Hong Zhang <[EMAIL PROTECTED]> writes:

>> I think a minimal core is a feature.  Not minimal in the sense of
>> trying to squeeze out the last op, but minimal in the sense that an
>> operation whose time to completion is going to swamp the overhead of
>> calling out into native code is implemented externally rather than as
>> an op. Any kind of I/O falls into that category, IMO.

> I disagree on this. The IO operations themselves are generally slow,
> period.  On general, I prefer to having a very fast mechanism to call
> native code, which is an important feature in common sense. If it can be
> proved that the native functions are still to slow for IO calls, we can
> use opcode for them.  But I don't see the need for now.

Okay, you just agreed with everything I said, I think.  :)

Exactly, the I/O operations are slow, which means that calling out to
native code will not make a noticeable performance difference, which I
think is a strong indicator that I/O operations should not be ops.

-- 
Russ Allbery ([EMAIL PROTECTED])             <http://www.eyrie.org/~eagle/>

Reply via email to