threerings / clyde Goto Github PK

I noticed the JVM's sine and cosine are used inside https://github.com/threerings/clyde/blob/609b95ac8694853b3c95a4feaa701ba5e2c01960/core/src/main/java/com/threerings/math/FloatMath.java

The problem with the JVM, is it tends to stick to IEEE accuracy too much. I was wondering if this was impacting Clyde significantly? Probably some sort of lookup table could do, but I don't know how much the higher error margins would propagate visually by.

Cursory googlin':
http://blog.element84.com/improving-java-math-perf-with-jafama.html
http://stackoverflow.com/questions/16930581/fast-sine-and-cosine-function-in-java

The error occurs in the read method that readObject() calls. The int size variable that _fieldIdReader returns in that function seems to be faulty, as the next line that creates the Map with the specified Illegal initial capacity throws this error:
java.lang.IllegalArgumentException: Illegal initial capacity: -1318888063

How to use this script?BIP or Dummy type, what's version the 3dsmax version required?

Specify "--add-opens=java.*****=ALL-UNNAMED" type arguments for every single internal JDK API used. In Java 16+, due to JEP 261 (https://openjdk.org/jeps/261), strong encapsulation is enabled by default, and assumed internal JDK APIs are illegal to be utilized via reflection. You can manually break this still via "--add-opens" or "--add-exports" commands.

This issue breaks legacy downstream product(s) by Three Rings.

Description Resource Path Location Type
Reference colorpos.classpath not found. build.xml /clyde-master line 112 Ant Buildfile Problem
Reference testlibs.classpath not found. build.xml /clyde-master line 122 Ant Buildfile Problem
Reference tools.classpath not found. build.xml /clyde-master line 57 Ant Buildfile Problem

Causing 65 errors during compile
I think...
if this is not the case, someone tell me how to compile clyde.

Seeing how it's frozen in development despite having some bad bugs, it'd be a good idea to move up to LWJGL 3.

The OpenAL and JInput JNI files seen in downstream derivative products need to be updated to the latest versions found bundled with LWJGL 2.9.3 to be compatible with recent Java versions.

Additionally, the file jinput.jar under the code directory as well needs replacement in the same downstream product(s).

Sifting through hashCode implementations within clyde, there's excess arithmetic in some cases, and a lack of allowing for some collisions to occur within the bucketing, caused by excess precision in the hashCode functions. Some of the hash codes should collide if it means the O(1) time is lowered overall. Hashes that collide though should have a limit to the collision count in each bucket, causing a compare, if optimized.

tl;dr the hashCode implementations might need a thorough pass.

http://forums.spiralknights.com/en/node/115634
Is it possible to fix it, maybe you can locate new thread for that.

threerings/getdown#272

Causes a failure to launch in recent JVMs for legacy threerings products based on clyde using getdown deployment.

I'd rather give a helping hand to assist in updating it, but I need a reference to build it with.

I'm interested in looking into placing individual models & bounds checking in separate threads, and possibly improve some backing store usage.