Topic : Incorrect initialization of Y register

Forum : ST7/STM8

Original Post
Post Information Post
October 19, 2009 - 3:01pm
Guest

The following code:

unsigned char RxBuf[17];

short _Balance[6];
short _PeakP[6];
short WIP_Balance[6];

void ParseBalanceData(void);

int main(void)
{
    ParseBalanceData();
}

void ParseBalanceData(void)
{
   unsigned char ptr = RxBuf[0]*3;
   short bal;

   bal = (RxBuf[2] << 8) | (RxBuf[3]);
    if (_PeakP[ptr] < bal) _PeakP[ptr] = bal;
   _Balance[ptr] = (WIP_Balance[ptr]) >> 4;
   WIP_Balance[ptr] -= _Balance[ptr];
   WIP_Balance[ptr] += bal;

}

Uses the Y register for line 22 "WIP_Balance[ptr] -= _Balance[ptr];", but incorrectly initializes it (uses tye value held in A, instead of the one already in X.

0058 BE00       F                      LD     X,ptr
005A 58                                SLL    X
005B 9097                              LD     Y,A
005D D6002A     F                      LD     A,(WIP_Balance + 01H,X)
0060 90D00012   F                      SUB    A,(_Balance + 01H,Y)
0064 88                                PUSH   A
0065 D60029     F                      LD     A,(WIP_Balance + 00H,X)
0068 90D20011   F                      SBC    A,(_Balance + 00H,Y)
006C D70029     F                      LD     (WIP_Balance + 00H,X),A
006F 84                                POP    A
0070 D7002A     F                      LD     (WIP_Balance + 01H,X),

For now i'm using #pragma ot(5), which removes the issue but produces some really ugly output

00A2 B600       F                      LD     A,ptr
00A4 48                                SLL    A
00A5 3F03       F                      CLR    ?BH
00A7 3F04       F                      CLR    ?BL
00A9 B704       F                      LD     ?BL,A
00AB B604       F                      LD     A,?BL
00AD 97                                LD     X,A
00AE B604       F                      LD     A,?BL
00B0 9097                              LD     Y,A

Hopefully this will be fixed in the next version?

--
Blasio

Replies
Post Information Post
+1
0
-1
October 20, 2009 - 12:01pm
Raisonance Support Team

Hi Blasio,

Thansk for this detailed report. We reproduced the problem in our labs. It is indeed cause by an erroneous optimization in the peephole -- OT(7).

It will be fixed in the next RKit-STM8 release.

REgards,
Bruno