Advertisement
Front, Kernel, Security, Distributions, Development. See your byline here on LWN.net.
| |||||||||||||
|
| |||||||||||||
User space atomic ops on ARMv5 and earlierUser space atomic ops on ARMv5 and earlierPosted Jan 13, 2009 16:44 UTC (Tue) by endecotp (guest, #36428)In reply to: User space atomic ops on ARMv5 and earlier by npitre Parent article: ARM SoC launched with Linux support (Linux Devices)
It would be great if gcc could make use of this when you use its __sync_* atomic builtins, or if glibc used it in its pthread_* implementations. I'm pretty sure that neither does at present, though I'd love to be corrected if they do!
(Log in to post comments)
User space atomic ops on ARMv5 and earlier Posted Jan 13, 2009 19:39 UTC (Tue) by npitre (subscriber, #5680) [Link] NPTL support for ARM in glibc certainly does. I developed the kernel part in collaboration with the person who did the glibc part. As to gcc, I don't see any specific ARM support for the __sync_* primitives, not even for ARMv6+ which has native instructions that could be used for that purpose. However this should be easy to implement following the PA model. Incidentally, the file gcc/config/pa/linux-atomic.c contains this note: /* Linux-specific atomic operations for PA Linux. Copyright (C) 2008 Free Software Foundation, Inc. Based on code contributed by CodeSourcery for ARM EABI Linux. Modifications for PA Linux by Helge Deller <deller@gmx.de> [...] So maybe that support does exist somewhere already?
User space atomic ops on ARMv5 and earlier Posted Jan 13, 2009 22:18 UTC (Tue) by endecotp (guest, #36428) [Link]
I've just done a quick test and it looks like gcc emits a function call when you invoke __sync_* for ARM. So presumably it would be possible to write a small library of things with the right names that calls the kernel helper code. Then near-optimal portable code would be possible.
Maybe someone has already done this....
|
|
||||||||||||