Re: [PATCH] net: filter: Convert the BPF VM to threaded code

[Eric Dumazet <eric.dumazet@gmail.com>]

Le vendredi 29 juillet 2011 à 01:10 -0700, Rui Ueyama a écrit :
> Convert the BPF VM to threaded code to improve performance.
> 
> The BPF VM is basically a big for loop containing a switch statement.  That is
> slow because for each instruction it checks the for loop condition and does the
> conditional branch of the switch statement.
> 
> This patch eliminates the conditional branch, by replacing it with jump table
> using GCC's labels-as-values feature. The for loop condition check can also be
> removed, because the filter code always end with a RET instruction.
> 

Well...


> +#define NEXT goto *jump_table[(++fentry)->code]
> +
> +	/* Dispatch the first instruction */
> +	goto *jump_table[fentry->code];

This is the killer, as this cannot be predicted by the cpu.

Do you have benchmark results to provide ?

We now have BPF JIT on x86_64 and powerpc, and possibly on MIPS and ARM
on a near future.