diff --git a/lib/arm-mes-gcc/setjmp.c b/lib/arm-mes-gcc/setjmp.c
index 5e6d1fb6..147d9066 100644
--- a/lib/arm-mes-gcc/setjmp.c
+++ b/lib/arm-mes-gcc/setjmp.c
@@ -1,6 +1,6 @@
 /* -*-comment-start: "//";comment-end:""-*-
  * GNU Mes --- Maxwell Equations of Software
- * Copyright © 2017,2018,2019 Jan (janneke) Nieuwenhuizen <janneke@gnu.org>
+ * Copyright © 2017,2018,2019,2020 Jan (janneke) Nieuwenhuizen <janneke@gnu.org>
  * Copyright © 2019,2020 Danny Milosavljevic <dannym@scratchpost.org>
  *
  * This file is part of GNU Mes.
@@ -22,6 +22,7 @@
 #include <setjmp.h>
 #include <stdlib.h>
 
+#if !__TINYC__
 void
 __attribute__ ((noinline))
 longjmp (jmp_buf env, int val)
@@ -42,7 +43,31 @@ longjmp (jmp_buf env, int val)
   // *INDENT-ON*
   // not reached
 }
+#else //__TINYC__
+__asm__ (".global longjmp\n");
+__asm__ ("longjmp:\n");
+__asm__ (".int 0xe52db004\n"); //push  {fp}      ; (str fp, [sp, #-4]!)
+__asm__ (".int 0xe28db000\n"); //add   fp, sp, #0
+__asm__ (".int 0xe24dd00c\n"); //sub   sp, sp, #12
+__asm__ (".int 0xe50b0008\n"); //str   r0, [fp, #-8]
+__asm__ (".int 0xe50b100c\n"); //str   r1, [fp, #-12]
+__asm__ (".int 0xe51b3008\n"); //ldr   r3, [fp, #-8]
+__asm__ (".int 0xe51b200c\n"); //ldr   r2, [fp, #-12]
+__asm__ (".int 0xe1a00003\n"); //mov   r0, r3
+__asm__ (".int 0xe1a01002\n"); //mov   r1, r2
+__asm__ (".int 0xe3510000\n"); //cmp   r1, #0
+__asm__ (".int 0x03a01001\n"); //moveq r1, #1
+__asm__ (".int 0xe490d004\n"); //ldr   sp, [r0], #4
+__asm__ (".int 0xe490e004\n"); //ldr   lr, [r0], #4
+__asm__ (".int 0xe8b00ff0\n"); //ldm   r0!, {r4, r5, r6, r7, r8, r9, sl, fp}
+#if HAVE_FLOAT
+// TODO: If using VFP, vldmia r0!, {d8-d15}
+#endif
+__asm__ (".int 0xe1a00001\n"); //mov   r0, r1
+__asm__ (".int 0xe12fff1e\n"); //bx    lr
+#endif //__TINYC__
 
+#if !__TINYC__
 int
 __attribute__ ((noinline))
 setjmp (jmp_buf env)
@@ -60,3 +85,25 @@ setjmp (jmp_buf env)
   // *INDENT-ON*
   return 0;
 }
+#else //__TINYC__
+__asm__ (".global setjmp\n");
+__asm__ ("setjmp:\n");
+__asm__ (".int 0xe52db004\n"); //push  {fp}      ; (str fp, [sp, #-4]!)
+__asm__ (".int 0xe28db000\n"); //add   fp, sp, #0
+__asm__ (".int 0xe24dd00c\n"); //sub   sp, sp, #12
+__asm__ (".int 0xe50b0008\n"); //str   r0, [fp, #-8]
+__asm__ (".int 0xe51b3008\n"); //ldr   r3, [fp, #-8]
+__asm__ (".int 0xe1a00003\n"); //mov   r0, r3
+__asm__ (".int 0xe480d004\n"); //str   sp, [r0], #4
+__asm__ (".int 0xe480e004\n"); //str   lr, [r0], #4
+__asm__ (".int 0xe28bd000\n"); //add   sp, fp, #0
+__asm__ (".int 0xe49db004\n"); //pop   {fp}      ; (ldr fp, [sp], #4)
+__asm__ (".int 0xe8a00ff0\n"); //stmia r0!, {r4, r5, r6, r7, r8, r9, sl, fp}
+#if HAVE_FLOAT
+// TODO: If using VFP, vstmia r0!, {d8-d15}
+#endif
+__asm__ (".int 0xe1a0b00d\n"); //mov   fp, sp
+__asm__ (".int 0xe3a03000\n"); //mov   r3, #0
+__asm__ (".int 0xe1a00003\n"); //mov   r0, r3
+__asm__ (".int 0xe12fff1e\n"); //bx    lr
+#endif //__TINYC__