Dear friends of the Freescale Forum,
Although I am grateful for the intelligent tools that Freescale has developed for compiling very efficient code, sometimes I am frustrated when the intelligence prevents me from executing redundant tasks like testing my static ram array attched to the mini-flex bus.
During the development stage I discovered that if compiling under global optimization of 4 the intelligence would prevent me from taking a value, writing to memory and reading it back again.
It would ignore the code altogether because I guess it felt it was reduntant.
To circumvent the intelligence I wrote the following code:
uint_16 test_memory(void){ mem_slct[0] = (uint_8 *)0x100ffff0 ; mem_slct[1] = (uint_8 *)0x101ffff2 ; mem_slct[2] = (uint_8 *)0x102ffff4 ; mem_slct[3] = (uint_8 *)0x103ffff6 ; mem_slct[4] = (uint_8 *)0x104ffff7 ; mem_slct[5] = (uint_8 *)0x105ffffa ; mem_slct[6] = (uint_8 *)0x100ffffc ; mem_slct[7] = (uint_8 *)0x107ffffe ; for(num_chips = 0;num_chips < 8; num_chips ++ ) { mem_pointer = mem_slct[num_chips]; select_ram_page(mem_pointer); old_memory[num_chips] = *mem_pointer; mem_pointer++; test_data = *mem_pointer; mem_pointer --; *mem_pointer =0x55; test_data = *mem_pointer; *mem_pointer = old_memory[num_chips]; if( test_data != 0x55 ) break; } select_ram_page(0); return (num_chips); } I used global variables so that I could view the results in the IDE.
I had to actually use the mem_pointer so that it wouldn't be optimized out of use.
During our first production run of a 52259 based product we ran into defects where address and data lines were disconnected. Sometimes a defect of this nature can go undetected if the memory location affected isn't reached.
I now need to come up with a full memory test (not just a few locations) that will be able to indicate when a defect is discovered.
I will post my code when finished but I am open to suggestions.
Thanks
