perf: 6 performance optimizations for STM32WB55 (Cortex-M4 @ 64MHz)

[WonderMr]

What's new

1. Compiler: -Og → -Os (release builds) Replace debug-level optimization with size-optimized -Os, which enables most -O2 passes: function inlining, dead code elimination, loop optimization, aggressive register allocation, tail call optimization, and common subexpression elimination. File: site_scons/firmwareopts.scons

2. Disable heap memset on free() in release configHEAP_CLEAR_MEMORY_ON_FREE was always 1, causing FreeRTOS Heap_4 to memset() every freed block to zero. Useful for catching use-after-free in debug, but pure waste in release. Now conditional on FURI_DEBUG. Saves ~500+ memset calls per second during active GUI/protocol work. File: targets/f7/inc/FreeRTOSConfig.h

3. SPI TX via DMA instead of busy-wait polling furi_hal_spi_bus_tx() polled TXE flag byte-by-byte, keeping the CPU in a tight loop for entire SPI transfers. Now delegates to furi_hal_spi_bus_trx_dma() which uses DMA2_Channel7 and FreeRTOS semaphore-based sleep, freeing CPU during display updates (~1KB/ frame @ 20fps) and radio TX operations. File: targets/f7/furi_hal/furi_hal_spi.c

4. Fix realloc() to copy min(old_size, new_size) Original realloc() copied size (new) bytes from old block via memcpy, which reads past allocation when growing. Added memmgr_heap_get_block_size() that reads usable size from Heap_4 BlockLink_t header. Now copies min(old_size, new_size) bytes, fixing potential UB and reducing unnecessary copying. Files: furi/core/memmgr.c, furi/core/memmgr_heap.c/.h, targets/f7/api_symbols.csv

5. Fix calloc() to explicitly zero memory Original calloc() just called pvPortMalloc() without memset, relying on configHEAP_CLEAR_MEMORY_ON_FREE=1 for zero-initialized returns. With optimization №2 disabling that in release, calloc() would return uninitialized memory. Added explicit memset(0). File: furi/core/memmgr.c

6. Branch prediction hints on furi_check/assert/break Added __builtin_expect(!(__e), 0) to all assertion macros. Tells GCC that error path is cold: crash code moves to end of function, hot path becomes fall-through (0 pipeline penalty on Cortex-M4 3-stage pipeline). Affects ~2300+ call sites across the firmware. File: furi/core/check.h

Also: strncpy → strlcpy in subghz_scene_save_name.c (-Os exposed -Werror=stringop-truncation warning).

Verification

• Build: ./fbt COMPACT=1 DEBUG=0
• Boot device, navigate Settings → About — FW version shown
• Open SubGHz, NFC, IR apps — UI responsive, no hangs

Checklist (For Reviewer)

• PR has description of feature/bug
• Description contains actions to verify feature/bugfix
• I've built this code, uploaded it to the device and verified feature/bugfix

[xMasterX]

Checklist (For Reviewer) clearly says for who its made, stop editing it pls, thats for repo maintainers, not PR author