Intel® C++ Compiler Classic Developer Guide and Reference

ID 767249
Date 3/31/2023
Public

A newer version of this document is available. Customers should click here to go to the newest version.

Document Table of Contents

Intrinsics for Miscellaneous FP Operations

The prototypes for Intel® Advanced Vector Extensions 512 (Intel® AVX-512) intrinsics are located in the zmmintrin.h header file.

To use these intrinsics, include the immintrin.h file as follows:

#include <immintrin.h>


Intrinsic Name

Operation

Corresponding
Intel® AVX-512 Instruction

_mm512_fixupimm_pd, _mm512_mask_fixupimm_pd, _mm512_maskz_fixupimm_pd

_mm512_fixupimm_round_pd, _mm512_mask_fixupimm_round_pd, _mm512_maskz_fixupimm_round_pd

Fixes up packed float64 elements.

VFIXUPIMMPD

_mm512_fixupimm_ps, _mm512_mask_fixupimm_ps, _mm512_maskz_fixupimm_ps

_mm512_fixupimm_round_ps, _mm512_mask_fixupimm_round_ps, _mm512_maskz_fixupimm_round_ps

Fixes up packed float32 elements.

VFIXUPIMMPS

_mm_fixupimm_round_sd, _mm_mask_fixupimm_round_sd, _mm_maskz_fixupimm_round_sd

Fixes up scalar float64 elements.

VFIXUPIMMSD

_mm_fixupimm_round_ss, _mm_mask_fixupimm_round_ss, _mm_maskz_fixupimm_round_ss

Fixes up scalar float32 elements.

VFIXUPIMMSS

_mm_getexp_round_pd, _mm_mask_getexp_round_pd, _mm_maskz_getexp_round_pd

_mm_maskz_getexp_round_pd

Converts exponent of each packed float64 element to a rounded float64 number representing the integer exponent.

VGETEXPPD

_mm_getexp_round_ps, _mm_mask_getexp_round_ps, _mm_maskz_getexp_round_ps

_mm_maskz_getexp_round_ps

Converts exponent of each packed float32 element to a rounded float32 number representing the integer exponent.

VGETEXPPS

_mm_getexp_round_sd, _mm_mask_getexp_round_sd, _mm_maskz_getexp_round_sd

_mm_getexp_round_sd, _mm_mask_getexp_round_sd, _mm_maskz_getexp_round_sd

Converts exponent of each scalar float64 element to a rounded scalar float64 number representing the integer exponent.

VGETEXPSD

_mm_getexp_round_ss, _mm_mask_getexp_round_ss, _mm_maskz_getexp_round_ss

Converts exponent of each scalar float32 element to a rounded scalar float32 number representing the integer exponent.

VGETEXPSS


variable definition
k

writemask used as a selector

a

first source vector element

b

second source vector element

c

third source vector element

src

source element to use based on writemask result

round

Rounding control values; these can be one of the following (along with the sae suppress all exceptions flag):

  • _MM_FROUND_TO_NEAREST_INT - rounds to nearest even
  • _MM_FROUND_TO_NEG_INF - rounds to negative infinity
  • _MM_FROUND_TO_POS_INF - rounds to positive infinity
  • _MM_FROUND_TO_ZERO - rounds to zero
  • _MM_FROUND_CUR_DIRECTION - rounds using default from MXCSR register
src

source element

imm

reporting flag

src

source element


_mm512_fixupimm_pd

extern __m512d __cdecl _mm512_fixupimm_pd(__m512d a, __m512d b, __m512i c, int imm);

Fixes up packed float64 elements in a and b using packed 64-bit integers in c, and stores the result.

imm is used to set the required flags reporting.


_mm512_mask_fixupimm_pd

extern __m512d __cdecl _mm512_mask_fixupimm_pd(__m512d a, __mmask8 k, __m512d b, __m512i c, int imm);

Fixes up packed float64 elements in a and b using packed 64-bit integers in c, and stores the result using writemask k (elements are copied from a when the corresponding mask bit is not set).

imm is used to set the required flags reporting.


_mm512_maskz_fixupimm_pd

extern __m512d __cdecl _mm512_maskz_fixupimm_pd(__mmask8 k, __m512d a, __m512d b, __m512i c, int imm);

Fixes up packed float64 elements in a and b using packed 64-bit integers in c, and stores the result using zeromask k (elements are zeroed out when the corresponding mask bit is not set).

imm is used to set the required flags reporting.



