horkyze / cudasha256 Goto Github PK

As the title says, the file API used does not allow working with files larger than 4 GB. With cards that have up to 24 GB of VRAM like 3090 RTX and with fast SSDs that would probably be a better use case than doing sha256 on a GPU for a bunch of small files where just the latency of the request is going to hide all the benefits of fast calculation.

Consider changing file API to something that can handle 64-bit file sizes, and consider using ptrdiff_t instead of char for pointer arithmetic.

I was trying to use your code as a starting point to make a sha256 cuda hasher that takes lines from stdin (Or any text file) and hashes each line as a job (Outputing HASH and Original Line in the output separated by a space or other delimiter)

Your program is so close to doing this (You hash filesystem files), but my c++ is weak. Can you assist me? The way you read the files in for the –f switch, but instead of getting the file data from get_file_data, could you just hash the string?

Hashing "hello world!" should return 7509e5bda0c762d2bac7f90d758b5b2263fa01ccbc542ab5e3df163be08e6ca9, and hashing that in a file works, when i try to use the api to hash "hello world!", it gives a totally incorrect hash.

(code that produces incorrect result)

#include <stdio.h>
#include <stdint.h>
#include "sha256.cuh"


__global__ void seedHash()
{
    uint8_t testInput[] = "hello world!";

    //memcpy(byteArray, &seed, sizeof(uint64_t));

    uint8_t byteArray[32];
    SHA256_CTX sha;
    sha256_init(&sha);
    sha256_update(&sha, testInput, 12);
    sha256_final(&sha, byteArray);

    printf("Hash output: ");
    for (int i = 0; i < 32; i++) printf("%.2x", byteArray[i]);
    printf("\n");
}

int main()
{
    //cudaSetDevice(1);
    seedHash<<<1,1>>>();
    cudaDeviceSynchronize();
}

HI @Horkyze!

I'd like to give this a go in a school project I'm working on, however there is no licence.md or information on reuse.

May I use it?

Hi,

I get cudaError 30 (unknown error) while calling cudaMallocManaged.

I have Quadro K420 card with cc 3.0.

Do you have any idea about this error?

Thanks.

Hello, may I ask for your help?

After I modify the code, I get an incorrect hash value.

		buff = (BYTE *)"abc";
		checkCudaErrors(cudaMallocManaged(&buff, (3)*sizeof(char)));
		jobs[i] = JOB_init(buff, 3, argv[index]);