On Mon, 09 Feb 2015 14:27:02 -0200 Breno Leitao <bren...@br.ibm.com> wrote:
> Hi Konstantinos, Hi Breno, > Thanks for being interested in it. I am also willing to help. > > Why not using a remote VM on a new POWER machine hosted at Unicamp? Already have 2 of those, one ppc64el and one ppc32 (but on ppc64 hardware). The thing is that I want to actually test things that involve local testing (for example, I've been contracted doing some ffmpeg altivec optimizations [1] (ppc32 for now) and I found that some code passes testing (make fate passes) but produces wrong visual artifacts, and without an actual display I can't really see the problem). > You have been doing so using the IBM POWER mini cloud[1], and the > mini cloud will continue to exist for long. In fact, the minicloud > has some idle resources for now, and it should grow soon. We are > expecting to receive some new POWER8 to add to the mini cloud. > > Other than that, if you have any special requirement (as a full > machine), I can work to have it for you, during the bootstrap > process. Let me know. Thank you, for now my needs are quite modest, I definitely appreciate being able to test VSX code on both LE/BE systems, and being able to test Eigen nightlies on the systems (ppc64le VM already runs nightlies, need to setup the BE VM as well to do that). > Also, We are trying to replace Debian infrastructure removing old > powerpc hardware used to do building and replace it with VMs on the > new POWER machines. If ppc64 gets accept in Debian, I understand that > they would use VM on the current POWER machines that the DSA owns, > other than these old hardware. Yes, and it would definitely make sense too. As ppc64le port has proven, build speeds are severely reduced on this hardware :) Regards Konstantinos [1]: https://github.com/markos/FFmpeg/tree/feature/morealtivec -- Konstantinos Margaritis email: mar...@freevec.org, mar...@debian.org blog: http://freevec.org gpg: 3FC3 7391 FECD D3CB 5501 7121 6606 027F 6437 124C
pgpVkcLhQLvx9.pgp
Description: PGP signature
