Piero Bosio Social Web Site Personale

Days since my threadripper workstation was turned into a single core babby by NTFS filter driver: 0

So fundamentally speaking, bulkheads are deliberately introducing significant inefficiency into your system, in order to improve its ability to survive in degraded mode. This is a really good idea, but you need to make people actually *utilize* it somehow or they will take shortcuts. Doing this stuff 'properly' is something which is going to look bad on a cost sheet in the short term, and then in the long term if it works, is going to look like nothing happened.

even if i imagine myself the most business brained shitstain motherfucker middle manager boss in the world I feel like i'd be literally yelling at these people to get back to fucking work and stop wasting time.

The current scene in all its glory. big stonks.

segfault #2 in slang shader compiler unlocked! woot!

if only there were some kind of 'scalar load' instruction

btw has anyone actually written anything cool in unreal's verse, that's public? just out of curiosity, it's been a while since i've heard anything on that front but i assume it's mostly because im not writing fortnite mods

good news everyone, i've been promoted to "Expert Programmer" at work. this means my posts are all made by an Expert and thus can no longer be challenged by those who are not at least Expert Level. unfortunately one of my colleagues also pointed out that there exists Senior Experts, which I am not. However soon, with one more title, i will beenough

omg the fucking thing is never initalized. c is such a joy

omg people keep posting the llm generated trash article. okay well lets try to take a useful lesson from it:WHEN YOU LEAK MEMORY IT LEAKS UNTIL YOU HAVE NO MEMORY LEFT SO THE BIG NUMBER IS JUST AN APPROXIMATION OF YOUR COMPUTER'S TOTAL RAMThis is of course, a minor thing, but it's important to keep in mind when reading clickbait that they're just using numbers to bamboozle, not to inform.

I wish there were a way to just annotate your loop with "please specialize the entire loop for this interior condition", instead of writing out two versions or doing some cursed generics thing.

thankfully at the spir-v level this is answerable:1) shifting out of range produces an unspecified value2) if you annotate a shift instruction with `NoUnsignedWrap` or `NoSignedWrap`, then it becomes undefined behavior.at the slang level it's specified as: <crickets>edit: Reading the code a left shift in slang ir just produces OpShiftLeftLogical, so in practice, on the spir-v backend, it should behave as 1, barring any potential shenanigans in the slang optimisation pipeline itself.

In vulkan subgroup indices are not guaranteed to map to local invocation indices...e.g. for wave32 and a 64 lane workgroup this is NOT guaranteed```workgroup: [ 0 ... 31 32 ... 63 ]subgroup: 0: [ 0 ... 31 ], 1: [ 0 ... 31 ]```so to be compliant you need to manually reconstruct the local invocation id from the subgroup index and subgroup size. e.g.`let local_id = subgroup_index() * subgroup_size() + subgroup_id();`So,1) why the heck (intel? arm?)2) how 2 HLSL?

i'm not really sure if there's a better way to handle this kind of stuff as part of a Social Website, which sits somehow between obnoxious "actually ur posting cringe" public replies, vague subposting, and ignoring it. it's not really possible i think to expect people to always notice that they're falling victim to what is more or less a scam. :/

Does HLSL/slang have no equivalent to `gl_SubgroupID` / `gl_NumSubgroups`, or am I missing something obvious?

rickrolling is one of the most continually amusing things which exist, having gone through many cycles of cringe and haha. i want to know if it can be killed. can the power of microsoft make anything unfunny? is there a limit?

@zeux @aras you can kinda understand why people are so hesitant about quote posts on here when you look a bit at the engagement farming over yonder. It was a bit like that when I left twitter and I can only imagine it's getting worse.

There's a dish on the menu at our regular Indian takeaway place called "Ghandi's Revenge". I have not yet tried it

The reason you can iterate easily in shadertoy is because the data model it's giving you access to is extremely limited. And so shaders tend to reconstruct their entire state per pixel per frame. That's fantastic for low iteration times, but it's not fast. Doing this on the CPU without any data model is immediately the worst case for parallel debugging, with or without a stack, because any job can access any data. Regular bindful shader programming has a lot more information about data access.

revert your broken code challenge(please please just revert your broken code when it's broken, you can submit it again after it's not broken)