diff --git a/libcpu/aarch64/common/include/armv8.h b/libcpu/aarch64/common/include/armv8.h
index 9063a8337d..7388b078a2 100644
--- a/libcpu/aarch64/common/include/armv8.h
+++ b/libcpu/aarch64/common/include/armv8.h
@@ -11,6 +11,12 @@
 #ifndef __ARMV8_H__
 #define __ARMV8_H__
 
+#include <rtconfig.h>
+
+#ifdef ARCH_USING_HW_THREAD_SELF
+#define ARM64_THREAD_REG tpidr_el1
+#endif /* ARCH_USING_HW_THREAD_SELF */
+
 #ifdef __ASSEMBLY__
 
 /*********************
diff --git a/libcpu/aarch64/common/include/cpuport.h b/libcpu/aarch64/common/include/cpuport.h
index e5e256b57b..e5216f3b89 100644
--- a/libcpu/aarch64/common/include/cpuport.h
+++ b/libcpu/aarch64/common/include/cpuport.h
@@ -57,14 +57,14 @@ void _thread_start(void);
 rt_inline struct rt_thread *rt_hw_thread_self(void)
 {
     struct rt_thread *thread;
-    __asm__ volatile ("mrs %0, tpidr_el1":"=r"(thread));
+    __asm__ volatile ("mrs %0, " RT_STRINGIFY(ARM64_THREAD_REG) :"=r"(thread));
 
     return thread;
 }
 
 rt_inline void rt_hw_thread_set_self(struct rt_thread *thread)
 {
-    __asm__ volatile ("msr tpidr_el1, %0"::"r"(thread));
+    __asm__ volatile ("msr " RT_STRINGIFY(ARM64_THREAD_REG) ", %0"::"r"(thread));
 }
 
 #endif /* ARCH_USING_HW_THREAD_SELF */
diff --git a/libcpu/aarch64/common/mp/context_gcc.S b/libcpu/aarch64/common/mp/context_gcc.S
index adc3587651..96743391bf 100644
--- a/libcpu/aarch64/common/mp/context_gcc.S
+++ b/libcpu/aarch64/common/mp/context_gcc.S
@@ -27,6 +27,12 @@
 
 .globl rt_hw_context_switch_to
 
+.macro update_tidr, srcx
+#ifdef ARCH_USING_HW_THREAD_SELF
+    msr     ARM64_THREAD_REG, \srcx
+#endif /* ARCH_USING_HW_THREAD_SELF */
+.endm
+
 /*
  * void rt_hw_context_switch_to(rt_uint3 to, struct rt_thread *to_thread);
  * X0 --> to (thread stack)
@@ -35,6 +41,7 @@
 rt_hw_context_switch_to:
     ldr     x0, [x0]
     mov     sp, x0
+    update_tidr x1
 
     /* reserved to_thread */
     mov     x19, x1
@@ -62,6 +69,7 @@ rt_hw_context_switch:
     str     x3, [x0]            // store sp in preempted tasks TCB
     ldr     x0, [x1]            // get new task stack pointer
     mov     sp, x0
+    update_tidr x2
 
     /* backup thread self */
     mov     x19, x2
@@ -119,6 +127,7 @@ rt_hw_context_switch_interrupt:
     /* setup SP to to-thread's */
     ldr     x0, [TO_SPP]
     mov     sp, x0
+    update_tidr TO_TCB
 
     mov     x0, TO_TCB
     bl      rt_cpus_lock_status_restore
diff --git a/src/scheduler_mp.c b/src/scheduler_mp.c
index 7a7bd61e00..6a228e6b53 100644
--- a/src/scheduler_mp.c
+++ b/src/scheduler_mp.c
@@ -1069,9 +1069,6 @@ void rt_sched_post_ctx_switch(struct rt_thread *thread)
     }
     /* safe to access since irq is masked out */
     pcpu->current_thread = thread;
-#ifdef ARCH_USING_HW_THREAD_SELF
-    rt_hw_thread_set_self(thread);
-#endif /* ARCH_USING_HW_THREAD_SELF */
 }
 
 #ifdef RT_DEBUGING_CRITICAL