mpsquark / bavcl Goto Github PK

Add dot and cross product calculation for Vector3 class

Add functions to allow writing Vectors to bitmap.

Add the Tabular/Table/Dataframe data structure which will act similar to pandas dataframes.
-basic functionality

Ability to manually choose when to sync the data to the CPU, so it doesn't sync to CPU after each operation improving performance (raw speed).

Add Extensions methods for;
float
float[]
float[,]
double
double[]
double[,]
int
int[]
int[,]
long
long[]
long[,]
bool
bool[]
bool[,]

To Print those cleanly to the Console for improved debugging experience.

currently DeCache accepts a hashset of ID's it is to not remove during decache operation. This is to be moved to a public variable in the gpu class and to be handled by the GPU class.

Replace seperate dictionaries storing memory size data, memory buffer data and other related info into a singular dictionary of structs.

Add Indexing functionality to the Geometric.Vector3 class

Add a data type optimised for performing data masking.
Also known as Vector

Update the version of ILGPU used in the library to 1.0.0 Release

Add normalisation calculation to Vector3 class

add overloads for OP and OP_IP for Vector3 class

Found an issue here that makes it so that UpdateBuffer is constantly allocating new buffers instead of using the buffer that already exists.

Add functionality for converting to & from vector (singular and multiple) to XML.

Magnitude Calculations for vector3 class

Add functionality for converting to and from vector (singular and multiple) to JSON.

Add an additional optional parameter to in-place functions to further optimise chained functions by omitting repeated code sections.

Incrementing live count will be exclusive to chain beginning
Get Buffer will only be called once for the inplace variable at chain beginning.
Decrementing live count will be exclusive to chain ending

Sort out the structure of the project
Seperating the Vector and Vector3 classes into separate namespaces
Improving base classes and refactoring to improve readability.

Add Distance calculation for Vector3 class

Add matrix multiplication (and other operations like addition, subtraction and division, to be performed 'matrix-wise') to the Vector class

Test if generating an output Vector using Copy and Cache Methods

Test:

Vector Output = new Vector(gpu, new float[length], columns) 

Output.IncrementLiveCount(); 

Output.GetBuffer();  

VS

Vector Output = Vector.Empty(gpu).IncrementLiveCount().Cache(new float[length], out Memorybuffer buffer); 

Provide methods for integration between the Dataframe data structure provided by Data Science Library and SQL and vice versa.

Provide functions for Lambda Cold Dark Matter model Cosmology calculations in Cosmology namespace.

Currently any infinity value, will be displayed as 8, to avoid confusion with the number eight this will be changed to Inf. Likewise for negative infinity.

Add vectors which can hold string data. Important update for a future feature which will add Tabular functionality (similar to pandas.DataFrame)

Write Unit Tests

The function optimiser would accept an array of function commands Fx[], an array of inputs In[] and would execute the functions in the most optimal way, reducing GetBuffer() calls, and other overhead, and especially reducing the quantity of new Vector objects generated.
The function called OK or COMMIT provided by this class would be used to trigger the functions to be called.

Similar to to how queries in SQL can be called but execute together upon the commit call.

Update ToString to use the highest & lowest NON-infinite value to determine digit parameter

Currently calling AccessRow or AccessColumn returns a new Vector. Making a new Vector object is slow as it needs to be cached, sometimes the user just wants the raw data, hence AsArray option needs to be added.

Completely overhaul the geometric vectors class called Vector3 to all use the new Least Recently Used caching system for performance improvement.

Will allow running multiple functions on a single dataset or multiple functions on multiple data sets.
Most likely using streams. Maybe using 2D array views.