Hello,
recently I ran into a problem while working with 64-bit integers on MSP430G2553. The code listed below does some manipulations with 64-bit unsigned integers. The problem only occurs on "cold start", i. e. after powering up the board, flashing the firmware and running it from CCS. However, this is a result of my attempt of minimizing the amount of code required to reproduce the problem; in my current project this problem happens every time the partucular piece of code is executed. Anyway, here is the code:
main.c:
#include <msp430.h>
#include <stdint.h>
#include "test-file.h"
static test_struct test_struct_array[] = {
{
.data = 0xDEADBEEFDEADBEEF
},
{
.data = 0xDEADBEEFDEADBEEF
}
};
/*
* main.c
*/
void main(void) {
WDTCTL = WDTPW | WDTHOLD;
f_init(
(test_struct *) &test_struct_array,
ARRAY_SIZE(test_struct_array)
);
LPM3;
}
test-file.h:
#ifndef TEST_FILE_H
#define TEST_FILE_H
#include <msp430.h>
#include <stdint.h>
#define ARRAY_SIZE(a) (sizeof(a) / sizeof((a)[0]))
typedef struct {
uint64_t data;
} test_struct;
void f_init(test_struct * d, uint8_t s);
#endif
test-file.c:
#include "test-file.h"
test_struct * t;
uint8_t c;
void f_b(uint8_t a) {
volatile int b = a;
}
void f_a() {
uint64_t a = t[c - 1].data;
uint64_t b = t[c - 1].data;
//
uint8_t i = 63;
//
do {
f_b(1 && (a & ((uint64_t) 1 << i)));
f_b(1 && (b & ((uint64_t) 1 << i)));
} while (i--);
}
void f_init(test_struct * d, uint8_t s) {
t = d;
c = s;
f_a();
}
Expected result and the problem
There is a breakpoint in the screenshot below. At this point, both a and b should be equal 0xDEADBEEFDEADBEEF. However, b is 0xFFFF847DDEADBEEF (upper 32 bits differ from time to time, so I guess these bits are kind of random). Screenshot:
When volatile keyword is added to the variable declarations, everything works as excepted.
Is this a compiler or code problem?
Optimization settings: 0 (Register optimizations)
Compiler: MSP430 Compiler Tools 4.3.5
CCS: 6.0.1.00040
Regards,
-- Vlad
