diff --git a/src/arch.rs b/src/arch.rs
index 33b3811..140d079 100644
--- a/src/arch.rs
+++ b/src/arch.rs
@@ -70,11 +70,30 @@ mod aarch64 {
 #[cfg(target_arch = "aarch64")]
 pub use aarch64::*;
 
+#[cfg(target_arch = "mips")]
+mod mips {
+    use gimli::{Register, MIPS};
+
+    pub struct Arch;
+
+    #[allow(unused)]
+    impl Arch {
+        pub const SP: Register = MIPS::SP;
+        pub const RA: Register = MIPS::RA;
+
+        pub const UNWIND_DATA_REG: (Register, Register) = (MIPS::A0, MIPS::A1);
+        pub const UNWIND_PRIVATE_DATA_SIZE: usize = 2;
+    }
+}
+#[cfg(target_arch = "mips")]
+pub use mips::*;
+
 #[cfg(not(any(
     target_arch = "x86_64",
     target_arch = "x86",
     target_arch = "riscv64",
     target_arch = "riscv32",
-    target_arch = "aarch64"
+    target_arch = "aarch64",
+    target_arch = "mips",
 )))]
 compile_error!("Current architecture is not supported");
diff --git a/src/lib.rs b/src/lib.rs
index 23e776e..e6de373 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -2,6 +2,7 @@
 #![feature(c_unwind)]
 #![feature(naked_functions)]
 #![feature(non_exhaustive_omitted_patterns_lint)]
+#![cfg_attr(target_arch = "mips", feature(asm_experimental_arch))]
 // lang_items is an internal feature. `internal_features` lint is added recently
 // so also allow unknown lints to prevent warning in older nightly versions.
 #![cfg_attr(
diff --git a/src/unwinder/arch/mips.rs b/src/unwinder/arch/mips.rs
new file mode 100644
index 0000000..f369641
--- /dev/null
+++ b/src/unwinder/arch/mips.rs
@@ -0,0 +1,289 @@
+use core::arch::asm;
+use core::fmt;
+use core::ops;
+use gimli::{Register, MIPS};
+
+// Match DWARF_FRAME_REGISTERS in libgcc
+pub const MAX_REG_RULES: usize = 188;
+
+#[repr(C)]
+#[derive(Clone, Default)]
+pub struct Context {
+    pub gp: [usize; 32],
+    #[cfg(target_feature = "single-float")]
+    pub fp: [usize; 32],
+}
+
+impl fmt::Debug for Context {
+    fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
+        let mut fmt = fmt.debug_struct("Context");
+        for i in 0..=31 {
+            fmt.field(MIPS::register_name(Register(i as _)).unwrap(), &self.gp[i]);
+        }
+        #[cfg(target_feature = "single-float")]
+        for i in 0..=31 {
+            fmt.field(
+                MIPS::register_name(Register((i + 32) as _)).unwrap(),
+                &self.fp[i],
+            );
+        }
+        fmt.finish()
+    }
+}
+
+impl ops::Index<Register> for Context {
+    type Output = usize;
+
+    fn index(&self, reg: Register) -> &usize {
+        match reg {
+            Register(0..=31) => &self.gp[reg.0 as usize],
+            #[cfg(target_feature = "single-float")]
+            Register(32..=63) => &self.fp[(reg.0 - 32) as usize],
+            _ => unimplemented!(),
+        }
+    }
+}
+
+impl ops::IndexMut<gimli::Register> for Context {
+    fn index_mut(&mut self, reg: Register) -> &mut usize {
+        match reg {
+            Register(0..=31) => &mut self.gp[reg.0 as usize],
+            #[cfg(target_feature = "single-float")]
+            Register(32..=63) => &mut self.fp[(reg.0 - 32) as usize],
+            _ => unimplemented!(),
+        }
+    }
+}
+
+macro_rules! code {
+    (save_gp) => {
+        "
+        sw $zero,   0x00($sp)
+        sw $s0,     0x40($sp)
+        sw $s1,     0x44($sp)
