Sentinel floating-point conversion functions

Article
10/17/2023

Microsoft Specific

Conversion functions between floating-point types and integral types that use an Intel Architecture (IA) AVX-512 compatible sentinel strategy.

Syntax

int _cvt_ftoi_sent(float value);
long long _cvt_ftoll_sent(float value);
unsigned _cvt_ftoui_sent(float value);
unsigned long long _cvt_ftoull_sent(float value);
int _cvt_dtoi_sent(double value);
long long _cvt_dtoll_sent(double value);
unsigned _cvt_dtoui_sent(double value);
unsigned long long _cvt_dtoull_sent(double value);

Parameters

value
[in] A floating-point value to convert.

Return value

The integer-typed result of the conversion.

Requirements

Header: <intrin.h>

Architecture: x86, x64

Remarks

These intrinsics are floating-point to integral type conversion functions that use a sentinel strategy: They return the result value farthest from zero as a proxy sentinel value for NaN. Any invalid conversion returns this sentinel value. The specific sentinel value returned depends on the result type.

Result type	Sentinel	`<limits.h>` constant
`int`	-2147483648 (0x80000000)	`INT_MIN`
`unsigned int`	4294967295 (0xFFFFFFFF)	`UINT_MAX`
`long long`	-9223372036854775808 (0x8000000000000000)	`LLONG_MIN`
`unsigned long long`	18446744073709551615 (0xFFFFFFFFFFFFFFFF)	`ULLONG_MAX`

The sentinel conversion intrinsics are available starting in Visual Studio 2019 version 16.10.

END Microsoft Specific