egg|nomz|egg changed the topic of #kspacademia to: https://gist.github.com/pdn4kd/164b9b85435d87afbec0c3a7e69d3e6d | Dogs are cats. Spiders are cat interferometers. | Космизм сегодня! | Document well, for tomorrow you may get mauled by a ネコバス. | <UmbralRaptor> egg|nomz|egg: generally if your eyes are dewing over, that's not the weather. | <ferram4> I shall beat my problems to death with an engineer.
<egg|zzz|egg>
!wpn -add:wpn paranotion
<Qboid>
egg|zzz|egg: Weapon added!
<egg|zzz|egg>
!wpn -add:wpn metanotion
<Qboid>
egg|zzz|egg: Weapon added!
<egg|zzz|egg>
!wpn -add:wpn notion
<Qboid>
egg|zzz|egg: Weapon added!
<egg|zzz|egg>
!wpn whitequark
* Qboid
gives whitequark a peculiar kindle-like frustum
<UmbralRaptor>
!wpn egg|zzz|egg
* Qboid
gives egg|zzz|egg a FORTH-like hedgehog
<UmbralRaptor>
!wpn -add:wpn Algol
<Qboid>
UmbralRaptor: Weapon added!
<UmbralRaptor>
!wpn -add:adj Algol
<Qboid>
UmbralRaptor: Adjective added!
<UmbralRaptor>
!wpn -add:wpn stack
<Qboid>
UmbralRaptor: Weapon already added!
<UmbralRaptor>
!wpn -list
* Qboid
gives -list a transitive stinger
<UmbralRaptor>
!wpn -add:adj thermal
<Qboid>
UmbralRaptor: Adjective already added!
<UmbralRaptor>
!wpn -add:adj turbulent
<Qboid>
UmbralRaptor: Adjective added!
<UmbralRaptor>
!wpn -add:adj ordered
<Qboid>
UmbralRaptor: Adjective already added!
<UmbralRaptor>
!wpn -add:adj chaotic
<Qboid>
UmbralRaptor: Adjective added!
<egg|zzz|egg>
whitequark: "deleting dtor"? Ꙩ_ꙩ
<soundnfury>
egg|zzz|egg: ~book();
<soundnfury>
/ literary deconstruction
<soundnfury>
s/\//\/\/
<Qboid>
soundnfury meant to say: // literary deconstruction
<whitequark>
egg|zzz|egg: yeah
<whitequark>
deleting destructor.
<whitequark>
destructs and calls delete.
<egg|zzz|egg>
huh
<whitequark>
I think it's necessary because of exception handling or something
<egg|zzz|egg>
hmm
<soundnfury>
!wpn whitequark
* Qboid
gives whitequark a stellated pharmacology with a kukri attachment
<kmath>
<fchollet> "Neural networks" are a sad misnomer. They're neither neural nor even networks. They're chains of differentiable, p… https://t.co/IwTVnOFAUZ
<UmbralRaptor>
Via the helvetic corvid
<egg|zzz|egg>
awang: still can't get KSP 1.3.1 to load with your rebuilt RSS/RO etc.
<e_14159>
UmbralRaptor: Well, yes (though I would say "transformation" instead of "geometric" functions. But you can't, apparently, change terminology that's been used for decades.
<Qboid>
U+00F8 LATIN SMALL LETTER O WITH STROKE (ø)
<egg|zzz|egg>
bofh: ?
<bofh>
Ahh.
<egg|zzz|egg>
bofh: uniformly?
<egg|zzz|egg>
bofh: ah, for the polynomials, yes
<egg|zzz|egg>
bofh: note that there are no branches, this is templatized and the degree is thus known at compile time
<bofh>
so I've always only compared fully unrolled branch-free Estrin & Horner, degree known at compile-time is really by far what you want.
<egg|zzz|egg>
okay so we're comparing the same thing
<egg|zzz|egg>
bofh: semi-unrelated, we looked at using sse2 intrinsics for the vector addition and scalar-vector multiplication, it seems it improves things a bit (MSVC doesn't vectorize those things even when you ask nicely)
<egg|zzz|egg>
bofh: and now I went into a rabbit hole of staring at the list of intrinsics
<egg|zzz|egg>
bofh: I wonder what the fastest way of computing a 3d inner product is :-p
<bofh>
probably unrolled scalar if your 3d vector stride is 3. if it's padded to 4, movaps/movsldup/addps/movhlps/addss is likely faster.
<egg|zzz|egg>
bofh: vector stride?
<bofh>
as in are your 3d vectors struct _vec3 { float[3]; } __attribute__((packed)); or struct _vec3 { float[4]; } __attribute__((packed)); ?
