@rygorous@mastodon.gamedev.place avatar

rygorous

@rygorous@mastodon.gamedev.place

Abstraction maker, abstraction breaker. FUN FACT: things I prefix with FUN FACT are sometimes fun and sometimes factual, but very rarely both.

This profile is from a federated server and may be incomplete. Browse more on the original instance.

dougbinks, to random
@dougbinks@mastodon.gamedev.place avatar

I should probably invert my normal research approach for fast algorithms from.

  1. I search the internet for information.
  2. I read @rygorous' blog.

This time it's for fast blurs.

https://fgiesen.wordpress.com/2012/08/01/fast-blurs-2/

rygorous,
@rygorous@mastodon.gamedev.place avatar

@dougbinks FWIW this is nice and constant-time for very large kernels but for small to medium sizes you're better off doing something else

rygorous,
@rygorous@mastodon.gamedev.place avatar

@dougbinks Yup, you're right! Will fix!

pervognsen, to random
@pervognsen@mastodon.social avatar

No need to fear, the optimizing compiler is here:

mov r8d, 1
test r8, r8
jne .LBB1_2

rygorous,
@rygorous@mastodon.gamedev.place avatar

@pervognsen it's defensive optimizing

pervognsen, to random
@pervognsen@mastodon.social avatar
rygorous,
@rygorous@mastodon.gamedev.place avatar

@pervognsen The FE changes don't sound minor to me.

aeva, to random
@aeva@mastodon.gamedev.place avatar

which zelda game is link rot from

rygorous,
@rygorous@mastodon.gamedev.place avatar

@aeva pikmin. duh

lritter, to random
@lritter@mastodon.gamedev.place avatar

so what's his punishment? i expect it to be tough considering in the US you can get 1000 years of jailtime for littering or smth

rygorous,
@rygorous@mastodon.gamedev.place avatar

@aeva @lritter yeah, never forget, owning like $40 worth of weed can get you jail time if you're black, but if you're white and/or a corporation, laundering over $860 million of drug money is a $1200 million fine (if you're caught)

rygorous,
@rygorous@mastodon.gamedev.place avatar

@breakin @aeva @lritter In practical terms, I don't expect this to have much if any immediate consequences, because

  1. of course they'll be going for an appeal, and it's going to be a long time for that dust to settle
  2. election-wise, things being as they are, I find it hard to imagine a significant chunk of the population thinking "well I was gonna vote for him but now that he's been convicted in a jury trial I've changed my mind"
rygorous,
@rygorous@mastodon.gamedev.place avatar

@aeva @lritter we're losing money on every transaction but making it up in volu.... never mind I was just informed we're apparently on mute

pervognsen, to random
@pervognsen@mastodon.social avatar

I stopped paying attention to process nodes around 12 nm but I just noticed Arm Cortex X925 is advertised as designed for a 3 nm process. I'm assuming this has even less to do with lambda than it once did?

rygorous,
@rygorous@mastodon.gamedev.place avatar

@pervognsen This talk is now 11+ years old https://youtu.be/_bhEDQzNQ-c?si=He4dpsBsOCo3330U&t=961 and the editing is terrible but it answered a lot of questions for me about why even SEM images of circuits look very different for newer designs than older ones

rygorous,
@rygorous@mastodon.gamedev.place avatar

@pervognsen Yes, absolutely nothing.

As per IRDS (2021), expected feature sizes for "3nm" generation processes are ~48nm gate pitch, ~24nm metal pitch.

Doomed_Daniel, to random
@Doomed_Daniel@mastodon.gamedev.place avatar

I've got a (probably simple) graphics programming-related question:
Is it correct that the only useful values for GL_TEXTURE_MAX_ANISOTROPY_EXT are 2, 4, 8 and 16 (and maybe 1 for "don't use anisotropic filtering")?

At least as far as I can remember I've never seen other values configurable in games; however, for some reason, GL_TEXTURE_MAX_ANISOTROPY_EXT is used with floats (glTexParameterf()), and the spec only says "float greater or equal to 1.0"

rygorous,
@rygorous@mastodon.gamedev.place avatar

@pervognsen @Doomed_Daniel no I'm pretty sure it does support npot2 values, it's just that the limits are usually picked as pow2s. IIRC the actual sample count is just any integer. (but it is ints.)

rygorous,
@rygorous@mastodon.gamedev.place avatar

@zeux @pervognsen @Doomed_Daniel For the one HW where I do know the exact details (the souped-up version of Intel Gen samplers that ended up in LRB1) the way the samplers work internally is that everything other than point filtering is internally sequenced into multiple "bilinear" taps which complete at a rate of (at most) one per cycle and can add into an accumulator over multiple cycles, which is how all the more complex filtering modes work.

rygorous, to random
@rygorous@mastodon.gamedev.place avatar

"Animal Kingdom". Bullshit.