_mm512_fixupimm_round_pd

extern __m512d __cdecl _mm512_fixupimm_round_pd(__m512d a, __m512d b, __m512i c, int imm, int round);

Fixes up packed float64 elements in a and b using packed 64-bit integers in c, and stores the result.

imm is used to set the required flags reporting.


_mm512_mask_fixupimm_round_pd

extern __m512d __cdecl _mm512_mask_fixupimm_round_pd(__m512d a, __mmask8 k, __m512d b, __m512i c, int imm, int round);

Fixes up packed float64 elements in a and b using packed 64-bit integers in c, and stores the result using writemask k (elements are copied from a when the corresponding mask bit is not set).

imm is used to set the required flags reporting.


_mm512_maskz_fixupimm_round_pd

extern __m512d __cdecl _mm512_maskz_fixupimm_round_pd(__mmask8 k, __m512d a, __m512d b, __m512i c, int imm, int round);

Fixes up packed float64 elements in a and b using packed 64-bit integers in c, and stores the result using zeromask k (elements are zeroed out when the corresponding mask bit is not set).

imm is used to set the required flags reporting.



_mm512_fixupimm_ps

extern __m512 __cdecl _mm512_fixupimm_ps(__m512 a, __m512 b, __m512i c, int imm);

Fixes up packed float32 elements in a and b using packed 32-bit integers in c, and stores the result. imm is used to set the required flags reporting.


_mm512_mask_fixupimm_ps

extern __m512 __cdecl _mm512_mask_fixupimm_ps(__m512 a, __mmask16 k, __m512 b, __m512i c, int imm);

Fixes up packed float32 elements in a and b using packed 32-bit integers in c, and stores the result using writemask k (elements are copied from a when the corresponding mask bit is not set).

imm is used to set the required flags reporting.


_mm512_maskz_fixupimm_ps

extern __m512 __cdecl _mm512_maskz_fixupimm_ps(__mmask16 k, __m512 a, __m512 b, __m512i c, int imm);

Fixes up packed float32 elements in a and b using packed 32-bit integers in c, and stores the result using zeromask k (elements are zeroed out when the corresponding mask bit is not set).

imm is used to set the required flags reporting.



_mm512_fixupimm_round_ps

extern __m512 __cdecl _mm512_fixupimm_round_ps(__m512 a, __m512 b, __m512i c, int imm, int round);

Fixes up packed float32 elements in a and b using packed 32-bit integers in c, and stores the result. imm is used to set the required flags reporting.


_mm512_mask_fixupimm_round_ps

extern __m512 __cdecl _mm512_mask_fixupimm_round_ps(__m512 a, __mmask16 k, __m512 b, __m512i, int imm, int round);

Fixes up packed float32 elements in a and b using packed 32-bit integers in c, and stores the result using writemask k (elements are copied from a when the corresponding mask bit is not set).

imm is used to set the required flags reporting.


_mm512_maskz_fixupimm_round_ps

extern __m512 __cdecl _mm512_maskz_fixupimm_round_ps(__mmask16 k, __m512 a, __m512 b, __m512i, int imm, int round);

Fixes up packed float32 elements in a and b using packed 32-bit integers in c, and stores the result using zeromask k (elements are zeroed out when the corresponding mask bit is not set).

imm is used to set the required flags reporting.



_mm_fixupimm_sd

extern __m128d __cdecl _mm_fixupimm_sd(__m128d a, __m128d b, __m128i c, int imm);

Fixes up lower float64 elements in a and b using lower 64-bit integer in c, stores the result in lower destination element, and copies upper element from a to upper destination element.

imm is used to set the required flags reporting.


_mm_mask_fixupimm_sd

extern __m128d __cdecl _mm_mask_fixupimm_sd(__m128d a, __mmask8 k, __m128d b, __m128i c, int imm);

Fixes up lower float64 elements in a and b using lower 64-bit integer in c, stores the result in lower destination element using writemask k (the element is copied from a when mask bit 0 is not set), and copies upper element from a to upper destination element.

imm is used to set the required flags reporting.


