From my days as a graphics programmer (OpenGL/win32) for the demoscene. It was a collaboration with Joseph Mocanu.

## Wednesday, January 28, 2015

### Wednesday Night Non-Hack #10

Worked late tonight (abstract for SNUG this year) and I am too tired to spend any time working for fun.

I received the book I mentioned last week and learned quite a bit about how constraint programming works in the real world (not these clumsy for-fun hacks that I have been playing with). The next logical step would be to find a way to represent a constraint as a BDD. There appears to be a way forward with pyeda.

I don't think I will spend much more time on constraint programming, but I remain committed to bring this side project to some logical conclusion.

I received the book I mentioned last week and learned quite a bit about how constraint programming works in the real world (not these clumsy for-fun hacks that I have been playing with). The next logical step would be to find a way to represent a constraint as a BDD. There appears to be a way forward with pyeda.

I don't think I will spend much more time on constraint programming, but I remain committed to bring this side project to some logical conclusion.

## Wednesday, January 21, 2015

### Wednesday Night Hack #9 - Generation of constrained random stimulus using python-constraint with bit-level approach

One of the issues faced my in last experiment with constraint programming for random stimulus generation was that the constraint solver library only accepted input variables of finite domain. As it turns out, all available constraint solvers share this same constraint. The state space explodes when specifying 16-bit variables and it became difficult to solve a simple constraint. For the purpose of stimulus generation, it is absolutely necessary to randomize large payloads;128 or 256 bit wide buses is not uncommon. One approach is model one 2-bit variable for each bit of the random variable. As it turns out, it works well. This is commonly called, if I am not mistaken, a bit level approach.

On some other related notes :

I ordered this book on constraint based verification. I suspect it will heavy on the computer science, but for $17.99 I couldn't pass it up. I will give it a quick read.

Also, I stumbled on this web site: http://www.personalkanban.com/ - be forewarned, it's new "agey", similar concepts to "GTD". Well worth a read through if you are interested in personal growth. Quoting: "Personal Kanban gives us clarity in our work and our lives by visualizing those tasks, expectations, and commitments we have and helping us prioritize and complete. With only two simple rules, visualize your work and limit your work in progress."

Without further ado, here is the source code to tonight's experiment. Enjoy!

On some other related notes :

I ordered this book on constraint based verification. I suspect it will heavy on the computer science, but for $17.99 I couldn't pass it up. I will give it a quick read.

Also, I stumbled on this web site: http://www.personalkanban.com/ - be forewarned, it's new "agey", similar concepts to "GTD". Well worth a read through if you are interested in personal growth. Quoting: "Personal Kanban gives us clarity in our work and our lives by visualizing those tasks, expectations, and commitments we have and helping us prioritize and complete. With only two simple rules, visualize your work and limit your work in progress."

Without further ado, here is the source code to tonight's experiment. Enjoy!

#!/usr/bin/env python2.7 from constraint import * from random import seed class Solver: def __init__(self): self.solver = Problem() self.solver.setSolver(MinConflictsSolver(100)) def add_variable(self, name, length): # Each bit of input variable is in binary domain, internal format is name[bit position] variables = ['{}[{}]'.format(name, i) for i in range(length)] for variable in variables: self.solver.addVariable(variable, [0, 1]) def add_constraint(self, *args): self.solver.addConstraint(*args) def randomize(self): self.solution = self.solver.getSolution() def get_variable(self, name): # generate decimal representation from internal bitwise representation remap = 0 for variable in self.solution: if name in variable: index = variable.replace('[', '_').replace(']', '').split('_')[1] remap += 2**int(index) * int(self.solution[variable]) return remap #seed(1) # fix random seed, if ndeed solver = Solver() # init solver object solver.add_variable('xx', 32) # rand bit [31:0] xx solver.add_variable('yy', 15) # rand bit [15:0] yy # x is EVEN solver.add_constraint(lambda x: x == 0, ['xx[0]']) # x is EVEN # y is ODD solver.add_constraint(lambda x: x == 1, ['yy[0]']) # y is EVEN # solve the problem solver.randomize() # return random solutions for variable xx and yy print(solver.get_variable('xx')) print(solver.get_variable('yy'))

## Sunday, January 18, 2015

### Source code for M.Sc thesis (2007) released

I have uploaded the source code to my M.Sc thesis project. The main reason why this is worth checking out is because of the use of SystemVerilog features which were used in the feature. The features include packages, typedef, interfaces, modports, always_comb blocks, and a good amount of parameterization. Considering most of this work was done in 2006-2007, I think its pretty impressive.

Please visit the following repository : https://bitbucket.org/ecote/system99/src

Please visit the following repository : https://bitbucket.org/ecote/system99/src

## Friday, January 16, 2015

### Early January Hack - Practical RISC-V Random Test Generation using Constraint Programming

As hinted in my previous blog post, here is the presentation I did for RISC-V workshop :

I will not release the test generator quite yet. The method I chose to input the problem to the constraint solver is too verbose. I am already looking at an alternative solution using Python introspection and or-tools from Google.

I do have a patch for spike to allow the tool to be controlled from a different process. I can release that if there is interest.

I will not release the test generator quite yet. The method I chose to input the problem to the constraint solver is too verbose. I am already looking at an alternative solution using Python introspection and or-tools from Google.

I do have a patch for spike to allow the tool to be controlled from a different process. I can release that if there is interest.

## Saturday, January 3, 2015

### Mako templates for Verilog, now released

I cleaned up and published code I presented earlier this year on the use of Mako templates for Verilog.

Here is a link to the repository: https://bitbucket.org/ecote/vmako/overview

### Update to SimpleMIPS to use elftools, improved tracing, and new instructions

In the spirit of the new year, I am publishing some code that I have laying around on my workstation.

SimpleMIPS is an existing Python-based instruction set simulator for MIPS. I wrote some code to replace its roll-your-own ELF parser with code from elftools library, to add simple for instruction trace generation, and add support for handful of new instructions.

Here is a link to my the repository : https://bitbucket.org/ecote/miss/overview

SimpleMIPS is an existing Python-based instruction set simulator for MIPS. I wrote some code to replace its roll-your-own ELF parser with code from elftools library, to add simple for instruction trace generation, and add support for handful of new instructions.

Here is a link to my the repository : https://bitbucket.org/ecote/miss/overview

Subscribe to:
Posts (Atom)