complete of pass two for measurments and metirics.1) Abandned a decorator memeber fo Measuremnt class to accumulate
metrics. Becasue I did not know how to pass in per-funciton call paramerters.
3) measurements.json is a file made up of various json records. jplot.py
is a small plotter oto produce initial plots that help validate the simulation
and give you something to show. (for example, plot deviaions of function calls
form the model strace...
removed the old code record and replaced it with JSON to propagate measuremnts information to code generations, (adding one such feature last time showed how tedious this could be without records havign more strucuture
This is the frsit version of the code that attempts to store measurementsHow far di dI get?
Implemented an object, Measurement, to support mesurement collection,
thsi colelsct a few summary items, and a short log.
The bulk is in simsupportlib.py.
The measurents are impemented in terms of a decorater. Havig taken a
look, I'm not sure a desorate of what we ant. (remember, we autogenerate
code, and I'm not sure we need the notational elegancesimplicity/maintainbai...
now works wiht a custom file squashed (fro mindependent trace files -- one fromeach thread" THsi version contains hacks that need to be removed under the
assumption the threads wwere like fored children -- the issue is shared file
descriptor in thread v.s non shraed, inherited under true forks.
last of a block of changed speeding up testing. mkfiles.sh checkto see if a file of prport size exists; also the list fo all files to be made is checked so that fikes are made once, not once each time for each time they are opened
improves sleep between IO-- The sleep now works
== simdriver has switches that speedup sleep by a factor, simulating fsater or slower CPUS or sppress sleeps alltogether
-- Spent tiem investigating wy simdrive is making no contributions to logging.
no clude == If I cut out the main program, that makes log entries.
Can now make and test the generated simualtion files in an isolated directory.mkfiles.sh is completely stand alone.
simulation.py is completely stand alone
simulation.c is completely stand alone.
got rid of the toxic from os import *Ths is a staep on building a minimal set of files.
simsuportlib.py now hs definitions from loni xof the
O_ flrgs needed for opens and x_OK flags needed for access.
at th emoment these are not used in the python simulation.
working on making small stand along files for evalaution -- make a ups dir and now pyhsically inlcude suport libs in C an dShell -- python is complicated
tweaked tests into two parts, put high level execption handler in, makes the right files for cahins, but iteration over the chains is not occurring -- only chain one is iterated over
first support of chains.Simdriver now takes a switch indicatng now many chain links the simulation
should run for. Support for looking for the right paths is coded.
Tested for just a chain of a single link.
The control in simdriver is for the total number of links. I Now need
to generate chains of specified length when generating the shell script
that makes files. Simdriver just loops over the links unti the numb...
Donald Petravick
Matias Carrasco Kind
Don Petravick
Don Petravick