arch: arm: cortex_m: Allow VTOR to be relocated to TCM

Relocate VTOR from SRAM to ITCM/DTCM to minimize interrupt latency.
TCM offers single-cycle access compared to multi-cycle SRAM reads.
This improves exception handling speed for real-time workloads.

Signed-off-by: Peter van der Perk <[email protected]>

Author: PetervdPerk-NXP

arch/arm/Kconfig

@@ -190,4 +190,40 @@ config HAS_SWO
 	help
 	  When enabled, indicates that SoC has an SWO output
 
+DT_CHOSEN_Z_DTCM := zephyr,dtcm
+DT_CHOSEN_Z_ITCM := zephyr,itcm
+
+choice
+	prompt "Vector table memory location"
+	depends on SRAM_VECTOR_TABLE
+	default VECTOR_TABLE_SRAM
+
+config VECTOR_TABLE_SRAM
+	bool "SRAM"
+	help
+	  Place the vector table in SRAM.
+
+config VECTOR_TABLE_DTCM
+	bool "DTCM"
+	depends on $(dt_chosen_enabled,$(DT_CHOSEN_Z_DTCM))
+	help
+	  Place the interrupt vector table in Data Tightly Coupled Memory (DTCM).
+
+	  While the vector table is instruction-fetched during exception entry, a DTCM
+	  option is provided for systems where ITCM is unavailable. DTCM still offers
+	  low-latency, deterministic access compared to normal RAM, but is not the
+	  optimal location for instruction fetch performance.
+
+config VECTOR_TABLE_ITCM
+	bool "ITCM"
+	depends on $(dt_chosen_enabled,$(DT_CHOSEN_Z_ITCM))
+	help
+	  Place the interrupt vector table in Instruction Tightly Coupled Memory (ITCM).
+
+	  ITCM is accessed via the CPU instruction bus and provides single-cycle,
+	  deterministic instruction fetches. This typically yields the lowest possible
+	  interrupt latency and is the preferred location when available.
+
+endchoice
+
 endmenu

arch/arm/core/cortex_m/ram_vector_table.ld

@@ -28,5 +28,11 @@ SECTION_PROLOGUE(.sram_vt,,)
 	. += _vector_end - _vector_start;
 	MPU_ALIGN(_sram_vector_size);
 	_sram_vector_end = .;
+#if defined(CONFIG_VECTOR_TABLE_ITCM)
+} GROUP_DATA_LINK_IN(ITCM, ROMABLE_REGION)
+#elif defined(CONFIG_VECTOR_TABLE_DTCM)
+} GROUP_DATA_LINK_IN(DTCM, ROMABLE_REGION)
+#else
 } GROUP_DATA_LINK_IN(RAMABLE_REGION, ROMABLE_REGION)
+#endif
 _sram_vector_size = _sram_vector_end - _sram_vector_start;