diff --git a/include/unicorn/unicorn.h b/include/unicorn/unicorn.h
index 4619c7c3..5f04ec10 100644
--- a/include/unicorn/unicorn.h
+++ b/include/unicorn/unicorn.h
@@ -624,6 +624,35 @@ uc_err uc_mem_protect(uc_engine *uc, uint64_t address, size_t size, uint32_t per
 UNICORN_EXPORT
 uc_err uc_mem_regions(uc_engine *uc, uc_mem_region **regions, uint32_t *count);
 
+/*
+ Save a copy of the current state's registers
+ This API should be used to efficiently make or update a saved copy of the
+ state's registers.
+
+ @uc: handle returned by uc_open()
+ @buffer: pointer to the region to store the registers in. The first call to
+   this function should pass NULL in this parameter, so a region of the
+   appropriate size for the current architecure can be allocated. Further calls
+   to this function may pass in the return value of previous calls.
+
+ @return a pointer to the region the registers were saved in. If buffer was
+   NULL, this is a newly allocated region, otherwise it is the same as buffer.
+   Any allocation performed by this function must be freed by the user.
+*/
+UNICORN_EXPORT
+void *uc_save_regstate(uc_engine *uc, void *buffer);
+
+/*
+ Restore the current state's registers from a saved copy
+ This API should be used to roll the CPU register state back to a previous
+ state saved by uc_save_regstate().
+
+ @uc: handle returned by uc_open()
+ @buffer: pointer returned by uc_save_regstate()
+*/
+UNICORN_EXPORT
+void uc_restore_regstate(uc_engine *uc, void *buffer);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/qemu/target-arm/unicorn.h b/qemu/target-arm/unicorn.h
index 0b3fb93d..d8a02505 100644
--- a/qemu/target-arm/unicorn.h
+++ b/qemu/target-arm/unicorn.h
@@ -19,4 +19,7 @@ void arm_uc_init(struct uc_struct* uc);
 __attribute__ ((visibility ("default")))
 void arm64_uc_init(struct uc_struct* uc);
 
+extern const int ARM_REGS_STORAGE_SIZE;
+extern const int ARM64_REGS_STORAGE_SIZE;
+
 #endif
diff --git a/qemu/target-arm/unicorn_aarch64.c b/qemu/target-arm/unicorn_aarch64.c
index e2c593a6..bb2610fd 100644
--- a/qemu/target-arm/unicorn_aarch64.c
+++ b/qemu/target-arm/unicorn_aarch64.c
@@ -10,6 +10,8 @@
 #include "uc_priv.h"
 
 
+const int ARM64_REGS_STORAGE_SIZE = offsetof(CPUARMState, tlb_table);
+
 static void arm64_set_pc(struct uc_struct *uc, uint64_t address)
 {
     ((CPUARMState *)uc->current_cpu->env_ptr)->pc = address;
diff --git a/qemu/target-arm/unicorn_arm.c b/qemu/target-arm/unicorn_arm.c
index bdf2222c..706cf7bb 100644
--- a/qemu/target-arm/unicorn_arm.c
+++ b/qemu/target-arm/unicorn_arm.c
@@ -10,6 +10,8 @@
 #include "uc_priv.h"
 
 
+const int ARM_REGS_STORAGE_SIZE = offsetof(CPUARMState, tlb_table);
+
 static void arm_set_pc(struct uc_struct *uc, uint64_t address)
 {
     ((CPUARMState *)uc->current_cpu->env_ptr)->pc = address;
diff --git a/qemu/target-i386/unicorn.c b/qemu/target-i386/unicorn.c
index 004e2a4a..a4d7a640 100644
--- a/qemu/target-i386/unicorn.c
+++ b/qemu/target-i386/unicorn.c
@@ -12,6 +12,8 @@
 #include "uc_priv.h"
 
 
+const int X86_REGS_STORAGE_SIZE = offsetof(CPUX86State, tlb_table);
+
 static void x86_set_pc(struct uc_struct *uc, uint64_t address)
 {
     ((CPUX86State *)uc->current_cpu->env_ptr)->eip = address;
diff --git a/qemu/target-i386/unicorn.h b/qemu/target-i386/unicorn.h
index 37e38b4f..cb292001 100644
--- a/qemu/target-i386/unicorn.h
+++ b/qemu/target-i386/unicorn.h
@@ -12,4 +12,6 @@ void x86_reg_reset(struct uc_struct *uc);
 
 void x86_uc_init(struct uc_struct* uc);
 int x86_uc_machine_init(struct uc_struct *uc);
+
+extern const int X86_REGS_STORAGE_SIZE;
 #endif
diff --git a/qemu/target-m68k/unicorn.c b/qemu/target-m68k/unicorn.c
index d2818cfe..849eb08f 100644
--- a/qemu/target-m68k/unicorn.c
+++ b/qemu/target-m68k/unicorn.c
@@ -10,6 +10,8 @@
 #include "uc_priv.h"
 
 
+const int M68K_REGS_STORAGE_SIZE = offsetof(CPUM68KState, tlb_table);
+
 static void m68k_set_pc(struct uc_struct *uc, uint64_t address)
 {
     ((CPUM68KState *)uc->current_cpu->env_ptr)->pc = address;
diff --git a/qemu/target-m68k/unicorn.h b/qemu/target-m68k/unicorn.h
index 7cf8e21a..59471865 100644
--- a/qemu/target-m68k/unicorn.h
+++ b/qemu/target-m68k/unicorn.h
@@ -12,4 +12,5 @@ void m68k_reg_reset(struct uc_struct *uc);
 
 void m68k_uc_init(struct uc_struct* uc);
 
