Register allocation

I read about LLVMs register allocator <http://blog.llvm.org/2011/09/greedy-register-allocation-in-llvm-30.html>

Register allocation is a very difficult part of a compiler.. The chordal approach is beautiful but impractical for instruction sets like x86 that have fixed registers in mul and div. Puzzle solving looks like the most efficient general algorithm. I like the approach LLVM describes a lot though, take it really simple but then having a lot of flexibility to tweak it and try out new optimizations and heuristics.


Math

I made a mistake in my mathematics again..

I've been working hard on this thing called Discrete Valuation Rings. It's a way focus in on a particular prime in algebraic number theory. This "local" approach it's called can be used to give a really nice proof that ideals have unique factorization - on the other hand the easiest way to define a DVR is to use unique factorization of ideals.

So you have to pick which one you want to do the hard way, then the other comes out easily. I've been looking into doing the valuation one first because I already know about the other route.

I had been working on a proof for this for a while and I finaly thought I came up with one, but there was something wrong - I could spot no mistakes what so ever in my proof at all but I knew it wasn't valid because the proof didn't "use" enough tricks and difficult parts.

All of the proofs of this result that I've seen in books had a lot more to them so I was sure something was wrong. After a really long time of not seeing any mistakes at all I asked a friend and he showed me that I had assumed ideals were totally ordered in my ring: This is actually true and it's equivalent to the ring being a valuation ring (which it is) but it was an unproved assumption.


Textpunk

I read about some new scripts that have been included in unicode. <http://babelstone.blogspot.co.uk/2016/01/whats-new-in-unicode-90.html>

I'm not very interested in the emojis and I don't think they belong in unicode at all.


Every script (writing system) included in Unicode is given a four letter code. Here's a list of them all <http://www.unicode.org/iso15924/iso15924-codes.html>


Osage

From wikipedia <https://en.wikipedia.org/wiki/Osage_Nation>


The Osage are a Midwestern Native American Siouan-speaking tribe of the Great Plains who originated in the Ohio River valley in the area of present day Arkansas and Missouri.

Osage /ˈoʊseɪdʒ/[2] is a Siouan language spoken by the Osage people of Oklahoma. The last native speaker, Lucille Roubedeaux, died ca. 2005.[3]

In February 2014 a conference was held by the script's creator, Herman Mongrain Lookout, and the staff at the Osage Nation Language Department, along with UCS expert Michael Everson. They settled on a more detailed representation of the sounds in Osage and accounted for and documented changes in the Osage language. The result of the conference was a script reform; this included introduction of lower case, abolition of two ligatures, and the addition of at least one new character. The reformed alphabet is phonemic.[2]

Bhaiksuki

<http://skyknowledge.com/bhaiksuki.htm>


The rarely seen Bhaikṣukī (or Bhaikshuki) script was used in Odantapuri, North India, to write Buddhist texts around 11th–12th century AD. It is clearly Brahmic, and has a calligraphic charm coming from the use of a thick nibbed pen, with distinct “arrowheads” upon many letters. Indirect source for this information is Anshuman Pandey’s Unicode proposal. At this stage I am just presenting the manuscript letters; later I shall develop a font and/or revision based on the essentials of this script.

Zhang Zhung

<http://babelstone.blogspot.co.uk/2007/05/zhang-zhung-scripts.html>


The Zhang Zhung culture (pronounced shang shung in the modern Lhasa dialect) is an ancient culture that flourished in the western and northern parts of Tibet before the introduction of Buddhism into the country during the 7th century. For a long time there was little concrete evidence of the existence of the Zhang Zhang kingdom described in Tibetan sources, and it was regarded as semi-legendary by some Western scholars, but in recent years archaeological investigation has confirmed the existence of an extensive early Tibetan culture that has been identified with that of the Zhang Zhung.

Tangut

<https://en.wikipedia.org/wiki/Tangut_script>


The Tangut script (Chinese: 西夏文 xī xià wén) was a logographic writing system, used for writing the extinct Tangut language of the Western Xia Dynasty

According to the latest count, 5863 Tangut characters are known, excluding variants

According to the History of Song (1346), the script was designed by the high-ranking official Yeli Renrong under Western Xia Emperor Li Yuanhao's supervision in 1036

Government schools were founded to teach the script. Official documents were written in the script (with diplomatic ones written bilingually). A great number of Buddhist scriptures were translated from Tibetan and Chinese, and block printed in the script

Isn't the Seal script mysterious and beautiful? <https://en.wikipedia.org/wiki/Chinese_character_classification>


Summary

Just looking into the background of the scripts of a few languages and I haven't even scratched the surface and the richness of all these diverse communities and cultures is already overwhelming me.

I would like to make a stronger effort to study history of tribes/dynasties/lost civilizations via writing systems in the future.


P.S. I added a new blockquote feature to my blogging software.