when using an A-T cut crystal,a drive limiting resistor MAY be need between osc2 and one pin of crystal to get the specified frequecy, but it doesn't suggest a typical resistence value for a 16M crystal.and i also wander even if found the correct value resistor and connect it,will the system work?

Meanwhile, I have had a look at datasheet. Oh dear, what a shitty chapter about oscillator! Is feedback resistor of about 2.2 to 10MOhm internally connected, or not? What about the recommended capacitors? Very, very bad datasheet, a shame for microchip ... Did you already contact manufacturer for asking for details?

Have you read my reply in thread 'external oscillator drive'? Here, I explained in detail, that Pierce oscillator NEEDS some resistance at output of inverter! If output transistors of inverter (PMOS/NMOS) show very low drain source channel impedance (in the 50 Ohm range), then, an additional resistance is needed for achieving necessary phase lag AND for limiting drive level.
If manufacturer of chip recommends additional resistor, that 'MAY be needed', then very probably it IS needed!!

Ok, let's solve problem. Add in any case 2.2MOhm between both oscillator pins, as long as we do not surely know, whether it's internally connected or not. Unprobably, but how knows...

Use 30pF capacitors at both oscillator pins to ground. Many other manufacturers recommend something in this range.
Also, add this series drive limiting and phase lag introducing resistor. But you must find out correct range of value by experiment: Make conditions such that oscillator will have problems to start and increase resistor up to the moment, when oscillator is no longer able to start. Then take the half of this resistor value.

So, first start with 50 Ohm. Cool down your chip to the lowest ambient temperature your circuit will see in application, then transconductance of inverter (something like gain) heavily decreases and start-up of oscillator and maintaining of oscillation is most difficult. Power-on your application and find out, whether oscillator has started. Don't connect anything to oscillator parts, like oscilloscope, etc. Added stray capacitance would cause wrong results!

If oscillator has started, try a bigger resistance, like 100 Ohm e.g. Again power-on your application and find out whether oscillator has started.

Repeat this game up to the value, where oscillator is not longer able to start. Then, take the half (or third) of this resistor. This gives you the optimum compromise between lowest drive level and highest start-up guarantee. Your series resistor will be in the range of 100...500Ohm, I assume.

Of course, it's possible to measure crystal current directly by HF current probe. But that's a job for a master. Above procedure may sound amateurish to some people, but it's just, what many manufacturers suggest, not necessarily in datasheet, but when you have contact to an application engineer.

Good luck,
Kai