Date: Wed, 10 Nov 2010 20:28:33 +0200 From: App Deb <appdebgr@gmail.com> To: =?UTF-8?Q?C=2E_Bergstr=C3=B6m?= <cbergstrom@pathscale.com> Cc: "O. Hartmann" <ohartman@zedat.fu-berlin.de>, freebsd-questions@freebsd.org Subject: Re: FreeBSD-similar build-from-source Linux? Message-ID: <AANLkTi=GKjGCpckXYrkqBRJc-8OUXjKg5jQz3=isOOfX@mail.gmail.com> In-Reply-To: <4CDAE252.1060805@pathscale.com> References: <4CDA8F15.506@zedat.fu-berlin.de> <AANLkTinBgFrHLU5FaknGAeXdixKn_j%2BoTiOOEi7uV_AB@mail.gmail.com> <4CDAE252.1060805@pathscale.com>
next in thread | previous in thread | raw e-mail | index | archive | help
2010/11/10 "C. Bergstr=C3=B6m" <cbergstrom@pathscale.com>: > App Deb wrote: >> >> 2010/11/10 O. Hartmann <ohartman@zedat.fu-berlin.de>: >> >>> >>> Hello out there, >>> >>> well, my question may sound heretic, but since we use mostly Linux base= d >>> systems in our scientific environment and FreeBSD seems to lack in seve= re >>> support in GPGPU/CUDA capable graphics boards I need to setup a kind of >>> Linux facility to ensure having the software and tools I need for my >>> work. >>> I'm looking for a Linux distribution that is similar handled like >>> FreeBSD, >>> where I'm able to rebuild the whole system from sources, not even the t= he >>> Linux kernel, also the GNU tools and the packages. Maybe there are some >>> people out here having already taken this step. >>> Any suggestion is appreciated, >>> >> >> The NVIDIA FreeBSD driver provides the cuda libraries for linux >> compatibilty. >> >> So 32-bit Linux Cuda applications should work on FreeBSD. >> > > Have you tested this first hand? > > Is there 32bit Fermi? =C2=A0(Sorry, but I don't think the HPC world can r= eally > live off 32bit applications) > > There's 2 types of code in a "cuda application" there's host code produce= d > by the native compiler (gcc/intel/pathscale.. etc) and the device code.. > produced by nvidia's toolchain.. =C2=A0I would make a pretty nice bet tha= t the > nvidia userland toolchain will not compile code on freebsd right now.. = =C2=A0(not > be negative here, but I'm intimately familiar with it..) =C2=A0Worse case= you'd > be copying code built on linux over and also the runtime libs.. =C2=A0(no= t to > mention debugger, profiler. etc) > > Anyway.. as mentioned before we'll have a native working solution before = the > years end and anyone interested to help alpha/beta test is welcome to > contact me offlist.. > > ./C > No sorry I don't have any experience with cuda. I have only tested this: http://cuda-z.sourceforge.net/ And the linux binary seems to recognize everything under FreeBSD (linuxulator + nvidia). You will probably need to setup a linux toolchain to do the work. I know this is not a good solution.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?AANLkTi=GKjGCpckXYrkqBRJc-8OUXjKg5jQz3=isOOfX>