_mspi_workaround_seg:           0 B          0 B
      lp_ram_seg:         104 B      32744 B      0.32%
 lp_reserved_seg:          24 B         24 B    100.00%
  extern_ram_seg:          2 MB        64 MB      3.12%
CP: nuttx.hex

And the final firmware size:

-rw-r--r-- 1 tiago tiago 15422800 jun 15 11:41 nuttx.bin

[xiaoxiang781216]

should we set OPT to empty string and let CFLAGS at line 160 control the optimization level

[tmedicci]

I couldn't investigate it further, but it wasn't working as expected, so I set it directly on OPT for Python's configure script.

We can set this as an issue and investigate it later.

[acassis]

@tmedicci @xiaoxiang781216 I think the optimization flags should come from NuttX building system, because it was common in the past have some optimizations which crash the final firmware for some compilers, so fixing this OPT could cause that issue on some arch. But the idea of reducing size is very good, but it could come from defconfig

[cederom]

+1 from me, out of lab ;-)

…

-- CeDeROM, SQ7MHZ, http://www.tomek.cedro.info