Logo

Intel Tutorial- Thursday Afternoon - Shared screen with speaker view
Jeffrey Rodgers - Intel
03:22
jeff.rodgers@Intel.com
Rakshith Krishnappa - Intel
13:09
Link to signup for DevCloud account: https://intelsoftwaresites.secure.force.com/devcloud/oneapi
Rakshith Krishnappa - Intel
16:38
command to get training notebooks: /data/oneapi_workshop/get_jupyter_notebooks.sh
danconnors
18:04
What was the terminal command? /data/oneapi_workshop…
Praveen Kundurthy- Intel corp
18:22
Highlights poor performing loopsShows performance ‘headroom’ for each loopWhich can be improvedWhich are worth improvingShows likely causes of bottlenecksMemory bound vs. compute boundSuggests next optimization steps
Rakshith Krishnappa - Intel
18:35
/data/oneapi_workshop/get_jupyter_notebooks.sh
Praveen Kundurthy- Intel corp
18:41
sorry for the typo
Brandon Biggs - Idaho National Lab
21:54
Is there a way to use the c++ kernels instead of having to write out to a file and then execute?
Brandon Biggs - Idaho National Lab
22:55
Gotcha. Thank you
Ben Odom -- Intel
24:11
Please ask questions as we go, we will do our best to answer.
Martin Cuma
28:01
Can you please say a few words on how DPC++ differs, or is the same, from C++17, 20, 2x?
Martin Cuma
30:24
sounds good, thanks
Ben Odom -- Intel
47:46
Hopefully you all are having a good experience on the devcloud, let us know if you are not.
Thomas Hauser
56:53
The devcloud is great.
Ben Odom -- Intel
57:06
Please do not hesitate to ask questions
Martin Cuma
01:01:58
How do you embed the Advisor output into the Jupyter notebook? That looks nice.
Martin Cuma
01:02:49
Also, is this a new feature in the Advisor tool? I missed that at the beginning if you said that.
Ben Odom -- Intel
01:03:29
we point the markdown cell at the html output file
Ben Odom -- Intel
01:03:37
it's why we like jupyter!
Martin Cuma
01:05:48
Ah, OK, I see. I like the vectorizing and threading Advisor features so it's good to see how one can get the html report and use it like in Jupyter
Ben Odom -- Intel
01:06:34
vtune also has an html export feature, not as rich as the gui version but you can see that in our other module
Aaron Holt - CU Boulder
01:06:53
How does Advisor calculate performance for future/other devices? How accurate is it?
Ben Odom -- Intel
01:09:38
it's very accurate
Ben Odom -- Intel
01:10:32
it uses trip counts, loop correctness, memory access patterns and a roofline analysis for bandwidth
Ben Odom -- Intel
01:11:09
it works with gcc and visualstudio, but it is more feature rich with Intel compilers
Rakshith Krishnappa - Intel
01:11:51
you can specify the future device's frequency, EU count, DRAM, L3 cache sizes and accordingly Intel Advisor tool estimates the performance prediction.
Aaron Holt - CU Boulder
01:12:17
Thanks!
Matthew Sgambati - Idaho National Laboratory
01:24:09
Thanks! Great presentation!
Jeffrey Rodgers - Intel
01:24:38
jeff.rodgers@intel.com
Chris Chang - NREL
01:25:17
Which ISO C++ standard does DPC++ extend?
Chris Chang - NREL
01:25:49
C++17?
Chris Chang - NREL
01:26:16
Thanks!
Martin Cuma
01:26:23
Thanks, great talk.
Tzu-Chi Yen
01:26:30
Thanks!