Re: [Lightning] JIT_R11 trashed by jit_callr on MIPS32

[Paulo César Pereira de Andrade]

Em 27/10/2014 17:52, "Paul Cercueil" <address@hidden> escreveu:
>
> Hi,

  Hi,

> I've been working for some time now on my spare time on a MIPS-to-Lightning dynamic recompiler: https://github.com/pcercuei/lightrec/ 

  It looks very nice. On a quick look I only suggest not using
JIT_RA0, as you should be aware it is not available on all ports,
on ix86 parameters are passes on stack, and on ia64 the first
register argument has a variable value.

> It abuses Lightning a little bit, in the way that it doesn't use functions - the generated code directly jumps to the next block of code, recompiling it if needed. It works fine on MIPS32 for now, except for one point: jit_callr trashes the JIT_R11 register...

  I am afraid this is expected :(

> Here's a minimal test case:
>
> <paul:~/dev/gcw0/lightrec> $ cat minimal.c
> #include <lightning.h>
>
> static void get_new_function(void) { return &end_execution; }
>
> int main(int argc, char **argv)
> {
>         jit_state_t *_jit;
>         init_jit(argv[0]);
>
>         _jit = jit_new_state();
>
>         jit_movi(JIT_R11, 54);
>         jit_movi(JIT_R0, &get_new_function);
>         jit_callr(JIT_R0);
>
>         jit_retval(JIT_R0);
>         jit_addr(JIT_R0, JIT_R0, JIT_R11);
>         jit_jmpr(JIT_R0);
>
>         jit_emit();
>         jit_disassemble();
>         jit_clear_state();
>         jit_destroy_state();
>         finish_jit();
>
>         return 0;
> }
>
> The code basically sets the JIT_R11 register to 54, call "get_new_function", adds JIT_R11 to the value returned and jumps to that new address.
>
> This is what I obtain:
>
> opendingux:/media/data/local/home # ./minimal
>         0x77295000      li      t9,54
>         0x77295004      lui     v0,0x40
>         0x77295008      ori     v0,v0,0x9d0
>         0x7729500c      move    t9,v0
>         0x77295010      jalr    v0
>         0x77295014      nop
>         0x77295018      addu    v0,v0,t9
>         0x7729501c      jr      v0
>         0x77295020      nop
>
> The LUI/ORI correspond the load of the address of get_new_function. The jalr is the jit_callr. But between those, the JIT_R11 (register $t9 here) is trashed for no good reason.
>
> My current workaround is to completely avoid the use of JIT_R11, but it would be great if it could be fixed :).

  Remember that JIT_R11 is not callee save, so, lightning is
supposed to be allowed to clobber it. A backend specific comment
should be added to the documentation about such special cases.

  t9 is the pic register. Lightning does not know if it is calling a
pic C function so if the jit_callr argument is not t9, it will set it to
the function pointer. A workaround in your code would be to
know about that.

> Thanks,
>
> Paul
>

Thanks,
Paulo