|
| 1 | +What's new in v2.1: |
| 2 | + * Various bug fixes |
| 3 | + * Documentation has been significantly improved. The public API is now almost |
| 4 | + fully documented. |
| 5 | + * Added support for MIPS MSA instruction set. |
| 6 | + * Added support for PowerPC VSX v2.06 and v2.07 instruction sets. |
| 7 | + * Added support for x86 AVX512BW, AVX512DQ and AVX512VL instruction sets. |
| 8 | + * Added support for 64-bit little-endian PowerPC. |
| 9 | + * Added support for arbitrary width vectors in extract() and insert() |
| 10 | + * Added support for arbitrary source vectors to to_int8(), to_uint8(), |
| 11 | + to_int16(), to_uint16(), to_int32(), to_uint32(), to_int64(), to_uint64(), |
| 12 | + to_float32(), to_float64(). |
| 13 | + * Added support for per-element integer shifts to shift_r() and shift_l(). |
| 14 | + Fallback paths are provided for SSE2-AVX instruction sets that lack |
| 15 | + hardware per-element integer shift support. |
| 16 | + * Make shuffle_bytes16(), shuffle_zbytes16(), permute_bytes16() and |
| 17 | + permute_zbytes() more generic. |
| 18 | + * New functions: popcnt, reduce_popcnt, for_each, to_mask() |
| 19 | + * Xcode is now supported. |
| 20 | + * The library has been refactored in such a way that older compilers are able |
| 21 | + to optimize vector emulation code paths much better than before. |
| 22 | + * Deprecation: implicit conversion operators to native vector types has been |
| 23 | + deprecated and a replacement method has been provided instead. The implicit |
| 24 | + conversion operators may lead to wrong code being accepted without a |
| 25 | + compile error on Clang. |
| 26 | + |
1 | 27 | What's new in v2.0: |
2 | 28 |
|
3 | 29 | * Various bug fixes. |
| 30 | + * Intel compiler is now supported on Windows. Newer versions of other compilers |
| 31 | + are now supported. |
4 | 32 |
|
5 | 33 | What's new in v2.0~rc2: |
6 | 34 |
|
|
0 commit comments