_mm_com_epi16

Visual Studio 2010

Visual Studio 2010 SP1 is required.

Microsoft Specific

Generates the XOP instruction vpcomw to return a bitmask generated by a word-by-word signed comparison of its first two sources using the condition given by the third source.

__m128i _mm_com_epi16 (
   __m128i src1,
   __m128i src2,
   int condition
);

[in] src1

Packed 128-bit array of eight 16-bit signed integers.

[in] src2

Packed 128-bit array of eight 16-bit signed integers.

[in] condition

A 32-bit integer condition to be used in the comparison. Legal values are 0 through 7.

A packed 128-bit array r of eight 16-bit unsigned integers, each of which is either 0x0000 or 0xFFFF. If cond is the logical operator implied by the value of condition, then

r[i] := src1[i] cond src2[i] ? 0xFFFF : 0x0000;

Intrinsic

Architecture

_mm_com_epi16

XOP

Header file <intrin.h>

This instruction compares each byte of src1 to the corresponding byte of src2, using the comparison operator described by condition. (Each pair of bytes uses the same comparison operator.) The legal values condition are 0 through 7. By including <intrin.h> you will be able to use the following names rather than numerical values for condition:

Integer Value

Symbolic Name

Meaning

0

_MM_PCOMCTRL_LT

r[i] = (src1[i] < src2[i]) ? 0xFFFF : 0x0000

1

_MM_PCOMCTRL_LE

r[i] = (src1[i] <= src2[i]) ? 0xFFFF : 0x0000

2

_MM_PCOMCTRL_GT

r[i] = (src1[i] > src2[i]) ? 0xFFFF : 0x0000

3

_MM_PCOMCTRL_GE

r[i] = (src1[i] >= src2[i]) ? 0xFFFF : 0x0000

4

_MM_PCOMCTRL_EQ

r[i] = (src1[i] == src2[i]) ? 0xFFFF : 0x0000

5

_MM_PCOMCTRL_NEQ

r[i] = (src1[i] != src2[i]) ? 0xFFFF : 0x0000

6

_MM_PCOMCTRL_FALSE

r[i] = 0x0000 (mask of all zeros)

7

_MM_PCOMCTRL_TRUE

r[i] = 0xFFFF (mask of all ones)

If you prefer, you may use the following macros, which will be defined when <intrin.h> is included:

Macro

Meaning

_mm_comlt_epi16(src1, src2)

_mm_com_epi16(src1, src2, 0)

_mm_comle_epi16(src1, src2)

_mm_com_epi16(src1, src2, 1)

_mm_comgt_epi16(src1, src2)

_mm_com_epi16(src1, src2, 2)

_mm_comge_epi16(src1, src2)

_mm_com_epi16(src1, src2, 3)

_mm_comeq_epi16(src1, src2)

_mm_com_epi16(src1, src2, 4)

_mm_comneq_epi16(src1, src2)

_mm_com_epi16(src1, src2, 5)

_mm_comfalse_epi16(src1, src2)

_mm_com_epi16(src1, src2, 6)

_mm_comtrue_epi16(src1, src2)

_mm_com_epi16(src1, src2, 7)

The vpcomb instruction is part of the XOP family of instructions. Before you use this intrinsic, you must ensure that the processor supports this instruction. To determine hardware support for this instruction, call the __cpuid intrinsic with InfoType = 0x80000001 and check bit 11 of CPUInfo[2] (ECX). This bit is 1 when the instruction is supported, and 0 otherwise.

#include <stdio.h>
#include <intrin.h>
main()
{
    __m128i a, b, d;
    int i;
    for (i = 0; i < 8; i++) {
        a.m128i_i16[i] = ((20011*i) % 65535) - 32768;
        b.m128i_i16[i] = ((32767*i) % 65535) - 32768;
    }
    printf("src1:   ");
    for (i = 0; i < 8; i++) printf_s(" %6d", a.m128i_i16[i]);
    printf_s("\nsrc2:   ");
    for (i = 0; i < 8; i++) printf_s(" %6d", b.m128i_i16[i]);
    printf_s("\n<  mask:");
    d = _mm_com_epi16(a, b, _MM_PCOMCTRL_LT);
    for (i = 0; i < 8; i++) printf_s("   %04x", d.m128i_u16[i]);
    printf_s("\n== mask:");
    d = _mm_com_epi16(a, b, _MM_PCOMCTRL_EQ);
    for (i = 0; i < 8; i++) printf_s("   %04x", d.m128i_u16[i]);
    printf("\n");
}
src1:    -32768 -12757   7254  27265 -18259   1752  21763 -23761
src2:    -32768     -1  32766     -2  32765     -3  32764     -4
<  mask:   0000   ffff   ffff   0000   ffff   0000   ffff   ffff
== mask:   ffff   0000   0000   0000   0000   0000   0000   0000

Community Additions

ADD
Show: