== Progress == * Neon vext support for builtin_shuffle: * Committed vext patch upstream, as well as a small cleanup patch. * Merged vext support into gcc-linaro/4.7 branch. * Posted upstream a follow-up patch to make vext tests support big-endian. Filed PR 54517: wrong code generation in big-endian with inline in these tests. * Updated patch to fix 3 testcases in big-endian after upstream comments. * Implement builtin_bswap16: * Posted upstream a patch to implement bswap16. Discussion on-going, to have less duplication between arm and thumb patterns. * Investigating how to make GCC catch the (x<<8)|(x>>8) construct (where x is unsigned short) and map it to rev16, like builtin_bswap16.
== Next == * Continue with bswap16 support.
On 8 September 2012 02:17, Christophe Lyon christophe.lyon@linaro.org wrote:
== Progress ==
- Neon vext support for builtin_shuffle:
- Committed vext patch upstream, as well as a small cleanup patch.
- Merged vext support into gcc-linaro/4.7 branch.
I saw that, thanks. To finish up could you go to the merge request: https://code.launchpad.net/~christophe-lyon/gcc-linaro/gcc-4.7-vec-permute-v...
and mark it as 'merged', then go to the branch: https://code.launchpad.net/~christophe-lyon/gcc-linaro/gcc-4.7-vec-permute-v...
and mark the status as 'Merged'.
Launchpad used to do this automatically, but Linaro GCC is so big that we broke it!
-- Michael
linaro-toolchain@lists.linaro.org
-
Christophe Lyon -
Michael Hope