VSQRTPD (XMM, XMM) - Latency

Experiment 1

• Instruction: VSQRTPD XMM0, XMM2
• Chain instructions: VORPD XMM2, XMM0, XMM0; VORPD XMM2, XMM2, XMM1; VANDPD XMM2, XMM2, XMM1; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2;
• Chain latency: ≥15
• Code:

     0:	c5 f9 51 c2          	vsqrtpd xmm0,xmm2
     4:	c5 f9 56 d0          	vorpd  xmm2,xmm0,xmm0
     8:	c5 e9 56 d1          	vorpd  xmm2,xmm2,xmm1
     c:	c5 e9 54 d1          	vandpd xmm2,xmm2,xmm1
    10:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    14:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    18:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    1c:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    20:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    24:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    28:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    2c:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    30:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    34:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    38:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    3c:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
  

• Init:

  MOV R15, 10000;
  L: VADDPS YMM0, YMM1, YMM1;
  VADDPS YMM0, YMM1, YMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x465a61fe1acdc21c;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  VMOVUPD XMM2, [R14];
  VMOVUPD XMM1, [R14]

• Show nanoBench command
• Results:
  • MPERF: 29.23
  • APERF: 35.49
  • UOPS: 16.0

Experiment 2

• Instruction: VSQRTPD XMM2, XMM2
• Code:

     0:	c5 f9 51 d2          	vsqrtpd xmm2,xmm2
  

• Init:

  MOV R15, 10000;
  L: VADDPS YMM0, YMM1, YMM1;
  VADDPS YMM0, YMM1, YMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x3ff0000000000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  VMOVUPD XMM2, [R14];
  VMOVUPD XMM1, [R14]

• Show nanoBench command
• Results:
  • MPERF: 16.71
  • APERF: 20.51
  • UOPS: 1.0

Experiment 3

• Instruction: VSQRTPD XMM0, XMM2
• Chain instructions: VORPD XMM2, XMM0, XMM0; VORPD XMM2, XMM2, XMM1; VANDPD XMM2, XMM2, XMM1; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2; VORPD XMM2, XMM2, XMM2;
• Chain latency: ≥15
• Code:

     0:	c5 f9 51 c2          	vsqrtpd xmm0,xmm2
     4:	c5 f9 56 d0          	vorpd  xmm2,xmm0,xmm0
     8:	c5 e9 56 d1          	vorpd  xmm2,xmm2,xmm1
     c:	c5 e9 54 d1          	vandpd xmm2,xmm2,xmm1
    10:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    14:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    18:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    1c:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    20:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    24:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    28:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    2c:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    30:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    34:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    38:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
    3c:	c5 e9 56 d2          	vorpd  xmm2,xmm2,xmm2
  

• Init:

  MOV R15, 10000;
  L: VADDPS YMM0, YMM1, YMM1;
  VADDPS YMM0, YMM1, YMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x3ff0000000000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  VMOVUPD XMM2, [R14];
  VMOVUPD XMM1, [R14]

• Show nanoBench command
• Results:
  • MPERF: 29.17
  • APERF: 35.49
  • UOPS: 16.0