bgb system News

if anyone cares, there is a reason why I haven't uploaded anything in the last month or so.

mostly they were related to vague attempts to improve my physics system, and now the process of writing a c99 interpreter.

I will probably upload something when I have more implemented and working or such...

if anyone wants to email me, they can use my real email:
cr88192@hotmail.com

if any email was sent via the sourceforge mechanism likely I wasn't recieving it (it was being caught by spamfilter). this has been fixed now (at least hopefully).

if anyone wonders why there has been no activity since about march, it was because the month of april I was "in transit" pretty much the entire month (about the end of march to the end of april), with no real computer or internet access.

I am back so activity may generally start up again.

a documentation effort was completed.
a vfs type mechanism was also added.

the lbcc project is a compiler for my language. recently most work has been going twards this.

I plan to reimplement a lot of the vmsys features here and then drop this onto my os...
much work is still needed, eg: net, threads, message passing, compiled<->interpreted code integration, ...

this could be cool I think.

pdlib is a small c library I wrote for persistence features (intended more for stuff not related to my main project...).
it intends to try to avoid interfering with normal programming practice as much as possible, mostly one just defines structures and uses special alloc functions.
it is not really gueranteed to work that well at this time, it was beaten together over the past few days and has had very little testing...... read more

implemented sprites, they have angles and such but at present don't have multiple frames.

the problem with the net was actually apparently that my net code was too slow... I had underestimated how slow it was and assumed it was not the problem...

I have typex mirroring working. basically this was a lot of detail stuff.

I was also working on sharing 3d worlds and such. I have been generally battling with tcp performance.
I had implemented a kind of ping/pong like feature for preventing sending updates too fast.
I have yet to figure why it is so slow for such small amounts of data (why should 200 bytes take 200-400ms to get from one process to another?...).
TCP_NODELAY does not seem to have much effect for loopback.
I am thinking: maybe loopback is horridly inefficient for this?...... read more

I have a new protocol which I call "MRP 0.1", which (though I have a more gradual approach to things normally) involved a lot of changes from the last one (which I called "MRP 0").
I added lazy mirroring. this should fix some of the problems I had before. circular structures/object identity work more correctly now.

I printed a book as well ("Concurrent Programming in Erlang"), this was legal (as far as I know), the book was freely released albeit without "part 2"...... read more

I had realized that there is a problem:
I have implemented eager mirroring, where I need lazy mirroring...
I will likely have to redesign the mirroring to make it lazy.
the idea will be that an object is first created as a placeholder, where it will notify mirrors of existance but they will not send updates.
once the object is applied it will be mirrored from another node (possibly pulling in more placeholder objects).... read more

progress is slow recently, I have been getting distracted playing this game a lot recently...

I at least contribute some money to them, I bought the game, as opposed to sign and udeden ("legend of twilight's bracelet")...
I am probably the only one around here (family) who likes this...

yesterday I added hash tables, like other things they have the notation (<hash> <key>) for reference and (<hash> <key> <value>) for binding.

they are created with the 'hash-new' builtin.

I was optimizing the interpreter (what I could). I got it about 3x faster in some cases (script-heavy stuff mostly).
I was not able to get rid of consing arguments though. if pulled off this could have a reasonable impact on performance.... read more

I have been making some changes to the object system.
I have merged slots and handlers, which doesn't really change all that much but is a minor improvement.

I have been implementing stuff to try to handle synchronizing "mirrors" of objects. the objects will synchronize in terms of mirrors.
each slot could have a timestamp, and updates are sent with associated timestamps. it will be a "newest wins out" scenario, so inconsistencies will be allowed but it is assumed that minor ones will be beaten out during the process of syncing.... read more

ok, this morning was dealing with bouncing and energy transfer, was working on an annoyance of mysterious energy gain.
kind of cool is that now the:
A is moving, B is stopped;
A hits B;
A is stopped, B is moving; effect works acceptably well.

I have variable friction and bounce.
a lot of the basic stuff is done now.

I have a general means by which the camera can control/be an entity (the world is mildly interactive).

was considering allowing transfer of force between objects.

I could also go and implement a texture system though as well...

I don't know, for me now it is late...

ok, this morning (spending the rest of the day having to do other stuff) I had implemented a physics system.
sliding was a problem, solutions are still being worked on (I had got sliding basically working, but I am fiddling with it to do it better/more generally).
current ideas are mostly along the lines of slightly adjusting the box sizes and jittering the object away from the surface.
other aspects (like mass or generalized friction) have not been done yet. impacts exerting force and such has also not been done (at present I am not sure how this would be pulled off).... read more

I have added 3d objects type stuff to the gui code (currently non-interactive). it was implemented somewhat differently than bgf, and has a much ugglier syntax (no supprise, this time it is code in my lang). the bgf syntax would have been difficult in my lang given the evaluation rules and such (I could have implemented them via macros, but this is a technical kludge in my mind). it would not be too difficult to write a bgf parser though.
the primary syntactic difference is that properties are given as 'name: arg' as opposed to '(name arg)', and the names of things are different.... read more

yes!...
I now have persistent links working after a long period of screwing with it (and doing work/watching tv...).
some bits were a little hacky, but this is acceptable.

I can imagine a fair amount of stuff that needs improvement as well (such as possibly using guid's to identify nodes within the protocol...).

for now I may update and then continue fiddling with it.

I fixed some more problems with the store (a typex related bug could cause data corruption and such).

I almost have persistent links working, most of the code is in place but at present there seems to be a bug reconnecting after restore...

I will likely post an updated version once I get this working...

guids are now used to identify nodes. the guids are persistent. unlike net addresses or node names guids can be viewed as stable and unique (unless one copies the node).

I have added a set of input links ("scions"), other nodes recieve a scion number as a reference. these are viewed as consistent between runs (the swizzler and compactor make memory addresses not particularly stable...).... read more

I was able to get the compactor working. it was a result of a yet not understood bug:
x&=~f;
should clear the bits which are 1 in f (as opposed to setting x to 0). I have encountered this before and I think it is related to gcc.

I had disabled the gc during compaction as it seems a badly timed gc (in the middle of running a compaction) will do a fair amont of damage...

otherwise with a little fiddling it was working right.
tommorow I can probably work on it more and see if bugs pop up...
more detail stuff is needed as well though.

I am implementing a compactor.
I am having difficulty getting it to work, somehow the chunk flags are apparently getting cleared during compaction. this causes the chunks not to be properly unmapped thus causing store inflation.

if I can get it working I can make more use of the store (as it will stay a more constant size).

oddly getting relocation of objects working was not very difficult, it is some other stuff being a pain...

whether I do anything or not no one cares...
my existance is a waste of others' time.

I work on project and no one downloads it, really there is probably little use in having made it.

I am driven by impulse, impulse is all that keeps me going. I have little point or reason otherwise to do much of anything...

ok, for the past few days I have been busy and have not done much (so the image from the 27th is fairly up to date...).
before that I had been working on the distributed stuff, I still have a number of features I had wanted to code (message flow control and "splinter evaluation") but have been busy.

the root/scripts directory contains examples which demonstate a lot of the general workings of the system (many were constructed to test things).... read more