On 05/20/2011 11:12 AM, Richard Henderson wrote:
> On 05/20/2011 05:39 AM, Kirill Batuzov wrote:
>> + i = (op == INDEX_op_call) ?
>> + (args[0] >> 16) + (args[0] & 0xffff) + 3 :
>> + def->nb_args;
>> + while (i) {
>> + *gen_args = *args;
>> + args++;
>> + gen_args++;
>> + i--;
>> + }
>
> If you use the correct NOP, i.e. nop vs nop[123n], then
> I don't believe you need to compact the arguments like this.
Bah, nevermind. I forgot that we'd have to do something else
odd to replace n-operand operation opcodes with 2-operand movi.
I went back and saw Aurelien did this with a memmove in his
patch; it's probably more efficient to move pieces at a time
to fill in holes, as you do here.
r~
