iiinova — where your digital self keeps evolving

Giuseppe Bilotta

@giuseppebilotta.fediscience.org.ap.brid.gy

10 followers 0 following 200 posts

Researcher @ INGV-OE. Opinions my own. Bluesky users: if you wish to follow/interact with me, please follow @ap.brid.gy first or your interactions will not […] 🌉 bridged from ⁂ fediscience.org/@giuseppebilotta, follow @ap.brid.gy to interact

Posts Replies Media

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 10h

What a wonderful piece of software #VisiData is, although like #Vim it needs some getting adjusted to.

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 02/10/2026

The more I learn about #OpenACC the less I like it, and the less I understand why it's a thing in the first place. (No, I know why it's a thing; but really wish it wasn't.)

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 02/10/2026

An update about the #Clang optimization that broke my code: I have submitted a bug report github.com/llvm/llvm-project/issues… and apparently it has been acknowledged as a bug and is currently being investigated. This is a pleasant surprise. Juggling uninitialized values is […]

fediscience.org

Original post on fediscience.org

000

Reposted by Giuseppe Bilotta

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 02/10/2026

I have just pushed an update to #clinfo with preliminary (untested) support for a bunch of QCOM extensions. If you have an #Adreno #Qualcomm #Snapdragon device with #OpenCL support, I would appreciate some #testing (and a sample output, especially if your device/driver version supports the […]

fediscience.org

Original post on fediscience.org

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 02/10/2026

I'm not saying vendors should provide me with (access to) every device that would allow me to test clinfo support for each of their extensions (although, hey, I'm not going to say no to an AMD GPU with cl_amd_liquid_flash support), but they could at least document the size and type of the new […]

fediscience.org

Original post on fediscience.org

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 02/10/2026

I have just pushed an update to #clinfo with preliminary (untested) support for a bunch of QCOM extensions. If you have an #Adreno #Qualcomm #Snapdragon device with #OpenCL support, I would appreciate some #testing (and a sample output, especially if your device/driver version supports the […]

fediscience.org

Original post on fediscience.org

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 01/10/2026

Thanks, we're working on it with the package author but it isn't as simple as it seemed initially. (BTW, please follow @ap.brid.gy for easier interaction w/ me and possibly other accounts federated from the Fediverse).

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 30/09/2026

OK for anyone who is curious about this #Clang #compilerBug, I've set up a godbolt: godbolt.org/z/bTvaEf9sq Compare the assembler output from GCC, or from Clang version 18 and earlier, to that of Clang 19 and later. Can you spot where the issue is, without running? Hint: it's in how […]

fediscience.org

Original post on fediscience.org

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 30/09/2026

Found the issue, and it's most definitely a miscompilation, a greedy optimization caused by uninitialized memory that however should be NEVER accessed (if the conditional that Clang is optimizing out was actually respected). But at least this should allow me to get a repro much faster than […]

fediscience.org

Original post on fediscience.org

100

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 30/09/2026

The cellphone has finally finished upgrading to the latest latest firmware revision (September) and I was in fact able to find the missing library. So it's there, but the opencl-vendor-driver installation script does not find it, so I've submitted an issue ticket to termux-packages. Looking […]

fediscience.org

Original post on fediscience.org

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 29/09/2026

Oh wow managed to reproduce the issue with Clang 21 on my laptop! This is opening to some interesting debugging opportunities.

100

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 29/09/2026

a printf("%d\n, numParticles) before the loop that compares the indices to numParticles makes it behave correctly. How.

100

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 29/09/2026

This is the weirdest bug I've ever come across. Comparing against a constant with the same value as numParticles, regardless of signedness, works. So strange.

100

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 29/09/2026

From time to time I test (for fun, mostly) whether GPUSPH still builds in Termux. It's an interesting stress test because (1) it's the CPU backend, that we rarely use ourselves (2) it's on an ARM architecture (3) the performances are abysmal. Point (2) in particular has been very interesting in […]

fediscience.org

Original post on fediscience.org

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 29/09/2026

The #OpenCL on #Android saga continues. I updated my phone to the latest firmware: it had been nagging me for a while, and security-wise it's a generally a good idea anyway; plus, I was curious if the firmware update would bring with itself a driver update too for the GPU. And well, it did, but […]

fediscience.org

Original post on fediscience.org

100

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 27/09/2026

OK, so I'm found a bug in my checks (good), but the CL_DEVICE_MEMORY_CAPABILITIES_IMG device query that should be exposed by cl_img_mem_properties gives me a CL_INVALID_VALUE (-30) error, because the properties was added only in later revisions of the extension. This means I should actually […]

fediscience.org

Original post on fediscience.org

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 27/09/2026

Jokes aside, apparently the termux developers have thought of everything! There's an "opencl-vendor-driver" that exposes the vendor driver to the termux installation. However, when this is installed, clinfo should be compiled “regularly”, i.e. without the Android termux hack I had implemented […]

fediscience.org

Original post on fediscience.org

200

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 27/09/2026

The I consent / I don't meme, with me, Imagination Technology and Motorola consenting, and Google saying “I don't”.

