What I did today

Miscellaneous Topics
#1601

Sticking to 2 for now until LWJGL3 moves at least to beta. I can’t keep on shifting the goalposts or I’ll never get this bloody game done.
I’ve fixed all the horribly static-ridden code in the libgdx vecmath package with ThreadLocals but I have no idea how to submit a patch or anything. Or even why it was done with statics in the first place (don’t you guys use threads!?)

Cas :slight_smile:

#1602

ThreadLocals have quite an impact on performance. Whether that is relevant to your use is another matter :slight_smile:

For LibStruct I wrote this hacky, yet super fast workaround:


(see .get() / .set(…))

#1603

It certainly is relevant… I need to transform about 30,000 sprites’ vertices by each sprite’s matrix and write the results out to VBOs - I spend a lot of time doing Matrix4.mul() via various routes. I’ll see if that helps.

Cas :slight_smile:

#1604

@Riven
Would inflate memory usage a little bit, but multiplying/dividing by some factor (>= 8) to index into the threadid2value array might improve performance under contention by eliminating false sharing. I would expect it to be faster anyway.

EDIT: if/when writes are occuring

#1605

@BurntPizza: In LibStruct I mainly use .initialValue() and .get(), so it’s pretty much read-only, which means contention is not an issue. Feel free to benchmark / optimize for your use case.

@Cas: FastThreadLocal instances are rather memory hungry, so I’d advice you to make only 1 FastThreadLocal with a pointer to some class that holds all your (previously static) references to ‘scratch’ objects.


myFastTL.get().vec3tmp.x += 4.1f;
myFastTL.get().mat4tmp.m33 += 4.2f;
#1606

A good plan.

Cas :slight_smile:

#1607

Found out Tom Forsyth agrees with me about the value of unit-tests:

1200×632 730 KB

#1608

I’m glad that I’m not the only one who thinks TDD is a hilariously poor methodology.

Cas :slight_smile:

#1609

You’re not the only one

ME <-----

Kev

#1610

Of course TDD is rubbish, but it is less rubbish than the alternative. My tests regularly catch errors and regressions, and writing the tests improves the initial code too.

#1611

Now, what would you say the alternative was? (I notice you used the singular)

Cas :slight_smile:

#1612

Either you test, or you don’t. It’s a boolean! (I know there are many different ways of not testing, but all of those are compatible with testing too).

#1613

This will rapidly devolve into an “exciting thread” so I shall make a new one! :point:

Cas :slight_smile:

#1614

Kopied the kinda popular game “Dont touch the Spikes!”.

Implemented in haskell using my libs and a tiny little bit of FRP.
(Code: https://github.com/matheus23/DontTouchTheSpikesHaskell)

#1615

Got my new SSD in, and installed ubuntu. I can’t play space engineers, but I’m learning allot about linux. :smiley:

Also got a Rosewill THOR chassis too… Thing’s huge…

#1616

I finally got this on my hackintosh:

A CPU Panic!! Seeing it for the first time in my life! Problem is due to the wrong kernel extension for CPUPowerManagement.

EDIT: As a bonus, I also got my 1 TB HDD completely broken. I can’t see any partitions in it now.

#1617

It’s a good bit of fun, isn’t it? (learning Linux). I did similar a few months ago, and haven’t looked back. I was a bit reluctant to take the plunge, but glad I did.

#1618

I applied to a volunteer organisation that helps young, disabled and elderly people with little to no IT skills learn how to use computers for everyday use. I will be teaching classes starting next Thursday. Really looking forward to it!

It is though a 7.5 mile cycle from my home to there, then another 7 mile to college afterwards then 8 mile home from college. Yay! Fitness \o/.

#1619

I’ve always had bad luck with ubuntu. Mint and centos seems to work better for me.

#1620

Finished off upgrading one of the tile sets for a guys roguelike game.

1280×576 52.8 KB