diff --git a/bindings/python/unicorn/unicorn.py b/bindings/python/unicorn/unicorn.py
index e2cb6a5d..fa50ad1f 100644
--- a/bindings/python/unicorn/unicorn.py
+++ b/bindings/python/unicorn/unicorn.py
@@ -9,7 +9,7 @@ import os.path
 import sys
 import weakref
 
-from . import x86_const, unicorn_const as uc
+from . import x86_const, arm64_const, unicorn_const as uc
 
 if not hasattr(sys.modules[__name__], "__file__"):
     __file__ = inspect.getfile(inspect.currentframe())
@@ -223,6 +223,13 @@ class uc_x86_xmm(ctypes.Structure):
         ("high_qword", ctypes.c_uint64),
     ]
 
+class uc_arm64_neon128(ctypes.Structure):
+    """128-bit neon register"""
+    _fields_ = [
+        ("low_qword", ctypes.c_uint64),
+        ("high_qword", ctypes.c_uint64),
+    ]
+
 # Subclassing ref to allow property assignment.
 class UcRef(weakref.ref):
     pass
@@ -317,6 +324,14 @@ class Uc(object):
                     raise UcError(status)
                 return reg.value
 
+        if self._arch == uc.UC_ARCH_ARM64:
+            if reg_id in range(arm64_const.UC_ARM64_REG_Q0, arm64_const.UC_ARM64_REG_Q31+1) or range(arm64_const.UC_ARM64_REG_V0, arm64_const.UC_ARM64_REG_V31+1):
+                reg = uc_arm64_neon128()
+                status = _uc.uc_reg_read(self._uch, reg_id, ctypes.byref(reg))
+                if status != uc.UC_ERR_OK:
+                    raise UcError(status)
+                return reg.low_qword | (reg.high_qword << 64)
+
         # read to 64bit number to be safe
         reg = ctypes.c_uint64(0)
         status = _uc.uc_reg_read(self._uch, reg_id, ctypes.byref(reg))
@@ -349,6 +364,12 @@ class Uc(object):
                 reg.rid = value[0]
                 reg.value = value[1]
 
