MSP430FR2433: TimerA --> to reduce the time drift

I assume that ACLK is sourced from a 32KHz crystal. The first step is to make sure that crystal meets your accuracy requirements. Your error appears to exceed even the worst (100PPM) crystals.

The timer is going to generate interrupts on schedule. Unless you disable interrupts for a very long time, this should increment your counter just fine. Hopefully you have MCLK set to a high speed so the CPU can get its work done in a timely fashion.

I'm not quite sure where you got 3285 from (I would expect 3276 or 3277).

That said, getting a factor of 10 (really a factor of 5) out of a 2^15-Hz clock is a bit of a pain. What you have going for you is that the error is completely predictable, due to the fraction part of 32768/10.

One solution is to distribute that error over your period. One formulation: (2*3276=6552 ticks)+(8*3277=26216 ticks) = 32768 ticks. In code terms: start with CCR0=3276-1; when your 0.1sec counter counts to 2, set CCR0=3277-1; when that counter counts to 10 (wraps to 0) reset CCR0=3276-1.

With this, your 0.1 sec counter would jitter (.03%), but your 1 sec counter would be correct.

Actually, Previously i have tried with  CCR0=3276-1 this value but at that time i got more time drift.

to gave ACLK as Source, are we need to give any pin selection instructions like, ACLK is a secondary function of P2.2, so are we need to give instruction like below
P2SEL0=BIT2;
P2SEL1=BIT2;
P2DIR=BIT2;

You could of course output ACLK to a pin that way. But where does it come from?

I suspect that you are using the internal oscillator. While good, it isn't nearly as accurate as a crystal.

are we need to use external crystal MSP430FR2433 EVK. are we not have any internal crystal. if we have how to configure this to ACLK

If you're using the FR2433 Launchpad, there is a crystal, but you'll have to do a bit of soldering.

Once the crystal is physically connected, you'll need some software to enable it. See Example msp430fr243x_CS_04.c (lines 79-94):

https://dev.ti.com/tirex/explore/node?node=AGi0Jkf6165naoMtm7ibJg__IOGqZri__LATEST

If we did soldering can't we use those as normal gpios ever?

i don't want to use external crystal, then what is solution like to use internal crystal

You're probably using the REFO (32kHz) oscillator now, which is only rated to +/-3.5% (>2 min/hr). The real problem is that without a "known good" reference you can't make adjustments (calibrate) to compensate.

If you're using this board in a lab (more or less constant temperature/voltage), you have the option to output ACLK to P2.2 (corrected sequence "P2SEL1 |= BIT2; P2DIR |= BIT2;") and use an oscilloscope to measure ACLK and adjust your CCR0 value.

Thank you Bruce for your clear explanation. 

So by your explanation can i conclude like REFO is an internal Oscillator.

If you connect the crystal (solder-bridge R2/R3) and (a) remove R4/R5, you can't use the header pins any more (b) don't remove R4/R5, I suspect you can't use the header pins since the crystal circuit will interfere (but ask a EE about this one).