/g/ - OpenCL - Just learning it right no...

Sitewide Rules
You must be 18 or older to post.
Sharing personal details or engaging in doxing is strictly prohibited.
Political discussions should be confined to /pol/.
Off-topic discussions, thread derailment, or spam may result in a ban and IP blacklist.
Pornographic content is strictly prohibited.
Any activity violating local laws is not allowed.
If you are not an Indian, you can only post in /int/. Or create and account and ask for approval to post in other boards.

OpenCL

Duck

17tB9B

24/09/25 (Wed) 17:00| Last Reply: 5 months 4 days ago

No.2157

Select Post

Just learning it right now, will add whatever I make here.

Wrote some matrix multiplication code today.


#include <CL/cl.h>
#include <stdio.h>
#include <stdlib.h>


int main(int argc, char** argv){
 cl_platform_id platform;
    cl_device_id device;
    clGetPlatformIDs(1, &platform, NULL);
    clGetDeviceIDs(platform, CL_DEVICE_TYPE_GPU, 1, &device, NULL);

    // 2. Create context and command queue
    cl_context context = clCreateContext(NULL, 1, &device, NULL, NULL, NULL);
    cl_command_queue queue = clCreateCommandQueue(context, device, 0, NULL);

   FILE* f = fopen("test_cl.cl", "rb");
fseek(f, 0, SEEK_END);
size_t size = ftell(f);
rewind(f);

char* source = malloc(size+1);
fread(source, 1, size, f);
source[size] = '\0';
fclose(f);

cl_program program = clCreateProgramWithSource(context, 1,
                          (const char**)&source, &size, NULL);
clBuildProgram(program, 1, &device, NULL, NULL, NULL);

 cl_kernel kernel = clCreateKernel(program, "test_kernel", NULL);
  
 
 
  size_t m=3;
 size_t n=2;
 size_t a=2;
 size_t b=4;


 
  float *A = (float*)malloc(sizeof(float) * m * n);
    float *B = (float*)malloc(sizeof(float) * a * b);
    float *C = (float*)malloc(sizeof(float) * m * b);
 

for (int i = 0; i < m*n; i++) {
        A[i] = (float)i;
      
    }

for (int i = 0; i < a*b; i++) {
        B[i] = 1024 - (float)i;
      
    }



     cl_mem d_A = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR, sizeof(float)*(m*n), A, NULL);
    cl_mem d_B = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR, sizeof(float)*(a*b), B, NULL);
    cl_mem d_C = clCreateBuffer(context, CL_MEM_WRITE_ONLY, sizeof(float)*(m*b), NULL, NULL);
 
  clSetKernelArg(kernel, 0, sizeof(cl_mem), &d_A);
    clSetKernelArg(kernel, 1, sizeof(cl_mem), &d_B);
    clSetKernelArg(kernel, 2, sizeof(cl_mem), &d_C);
    clSetKernelArg(kernel, 3, sizeof(unsigned int), &m);
   clSetKernelArg(kernel, 4, sizeof(unsigned int), &b);
  clSetKernelArg(kernel, 5, sizeof(unsigned int), &a);
    size_t globalSize[2] = {m, b};
    clEnqueueNDRangeKernel(queue, kernel, 2, NULL, globalSize, NULL, 0, NULL, NULL);
     clFinish(queue);


clEnqueueReadBuffer(queue, d_C, CL_TRUE, 0, m*b * sizeof(float), C, 0, NULL, NULL);


     
int k=0; int i = 0; int z =0;
   
       while(k<m){
         printf("%f\t", C[z]);
         z=z+1;
        i++;
         if(i==(b)){k=k+1; i=0; printf("\n");}
       
       
    }
         clReleaseMemObject(d_A);
    clReleaseMemObject(d_B);
    clReleaseMemObject(d_C);
    clReleaseProgram(program);
    clReleaseKernel(kernel);
    clReleaseCommandQueue(queue);
    clReleaseContext(context);
    free(A);
    free(B);
    free(C);
 

