_mm_com_epi32
Visual Studio 2010 SP1 is required.
Microsoft Specific
Generates the XOP instruction vpcomd to return a bitmask generated by a doubleword-by-doubleword signed comparison of its first two sources using the condition given by the third source.
__m128i _mm_com_epi32 (
__m128i src1,
__m128i src2,
int condition
);
Parameters
[in] src1
Packed 128-bit array of four 32-bit signed integers.[in] src2
Packed 128-bit array of four 32-bit signed integers.[in] condition
A 32-bit integer condition to be used in the comparison. Legal values are 0 through 7.
Return value
A packed 128-bit array r of four 32-bit unsigned integers, each of which is either 0x00000000 or 0xFFFFFFFF. If cond is the logical operator implied by the value of condition, then
r[i] := src1[i] cond src2[i] ? 0xFFFFFFFF : 0x00000000;
Requirements
Intrinsic |
Architecture |
|---|---|
_mm_com_epi32 |
XOP |
Header file <intrin.h>
Remarks
This instruction compares each doubleword of src1 to the corresponding doubleword of src2, using the comparison operator described by condition. (Each pair of doublewords 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]) ? 0xFFFFFFFF : 0x00000000 |
1 |
_MM_PCOMCTRL_LE |
r[i] = (src1[i] <= src2[i]) ? 0xFFFFFFFF : 0x00000000 |
2 |
_MM_PCOMCTRL_GT |
r[i] = (src1[i] > src2[i]) ? 0xFFFFFFFF : 0x00000000 |
3 |
_MM_PCOMCTRL_GE |
r[i] = (src1[i] >= src2[i]) ? 0xFFFFFFFF : 0x00000000 |
4 |
_MM_PCOMCTRL_EQ |
r[i] = (src1[i] == src2[i]) ? 0xFFFFFFFF : 0x00000000 |
5 |
_MM_PCOMCTRL_NEQ |
r[i] = (src1[i] != src2[i]) ? 0xFFFFFFFF : 0x00000000 |
6 |
_MM_PCOMCTRL_FALSE |
r[i] = 0x00000000 (mask of all zeros) |
7 |
_MM_PCOMCTRL_TRUE |
r[i] = 0xFFFFFFFF (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_epi32(src1, src2) |
_mm_com_epi32(src1, src2, 0) |
_mm_comle_epi32(src1, src2) |
_mm_com_epi32(src1, src2, 1) |
_mm_comgt_epi32(src1, src2) |
_mm_com_epi32(src1, src2, 2) |
_mm_comge_epi32(src1, src2) |
_mm_com_epi32(src1, src2, 3) |
_mm_comeq_epi32(src1, src2) |
_mm_com_epi32(src1, src2, 4) |
_mm_comneq_epi32(src1, src2) |
_mm_com_epi32(src1, src2, 5) |
_mm_comfalse_epi32(src1, src2) |
_mm_com_epi32(src1, src2, 6) |
_mm_comtrue_epi32(src1, src2) |
_mm_com_epi32(src1, src2, 7) |
The vpcomd 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.
Example
#include <stdio.h>
#include <intrin.h>
main()
{
__m128i a, b, d;
int i;
for (i = 0; i < 4; i++) {
a.m128i_i32[i] = 1000*i - 1500;
b.m128i_i32[i] = 2000*i - 2500;
}
printf("src1: ");
for (i = 0; i < 4; i++) printf_s(" %8d", a.m128i_i32[i]);
printf_s("\nsrc2: ");
for (i = 0; i < 4; i++) printf_s(" %8d", b.m128i_i32[i]);
printf_s("\n!= mask: ");
d = _mm_com_epi32(a, b, _MM_PCOMCTRL_NEQ);
for (i = 0; i < 4; i++) printf_s(" %08x", d.m128i_u32[i]);
printf_s("\ntrue mask: ");
d = _mm_com_epi32(a, b, _MM_PCOMCTRL_TRUE);
for (i = 0; i < 4; i++) printf_s(" %08x", d.m128i_u32[i]);
printf("\n");
}
src1: -1500 -500 500 1500 src2: -2500 -500 1500 3500 != mask: ffffffff 00000000 ffffffff ffffffff true mask: ffffffff ffffffff ffffffff ffffffff