right shift issue

[tpircher-zz]

I think I'm running into a problem with the right-shift operation. It seems that the rrc operation in-fills zeros or ones, independently of the carry flag.

I'm trying to emulate the following C code:

#include <stdlib.h>
#include <stdint.h>

void xprintf(const char *fmt, unsigned int val)
{
    (void)fmt;
    volatile unsigned int xxx = val;
    (void)xxx;
}

int main(void)
{
    volatile uint16_t val1 = 0xf001;
    volatile uint16_t val2 = 0x4f85;
    xprintf("res: %#x\n", val1 ^ (val2>>4));

    return 0;
}

I think I run into a problem with the right-shift (val2>>4). This code is compiled as:

    c04a:       b4 40 85 4f     mov     #20357, -4(r4)  ;#0x4f85, 0xfffc(r4)
    c04e:       fc ff 
    c050:       1f 44 fc ff     mov     -4(r4), r15     ;0xfffc(r4)
    c054:       12 c3           clrc
    c056:       0f 10           rrc     r15
    c058:       12 c3           clrc
    c05a:       0f 10           rrc     r15
    c05c:       12 c3           clrc
    c05e:       0f 10           rrc     r15
    c060:       12 c3           clrc
    c062:       0f 10           rrc     r15

In the online emulator I get the sequence of values of:

0x4f85
0xa7c2
0xd4e1
0x69f0
0xb4f8

I.e. sometimes I get a zero infill, sometimes I get a 1 infill, even though the carry flag should be cleared by the clrc instruction.

I compile with msp430-gcc version 4.6.3:

msp430-gcc -mmcu=msp430g2553 -O0 -g -W -Wall -std=c99   -c -o main.o main.c
msp430-gcc -mmcu=msp430g2553 -O0 -g -o firmware.elf main.o

[RudolfGeosits]

tpircher,

Thank you for bringing this bug to my attention, I'll be taking a look at it this week!

• Rudolf