milancurcic

15 ms·

121.
by milancurcic 9y ago
Thank you for the feedback! At this point we don't target end consumers, but rather small businesses that create these kind of products. However, we do get a lot of requests for specialized and intuitive graphics/visualization and
122.
by milancurcic 9y ago
Unfortunately, at this early stage, Cloudrun still requires some level of WRF expertise, needed to create these input files. We are working on automating these steps so you don't have to, but it will take another 2 months or so. In the
123.
by milancurcic 9y ago
At the moment, I can't say much about the wave model that we will offer, other than that it will support both global and any custom regions, and it will not be WW3. NOAA already does as good job as you can get with WW3.
124.
by milancurcic 9y ago
For now, we only offer ARW in real data case mode (em_real). We will slowly enable other modes and WRF components if there is demand. All runs go on Intel Xeon Platinum 8168. We currently restrict the execution to shared-memory mode while w
125.
by milancurcic 9y ago
It's possible, we had a few people ask for MPAS already. Beyond WRF, the priority will probably be on surface wave modeling, but ultimately prospective users will decide. Do you have preference for MPAS Atmosphere, Ocean, or both?
126.
by milancurcic 9y ago
Fantastic, thank you, I will follow up!
127.
by milancurcic 9y ago
Nice! The terrain issue that you mention is the limitation of the grid resolution. For example, HRRR, the operational model used in the charts in your link, has 3 km horizontal spacing. Individual peaks and slopes are poorly resolved at tha
128.
by milancurcic 9y ago
Kudos for sticking through with it for your class project! I have been working with and have contributed to WRF code since 2009, and it's sure not straightforward. Porting it to different HPC systems can be especially painful. This pai
129.
by milancurcic 9y ago
Thank you for reporting the CC issue. We are looking into it. Also, great to know about your use case! We have been looking into applying this for snowfall prediction for ski resorts, but I wasn't thinking about mountain expeditions in
130.
by milancurcic 9y ago
Thank you, mino! At this point, we are targeting existing WRF users and newcomers into the field. These are either academic meteorologists without access to big hardware, or SMBs with focus on weather intelligence for agriculture, commoditi
131.
by milancurcic 9y ago
Cofounder of Cloudrun here. Cloudrun is a SaaS platform for the Weather Research and Forecasting (WRF) model. WRF is by far the most widely used numerical weather prediction model, but it currently requires access to supercomputers and doma
132.
Show HN: Cloudrun – Numerical weather prediction in the cloud(cloudrun.co)
41 points by milancurcic 9y ago|37 comments
133.
Doctor Fortran in “Eighteen Is the New Fifteen”(software.intel.com)
1 points by milancurcic 9y ago|0 comments
134.
by milancurcic 9y ago
Yes it can, in which case you embed OpenMP parallelism in low-level library calls, and is another way to accomplish what the article suggests. But it's still a fine-grained parallelism so numpy can match plain Fortran performance in
135.
by milancurcic 9y ago
False. OpenMP does shared-memory parallelism and has nothing to do with LAPACK. numpy is completely serial.
136.
by milancurcic 9y ago
> Once you’ve set up Travis, it will automatically test every commit to every branch Small correction: It will automatically test the latest commit on every push to every branch
137.
by milancurcic 9y ago
I have used the multipart package: https://github.com/defnull/multipart Does your library add have any extra functionality beyond those in multipart? Specifically, is it possible to get access to the file-like object b
138.
by milancurcic 9y ago
wavy is a spectral ocean wave modeling framework. It is a Fortran library you can use to simulate ocean surface waves. Why wavy? Existing ocean wave models (e.g. Wavewatch III, SWAN, WAM) are built as opaque black-boxes, and they are diffic
139.
Show HN: A spectral ocean wave modeling framework(github.com)
1 points by milancurcic 9y ago|1 comments
140.
by milancurcic 10y ago
Thank you jerf and sampo for your very helpful comments. This is something that I had on my mind during the implementation but focused primarily on expressiveness. While the array slicing in Fortran is O(1), the implementation of tail does
141.
by milancurcic 10y ago
Unfortunately, Fortran standard does not provide good support for generic programming so some level of repetition is necessary :/. I don't see this as a major issue though - boilerplate coding happens once. In any case, pick any l
142.
by milancurcic 10y ago
Thank you for the kind words. I cannot comment on the performance hit yet but this is something I plan to examine soon. Based on experience I can guess that the performance decrease from using recursion could be about an order of magnitude
143.
Show HN: Functional programming for modern Fortran(github.com)
98 points by milancurcic 10y ago|31 comments