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.
<kmath>
<tompeterken> Who fancies looking at pretty pictures of galaxies AND helping me with my PhD? Yeah that sounds great doesn't it?!… https://t.co/kxdJV5o1LQ
UmbralRaptop has quit [Read error: Connection reset by peer]
UmbralRaptor has joined #kspacademia
<egg|zzz|egg>
bofh: hm, it's going to be hard to bring that sincos below 50 cycles
<egg|zzz|egg>
(latency)
<egg|zzz|egg>
bofh: also I am fascinated at how consistently Horner is terrible :-p
UmbralRaptor has quit [Read error: -0x1: UNKNOWN ERROR CODE (0001)]
UmbralRaptop has joined #kspacademia
UmbralRaptop has quit [Read error: Connection reset by peer]
UmbralRaptop has joined #kspacademia
UmbralRaptop is now known as UnethicalRaptor
* UnethicalRaptor
pokes training videos with a stick.
UmbralRaptop has joined #kspacademia
UmbralRaptop has quit [Client Quit]
UmbralRaptop has joined #kspacademia
UnethicalRaptor has quit [Read error: Connection reset by peer]
<egg|zzz|egg>
!u ⟼
<Qboid>
U+27FC LONG RIGHTWARDS ARROW FROM BAR (⟼)
<egg|zzz|egg>
!u ↦
<Qboid>
U+21A6 RIGHTWARDS ARROW FROM BAR (↦)
<egg|zzz|egg>
bofh: btw do you have any idea why the following character has the name it has:
<egg|zzz|egg>
!u ⧧
<Qboid>
U+29E7 THERMODYNAMIC (⧧)
<UmbralRaptop>
current status: angry that the references include links to papers on closed access journals, but skip it for open access ones.
UmbralRaptop has quit [Read error: Connection reset by peer]
UmbralRaptop has joined #kspacademia
* UmbralRaptop
🔪 📶
<egg|zzz|egg>
UmbralRaptop: Ꙩ_ꙩ
* egg|zzz|egg
wonders whether bofh is sleeping
<UmbralRaptop>
egg|zzz|egg: online research ethics training thing
* UmbralRaptop
uses a ☕ to cast "summon bofh"
<egg|zzz|egg>
UmbralRaptop: so you're stabbing the online citation ethics of the training about online citation ethics?
<kmath>
<✔NASA_TESS> TESS Mission Update: Successfully downlinked through @NASASCaN #DSN the first 2 weeks of @NASA_TESS science data!… https://t.co/qT8zxodNXw
<kmath>
<devonzuegel> Vegetables 300kcal ⏎ Fruit 150kcal ⏎ Meat & fish 600kcal ⏎ Dark chocolate 3,600kcal ⏎ Grains 400kcal ⏎ someone who is good at… https://t.co/dRaRil8RJD
<bofh>
egg|zzz|egg: sec
<egg|zzz|egg>
bofh: also, I'm getting different numbers from IACA 2.x's latency analysis and from looking at the latency of the initial iteration in IACA 3's trace
<UmbralRaptop>
bofh: just burn 4750 kcal/day!
<egg|zzz|egg>
(IACA 3 doesn't directly have a latency analysis, but it shows you a trace of the first iterations of its throughput analysis so you can read the latency off that...)
<bofh>
egg|zzz|egg: so at this point I want to benchmark it using RDTSC-in-a-loop (i.e. what we actually use to benchmark lavc, since it gives more useful real-world numbers compared to, say, IACA).
<bofh>
UmbralRaptop: I mean I think my absurd chocolate consumption is actually good atm since I think I'd be running a caloric deficit otherwise. Keep in mind I sleep little, am on stimulants and have been walking, like, 3-7km a day for the past month, among other things. :p
<UmbralRaptop>
heh
<APlayer>
UmbralRaptop: Did you just calculate the wattage of bofh?
<UmbralRaptop>
Yes
<APlayer>
Which means that bofh is approximately 30.8% of a horse
<egg|phone|egg>
Bofh: I have benchmarks too, but it's hard to get cycle resolution there
<bofh>
egg|phone|egg: cycle resolution is too noisy to be useful, what you want is microsecond resolution for some large number of retries of the function (so it's guaranteed to be in cache or whatnot)
<bofh>
I mean if you want cycle resolution, Agner Fog has tables of throughput and latency per each instruction, but it takes a few minutes to realize if you add those numbers up your actual function ovten takes much more or much less and it's variable per run too.
egg|phone|egg has quit [Read error: Connection reset by peer]
<egg|work|egg>
bofh: yeah what i mean is that microsecond on large number of retries is too noisy to get cycles on the base function; also it means I have to be running it on the actual arch, so I need to switch computers to compare SNB and SKL, and I don't have a HSW
<egg|work|egg>
this makes it quite messy to try to figure out the effect of smol optimizations
<bofh>
and my point is "cycles on the base function" is almost always meaningless and quite often not even well-defined :P
<egg|work|egg>
bofh: e.g. evaluating a polynomial as ax + (bx + cx * x^2) * x^2 is has lower latency than ax + (b + cx^2) * (x^2 * x) on SNB according to IACA 2.x and same on HSW according to both IACAs, but checking that by hand is going to be tedious (in particular because I've been working on that on my SKL laptop :-p)
<egg|work|egg>
(also i might have pinged x)
mlbaker has quit [Ping timeout: 202 seconds]
<bofh>
egg|work|egg: also is that difference at all meaningful? (latency differences of 1 cycle virtually never are).
<Qboid>
[#1890] title: RSS has the weird inclinations. | I noticed that axial tilt works, but RSS still has the super weird inclinations even though axial tilt works, | https://github.com/mockingbirdnest/principia/issues/1890
UmbralRaptop has joined #kspacademia
<egg|zzz|egg>
!wpn UmbralRaptop
* Qboid
gives UmbralRaptop a devirtualized sharp Sagan
<egg|zzz|egg>
!wpn __awang__
* Qboid
gives __awang__ a mission
<UmbralRaptop>
!wpn egg|zzz|egg
* Qboid
gives egg|zzz|egg a stinging table/sumerogram hybrid
<egg|zzz|egg>
bofh: which is of course a follow-up on #1889, which has the appropriate branch name
<Qboid>
[#1889] title: Improve the latency benchmark | Compute (cos bitand 2 ** 52 - 1) bitxor sin to induce a data dependency without reading from precomputed random inputs or doing a couple of floating-point additions.... | https://github.com/mockingbirdnest/principia/issues/1889