Strange genes in pools distributing phenotypes is no basis for a system of government

pervognsen, (edited ) to random
@pervognsen@mastodon.social avatar

When a beverage proclaims "naturally & artificially flavored" on its label in bold letters you know they're not hiding anything. Or maybe they're trying to draw attention away from the supernatural flavoring.

rygorous,
@rygorous@mastodon.gamedev.place avatar

@pervognsen or the paranormal flavoring

rygorous,
@rygorous@mastodon.gamedev.place avatar

@pervognsen can't overemphasize how relieved I am that this is, evidently, genuine!

can you imagine if it was an imitation blue soda instead

rygorous,
@rygorous@mastodon.gamedev.place avatar

@koos303 @pervognsen just don't let yourself get scammed by the contradiction in terms that is zero-calorie "energy drinks"

aeva, to random
@aeva@mastodon.gamedev.place avatar

fuck I forgot to get high this weekend

rygorous,
@rygorous@mastodon.gamedev.place avatar

@aeva tired: get high during the weekend
wired: get high on a weekday
inspired: get high during a boring meeting

rygorous,
@rygorous@mastodon.gamedev.place avatar

@aeva (personally, I'm straight edge, but if you're gonna subscribe to the lifestyle, might as well commit)

rygorous, to random
@rygorous@mastodon.gamedev.place avatar

One of my favorite things for art to do is take a concept that is patently ridiculous but then commit so hard and execute so well you are forced to meet it on its own terms.

Anyway the Muse song "Knights of Cydonia" and the accompanying music video are like close to 18 years old now and they're still awesome

regehr, to random
@regehr@mastodon.social avatar

jfc is that an attempt at a beard or is he molding

rygorous,
@rygorous@mastodon.gamedev.place avatar

@regehr @pervognsen in my case, the reason I have the goatee but no mustache is that my mustache has what I call the Reverse Hitler with a very visible gap in the middle of my lip and it just looks ridiculous

regehr, to random
@regehr@mastodon.social avatar

my new favorite thing is this diagram of all 256 configurations of the ternlog instruction

https://www.sandpile.org/x86/ternlog.htm

rygorous,
@rygorous@mastodon.gamedev.place avatar

@regehr I'm annoyed that most descriptions don't bother to mention that "the immediate operand is just the truth table"

rygorous,
@rygorous@mastodon.gamedev.place avatar

@regehr I successfully lobbied some folks at Intel to add these constants a while back: https://clang.llvm.org/doxygen/avx512fintrin_8h.html#a955ea13667c676eca777bda9b753e93c

since it's a truth table, you can write
(_MM_TERNLOG_A | _MM_TERNLOG_B | _MM_TERNLOG_C)
to get the truth table for a 3-operand OR, or
(_MM_TERNLOG_A ^ _MM_TERNLOG_B) & ~_MM_TERNLOG_C
to get the truth table (and thus the correct immediate operand) to encode (a^b) & ~c, etc.

rygorous,
@rygorous@mastodon.gamedev.place avatar

@dougall @regehr major/minor are majority/minority vote. These are probably the two fundamental monotone functions of three arguments and an important building block.

The generalization to higher arity are threshold functions f_k(x_1,...,x_n) = [x_1 + ... + x_n >= k]

amonakov, to random
@amonakov@mastodon.gamedev.place avatar

(prompted by discussion of detecting bitwise and-not earlier in GCC's optimization pipeline)

My ideal compiler IR would not have and/or/xor as distinct bitwise ops, just generic ternlog and probably the corresponding two-operand function ("bilog"?) too.

rygorous,
@rygorous@mastodon.gamedev.place avatar

@amonakov @pervognsen It's just about how early in their development they were committed to having 3 source operands. Same with 2-source-reg-plus-index-reg shuffles. The generic crossbar network is usually already there: sometime past the 8th dedicated unpack/pack/shuffle pattern for different type sizes, it's easier to build the general crossbar and just supply constants for the index vector in the "canned shuffle" cases.

rygorous, to random
@rygorous@mastodon.gamedev.place avatar

You can tell that Chad really isn't much of a Chad at all going by how the usual unit in common parlance is actually the Gigachad.

and here I thought Pascals and Teslas were awkwardly normalized. This is like Farad levels of inconvenient scaling

  • All
  • Subscribed
  • Moderated
  • Favorites
  • megavids
  • thenastyranch
  • rosin
  • GTA5RPClips
  • osvaldo12
  • love
  • Youngstown
  • slotface
  • khanakhh
  • everett
  • kavyap
  • mdbf
  • DreamBathrooms
  • ngwrru68w68
  • provamag3
  • magazineikmin
  • InstantRegret
  • normalnudes
  • tacticalgear
  • cubers
  • ethstaker
  • modclub
  • cisconetworking
  • Durango
  • anitta
  • Leos
  • tester
  • JUstTest
  • All magazines