Hi,

 

The generated IR for vld3_dup and vld4_dup intrinsics was identical to that
for the vld2_dup intrinsic. 
The resulting code will only affect two vectors instead of 3 or 4. This is
because the correct vld/vldlane 

intrinsic used in the implementation of vld_dup is not selected.

 

The patch corrects this and adds a regression test.    

 

Please review this patch.

 

Thanks,

Silviu

Attachment: clang_vld_dup.diff
Description: Binary data

_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to