Module swiss_army_knife::simd_compatibility::x86_64::sse2

SSE2 support, including functions one might assume to be present as intrinsics provided by Intel but aren’t.

Functions

_mm_blend_epi64^⚠

This is efficient and compiles down to the same assembler as _mm_blend_epi32() but with a different immediate value used for the control word.

_mm_extract_epi8_constant^⚠

This is inefficient and should be avoided inside loops.

_mm_extract_epi8_variable^⚠

This is inefficient and should be avoided inside loops.

_mm_lane_crossing_left_shift_upto_64_bits^⚠

left_shift_number_of_bits must be between 0 and 64 inclusive.

_mm_lane_crossing_right_shift_upto_64_bits^⚠

right_shift_number_of_bits must be between 0 and 64 inclusive.

_mm_popcnt_epi8^⚠

Based on https://stackoverflow.com/questions/17354971/fast-counting-the-number-of-set-bits-in-m128i-register/17355341#17355341.

_mm_reduce_add_epu8^⚠

This is inefficient and should be avoided inside loops.

_mm_reduce_add_epu32^⚠

This is inefficient and should be avoided inside loops.

_mm_shuffle_epi64_constant^⚠

Shuffles.