VCOMPRESSPS - COMPRESS Packed Single

VCOMPRESSPS xmm1/m128{k1}{z}, xmm2    (V5+VL
__m128 _mm_mask_compress_sd(__m128 s, __mmask8 k, __m128 a)
__m128 _mm_maskz_compress_sd(__mmask8 k, __m128 a)
void _mm_mask_compressstoreu_sd(void* d, __mmask8 k, __m128 a)

If each bit of (2) is set, the corresponding element of (1) is copied to (3), packed tightly from the lowest position.

If (3) is XMM register and {z} is specified (_maskz_ intrinsic used), upper elements of (3) are zero cleared.
If (3) is XMM register and {z} is not specified (_mask_ intrinsic used), upper elements of (3) are left unchanged (copied from s).
If (3) is memory (_compressstoreu_ intrinsic used) , upper elements of (3) are untouched.
VCOMPRESSPS ymm1/m256{k1}{z}, ymm2    (V5+VL
__m256 _mm256_mask_compress_sd(__m256 s, __mmask8 k, __m256 a)
__m256 _mm256_maskz_compress_sd(__mmask8 k, __m256 a)
void _mm256_mask_compressstoreu_sd(void* d, __mmask8 k, __m256 a)

If each bit of (2) is set, the corresponding element of (1) is copied to (3), packed tightly from the lowest position.

If (3) is YMM register and {z} is specified (_maskz_ intrinsic used), upper elements of (3) are zero cleared.
If (3) is YMM register and {z} is not specified (_mask_ intrinsic used), upper elements of (3) are left unchanged (copied from s).
If (3) is memory (_compressstoreu_ intrinsic used) , upper elements of (3) are untouched.
VCOMPRESSPS zmm1/m512{k1}{z}, zmm2    (V5
__m512 _mm512_mask_compress_sd(__m512 s, __mmask16 k, __m512 a)
__m512 _mm512_maskz_compress_sd(__mmask16 k, __m512 a)
void _mm512_mask_compressstoreu_sd(void* d, __mmask16 k, __m512 a)

If each bit of (2) is set, the corresponding element of (1) is copied to (3), packed tightly from the lowest position.

If (3) is ZMM register and {z} is specified (_maskz_ intrinsic used), upper elements of (3) are zero cleared.
If (3) is ZMM register and {z} is not specified (_mask_ intrinsic used), upper elements of (3) are left unchanged (copied from s).
If (3) is memory (_compressstoreu_ intrinsic used) , upper elements of (3) are untouched.

x86/x64 SIMD Instruction List  Feedback