<egg|zzz|egg>
bofh: they're double, who do you think I am
<egg|zzz|egg>
bofh: and while right now they are 3 doubles, I might turn them into two _m128d
<egg|zzz|egg>
bofh: or one _m256d even? :-p
<bofh>
oh, right, sorry, I keep forgetting you actually *do* need the precision :P
<bofh>
(sorry, it just oddly happens that nearly all my handwritten vector asm is on floats to the point where even if assuming single precision would be daft, I still do so anyway as it's so hardcoded in my mind :P)
<egg|zzz|egg>
hah
<egg|zzz|egg>
bofh: also, should I assume I have AVX and SSE4.1 or am I likely to have users on potatoes
<bofh>
your reasonable targets are SSE3 for the potato case & AVX2 if you want to assume Ivy Bridge.
<egg|zzz|egg>
bofh: okay so I'm not going to assume Ivy
<egg|zzz|egg>
'cause I'm on a Sandy
<egg|zzz|egg>
bofh: what does SSE3 have that SSE2 doesn't?
<egg|zzz|egg>
bofh: SSE4.1 has DPPD, so if I had critical inner products that might make some sense
<kmath>
<FioraAeterna> as far as i know nobody's made this yet so i felt the need to fix this problem https://t.co/ZUBXYP2b97
<bofh>
well, for doubles it's just a movhlps (yes, p*s* here) followed by an addsd on 4-vectors. for AVX2 I think you want to use vpermilpd, addpd, vmovhlps, vaddsd optimally, but I could be wrong (Fiora, thoughts?)
<egg|zzz|egg>
bofh: what does your first "it" refer to here
<egg|zzz|egg>
bofh: ah, you mean doing a hadd?
<egg|zzz|egg>
bofh: but why is hadd slower than that
<kmath>
<FioraAeterna> .@stephentyrone @bofh453 here is a graph of the usability of Intel integer SIMD over time https://t.co/I9qnAjNAWM
<bofh>
and yeah. as for why it's slower (primarily on old CPUs) I have no clue tbh, but it genuinely is, check Agner Fog's tables if you want to confirm, the hadd*/hsub* operations are fuckass-slow.
<egg|zzz|egg>
fun
<egg|zzz|egg>
!tell awang I can't get KSP 1.3.1 with your RSS/RO etc. rebuilds to load, which versions of the other mods are you using?
<Qboid>
egg|zzz|egg: I'll redirect this as soon as they are around.
<egg|zzz|egg>
!tell awang also, please update the expected KSP version in the checkers
<Qboid>
egg|zzz|egg: I'll redirect this as soon as they are around.
<egg|zzz|egg>
!wpn bofh
* Qboid
gives bofh a noun constructor
<egg|zzz|egg>
did Qboid build an object-oriented language on top of the AGC
<egg|zzz|egg>
bofh: wait why is SSE2 a dip in the usability graph
<soundnfury>
there are two things called "kernel" programming. One uses all that SSE LDCMPWTFBBQ stuff... the other almost never does, because then we'd have to save and restore the userspace FPU state just in case they were using it
<kmath>
<bofh453> @rygorous @stephentyrone @FioraAeterna "Andy handed out pieces of swiss cheese [...] to mark the non-orthogonality of the MMX ISA" priceless
<egg|zzz|egg>
Ꙩ_ꙩ
<egg|zzz|egg>
!wpn bofh
* Qboid
gives bofh a median tofu
<bofh>
what's the moment-generating function of tofu anyhow?
<UmbralRaptor>
so(y)
awang has joined #kspacademia
egg|phone|egg has quit [Remote host closed the connection]
<awang>
egg|work|egg: Is it one of the dlls specifically that is causing KSP to not load?
<Qboid>
awang: egg|zzz|egg left a message for you in #kspacademia [14.01.2018 18:18:05]: "I can't get KSP 1.3.1 with your RSS/RO etc. rebuilds to load, which versions of the other mods are you using?"
<Qboid>
awang: egg|zzz|egg left a message for you in #kspacademia [14.01.2018 18:19:15]: "also, please update the expected KSP version in the checkers"
<awang>
Let me check the other versions...
<awang>
Been a while since I booted KSP
<awang>
Kopernicus is 1.3.1-3
<awang>
Uh
<awang>
8192 RSS textures
<awang>
MM 3.0.1
<awang>
lamont's PEG-10 MJ build
<awang>
Principia from git
<awang>
RealChute 1.4.5
<awang>
FAR 0.15.9.Liebe
<awang>
RealFuels from git
<awang>
RealPlume v11.0.0
icefire has joined #kspacademia
<awang>
SmokeScreen v2.7.5
<awang>
Proc parts v1.2.12
<awang>
B9 proc wings fork v0.40.13
<awang>
Deadly Reentry 7.6.2
<awang>
RSSTimeFormatter at update-for-KSP-1.3.0 branch
<awang>
Proc fairings v4.0
<awang>
Forgotten Real Engines v0.7.1.0
<awang>
Niche Parts v1.4.2
<awang>
RSS is at recompile-for-1.3.1, which should be the same as the PR on GitHub
<awang>
RO is at update-for-KSP-1.3.0, which should match the PR on GitHub
<awang>
RealHeat is at PhineasFreak/RH-KSP-1.3-Binaries
<awang>
Which I think matches his PR
<awang>
RealFuels HEAD for me is derived from the port-newCryo-1.3.0 branch
<awang>
So all in all not really sure what breaks things :(
<egg|zzz|egg>
awang: hm, you use FAR at Liebe rather than at head?
