| 1 |
On 11/15/2010 01:30 AM, Diego Elio Pettenò wrote: |
| 2 |
> Il giorno dom, 14/11/2010 alle 22.03 +0100, Thomas Kahle ha scritto: |
| 3 |
>> I have a package (sci-libs/mpir) whose configure supports building of |
| 4 |
>> fat binaries with both x86 and amd64 assembler in the binary. |
| 5 |
> |
| 6 |
> Oh the heck are they implemented? If they are FatELF, no they shouldn't |
| 7 |
> be used, ever, full stop. |
| 8 |
> |
| 9 |
> In most cases, this sounds fishy and almost a hack deemed to failure so |
| 10 |
> my default vote would be "do not expose this functionality to the user". |
| 11 |
|
| 12 |
It's not FatELF. It's pretty much what the Intel compiler does, but in |
| 13 |
this case it's done by hand; the binary includes several versions of the |
| 14 |
code. Which version is executed depends on the CPU detected at runtime. |
| 15 |
Has nothing to do with the FatELF brain damage :-) |
Archives