VEXTRACTI128 (M128, YMM, I8) - Port Usage

With blocking instructions for ports {'4', '5'}:

Code:

   0:	c4 c1 79 7e c8       	vmovd  r8d,xmm1
   5:	c4 c1 79 7e c9       	vmovd  r9d,xmm1
   a:	c4 c1 79 7e ca       	vmovd  r10d,xmm1
   f:	c4 c1 79 7e cb       	vmovd  r11d,xmm1
  14:	c4 c1 79 7e cc       	vmovd  r12d,xmm1
  19:	c5 f9 7e cb          	vmovd  ebx,xmm1
  1d:	c5 f9 7e c9          	vmovd  ecx,xmm1
  21:	c5 f9 7e ca          	vmovd  edx,xmm1
  25:	c5 f9 7e c8          	vmovd  eax,xmm1
  29:	c4 c1 79 7e c8       	vmovd  r8d,xmm1
  2e:	c4 c3 7d 39 06 02    	vextracti128 XMMWORD PTR [r14],ymm0,0x2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
VXORPS XMM1, XMM1, XMM1;
VXORPS YMM0, YMM0, YMM0

Show nanoBench command
Results:
- MPERF: 8.7
- APERF: 10.05
- FpuPipeAssignment.Total0: 0.62
- FpuPipeAssignment.Total1: 0.34
- FpuPipeAssignment.Total2: 0.03
- FpuPipeAssignment.Total3: 0.03
- FpuPipeAssignment.Total4: 5.51
- FpuPipeAssignment.Total5: 5.5

⇨ One μop that can only use ports {'4', '5'}

With blocking instructions for ports {'0', '1', '2', '3'}:

Code:

   0:	0f 6f c1             	movq   mm0,mm1
   3:	0f 6f d1             	movq   mm2,mm1
   6:	0f 6f d9             	movq   mm3,mm1
   9:	0f 6f e1             	movq   mm4,mm1
   c:	0f 6f e9             	movq   mm5,mm1
   f:	0f 6f f1             	movq   mm6,mm1
  12:	0f 6f f9             	movq   mm7,mm1
  15:	0f 6f c1             	movq   mm0,mm1
  18:	0f 6f d1             	movq   mm2,mm1
  1b:	0f 6f d9             	movq   mm3,mm1
  1e:	c4 c3 7d 39 06 02    	vextracti128 XMMWORD PTR [r14],ymm0,0x2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
PXOR MM1, MM1;
VZEROALL;
VXORPS YMM0, YMM0, YMM0

Show nanoBench command
Results:
- MPERF: 2.66
- APERF: 3.05
- FpuPipeAssignment.Total0: 3.0
- FpuPipeAssignment.Total1: 2.5
- FpuPipeAssignment.Total2: 3.0
- FpuPipeAssignment.Total3: 2.5
- FpuPipeAssignment.Total4: 0.0
- FpuPipeAssignment.Total5: 1.0

⇨ One μop that can only use ports {'0', '1', '2', '3'}

With an indexed addressing mode

With blocking instructions for ports {'0', '1'}:

Code:

   0:	0f ec c1             	paddsb mm0,mm1
   3:	0f ec d1             	paddsb mm2,mm1
   6:	0f ec d9             	paddsb mm3,mm1
   9:	0f ec e1             	paddsb mm4,mm1
   c:	0f ec e9             	paddsb mm5,mm1
   f:	0f ec f1             	paddsb mm6,mm1
  12:	0f ec f9             	paddsb mm7,mm1
  15:	0f ec c1             	paddsb mm0,mm1
  18:	0f ec d1             	paddsb mm2,mm1
  1b:	0f ec d9             	paddsb mm3,mm1
  1e:	c4 83 7d 39 04 2e 02 	vextracti128 XMMWORD PTR [r14+r13*1],ymm0,0x2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
PXOR MM0, MM0;
PXOR MM1, MM1;
PXOR MM2, MM2;
PXOR MM3, MM3;
PXOR MM4, MM4;
PXOR MM5, MM5;
PXOR MM6, MM6;
PXOR MM7, MM7;
VZEROALL;
VXORPS YMM0, YMM0, YMM0;
XOR R13, R13

Show nanoBench command
Results:
- MPERF: 4.23
- APERF: 5.01
- FpuPipeAssignment.Total0: 5.01
- FpuPipeAssignment.Total1: 5.0
- FpuPipeAssignment.Total2: 0.91
- FpuPipeAssignment.Total3: 0.09
- FpuPipeAssignment.Total4: 0.09
- FpuPipeAssignment.Total5: 0.92

With blocking instructions for ports {'4', '5'}:

Code:

   0:	c4 c1 79 7e c8       	vmovd  r8d,xmm1
   5:	c4 c1 79 7e c9       	vmovd  r9d,xmm1
   a:	c4 c1 79 7e ca       	vmovd  r10d,xmm1
   f:	c4 c1 79 7e cb       	vmovd  r11d,xmm1
  14:	c4 c1 79 7e cc       	vmovd  r12d,xmm1
  19:	c5 f9 7e cb          	vmovd  ebx,xmm1
  1d:	c5 f9 7e c9          	vmovd  ecx,xmm1
  21:	c5 f9 7e ca          	vmovd  edx,xmm1
  25:	c5 f9 7e c8          	vmovd  eax,xmm1
  29:	c4 c1 79 7e c8       	vmovd  r8d,xmm1
  2e:	c4 83 7d 39 04 2e 02 	vextracti128 XMMWORD PTR [r14+r13*1],ymm0,0x2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
VXORPS XMM1, XMM1, XMM1;
VXORPS YMM0, YMM0, YMM0;
XOR R13, R13

Show nanoBench command
Results:
- MPERF: 8.62
- APERF: 10.05
- FpuPipeAssignment.Total0: 0.45
- FpuPipeAssignment.Total1: 0.07
- FpuPipeAssignment.Total2: 0.03
- FpuPipeAssignment.Total3: 0.46
- FpuPipeAssignment.Total4: 5.51
- FpuPipeAssignment.Total5: 5.5

⇨ One μop that can only use ports {'4', '5'}

With blocking instructions for ports {'0', '1', '2', '3'}:

Code:

   0:	0f 6f c1             	movq   mm0,mm1
   3:	0f 6f d1             	movq   mm2,mm1
   6:	0f 6f d9             	movq   mm3,mm1
   9:	0f 6f e1             	movq   mm4,mm1
   c:	0f 6f e9             	movq   mm5,mm1
   f:	0f 6f f1             	movq   mm6,mm1
  12:	0f 6f f9             	movq   mm7,mm1
  15:	0f 6f c1             	movq   mm0,mm1
  18:	0f 6f d1             	movq   mm2,mm1
  1b:	0f 6f d9             	movq   mm3,mm1
  1e:	c4 83 7d 39 04 2e 02 	vextracti128 XMMWORD PTR [r14+r13*1],ymm0,0x2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
PXOR MM1, MM1;
VZEROALL;
VXORPS YMM0, YMM0, YMM0;
XOR R13, R13

Show nanoBench command
Results:
- MPERF: 2.59
- APERF: 3.05
- FpuPipeAssignment.Total0: 3.0
- FpuPipeAssignment.Total1: 2.5
- FpuPipeAssignment.Total2: 3.0
- FpuPipeAssignment.Total3: 2.5
- FpuPipeAssignment.Total4: 0.0
- FpuPipeAssignment.Total5: 1.0

⇨ One μop that can only use ports {'0', '1', '2', '3'}