diff --git a/Makefile b/Makefile
index 01b26d2..46ce829 100644
--- a/Makefile
+++ b/Makefile
@@ -20,7 +20,7 @@ ifneq (,)
   mabi := -mabi=$(if $(is_32bit),ilp32,lp64)
 endif
 
-CFLAGS        := -Wall -Werror  -fno-builtin -nostdlib -D__NO_INLINE__ -mcmodel=medany -g -Og -std=gnu99 -Wno-unused -Wno-attributes -fno-delete-null-pointer-checks -fno-PIE $(march)
+CFLAGS        := -Wall -Werror  -fno-builtin -nostdlib -D__NO_INLINE__ -mcmodel=medany -g -O0 -std=gnu99 -Wno-unused -Wno-attributes -fno-delete-null-pointer-checks -fno-PIE $(march)
 COMPILE       	:= $(CC) -MMD -MP $(CFLAGS) $(SPROJS_INCLUDE)
 
 #---------------------	utils -----------------------
@@ -63,21 +63,24 @@ SPIKE_INF_LIB   := $(OBJ_DIR)/spike_interface.a
 
 
 #---------------------	user   -----------------------
-USER_CPPS 		:= user/*.c 
+USER_CPP0 		:= user/app_alloc0.c user/user_lib.c
+USER_CPP1 		:= user/app_alloc1.c user/user_lib.c
 
-USER_CPPS  		:= $(wildcard $(USER_CPPS))
-USER_OBJS  		:= $(addprefix $(OBJ_DIR)/, $(patsubst %.c,%.o,$(USER_CPPS)))
+USER_CPP0  		:= $(wildcard $(USER_CPP0))
+USER_CPP1  		:= $(wildcard $(USER_CPP1))
+USER_OBJ0  		:= $(addprefix $(OBJ_DIR)/, $(patsubst %.c,%.o,$(USER_CPP0)))
+USER_OBJ1  		:= $(addprefix $(OBJ_DIR)/, $(patsubst %.c,%.o,$(USER_CPP1)))
 
-
-
-USER_TARGET 	:= $(OBJ_DIR)/app_sum_sequence
+USER_TARGET0 	:= $(OBJ_DIR)/app_alloc0
+USER_TARGET1 	:= $(OBJ_DIR)/app_alloc1
 #------------------------targets------------------------
 $(OBJ_DIR):
 	@-mkdir -p $(OBJ_DIR)	
 	@-mkdir -p $(dir $(UTIL_OBJS))
 	@-mkdir -p $(dir $(SPIKE_INF_OBJS))
 	@-mkdir -p $(dir $(KERNEL_OBJS))
-	@-mkdir -p $(dir $(USER_OBJS))
+	@-mkdir -p $(dir $(USER_OBJ0))
+	@-mkdir -p $(dir $(USER_OBJ1))
 
 $(OBJ_DIR)/%.o : %.c
 	@echo "compiling" $<
@@ -102,9 +105,14 @@ $(KERNEL_TARGET): $(OBJ_DIR) $(UTIL_LIB) $(SPIKE_INF_LIB) $(KERNEL_OBJS) $(KERNE
 	@$(COMPILE) $(KERNEL_OBJS) $(UTIL_LIB) $(SPIKE_INF_LIB) -o $@ -T $(KERNEL_LDS)
 	@echo "PKE core has been built into" \"$@\"
 
-$(USER_TARGET): $(OBJ_DIR) $(UTIL_LIB) $(USER_OBJS)
+$(USER_TARGET0): $(OBJ_DIR) $(UTIL_LIB) $(USER_OBJ0)
+	@echo "linking" $@	...	
+	@$(COMPILE) --entry=main $(USER_OBJ0) $(UTIL_LIB) -o $@
+	@echo "User app has been built into" \"$@\"
+	
+$(USER_TARGET1): $(OBJ_DIR) $(UTIL_LIB) $(USER_OBJ1)
 	@echo "linking" $@	...	
-	@$(COMPILE) --entry=main $(USER_OBJS) $(UTIL_LIB) -o $@
+	@$(COMPILE) --entry=main $(USER_OBJ1) $(UTIL_LIB) -o $@
 	@echo "User app has been built into" \"$@\"
 
