main.cpp

#include "../../commonfiles/ymmval.h"
#include <limits>

using namespace std;

extern "C" void AvxPfpCompare(const YmmVal* a, const YmmVal* b, YmmVal c[8]);

int main(int argc, char* argv[])
{
    char buff[256];
    __attribute__ ((aligned(32))) YmmVal a;
	__attribute__ ((aligned(32))) YmmVal b;
	__attribute__ ((aligned(32))) YmmVal c[8];

    const char* instr_names[8] =
    {
        "vcmpeqpd", "vcmpneqpd", "vcmpltpd", "vcmplepd",
        "vcmpgtpd", "vcmpgepd", "vcmpordpd", "vcmpunordpd"
    };

    a.r64[0] = 42.125;
    a.r64[1] = -36.875;
    a.r64[2] = 22.95;
    a.r64[3] = 3.75;

    b.r64[0] = -0.0625;
    b.r64[1] = -67.375;
    b.r64[2] = 22.95;
    b.r64[3] = numeric_limits<double>::quiet_NaN();

    AvxPfpCompare(&a, &b, c);

    printf("Results for AvxPackedFloatingPointCompare\n");
    printf("a: %s\n", a.ToString_r64(buff, sizeof(buff), false));
    printf("a: %s\n", a.ToString_r64(buff, sizeof(buff), true));
    printf("\n");
    printf("b: %s\n", b.ToString_r64(buff, sizeof(buff), false));
    printf("b: %s\n", b.ToString_r64(buff, sizeof(buff), true));

    for (int i = 0; i < 8; i++)
    {
        printf("\n%s results\n", instr_names[i]);
        printf("  %s\n", c[i].ToString_x64(buff, sizeof(buff), false));
        printf("  %s\n", c[i].ToString_x64(buff, sizeof(buff), true));
    }

    return 0;
}

avxpackedfloatingpointcompare.asm

; Name:     avxpackedfloatingpointcompare.asm
;
; Build:    g++ -c -m32 main.cpp -o main.o
;           nasm -f elf32 -o avxpackedfloatingpointcompare.o avxpackedfloatingpointcompare.asm
;           g++ -m32 -o avxpackedfloatingpointcompare avxpackedfloatingpointcompare.o main.o ../../commonfiles/ymmval.o
;
; Source:   Modern x86 Assembly Language Programming p. 386

global AvxPfpCompare

section .text

; extern "C" void AvxPfpCompare_(const YmmVal* a, const YmmVal* b, YmmVal c[8]);
;
; Description:  The following function demonstrates use of the
;               x86-AVX compare instruction vcmppd.
;
; Requires:     AVX

%define a   [ebp+8]
%define b   [ebp+12]
%define c   [ebp+16]

AvxPfpCompare:
    push    ebp
    mov     ebp,esp

; Load argument values
    mov     eax,a               ;eax = ptr to a
    mov     ecx,b               ;ecx = ptr to b
    mov     edx,c               ;edx = ptr to c
    vmovapd ymm0,[eax]          ;ymm0 = a
    vmovapd ymm1,[ecx]          ;ymm1 = b

; Compare for equality
    vcmpeqpd ymm2,ymm0,ymm1
    vmovapd  [edx],ymm2

; Compare for inequality
    vcmpneqpd ymm2,ymm0,ymm1
    vmovapd   [edx+32],ymm2

; Compare for less than
    vcmpltpd ymm2,ymm0,ymm1
    vmovapd  [edx+64],ymm2

; Compare for less than or equal
    vcmplepd ymm2,ymm0,ymm1
    vmovapd  [edx+96],ymm2

; Compare for greater than
    vcmpgtpd ymm2,ymm0,ymm1
    vmovapd  [edx+128],ymm2

; Compare for greater than or equal
    vcmpgepd ymm2,ymm0,ymm1
    vmovapd  [edx+160],ymm2

; Compare for ordered
    vcmpordpd ymm2,ymm0,ymm1
    vmovapd   [edx+192],ymm2

; Compare for unordered
    vcmpunordpd ymm2,ymm0,ymm1
    vmovapd     [edx+224],ymm2

; Zero upper 128-bit of all YMM registers to avoid potential x86-AVX
; to x86-SSE transition penalties.
    vzeroupper
    pop     ebp
    ret

build

g++ -c -m32 main.cpp -o main.o
nasm -f elf32 -o avxpackedfloatingpointcompare.o avxpackedfloatingpointcompare.asm
g++ -m32 -o avxpackedfloatingpointcompare avxpackedfloatingpointcompare.o main.o ../../commonfiles/ymmval.o