|
发表于 2011-5-20 18:09:03
|
显示全部楼层
回复 4# onesupeng 的帖子
I did not curse EXA or PowerFLOW. EXA sold PowerFLOW to NASA to conduct a test funded by Office of Naval Research (ONR) around 1995(?). For 6 months and 256 CPUs of SGI the cost was about USD$100,000 (cannot remember the exact figure). The test started with flow past a 3D prolate (football like), it quickly run out of memory thus could not do it. They then decided to do the flow past a 2D NACA-0012 airfoil with Re between 500 and 0.5 * 10^9. In some case PowerFLOW yielded negative drag coefficient. When NASA people asked EXA to explain, the answer was always "The Digital Physics of EXA is intrinsically better than the Navier-Stokes equations ..."
When I jointed ICASA at NASA LaRC in 1996, I was asked to find out what's going on. Because EXA would not tell us exactly what they did in PowerFLOW, we wrote our own code and compare our solutions with those of PowerFLOW and other Navier-Stokes solvers. What we found was that the boundary conditions in PowerFLOW caused the problem. So we published the results in J. Stat. Phys. (see attached paper).
Since then I was EXA's public enemy Number 1. End of the story.
[ 本帖最后由 luo@odu.edu 于 2011-5-23 08:25 编辑 ] |
|