Matrix Multiplication using hybrid of OpenMp and CUDA
$200-350 USD
Zrušeno
Zveřejněno před více než 6 roky
$200-350 USD
Zaplaceno při doručení
Hi , I have a part of code to do matrix multiplication by using Cuda & Openmp . I want run three copies of matrix multiplication (same inputs) at parallel on three kernel . Then comparing the results for three kernels .
Three versions of multiplication arrays are as follows:
One of kernel do with shared memory, other without shared memory and last one do like this equation:
A * B "transpose of B" = C.
- Review the previous code
- Run the three copies in parallel "use Openmp". Run each kernels on a separate threads then compare three results and show me the results in Commands windows .After verification of the preliminary results of the program
Create an error in one of the kernels like this:
Change the addition process to subtract:
Cvalue + = ([login to view URL] [row * width + e]) * ([login to view URL] [e * width + col]);
To ensure that the program can detect the error , change the addition sign to subtract __>
Cvalue - = ([login to view URL] [row * width + e]) * ([login to view URL] [e * width + col]);
And determine where the error occurred in which kernels one or two or three .
Create multithreading from correct kernel and run it on the kernel where the error located and compared the results with previous results for the same kernel .So that we can determine the type of error .
I'm computer engineering TA with 10+ years of experience.
I'm experienced with GPU programming over CUDA - Linear algebra over C/C++ / Java/Python
Relevant Skills and Experience
I'm experienced with data structures - communications between threads and kernels in CUDA to truly
parallize the code - Linear algebra (Matrix operations - LU,QR,SVD Factorization )etc
Proposed Milestones
$300 USD - work done perfectly
$250 USD v 10 dnech
5,0 (13 recenze)
3,5
3,5
7 freelanceři nabízejí v průměru $285 USD za tuto práci
hello
I have strong background and experience with CUDA C development, you can see many completed projects in my profile. I can solve this with clear and effective code. You can ask me anything ofc
Relevant Skills and Experience
CUDA C/C++ programmer, also experience with hybrid techniques like OpenMP and MPI
Proposed Milestones
$250 USD - we can discuss it
Do you have any deadline?
Thanks
Coming from a computational physics background I have lots of numerical analysis and programming experience.
I have done such work before when I was working on a research team building a computing cluster using Nvidia GPUs for the university I work for. I have written thousands of lines of functional code.
We are a Team of Technical Consultants and Data Engineers having healthy experience into Big Data technologies,IOT/Cloud/AWS and Python/AI+Machine Learning.
The Dynamic force has qualified engineers having expertise in solving complex problems.
We have Experienced Customer Service Representative with demonstrated history of working in industry.
I have experience developing CUDA software during my PhD with multithreading (pthreads) in the host machine to paralellise IO and GPGPU calculations. This helped in the development of a bioinformatics application.
I also studied an official master in parallel computing, learning to use OpenMP. I also working in adapting an OS kernel to support the pthread.h API.