วีดีโอ

We are good with CMake 3.23.0 or before. We have some software requiring 3.23 so it is widely available

A truly awful presentation 😅

Is anything happening regarding SWMR support for Windows? Is the VFD SWMR project still alive? Will this feature gain Windows support?

Can you please edit the audio of this video and re-post it, it is bot clear at all, so much echo which makes it hard to hear. Thank you.

How exactly do you guys support dynamically resizable datasets? You can't append data to HDF5 without the code getting messy.

The irony of this. Just make sure your format is ACID compliant. That's it. You won't have these problems

Save yourself the hassle. You can't append data reliably in HDF. Workarounds exist, but they fail if there's a power cut, corrupting your data. A format that can't safely handle incremental updates is pointless for dynamic datasets.

nice video

Thank you for the interesting talk! Is there a way to show multiple linear signals in one plot?

Thank you for this very useful video!

Thank you for this very useful video!

Thank you for this very useful video!

Thanks , How i solve this error in my hecras model please tell mes something. hdf5-diag error detected in hdf5 (1.8.11) thread 0 #0000: ..\..\ srcH5p

Thank you so mcuh!

Could you do a video going over setting user credentials and api key in the configuration.yml (and/or other) config files, then building the project? Are you able to update these configurations this way without building the project, or do you have to use hsconfig for something like that? Could you also give some examples of the get, put, post, and delete calls with curl commands after the user auth is set? Thanks!

,!!!!

finally building the good compressors in! looking forward to it and paraview. Bitshuffle+zstd is one of the most valuable ones I hope that becomes a builtin. Definitely interested in GPU / overriding a plugin type implementation in some way too. Maybe the client is the side that would do that though to do it targetted.

Direct link to slide deck: www.hdfgroup.org/wp-content/uploads/2023/08/HUG23_FlashX_Compression.pdf

"Promosm"

I've had the most difficult time using hdf5 in C/C++. I always resort to using xml, unfortunately.

Do you have a script that runs through repacking the files appropriately and using h5py in the appropriate ways to set the cache, etc?

I hear XDMF and I think paraview. Using the CPUs hard should have effect on the NVidia drivers ability to keep the GPUs fed with low latency or do other bookkeeping so assuming the CPUs are doing nothing really doesn't work out in practice IMO. Anyway I once did paraview with a few VFD support patches and it worked. It took longer to open but didn't need the copy. So the question is can HDF internally speed up the opening on distributed filesystems where 1024 ranks or something like this have their single private datasets so no one needs to copy for a single file view.

why arent image max and min not saved as attributes?

Question: why HDF5 ? Can't one simply use the filesystem instead ? It provides all the same capabilities -- metadata, heterogeneous data, big data, fast i/o, cross platform, easy sharing (tar, zip, etc) -- and doens't require learning something new (and a tutorial), maintenance HDFView is not working for my h5 file, nor is HDFCompass; Libraries are not well supported in every language; etc. Go figure ... That reminds me of the quote: “The bureaucracy is expanding to meet the needs of the expanding bureaucracy" ... :D

thanks a lot that's very helpful

Incase anyone else runs into an error following what appears to be the easiest and most simplest example of how to use HDF5, be sure to set the "mode" in the call to "h5py.File" to write. In other words, change "f = h5py.File("file5.hdf5")" to "f = h5py.File("file5.hdf5", 'w')".

Is there a new version of Kita Server available on AWS somewhere?

At HEPS, we are using SWMR as well in file writing

yes to simplifying+dramatic speedup of metadata cache. Sounds like that will give you guys so much back bugs wise and performance wise. Definitely question the heft of your c constructs, strongly recommend experimenting with the many many high quality c++ concurrency datastructures even if you recapture it later in c after you figure out what works.

I have joined a new job where I'm required to use the rhdf5 package. However, I have only always learned with CRAN. I can't seem to be able to install these packages despite multiple tries. It seems there are limited resources on rhdf5 to learn from. How do I install it ? how do I work with it ? what is the pattern of code it uses, because it seems very different from CRAN. Can you please share resources for the same ?

extremely quiet audio

Good explanation, but recording volume is extremely low. Maybe you can reupload with proper audio volume

Matlab uses internally hdf5 where images are also stored inside it but I’m unable to extract data from it using h5py module in python if you read it in time please tell me how to read that

Hi, thank you for the information. You saved me. Best wishes.

Thanks for the clear presentation. It seems to be the ideal format for my task.

thanks for the explanation but is would be better if you slowed down while you are speaking.

thanks! that's a very clear explanation!

thank you! maybe even it's too clear introduction xD

Nice talk!

Hi, I'm trying to build Parallel HDF5 on MacBook, but mpich my Mac installed was not correct, the mpi I/o for mpich shall be supported by romio, but even if I type --enable-romio for configure of mpich, it did not generate the important library libmpio.a, is there any solution? Thanks a lot!

Hi is this stable enough for operation ?

@7:40 Is the chunk size implicitly selected?

I can't tell you how frustrating HDF5, h5py and hdfql all are. Ridiculous level of complexity with very poor documentation. I'm trying to read a variable length of compound "records" from hdf5. h5py segfaults - can't handle this. hdfql - how do you even get at this once you have the cursor pointed to the first record ? e.g., how to parse the compound data from the datatype ? I see no doc/example anywhere.

Terrible audio, unfortunately,

Thank you very much for the help, Gerd!

I tried using h5cpp -- the first talk -- but there are some functions of documentation that are not really implemented and some examples end in segmentation fault. Example h5::gr_t group = fd_var["/path_to_group"]

Could you show a video working step by step with HDF5 in python or R?

lovely!