Today, we honor Sandy, the most curious guinea pig I knew, my closest friend for many, many years. She passed away early this morning, circling around once to check on the other pigs, then nestling herself into her favorite little spot one last time.
I’ll always miss the way you’d try and figure out anything. The way you notice everyone coming and demand to boop their nose. The way you’d always remind me that my left shoulder will always be your perch.
Sleep well, my beloved piggie.
idea: flowchart-oriented technical support wiki
kinda like a support forum, but instead of being oriented for threads containing posts, you ask a question, and replies come like flowchart blocks where you respond to spin it off in a direction
and then if someone else has the same question but a different cause, they can branch it off wherever their problem starts to differ
blocks can be edited, annotated with sources, marked out of date, etc. with the idea being that what starts as a forum-post like "i have this problem please help" grows into a living document for solving that problem
Fwahaha, I'm not sure whether to be proud or ashamed of this. While testing multistream RTP feeds with Janus, I've ended up rolling together an 1195-character-long ffmpeg command. It's ridiculously long hahaha
if you read the i486 manual, you'll find some really weird things that don't make sense. for example: the fpu instructions universally have faster timings if you are operating on memory addresses rather than fpu registers.
and surely that can't be right. right? surely that's going to be slower.
and that's the common wisdom of the day. if you find old usenet discussions about it, people will say that yes, the timings are faster, but in the real world it doesn't work out like that, the memory fetch cycles aren't included there, and of course working with registers is faster. there's no other way it could be. if you dont believe it write a benchmark test.
and like other common wisdom, that's just completely wrong.
you can write all the real world tests you like (i just did), and what you will find is that the fpu is nearly always faster with memory addresses than with it's own internal registers. why? who knows! but that's how intel made it, that's how they described it in the manual, and, counterintuitively, that really is how it works.
(For reference, that sub was a Fulcrum TS221 - fantastic little sub. I didn't get a chance to play Hypocrisy through the other PAs I demo'd, but out of the other subs there I found the D&B V-SUB and L'Acoustics KS28 traded blows throughout the listening session. V-SUB barely won on tightness, KS28 barely won on overall musicality. Wasn't as impressed with the Martin or Adamson subs there, though Martin's line array was wonderfully pleasant to listen to.)
Was doing a PA shootout a little while back. Turns out that FM bass in Hypocrisy by @envgen is a fantastic subwoofer test with all those sharp transients. It's super hard to make a sub tight, powerful, and musical all at the same time...
Oh yeah, last night's Keep Talking stream with @quarky was quite fun! Enjoy this clip of them definitely not pressing the detonate button: https://clips.twitch.tv/TalentedCrowdedClipzCurseLit--fRPR7B3CLh7eza0