Problem with Compiler Optimization

Hi,

I am designing a Hardware Accelerator for IDCT which requires 17 inputs. I am writing into the 17 registers with the last register a control register which starts the hardware accelerator FSM. Immediately I am reading the computed IDCT values from the registers. But for some reason (probably due to compiler optimization) the reads are going to the hardware accelerator first and then the writes are going. Because of this I am getting the wrong values. The compiler optimization level I am using is O0. Below is the part of code used.

//Writing

*(volatile int *)(IDCT_BASE + 0) = 0x17;

*(volatile int *)(IDCT_BASE + 4) = 0x17;

*(volatile int *)(IDCT_BASE + 8) = 0x17;

*(volatile int *)(IDCT_BASE + 0xc) = 0x17;

*(volatile int *)(IDCT_BASE + 0x10) = 0x17;

*(volatile int *)(IDCT_BASE + 0x14) = 0x17;

*(volatile int *)(IDCT_BASE + 0x18) = 0x17;

*(volatile int *)(IDCT_BASE + 0x1c) = 0x17;

*(volatile int *)(IDCT_BASE + 0x20) = 0x17;

*(volatile int *)(IDCT_BASE + 0x24) = 0x17;

*(volatile int *)(IDCT_BASE + 0x28) = 0x17;

*(volatile int *)(IDCT_BASE + 0x2c) = 0x17;

*(volatile int *)(IDCT_BASE + 0x30) = 0x17;

*(volatile int *)(IDCT_BASE + 0x34) = 0x17;

*(volatile int *)(IDCT_BASE + 0x38) = 0x17;

*(volatile int *)(IDCT_BASE + 0x3c) = 0x17;

*(volatile int *)(IDCT_BASE + 0x40) = 0x17;

//Reading

d0 = *(volatile int *)(IDCT_BASE + 0);

d1 = *(volatile int *)(IDCT_BASE + 4);

d2 = *(volatile int *)(IDCT_BASE + 8);

d3 = *(volatile int *)(IDCT_BASE + 0xc);

d4 = *(volatile int *)(IDCT_BASE + 0x10);

d5 = *(volatile int *)(IDCT_BASE + 0x14);

d6 = *(volatile int *)(IDCT_BASE + 0x18);

d7 = *(volatile int *)(IDCT_BASE + 0x1c);

The computed values are stored in the first 8 registers.

When I observed it in simulation the reads are going first in burst mode following by the write to the 17th register followed by write to first 8 registers and the write to the next 8 registers.

Can someone help me with this.

Thanks.