Benchmark (Overview)

Of primary interest to our users is how fast the gdsriplib can rasterize a large area data file at high resolution. This depends on many paramters:

• The size and complexity of the input data file

• the physical size of the area to rasterized

• the resolution or DPI

• The available hardware - CPU's, memory and disk speed

• proper use of threads and memory

• relative contribution of formatting/compression associated with the rasterization.

Each user will need to run their own benchmarks -- the data we provide here is focused on one application provided by an early adopter. However it will give other users some feel for the throughput and for some of the parameters selected.

Flat Panel Example

The project involves the rasterization of a large flat panel array. This particular flat panel is about 2000 mm x 2000 mm in physical size. The input GDSII data file is quite small since the flat panel data is highly regular and the GDSII file uses many nested arrays. (We cannot provide the data due to its proprietary nature.) An illustration of the panel array is shown below:

You can see that the array is made up of 15 panels - for time reasons, some benchmarks are run on the "pannel3" structure.

The Computer Hardware

Our primary machine for testing and development on this project is a dual Opteron workstation. Initially we configured it with:

• 2 x 1.6 GHz Opteron CPU

• 4 GB RAM

• dual SATA disk drives configured as RAID 0

• Red Hat Linux Enterprise 3.0 64 bit edition

Later, we modified it by adding 4GB of memory, changing the disk drives (not voluntarily, from SATA to ATA) and changing the OS to Red Hat Enterprise 4 (see Page 3)

gdsriplib Benchmark   1   2   3