+extern const int M68K_REGS_STORAGE_SIZE;
 #endif
diff --git a/qemu/target-mips/unicorn.c b/qemu/target-mips/unicorn.c
index 94c1819f..6d84a97b 100644
--- a/qemu/target-mips/unicorn.c
+++ b/qemu/target-mips/unicorn.c
@@ -9,6 +9,14 @@
 #include "unicorn_common.h"
 #include "uc_priv.h"
 
+// prevent the lines from being compiled twice
+#ifdef TARGET_WORDS_BIGENDIAN
+#ifdef TARGET_MIPS64
+const int MIPS64_REGS_STORAGE_SIZE = offsetof(CPUMIPSState, tlb_table);
+#else // MIPS32
+const int MIPS_REGS_STORAGE_SIZE = offsetof(CPUMIPSState, tlb_table);
+#endif
+#endif
 
 static uint64_t mips_mem_redirect(uint64_t address)
 {
diff --git a/qemu/target-mips/unicorn.h b/qemu/target-mips/unicorn.h
index c9639eb8..53b5c32a 100644
--- a/qemu/target-mips/unicorn.h
+++ b/qemu/target-mips/unicorn.h
@@ -15,4 +15,7 @@ void mipsel_uc_init(struct uc_struct* uc);
 void mips64_uc_init(struct uc_struct* uc);
 void mips64el_uc_init(struct uc_struct* uc);
 
+extern const int MIPS_REGS_STORAGE_SIZE;
+extern const int MIPS64_REGS_STORAGE_SIZE;
+
 #endif
diff --git a/qemu/target-sparc/unicorn.c b/qemu/target-sparc/unicorn.c
index a2a3dcaa..5cfb8d2a 100644
--- a/qemu/target-sparc/unicorn.c
+++ b/qemu/target-sparc/unicorn.c
@@ -10,6 +10,8 @@
 #include "uc_priv.h"
 
 
+const int SPARC_REGS_STORAGE_SIZE = offsetof(CPUSPARCState, tlb_table);
+
 static bool sparc_stop_interrupt(int intno)
 {
     switch(intno) {
diff --git a/qemu/target-sparc/unicorn.h b/qemu/target-sparc/unicorn.h
index 53bf1303..2140f286 100644
--- a/qemu/target-sparc/unicorn.h
+++ b/qemu/target-sparc/unicorn.h
@@ -13,4 +13,7 @@ void sparc_reg_reset(struct uc_struct *uc);
 void sparc_uc_init(struct uc_struct* uc);
 void sparc64_uc_init(struct uc_struct* uc);
 
+extern const int SPARC_REGS_STORAGE_SIZE;
+extern const int SPARC64_REGS_STORAGE_SIZE;
+
 #endif
diff --git a/qemu/target-sparc/unicorn64.c b/qemu/target-sparc/unicorn64.c
index 9a748936..b8a96dda 100644
--- a/qemu/target-sparc/unicorn64.c
+++ b/qemu/target-sparc/unicorn64.c
@@ -10,6 +10,8 @@
 #include "uc_priv.h"
 
 
+const int SPARC64_REGS_STORAGE_SIZE = offsetof(CPUSPARCState, tlb_table);
+
 static bool sparc_stop_interrupt(int intno)
 {
     switch(intno) {
diff --git a/uc.c b/uc.c
index d6f26ba4..d9a9282d 100644
--- a/uc.c
+++ b/uc.c
@@ -1160,3 +1160,36 @@ uc_err uc_query(uc_engine *uc, uc_query_type type, size_t *result)
 
     return UC_ERR_OK;
 }
+
+size_t cpu_regs_size(uc_arch arch, uc_mode mode);
+size_t cpu_regs_size(uc_arch arch, uc_mode mode) {
+    // each of these constants is defined by offsetof(CPUXYZState, tlb_table)
+    // tbl_table is the first entry in the CPU_COMMON macro, so it marks the end
+    // of the interesting CPU registers
+    switch (arch) {
+        case UC_ARCH_M68K:  return M68K_REGS_STORAGE_SIZE;
+        case UC_ARCH_X86:   return X86_REGS_STORAGE_SIZE;
+        case UC_ARCH_ARM:   return ARM_REGS_STORAGE_SIZE;
+        case UC_ARCH_ARM64: return ARM64_REGS_STORAGE_SIZE;
+        case UC_ARCH_MIPS:  return mode & UC_MODE_MIPS64 ? MIPS64_REGS_STORAGE_SIZE : MIPS_REGS_STORAGE_SIZE;
+        case UC_ARCH_SPARC: return mode & UC_MODE_SPARC64 ? SPARC64_REGS_STORAGE_SIZE : SPARC_REGS_STORAGE_SIZE;
+        default: return 0;
+    }
+}
+
+UNICORN_EXPORT
+void *uc_save_regstate(uc_engine *uc, void *buffer) {
+    size_t sz = cpu_regs_size(uc->arch, uc->mode);
+    if (!buffer) {
+        buffer = malloc(sz);
+    }
+
+    memcpy(buffer, uc->current_cpu->env_ptr, sz);
+    return buffer;
+}
+
+UNICORN_EXPORT
+void uc_restore_regstate(uc_engine *uc, void *buffer) {
+    size_t sz = cpu_regs_size(uc->arch, uc->mode);
+    memcpy(uc->current_cpu->env_ptr, buffer, sz);
+}