Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
milancurcic
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
13 ms
·
61.
▲
First Year of Fortran-Lang
(medium.com)
7 points
by
milancurcic
6y ago
|
0 comments
62.
▲
by
milancurcic
6y ago
Knuckles, wrists, elbows, toes, ankles, knees. My wife hates it, I can't live without it.
63.
▲
by
milancurcic
6y ago
Thank you for your support! I'm happy that it was helpful for you. If you have any questions or get stuck at any point feel free to email me at milancurcic@hey.com. I'll be happy to help.
64.
▲
by
milancurcic
6y ago
Thank you! :) I totally get it. Same with my book: a new revision of the standard (Fortran 2018) was published half-way through the writing. However, thanks to the glacial pace of the Fortran standards development, I was able to plan for th
65.
▲
by
milancurcic
6y ago
Regarding Con 5), it's generally true with some exceptions. I wrote a Fortran book ( https://www.manning.com/books/modern-fortran ) and expect it to stay fresh for at least several years thanks to the slow pace and
66.
▲
by
milancurcic
6y ago
I've been enjoying Stadia on a 2013 Linux PC with a ~30/8 (down/up) Mbps wi-fi. Both Destiny 2 and AC: Odyssey work great. Occasional hiccups, yes. No more than a few per hour. It doesn't bother me. The fact that I'
67.
▲
FortranCon2020 Presentation Recordings
(youtube.com)
8 points
by
milancurcic
6y ago
|
0 comments
68.
▲
by
milancurcic
6y ago
Standard? True, undefined. Compilers, however, happily overflow unless you tell them to catch it at run time. So, my original statement is at least half false. :)
69.
▲
by
milancurcic
6y ago
Fortran predictably overflows the result, in contrast to fptoui which gives a poison value. I agree that even a predictable overflow can bite you on the behind. But it's better than undefined behavior. I'm not a fan of saturated c
70.
▲
by
milancurcic
6y ago
Not quite, but at optimization levels set by default WRF configuration for each compiler (definitely not a fair comparison): gfortran: -O2 -ftree-vectorize -funroll-loops ifort: -O3 I don't have the timing results anymore. This was in
71.
▲
by
milancurcic
6y ago
Nice article. I'm curious which Fortran compiler did you use? One key difference that I see in the context of this article is that Fortran is natively parallel (both shared and distributed memory) via coarrays, teams, and events, where
72.
▲
by
milancurcic
6y ago
Indeed, and this depends between applications. For the WRF (Weather Research and Forecasting) model, I get 3-4x speed up with Intel Fortran compiler over gfortran. I don't see where the Author mentions the Fortran compiler used, though
73.
▲
History of coarrays and SPMD parallelism in Fortran
(dl.acm.org)
2 points
by
milancurcic
6y ago
|
0 comments
74.
▲
by
milancurcic
6y ago
Of course these are anecdotal. I got my IdeaPad U430 in 2014. Same year it slid off my bunk bed (6' high), first hitting the corner of my dresser before it fell on the hardwood floor. It cracked its spine in half and got a dent in the
75.
▲
Fortran-Keras Deep Learning Bridge for Scientific Computing
(arxiv.org)
7 points
by
milancurcic
6y ago
|
0 comments
76.
▲
Running for International Fortran Committee Chair
(ondrejcertik.com)
7 points
by
milancurcic
6y ago
|
0 comments
77.
▲
by
milancurcic
7y ago
Lead author here. We found further evidence of drag saturation in hurricane winds, which was independently discovered in the field and laboratory in 2003. Drag saturation means that the air-sea drag coefficient ceases to increase with wind
78.
▲
Revised estimates of ocean surface drag in strong winds
(arxiv.org)
2 points
by
milancurcic
7y ago
|
1 comments
79.
▲
by
milancurcic
7y ago
tl;dr If the used cup won't touch the pouring utensil, it's OK to reuse; otherwise, you get a new cup. In retrospect, this is what I observe as a customer in most cases.
80.
▲
by
milancurcic
7y ago
I had no idea and it makes sense. Most coffee shops and bars that I go to (South Florida) don't follow this.
81.
▲
by
milancurcic
7y ago
When I asked a Starbucks barista to fill my coffee cup (now empty) with water, it didn't compute. He insisted that I should get a fresh cup with a lid and straw and all just for water. It took a few back and forths but in the end he ag
82.
▲
by
milancurcic
7y ago
Some of my best ideas and realizations in science research and business I got while exercising, sleeping, or in the shower.
83.
▲
by
milancurcic
7y ago
Original thread: https://news.ycombinator.com/item?id=13024702
84.
▲
by
milancurcic
7y ago
> I haven't touched any later standards after 77; is Fortran still technically not Turing complete? Fortran has been Turing complete since its inception in 1957 :).
85.
▲
Show HN: A minimal Fortran TCP client and server
(github.com)
110 points
by
milancurcic
7y ago
|
45 comments
86.
▲
Implicit None and Carry On
(milancurcic.com)
2 points
by
milancurcic
7y ago
|
1 comments
87.
▲
Proposals for the Fortran Standard Committee
(github.com)
2 points
by
milancurcic
7y ago
|
0 comments
88.
▲
Cloudrun Forecast Scheduling
(medium.com)
1 points
by
milancurcic
7y ago
|
0 comments
89.
▲
by
milancurcic
7y ago
Yeah, we joked about it on their launch day, "welp, there goes our SEO!" :) Probably not a big deal as we're in a quite narrow niche.
90.
▲
by
milancurcic
7y ago
At peak capacity, bare metal pays off soon. Even when they run at near 100% of CPUs, weather models never run 24/7 but only fraction of the time. Examples: A national weather service (any country) may run at 20-50% capacity. In a resea
More ›