🔒 Locked

AngelScript 1.10.0 FINAL (2004/11/02)

Started by WitchLord Nov 2, 2004 at 3:42 PM 44 replies 11k views

Original Post

Nov 2, 2004 at 3:42 PM

This is the largest update ever made in one release, during the AngelScript history. A very shortened list of updates that were made is:

Further optimized the VM with perhaps 1.5x average performance improvement

Quite a few bug fixes

A new boolean operator, xor, was added

The library interface was improved, especially functions that returned strings

ExecuteStep() now works as before (if the correct compile time flags are set)

ExecuteString() has gotten a much needed overhaul. It's now possible to call it recursively, and also from multiple threads at the same time.

The much awaited native support for arrays have finally been implemented. Now script writers can enjoy the full power of arrays in their functions. Application writers can easily register the exact array objects that they need for their functions, which makes the interaction very easy.

Basic support for multithreading has been added, though it can be disabled at compile time for those who need the extra performance. Those are the most important points. You can read the complete list on the AngelScript site. Enjoy!

EddHead

Nov 3, 2004 at 11:40 PM

Aaaargh! All these updates, and No time to implement!

WitchLord

Nov 4, 2004 at 7:27 AM

No worries [wink]

lxnyce

Nov 4, 2004 at 10:21 AM

I finally got some time to mess with AngelScript again, and it seems that the threading is causing crashes all over. None of the tests ran for me, they all crashed at ENTERCRITICALSECTION(criticalSection);

The crashes went away when I declared NO_THREADS in the project. I believe multi-threading is still not working fully yet, but they are on by default in MSVC project found in angelscript\projects\msvc6. Also the "compiling and linking with the library" link at the top of http://www.angelcode.com/angelscript/ is broken.

Not griping here, I LOVE angelscript. Just pointing out the problems I ran into just loading it up after a long break.

Forgot to mention I am using MSVC 7 with MSVC 6 libraries if that had anything to do with it.

[Edited by - lxnyce on November 4, 2004 10:21:25 AM]

WitchLord

Nov 4, 2004 at 10:58 AM

The problem is most likely because you're using MSVC7. The library works just fine in all my tests with multithreading turned on.

Still I will set the NO_THREADS flag as on by default, since multithreading is not yet stable.

Thanks for pointing out the broken link as well.

Regards,
Andreas

caesar4

Nov 4, 2004 at 7:39 PM

andreas
take a look at http://softwire.sourceforge.net, it might help with the performance issues uve havin

Aggrav8d

Nov 5, 2004 at 12:50 AM

1. Congratulations! What's next on the agenda? :)

2. So, just to be clear:

int a[5](6);

would be the proper syntax in C++ to create an array of five floats initialized to the value 6. What is the syntax in angelscript?

3. I'm glad that you added the output message to Build(), but now I get an "Invalid Configuration" and an asINVALID_CONFIGURATION with no idea as to *why* it is invalid.

WitchLord

Nov 5, 2004 at 8:30 AM

ceasar4:

I'm aware of SoftWire, but AngelScript hasn't come so far as to compile directly to native code yet. Once it does I will most likely use SoftWire though.

Thanks anyway.

Aggrav8d:

1. Thanks. What's next is AS2, with the slightly changed syntax. AS1 will also be improved with a few things, like initialization lists, -> operator overload, better compile speed, etc.

2. AS still don't support automatic initialization of arrays, so in this case you'll have to write a loop to initialize each of the elements.

int[] a(5);
for( int n = 0; n < 5; n++ ) a[n] = 6;

3. AS doesn't store the function call that failed during configuration. You'll have to verify each of the return codes. I recommend doing it like this:

r = engine->RegisterObjectType(...); assert( r >= 0 );
r = engine->RegisterObjectMethod(...); assert( r >= 0 );
r = engine->RegisterObjectMethod(...); assert( r >= 0 );
r = engine->RegisterGlobalFunction(...); assert( r >= 0 );

This doesn't polute the code with lots of if statements, and it will let you easily find the problem in debug mode. In release mode the return codes aren't verified but they won't fail if they don't fail in debug mode (unless I made a mistake somewhere).

Aggrav8d

Nov 5, 2004 at 12:44 PM

I manage to hit the assert(false) at the end of Dereference() without getting any kind of error message or code. What should I be looking for?

Digging a little further I see that it is trying to do a conversion but the type->dataType.GetSizeInMemoryDWords() is equal to 16 (!!)

[Edited by - Aggrav8d on November 5, 2004 1:44:23 PM]

WitchLord

Nov 5, 2004 at 1:53 PM

You have probably stumbled upon a bug. I'm guessing it is related to registered types and the way asCDataType changed for 1.10.0. Is it possible for you to give me a sample code that reproduces the assert failure?

Aggrav8d

Nov 5, 2004 at 2:24 PM

Hmm...I can't really chop my code up to provide a simple example, but I can describe the situation. I have an array of pointers to a class with properties and methods. In a loop i call a method that returns new instances

MyClass *[] c(5);int i;for(i=0;i<5;++i) {  c=RegisteredAllocatorForMyClass();  c->RegisteredDoStuff(45.0f,1.0f);}

And I get two problems:
the first, execution dies inside the for loop (but where, i'm not sure yet)
the second, if I change to MyClass [] and get rid of the allocation then I get No matching signatures to 'RegisteredDoStuff(const float, const float)'. I registered as non-const. If I change the registartion and internal methods to const I still get the error.