_mm_maskz_fixupimm_sd

extern __m128d __cdecl _mm_maskz_fixupimm_sd(__mmask8 k, __m128d a, __m128d b, __m128i c, int imm);

Fixes up lower float64 elements in a and b using lower 64-bit integer in c, stores the result in lower destination element using zeromask k (the element is zeroed out when mask bit 0 is not set), and copies upper element from a to upper destination element.

imm is used to set the required flags reporting.



_mm_fixupimm_round_sd

extern __m128d __cdecl _mm_fixupimm_round_sd(__m128d a, __m128d b, __m128i c, int imm, int round);

Fixes up lower float64 elements in a and b using lower 64-bit integer in c, stores the result in lower destination element, and copies upper element from a to upper destination element.

imm is used to set the required flags reporting.


_mm_mask_fixupimm_round_sd

extern __m128d __cdecl _mm_mask_fixupimm_round_sd(__m128d a, __mmask8 k, __m128d b, __m128i c, int imm, int round);

Fixes up lower float64 elements in a and b using lower 64-bit integer in c, stores the result in lower destination element using writemask k (the element is copied from a when mask bit 0 is not set), and copies upper element from a to upper destination element.

imm is used to set the required flags reporting.



_mm_maskz_fixupimm_round_sd

extern __m128d __cdecl _mm_maskz_fixupimm_round_sd(__mmask8 k, __m128d a, __m128d b, __m128i c, int imm, int round);

Fixes up lower float64 elements in a and b using lower 64-bit integer in c, stores the result in lower destination element using zeromask k (the element is zeroed out when mask bit 0 is not set), and copies upper element from a to upper destination element.

imm is used to set the required flags reporting.



_mm_fixupimm_round_ss

extern __m128 __cdecl _mm_fixupimm_round_ss(__m128 a, __m128 b, __m128i c, int imm, int round);

Fixes up lower float32 elements in a and b using lower 32-bit integer in c, stores the result in lower destination element, and copies upper three packed elements from a to upper destination elements.

imm is used to set the required flags reporting.



_mm_mask_fixupimm_round_ss

extern __m128 __cdecl _mm_mask_fixupimm_round_ss(__m128 a, __mmask8 k, __m128 b, __m128i c, int imm, int round);

Fixes up lower float32 elements in a and b using lower 32-bit integer in c, stores the result in lower destination element using writemask k (the element is copied from a when mask bit 0 is not set), and copies upper three packed elements from a to upper destination elements.

imm is used to set the required flags reporting.



_mm_maskz_fixupimm_round_ss

extern __m128 __cdecl _mm_maskz_fixupimm_round_ss(__mmask8 k, __m128 a, __m128 b, __m128i c, int imm, int round);

Fixes up lower float32 elements in a and b using lower 32-bit integer in c, stores the result in lower destination element using zeromask k (the element is zeroed out when mask bit 0 is not set), and copies upper three packed elements from a to upper destination elements.

imm is used to set the required flags reporting.



_mm_fixupimm_ss

extern __m128 __cdecl _mm_fixupimm_ss(__m128 a, __m128 b, __m128i c, int imm);

Fixes up lower float32 elements in a and b using lower 32-bit integer in c, stores the result in lower destination element, and copies upper three packed elements from a to upper destination elements.

imm is used to set the required flags reporting.



_mm_mask_fixupimm_ss

extern __m128 __cdecl _mm_mask_fixupimm_ss(__m128 a, __mmask8 k, __m128 b, __m128i c, int imm);

Fixes up lower float32 elements in a and b using lower 32-bit integer in c, stores the result in lower destination element using writemask k (the element is copied from a when mask bit 0 is not set), and copies upper three packed elements from a to upper destination elements.

imm is used to set the required flags reporting.



_mm_maskz_fixupimm_ss

extern __m128 __cdecl _mm_maskz_fixupimm_ss(__mmask8 k, __m128 a, __m128 b, __m128i c, int imm);

Fixes up lower float32 elements in a and b using lower 32-bit integer in c, stores the result in lower destination element using zeromask k (the element is zeroed out when mask bit 0 is not set), and copies upper three packed elements from a to upper destination elements.

imm is used to set the required flags reporting.



_mm512_getexp_pd

extern __m512d __cdecl _mm512_getexp_pd(__m512d a);

Converts the exponent of each packed float64 element in a to a float64 number representing the integer exponent, and stores the result. This intrinsic essentially calculates floor(log2(x)) for each element.



