Tile Test for GPU (AMD)- scientific computation

Name: Tile Test for GPU (AMD)- scientific computation
Uploaded: Apr 17, 2026
Duration: 88 s
Description: Optimizing matrix subdivision is crucial for GPU-accelerated scientific computing. Testing an AMD Radeon RX 580 8GB with a 4096×4096 matrix across various tile sizes—including (8, 32), (16, 16), (32, 8), (32, 16), (16, 32), (64, 4), (4, 64)—showed that a 64×4 configuration provided the best performance. This project uses PyOpenCL for high-performance GPU communication. Code used :

Big Belly131 subscribers

25 views

Apr 17, 2026

1:28

Optimizing matrix subdivision is crucial for GPU-accelerated scientific computing. Testing an AMD Radeon RX 580 8GB with a 4096×4096 matrix across various tile sizes—including (8, 32), (16, 16), (32, 8), (32, 16), (16, 32), (64, 4), (4, 64)—showed that a 64×4 configuration provided the best performance. This project uses PyOpenCL for high-performance GPU communication. Code used :

Download

0 formats

No download links available.