4 ms·
> AI programs only need a small amount of different kernels That is not the case. What appears like a simple matmul operation actually requires these libraries
by david-gpu 2y ago
> AI programs only need a small amount of different kernels
That is not the case. What appears like a simple matmul operation actually requires these libraries to select which specific kernel out of the many internally available to execute.
If you are curious to learn more, NVidia open sourced a library called Cutlass some years ago. And remember that is only what they are willing to open source.
- littlestymaar 2y agoIs that really different from AV codecs in terms of scale though?
- david-gpu 2y agoI am not at liberty to discuss more than that.
- saagarjha 2y agoYes, you can peek under the hood of cuBLAS and notice that it has dozens of kernels for different problem sizes. It’s not generally the case that when you do h264 at a different crf you have a completely different tiling strategy that you have to implement.