_mm512_mask_getexp_pd

extern __m512d __cdecl _mm512_mask_getexp_pd(__m512d src, __mmask8 k, __m512d a);

Converts the exponent of each packed float64 element in a to a float64 number representing the integer exponent, and stores the result using writemask k (elements are copied from src when the corresponding mask bit is not set). This intrinsic essentially calculates floor(log2(x)) for each element.



_mm512_maskz_getexp_pd

extern __m512d __cdecl _mm512_maskz_getexp_pd(__mmask8 k, __m512d a);

Converts the exponent of each packed float64 element in a to a float64 number representing the integer exponent, and stores the result using writemask k (elements are copied from src when the corresponding mask bit is not set). This intrinsic essentially calculates floor(log2(x)) for each element.



_mm512_getexp_round_pd

extern __m512d __cdecl _mm512_getexp_round_pd(__m512d a, int round);

Converts the exponent of each packed float64 element in a to a float64 number representing the integer exponent, and stores the result. This intrinsic essentially calculates floor(log2(x)) for each element.



_mm512_mask_getexp_round_pd

extern __m512d __cdecl _mm512_mask_getexp_round_pd(__m512d src, __mmask8 a, __m512d, int round);

Converts the exponent of each packed float64 element in a to a float64 number representing the integer exponent, and stores the result using writemask k (elements are copied from src when the corresponding mask bit is not set). This intrinsic essentially calculates floor(log2(x)) for each element.



_mm512_maskz_getexp_round_pd

extern __m512d __cdecl _mm512_maskz_getexp_round_pd(__mmask8 k, __m512d a, int round);

Converts the exponent of each packed float64 element in a to a float64 number representing the integer exponent, and stores the result using zeromask k (elements are zeroed out when the corresponding mask bit is not set). This intrinsic essentially calculates floor(log2(x)) for each element.



_mm512_getexp_ps

extern __m512 __cdecl _mm512_getexp_ps(__m512 a);

Converts the exponent of each packed float32 element in a to a float32 number representing the integer exponent, and stores the result. This intrinsic essentially calculates floor(log2(x)) for each element.



_mm512_mask_getexp_ps

extern __m512 __cdecl _mm512_mask_getexp_ps(__m512 src, __mmask16 k, __m512 a);

Converts the exponent of each packed float32 element in a to a float32 number representing the integer exponent, and stores the result using writemask k (elements are copied from src when the corresponding mask bit is not set). This intrinsic essentially calculates floor(log2(x)) for each element.



_mm512_maskz_getexp_ps

extern __m512d __cdecl _mm512_maskz_getexp_ps(__mmask16 k, __m512 a);

Converts the exponent of each packed float32 element in a to a float32 number representing the integer exponent, and stores the result using writemask k (elements are copied from src when the corresponding mask bit is not set). This intrinsic essentially calculates floor(log2(x)) for each element.



_mm512_getexp_round_ps

extern __m512 __cdecl _mm512_getexp_round_ps(__m512 a, int round);

Converts the exponent of each packed float32 element in a to a float32 number representing the integer exponent, and stores the result. This intrinsic essentially calculates floor(log2(x)) for each element.



_mm512_mask_getexp_round_ps

extern __m512 __cdecl _mm512_mask_getexp_round_ps(__m512 src, __mmask16 k, __m512 a, int round);

Converts the exponent of each packed float32 element in a to a float32 number representing the integer exponent, and stores the result using writemask k (elements are copied from src when the corresponding mask bit is not set). This intrinsic essentially calculates floor(log2(x)) for each element.



_mm512_maskz_getexp_round_ps

extern __m512 __cdecl _mm512_maskz_getexp_round_ps(__mmask16 k, __m512 a, int round);

Converts the exponent of each packed float32 element in a to a float32 number representing the integer exponent, and stores the result using zeromask k (elements are zeroed out when the corresponding mask bit is not set). This intrinsic essentially calculates floor(log2(x)) for each element.



_mm_getexp_round_sd

extern __m128d __cdecl _mm_getexp_round_sd(__m128d a, __m128d b, int round);

Converts lower exponent of float64 element in b to a float64 number representing the integer exponent, stores the result in lower destination element, and copies upper element from a to upper destination element. This intrinsic essentially calculates floor(log2(x)) for lower element.



