| 02:02.54 | *** join/#debianppc camm` (~user@pool-173-61-191-70.cmdnnj.fios.verizon.net) |
| 02:27.22 | camm` | any doc on how to write a linker veneer for ppc64el? I have ppc64 working, but the function descriptors appear different. |
| 03:17.34 | *** join/#debianppc camm` (~user@pool-173-61-191-70.cmdnnj.fios.verizon.net) |
| 05:18.25 | *** join/#debianppc AbouEmre (~Thunderbi@41.87.3.26) |
| 06:43.36 | *** join/#debianppc markos__ (~markos@178-147-196-255.mobile.cosmote.net) |
| 06:59.22 | *** join/#debianppc apt (ibot@rikers.org) |
| 06:59.22 | *** topic/#debianppc is The Debian GNU/Linux PowerPC Port: http://www.debian.org/ports/powerpc/ | Debian CDs are not bootable on OldWorld PowerMacs |
| 07:06.08 | *** join/#debianppc markos__ (~markos@178-147-192-208.mobile.cosmote.net) |
| 08:40.25 | *** join/#debianppc cascardo (~cascardo@177.81.13.23) |
| 11:31.45 | *** join/#debianppc cascardo (~cascardo@177.117.8.99) |
| 11:51.02 | *** join/#debianppc mauricfo (~mauricfo@177.193.227.210) |
| 12:08.09 | *** join/#debianppc Mo (~Mo@unaffiliated/mo) |
| 12:13.46 | *** join/#debianppc tdaitx (~tdaitx@177.97.74.250) |
| 12:18.40 | *** join/#debianppc francsd (francisco@nat/ibm/x-yopvdqhyanwsgeyi) |
| 12:18.46 | *** join/#debianppc cascardo (~cascardo@177.117.8.99) |
| 12:38.01 | *** join/#debianppc camm` (~user@nat-tvwna-outside-visitornet2-a-114.princeton.org) |
| 12:40.11 | camm` | docs detailing abi differences between ppc64el and ppc64? Specifically in writing linker veneers/reloc docs |
| 13:05.29 | *** join/#debianppc AbouEmre (~Thunderbi@41.87.3.26) |
| 13:46.38 | *** join/#debianppc AbouEmre (~Thunderbi@41.87.3.26) |
| 13:51.39 | *** join/#debianppc hartb (~hartb@32.97.110.56) |
| 13:56.27 | *** join/#debianppc leitao (~breno@179-125-236-54.desktop.com.br) |
| 14:30.10 | *** join/#debianppc leitao (~breno@179-125-236-54.desktop.com.br) |
| 14:47.48 | camm` | ppc64el has no function descriptors? |
| 15:04.43 | *** join/#debianppc apt_ (ibot@rikers.org) |
| 15:04.43 | *** topic/#debianppc is The Debian GNU/Linux PowerPC Port: http://www.debian.org/ports/powerpc/ | Debian CDs are not bootable on OldWorld PowerMacs |
| 15:05.55 | *** join/#debianppc apt (ibot@rikers.org) |
| 15:05.55 | *** topic/#debianppc is The Debian GNU/Linux PowerPC Port: http://www.debian.org/ports/powerpc/ | Debian CDs are not bootable on OldWorld PowerMacs |
| 15:33.47 | camm` | harth: thanks so much -- scoping it out .... |
| 15:38.16 | camm` | so function descriptors are dropped? |
| 15:38.31 | camm` | is ppc64 at debian-ports now dead? |
| 15:39.07 | camm` | if the two will coexist, how does one distinguish using the standard autoconf tools? |
| 15:40.26 | *** join/#debianppc thierry-ibm1 (tfauck@nat/ibm/x-kvqxzgnmyyhyqdhc) |
| 15:48.12 | *** join/#debianppc AbouEmre1 (~Thunderbi@41.87.3.26) |
| 16:35.51 | *** join/#debianppc pankid (~pan@73.191.253.36) |
| 17:21.00 | *** join/#debianppc markos__ (~markos@109-178-219-146.mobile.cosmote.net) |
| 17:25.03 | hartb | camm`: Sorry! I missed your messages. Function descriptors are dropped in the new ABI. Both big-endian and little-endian linux will be supported on IBM Power machines. I don't know about the state of the ppc64 debian port specifically. And I'm afraid I'm not sure about how to distinguish in autoconf |
| 17:33.09 | hartb | camm`: And I should say that big-endian linux on power still uses the same ABI it always has. No change there as far as I know. |
| 17:33.18 | hartb | lunch; biab |
| 17:45.54 | *** join/#debianppc pankid (~pan@173-162-51-174-michigan.hfc.comcastbusiness.net) |
| 18:40.39 | *** join/#debianppc sloopy (~sloopy@unaffiliated/sloopy) |
| 19:22.56 | *** join/#debianppc pankid (~pan@173-162-51-174-michigan.hfc.comcastbusiness.net) |
| 19:25.43 | camm` | hartb: thanks so much. I think I see how it is working. Perhaps a question or two a bit later. |