}

OpenCL

Duck

loc-IN

51q9xuub4j7.png

Just learning it right now, will add whatever I make here. Wrote some matrix multiplication code today. [code] #include <CL/cl.h> #include <stdio.h> #include <stdlib.h> int main(int argc, char** argv){ cl_platform_id platform; cl_device_id device; clGetPlatformIDs(1, &platform, NULL); clGetDeviceIDs(platform, CL_DEVICE_TYPE_GPU, 1, &device, NULL); // 2. Create context and command queue cl_context context = clCreateContext(NULL, 1, &device, NULL, NULL, NULL); cl_command_queue queue = clCreateCommandQueue(context, device, 0, NULL); FILE* f = fopen("test_cl.cl", "rb"); fseek(f, 0, SEEK_END); size_t size = ftell(f); rewind(f); char* source = malloc(size+1); fread(source, 1, size, f); source[size] = '\0'; fclose(f); cl_program program = clCreateProgramWithSource(context, 1, (const char**)&source, &size, NULL); clBuildProgram(program, 1, &device, NULL, NULL, NULL); cl_kernel kernel = clCreateKernel(program, "test_kernel", NULL); size_t m=3; size_t n=2; size_t a=2; size_t b=4; float *A = (float*)malloc(sizeof(float) * m * n); float *B = (float*)malloc(sizeof(float) * a * b); float *C = (float*)malloc(sizeof(float) * m * b); for (int i = 0; i < m*n; i++) { A[i] = (float)i; } for (int i = 0; i < a*b; i++) { B[i] = 1024 - (float)i; } cl_mem d_A = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR, sizeof(float)*(m*n), A, NULL); cl_mem d_B = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR, sizeof(float)*(a*b), B, NULL); cl_mem d_C = clCreateBuffer(context, CL_MEM_WRITE_ONLY, sizeof(float)*(m*b), NULL, NULL); clSetKernelArg(kernel, 0, sizeof(cl_mem), &d_A); clSetKernelArg(kernel, 1, sizeof(cl_mem), &d_B); clSetKernelArg(kernel, 2, sizeof(cl_mem), &d_C); clSetKernelArg(kernel, 3, sizeof(unsigned int), &m); clSetKernelArg(kernel, 4, sizeof(unsigned int), &b); clSetKernelArg(kernel, 5, sizeof(unsigned int), &a); size_t globalSize[2] = {m, b}; clEnqueueNDRangeKernel(queue, kernel, 2, NULL, globalSize, NULL, 0, NULL, NULL); clFinish(queue); clEnqueueReadBuffer(queue, d_C, CL_TRUE, 0, m*b * sizeof(float), C, 0, NULL, NULL); int k=0; int i = 0; int z =0; while(k<m){ printf("%f\t", C[z]); z=z+1; i++; if(i==(b)){k=k+1; i=0; printf("\n");} } clReleaseMemObject(d_A); clReleaseMemObject(d_B); clReleaseMemObject(d_C); clReleaseProgram(program); clReleaseKernel(kernel); clReleaseCommandQueue(queue); clReleaseContext(context); free(A); free(B); free(C); } [/code]

Duck

17tB9B

24/09/25 (Wed) 17:01

No.2158

Select Post

forgot to attach kernel

Anonymous

amIs5S

24/09/25 (Wed) 17:04

No.2159

Select Post

>>2158

love you anon

Duck

O//58R

24/09/25 (Wed) 18:12

No.2160

Select Post

>>2159

Love you too saar

Duck

E1cncT

24/09/25 (Wed) 19:49

No.2161

Select Post

Downloaded some openCL book, none of their example codes work. Bruh

Anonymous

eQcw0H

24/09/25 (Wed) 19:58

No.2162

Select Post

what are you writing opencl for? platform? each vendor has their own big fat libraries implementing the spec. those libs contains tons of test cases and example programs.

Anonymous

eQcw0H

24/09/25 (Wed) 19:59

No.2163

Select Post

contain*

Duck

E1cncT

24/09/25 (Wed) 20:01

No.2164

Select Post

>>2162

Well I was planning to pair it with openGL, doing some simulations etc. The amd device isn't rocm supported so openCL is basically the only option

Anonymous

eQcw0H

24/09/25 (Wed) 20:09

No.2165

Select Post

>>2164

hmm, your compiler sources are the best documentation btw

Anonymous

eQcw0H

24/09/25 (Wed) 20:20

No.2166

Select Post

cant you sort of do a static analysis of the binary produced by the compiler to reconstruct the program logic?

Anonymous

o5RdgU

25/09/25 (Thu) 14:53

No.2187

Select Post

join #indiachan on irc.libera.chat

Duck

E1cncT

25/09/25 (Thu) 15:41

No.2188

Select Post

>>2187

Nahi karni bhai

Anonymous

TwD3PD

25/09/25 (Thu) 15:50

No.2189

Select Post

>>2157(OP)

What's the use case of opencl?

Duck

E1cncT

25/09/25 (Thu) 15:55

No.2190

Select Post

>>2189

Writing gpu kernels but you don't have to write separate code for amd and nvidia. Device can be selected at runtime

Anonymous

C9bAm4

25/09/25 (Thu) 16:03

No.2191

Select Post

>>2190

Sounds really cool yaar. You are into low level programming?