 -include $(wildcard $(OBJ_DIR)/*/*.d)
@@ -112,16 +120,16 @@ $(USER_TARGET): $(OBJ_DIR) $(UTIL_LIB) $(USER_OBJS)
 
 .DEFAULT_GOAL := $(all)
 
-all: $(KERNEL_TARGET) $(USER_TARGET)
+all: $(KERNEL_TARGET) $(USER_TARGET0) $(USER_TARGET1)
 .PHONY:all
 
-run: $(KERNEL_TARGET) $(USER_TARGET)
+run: $(KERNEL_TARGET) $(USER_TARGET0) $(USER_TARGET1)
 	@echo "********************HUST PKE********************"
-	spike $(KERNEL_TARGET) $(USER_TARGET)
+	spike -p2 $(KERNEL_TARGET) $(USER_TARGET0) $(USER_TARGET1)
 
 # need openocd!
-gdb:$(KERNEL_TARGET) $(USER_TARGET)
-	spike --rbb-port=9824 -H $(KERNEL_TARGET) $(USER_TARGET) &
+gdb:$(KERNEL_TARGET) $(USER_TARGET0) $(USER_TARGET1)
+	spike --rbb-port=9824 -H -p2 $(KERNEL_TARGET) $(USER_TARGET0) $(USER_TARGET1) &
 	@sleep 1
 	openocd -f ./.spike.cfg &
 	@sleep 1
@@ -135,7 +143,8 @@ gdb_clean:
 
 objdump:
 	riscv64-unknown-elf-objdump -d $(KERNEL_TARGET) > $(OBJ_DIR)/kernel_dump
-	riscv64-unknown-elf-objdump -d $(USER_TARGET) > $(OBJ_DIR)/user_dump
+	riscv64-unknown-elf-objdump -d $(USER_TARGET0) > $(OBJ_DIR)/app_alloc0_dump
+	riscv64-unknown-elf-objdump -d $(USER_TARGET1) > $(OBJ_DIR)/app_alloc1_dump
 
 cscope:
 	find ./ -name "*.c" > cscope.files
@@ -148,4 +157,4 @@ format:
 	@python ./format.py ./
 
 clean:
-	rm -fr ${OBJ_DIR}
+	rm -fr ${OBJ_DIR}
\ No newline at end of file
diff --git a/kernel/config.h b/kernel/config.h
index bdc8270..13f51c2 100644
--- a/kernel/config.h
+++ b/kernel/config.h
@@ -1,8 +1,8 @@
 #ifndef _CONFIG_H_
 #define _CONFIG_H_
 
-// we use only one HART (cpu) in fundamental experiments
-#define NCPU 1
+// we use two HART (cpu) in challenge3
+#define NCPU 2
 
 //interval of timer interrupt. added @lab1_3
 #define TIMER_INTERVAL 1000000
diff --git a/kernel/elf.c b/kernel/elf.c
index 7d7fa94..245c840 100644
--- a/kernel/elf.c
+++ b/kernel/elf.c
@@ -123,7 +123,7 @@ void load_bincode_from_host_elf(process *p) {
   size_t argc = parse_args(&arg_bug_msg);
   if (!argc) panic("You need to specify the application program!\n");
 
-  sprint("Application: %s\n", arg_bug_msg.argv[0]);
+  sprint("hartid = ?: Application: %s\n", arg_bug_msg.argv[0]);
 
   //elf loading. elf_ctx is defined in kernel/elf.h, used to track the loading process.
   elf_ctx elfloader;
@@ -148,5 +148,5 @@ void load_bincode_from_host_elf(process *p) {
   // close the host spike file
   spike_file_close( info.f );
 
-  sprint("Application program entry point (virtual address): 0x%lx\n", p->trapframe->epc);
+  sprint("hartid = ?: Application program entry point (virtual address): 0x%lx\n", p->trapframe->epc);
 }
diff --git a/kernel/kernel.c b/kernel/kernel.c
index d633654..1edbf7c 100644
--- a/kernel/kernel.c
+++ b/kernel/kernel.c
@@ -52,7 +52,7 @@ void load_user_program(process *proc) {
   // USER_STACK_TOP = 0x7ffff000, defined in kernel/memlayout.h
   proc->trapframe->regs.sp = USER_STACK_TOP;  //virtual address of user stack top
 
