Code breaks when optimization is used (msp430-gcc)

Up until now I've been using Energia for projects that required the use of UART, but I decided to get my hands dirty and just use the msp430-gcc toolchain.  

The below program is simply a test program I made to send a number over UART at 9600.  when I compile it without optimization, everything works just fine("msp430-gcc -mmcu=msp430g2553 -g -o adc adc.c"), but when I use an optimization flag such as -Os, -O2, or -O1 it totally breaks my code.  What am I doing wrong in my code that's causing it to break when I try to do any optimization?

I'm using msp430-gcc version 4.6.3 20120301 

#include <msp430.h> 
#include <stdio.h> //for sprintf
void delay(int dly);
void printnumb();
void UART_Write(char dataq);
int flag=0;

char buffer[4];
int place = 2024;

void main(void) {
    WDTCTL = WDTPW | WDTHOLD;	// Stop watchdog timer
          IE1 |= WDTIE; //enable interrupt
	 BCSCTL1 = CALBC1_16MHZ;
	 DCOCTL = CALDCO_16MHZ;
 P1DIR |=BIT2;
  P1OUT &=BIT2;
  P1SEL |=0b00000110;//page 43 on msp430g2553 datasheet
  P1SEL2|=0b00000110;//page 329  
  UCA0CTL1 |= UCSWRST + UCSSEL_2;
  UCA0BR0=104;
  UCA0BR1=0;
  UCA0MCTL =0b00110001;//modulation control p 431
  UCA0CTL1 &=~UCSWRST;
	 __enable_interrupt();
while(1){
printnumb();
delay(2000);

}

}
#pragma vector=WDT_VECTOR
__interrupt void watchdog_timer(void)
{
flag=1;
}

void delay(int dly){
dly= dly*2;
int cycle=0;
WDTCTL = WDTPW | WDTTMSEL| WDTIS0; 
while(cycle<dly){
if(flag==1){
cycle++;
flag=0;
}
}

WDTCTL = WDTPW | WDTHOLD;
}

void printnumb(){
 sprintf(buffer, "%i", place);
 int cnt = 0;
 while(cnt < 4){
  UART_Write(buffer[cnt]);
  cnt=cnt+1;
 }
 UART_Write('\r');
 UART_Write('\n');
}

void UART_Write(char dataq){
  while((IFG2&UCA0TXIFG)==0){
}
 UCA0TXBUF=dataq;
}