diff --git a/arm-gen.c b/arm-gen.c
index 81fa185a..b132e3e3 100644
--- a/arm-gen.c
+++ b/arm-gen.c
@@ -1396,6 +1396,20 @@ void gfunc_call(int nb_args)
   float_abi = def_float_abi;
 }
 
+void tcc_run_start(int (*prog_main)(int, char **, char **), int cnt, char **var)
+{
+#ifdef __arm__
+    void *sp;
+
+    asm("sub sp, sp, %1\n"
+        "\tmov %0, sp"
+        : "=r" (sp)
+        : "r" ((((size_t) cnt + 1) & -2) * sizeof(char *)));
+    memcpy(sp, var, cnt * sizeof(char *));
+    asm("mov pc, %0" : : "r" (prog_main));
+#endif
+}
+
 /* generate function prolog of type 't' */
 void gfunc_prolog(Sym *func_sym)
 {
diff --git a/arm64-gen.c b/arm64-gen.c
index 03ae56c3..8e7e2f36 100644
--- a/arm64-gen.c
+++ b/arm64-gen.c
@@ -1167,6 +1167,25 @@ ST_FUNC void gfunc_call(int nb_args)
     tcc_free(t);
 }
 
+void tcc_run_start(int (*prog_main)(int, char **, char **), int cnt, char **var)
+{
+#if defined(__aarch64__)
+#if defined(__TINYC__)
+    // FIXME: immplement arm64 assembler
+    fprintf(stderr, "tcc -nostdlib -run not implement for arm64\n");
+#else
+    void *sp;
+
+    asm("sub sp, sp, %1\n"
+        "\tmov %0, sp"
+        : "=r" (sp)
+        : "r" ((((size_t) cnt + 1) & -2) * sizeof(char *)));
+    memcpy(sp, var, cnt * sizeof(char *));
+    asm("br %0" : : "r" (prog_main));
+#endif
+#endif
+}
+
 static unsigned long arm64_func_va_list_stack;
 static int arm64_func_va_list_gr_offs;
 static int arm64_func_va_list_vr_offs;
diff --git a/i386-gen.c b/i386-gen.c
index c8d8af53..b9da31cd 100644
--- a/i386-gen.c
+++ b/i386-gen.c
@@ -504,6 +504,24 @@ ST_FUNC void gfunc_call(int nb_args)
     vtop--;
 }
 
+void tcc_run_start(int (*prog_main)(int, char **, char **), int cnt, char **var)
+{
+#ifdef __i386__
+#ifdef TCC_TARGET_PE
+    fprintf(stderr, "tcc -nostdlib -run not implement for TCC_TARGET_PE\n");
+#else
+    void *sp;
+
+    asm("sub %1, %%esp\n"
+        "\tmov %%esp, %0"
+        : "=r" (sp)
+        : "r" ((((size_t) cnt + 1) & -2) * sizeof(char *)));
+    memcpy(sp, var, cnt * sizeof(char *));
+    asm("jmp *%0" : : "r" (prog_main));
+#endif
+#endif
+}
+
 #ifdef TCC_TARGET_PE
 #define FUNC_PROLOG_SIZE (10 + USE_EBX)
 #else
diff --git a/riscv64-gen.c b/riscv64-gen.c
index c0db5613..2226802c 100644
--- a/riscv64-gen.c
+++ b/riscv64-gen.c
@@ -769,6 +769,20 @@ done:
    tcc_free(info);
 }
 
+void tcc_run_start(int (*prog_main)(int, char **, char **), int cnt, char **var)
+{
+#ifdef __riscv
+    void *sp;
+
+    asm("sub sp, sp, %1\n"
+        "\tmv %0, sp"
+        : "=r" (sp)
+        : "r" ((((size_t) cnt + 1) & -2) * sizeof(char *)));
+    memcpy(sp, var, cnt * sizeof(char *));
+    asm("jalr %0" : : "r" (prog_main));
+#endif
+}
+
 static int func_sub_sp_offset, num_va_regs, func_va_list_ofs;
 
 ST_FUNC void gfunc_prolog(Sym *func_sym)
diff --git a/tcc.h b/tcc.h
index 410a5d8f..310409df 100644
--- a/tcc.h
+++ b/tcc.h
@@ -1797,6 +1797,7 @@ ST_FUNC const char *dlerror(void);
 ST_FUNC void *dlsym(void *handle, const char *symbol);
 #endif
 ST_FUNC void tcc_run_free(TCCState *s1);
+ST_FUNC void tcc_run_start(int (*prog_main)(int, char **, char **), int cnt, char **var);
 #endif
 
 /* ------------ tcctools.c ----------------- */
diff --git a/tccrun.c b/tccrun.c
index 8ef0b659..8e9da653 100644
--- a/tccrun.c
+++ b/tccrun.c
@@ -249,8 +249,30 @@ LIBTCCAPI int tcc_run(TCCState *s1, int argc, char **argv)
     fflush(stderr);
 
     ret = tcc_setjmp(s1, main_jb, tcc_get_symbol(s1, top_sym));
-    if (0 == ret)
-        ret = prog_main(argc, argv, envp);
+    if (0 == ret) {
+        if (s1->nostdlib) {
+	    int n = 1;
+	    char **p, **e = envp;
+
+	    /* create sysv memory layout: argc, argv[], NULL, envp[], NULL */
+	    if (envp)
+	        while (*e++)
+		    n++;
+	    p = tcc_malloc((argc + n + 2) * sizeof(char **));
+	    p[0] = (char *) (size_t) argc;
+	    memcpy(p + 1, argv, argc * sizeof(char *));
+	    p[argc + 1] = NULL;
+	    if (envp)
+	        memcpy(p + argc + 2, envp, n * sizeof(char *));
+	    else
+	        p[argc + 2] = NULL;
+	    /* Probably never returns */
+	    tcc_run_start(prog_main, argc + n + 2, p);
+	    tcc_free(p);
+	}
+	else
+            ret = prog_main(argc, argv, envp);
+    }
     else if (RT_EXIT_ZERO == ret)
         ret = 0;
 
diff --git a/x86_64-gen.c b/x86_64-gen.c
index c679db28..d3b07bb9 100644
--- a/x86_64-gen.c
+++ b/x86_64-gen.c
@@ -933,6 +933,10 @@ void gfunc_call(int nb_args)
     vtop--;
 }
 
+void tcc_run_start(int (*prog_main)(int, char **, char **), int cnt, char **var)
+{
+    fprintf(stderr, "tcc -nostdlib -run not implement for TCC_TARGET_PE\n");
+}
 
 #define FUNC_PROLOG_SIZE 11
 
@@ -1434,6 +1438,20 @@ void gfunc_call(int nb_args)
     vtop--;
 }
 
+void tcc_run_start(int (*prog_main)(int, char **, char **), int cnt, char **var)
+{
+#ifdef __x86_64__
+    void *sp;
+
+    asm("subq %1, %%rsp\n"
+	"\tmovq %%rsp, %0"
+	: "=r" (sp)
+	: "r" ((((size_t) cnt + 1) & -2) * sizeof(char *)));
+    memcpy(sp, var, cnt * sizeof(char *));
+    asm("jmp *%0" : : "r" (prog_main));
+#endif
+}
+
 #define FUNC_PROLOG_SIZE 11
 
 static void push_arg_reg(int i) {