Reading of const-array (2-dim) with PCH 3.202, Opt-9

[UKH]

Hello together,

during compiler-update (3.189 -> 3.202) one of a function creates
troubles:

byte const CAN_SPEED[4][5] =
{
// BRG, SYNC, PROP, PH1, PH2
{ 0, 0, 0, 3, 3}, // 1 MBit -> 30m (10 Tq)
{ 1, 0, 0, 3, 3}, // 500 kBit -> 100m (10 Tq)
{ 4, 0, 2, 5, 5}, // 125 kBit -> 500m (16 Tq)
{ 9, 0, 2, 5, 5} // 62,5 kBit -> 1000m (16 Tq)
};

void can_set_baud(byte b_nr) {
byte a;
b_nr = 0;
a =CAN_SPEED[b_nr][0]; // a=3 ? should be 0
a =CAN_SPEED[b_nr][1]; // a=1 ? should be 0
a =CAN_SPEED[b_nr][2]; // a=0 ? should be 0
a =CAN_SPEED[b_nr][3]; // a=0 ? should be 3
a =CAN_SPEED[b_nr][4]; // a=3 ? should be 3
b_nr=1;
a =CAN_SPEED[b_nr][0]; // a=3 ? should be 1
a =CAN_SPEED[b_nr][1]; // a=4 ? should be 0
a =CAN_SPEED[b_nr][2]; // a=0 ? should be 0
a =CAN_SPEED[b_nr][3]; // a=2 ? should be 3
a =CAN_SPEED[b_nr][4]; // a=5 ? should be 3
b_nr=2;
a =CAN_SPEED[b_nr][0]; // a=5 ? should be 4
a =CAN_SPEED[b_nr][1]; // a=9 ? should be 0
a =CAN_SPEED[b_nr][2]; // a=0 ? should be 2
a =CAN_SPEED[b_nr][3]; // a=2 ? should be 5
a =CAN_SPEED[b_nr][4]; // a=5 ? should be 5
b_nr=3;
a =CAN_SPEED[b_nr][0]; // a=5 ? should be 9
a =CAN_SPEED[b_nr][1]; // a=f2 ? should be 0
a =CAN_SPEED[b_nr][2]; // a=cf ? should be 2
a =CAN_SPEED[b_nr][3]; // a=0e ? should be 5
a =CAN_SPEED[b_nr][4]; // a=f0 ? should be 5
}

The reading of the const-array doesn't works.
Has someone found the same ?
If I use OPT-8 (+Y=8) it works correctly ????


UKH

[Haplo]

Did the same code work correctly under 3.189?

If you look at the pattern of the data you are getting in the variable a, it is obvious that the compiler is skipping the first four bytes of your array and treating the fifth byte as the first. And that is also the reason why your last four readings are totally wrong.

[UKH]

Yes, same works with 3.189.

UKH

[jventerprises]

i found that i could not use ICD and look at any values in the const array, but when i ran the code, it worked just fine. i could not step over the code that used the CONST, though.

i suspect it has to do with the compiler actually replacing these values at compile time, so they dont actually exist.
_________________
Jon

[Haplo]

Yes, that is because the compiler implements a look-up table with RETLWs for constant arrays. It doesn't store the actual values in the memory.

[UKH]

Yesterday I've passed my problem to CCS. Currently I turned back to version 3.189 which works fine in this case.

[UKH]

Bug was confirmed and will be fixed in 3.203

UKH