smp: Move for loops to use arch_num_cpus instead of CONFIG_MP_NUM_CPUS

Change for loops of the form:

for (i = 0; i < CONFIG_MP_NUM_CPUS; i++)
   ...

to

unsigned int num_cpus = arch_num_cpus();
for (i = 0; i < num_cpus; i++)
   ...

We do the call outside of the for loop so that it only happens once,
rather than on every iteration.

Signed-off-by: Kumar Gala <kumar.gala@intel.com>

arch/arc/core/arc_smp.c

@@ -133,7 +133,9 @@ void arch_sched_ipi(void)
 	/* broadcast sched_ipi request to other cores
 	 * if the target is current core, hardware will ignore it
 	 */
-	for (i = 0U; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (i = 0U; i < num_cpus; i++) {
 		z_arc_connect_ici_generate(i);
 	}
 }

arch/arm64/core/fpu.c

@@ -96,7 +96,9 @@ static void flush_owned_fpu(struct k_thread *thread)
 	int i;
 
 	/* search all CPUs for the owner we want */
-	for (i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (i = 0; i < num_cpus; i++) {
 		if (_kernel.cpus[i].arch.fpu_owner != thread) {
 			continue;
 		}

arch/arm64/core/smp.c

@@ -160,7 +160,9 @@ static void broadcast_ipi(unsigned int ipi)
 	/*
 	 * Send SGI to all cores except itself
 	 */
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		uint64_t target_mpidr = cpu_node_list[i];
 		uint8_t aff0 = MPIDR_AFFLVL(target_mpidr, 0);
 

arch/riscv/core/smp.c

@@ -67,7 +67,9 @@ void arch_sched_ipi(void)
 	key = arch_irq_lock();
 
 	id = _current_cpu->id;
-	for (i = 0U; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (i = 0U; i < num_cpus; i++) {
 		if (i != id) {
 			volatile uint32_t *r = (uint32_t *)get_hart_msip(i);
 			*r = 1U;

arch/x86/core/prep_c.c

@@ -67,7 +67,9 @@ FUNC_NORETURN void z_x86_prep_c(void *arg)
 #endif
 
 #if CONFIG_X86_STACK_PROTECTION
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		z_x86_set_stack_guard(z_interrupt_stacks[i]);
 	}
 #endif

drivers/interrupt_controller/intc_dw_ace.c

@@ -85,7 +85,9 @@ void dw_ace_irq_enable(const struct device *dev, uint32_t irq)
 	ARG_UNUSED(dev);
 
 	if (is_dw_irq(irq)) {
-		for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+		unsigned int num_cpus = arch_num_cpus();
+
+		for (int i = 0; i < num_cpus; i++) {
 			ACE_INTC[i].irq_inten_l |= BIT(ACE_IRQ_FROM_ZEPHYR(irq));
 			ACE_INTC[i].irq_intmask_l &= ~BIT(ACE_IRQ_FROM_ZEPHYR(irq));
 		}
@@ -99,7 +101,9 @@ void dw_ace_irq_disable(const struct device *dev, uint32_t irq)
 	ARG_UNUSED(dev);
 
 	if (is_dw_irq(irq)) {
-		for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+		unsigned int num_cpus = arch_num_cpus();
+
+		for (int i = 0; i < num_cpus; i++) {
 			ACE_INTC[i].irq_inten_l &= ~BIT(ACE_IRQ_FROM_ZEPHYR(irq));
 			ACE_INTC[i].irq_intmask_l |= BIT(ACE_IRQ_FROM_ZEPHYR(irq));
 		}

drivers/interrupt_controller/intc_esp32.c

@@ -202,7 +202,9 @@ static struct vector_desc_t *find_desc_for_source(int source, int cpu)
 
 void esp_intr_initialize(void)
 {
-	for (size_t i = 0; i < (ESP_INTC_INTS_NUM * CONFIG_MP_NUM_CPUS); ++i) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (size_t i = 0; i < (ESP_INTC_INTS_NUM * num_cpus); ++i) {
 		intr_alloc_table[i].handler = default_intr_handler;
 		intr_alloc_table[i].arg = (void *)i;
 	}

drivers/interrupt_controller/intc_gic.c

@@ -138,9 +138,11 @@ static void gic_dist_init(void)
 
 	/*
 	 * Enable all global interrupts distributing to CPUs listed
-	 * in dts with the count of CONFIG_MP_NUM_CPUS.
+	 * in dts with the count of arch_num_cpus().
 	 */
-	for (i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (i = 0; i < num_cpus; i++) {
 		cpu_mask |= BIT(cpu_mpid_list[i]);
 	}
 	reg_val = cpu_mask | (cpu_mask << 8) | (cpu_mask << 16)

drivers/ipm/ipm_cavs_idc.c

@@ -36,7 +36,9 @@ static void cavs_idc_isr(const struct device *dev)
 	bool do_sched_ipi = false;
 #endif
 
