Comments for teideal glic deisbhéalach

Comment on Here be dragons: advances in problems you didn’t even know you had by these dresses may be long and flowing gowns9

these dresses may be long and flowing gowns9 — Tue, 14 May 2013 13:56:55 +0000

you have got a terrific blog right here! would you wish to make some invite posts on my blog?

Comment on github is making me feel stupid(er) by Jim F

Jim F — Wed, 08 May 2013 21:45:09 +0000

I’m new to GitHub. I didn’t verbalize it as such in my mind, but the “lots of horizontal bars of information” is confusing to me, for sure.

Comment on Big fucking deal by sinelaw

sinelaw — Fri, 03 May 2013 15:18:56 +0000

Hah! Your version is actually easier to understand. Made me laugh.

Comment on Big fucking deal by Thorsten Claus

Thorsten Claus — Thu, 02 May 2013 17:54:14 +0000

Very nice, can I rebase that?

Comment on Big fucking deal by bbb

bbb — Wed, 01 May 2013 21:10:52 +0000

This made my shitty day…

Comment on Big fucking deal by BMeph

BMeph — Wed, 01 May 2013 14:32:53 +0000

Let me guess: someone switched your tea over to decaf? 8D

Still, nice job on the useless fluff-word winnowing – you wield a fine pitchfork, indeed!

Comment on Two-dimensional spatial hashing with space-filling curves by Carter T Schonwald

Carter T Schonwald — Tue, 16 Apr 2013 19:45:12 +0000

I”ll actually have some benchmarks on morton vs hilbert layouts for linear algebra soon. It also looks like a straightline loopless hilbert algorithm *may* exist, trying to sus out the details shortly. (will certainly be bigger than the morton algorithm, but thats a small price for better locality!)

Comment on What’s in a number? criterion edition by Alexey Khudyakov

Alexey Khudyakov — Sun, 14 Apr 2013 14:39:23 +0000

I think this effect is not very important by itself. It’s just too small in most
cases. But it highlight quite clearly that heap size is important variable and
should be taken into account. Heap increase from 120k to 350k inflated results
by ~5%! Results changing due to benchmark reordering could be linked to the
changes of heap size among other things.

Also there’re other problems which analysis of time series could uncover. I’ve
seen sudden jumps of run time. E.g. first 50 measurement consistently get say
500ns and last 50 – 520ns. It’s very apparent on time series plot.

Is it possible to add dumping of raw measurements? It will make analyses of
criterion’s performance much more powerful.

Comment on What’s good for C++ is good for … Haskell!? by Yves Pares

Yves Pares — Sat, 13 Apr 2013 06:20:25 +0000

This may be overrated, depending on how log performs but you can count digits this way :

baseLog = log 10 — computed once and reused

countDigits :: (Integral a) => a -> Int
countDigits x = log x/baseLog + 1

Comment on Making GHC’s I/O manager more scalable by What’s the status of multicore programming in Haskell? | Everyday I'm coding

Tue, 09 Apr 2013 05:26:16 +0000

[...] there is work to make the IO manager more scalable — now with a paper on the design :: [...]