<egg|zzz|egg>
though tbh I had found no issues with head....
<awang>
egg|zzz|egg: I wasn't aware FAR had updates
<awang>
Don't know if I did
<awang>
git log says that I installed Far Liebe 2017-10-28
<egg|zzz|egg>
yeah still hanging, it's not FAR's fault
<egg|zzz|egg>
what the hell am I screwing up
<awang>
Just finished booting on my machine
<awang>
I'm pretty sure that the uncommited changes don't touch things that would freeze KSP
<awang>
I hope
<egg|zzz|egg>
awang: also, fun, we both use FRE and Niche parts as our engine set :-p
<awang>
Nah, I have a bajillion other mods installed, too
<awang>
I just listed the mods you did in your hastebin link in #RO
<awang>
Hmmm
<awang>
Maybe I should try just cherry-picking commits to match your install
<Ellied>
UmbralRaptor: Am I supposed to be banned in KSPO?
<Ellied>
I don't plan on lingering either way but I'm admittedly curious
<Ellied>
[#KSPOfficial] GreeningGal*!*@* banned by ipo.esper.net on Sat, 01 Apr 2017 15:25:22
<UmbralRaptor>
Uh, unsure. I vaguely recall something with duodex?
<Ellied>
figures.
<awang>
egg|zzz|egg: Do you have MFI installed?
<egg|zzz|egg>
awang: yeah
<egg|zzz|egg>
comes with Kopernicus
<awang>
Ok, just checking
<awang>
Didn't know that
<awang>
Going to try to replicate your setup
<awang>
Which means dropping all the other commits
<awang>
And I'm pretty sure that I'm going to delete a dependency somewhere by accident
<Ellied>
UmbralRaptor: And not even a PM on the forums either, wow.
<Ellied>
admittedly I'm used to the SCP Wiki community where you have an entire disciplinary report written up about you (sometimes multiple such reports) if you get banned for so much as an hour
<egg|zzz|egg>
awang: um, I should push those egg-something branches
<awang>
Looks like it doesn't include implicit dependencies?
<awang>
I also updated the RO/RSS checkers
<awang>
Or at least I think
<awang>
RealHeat PR isn't mine
<awang>
What changes did you make in those branches?
<egg|zzz|egg>
awang: pulled yours, in the RO one also the blowfish MM update
<egg|zzz|egg>
awang: and you didn't update the checkers, or at least you didn't push the source for that update
<egg|zzz|egg>
okay you did
<egg|zzz|egg>
but recently :-p
<egg|zzz|egg>
yeah, "committed an hour ago"
<egg|zzz|egg>
awang: also my moon timewarp altitude change
<egg|zzz|egg>
awang: which implicit dependencies
StCypher has quit [Ping timeout: 186 seconds]
<awang>
egg|zzz|egg: Yep, the checker update was recent
<awang>
Implicit dependencies like MFI included with Kopernicus
<awang>
Because I installed mods one by one
<awang>
So MFI and Kopernicus were two separate commits, for example
<UmbralRaptor>
Ellied: what timeframe are you looking at for graduation and grad schools?
<Ellied>
basically the normal cycle (applications in January) shifted ahead to next year
<Ellied>
graduation is ideally going to be at the end of the summer for me
<UmbralRaptor>
Hm
<egg|zzz|egg>
meow
<egg|zzz|egg>
meo
<UmbralRaptor>
Uh, I'd as much as possible aim for being ready with the applications by December 1st, unless all the schools have January deadlines?
* UmbralRaptor
meows at egg
<UmbralRaptor>
Yada, yada, 3x letters of recommendation, get this lined up as soon as you have a list of schools, etc.
<egg|zzz|egg>
喵
<Ellied>
UmbralRaptor: something like that. I have lots of time to prepare at the moment and a heavy course load so it's hard to find time to work on that, but I do have several people who are attentive to those deadlines watching over me and making sure I don't forget.
<Ellied>
I also plan to enroll here as a grad student, at least temporarily (probably for a Master's, but there's a chance I'll get into a Ph.D. program sooner) and that's all deadlines in April
<kmath>
<FakeUnicode> @atatabitata It may not be easier, but it would probably be faster to just re-form the Soviet Union as a sovereign… https://t.co/32pfspIpXQ