Strange timing behavior in my code

I write two codes for generating greq at PA0-Pa7: Diff in code 1 & code 2 is that only BSSR register is moved to define macro.

code1

#define WRITE_TFT_DATA_PINS(VALUE)                     {                                                                 \ 
                                                            GPIOA->BSRRL = (uint16_t)( VALUE & 0x000000ff);              \ 
                                                            GPIOA->BSRRH = (uint16_t)(~VALUE & 0x000000ff);              \ 
                                                       }

void main()
{
  for( i = 8; i < 255; i++ )
  {
    WRITE_TFT_DATA_PINS(i);
  }
}

code2

void main()
{
  for( i = 8; i < 255; i++ )
  {
    GPIOA->BSRRL = (uint16_t)( VALUE & 0x000000ff);
    GPIOA->BSRRH = (uint16_t)(~VALUE & 0x000000ff);
  }
}

Now on checking freq on PA7 pin for code 1 = 55Khz & for code 2 = 65Khz (approx.)

I checked the preprocessor file of both code, both are almost same
code1

    while(1)
    {
        uint32_t i;
        for( i = 8; i < 255; i++ )
        {
            { ((GPIO_TypeDef *) ((((uint32_t)0x40000000) + 0x00020000) + 0x0000))->BSRRL = (uint16_t)( i & 0x000000ff); (GPIO_TypeDef *) ((((uint32_t)0x40000000) + 0x00020000) + 0x0000))->BSRRH = (uint16_t)(~i & 0x000000ff); };


        }
    }

code2

    while(1)
    {
        uint32_t i;
        for( i = 8; i < 255; i++ )
        {

             ((GPIO_TypeDef *) ((((uint32_t)0x40000000) + 0x00020000) + 0x0000))->BSRRL = (uint16_t)( i & 0x000000ff);
             ((GPIO_TypeDef *) ((((uint32_t)0x40000000) + 0x00020000) + 0x0000))->BSRRH = (uint16_t)(~i & 0x000000ff);
        }
    }

Then i checked its disassembly , after for loop:

code1

0x0800020E 2008      MOVS     r0,#0x08
0x08000210 B2C2      UXTB     r2,r0
0x08000212 830A      STRH     r2,[r1,#0x18]
0x08000214 43C2      MVNS     r2,r0
0x08000216 B2D2      UXTB     r2,r2
0x08000218 834A      STRH     r2,[r1,#0x1A]
0x0800021A 1C40      ADDS     r0,r0,#1
0x0800021C 28FF      CMP      r0,#0xFF
0x0800021E D3F7      BCC      0x08000210
0x08000220 E7F5      B        0x0800020E

code2

0x0800021A 2109      MOVS     r1,#0x09
0x0800021C B2CA      UXTB     r2,r1
0x0800021E 8302      STRH     r2,[r0,#0x18]
0x08000220 43CA      MVNS     r2,r1
0x08000222 B2D2      UXTB     r2,r2
0x08000224 8342      STRH     r2,[r0,#0x1A]
0x08000226 1C4A      ADDS     r2,r1,#1
0x08000228 B2D4      UXTB     r4,r2
0x0800022A 8304      STRH     r4,[r0,#0x18]
0x0800022C 43D2      MVNS     r2,r2
0x0800022E B2D2      UXTB     r2,r2
0x08000230 8342      STRH     r2,[r0,#0x1A]
0x08000232 1C89      ADDS     r1,r1,#2
0x08000234 29FF      CMP      r1,#0xFF
0x08000236 D3F1      BCC      0x0800021C
0x08000238 E7EC      B        0x08000214

My question is when both coeds are same except define which I don't think should change the code assembly.
Also optimzation and other parameter are same for both