pheon

luajit buffer serialization is amazing fast and also surprisingly very space efficient. this should really be at the top of the thread.

long time, bravo Mike! is there a youtube of all the VU coding comps submissions while it was running. think it was 2 or 3 years the comp ran for?

we are actually doing that now at https://fmad.io for SDR instead of raw oscilloscope ADC. SDR is about an order of magnitude less data e.g. 6GS/sec instead of 60GS/sec and Matlab as the backend instead of Python.. as its a bit faster when mining terabytes of data for that ah-ha wtf bit.

Surprisingly cool going from raw IQ/analog data to ethernet frames much like the op has written.

Indeed awsome hardware to code for. It was amusing when the Sony guys joked people are writing their VU assembler in Excel. To make it easier to mentally keep track of the instructions in each pipe.

The architecture and design of PS1/2/3 were driven from Tokyo, where as PS4/5/+ are driven from the US. It happened because the internal corporate structure changed significantly, leading the design to be heavily influenced by game dev community vs hardware engineers.

As such... its sad as we will never see any new crazy ass hardware architectures going forward.

God dam wtf! Catch up with you offline :)

Is there any throughput performance numbers published?

Not possible, write bandwidth of tape is not very high.

My guess is they have a cabinet / 42U worth of HDD arrays, by the numbers they need ~ 176Gbps write bandwidth which is totally possible even in half a cabinet.

400TB worth of SSDs @ 176Gbps write bandwidth is also very possible, would be ~ 4U expect but extremely expensive.

Yup, there are exceptional engineering talent in these mega corps, rare but it does exist. Usually the talent gets filtered by Division X progression. Eventually the best end up in Division 1, which is good but it takes a long time for this process to do its magic... but thats ok you`ve signed up for 30 years of service.

Thats the theory, sometimes it works in practice, sometimes other shit screws it all up.

In corporate Japan, its even numbered too. Software Division 1, 2, 3, 4, 5

Platoon, Company, Battalion 1, 2, 3, 4, 5 etc.

Of course the goal is to get into Division 1, as its the most prestigious.

Peanuts. Japans labor market structure is based on the army/military mode of operation.

Imagine your skilled with a riffle because.. you like to shoot stuff out on the farm. Sign up for the army and you will be shipped to bootcamp where they teach you how to shoot "correctly", polish your shoes and make your bed and most importantly chain of command.

Note: commanding office says jump, you say how high.

You have your platoon where everyone entered at the same time, your all buddies, get shat on by the peps one year ahead of you and everyone progress at the same pace.

Note: skill as a good shot has no relevance to your rank and pay.

After you have put in time you can advanced in rank and thus pay. Get married / kids theres additional benefits.

Note: your rank and pay are directly related to years of service.

Unlike the army, you will not get any medals, heroics of jumping on a grenade to save the team are frowned upon. Its better to hold a meeting with the team, to clarify what the grenade is, check all possible outcomes, investigate every tiny detail by which time everyone is dead. Key point is everyone, the dead part is largely irrelevant.

After some time, your buddies seem to like you and your leading the platoon. Congratulations you have now advanced your career to the fast track lane. Moving slightly faster than your peers as they are "workers" and not management.

Note: fast track has nothing to do with your shooting skill.

.. and so on and so on.

source: lived and worked Tokyo for 10+ years.

Am seriously pissed at this. MS has been slowly acquiring all the tools used for my business, its like surround and siege warfare. Skype, Linkedin, now GitHub..

The time horizons on these kinds of acquisitions are 5 years out. In that time you`ll log into Windows 20, single sign in to Skype, Linkedin, Github all re-branded, all re-written code, with messed up ELUA and a shitty product.

Dont believe todays fluff and BS, the end game is many years out.

According to the recent update of the Japanese Telecommunications act, its illegal to record the packet payloads without a court order. However recording the headers, e.g. TCP headers if fair game.

So its perfectly legal for ISPs and agencies to record the metadata. Then again Japan dosent have freedom of the press and many other things you might expect.

Yup or your design is just large. Due to the non-deterministic nature of the compiler guiding it at a high level makes it less likely to choose resources in weird locations. e.g. I roughly map out block ram assignments for some of the top level modules but still give it plenty of wiggle space.

