Papers of the day   All papers

The Ecological Impact of High-performance Computing in Astrophysics

Comments

Roberto Trotta: "Educators may want to reconsider teaching Python to University students. There are plenty environmentally friendly alternatives." 🤔 Python is the more CO2-intensive and less efficient of the languages in astronomy, argues https://arxiv.org/pdf/2009.11295v1.pdf https://t.co/lfiB60kgnr

126 replies, 522 likes


JD Long: This sort of noise is remarkably silly. God help the astronomer who gets taught C++ instead of Python because her professor thinks that’s a more environmentally conscious choice.

30 replies, 325 likes


Bryce | BlackLivesMatter: This is a bad take. Developer productivity matters The C++, CUDA, Fortran, and Swift datapoints on that chart likely required more effort & there are fewer people with the skills to make that effort The code you never have time to write is infinitely slower than the code you do

19 replies, 152 likes


J. Daniel Garcia: Another reason to learn C++. C++ causes much less CO2 than other options.

7 replies, 78 likes


Dr. Kaitlin Rasmussen: oh gee huh I guess I’ll stop running *checks paper* stellar evolution models with python scripts 🙄🙄🙄

17 replies, 71 likes


Antifa Comptroller: Just think: if you’re an astronomer somehow hand-rolling your algorithms in python, you should stop, because you’ll be doing as much damage in a year as... 1/6th of a roundtrip flight from NY to London. https://t.co/kPL8dJ5pUe

3 replies, 63 likes


Dr. Phil Metzger: I knew there was a reason I still use fortran... (Well, probably it’s because I never had time to switch.)

11 replies, 45 likes


James: Any time I’ve worked on a project the Python-bound computations (e.g. pegging CPU at 100%) are replaced w/ a performant variants within few months as they’re most obvious performance bottleneck. But 9/10 components are not bound this way & it’s much faster to code them in Python.

2 replies, 31 likes


Dave 🦔: I don't think millions or billions of people learning to code in a less efficient language would do as much harm as completely unnecessary computation like high frequency trading and bitcoin mining... Or doing ML when you could use a deterministic rule...

3 replies, 30 likes


Sarah Kendrew: Yet another thread about the “your python code is killing the planet, stop teaching kids python” article. Maybe we could just post these many tweet threads to the arxiv as a rebuttal?

2 replies, 29 likes


David O'Regan: My Group and I almost exclusively use environmentally friendlier modern Fortran, to develop environmentally (much!) friendlier approaches to the quantum many-electron problem, to enable the development and discovery of materials for energy storage, solar cells, CO2 reduction, etc

1 replies, 25 likes


Alexander Paschall: Can someone please find out the C02 emission of a UE4 Blueprint, asking for a friend

3 replies, 21 likes


Dr Amruta 🌟🔭😍👩🏾‍💻: If you really want educators/universities caring for environment maybe first ask them to disinvest from fossil fuels.

3 replies, 20 likes


Catherine Walsh: Now I have a really good excuse for (still) using Fortran ...

2 replies, 17 likes


(((ل()(ل() 'yoav)))): these damn astronomers and their carbon emitting habits!!!1

3 replies, 11 likes


Maded #JoinAghamYouth: While big oil companies and huge ass countries are killing our environment, this guy is passing off computational speed of PLs as environmental friendliness skskskssk

1 replies, 11 likes


Leo C. Stein 🦁: In case you don't know what I'm subtweeting, the arXiv preprint is linked here (twitter poster is not paper author) https://twitter.com/R_Trotta/status/1309457430267523072

2 replies, 11 likes


Gerardo Cisneros: And this, boys and girls, is one reason why #Fortran is still the language of choice in #HPC. #FortranForever!

1 replies, 10 likes


Teppei Katori (香取哲平): Yes, #SuperKamiokande knows this many years and insist to use F77

2 replies, 9 likes


Karen: Dunno...I think in the US the carbon footprint from the drive to the lab probably dwarves that of any compute

1 replies, 8 likes


Stefano Sanvito: I'm a proud 90s Fortran man!!

0 replies, 8 likes


Chenna: Considering Python's capabilities, the outcome is obvious. If you are using Python for computational science and engineering, then you are doing it all wrong. This should encourage others to ditch #Python for computing intensive applications.

