Wow! With one exception, those are great bunch of posts and I appreciate it. I have not tried Virtual Box yet, just downloaded it. I note as you have that earlier versions of Linux going back ten years had no problem working with other file systems. In taking that approach, though, what would be the purpose of OpenSolaris, in case I missed something? As I understand it, just opening MS Windows in another window, I still could not transfer data between OpenSolaris' and other file systems. Unless maybe I sent it to another computer on my LAN in Windows and brought it back in OpenSolaris.
Perhaps unlike many others here, my interest is not pure I.T.. My primary interest, for now, is in writing computer models of physical systems. Specifically, testing an idea I have for a solar energy concentrator. A ray-tracing model, for example, would be part of it. One which I will have to write, in Fortran, so that I will have access to its innards. If the first feasibility studies work out, I intend to learn how to use the nVidia CUDA technology, which has been benchmarked in separate studies at 100 to 170 gigaflops per second. The CUDA tools are free, but they are predicated on using either gcc in Linux or c++ in MS Visual Studio, with a significant tilt towards VS. Maybe c programmers are not daunted by this, but as a fossil fortran programmer, I'm used to being able to write my code in much fewer lines, more directly targeted to numerical algorithms and plotting. Looking at OpenGL code just boogles my mind. nVidia's nvcc compiler is a multi-pass system that takes me back over 40 years to when I was trying to get an IBM 650 vacuum tube computer to compile a very simple Fortran II program with multiple stacks of punch cards. It was painful and never succeeded. The nVidia PTX language takes me back to GOTRAN and IBM 70xx assembly language. Ouch. Maybe the IT gods think that physical simulation is an insignificant part of the market compared to databases, but I think that Sun is missing a bet here. One paper I found (on the nVidia site, I think) simulated a galaxy of 16,384 objects with gravity using an nVidia card at 100 gfps or better. The nVidia CUDA site demonstrates an impressive range of complex applications. I can see how they might benefit from a large file system. This market also begs for a higher-level language that puts that kind of computing power within reach of everyone. But so far, I can't see how any of this fits with OpenSolaris. I was looking at it because I have an old PCI+AGP computer that won't take any but one of the nVidia-based Ge8 GT cards (Sparkle 8500GT PCI). If I want to use the less expensive 8600 to 9800 series cards, I have to get a barebones computer with a PCI-express bus. RAID is now tending toward SATA HD. CUDA works on XP, Vista or Linux, and comes with some Fortran wrappers. I once discarded XP for Win2000 because I found that I hated its constant needy security demands. My Fortran compilers are either 10 years old or buggy, and I can't afford $800 for a new one. So, maybe a new dual-processor (possibly SLI-ready), barebones computer with a freeware (Ubuntu or OpenSolaris) OS and freeware fortran that connects with an nVidia 8800 or 9800 card. All less than $1000 if I put it together myself, which I can. I like the idea of OpenSolaris and the Fortran IDE studio that comes with it. But right now, it's looking more like using my old Absoft Pro Fortran 5.0 compiler or MS Visual Studio with XP, and turning off all the security popups. Since I'm still trying to find a graphics package I can use (DISLIN went away when I found a serious bug in OpenWatcom Fortran; it executed a comment line and failed to execute the execution line above it), I'm willing to be convinced. But it all has to connect without costing so much pain and money that I can't get the job done. My old Absoft Pro apparently can compile Fortran 77 (my favorite), Fortran 90 and C++ modules and code in the same program, and has a workable IDE. If I want to get into CUDA, I can put XP and the one 8000-series card that fits my old computer for about $240+shipping. Maybe with an upgrade to the power supply. I would still have either to buy VS or to figure out how to hack nVidia's nvcc compiler to work with my Absoft compiler, and then spend more money on hardware later. If Sun can show me a shorter route through the shallow part of the swamp than I see now, with written tutorials, thank you, I'd be happy to go that way. Live long and prosper, an old fossil fortran programmer -- This message posted from opensolaris.org
