@万苟仙
Your code at 19:07 is wrong, the index doesn't correctly locate the thread within blocks and the stride is also wrong!!! it should be : int index = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x;
@omri9325
Can you talk about some of your chrome extensions? I got curious from these icons up there.
@7Trident3
Next video suggestion!! How to run plaidML and access AMD and Intel GPU's too!
@mrtynxx
Amazing video! Thanks
@ykochubeev
Thank you fo such great video! Could you please also share info where could we download supplementary *.cu source files? For example "simple.cu" which was used in this colab notebook
@abhishekborikar4958
Where should I find some study buddy online? Please share link.
@RAJATTHEPAGAL
😅 Maybe nitpicking but for those who read 1<<20 as shorthand for 1 million, it's not it's actually a bit shift operator which turned 32 bit 1 i.e 31 0s followed by 1 , into 12 0s followed 1 and then again 0s. Basically shifting the 1 bit 20 places ahead. Which 2^20, and it about a million. Hence the shorthand. But it's bitshift for those who may be wondering.
@eduardojreis
0:39 - I think you meant this picture is a Multi-Layer Perceptron.
@trendshift-e4u
Question here : - Without threading the execution time was 126.13ms - with one block and 256 threads the execution time was 3.9ms - with 4056 blocks and having 256 threads in each block the execution time was 3.2ms How come having one million threads instead of 256 threads optimezes only 0.7ms... it should have been alot better number in think ?
@ScriptureFirst
3:50. Metcalfe’s Law.
@basilgeorge5657
If he only made course on how to present knowledge through video he could have avoided all issues happened in past.
@ahn_buguei
Wow. CUDA is not impossible. Thanks!
@lennartvandergoten6592
What about AVX, FMA, SSE. With them it would be a lot faster
@In-N-Out333
This guy is the ITT Tech of YouTube
@arsen_y5
good content, thanks!
@bhuvaneshs.k638
Hello Siraj..... love ur wrk...!! Can u do a video on new technique called Neural Structure Learning maybe give a benchmark b/w usual way n Neural Structure Learning
@tomtom-wv3hc
Great stuff Siraj :) absolutely amazing !!! Can you make a video on new technique called "Neural Network Surgery" ? All the best :)
@empiricistsacademy7181
Like/Dislike ratio is increasing, keep up the consistency Siraj. Nice to see you are continuing to make videos in areas you are interested in. It might be just me, but looks like you got a bit chubby over time, hope you are eating healthy and exercising. I have a weight loss motivational video on my channel if you need any motivation to get started.
@akshatshah4238
Kindly make a video on CUDA in python and the numba library.
@RobertMagier