1 replies, 8 likes


Tom Ruen: Python is slow compared to everything else, but it depends what you're doing. Like if you're spending 20 minutes to write a Python script that is run once and finishes in 10 second versus 60 minutes for a C++ program that runs in 0.1 seconds, take your pick I guess.

1 replies, 6 likes


Christian Málaga-Chuquitaype: Interesting! The good old FORTRAN still rocks ;)

1 replies, 5 likes


Malcolm Fairbairn: 🤫😎😇⏩👍

3 replies, 5 likes


Albert Vilella: Wow! #nvidia #CUDA almost bookending the chart depending on single or multicore

1 replies, 5 likes


Chris Chapman: Ok, that nails it -- I will code my next Bayesian model and CFA in assembly language. Was on the fence.

0 replies, 5 likes


Jim Geach: But it’s a pretty disingenuous analysis - this demonstrating using Python to run an N-body simulation, so of course it’s going to be inefficient.

2 replies, 5 likes


Tobias Wood: I would love to see the Neuroimaging equivalent of this analysis...

2 replies, 5 likes


André David: Educators please make sure to teach that different languages and different libraries have different use cases. Also that comparisons like this one do not reflect real life codebases. Python’s efficient at gluing together disparate parts. (Tweet below not by the author)

1 replies, 5 likes


Nuclear Physics: Fortran for ever! Never ever tease me again! I am not old, I am ecologic! (Source: https://arxiv.org/pdf/2009.11295v1.pdf) https://t.co/VsKqgueOfn

1 replies, 4 likes


Energy Book Nerd: "The runtime performance/CO2 emission of #Python can be improved in a myriad of ways [#NumPy libraries]...but optimization is generally stopped as soon as the calculation runs within ... reasonable amount of time, such as the coffee-refill time-scale or a holiday weekend." 🤦🤣

0 replies, 3 likes


Concettina Sfienti: ...and they make fun of me when I use fortran for quick debugs!

0 replies, 3 likes


Néstor Espinoza: This recommendation starts getting ridiculous when the author states that “instead of python” you should use “numpy” as an alternative. That’s like saying “don’t use the sink for cleaning the dishes” —- “use dishwasher soap”.

0 replies, 3 likes


Dennis Mabrey: The environmental impact/carbon footprint of different programming languages written from an astronomy researcher's point of view... Let's just say Python isn't very green compared to C++. @lefticus @meetingcpp

0 replies, 2 likes


Tristan Moody: There are so many confounding variables in this graph I don't know where to begin. Are we talking compiled vs interpreted vs JIT compiled vs bytecode languages? CPU vs GPU?

1 replies, 2 likes


Marco Olivieri: @cjunkk @la_gata_lova fortran rulez

2 replies, 2 likes


James Monk: "The growing popularity of Python is disquieting." Wait until they discover how much javascript emits!

0 replies, 1 likes


José Vines: "There are several excellent alternatives to Python. The first choice is to utilize high-performance libraries, such as NumPy and Numba." Avoid using python by using python instead!

1 replies, 1 likes


Austin McShan: I think this kind of individualistic CO2 reduction approach is not conducive. Electrical production and large corporate energy use should be the focus. The consumer is not the major problem; large scale supply chains and fossil fuels are.

0 replies, 1 likes


Fabio Fracassi: Performance should be measured in watts/task...

0 replies, 1 likes


Tiago: Scripted languages, such as Python and Ruby are way more easy to write, read and understand code written in Python than in compiled languages such as C/C++. But they should reconsider teaching Python in favor of Ruby 🤓

0 replies, 1 likes


Stephen Eglen: Considerations of language choice based on CO2 production... hate to think where #rstats might be! But I'm guessing Julia language will fit in the bottom left nicely...

1 replies, 1 likes


Florian Wagner, PhD: The bigger problem is when researchers think that students should learn programming by starting with some specialized language that happens to be popular in their field, because that shows that they probably aren't very good programming teachers. 😔

0 replies, 1 likes


Dr Owain Kenway: CPython is crap. We know this.

1 replies, 0 likes


Content

Found on Sep 25 2020 at https://arxiv.org/pdf/2009.11295v1.pdf

PDF content of a computer science paper: The Ecological Impact of High-performance Computing in Astrophysics