CUDA Part A: GPU Architecture Overview and CUDA Basics; Peter Messmer (NVIDIA)

[@shezijalal1774]

Great presentation and a good style to express the key concepts in a simple manner, really like it, it is good that be slow rather to go fast, so every one can understand clearly, people feel's like the person is scared but that's not the case !

[@wyatt5984]

Fantastic insights and great video to learn the basics of CUDA and still relevant information. Thanks so much for this video.

[@paulgrad1348]

What's an xtor? It doesn't show up anywhere. Edit: Oh, transistor duh. That's a needlessly obscure abbreviation though.

[@webgpu]

Thanks, i didn't know either (the closest i knew was "xtal" - crystal)

[@dedenjama7173]

am just trying to figure out whether i should buy their stock now that its split. CUDA WUDA SHUDA

[@robertpearson8546]

One defect in C is that your main routine can be anything as long as it is "main". Your compiler output can be anything as long as it is "a.out".

[@nick9198]

wut

[@AssassinGrudge]

Daynamic paralleisme is not supported on kerpel with sm archi = 30 and i can't affort to by new peer of GTX with higher sm_3.x what should i do ?

[@LearnitCS]

Check out the entire CUDA programming course here: kzbin.info/www/bejne/mafSZJqkftOEba8

[@wolfram77]

In code example 1; i see an error. Can anyone else find it out?

[@OttoFazzl]

So a kernel is a C++ function that is mapped onto the grid of thread blocks? This is why we avoid all the loops. But I did not hear the presenter to say it explicitly, it is just a mapping. Since each block/thread will have their own indices, the mapped kernel will act on a different place in memory.

[@OttoFazzl]

I couldn't wrap my head around about the absence of loops until I realized that was just function mapping to an array of blocks/threads. Why didn't he say so explicitly?

[@darylallen2485]

Seems to me that a warp and a grid are synonyms.

[@patrickjones7434]

Great lecture

[@shanku444]

There should be a 3d visualization of blocks and grids

[@kansairobot2015]

I see you are using a virtual environment. Can you use CUDA and access the GPU from linux running on a virtual environment?

[@vasanthplays8776]

In certain scenarios you could

[@kansairobot2015]

@@vasanthplays8776 care to explain how? so far i have been told you cant

[@robertpearson8546]

Your code assumes float * means float[] *

[@临赵-u5s]

how does this guy compile his cuda program in ubuntu box? is that even possible?

[@alsa010]

+临赵 Why not? CUDA SDK exists for Linux too. The problem would be running the compiled binary inside VirualBox :)

[@patrickjones7434]

What is unbunto

[@epiqboy]

@@patrickjones7434 a Linux operating system distribution.

[@patrickjones7434]

Hi doe this mean I will have to go to university to become a coder

[@bruceb85]

this is why the quadro cost so much

[@quantum.9883]

Quadro's are often sold to companies with much larger budgets for computer systems than your average pc enthuasiast.

[@th1alb]

he sounds like he's scared and intimidated by the audience.

[@complicated2359]

the presentation is completely ruined by constant stutters.