100

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 27/09/2026

Oh, actually my cellphone (a Motorola g56) does have a PowerVR GPU, and the OpenCL library to go with it, but it's not accessible from termux. I do not like how unnecessarily restricted Android is becoming.

100

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 27/09/2026

I've just pushed a couple of commits in #clinfo that introduce support for a couple of Imagination (#PowerVR I think?) extensions, although I don't have any device that supports it, so if any of my followers or extended network happen to have one, I'd be curious to see the output if they could […]

fediscience.org

Original post on fediscience.org

100

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 26/09/2026

Thanks for the recommendations received so far, they gave me a pretty good idea of how to proceed. I'm going to trigger all the fences I know about in existing models, and hopefully some I don't know about yet. My AGENTS will be extremely political, sorry if it ends up ruffling some of your […]

fediscience.org

Original post on fediscience.org

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 26/09/2026

I've been pondering about moving my #FLOSS projects off from GitHub. I'm slow in taking action, but since I've been working on #clinfo a lot these days, I'm thinking this would be a good opportunity to make a, shall we say, less slop-friendly hub the primary one (I plan on keeping the GH one […]

fediscience.org

Original post on fediscience.org

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 26/09/2026

I'm distracting myself from some obnoxious work & life stuff by going over bugs and feature requests for clinfo and looking into finalize #OpenCL 3.1 support (which from what I can see just makes some of the extensions obligatory, but doesn't actually introduce new device properties itself)

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 24/09/2026

YouTube force-classifying videos as “shorts” and preventing the uploader from changing this classification is a misfeature.

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 23/09/2026

@karolherbst is there a way to make rusticl less chatty? Getting pci id for fd 14: 10de:25b8, driver (null) pci id for fd 15: 10de:25b8, driver (null) === Rusticl warning: Patched Mesa libclc not detected. Upstream libclc may contain known bugs or breaking changes and isn't guaranteed to work […]

fediscience.org

Original post on fediscience.org

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 01/09/2026

Sono le 7:32 del mio primo giorno di rientro dalle ferie e voglio già tornare in ferie. #ferie #lavoro

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 12/08/2026

No #eclipse here. Sun set before the moon had the time to cover it, it was cloudy anyway, and when it didn't rain rain, it rained volcanic ash. #Etna #MtEtna #eruption

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 11/08/2026

I've mentioned in the past that among other things, we rely on the #Meteosat geostationary satellites to monitor #MtEtna. Even though the spatial resolution is low (the sub-satellite pixel has a nominal resolution of 3km), these satellites provide information at a high temporal rate: MSG […]

fediscience.org

Original post on fediscience.org

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 10/08/2026

So, I'm spending my vacation preparing #lava flow inundation #hazard maps for the ongoing #MtEtna #eruption, how's your summer going? #Etna

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 09/08/2026

Come giustamente ha detto la mia collega nella nostra chat di lavoro, «Ma sempre a #Ferragosto?» #eruzioni #eruzione #Etna #MtEtna

Attività termica sull'Etna vista da SEVIRI nelle ultime 24 ore.

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 31/07/2026

Need to prepare the images for the extended abstract for the 20th #SPHERIC. My mind is already on vacation. These are going to be the hardest-to-make images I've ever made.

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 31/07/2026

Here we go again … #Etna #eruption

Plot of the radiative power from Mt Etna as detected from MSG SEVIRI with our HOTSAT algorithm. There has been a huge spike (peaking at 10GW) late yesterday evening, and it's still going strong even if decreasing over time (down to 6GW)

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 29/07/2026

Our latest submission to JCP has been rejected as “out of scope” because the editor has interpreted the paper as being “an application”. A curious decision, since the paper is on how to compute acoustic terms efficiently and with higher precision compared to how it's usually done, and there are […]

fediscience.org

Original post on fediscience.org

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 27/07/2026

I want to simulate this with GPUSPH dmv.community/@jcrabapple/116988909…

000

Reposted by Giuseppe Bilotta

2 Spoopy 4 You @jcrabapple.dmv.community.ap.brid.gy · 27/07/2026

🌊 In September 2023, seismometers around the world detected a mysterious signal repeating every 92 seconds for nine straight days. The source: a standing wave trapped inside a Greenland fjord. Here is what happened. A glacier in East Greenland had been thinning for years, no longer supporting […]

Original post on dmv.community

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 10/07/2026

From time to time I get a student whose exam project is good enough to be a thesis for their degree, and I feel embarrassed because mine is only a 6 CFU subject and I the only I can do is give them top marks. (Yes, they usually already have a thesis supervisor.)

