SPARK

Add restart function ,if some error happened somehow , you can restart the program and continu compute from the latest failed iteration

SPARK

1.save picture and data to local disk instead show them in the java window
2.add some connectors

java thread pool edition

This version is not recommended , because it's robustness is not as strong as GPU version( GPU version can also use java thread pool when there is no GPU card on computer).
there may be some import parameter to change to realize high robust.

First Spark Version

you can use local mode which can set parallelism by setting "local[N]"
Or you can run spark on standalone or yarn cluster manager to reallize distributed parallelism.

GPU version

GPU version can only use cpu device in this time, when using cpu ,it compute with java thread pool.
we may design the code that can be compiled to openCL code in the furture.So the application can utilize GPU computing.
But even though in java thread pool mode ,it can show much more rubost than our parallel version which use java thread pool too.

Serial version

1. This version contain three benchmark,boundary condition 1 and 2 are cantilever beams and boundary condition 3 is MMB
2. Optimization method in macro and micro are both SIMP method.
3. The effective material properties is computed by homogenization method.
4. All parameters you can change can be configured in file parameter.xml.