Closed berkus closed 3 years ago
@bors-servo r+ Looks reasonable; thanks!
:pushpin: Commit bd6d015 has been approved by jdm
:hourglass: Testing commit bd6d015e6ff7ccc68fa50662e5b714aa4d5e9d45 with merge 20adbeed149713ae4354cede3a30ca0f622c9943...
:sunny: Test successful - checks-travis Approved by: jdm Pushing 20adbeed149713ae4354cede3a30ca0f622c9943 to master...
According to packed_simd docs, the syntax for simd_shuffle4() is as follows:
"The indices must be in range [0, M N) where M is the number of input vectors (1 or 2)and N is the number of lanes of the input vectors. The indices i in range [0, N) refer to the i-th element of vec0, while the indices in range [N, 2N) refer to the i - N-th element of vec1."
I did not find implementation or documentation for simd_shuffle4() but I believe packed_simd implements exactly the same interface.
Plus, implementing this change has fixed font-kit glyph output on an Apple M1 mac (64-bit arm).
Closes #450