AURIX™ Forum Discussions

[User19104]

Hi,

I'm using the hightec compiler, and I'm seeing the following:


For addind a value, the compiler adds a shift left of 3.... why is that?

		offset = (j<<5);
  0x80000C2E 19 EF FC FF  LD.W      d15,[a14]-0x4
  0x80000C32 06 5F        SH        d15,0x5
  0x80000C34 53 1F 40 20  MUL.U     e2,d15,0x1
  0x80000C38 89 E2 68 F9  ST.D      [a14]-0x18,e2
		offset += destLoc;
  0x80000C3C 19 EF E8 FF  LD.W      d15,[a14]-0x18
  0x80000C40 06 3F        SH        d15,0x3
  0x80000C42 19 E2 F4 FF  LD.W      d2,[a14]-0xC
  0x80000C46 42 2F        ADD       d15,d2
  0x80000C48 0B F0 00 28  MOV       e2,d15
  0x80000C4C 89 E2 68 F9  ST.D      [a14]-0x18,e2

this causes the value of offset to be larger than I intend it to be.

[NeMa_4793301]

Can you include more code here - how are offset and destLoc declared?

[User19104]

Hi,

The code:

uint16 password = IfxScuWdt_getSafetyWatchdogPassword();
    uint32 j = 0, i=0;
	 uint64 *dst;
	 uint64 *destLoc = 0;
	 volatile uint32 delay = 0;
	 uint32 offset = 0;
	

	if (((((uint32)dest & 0xff000000) == 0xa0000000) || ((uint32)dest & 0xff000000) == 0x80000000))    // program flash
    {
        destLoc = (uint64 *) (0xa0000000 | (uint32)dest);
    }
    else if (((uint32)dest & 0xff000000) == 0xaf000000)       // data flash
    {
        destLoc = (uint64 *) (0xaf000000 | (uint32)dest);
    }	 

	for (j= 0; j < size/0x20; j++)
    {
		IfxFlash_clearStatus(0);
		offset = (j<<5);
		offset += destLoc;

[NeMa_4793301]

Why use uint64? The AURIX only supports a 32-bit address space.

I'd use an explicit typecast for the line offset += destLoc:

offset += (uint32) destLoc;

[User19104]

Thought about it, but dn't understand why does it matter.

works anyway.

[MoD]

This comes from the pointer arithmetic:
pointer 64 bit -> right shift 3
pointer 32 bit -> right shift 2
pointer 16 bit -> right shift 1
pointer 8 bit -> right shift 0