VSQRTPS (XMM, XMM) - Latency

Experiment 1

• Instruction: VSQRTPS XMM0, XMM2
• Chain instructions: VORPS XMM2, XMM0, XMM0; VORPS XMM2, XMM2, XMM1; VANDPS XMM2, XMM2, XMM1; VORPS XMM2, XMM2, XMM2; VORPS XMM2, XMM2, XMM2; VORPS XMM2, XMM2, XMM2; VORPS XMM2, XMM2, XMM2; VORPS XMM2, XMM2, XMM2; VORPS XMM2, XMM2, XMM2;
• Chain latency: ≥9
• Code:

     0:	c5 f8 51 c2          	vsqrtps xmm0,xmm2
     4:	c5 f8 56 d0          	vorps  xmm2,xmm0,xmm0
     8:	c5 e8 56 d1          	vorps  xmm2,xmm2,xmm1
     c:	c5 e8 54 d1          	vandps xmm2,xmm2,xmm1
    10:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
    14:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
    18:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
    1c:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
    20:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
    24:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
  

• Init:

  MOV R15, 10000;
  L: VADDPS YMM0, YMM1, YMM1;
  VADDPS YMM0, YMM1, YMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x72d30ff172d30ff1;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  VMOVUPS XMM2, [R14];
  VMOVUPS XMM1, [R14]

• Show nanoBench command
• Results:
  • MPERF: 20.06
  • APERF: 23.0
  • UOPS: 10.0

Experiment 2

• Instruction: VSQRTPS XMM2, XMM2
• Code:

     0:	c5 f8 51 d2          	vsqrtps xmm2,xmm2
  

• Init:

  MOV R15, 10000;
  L: VADDPS YMM0, YMM1, YMM1;
  VADDPS YMM0, YMM1, YMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x3f8000003f800000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  VMOVUPS XMM2, [R14];
  VMOVUPS XMM1, [R14]

• Show nanoBench command
• Results:
  • MPERF: 6.91
  • APERF: 8.0
  • UOPS: 1.0

Experiment 3

• Instruction: VSQRTPS XMM0, XMM2
• Chain instructions: VORPS XMM2, XMM0, XMM0; VORPS XMM2, XMM2, XMM1; VANDPS XMM2, XMM2, XMM1; VORPS XMM2, XMM2, XMM2; VORPS XMM2, XMM2, XMM2; VORPS XMM2, XMM2, XMM2; VORPS XMM2, XMM2, XMM2; VORPS XMM2, XMM2, XMM2; VORPS XMM2, XMM2, XMM2;
• Chain latency: ≥9
• Code:

     0:	c5 f8 51 c2          	vsqrtps xmm0,xmm2
     4:	c5 f8 56 d0          	vorps  xmm2,xmm0,xmm0
     8:	c5 e8 56 d1          	vorps  xmm2,xmm2,xmm1
     c:	c5 e8 54 d1          	vandps xmm2,xmm2,xmm1
    10:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
    14:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
    18:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
    1c:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
    20:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
    24:	c5 e8 56 d2          	vorps  xmm2,xmm2,xmm2
  

• Init:

  MOV R15, 10000;
  L: VADDPS YMM0, YMM1, YMM1;
  VADDPS YMM0, YMM1, YMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x3f8000003f800000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  VMOVUPS XMM2, [R14];
  VMOVUPS XMM1, [R14]

• Show nanoBench command
• Results:
  • MPERF: 14.73
  • APERF: 17.0
  • UOPS: 10.0