mirror of
https://github.com/VectorCamp/vectorscan.git
synced 2025-06-28 16:41:01 +03:00
Fix x86 case, use alignr instead
This commit is contained in:
parent
c95e66144a
commit
e0e972d498
@ -72,8 +72,8 @@ SuperVector<S> blockDoubleMask(SuperVector<S> mask1_lo, SuperVector<S> mask1_hi,
|
|||||||
// offset c1 so it aligns with c2. The hole created by the offset is filled
|
// offset c1 so it aligns with c2. The hole created by the offset is filled
|
||||||
// with the last elements of the previous c1 so no info is lost.
|
// with the last elements of the previous c1 so no info is lost.
|
||||||
// If bits with value 0 lines up, it indicate a match.
|
// If bits with value 0 lines up, it indicate a match.
|
||||||
c2.template vshr_imm<1>().print8("c2.vshr_128(1)");
|
SuperVector<S> offset_c1 = inout_c1->alignr(new_c1, S-1);
|
||||||
SuperVector<S> c = new_c1 | (c2.template vshr_imm<1>());
|
SuperVector<S> c = offset_c1 | c2;
|
||||||
c.print8("c");
|
c.print8("c");
|
||||||
|
|
||||||
*inout_c1 = new_c1;
|
*inout_c1 = new_c1;
|
||||||
|
Loading…
x
Reference in New Issue
Block a user