Summary of Day 91:

Kernels of the day:

1. Symmetric Matrix Multiplication:

Kernel: 1

Click Here to redirect towards code

[!note]

• Performance: $4.8 \text{ TFLOPs}$
• Runtime: $166.12 \text{ ms}$
• GPU: A100-80GB

Kernel: 2

Click Here to redirect towards code

[!note]

• Performance: $8.5 \text{ TFLOPs}$
• Runtime: $93.81 \text{ ms}$
• GPU: A100-80GB

[!caution] For some reason, I could not get it working on H100 GPUs. It failed in last test case. I’ll try to optimize that in the future

1. Gemm with bias and RELU:

Click Here to redirect towards code

[!important] I went through HAMDI’s code for this. I just could not get it right on first time. Will try to give my own touch some day. I optimized his code basically

[!note]

• Performance: $10.6 \text{ TFLOPs}$
• Runtime: $1.43 \text{ ms}$
• GPU: A100-80GB