Intel® C++ Compiler Classic Developer Guide and Reference

ID 767249
Date 12/16/2022
Public

A newer version of this document is available.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

Document Table of Contents

<span class='option'>_mm256_round_ps</span>

Rounds off single-precision floating point values to nearest upper/lower integer depending on rounding mode. The corresponding Intel® AVX instruction isVROUNDPS.

Syntax

extern __m256 _mm256_round_ps(__m256 a, int iRoundMode );

Arguments

a

float32 vector

iRoundMode

A hexadecimal value dependent on rounding mode:

  • For rounding off to upper integer, the value is 0x0A
  • For rounding off to lower integer, the value is 0x09
Description

Rounds off the elements of a float32 vector a to the nearest upper/lower integer value. Two shortcuts, in the form of #defines, are used to achieve these two separate operations:

#define _mm256_ceil_ps(a)   _mm256_round_ps((a), 0x0A)
#define _mm256_floor_ps(a)  _mm256_round_ps((a), 0x09)

These #defines tells the preprocessor to replace all instances of _mm256_ceil_ps(a) with _mm256_round_ps((a), 0x0A) and all instances of _mm256_floor_ps(a) with _mm256_round_ps((a), 0x09).

For example, if you write the following:

__m256 a, b;
a = _mm256_ceil_ps(b);

the preprocessor will modify it to:

a = _mm256_round_ps(b, 0x0a);

The Precision Floating Point Exception is signaled according to the (immediate operand) iRoundMode value.

Returns

Result of the rounding off operation as a vector with single-precision floating point values.