+        if self._arch == uc.UC_ARCH_ARM64:
+            if reg_id in range(arm64_const.UC_ARM64_REG_Q0, arm64_const.UC_ARM64_REG_Q31+1) or range(arm64_const.UC_ARM64_REG_V0, arm64_const.UC_ARM64_REG_V31+1):
+                reg = uc_arm64_neon128()
+                reg.low_qword = value & 0xffffffffffffffff
+                reg.high_qword = value >> 64
+
         if reg is None:
             # convert to 64bit number to be safe
             reg = ctypes.c_uint64(value)
@@ -364,7 +385,7 @@ class Uc(object):
     # write to MSR - X86 only
     def msr_write(self, msr_id, value):
         return self.reg_write(x86_const.UC_X86_REG_MSR, (msr_id, value))
- 
+
     # read data from memory
     def mem_read(self, address, size):
         data = ctypes.create_string_buffer(size)
diff --git a/qemu/target-arm/unicorn_aarch64.c b/qemu/target-arm/unicorn_aarch64.c
index 9955b42e..5b9980f1 100644
--- a/qemu/target-arm/unicorn_aarch64.c
+++ b/qemu/target-arm/unicorn_aarch64.c
@@ -50,10 +50,27 @@ int arm64_reg_read(struct uc_struct *uc, unsigned int *regs, void **vals, int co
     for (i = 0; i < count; i++) {
         unsigned int regid = regs[i];
         void *value = vals[i];
+        // V & Q registers are the same
+        if (regid >= UC_ARM64_REG_V0 && regid <= UC_ARM64_REG_V31) {
+            regid += UC_ARM64_REG_Q0 - UC_ARM64_REG_V0;
+        }
         if (regid >= UC_ARM64_REG_X0 && regid <= UC_ARM64_REG_X28) {
             *(int64_t *)value = ARM_CPU(uc, mycpu)->env.xregs[regid - UC_ARM64_REG_X0];
         } else if (regid >= UC_ARM64_REG_W0 && regid <= UC_ARM64_REG_W30) {
             *(int32_t *)value = READ_DWORD(ARM_CPU(uc, mycpu)->env.xregs[regid - UC_ARM64_REG_W0]);
+        } else if (regid >= UC_ARM64_REG_Q0 && regid <= UC_ARM64_REG_Q31) {
+            float64 *dst = (float64*) value;
+            uint32_t reg_index = 2*(regid - UC_ARM64_REG_Q0);
+            dst[0] = ARM_CPU(uc, mycpu)->env.vfp.regs[reg_index];
+            dst[1] = ARM_CPU(uc, mycpu)->env.vfp.regs[reg_index+1];
+        } else if (regid >= UC_ARM64_REG_D0 && regid <= UC_ARM64_REG_D31) {
+            *(float64*)value = ARM_CPU(uc, mycpu)->env.vfp.regs[2*(regid - UC_ARM64_REG_D0)];
+        } else if (regid >= UC_ARM64_REG_S0 && regid <= UC_ARM64_REG_S31) {
+            *(int32_t*)value = READ_DWORD(ARM_CPU(uc, mycpu)->env.vfp.regs[2*(regid - UC_ARM64_REG_S0)]);
+        } else if (regid >= UC_ARM64_REG_H0 && regid <= UC_ARM64_REG_H31) {
+            *(int16_t*)value = READ_WORD(ARM_CPU(uc, mycpu)->env.vfp.regs[2*(regid - UC_ARM64_REG_H0)]);
+        } else if (regid >= UC_ARM64_REG_B0 && regid <= UC_ARM64_REG_B31) {
+            *(int8_t*)value = READ_BYTE_L(ARM_CPU(uc, mycpu)->env.vfp.regs[2*(regid - UC_ARM64_REG_B0)]);
         } else {
             switch(regid) {
                 default: break;
@@ -84,10 +101,26 @@ int arm64_reg_write(struct uc_struct *uc, unsigned int *regs, void* const* vals,
     for (i = 0; i < count; i++) {
         unsigned int regid = regs[i];
         const void *value = vals[i];
+        if (regid >= UC_ARM64_REG_V0 && regid <= UC_ARM64_REG_V31) {
+            regid += UC_ARM64_REG_Q0 - UC_ARM64_REG_V0;
+        }
         if (regid >= UC_ARM64_REG_X0 && regid <= UC_ARM64_REG_X28) {
             ARM_CPU(uc, mycpu)->env.xregs[regid - UC_ARM64_REG_X0] = *(uint64_t *)value;
         } else if (regid >= UC_ARM64_REG_W0 && regid <= UC_ARM64_REG_W30) {
             WRITE_DWORD(ARM_CPU(uc, mycpu)->env.xregs[regid - UC_ARM64_REG_W0], *(uint32_t *)value);
+        } else if (regid >= UC_ARM64_REG_Q0 && regid <= UC_ARM64_REG_Q31) {
+            float64 *src = (float64*) value;
+            uint32_t reg_index = 2*(regid - UC_ARM64_REG_Q0);
+            ARM_CPU(uc, mycpu)->env.vfp.regs[reg_index] = src[0];
+            ARM_CPU(uc, mycpu)->env.vfp.regs[reg_index+1] = src[1];
+        } else if (regid >= UC_ARM64_REG_D0 && regid <= UC_ARM64_REG_D31) {
+            ARM_CPU(uc, mycpu)->env.vfp.regs[2*(regid - UC_ARM64_REG_D0)] = * (float64*) value;
+        } else if (regid >= UC_ARM64_REG_S0 && regid <= UC_ARM64_REG_S31) {
+            WRITE_DWORD(ARM_CPU(uc, mycpu)->env.vfp.regs[2*(regid - UC_ARM64_REG_S0)], *(int32_t*) value);
+        } else if (regid >= UC_ARM64_REG_H0 && regid <= UC_ARM64_REG_H31) {
+            WRITE_WORD(ARM_CPU(uc, mycpu)->env.vfp.regs[2*(regid - UC_ARM64_REG_H0)], *(int16_t*) value);
+        } else if (regid >= UC_ARM64_REG_B0 && regid <= UC_ARM64_REG_B31) {
+            WRITE_BYTE_L(ARM_CPU(uc, mycpu)->env.vfp.regs[2*(regid - UC_ARM64_REG_B0)], *(int8_t*) value);
         } else {
             switch(regid) {
                 default: break;