Kyle Freeman

I know AI is sort of random, but damn Google you're so much worse with AI. Just got this response.. "Because compact builds can be tight, 32GB kits featuring tight timings are the ideal choice."

I love when technology solves a rare (for me) task. I recently needed to bring 100 year old mono recordings up to multi-track. Did it with the new Intel AI plug-in for Audacity. Worked great and I could move on to the next thing. AI certainly crushes it for some applications.

@NOTimothyLottes @winning_tactic The issue with Hyperthreading was complex. It drove core temperatures up, took extra silicon, and trashed the core specific cache, particularly the code cache. It was dropped with the recent CPUs. Gamers often turned it off.

@TheKyleFreeman @winning_tactic Hyper-threading is simply a mechanism for remaining efficient while hiding latency. Massive hyper-threading is how GPUs work. And pointer size effects data cache mostly, x86 opcode encoding uses 32-bit displacements and either [8-bit or 32-bit] imms regardless of 32/64-bit modes

Which is faster 32-bit or 64-bit apps? Ten years ago it was 32-bit apps, but today, it's 64-bit apps. Better CPUs with larger caches and compilers that make smart use of the extra registers in 64-bit mode offset the slight disadvantage of larger pointers and larger executables.

@NOTimothyLottes @winning_tactic Hyperthreading allowed a second code stream to be held and to attempt to execute in the portions of the CPU not being used by the main thread. In theory, FPU instructions from one thread could execute in parallel with integer commands in another and thread switching was instant.

@NOTimothyLottes @winning_tactic The L1 cache has a fixed 64k (previously 32k) for code. There's a separate L1 data cache. Both code and data share the L2 and L3 caches. The issue with code is that it's broken into instructions as it enters the L1 code cache, so thrashing the code cache is rather costly.

@NOTimothyLottes @winning_tactic The biggest hit from a 64-bit address space is loading a variable or pointer takes an extra 4 bytes. This is only bad if you run out of code cache. Luckily hyper-threading is gone and the code cache jumped from 32k to 64k per core on more recent CPUs.

@winning_tactic @TheKyleFreeman It’s just lower overhead for pointers. Cuts down the perf tax.

@NOTimothyLottes True, if you're in Linux that supports it. Windows, alas, doesn't give one the option.

@TheKyleFreeman 64-bit mode with fixed low 32-bit code address space is the fastest

Dark Lord is almost 40 years old and is still getting positive reviews. This playthrough meanders long enough for the sun to set, which was an uncommon feature for an adventure game. youtube.com/watch?v=7KnkYc…

I like AI for programming. It can do the boring stuff, and when it fails, it usually got close enough for a quick fix. I still don't trust it for high performance tasks, aligning cache usage between cores, fixed point math, inline SMD assembly, stuff that I really like doing.

AI is getting better at code, although it is picking up some of the bloatware mindset of the training data. Visual Studio's compiling is getting more optimized as well. I wonder if they're using AI to optimize the compiler output?

Can you tell which image from my 80's Animated Adventure game Dark Lord has been enhanced with AI?

It must be rare to have a subwoofer. Watched the Artimis launch, no rumble. Watched a press briefing, and my dishes rattled on every 'p'. News can do better.

@adahstwt 6502 Assembly.

name the programming language that made you love coding.

Dark Lord: An Animated Adventure opened with a ticking clock in an old secret room. Even in the 80's, ticking clocks were rare as silent quartz movements replaced almost all mechanical clocks in the 70's. Now the game itself is more of a dinosaur than ticking clocks.

The villain in my Apple ][ adventure game Dark Lord was named Lord Nequam. This was before the internet so I used a 400 year old Latin dictionary. Nequam means worthless, good for nothing · [of character] worthless, vile, bad.

Dear Alexa+, I'm curious. I buy weird and random stuff. I don't need a running commentary with pointless speculation. You're worse than that one Costco receipt checker.

If you really care about lighting and spend a crazy amount on it, are you a photonophile?

@karthikponna19 What if you started on things with only a numeric display? Then it's less "Hello World" and more 07734.

in which language did you write your first HELLO WORLD ?

The LLMs are the least amazing thing AI can do. They're pretty worthless for serious stuff, but man are they fun. You can now easily build all the computers in the original Star Trek and then argue with them like Kirk. Only they'll agree they're illogical and not blow up.