Re: [Lightning] On the benefits of inlining

[Ludovic Courtès]

Paolo Bonzini <address@hidden> writes:

> That's why a register would be preferable.  Patching is about as 
> expensive as writing one instruction, or only slightly less.  That is, 
> jit_movi + jit_patch_movi is about as expensive as writing 2 movs.

I see.

> Yes... I have to think about it, I must admit I don't like the idea of 
> providing asm statements within lightning.h.  Even though the way you'd 
> be using it is indeed cute.

The most rudimentary support needed would be macros for
registers/instructions, like:

  #define NOJIT_V0  "eax"
  ...

This doesn't imply providing asm statements.

"Higher-level" support could consist in macros like `nojit_movi',
`nojit_jmp', etc.  This would make Lightning very closely tight to GCC.

Thanks,
Ludovic.