[PATCH] Refactored the implementation of AArch64 NEON instruction ZIP, UZP and TRN and fix a bug when mixed use of vget_high_u8() and vuzp_u8().
kevinqindev at gmail.com
Mon Nov 25 02:19:28 PST 2013
This problem is exposed by a test case mixed use of vget_high_u8() and vuzp_u8(). To fix this problem, I refactored the implementation of AArch64 NEON instruction ZIP, UZP and TRN, to get them recognized when lowering shuffle_vector and build_vector, instead of matching a long sequence of insert_vector_elt.
Please reivew, thanks.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 24936 bytes
Desc: not available
More information about the llvm-commits