Parshav

Interested in Computer Systems and Machine Learning. Computer Science @NCState

[Parshav @gParshavv]

Day 1 was implementing a simple kernel to copy a 1D array. https://t.co/haKDmnPr98

[Parshav @gParshavv]

I have started the 100 days CUDA challenge by @hkproj. https://t.co/qCC3ZXOeEM Super initiative and pretty excited for this.

[Parshav @gParshavv]

@WarnerTeddy Air mattress is all we need

[Parshav @gParshavv]

@doesdatmaksense Sure

[Parshav @gParshavv]

If you’re interested in programming GPUs or just starting out, I think you’ll find this helpful. Check it out here:

https://t.co/pKmTnchxzA

[Parshav @gParshavv]

I’ve taken some time to write a small blog on how the execution of a CUDA program works behind the scenes and a gentle introduction to GPU architecture.
When I started learning GPU programming, I wished for something like this to help me understand the basics.

[Parshav @gParshavv]

@maharshii Next is DL models making themselves learn?

[Parshav @gParshavv]

That’s basically it. I am assuming we can replicate it with whatever task we are working on

[Parshav @gParshavv]

Was trying to learn CUDA and was finding it real hard to understand indexing threads correctly in higher dimensions. Just understood something that feels nice and correct,

[Parshav @gParshavv]

every element in the data that you are working with will also be identified by 2 dimensions(i and j). Now inside the kernel you can do, i = blockIdx.x*blockDim.x + threadIdx.x and j = blockIdx.y*blockDim.y+ threadIdx.y

[Parshav @gParshavv]

@tensor_fusion what if they already have it

[delian @zebulgar]

when I was showering this morning

somehow my apple watch triggered my phone to play just this video, and this video only

clearly a prophecy for the...

2025 MOOD!!!!

[Parshav @gParshavv]

Dynamic programming is basically trying to speed up recursions by using extra space. We optimize on time using more space.

[Parshav @gParshavv]

@elliotarledge let’s fucking go

[Parshav @gParshavv]

@anpaure Maintain a prefix sum in a hash map.