-	for (i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (i = 0; i < num_cpus; i++) {
 		if (i == curr_cpu_id) {
 			/* skip current core */
 			continue;
@@ -94,7 +96,9 @@ static int cavs_idc_send(const struct device *dev, int wait, uint32_t id,
 
 	/* Check if any core is still busy */
 	busy = false;
-	for (i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (i = 0; i < num_cpus; i++) {
 		if (i == curr_cpu_id) {
 			/* skip current core */
 			continue;
@@ -116,7 +120,7 @@ static int cavs_idc_send(const struct device *dev, int wait, uint32_t id,
 	ext &= IPC_IDCIETC_MSG_MASK;
 	ext |= IPC_IDCIETC_DONE; /* always clear DONE bit */
 
-	for (i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	for (i = 0; i < num_cpus; i++) {
 		if (i == curr_cpu_id) {
 			/* skip current core */
 			continue;
@@ -173,11 +177,13 @@ static int cavs_idc_set_enabled(const struct device *dev, int enable)
 	}
 #endif
 
-	for (i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (i = 0; i < num_cpus; i++) {
 		mask = 0;
 
 		if (enable) {
-			for (j = 0; j < CONFIG_MP_NUM_CPUS; j++) {
+			for (j = 0; j < num_cpus; j++) {
 				if (i == j) {
 					continue;
 				}

kernel/sched.c

@@ -593,7 +593,9 @@ static bool thread_active_elsewhere(struct k_thread *thread)
 #ifdef CONFIG_SMP
 	int currcpu = _current_cpu->id;
 
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		if ((i != currcpu) &&
 		    (_kernel.cpus[i].current == thread)) {
 			return true;
@@ -1291,7 +1293,9 @@ void init_ready_q(struct _ready_q *rq)
 void z_sched_init(void)
 {
 #ifdef CONFIG_SCHED_CPU_MASK_PIN_ONLY
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		init_ready_q(&_kernel.cpus[i].ready_q);
 	}
 #else

kernel/smp.c

@@ -112,7 +112,10 @@ void z_smp_start_cpu(int id)
 void z_smp_init(void)
 {
 	(void)atomic_clear(&start_flag);
-	for (int i = 1; i < CONFIG_MP_NUM_CPUS; i++) {
+
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 1; i < num_cpus; i++) {
 		start_cpu(i, &start_flag);
 	}
 	(void)atomic_set(&start_flag, 1);

kernel/thread.c

@@ -1100,7 +1100,9 @@ int k_thread_runtime_stats_all_get(k_thread_runtime_stats_t *stats)
 #ifdef CONFIG_SCHED_THREAD_USAGE_ALL
 	/* Retrieve the usage stats for each core and amalgamate them. */
 
-	for (uint8_t i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (uint8_t i = 0; i < num_cpus; i++) {
 		z_sched_cpu_usage(i, &tmp_stats);
 
 		stats->execution_cycles += tmp_stats.execution_cycles;

kernel/usage.c

@@ -296,7 +296,9 @@ void k_sys_runtime_stats_enable(void)
 
 	/* Enable gathering of runtime stats on each CPU */
 
-	for (uint8_t i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (uint8_t i = 0; i < num_cpus; i++) {
 		_kernel.cpus[i].usage.track_usage = true;
 #ifdef CONFIG_SCHED_THREAD_USAGE_ANALYSIS
 		_kernel.cpus[i].usage.num_windows++;
@@ -328,7 +330,9 @@ void k_sys_runtime_stats_disable(void)
 
 	uint32_t now = usage_now();
 
-	for (uint8_t i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (uint8_t i = 0; i < num_cpus; i++) {
 		cpu = &_kernel.cpus[i];
 		if (cpu->usage0 != 0) {
 			sched_cpu_update_usage(cpu, now - cpu->usage0);

soc/xtensa/intel_adsp/ace/multiprocessing.c

@@ -46,7 +46,9 @@ void soc_mp_init(void)
 
 	irq_enable(ACE_IRQ_TO_ZEPHYR(ACE_INTL_IDCA));
 
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		/* DINT has one bit per IPC, unmask only IPC "Ax" on core "x" */
 		ACE_DINT[i].ie[ACE_INTL_IDCA] = BIT(i);
 
@@ -111,7 +113,9 @@ void arch_sched_ipi(void)
 	uint32_t curr = arch_proc_id();
 
