compiler making illegal opcodes

[josh_outerspace]

Heres some debug listing, it has invalid opcodes neatly commented. How can this happen? I must be doing something incorrectly. Below that is my source file. When I run the program, ADC always reads 0 and it crashes after a while. Sorry about tabs being lost in the paste, they only ever go 1 level deep.

Thanks,
J JORDAN

20002DD6: 0A004572 eori.b #0x72,d0
20002DDA: 726F moveq #111,d1
20002DDC: 7220 moveq #32,d1
20002DDE: 6F6E ble.s 0x20002E4E (0x20002e4e) ; 0x20002e4e
20002DE0: 206F7065 movea.l 28773(a7),a0
20002DE4: 7261 moveq #97,d1
20002DE6: 6E64 bgt.s 0x20002E4C (0x20002e4c) ; 0x20002e4c
20002DE8: 20726561640A movea.l ([25610,a2]),a0
20002DEE: 00005265 ori.b #0x65,d0
20002DF2: 7365 dc.w 0x7365 ; Invalid opcode
20002DF4: 7276 moveq #118,d1
20002DF6: 6564 bcs.s 0x20002E5C (0x20002e5c) ; 0x20002e5c
20002DF8: 2046 movea.l d6,a0
20002DFA: 6175 bsr.s 0x20002E71 (0x20002e71) ; 0x20002e71
20002DFC: 6C74 bge.s 0x20002E72 (0x20002e72) ; 0x20002e72
20002DFE: 2053 movea.l (a3),a0
20002E00: 7461 moveq #97,d2
20002E02: 7475 moveq #117,d2
20002E04: 7320 dc.w 0x7320 ; Invalid opcode
20002E06: 456E dc.w 0x456e ; Invalid opcode
20002E08: 636F bls.s 0x20002E79 (0x20002e79) ; 0x20002e79
20002E0A: 6469 bcc.s 0x20002E75 (0x20002e75) ; 0x20002e75

#include "common.h"
#include "mcf5xxx.h"
#include
#include "low_level.h"

__interrupt__ void pit0_irq(void)
{//vector 64+55=119, source =55
//take a read from the ADC
unsigned int adc_result; //store adc result here
PIT0_PMR = 0x6000; //reset flag
ADC_CTRL1 |= (113); //start conversion
adc_result = (ADC_ADRSLT0>>3); //read result register
printf("%d\n\r", adc_result);
fflush(stdout);

}

int main()
{
GPIO_DDRTC = 0xff; //init output port
GPIO_PORTTC = 0x00; //init port pins
mcf5xxx_set_handler(64 + 55, pit0_irq); //init irq

//PWM module setup, outputs 1, 3, 5, 7
GPIO_PTDPAR = 0x0f; //pwm outputs on port td
PWM_PWMPOL = 0xaa; //pulses high
PWM_PWMCTL = 0xf0; //concatenate all 4 channels to 16bit
PWM_PWMDTY01 = 0x0010;
PWM_PWMPER01 = 0xffff;
PWM_PWMDTY23 = 0x1000;
PWM_PWMPER23 = 0xffff;
PWM_PWMDTY45 = 0x1000;
PWM_PWMPER45 = 0xffff;
PWM_PWMDTY67 = 0x1000;
PWM_PWMPER67 = 0xffff;
PWM_PWMSCLA = PWM_PWMSCLB = 8;
PWM_PWMCLK = 0xaa; //select scaled clocks
PWM_PWME = 0xaa; //enable all channels


//PIT module setup
PIT0_PCSR = PIT_PCSR_PRE(13)|PIT_PCSR_PIE|
PIT_PCSR_RLD|PIT_PCSR_EN;
PIT0_PMR = 0x6000;

//configure analog to digital converter
ADC_CTRL2 = 3; //clock div assuming fast clock
ADC_ADSDIS = (11); //stop at AN1, only perform sample0 with AN0
ADC_POWER &= ~(0x0001); //power up adc A
GPIO_PANPAR = 1; //setup AN0 to primary pin function


INTC_ICR55 = 0 | INTC_ICR_IP(6) | INTC_ICR_IL(6);
//set priorities for pit0 irq55
INTC_IMRH &= ~INTC_IMRH_MASK55;
//unmask that specific interrupt

mcf5xxx_irq_enable();

while(1)
{

}

}

[SimonMarsden_de]

The whole disassembly listing looks like garbage. As well as the illegal opcodes, even the bytes which disassemble to instructions don't look sensible. For example,

moveq #111,d2
moveq #32,d2
ble.s 0x20002E4E

This would load 111 into the D2 register, immediately throw that away and then load 32 into D2. Finally, the 'Branch on Less than or Equal (BLE)' would never be taken, since 32 is not less than or equal to 0 - in other words, this code makes no sense.

In conclusion, I doubt that that's compiler generated code. It looks as though you are trying to execute data. Perhaps the program is slowly being corrupted during execution?

[josh_outerspace]

Maybe it was the result of trying to treat the 8 bit pwm registers as 16 bit, I hacked up the header file, see the third line is my own invention. Is my goal of treating the two registers together as a 16 bit register possible? Also what Ive done is take away all the MCF_ from all the defines.

#define PWM_PWMDTY0 (*(vuint8 *)(&__IPSBAR[0x1B001C]))
#define PWM_PWMDTY1 (*(vuint8 *)(&__IPSBAR[0x1B001D]))
#define PWM_PWMDTY01 (*(vuint16 *)(&__IPSBAR[0x1B001C]))

The illegal opcodes and crashing seems to go away when my modifications are removed, but the ADC still reads zero always.

Message Edited by J JORDAN on 2006-12-0703:16 PM

[CompilerGuru]

As already mentioned, the decoded area does not contain code, but it is not corrupted either.
Instead it just contains some strings, the first few bytes (45 72 ....) read as
"Error on operand". It contains string literals for error handling.

Daniel

[peg]

Hi J Jordan,

See here regarding tab problem:

http://forums.freescale.com/freescale/board/message?board.id=FORUMUSE&message.id=294

Regards

Peg