v4df and v8df

add_v4df_O0.s

add_v4df(double __vector, double __vector):
# parameter 1: %ymm0
# parameter 2: %ymm1
..B1.1:                         # Preds ..B1.0
        .cfi_startproc
        .cfi_personality 0x3,__gxx_personality_v0
..___tag_value__Z8add_v4dfU8__vectordS_.1:
..L2:
                                                          #9.25
        pushq     %rbp                                          #9.25
        .cfi_def_cfa_offset 16
        movq      %rsp, %rbp                                    #9.25
        .cfi_def_cfa 6, 16
        .cfi_offset 6, -16
        subq      $112, %rsp                                    #9.25
        vmovupd   %ymm0, -64(%rbp)                              #9.25
        vmovupd   %ymm1, -32(%rbp)                              #9.25
        movl      $4, -112(%rbp)                                #10.14
        movl      $0, -108(%rbp)                                #10.14
                                # LOE rbx rbp rsp r12 r13 r14 r15 rip
..B1.2:                         # Preds ..B1.2 ..B1.1
        lea       -64(%rbp), %rax                               #10.14
        movl      -108(%rbp), %edx                              #10.14
        movslq    %edx, %rdx                                    #10.14
        imulq     $8, %rdx, %rdx                                #10.14
        addq      %rdx, %rax                                    #10.14
        lea       -32(%rbp), %rdx                               #10.14
        movl      -108(%rbp), %ecx                              #10.14
        movslq    %ecx, %rcx                                    #10.14
        imulq     $8, %rcx, %rcx                                #10.14
        addq      %rcx, %rdx                                    #10.14
        vmovsd    (%rax), %xmm0                                 #10.14
        vmovsd    (%rdx), %xmm1                                 #10.14
        vaddsd    %xmm1, %xmm0, %xmm0                           #10.14
        lea       -104(%rbp), %rax                              #10.14
        movl      -108(%rbp), %edx                              #10.14
        movslq    %edx, %rdx                                    #10.14
        imulq     $8, %rdx, %rdx                                #10.14
        addq      %rdx, %rax                                    #10.14
        vmovsd    %xmm0, (%rax)                                 #10.14
        movl      $1, %eax                                      #10.14
        addl      -108(%rbp), %eax                              #10.14
        movl      %eax, -108(%rbp)                              #10.14
        movl      -108(%rbp), %eax                              #10.14
        movl      -112(%rbp), %edx                              #10.14
        cmpl      %edx, %eax                                    #10.14
        jl        ..B1.2        # Prob 50%                      #10.14
                                # LOE rbx rbp rsp r12 r13 r14 r15 rip
..B1.3:                         # Preds ..B1.2
        vmovupd   -104(%rbp), %ymm0                             #11.1
        leave                                                   #10.14
        .cfi_restore 6
        ret                                                     #10.14

add_v4df_O3.s

add_v4df(double __vector, double __vector):
# parameter 1: %ymm0
# parameter 2: %ymm1
..B2.1:                         # Preds ..B2.0
        .cfi_startproc
        .cfi_personality 0x3,__gxx_personality_v0
..___tag_value__Z8add_v4dfU8__vectordS_.35:
..L36:
                                                         #9.25
        vaddpd    %ymm1, %ymm0, %ymm0                           #10.14
        ret                                                     #10.14

makefile

all: test_O0.s test_O3.s

test_O0.s: test_s0.s
        c++filt < $< > $@

test_O3.s: test_s3.s
        c++filt < $< > $@

test_s0.s: test.cpp
        icpc -std=c++11 -xHOST -O0 -S $< -o $@

test_s3.s: test.cpp
        icpc -std=c++11 -xHOST -O3 -S $< -o $@

clean:
        rm -f *.s

test.cpp

#include <stdio.h>
#include <stdlib.h>
#include <immintrin.h>

typedef double v4df __attribute__((vector_size(32)));
typedef double v8df __attribute__((vector_size(64)));

v4df
add_v4df(v4df a, v4df b){
  return a + b;
}

v8df
add_v8df(v8df a, v8df b){
  return a + b;
}

void
put_v4df(v4df &a){
  double *x = (double*)(&a);
  for(int i=0;i<4; i++) printf("%f ",a[i]);
  printf("\n");
}

void
put_v8df(v8df &a){
  double *x = (double*)(&a);
  for(int i=0;i<8; i++) printf("%f ",a[i]);
  printf("\n");
}

v4df
set1_v4df(double v){
  v4df a;
  for(int i=0;i<4; i++) a[i] = v;
  return a;
}

v8df
set1_v8df(double v){
  v8df a;
  for(int i=0;i<8; i++) a[i] = v;
  return a;
}


int
main(int argc, char **argv){
  const double a = atof(argv[1]);
  const double b = atof(argv[2]);
  v4df v4a = set1_v4df(a);
  v4df v4b = set1_v4df(b);
  v4df v4c = add_v4df(v4a,v4b);
  put_v4df(v4c);

  v8df v8a = set1_v8df(a);
  v8df v8b = set1_v8df(b);
  v8df v8c = add_v8df(v8a,v8b);
  put_v8df(v8c);
}