Verilog its like going from 2D -> 4D

And thats for real complex projects, not making LED`s blnik.

ah your siding on the lisp machine vs todays hardware arch. I think if the lisp hardware architecture has significant advantages we`ll see that emerging in "soft cpu`s" on fpga`s. Particularly as the Xeon/FPGA gear gathers steam.

Precisely. Also the instruction sets have been designed for compilers vs minimizing gate counts. Making it easier for compilers to schedule optimally - alot less weird and bizzare shit they have to deal with.

As for 2) totally agree, GPU and FPGA programming has become "the new assembly language".

It use to be you could drop from C/C++ to assembler and gain massive performance boosts. These days with C intrinsics there`s no need for pure assembly code. But droping to GPU or FPGA code is a total must now, if you need any significant juice from your system.

Its interesting reading all the comments here, and think most of it misses the OP`s point. That software development as a JOB means working coherently with other people / teams. The result: you need to be complicit to work effectively within that team/companies imaginary tower/religion. Otherwise you`ll loose your job or become "that" PITA engineer.

What`s worked well for myself is starting a small niche tech business. We`re small and focused on resolving customer problems. Not bickering over class hierarchy, design rules or other crap thats irrelevant to building products that customers really enjoy.

As someone who also a "foreigner" who also owns and runs a company here in Japan, much of the article was thats so entertaining... now... but was so not funny at the time.

Unfortunately the vast majority of foreigners here are 20-something year old assholes who demand Japan to be the same as their home country or clueless tourists enjoying the sights and sounds. Its though this lens that Japanese people experience either directly or indirectly foreigners and like everywhere on the planet, people remember (and exaggerate) the negative aspects like it was yesterday, but quickly forget the neutral and positive experiences.

If you can digest the above, then its easier to understand the Japanese perspective which is, all interaction with foreigners has substantial risk attached. All this means is, if a task has a local interaction then its likely got an additional risk management component to it.

This risk component is probably different, possibly offensive based on your own cultural norms, but if everyone has the same expectations and culture as you, then your not living in a foreign country.

didnt RTFM thus missed an important detail.

Put another way, hes so smart but losing money in the market thus the market is "wrong" and someone is cheating.

modify the luajit frontend. Added continue and != and some other weirder syntactic sugar, was only a half dozzen lines of code.

i think your analysis is slightly flawed - you have no anchor / point of reference.

maybe a better question would be, for ever hour worked, how many ticks are issued. aka normalizing against the number of patrol men/women active for a day or an hour.

e.g. more police on patrol == more tickets, but does not mean each police man/woman is biased to issue more tickets near the end of the month to meet their quota.

also consider counting backwards. e.g end of the month = 0, 1 day before end of month = 1, 2 days before end of month = 2 etc..

my 2cents.

being a good data scientist is about having enough intuition about the dataset to ask the right question aka form the hypothesis.

working out the question is what makes it a hard(and creative) process, and then you can apply your ML toolbox.

edit: whats different from a data scientist vs analyst/statistican is they build their own tools as the datasets are too massive & non-standard for the usual toolset.

any position that requires phd level math/physics/chemistry etc is extremely rare even in the west.

job descriptions like to say phd level blah blah blah but rarely do they actually require such a theoretical background.

My experience is the best coders ive met who get shit done, dont have CS degrees. If they do have a degree then its either in real engineering(building hardware) or pure math/physics.

It shows how corrupt the Japanese government here is. Not only did they sign RIAA into law with little fan-fare.. they made it a criminal offense by default.

If you asked the RIAA exactly what they dream of, this is it.

Because Korean is hungry, just like China.

Japan thinks nothing will change, everything continue smoothly, nothing bad will happen.

... and yes I live and work here(Tokyo) for many years.

5 years Tokyo here. That is one of the most concise posts on the topic ive yet read, well done.

I see the problem as the time scales of events. If reactors blew up once ever 3-5 years everyone would be watching and monitoring, lobbying, training, drilling, ranting etc.. but they dont.

Humans just are not designed to train and prepare for 1 event in 20 years. Maybe if Buddhist monks were in charge of the reactors we might have some hope... but they are not.

its the secret handshake that only CS grads know. If you cant` recite the CS pledge then you must be lame..