| jtaylor | so what is ppc64el? whats the difference to the other arches? | 10:58 |
|---|---|---|
| jtaylor | have I missed a mail on -devel about it? | 10:58 |
| jtaylor | so its little endian, why is it named el not le? oO | 11:02 |
| rbasak | Debian uses el for whatever reason. armel, mips64el. | 11:08 |
| rbasak | I don't know what it means or the history around it. | 11:08 |
| Laney | I thought it was an endianness joke, but eb exists so maybe not | 11:10 |
| === dpm_ is now known as dpm | ||
| === dholbach_ is now known as dholbach | ||
| xnox | jtaylor: rbasak: Laney: i always thought it was a clever pun: little endian - le - 6C 65, would be stored in memory as 65 6C (little endian) and thus it's "el" most significant letter last ;-) | 12:27 |
| Laney | yeah | 12:27 |
| rbasak | Well, that's debatable :) | 12:28 |
| xnox | rbasak: if one doesn't know endianess a self-declared "le" appears as "el" value ;-)))) or rather your BigEndian machine will see it as such ;-) | 12:29 |
| rbasak | It's only gets reversed in memory because you did the initial conversion in big endian :) | 12:29 |
| xnox | rbasak: memory addresses are sequential ;-) http://en.wikipedia.org/wiki/Endianness | 12:29 |
| rbasak | Yeah sure, but you read 6C 65 out of memory, and convert it to an integer, and then to hex, and then print it here. | 12:30 |
| rbasak | The endianness comes into play when you convert to an integer. | 12:30 |
| rbasak | Nothing gets swapped. | 12:30 |
| rbasak | It's ASCII->integer when you have the endianness question. So saying that le is el in little endian doesn't really work. | 12:31 |
| rbasak | (or 2-byte -> 16 bit integer, if you like) | 12:31 |
| jtaylor | wgrant: so ppc64el is long double = 16 byte but 80 bit precision? | 12:36 |
| jtaylor | I'm trying to understand your numpy patch | 12:37 |
| wgrant | jtaylor: It's the same as ppc64, except the bytes within each double are flipped. | 12:39 |
| jtaylor | are you planing on forwarding it upstream? | 12:42 |
| jtaylor | also are there porterboxes available numpys test segfaults with python-dbg | 12:43 |
| wgrant | jtaylor: Modern qemu-system-ppc64 can do LE | 12:45 |
| wgrant | jtaylor: We don't have a pregenerated image for it, though. | 12:46 |
| jtaylor | qemu sucks for floating point | 12:46 |
| jtaylor | it doesn'T seem to respect ieee754 | 12:46 |
| wgrant | Hm | 12:47 |
| wgrant | Anyway, we barely have buildds atm, no porter boxes yet. qemu *should* work fine. | 12:48 |
| wgrant | The primary archive isn't quite debootstrappable yet, but http://people.ubuntu.com/~ubuntu-archive/bootstrap/ppc64el is. | 12:48 |
| jtaylor | I'll give it a try | 12:50 |
| jtaylor | if you have access to a machine, it would be grat if you could just give me a backtrace | 12:50 |
| wgrant | Let me see what I can do. | 12:51 |
| wgrant | Thanks for looking | 12:51 |
| === j_f-f_ is now known as j_f-f | ||
| === freeflying is now known as freeflying_away | ||
| === Zic is now known as Guest7180 | ||
| === debfx_ is now known as debfx | ||
| === Guest7180 is now known as Zic | ||
| glenford | sd | 18:47 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!