-  sprint("user frame 0x%lx, user stack 0x%lx, user kstack 0x%lx \n", proc->trapframe,
+  sprint("hartid = ?: user frame 0x%lx, user stack 0x%lx, user kstack 0x%lx \n", proc->trapframe,
          proc->trapframe->regs.sp, proc->kstack);
 
   // load_bincode_from_host_elf() is defined in kernel/elf.c
@@ -77,7 +77,7 @@ void load_user_program(process *proc) {
 // s_start: S-mode entry point of riscv-pke OS kernel.
 //
 int s_start(void) {
-  sprint("Enter supervisor mode...\n");
+  sprint("hartid = ?: Enter supervisor mode...\n");
   // in the beginning, we use Bare mode (direct) memory mapping as in lab1.
   // but now, we are going to switch to the paging mode @lab2_1.
   // note, the code still works in Bare mode when calling pmm_init() and kern_vm_init().
@@ -97,7 +97,11 @@ int s_start(void) {
   // the application code (elf) is first loaded into memory, and then put into execution
   load_user_program(&user_app);
 
-  sprint("Switch to user mode...\n");
+  sprint("hartid = ?: Switch to user mode...\n");
+  
+  uint64 hartid = 0;
+  
+  vm_alloc_stage[hartid] = 1;
   // switch_to() is defined in kernel/process.c
   switch_to(&user_app);
 
diff --git a/kernel/machine/mtrap.c b/kernel/machine/mtrap.c
index 72566ba..e85069f 100644
--- a/kernel/machine/mtrap.c
+++ b/kernel/machine/mtrap.c
@@ -16,7 +16,7 @@ static void handle_misaligned_store() { panic("Misaligned AMO!"); }
 
 // added @lab1_3
 static void handle_timer() {
-  int cpuid = 0;
+  int cpuid = read_csr(mhartid);
   // setup the timer fired at next time (TIMER_INTERVAL from now)
   *(uint64*)CLINT_MTIMECMP(cpuid) = *(uint64*)CLINT_MTIMECMP(cpuid) + TIMER_INTERVAL;
 
diff --git a/kernel/pmm.c b/kernel/pmm.c
index c10e007..9517d2a 100644
--- a/kernel/pmm.c
+++ b/kernel/pmm.c
@@ -15,6 +15,7 @@ extern uint64 g_mem_size;
 static uint64 free_mem_start_addr;  //beginning address of free memory
 static uint64 free_mem_end_addr;    //end address of free memory (not included)
 
+int vm_alloc_stage[NCPU] = { 0 }; // 0 for kernel alloc, 1 for user alloc
 typedef struct node {
   struct node *next;
 } list_node;
@@ -51,8 +52,11 @@ void free_page(void *pa) {
 //
 void *alloc_page(void) {
   list_node *n = g_free_mem_list.next;
+  uint64 hartid = 0;
+  if (vm_alloc_stage[hartid]) {
+    sprint("hartid = %ld: alloc page 0x%x\n", hartid, n);
+  }
   if (n) g_free_mem_list.next = n->next;
-
   return (void *)n;
 }
 
diff --git a/kernel/pmm.h b/kernel/pmm.h
index bce9075..bd5ba21 100644
--- a/kernel/pmm.h
+++ b/kernel/pmm.h
@@ -1,6 +1,8 @@
 #ifndef _PMM_H_
 #define _PMM_H_
 
+#include "config.h"
+
 // Initialize phisical memeory manager
 void pmm_init();
 // Allocate a free phisical page
@@ -8,4 +10,6 @@ void* alloc_page();
 // Free an allocated page
 void free_page(void* pa);
 