 	/* Signal agent B[n] to cause an interrupt from agent A[n] */
-	for (int core = 0; core < CONFIG_MP_NUM_CPUS; core++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int core = 0; core < num_cpus; core++) {
 		if (core != curr && soc_cpus_active[core]) {
 			IDC[core].agents[1].ipc.idr = INTEL_ADSP_IPC_BUSY;
 		}

soc/xtensa/intel_adsp/cavs/multiprocessing.c

@@ -29,7 +29,9 @@ __imr void soc_mp_startup(uint32_t cpu)
 	 * spurious IPI when we enter user code).  Remember: this
 	 * could have come from any core, clear all of them.
 	 */
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		IDC[cpu].core[i].tfc = BIT(31);
 	}
 
@@ -108,7 +110,9 @@ void soc_start_core(int cpu_num)
 	 * some platforms will mess it up.
 	 */
 	CAVS_INTCTRL[cpu_num].l2.clear = CAVS_L2_IDC;
-	for (int c = 0; c < CONFIG_MP_NUM_CPUS; c++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int c = 0; c < num_cpus; c++) {
 		IDC[c].busy_int |= IDC_ALL_CORES;
 	}
 
@@ -126,8 +130,9 @@ void soc_start_core(int cpu_num)
 void arch_sched_ipi(void)
 {
 	uint32_t curr = arch_proc_id();
+	unsigned int num_cpus = arch_num_cpus();
 
-	for (int c = 0; c < CONFIG_MP_NUM_CPUS; c++) {
+	for (int c = 0; c < num_cpus; c++) {
 		if (c != curr && soc_cpus_active[c]) {
 			IDC[curr].core[c].itc = BIT(31);
 		}
@@ -148,7 +153,10 @@ void idc_isr(const void *param)
 	 * of the ITC/TFC high bits, INCLUDING the one "from this
 	 * CPU".
 	 */
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		IDC[arch_proc_id()].core[i].tfc = BIT(31);
 	}
 }
@@ -161,7 +169,9 @@ __imr void soc_mp_init(void)
 	 * every other CPU, but not to be back-interrupted when the
 	 * target core clears the busy bit.
 	 */
-	for (int core = 0; core < CONFIG_MP_NUM_CPUS; core++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int core = 0; core < num_cpus; core++) {
 		IDC[core].busy_int |= IDC_ALL_CORES;
 		IDC[core].done_int &= ~IDC_ALL_CORES;
 
@@ -172,8 +182,8 @@ __imr void soc_mp_init(void)
 	}
 
 	/* Clear out any existing pending interrupts that might be present */
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
-		for (int j = 0; j < CONFIG_MP_NUM_CPUS; j++) {
+	for (int i = 0; i < num_cpus; i++) {
+		for (int j = 0; j < num_cpus; j++) {
 			IDC[i].core[j].tfc = BIT(31);
 		}
 	}

soc/xtensa/intel_adsp/common/clk.c

@@ -66,7 +66,10 @@ int cavs_clock_set_freq(uint32_t freq_idx)
 	k = k_spin_lock(&lock);
 
 	select_cpu_clock_hw(freq_idx);
-	for (i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (i = 0; i < num_cpus; i++) {
 		platform_clocks[i].current_freq = freq_idx;
 	}
 
@@ -94,7 +97,9 @@ void cavs_clock_init(void)
 	}
 #endif
 
-	for (i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (i = 0; i < num_cpus; i++) {
 		platform_clocks[i].default_freq = CAVS_CLOCK_FREQ_DEFAULT;
 		platform_clocks[i].current_freq = CAVS_CLOCK_FREQ_DEFAULT;
 		platform_clocks[i].lowest_freq = platform_lowest_freq_idx;

soc/xtensa/intel_adsp/common/ipc.c

@@ -169,7 +169,9 @@ bool intel_adsp_ipc_send_message_sync(const struct device *dev,
 #if defined(CONFIG_SOC_SERIES_INTEL_ACE)
 static inline void ace_ipc_intc_unmask(void)
 {
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		ACE_DINT[i].ie[ACE_INTL_HIPC] = BIT(0);
 	}
 }

subsys/debug/thread_analyzer.c

@@ -131,7 +131,9 @@ K_KERNEL_STACK_ARRAY_DECLARE(z_interrupt_stacks, CONFIG_MP_MAX_NUM_CPUS,
 
 static void isr_stacks(void)
 {
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		const uint8_t *buf = Z_KERNEL_STACK_BUFFER(z_interrupt_stacks[i]);
 		size_t size = K_KERNEL_STACK_SIZEOF(z_interrupt_stacks[i]);
 		size_t unused;

subsys/pm/pm_stats.c

@@ -35,7 +35,9 @@ static int pm_stats_init(const struct device *dev)
 {
 	ARG_UNUSED(dev);
 
