Compiler/CC2541: Wrong Output of the delay function

Part Number: CC2541

Tool/software: TI C/C++ Compiler

About the delay function, the output unit is microsecond,  it could not output the right time. Following is detail code:

/**************************************************************************************************
 * @fn      HalLcd_HW_WaitUs
 *
 * @brief   wait for x us. @ 32MHz MCU clock it takes 32 "nop"s for 1 us delay.
 *
 * @param   x us. range[0-65536]
 *
 * @return  None
 **************************************************************************************************/
void HalLcd_HW_WaitUs(uint16 microSecs)
{
  while(microSecs--)
  {
    /* 32 NOPs == 1 usecs */
    asm("nop"); asm("nop"); asm("nop"); asm("nop"); asm("nop");
    asm("nop"); asm("nop"); asm("nop"); asm("nop"); asm("nop");
    asm("nop"); asm("nop"); asm("nop"); asm("nop"); asm("nop");
    asm("nop"); asm("nop"); asm("nop"); asm("nop"); asm("nop");
    asm("nop"); asm("nop"); asm("nop"); asm("nop"); asm("nop");
    asm("nop"); asm("nop"); asm("nop"); asm("nop"); asm("nop");
    asm("nop"); asm("nop");
  }
}

The examination method is using  Oscilloscope to check the high level duration of P0_1. Following is detail code:

P0_1 = 1;

HalLcd_HW_WaitUs(1);

P0_1 = 0;

The actual result is around 580us--620us.

Later the duration is around 600us even I did not execute  Function HalLcd_HW_WaitUs() and just execute P0_1 = 1; and P0_1 = 0，The clock = 32MHz。

I doubted the BLE Statck  affected the execute time, so I try to use the demo in ..CC2541_43_44_45_Peripherals_Software_Examples\software_examples\io to just set time clock and control P0_1, the result is the same, the time cost is above 500us by just operating GPIO！

int main(void)
{
    /***************************************************************************
     * Setup I/O
     *
     */
    // Disable the 32 kHz RCOSC calibration
    SLEEPCMD |= OSC32K_CALDIS;
  
    // Change the system clock source to HS XOSC and set the clock speed to 32 MHz.
    CLKCONCMD = (CLKCONCMD & ~(CLKCON_CLKSPD | CLKCON_OSC)) | CLKCON_CLKSPD_32M;

    // Wait until system clock source has changed to HS XOSC (CLKCONSTA.OSC = 0).
    while(CLKCONSTA & CLKCON_OSC);
    
#if(chip==2541 || chip==2543 || chip==2545)

    // Initialize P0_1 for SRF05EB S1 button.
    P0SEL &= ~BIT1;           // Function as General Purpose I/O.
    P0DIR |= BIT1;            // Input.
    //P0INP |= BIT1;            // Select input type as 3-state, S1 has external pulldown.
    P0_1 = 0;                 // LED1 on.
#endif    
    
    /***************************************************************************
     * Main loop
     *
     */
  
    while(1)
    {
        P0_1 = 1;
        
        P0_1 = 0;
    }
}

Could someone tell me why?

Compiler Environmen: IAR