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.
<UmbralRaptop>
B787_300: well, the focal length of NaN meters makes things a bit more complicated…
<egg|anbo|egg>
!wpn whitequark
* Qboid
gives whitequark a nuclear hydrochlorofluorocarbon
<egg|anbo|egg>
!wpn bofh
* Qboid
gives bofh an electrical Helmholtz circle
tawny- has quit [Remote host closed the connection]
tawny- has joined #kspacademia
Technicalfool_ has joined #kspacademia
Technicalfool has quit [Ping timeout: 194 seconds]
tawny- has quit [Ping timeout: 182 seconds]
<bofh>
A tight upper bound on number of parts required to hold the value
<bofh>
pow(5, power) is
<bofh>
power * 815 / (351 * integerPartWidth) + 1
Qboid has joined #kspacademia
Thomas has joined #kspacademia
APlayer has joined #kspacademia
awang has quit [Ping timeout: 186 seconds]
Thomas has quit [Ping timeout: 194 seconds]
Thomas has joined #kspacademia
tawny- has joined #kspacademia
awang has joined #kspacademia
tawny- has quit [Ping timeout: 198 seconds]
<awang>
hUH
<awang>
Compiling with -march=native slowed down my program
<APlayer>
Time for benchmarking?
<awang>
Wonder where the bottleneck came from
<awang>
Yeah
<awang>
Well, sort of
<awang>
Got a regression suite-ish thing that I repurpose as a low-precision benchmark
<awang>
Runtime dropped from 140ish seconds to 160ish seconds on the low end after adding march=native
<APlayer>
Runtime increased, you mean?
<awang>
Er, yes
<awang>
Mixing up runtime and performance, apparently
<APlayer>
Because I too wish that dropping from the 3rd floor to the 5th was possible :P
<awang>
Got stairs on your way to/from work?
<awang>
Hmmm
<APlayer>
No, but perpetual motion machines would be nice ;-)
<awang>
At least until all the energy fries us :(
<awang>
Hm
<awang>
Running tests individually reveals practically no difference with march=native
<APlayer>
I guess if you have violation of conservation of energy, you also have decreasing entropy, don't you?
<awang>
idk?
<awang>
I'm inclined to say not necessarily, but this is way out of the areas I'm familiar with
<awang>
Just seems to me that you can make entropy increase faster when violating conservation of energy, but entropy doesn't necessarily have to decrease
<UmbralRaptop>
Is the core erosion accurately modeled?
* UmbralRaptop
may have to start using IDL =\
<SnoopJeDi>
UmbralRaptop, do you have any thoughts on this IceCube blazar thing?
<UmbralRaptop>
SnoopJeDi: I don't know enough HEP to say much beyond it seems cool.
tawny- has quit [Ping timeout: 182 seconds]
<SnoopJeDi>
yea, my big takeaway is "oh dang, I didn't realize IceCube had a multi-messenger program" hah
ferram4_ has joined #kspacademia
ferram4 has quit [Ping timeout: 202 seconds]
bofh_ has joined #kspacademia
<awang>
UmbralRaptop: IDL as in that Microsoft thing?
bofh has quit [Ping timeout: 202 seconds]
e_14159 has quit [Ping timeout: 202 seconds]
e_14159 has joined #kspacademia
dx has quit [Ping timeout: 202 seconds]
dx has joined #kspacademia
<UmbralRaptop>
awang: nah, random proprietary programming language that is inexplicably popular in astronomy.
<SnoopJeDi>
I think the explanation is that it was a DSL on-the-scene when things were just getting going. People got used to it, wrote tools, and then the normal academic pathways took over
<awang>
Sounds like the evolution of the typical legacy system
<SnoopJeDi>
I think of https://xkcd.com/297/ everytime someone mentions analysis tools that get propagated mostly by "here's a copy"
<SnoopJeDi>
awang, everything that is now "legacy" was once fresh and useful (and probably still is the latter)
* SnoopJeDi
glances at Perl
pthag has joined #kspacademia
<UmbralRaptop>
SnoopJeDi: you don't need a licence to run .pl files, though.
<SnoopJeDi>
UmbralRaptop, aren't there free implementations of IDL? (not that people *use* them...)
<UmbralRaptop>
Yeah, will probably be finding out how well they work.
<SnoopJeDi>
heh.
<UmbralRaptop>
Perl's sort of the FORTRAN of shell scripts. >_>
* SnoopJeDi
glances at today's FORTRAN reading assignment
* SnoopJeDi
cries
<SnoopJeDi>
this is one line in numpy :(
* UmbralRaptop
assumes that SnoopJeDi is looking at FORTRAN called by numpy?
<SnoopJeDi>
nah, ad-hoc code written by my supervisor. I have a suspicion that the pseudoinverse he constructs step-by-step is different from the one given by numpy.linalg.pinv() (which uses SVD)
<awang>
SnoopJeDi: Very true
<awang>
Fortran reading assignment?