+extern int vm_alloc_stage[NCPU];
+
 #endif
\ No newline at end of file
diff --git a/kernel/sync_utils.h b/kernel/sync_utils.h
new file mode 100644
index 0000000..988fc4c
--- /dev/null
+++ b/kernel/sync_utils.h
@@ -0,0 +1,20 @@
+#ifndef _SYNC_UTILS_H_
+#define _SYNC_UTILS_H_
+
+static inline void sync_barrier(volatile int *counter, int all) {
+
+  int local;
+
+  asm volatile("amoadd.w %0, %2, (%1)\n"
+               : "=r"(local)
+               : "r"(counter), "r"(1)
+               : "memory");
+
+  if (local + 1 < all) {
+    do {
+      asm volatile("lw %0, (%1)\n" : "=r"(local) : "r"(counter) : "memory");
+    } while (local < all);
+  }
+}
+
+#endif
\ No newline at end of file
diff --git a/kernel/syscall.c b/kernel/syscall.c
index a7a9b1d..2d666d6 100644
--- a/kernel/syscall.c
+++ b/kernel/syscall.c
@@ -30,9 +30,10 @@ ssize_t sys_user_print(const char* buf, size_t n) {
 // implement the SYS_user_exit syscall
 //
 ssize_t sys_user_exit(uint64 code) {
-  sprint("User exit with code:%d.\n", code);
+  sprint("hartid = ?: User exit with code:%d.\n", code);
   // in lab1, PKE considers only one app (one process). 
   // therefore, shutdown the system when the app calls exit()
+  sprint("hartid = ?: shutdown with code:%d.\n", code);
   shutdown(code);
 }
 
@@ -45,7 +46,7 @@ uint64 sys_user_allocate_page() {
   g_ufree_page += PGSIZE;
   user_vm_map((pagetable_t)current->pagetable, va, PGSIZE, (uint64)pa,
          prot_to_type(PROT_WRITE | PROT_READ, 1));
-
+  sprint("hartid = ?: vaddr 0x%x is mapped to paddr 0x%x\n", va, pa);
   return va;
 }
 
diff --git a/user/app_alloc0.c b/user/app_alloc0.c
new file mode 100644
index 0000000..6e491c9
--- /dev/null
+++ b/user/app_alloc0.c
@@ -0,0 +1,24 @@
+#include "user_lib.h"
+#include "util/types.h"
+
+#define N 5
+#define BASE 0
+
+int main(void) {
+  void *p[N];
+
+  for (int i = 0; i < N; i++) {
+    p[i] = naive_malloc();
+    int *pi = p[i];
+    *pi = BASE + i;
+    printu("=== user alloc 0 @ vaddr 0x%x\n", p[i]);
+  }
+
+  for (int i = 0; i < N; i++) {
+    int *pi = p[i];
+    printu("=== user0: %d\n", *pi);
+    naive_free(p[i]);
+  }
+
+  exit(0);
+}
\ No newline at end of file
diff --git a/user/app_alloc1.c b/user/app_alloc1.c
new file mode 100644
index 0000000..440765f
--- /dev/null
+++ b/user/app_alloc1.c
@@ -0,0 +1,24 @@
+#include "user_lib.h"
+#include "util/types.h"
+
+#define N 5
+#define BASE 5
+
+int main(void) {
+  void *p[N];
+
+  for (int i = 0; i < N; i++) {
+    p[i] = naive_malloc();
+    int *pi = p[i];
+    *pi = BASE + i;
+    printu(">>> user alloc 1 @ vaddr 0x%x\n", p[i]);
+  }
+
+  for (int i = 0; i < N; i++) {
+    int *pi = p[i];
+    printu(">>> user 1: %d\n", *pi);
+    naive_free(p[i]);
+  }
+
+  exit(0);
+}
\ No newline at end of file
diff --git a/user/app_sum_sequence.c b/user/app_sum_sequence.c
deleted file mode 100644
index 809d341..0000000
--- a/user/app_sum_sequence.c
+++ /dev/null
@@ -1,28 +0,0 @@
-/*
- * The application of lab2_3.
- */
-
-#include "user_lib.h"
-#include "util/types.h"
-
-//
-// compute the summation of an arithmetic sequence. for a given "n", compute
-// result = n + (n-1) + (n-2) + ... + 0
-// sum_sequence() calls itself recursively till 0. The recursive call, however,
-// may consume more memory (from stack) than a physical 4KB page, leading to a page fault.
-// PKE kernel needs to improved to handle such page fault by expanding the stack.
-//
-uint64 sum_sequence(uint64 n) {
-  if (n == 0)
-    return 0;
-  else
-    return sum_sequence( n-1 ) + n;
-}
-
-int main(void) {
-  // we need a large enough "n" to trigger pagefaults in the user stack
-  uint64 n = 1000;
-
-  printu("Summation of an arithmetic sequence from 0 to %ld is: %ld \n", n, sum_sequence(1000) );
-  exit(0);
-}