gistfile1.txt

josh@ubuntu:~/hasher$ dd if=/dev/zero bs=65536 count=16384 | ./xxHash/xxh128sum 
16384+0 records in
16384+0 records out
16024760318c6298efd1151033ad2e9f  stdin
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 0.23966 s, 4.5 GB/s
josh@ubuntu:~/hasher$ dd if=/dev/zero bs=65536 count=16384 | ./a.out 
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB, 1.0 GiB) copied, 0.210685 s, 5.1 GB/s
16024760318c6298efd1151033ad2e9f

hasher.c

// gcc -O3 -march=native -g -I ./xxHash hasher.c
#define _GNU_SOURCE
#include "xxh3.h"
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>

void print_xxh128(unsigned char * hash) {
    for(int i = 0; i < 16; i++) {
      printf("%02x", hash[i]);
    }
    printf("\n");
}

void calcul_hash_streaming(int fd)
{
    int pipes[2];
    if(pipe(pipes) != 0) {
        perror("pipe()");
	exit(1);
    }

    size_t const bufferSize = 64 * 1024;
    void* const buffer = malloc(bufferSize);
    if (buffer==NULL) abort();

    /* create a hash state */
    XXH3_state_t state;

    /* Initialize state with selected seed */
    if (XXH3_128bits_reset(&state) == XXH_ERROR) abort();

    size_t length;
    while(1) {
	// splice() has a shorter critical section than a simple read(), so
	// pre-splicing reduces lock contention
	length = splice(STDIN_FILENO, NULL, pipes[1], NULL, bufferSize, 0);
	if(length == 0) {
	    break;
	}
	read(pipes[0], buffer, length);
        if (XXH3_128bits_update(&state, buffer, length) == XXH_ERROR) abort();
    }

    /* Produce the final hash value */
    XXH128_hash_t const hash = XXH3_128bits_digest(&state);
    unsigned char digest[16];
    XXH128_canonicalFromHash((XXH128_canonical_t *) digest, hash);
    print_xxh128(digest);

    free(buffer);
}

int main() {
    calcul_hash_streaming(0);
}