Also note that if I allocate like this

MyClass *ptr=RegisteredAllocatorForMyClass();

It does allocate (but I still get the No Matching signatures problem). That's all I've been able to figure out so far.

WitchLord

Nov 5, 2004 at 2:36 PM

Ok, that information might be enough for me to find the problem. I'll give it a try and let you know.

But before that I'd like to know if you noticed the post by Licu about the pointer related bug in AS 1.10? If you didn't then this may actually be what is causing your problems.

You'll find the bug fix for the -> operator in that thread.

Aggrav8d

Nov 5, 2004 at 2:51 PM

I added the fix for the -> operator and now I get an assert about

instr->StackSize == StackSize

WitchLord

Nov 5, 2004 at 3:25 PM

There seems to be a problem with arrays of pointers. I've written a test case using your example and it gives errors it shouldn't.

I'm looking into this right now, but I might not be able to fix it until Sunday.

Aggrav8d

Nov 5, 2004 at 3:33 PM

Well that throws my schedule off a couple of days, but I'll find things to work on in the meantime. I wish you godspeed with whatever you're doing :)

WitchLord

Nov 5, 2004 at 4:19 PM

The problem I had found was related to me using integer to index the arrays. Once I change the type of n to uint it worked fine. (NOTE: I will improve the implicit casting rules for the next version)

void Test()                                {                                            class1*[] c(5);                            uint i;                                    for( i = 0; i < 5; ++i )                   {                                            c = class1Factory();                    c->function();                        }                                        }

This compiles and executes without problem (except that the compiler reports c as uninitialized, which is a bug but not a critical one).

The assert(instr->StackSize == StackSize) is probably related to the first problem you mentioned. It seems the compiler is trying to dereference a registered object even though it shouldn't.

This might be related to how you have registered your type. Can you show me how you registered the type in question? The one that was being dereferenced?

Aggrav8d

Nov 5, 2004 at 5:42 PM

I meant to type uint. Sorry.

  asEngine->RegisterObjectType("ObjectInstance", sizeof(ObjectInstance), asOBJ_CLASS);  asEngine->RegisterGlobalFunction("ObjectInstance *CreateObjectInstance(ObjectType *type)",asFUNCTION(::CreateObjectInstance),asCALL_CDECL);  asEngine->RegisterGlobalFunction("void ReleaseObjectInstance(ObjectInstance *)",asFUNCTION(::ReleaseObjectInstance),asCALL_CDECL);  asEngine->RegisterObjectMethod("ObjectInstance","GfxTexture *GetTexture()",asMETHOD(ObjectInstance, GetTexture), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","void SetTexture(GfxTexture *)",asMETHOD(ObjectInstance, SetTexture), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","GfxMesh *GetMesh()",asMETHOD(ObjectInstance, GetMesh), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","void SetMesh(GfxMesh *)",asMETHOD(ObjectInstance, SetMesh), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","bool ComesBefore(ObjectInstance *)",asMETHOD(ObjectInstance, ComesBefore), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","ObjectType *GetType()",asMETHOD(ObjectInstance, GetType), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","void SetPosition(float,float)",asMETHOD(ObjectInstance, SetPosition), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","void GetPosition(float&,float&)",asMETHOD(ObjectInstance, GetPosition), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","void SetVelocity(float,float)",asMETHOD(ObjectInstance, SetVelocity), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","void SetAngle(float)",asMETHOD(ObjectInstance, SetAngle), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","void SetAngularVelocity(float)",asMETHOD(ObjectInstance, SetAngularVelocity), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","void AssignPhysics(ObjectPhysicsDescription &opd)",asMETHOD(ObjectInstance, AssignPhysics),asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","void SetName(string &)",asMETHOD(ObjectInstance, SetName), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","void SetStack(int pos,float v)",asMETHOD(ObjectInstance, SetStack), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","float GetStack(int pos)",asMETHOD(ObjectInstance, GetStack), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","void SetImmaterial(bool state)",asMETHOD(ObjectInstance, SetImmaterial), asCALL_THISCALL);  asEngine->RegisterObjectMethod("ObjectInstance","bool GetImmaterial()",asMETHOD(ObjectInstance, GetImmaterial), asCALL_THISCALL);

WitchLord

Nov 6, 2004 at 11:21 AM

Hmm, I can't see anything wrong with the way you registered the object.

I think I'll need to know what script function it is that is causing the assert failure. Would it be possible for you to find this? Simply comment out the script functions one by one until you find the one that fails compiling.

In the meanwhile I will look through the library code for possible causes.

Aggrav8d

Nov 6, 2004 at 1:30 PM

ObjectInstance *[] instanceArray(5);uint i=0;instanceArray=ObtainObjectInstance(type);

is causing the problem.

WitchLord

Nov 6, 2004 at 2:07 PM

Unfortunately I don't seem to be able to reproduce the problem. Everything seems to work fine for me.

I will continue to search but without being able to reproduce the problem myself it is highly unlikely I'll find it.

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.