+        sw $s2,     0x48($sp)
+        sw $s3,     0x4C($sp)
+        sw $s4,     0x50($sp)
+        sw $s5,     0x54($sp)
+        sw $s6,     0x58($sp)
+        sw $s7,     0x5C($sp)
+        sw $k0,     0x68($sp)
+        sw $k1,     0x6C($sp)
+        sw $gp,     0x70($sp)
+        sw $t0,     0x74($sp)
+        sw $fp,     0x78($sp)
+        sw $ra,     0x7C($sp)
+        "
+    };
+    (save_fp) => {
+        "
+        swc1 $f0,   0x80($sp)
+        swc1 $f1,   0x84($sp)
+        swc1 $f2,   0x88($sp)
+        swc1 $f3,   0x8C($sp)
+        swc1 $f4,   0x90($sp)
+        swc1 $f5,   0x94($sp)
+        swc1 $f6,   0x98($sp)
+        swc1 $f7,   0x9C($sp)
+        swc1 $f8,   0xA0($sp)
+        swc1 $f9,   0xA4($sp)
+        swc1 $f10,  0xA8($sp)
+        swc1 $f11,  0xAC($sp)
+        swc1 $f12,  0xB0($sp)
+        swc1 $f13,  0xB4($sp)
+        swc1 $f14,  0xB8($sp)
+        swc1 $f15,  0xBC($sp)
+        swc1 $f16,  0xC0($sp)
+        swc1 $f17,  0xC4($sp)
+        swc1 $f18,  0xC8($sp)
+        swc1 $f19,  0xCC($sp)
+        swc1 $f20,  0xD0($sp)
+        swc1 $f21,  0xD4($sp)
+        swc1 $f22,  0xD8($sp)
+        swc1 $f23,  0xDC($sp)
+        swc1 $f24,  0xE0($sp)
+        swc1 $f25,  0xE4($sp)
+        swc1 $f26,  0xE8($sp)
+        swc1 $f27,  0xEC($sp)
+        swc1 $f28,  0xF0($sp)
+        swc1 $f29,  0xF4($sp)
+        swc1 $f30,  0xF8($sp)
+        swc1 $f31,  0xFC($sp)
+        "
+    };
+    (restore_gp) => {
+        "
+        lw $at,     0x04($a0)
+        lw $v0,     0x08($a0)
+        lw $v1,     0x0C($a0)
+        lw $a1,     0x14($a0)
+        lw $a2,     0x18($a0)
+        lw $a3,     0x1C($a0)
+        lw $t0,     0x20($a0)
+        lw $t1,     0x24($a0)
+        lw $t2,     0x28($a0)
+        lw $t3,     0x2C($a0)
+        lw $t4,     0x30($a0)
+        lw $t5,     0x34($a0)
+        lw $t6,     0x38($a0)
+        lw $t7,     0x3C($a0)
+        lw $s0,     0x40($a0)
+        lw $s1,     0x44($a0)
+        lw $s2,     0x48($a0)
+        lw $s3,     0x4C($a0)
+        lw $s4,     0x50($a0)
+        lw $s5,     0x54($a0)
+        lw $s6,     0x58($a0)
+        lw $s7,     0x5C($a0)
+        lw $t8,     0x60($a0)
+        lw $t9,     0x64($a0)
+        lw $k0,     0x68($a0)
+        lw $k1,     0x6C($a0)
+        lw $gp,     0x70($a0)
+        lw $sp,     0x74($a0)
+        lw $fp,     0x78($a0)
+        lw $ra,     0x7C($a0)
+        "
+    };
+    (restore_fp) => {
+        "
+        lwc1 $f0,   0x80($a0)
+        lwc1 $f1,   0x84($a0)
+        lwc1 $f2,   0x88($a0)
+        lwc1 $f3,   0x8C($a0)
+        lwc1 $f4,   0x90($a0)
+        lwc1 $f5,   0x94($a0)
+        lwc1 $f6,   0x98($a0)
+        lwc1 $f7,   0x9C($a0)
+        lwc1 $f8,   0xA0($a0)
+        lwc1 $f9,   0xA4($a0)
+        lwc1 $f10,  0xA8($a0)
+        lwc1 $f11,  0xAC($a0)
+        lwc1 $f12,  0xB0($a0)
+        lwc1 $f13,  0xB4($a0)
+        lwc1 $f14,  0xB8($a0)
+        lwc1 $f15,  0xBC($a0)
+        lwc1 $f16,  0xC0($a0)
+        lwc1 $f17,  0xC4($a0)
+        lwc1 $f18,  0xC8($a0)
+        lwc1 $f19,  0xCC($a0)
+        lwc1 $f20,  0xD0($a0)
+        lwc1 $f21,  0xD4($a0)
+        lwc1 $f22,  0xD8($a0)
+        lwc1 $f23,  0xDC($a0)
+        lwc1 $f24,  0xE0($a0)
+        lwc1 $f25,  0xE4($a0)
+        lwc1 $f26,  0xE8($a0)
+        lwc1 $f27,  0xEC($a0)