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 09/07/2026

After sitting for two months in the editor's hand, our latest manuscript has been finally sent out to the reviewers. Coincidentally, just a couple of days after I sent an email asking about the status of manuscript. #academia #academicPublishing #publishOrPerish

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 05/07/2026

OK I'm going to need ask the Fediverse for some math help here, because I have a feeling I'm forgetting something obvious and can't find the resources I'd need to answer my own question. Let's say I have an underdetermined linear system. The methods generally employed to solve these find the […]

fediscience.org

Original post on fediscience.org

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 22/06/2026

I think I got my first #genAI / #LLM exam project submission. I do not remember if I made a clear policy statement about this during my lessons, and the students are foreigners (their native language is neither Italian nor English), so I can understand that they might have felt that having an […]

fediscience.org

Original post on fediscience.org

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 08/06/2026

@simontol uno di questi giorni i giornalisti impareranno a smettere di usare l'espressione impropria “scossa di assestamento” per usare quella corretta (replica), ma il giorno evidentemente non è oggi 😞 (Notavo peraltro che le repliche nel senso sismologico hanno una loro pagina nella […]

fediscience.org

Original post on fediscience.org

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 08/06/2026

Does anybody happen to know the power series expansion for arcsin( k sin(x)) or do I have to derive it myself? #askFedi #lazyFedi #math #maths

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 27/05/2026

In a meeting to organize a new project, and I keep hearing “Mastodon” instead of “milestone”. I suspect my brain isn't entirely on the job today.

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 21/05/2026

I'm reading news of someone in academia complaining about the new no-AI policy by ArXiv that requires something as outlandish as co-authors to take responsibility for what gets published in a scientific paper and … I'm sorry, how is this unexpected? Academic publishing has a lot of issuers, but […]

fediscience.org

Original post on fediscience.org

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 16/05/2026

This was extremely frustrating because parallelizing step A would bring runtime down from 10+ seconds to under 1 second (yes, it's a beefy machine), but it was definitely not worth it to do that if step C would then be some 10 times (if not more) slower, with low CPU usage to boot. I have not […]

fediscience.org

Original post on fediscience.org

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 16/05/2026

I had a weird experience with #Julia and #parallelization. I had to collect all the pairs of items in two sets (S and G) that satisfied a particular relation function R. For Reasons™, it's easier to find these pairs looking at each element of S and find which elements of G are in relation with […]

fediscience.org

Original post on fediscience.org

100

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 15/05/2026

Has this always been a thing? I just got an email from a scientific journal publisher, informing me that an article of mine from 2024 was “among the most viewed articles” (in 2025, of the articles published in 2024). And I mean, OK, I'll take it, but I'm sitting here wondering: is this […]

fediscience.org

Original post on fediscience.org

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 14/05/2026

I'm currently looking for a way to bring together by experience with #SmoothedParticleHydroynamics and #remoteSensing, and as part of this process I also wrote a converter (in #Julia, because this has to fit within some existing code which I wrote in that language) that allows me to export […]

fediscience.org

Original post on fediscience.org

000

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 12/05/2026

A roomful of #ComputerScience professors fighting #GoogleGemini to do something that could be achieved with a sort -R | head -n 4 *sigh* #genAI #LLM

Giuseppe Bilotta @giuseppebilotta.fediscience.org.ap.brid.gy · 11/05/2026

For work I'm refreshing my knowledge on #OpenACC and I'm still wondering why anyone would want to touch that.

000

Reposted by Giuseppe Bilotta

Khronos Group @thekhronosgroup.fosstodon.org.ap.brid.gy · 05/05/2026

OpenCL 3.1 is here. The Khronos Group has moved several capabilities into the core spec, including SPIR-V kernels, subgroups, and integer dot products. Also includes improvements to the memory model and synchronization, plus better alignment with Vulkan via device UUID queries […]

Original post on fosstodon.org