Anonymous

o5RdgU

25/09/25 (Thu) 18:16

No.2193

Select Post

>>2188

np bhai, keep your shitter and pisscord

Anonymous

o5RdgU

25/09/25 (Thu) 18:17

No.2194

Select Post

>>2191

low-iq programming it's called, saar

Duck

E1cncT

25/09/25 (Thu) 23:29

No.2200

Select Post

I am fucking stupid, I have unified memory why am I copying the memory again like it's a discrete gpu

DUck

g8FUkT

29/09/25 (Mon) 01:16

No.2242

Select Post

Solved this today

https://codeforces.com/problemset/problem/2146/C

was fucking hard, atleast for me

DUck

g8FUkT

29/09/25 (Mon) 01:18

No.2243

Select Post

also watched reze arc today, no anime watching friends so went alone to watch it.

DUck

g8FUkT

29/09/25 (Mon) 01:24

No.2244

Select Post

By the end of the year I want

>1500 on cf

>1800 on chess blitz

DUck

g8FUkT

29/09/25 (Mon) 01:32

No.2245

Select Post

CS2 has such a shit experience man wtf, but I talked to some chinese guy studying in fucking NTU in singapore, also a man playing from a fking jewellery shop that his parents own and he is forced to sit.

Duck

ZN5IME

29/09/25 (Mon) 21:30

No.2273

Select Post

Solved some more CF today. Played laser tag today

Anonymous

kDnY+x

29/09/25 (Mon) 21:31

No.2275

Select Post

>>2273

do you play CTFs?

Duck

ZN5IME

29/09/25 (Mon) 21:32

No.2276

Select Post

>>2275

Not much I solved some problems during lockdowns like that mr robot challenge, and 4-5 more. Seem boring to me tbh

Anonymous

kDnY+x

29/09/25 (Mon) 21:37

No.2277

Select Post

>>2276

maybe it's just that you're retarded

Duck

858oaM

29/09/25 (Mon) 21:58

No.2280

Select Post

>>2277

Probably troo>>2277

Duck

E1cncT

30/09/25 (Tue) 14:24

No.2294

Select Post

Going home today. Sick of hearing some different southie language every morning

Duck

WVTAqR

01/10/25 (Wed) 06:01

No.2338

Select Post

Will have to learn threejs, can't handle the fact that someone can made that run in a browser and I can't

Duck

h4ZpI4

02/10/25 (Thu) 02:57

No.2359

Select Post

New found appreciation for home, will go to buy a laptop tomorrow. Hope some pajeet laptop salesman doesn't convince/scam my dad into buying some junk

Duck

wx3yDk

06/10/25 (Mon) 19:16

No.2399

Select Post

I want to kill myself but well

Duck

wx3yDk

08/10/25 (Wed) 22:21

No.2409

Select Post

What are you supposed to do when you solve a leetcode question but it isn't the best possible solution? Solve all of them again?

Anonymous

8BfR40

08/10/25 (Wed) 22:22

No.2410

Select Post

>>2409

look how can you optimize it, or just look the optimal solution and submit it.

Duck

wx3yDk

13/10/25 (Mon) 02:14

No.2446

Select Post

>I need to do better

>I need to kill myself

Fucking sucks that I can't bring myself to do either of the two. Feel like I am destined for mediocrity

/g/ - Technology

Board dedicated to discussions related to everyday Technology!

Notice

A poster on BharatChan must abide by the following rules:

Recent Posts

/desktop thread/

I am done. It's over.

Substack

4Chan bypass?

/emacs/ general

Forking jschan to submit a PR for captcha logic

This is Huge

PRICE OF INTERNET IN 1995

Motorola and Graphene linked up

HAPPENING

CPchads

JEEFICATION OF GSOC

Minecraft source code leaked

where do i get CRT monitor ??

Shifting to linux mint

Androidfags zara idhar aana

Sarvam is now proven to be a disappointment

XHDATA D-808 DX-ing setup, analogue modulation

i don't understand

RCE on Pocketbase possible?

AI Impact Summit 2026

Simple Linux General /slg/ - Useful Commands editi...

/wpg/ - Windows & Powershell General

Sarvam Apology Thread

Some cool tech in my college

the hmd touch 4g

Holy Shit

Saar american companies have best privacy saaar

/g/ related blogpost - backup thread

Android Hygiene

My favorite game rn

COOKED

Are we getting real 5g?

I want to create my own forum but I don't know how...

Is my psu not compatible with my mobo?

/i2pg/ - I2P general

OpenCL

OpenCL