+        lwc1 $f28,  0xF0($a0)
+        lwc1 $f29,  0xF4($a0)
+        lwc1 $f30,  0xF8($a0)
+        lwc1 $f31,  0xFC($a0)
+        "
+    };
+}
+
+#[naked]
+pub extern "C-unwind" fn save_context(f: extern "C" fn(&mut Context, *mut ()), ptr: *mut ()) {
+    unsafe {
+        #[cfg(target_feature = "single-float")]
+        asm!(
+            "
+            .set noreorder
+            .set nomacro
+            .set noat
+            move $t0, $sp
+            add $sp, $sp, -0x110
+            sw $ra, 0x100($sp)
+            ",
+            code!(save_gp),
+            code!(save_fp),
+            "
+            move $t9, $a0
+            move $a0, $sp
+            /* jalr must use $t9 in PIE code */
+            jalr $t9
+            nop
+            lw $ra, 0x100($sp)
+            add $sp, $sp, 0x110
+            jr $ra
+            nop
+            .set at
+            .set macro
+            .set reorder
+            ",
+            options(noreturn)
+        );
+        #[cfg(not(target_feature = "single-float"))]
+        asm!(
+            "
+            .set noreorder
+            .set nomacro
+            .set noat
+            move $t0, $sp
+            add $sp, $sp, -0x90
+            sw $ra, 0x80($sp)
+            ",
+            code!(save_gp),
+            "
+            move $t9, $a0
+            move $a0, $sp
+            /* jalr must use $t9 in PIE code */
+            jalr $t9
+            nop
+            lw $ra, 0x80($sp)
+            add $sp, $sp, 0x90
+            jr $ra
+            nop
+            .set at
+            .set macro
+            .set reorder
+            ",
+            options(noreturn)
+        );
+    }
+}
+
+pub unsafe extern "C" fn restore_context(ctx: &Context) -> ! {
+    unsafe {
+        #[cfg(target_feature = "single-float")]
+        asm!(
+            "
+            .set noreorder
+            .set nomacro
+            .set noat
+            ",
+            code!(restore_fp),
+            code!(restore_gp),
+            "
+            lw $a0, 0x10($a0)
+            jr $ra
+            nop
+            .set at
+            .set macro
+            .set reorder
+            ",
+            in("$a0") ctx,
+            options(noreturn)
+        );
+        #[cfg(not(target_feature = "single-float"))]
+        asm!(
+            "
+            .set noreorder
+            .set nomacro
+            .set noat
+            ",
+            code!(restore_gp),
+            "
+            lw $a0, 0x10($a0)
+            jr $ra
+            nop
+            .set at
+            .set macro
+            .set reorder
+            ",
+            in("$a0") ctx,
+            options(noreturn)
+        );
+    }
+}
diff --git a/src/unwinder/arch/mod.rs b/src/unwinder/arch/mod.rs
index 8158353..80e401e 100644
--- a/src/unwinder/arch/mod.rs
+++ b/src/unwinder/arch/mod.rs
@@ -23,11 +23,17 @@ mod aarch64;
 #[cfg(target_arch = "aarch64")]
 pub use aarch64::*;
 
+#[cfg(target_arch = "mips")]
+mod mips;
+#[cfg(target_arch = "mips")]
+pub use mips::*;
+
 #[cfg(not(any(
     target_arch = "x86_64",
     target_arch = "x86",
     target_arch = "riscv64",
     target_arch = "riscv32",
-    target_arch = "aarch64"
+    target_arch = "aarch64",
+    target_arch = "mips",
 )))]
 compile_error!("Current architecture is not supported");