_mm_mask_getexp_round_sd

extern __m128d __cdecl _mm_mask_getexp_round_sd(__m128d src, __mmask8 k, __m128d a, __m128d b, int round);

Converts lower exponent of float64 element in b to a float64 number representing the integer exponent, stores the result in lower destination element, and copies upper element from a to upper destination element. This intrinsic essentially calculates floor(log2(x)) for lower element.



_mm_maskz_getexp_round_sd

extern __m128d __cdecl _mm_maskz_getexp_round_sd(__mmask8 k, __m128d a, __m128d b, int round);

Converts lower exponent of float64 element in b to a float64 number representing the integer exponent, stores the result in lower destination element using writemask k (the element is copied from src when mask bit 0 is not set), and copies upper element from a to upper destination element. This intrinsic essentially calculates floor(log2(x)) for lower element.



_mm_getexp_sd

extern __m128d __cdecl _mm_getexp_sd(__m128d a, __m128d b);

Converts lower exponent of float64 element in b to a float64 number representing the integer exponent, stores the result in lower destination element, and copies upper element from a to upper destination element. This intrinsic essentially calculates floor(log2(x)) for lower element.



_mm_mask_getexp_sd

extern __m128d __cdecl _mm_mask_getexp_sd(__m128d src, __mmask8 k, __m128d a, __m128d b);

Converts lower exponent of float64 element in b to a float64 number representing the integer exponent, stores the result in lower destination element using writemask k (the element is copied from src when mask bit 0 is not set), and copies upper element from a to upper destination element. This intrinsic essentially calculates floor(log2(x)) for lower element.



_mm_maskz_getexp_sd

extern __m128d __cdecl _mm_maskz_getexp_sd(__mmask8 k, __m128d a, __m128d b);

Converts lower exponent of float64 element in b to a float64 number representing the integer exponent, stores the result in lower destination element using zeromask k (the element is zeroed out when mask bit 0 is not set), and copies upper element from a to upper destination element. This intrinsic essentially calculates floor(log2(x)) for lower element.



_mm_getexp_round_ss

extern __m128 __cdecl _mm_getexp_round_ss(__m128 a, __m128 b, int round);	

Converts lower exponent of float32 element in b to a float32 number representing the integer exponent, stores the result in lower destination element, and copies upper three packed elements from a to upper destination elements. This intrinsic essentially calculates floor(log2(x)) for lower element.



_mm_mask_getexp_round_ss

extern __m128 __cdecl _mm_mask_getexp_round_ss(__m128 src, __mmask8 k, __m128 a, __m128 b, int round);

Converts lower exponent of float32 element in b to a float32 number representing the integer exponent, stores the result in lower destination element using writemask k (the element is copied from src when mask bit 0 is not set), and copies upper three packed elements from a to upper elements. This intrinsic essentially calculates floor(log2(x)) for lower element.



_mm_maskz_getexp_round_ss

extern __m128 __cdecl _mm_maskz_getexp_round_ss(__mmask8 k, __m128 a, __m128 b, int round);	

Converts lower exponent of float32 element in b to a float32 number representing the integer exponent, stores the result in lower destination element using writemask k (the element is copied from src when mask bit 0 is not set), and copies upper three packed elements from a to upper elements. This intrinsic essentially calculates floor(log2(x)) for lower element.



_mm_getexp_ss

extern __m128 __cdecl _mm_getexp_ss(__m128 a, __m128 b);

Converts lower exponent of float32 element in b to a float32 number representing the integer exponent, stores the result in lower destination element, and copies upper three packed elements from a to upper destination elements. This intrinsic essentially calculates floor(log2(x)) for lower element.



_mm_mask_getexp_ss

extern __m128 __cdecl _mm_mask_getexp_ss(__m128 src, __mmask8 k, __m128 a, __m128b);

Converts lower exponent of float32 element in b to a float32 number representing the integer exponent, stores the result in lower destination element using writemask k (the element is copied from src when mask bit 0 is not set), and copies upper three packed elements from a to upper destination elements. This intrinsic essentially calculates floor(log2(x)) for lower element.


_mm_maskz_getexp_ss

extern __m128 __cdecl _mm_maskz_getexp_ss(__mmask8 k, __m128 a, __m128 b);

Converts lower exponent of float32 element in b to a float32 number representing the integer exponent, stores the result in lower destination element using zeromask k (the element is zeroed out when mask bit 0 is not set), and copies upper three packed elements from a to upper destination elements. This intrinsic essentially calculates floor(log2(x)) for lower element.