<awang>
How do coding reading assignments work?
<awang>
"Here's some code, figure out what it does"?
<SnoopJeDi>
it's self-assigned
<SnoopJeDi>
"I need to understand what X is doing, I'd better read some more of spaghetti.f95"
<UmbralRaptop>
awang: oh, those are always fun. Or Fun.
<SnoopJeDi>
Upshot: the deepest level of nested `do` loops in the block I am concerned with is only 4 deep!
<awang>
UmbralRaptop: I know that too well :(
<awang>
"Here's some code written by a rather unorganized employee a few years ago. Figure out how to add Feature A"
<awang>
SnoopJeDi: I don't think I want to know what codebase you're working with there "It's only 4 levels deep" is considered a good thing
<awang>
Although I'm hardly one to talk, with my Python class nesting abuse
<UmbralRaptop>
MANDELBROT.F77
<SnoopJeDi>
In my supervisor's defense, a lot of the nesting is for handling vectors, but...blegh.
<awang>
Also, the errors that gfortran spits out when confronted with ancient ifort code are quite interesting to look through
<awang>
Just gives me more appreciation for the poor folks who have to port VC++6-era code
<SnoopJeDi>
a line like `PIdPhaseMatrix(j,1:Nstart(istart)+1)=PIdPhaseMatrix(j,1:Nstart(istart)+1)-temp(l,1:Nstart(istart)+1)*anorm` just makes my skin crawl
<SnoopJeDi>
and that's not even a bad line with multiple operations on it!
<awang>
SnoopJeDi: AAAAAAAAAA
<SnoopJeDi>
once I unrolled the loops and worked out what the heck they're doing, I realized that a lot of the "loop on one line" stuff is just calculating inner products, but...I would assume that's something that can be made into a method? :|
<SnoopJeDi>
I dunno I just wanna figure out why I screw this problem up so much and he doesn't.
<awang>
You need a clang-format for Fortran
<awang>
Too bad one doesn't exist yet
<awang>
At least until flang gets going
<SnoopJeDi>
I suspect getting LLVM involved would tetrate my problems, given my complete ignorance of it
awang has quit [Ping timeout: 182 seconds]
awang has joined #kspacademia
<awang>
In theory, clang-format shouldn't require knowledge of LLVM itself
<awang>
Just differing formatting parameters
<awang>
Which is a rabbit hole in itself
tawny- has joined #kspacademia
<egg|anbo|egg>
whitequark: ANBOcat seemed to be fairly badly irritated under its flumethrin collar, so we removed it; but now three days later it is of course getting fleas... any thoughts?
<bofh_>
17:52:11 <@SnoopJeDi> Upshot: the deepest level of nested `do` loops in the block I am concerned with is only 4 deep!
<bofh_>
okay that is, like, normal by any standards. :P
<bofh_>
(I think my record that I've seen is 13 nested do loops, worst part is this was both sensible and unavoidable)
<SnoopJeDi>
yea, it's alarmingly close to "I can read this and understand the author's intent"
bofh_ is now known as bofh
<SnoopJeDi>
I'm just salty about having to translate this crap into the actual mathematical operation
<SnoopJeDi>
because I get some very conflicting answers from him when I ask him directly
<SnoopJeDi>
and I suspect there's something subtle in this construction that makes his pseudoinverse respect some degrees of freedom more than others, or something...
tawny- has quit [Ping timeout: 194 seconds]
tawny has joined #kspacademia
<awang>
bofh: Can you share what in the world was happening so that a 13-deep loop made sense?
* UmbralRaptop
wonders why the building just lost power for a few seconds.
<UmbralRaptop>
Also there appears to be lightning outside?
<SnoopJeDi>
UmbralRaptop, good day to post that link :P
<SnoopJeDi>
(considering the massive outage earlier)
<UmbralRaptop>
Hah!
<awang>
SnoopJeDi: ?
<SnoopJeDi>
awang, github had a large outage today
<bofh>
so we just had a completely inexplicable ~hour-long outage at ANBO just now
<bofh>
(also when the power came back on a random Albinoni CD containing Sinfonie e concerti a cinque started at the second in F major which opens with a fairly triumphant trumpet bit, which was hilarious).
<kmath>
<marccold> I am responsible for buying books on psychology AND necromancy because the creators of classification systems thoug… https://t.co/B9uYR3R4wH
<egg|anbo|egg>
so we just had another minute-long cut
<egg|anbo|egg>
!wpn UmbralRaptop
* Qboid
gives UmbralRaptop an astatine monorepo which strongly resembles an extinguisher
<UmbralRaptop>
egg|anbo|egg: don't confuse phosgene with hay?
<bofh>
16:42:20 * UmbralRaptop may have to start using IDL =\
<bofh>
my extreme condolences
* UmbralRaptop
also pokes pip and anaconda with sticks.