Stubbertville

@1600FILMZ It’s wild isn’t it? No better ad for payphones and landlines lmao

It’s been 6hrs and I still have no service

I pay Verizon $200+ a month for it to have an outage and not work at all ????

@32nds Long shot, can I get your autograph? I’m a huge fan of your work and have seen the halo 2 documentary you were in a dozen times lol

The muscle memory from this clip hit me like a truck. I must have rebuilt and tuned this encounter 1000+ times working on AI issues.

@falco_girgis Would this help most 3d games? Any of jnmartins projects or gta vc?

Holy CRAP, I just achieved a freaking 1.98x speedup for rotating a 3D vector by a quaternion on the Sega Dreamcast by using vectorized SH4 inline assembly over a standard C implementation! This one I am OVER THE MOON WITH. The SH4 has historically been said to only accelerate vector and matrix operations, with a lack of HW acceleration for quaternions traditionally being called "lamentable" due to quaternion math not becoming mainstream until a bit after the console was created. ...bullshit. Yet again, I was able to accelerate quaternion math using the SH4's vector instructions and FPU, this time for SUBSTANTIAL performance gainz. The trick? Why I was unable to achieve such gainz for so long? I spent so much time hyper fixating upon what has historically been accepted as the "optimal" floating-point algorithm for performing a quaternion rotation, based on minimal number of FP computations... The problem? What is considered "minimal" for scalar FP math is not necessarily minimal when you have SIMD and vector instructions performing multiple FP operations in parallel! One day I was verifying my implementation, when I ran into an alternate representation of the same operation, which was derived using a few linear algebra properties, yet was still mathematically equivalent to the original. This one being called "somewhat optimal," and "fairly fast," but was said to still be doing more work than necessary, compared to my previous implementation... But the majority of the computational work? Is performing the dot product operation between 3D vectors! Which we all know, with FIPR, is a single instruction for us to accelerate on the SH4! NOW, looking at the code, you can see the traditional algorithm on the top, implemented by the Vector3RotateByQuaternion() routine. This routine was taken from raylib--not because I'm picking on raylib, but because they typically have some of the prettiest, most straightforward implementations of linear algebra routines in pure, cross-platform C. Not only that, but in my experience, the way raymath is written also facilitates optimal code generation from the compiler. So I'm using raymath's routine here, because it represents a competent, vanilla, no bullshit C implementation to benchmark against. Now look at the balls-out SH4 optimized routine I implemented for SH4ZAM on the bottom pane. I am exploiting: 1) FIVE freaking 3D dot product patterns which I'm able to accelerate with FIPR either directly or through calling my special-case, chained dot-product routine, shz_vec3_dot3(), which carefully pipelines multiple FIPR calls together to calculate multiple dot products between a constant vector and batch of others. 2) CAREFUL register pinning, so that the compiler understands not only where operands must go for properly using them with the vector instructions, but also for knowing not to reload the invariants into FP regs across SH4 assembly boundaries. The end result? The pane on the right shows the execution time for multiple invocations of the two routines with the same operands, using the cycle-accurate performance counters on the SH4 CPU. Total gainz? About 100 clock cycles! 💪 GitHub commit and source code: github.com/gyrovorbis/sh4…

@electimon Hell yeah! Telegram still works for me but it’ll be nice to have a backup!

Coming in 999 years to a ancient iOS device near u

@poiitidis Will it run on windows 7?

nullDC 2.0.0 rewrite in rust is showing signs of life ... ;)

@electimon Holy

Waiter! Waiter! More HIG violations Please!!

@Stern_XD @TheBobPony Thanks for understanding, I’m actually quite good friends with the developer so I as well as him are very upset because he spends a large amount of time working on these patchers just for it to be shared for free by some asshole

@stubbertville @TheBobPony @stubbertville right here yeah, didn’t notice at first yeah that’s fucked.

@Stern_XD @TheBobPony No you have to pay for the Patreon

@stubbertville @TheBobPony No. They give it free on Patreon. He’s only providing a mirror.

@electimon Siiiick, now do mavericks 😂

@Anim8rJB Awesome!

was looking for my copy of Halo 2 on PC in the garage tonight, and found a whole box of 00’s stuff

@electimon It would be so neat to urban explore a subway system

@silly_lilah Don’t feel bad, Im running 7 on a 12th gen core i7, ddr5 and a 3080ti with an nvme 4th gen lol, people have told me I’m wasting the hardware

i’m addicted to daily driving old technology

@electimon Oooops

logged onto my server to see 5 cryptominers... i may have accidentally left the docker api port open LOL

@electimon Oled is the best

Is an HDR monitor worth it?

@poiitidis Awesome!

After much bug hunting, turns out the Corona's rendering was a simple fix after all! Headlights no longer inside the cars!

@falco_girgis @supersat @ModernVintageG Lord I hope it can happen, more dc games would be great!

I spent a loooot of time looking through the code, and.... God *dayum*, they REALLY reinvented so much of the wheel in those codebases, like... MANY of the core routines they had to implement from scratch are just builtin to KOS and our toolchains now as part of the stdlib, and they "just work" as-is... The main thing that would impede a port is that I can clearly see random x86 ASM interspersed throughout the codebase... that would all have to be converted to either C code or SH4 ASM for DC or another platform.

EA has uploaded fully recovered source code for Command & Conquer (aka, Tiberian Dawn). C&C Red Alert, C&C Renegade, and C&C Generals + Zero Hour to github. W move! github.com/electronicarts/

@electimon Looks good!

@poiitidis Looks fantastic!

One bug down, several more to go

@electimon Photography is great! Have any special ones you took you could share?

I take less photos now than at any point in the past 5 years, it's not that I don't want to, it's just harder to find the beauty in things now.