-	for (uint8_t i = 0U; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (uint8_t i = 0U; i < num_cpus; i++) {
 		for (uint8_t j = 0U; j < PM_STATE_COUNT; j++) {
 			snprintk(names[i][j], PM_STAT_NAME_LEN,
 				 "pm_cpu_%03d_state_%1d_stats", i, j);

subsys/shell/modules/kernel_service.c

@@ -211,7 +211,9 @@ static int cmd_kernel_stacks(const struct shell *shell,
 	 * kernel support, including dumping arch-specific exception-related
 	 * stack buffers.
 	 */
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		size_t unused;
 		const uint8_t *buf = Z_KERNEL_STACK_BUFFER(z_interrupt_stacks[i]);
 		size_t size = K_KERNEL_STACK_SIZEOF(z_interrupt_stacks[i]);

subsys/testsuite/ztest/src/ztress.c

@@ -120,8 +120,9 @@ static void control_load(void)
 	uint64_t idle_cycles = 0;
 	k_thread_runtime_stats_t rt_stats_all;
 	int err = 0;
+	unsigned int num_cpus = arch_num_cpus();
 
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	for (int i = 0; i < num_cpus; i++) {
 		k_thread_runtime_stats_t thread_stats;
 
 		err = k_thread_runtime_stats_get(idle_tid[i], &thread_stats);

tests/boards/intel_adsp/smoke/src/cpus.c

@@ -139,7 +139,9 @@ static void core_smoke(void *arg)
 
 ZTEST(intel_adsp_boot, test_4th_cpu_behavior)
 {
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		printk("Per-CPU smoke test %d...\n", i);
 		run_on_cpu(i, core_smoke, (void *)i, true);
 	}
@@ -196,7 +198,9 @@ static void halt_and_restart(int cpu)
 
 void halt_and_restart_thread(void *p1, void *p2, void *p3)
 {
-	for (int i = 1; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 1; i < num_cpus; i++) {
 		halt_and_restart(i);
 	}
 }

tests/boards/intel_adsp/smoke/src/smpboot.c

@@ -41,11 +41,13 @@ static void thread_fn(void *a, void *b, void *c)
  */
 ZTEST(intel_adsp_boot, test_1st_smp_boot_delay)
 {
+	unsigned int num_cpus = arch_num_cpus();
+
 	if (CONFIG_MP_NUM_CPUS < 2) {
 		ztest_test_skip();
 	}
 
-	for (int i = 1; i < CONFIG_MP_NUM_CPUS; i++) {
+	for (int i = 1; i < num_cpus; i++) {
 		printk("Launch cpu%d\n", i);
 		mp_flag = false;
 		k_thread_create(&cpu_thr, thr_stack, K_THREAD_STACK_SIZEOF(thr_stack),

tests/drivers/clock_control/cavs_clock/src/main.c

@@ -10,8 +10,9 @@
 static void check_clocks(struct cavs_clock_info *clocks, uint32_t freq_idx)
 {
 	int i;
+	unsigned int num_cpus = arch_num_cpus();
 
-	for (i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	for (i = 0; i < num_cpus; i++) {
 		zassert_equal(clocks[i].current_freq, freq_idx, "");
 	}
 }

tests/kernel/threads/thread_stack/src/main.c

@@ -391,7 +391,10 @@ ZTEST(userspace_thread_stack, test_stack_buffer)
 	       K_KERNEL_STACK_RESERVED);
 
 	printk("CONFIG_ISR_STACK_SIZE %zu\n", (size_t)CONFIG_ISR_STACK_SIZE);
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		printk("irq stack %d: %p size %zu\n",
 		       i, &z_interrupt_stacks[i],
 		       sizeof(z_interrupt_stacks[i]));

tests/lib/p4workq/src/main.c

@@ -180,7 +180,9 @@ ZTEST(lib_p4wq, test_fill_queue)
 	 * we can be sure to run).  They should all be made active
 	 * when added.
 	 */
-	for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+	unsigned int num_cpus = arch_num_cpus();
+
+	for (int i = 0; i < num_cpus; i++) {
 		zassert_true(add_new_item(p0), "thread should be active");
 	}
 
@@ -192,7 +194,7 @@ ZTEST(lib_p4wq, test_fill_queue)
 	 * we run out of threads.
 	 */
 	for (int pri = p0 - 1; pri >= p0 - 4; pri++) {
-		for (int i = 0; i < CONFIG_MP_NUM_CPUS; i++) {
+		for (int i = 0; i < num_cpus; i++) {
 			bool active = add_new_item(pri);
 
 			if (!active) {