calc.c

/*
 * Byte calculator. Sums bytes of a given file and displays to screen.
 *
 * Compile:
 *  gcc calc.c -o calc -O2
 *
 */
#include <stdlib.h>
#include <stdio.h>
#include <stdint.h>
#define BUFFER (1 << 20)

unsigned long long do_calc(FILE *f) {
    uint8_t buf[BUFFER];
    unsigned long long sum = 0;
    size_t nbytes, i;

    while ((nbytes = fread(buf, 1, BUFFER, f)) > 0)
        for (i = 0; i < nbytes; i++)
            sum += buf[i];

    return sum;
}

int main(int argc, const char* argv[]) {
    FILE *f;
    unsigned long long res;

    if (argc != 2)
        return 1;
   
    if ((f = fopen(argv[1], "rb")) == NULL) {
        perror("fopen");
        return 1;
    };

    res = do_calc(f);
    fclose(f);

    printf("%llu\n", res);
    return 0;
}

Ok, so a 20 minute hack gave this (ccacl is the C code, hcalc is the Haskell code)

[~/sandbox/calc:master]λ sudo su -c 'sync && echo 3 > /proc/sys/vm/drop_caches'
[~/sandbox/calc:master]λ du -h big_file 
4.4G    big_file
[~/sandbox/calc:master]λ sudo su -c 'sync && echo 3 > /proc/sys/vm/drop_caches'
[~/sandbox/calc:master]λ time ./ccalc big_file 
599364006345

real    0m14.595s
user    0m2.670s
sys 0m2.450s
[~/sandbox/calc:master]λ sudo su -c 'sync && echo 3 > /proc/sys/vm/drop_caches'
[~/sandbox/calc:master]λ time ./hcalc big_file 
599364006345

real    1m17.030s
user    1m14.010s
sys 0m2.943s

The implementation is almost naive, except for one use of BangPatterns, but I think I can get rid of it.

I will optimise and share the code, but in any case, at least it works fine.

Best I can do for now: https://gist.github.com/yfyf/6639162