| 19:26.07 | hartb | camm`: Good! Sure. |
| 19:31.49 | camm` | is gcc solid at the moment? |
| 19:39.20 | hartb | camm`: Yeah, I think so. Big-endian has long been supported, and LE support and the ABIv2 changes were in place early enough to support a LE ABIv2 user-space built with gcc 4.8.3 (I think it was) for Ubuntu 14.04. And the 3.16 kernel will be built with LE ABIv2 support with gcc 4.9 (I think) for Ubuntu 14.10 (and presumably some corresponding, underlying debian stream) |
| 19:39.22 | *** join/#debianppc tdaitx_ (~tdaitx@177.18.43.82) |
| 19:45.58 | *** join/#debianppc pankid (~pan@73.191.253.36) |
| 19:46.30 | camm` | hartb: thanks! A bit about my project -- I maintain gcl, and dependencies maxima/acl2/axiom/hol88. gcl compiles code at runtime, loads into brk'ed memory, relocates, mprotects, executes, saves to disk via unexec for later reexecution. |
| 19:46.54 | camm` | This requires a linker veneer |
| 19:49.10 | camm` | At the moment, I've basically ported to ppc64el, but see a few anomalies. For one, compiling without optimization appears less stable. Two, I need to understand if the linker should attempt to rewrite toc references against the global .got, as we do on some machines. Right now we just append the local .toc. .TOC. is a global symbol and needs special treatment. |
| 19:49.26 | hartb | camm`: Ah; yeah--I can see that you'd need to get the ABI specifics right for that. |
| 19:50.17 | camm` | I'm also seeing floating point errors in the maxima test suite. |
| 19:51.08 | camm` | anyway, you don't happen to be an expert on this, do you? :-) |
| 19:51.13 | hartb | camm`: I think the ABI itself and gcc support should be stable now. Of course ppc64le is still fairly new, so there are probably still a few issues in the darker corners, but it's been shipping in a distro for months now. The global .TOC. symbol is new and special, as you've found. |
| 19:52.10 | hartb | camm`: I don't; I'm just working on another project where we also had to deal with the ABI changes, though we're not using it as intimately as you are. |
| 19:52.15 | camm` | I take it a veneer can clobber r9, and need not touch r2, as the functions do that internally. |
| 19:52.29 | camm` | ok |
| 19:52.54 | camm` | are the docs freely available somewhere? |
| 19:53.24 | hartb | But we can probably get you answers to specific questions from our toolchain folks. I think the formal ABI doc is still undergoing some IBM review before it can be made available externally. |
| 19:53.52 | camm` | thanks so much. |
| 19:54.25 | hartb | camm`: you're welcome! |
| 20:01.03 | *** join/#debianppc wschmidt (wschmidt@nat/ibm/x-qdojinpkmsugpjyc) |
| 20:06.11 | *** join/#debianppc leitao (breno@nat/ibm/x-wnkblirfgdgcqpkc) |
| 20:08.46 | hartb | camm`: The ABI spec is available at: https://www-03.ibm.com/technologyconnect/tgcm/TGCMFileServlet.wss/ABI64BitOpenPOWER_21July2014_pub.pdf?id=B81AEC1A37F5DAF185257C3E004E8845&linkid=1n0000&c_t=ph1cdfft8hudicsekb9w8e669r4l4gjg |
| 20:09.19 | hartb | Unfortunately that site requires registration, but it's free. We're looking to move the spec to a registration-less site, but that's not done yet. |
| 20:09.31 | hartb | We apologize for the inconvenience |
| 20:10.18 | wschmidt | it will eventually be part of the OpenPOWER website that isn't done yet, from what I understand. |
| 20:12.17 | camm` | thanks so much! |
| 20:14.16 | *** join/#debianppc davemq (~user@32.97.110.54) |
| 20:21.54 | hartb | camm`: You're welcome! |
| 20:32.38 | *** join/#debianppc chdrsto (~chdrsto@77-57-110-225.dclient.hispeed.ch) |
| 20:44.07 | *** join/#debianppc TalkSoup (~chdrsto@77-57-110-225.dclient.hispeed.ch) |
| 20:44.27 | TalkSoup | Hi guys |
| 20:54.10 | hartb | howdy! |
| 23:58.41 | *** join/#debianppc sloopy (~sloopy@c-69-255-204-204.hsd1.md.comcast.net) |
| 23:58.48 | *** join/#debianppc sloopy (~sloopy@unaffiliated/sloopy) |