Mx?

1
Mx?

• A programming language for scientific
  computation.
• Related Languages
• Matlab
• IDL
• Maple, Mathcad, Mathematica

2
Goals

• Portability - interpreted and executed by a java
  interpreter.
• Efficiency - greatly improves programming
  efficiency. Program involving many matrix and
  vector operations are very short in Mx.
• Ease of use - very simple and quick to start

3
What can the language do?

• Basic matrix operations
• addition, multiplication
• Advanced operations matrix slicing and
  masking.
• Internal functions Print, input, load, save,
  plot, paint

4
Operations and Relations
Arithmetic Operations , /, , /, , , -
Logical Operations not and or
Relational Operations gt, gt, lt, lt, , !
5
Array and Range
Range Specifier i.e.. A110, 12 i.e.. A12
Array Constructor 1,23,4
6
Statements
Block Statement
iteration for (n 1100) loop
Assignments
Conditional Statements if (exp) ltstatgt if (exp) ltstatgt else ltstatgt
Break and Continue break continue
7
Functions and Function call
Self Defined Functions func ltidgt (para-list) exp func ltidgt (para-list)
Function Call id(para-list) Note can either be stand-lone or right value depends on if it has returned value
8
Internal functions

• Console output
• print() print arguments to the standard output
  one by one.
• Picture drawing
• paint() draws a matrix in a new window as an
  image
• plot() takes a matrix and plots it as a graph
• Color
• color() sets the current color (in RGB)
• colormap() take a matrix with rows as RGB, and
  sets a colormap.
• File I/O functions
• load( file, type, m, n )
• save( matrix, file, type )
• Matrix generator
• zeros( m, n )
• random( m, n )
• matrix operator
• inv( matrix )
• flip( matrix )

9
Example 1 Mr. Potato

• A load( "potato.dat", "byte", 128,128 )
• colormap(1)
• paint( A, flip(A), mirror(A), flip(mirror(A))
• A', flip(A'), mirror(A'), flip(mirror(A')) )
• return 0

10
Overview of the interpreter

• Currently the Mx programming language is
  implemented interpretively.
• The interpreter parses and executes the user
  input or programs in files, and generates printed
  output and/or pictures.
• The parser of the interpreter is written in
  Antlr, and the rest routines are in Java.
• For a small portion of our code, we tried macro
  expansion in Java.
• Adequate functionalities made our project quite
  large, luckily most of the functions are tested
  and work well (maybe we have forgotten the
  existences of some functions?)

11
Code Statistics
Parts Front-end Interpreter Matrix class test code
Statistics (lines) 481 2090 2731 941
Total 5891 lines (generated code not included) 5891 lines (generated code not included) 5891 lines (generated code not included) 5891 lines (generated code not included)

• (partial, currently in CVS)

12
Why another matrix class?

• The Mx programming language has important
  functionalities in-place matrix slicing and
  masking
• Existing Java matrix classes do not support these

13
What can we do in Mx with slicing and masking?

• A zeros(256,256)
• A1040, 1
• A2550,6050 random(50,50)
• AAgt0.75 or Alt0.25 0.5
• A100199,100199 random(100,100)
  A0100,0100A0100, 100100

14
An example of masking paint selected entries

• A load( "mri-anat.sdt", "short", 256, 256 )
• B load( "mri-func.fdt", "float", 256, 256 )
• A flip(A')
• B flip(B')
• paint( A )
• colormap(4)
• opaint( B, Bgt0.7, 0, 0, 0.7, 1.0 )

15
The features of our matrix class

• Arithmetic operations - / ...
• Sharing data between matrices
• Matrix slicing and masking
• Inversion and solving linear equations
• Comparing matrices
• Painting and plotting
• Transpose, flipping and mirroring
• And more...

16
The type system
17
Testing Plan

• 3 Stages
• Unit Testing
• Creation of our own library using Matlab as
  guidance.
• Library contains its own unit testing program.
• Regressive Testing
• Creation of a script to automate regressive
  testing written in tcl
• Takes a database of one or more .mx programs
• See example

18
Example of the Testing Database

• samples.tdb a sample of testing
• database for the Mx
• assign
• 1 2 3 4 a 1,23.4 return a
• transpose
• 1 3 2 4 a 1.23,4 return a
• transpose transpose
• 1 2 3 4 a 1,23,4 return a
• triple transposes
• 1 2 4 return 124
• matrix add
• 1 2 3 4 return 1,11,1 0,12,3

19
Testing Plan

• Advanced Testing
• Necessity for larger testing programs
• Integrated all the examples from the tutorial and
  used them as our advanced testing examples.

20
Lessons Learned

• Fully investigate the feasibility of the
  supporting tools
• Original Matrix classes lacked many functions
• Plotting, painting, masking, etc.
• Get advice from experienced
• When an easier method of implementing classes of
  function exists, use it