PCMPESTRM (XMM, XMM, I8) - Port Usage

With blocking instructions for port '0':

Code:

   0:	66 44 0f 50 c3       	movmskpd r8d,xmm3
   5:	66 44 0f 50 cb       	movmskpd r9d,xmm3
   a:	66 44 0f 50 d3       	movmskpd r10d,xmm3
   f:	66 44 0f 50 db       	movmskpd r11d,xmm3
  14:	66 44 0f 50 e3       	movmskpd r12d,xmm3
  19:	66 0f 50 db          	movmskpd ebx,xmm3
  1d:	66 0f 50 cb          	movmskpd ecx,xmm3
  21:	66 44 0f 50 c3       	movmskpd r8d,xmm3
  26:	66 44 0f 50 cb       	movmskpd r9d,xmm3
  2b:	66 44 0f 50 d3       	movmskpd r10d,xmm3
  30:	66 44 0f 50 db       	movmskpd r11d,xmm3
  35:	66 44 0f 50 e3       	movmskpd r12d,xmm3
  3a:	66 0f 3a 60 ca 02    	pcmpestrm xmm1,xmm2,0x2

Init:

MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOVUPD XMM3, [R14];
PXOR XMM1, XMM1;
PXOR XMM2, XMM2

Show nanoBench command
Results:
- Instructions retired: 13.0
- Core cycles: 12.0
- Reference cycles: 11.04
- UOPS_PORT_0: 12.05
- UOPS_PORT_1: 4.4
- UOPS_PORT_2: 0.0
- UOPS_PORT_3: 0.0
- UOPS_PORT_4: 0.0
- UOPS_PORT_5: 4.78
- UOPS_RETIRED.ANY: 21.0

With blocking instructions for port '1':

Code:

   0:	66 0f 58 dc          	addpd  xmm3,xmm4
   4:	66 0f 58 ec          	addpd  xmm5,xmm4
   8:	66 0f 58 f4          	addpd  xmm6,xmm4
   c:	66 0f 58 fc          	addpd  xmm7,xmm4
  10:	66 44 0f 58 c4       	addpd  xmm8,xmm4
  15:	66 44 0f 58 cc       	addpd  xmm9,xmm4
  1a:	66 44 0f 58 d4       	addpd  xmm10,xmm4
  1f:	66 44 0f 58 dc       	addpd  xmm11,xmm4
  24:	66 44 0f 58 e4       	addpd  xmm12,xmm4
  29:	66 0f 58 dc          	addpd  xmm3,xmm4
  2d:	66 0f 58 ec          	addpd  xmm5,xmm4
  31:	66 0f 58 f4          	addpd  xmm6,xmm4
  35:	66 0f 3a 60 ca 02    	pcmpestrm xmm1,xmm2,0x2

Init:

MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOVUPD XMM3, [R14];
MOVUPD XMM4, [R14];
MOVUPD XMM5, [R14];
MOVUPD XMM6, [R14];
MOVUPD XMM7, [R14];
MOVUPD XMM8, [R14];
MOVUPD XMM9, [R14];
MOVUPD XMM10, [R14];
MOVUPD XMM11, [R14];
MOVUPD XMM12, [R14];
PXOR XMM1, XMM1;
PXOR XMM2, XMM2

Show nanoBench command
Results:
- Instructions retired: 13.0
- Core cycles: 15.07
- Reference cycles: 13.88
- UOPS_PORT_0: 2.69
- UOPS_PORT_1: 14.92
- UOPS_PORT_2: 0.0
- UOPS_PORT_3: 0.0
- UOPS_PORT_4: 0.0
- UOPS_PORT_5: 3.68
- UOPS_RETIRED.ANY: 21.0

⇨ 3 μops that can only use port '1'

With blocking instructions for port '5':

Code:

   0:	66 0f 55 dc          	andnpd xmm3,xmm4
   4:	66 0f 55 ec          	andnpd xmm5,xmm4
   8:	66 0f 55 f4          	andnpd xmm6,xmm4
   c:	66 0f 55 fc          	andnpd xmm7,xmm4
  10:	66 44 0f 55 c4       	andnpd xmm8,xmm4
  15:	66 44 0f 55 cc       	andnpd xmm9,xmm4
  1a:	66 44 0f 55 d4       	andnpd xmm10,xmm4
  1f:	66 44 0f 55 dc       	andnpd xmm11,xmm4
  24:	66 44 0f 55 e4       	andnpd xmm12,xmm4
  29:	66 0f 55 dc          	andnpd xmm3,xmm4
  2d:	66 0f 55 ec          	andnpd xmm5,xmm4
  31:	66 0f 55 f4          	andnpd xmm6,xmm4
  35:	66 0f 3a 60 ca 02    	pcmpestrm xmm1,xmm2,0x2

Init:

PXOR XMM3, XMM3;
PXOR XMM4, XMM4;
PXOR XMM5, XMM5;
PXOR XMM6, XMM6;
PXOR XMM7, XMM7;
PXOR XMM8, XMM8;
PXOR XMM9, XMM9;
PXOR XMM10, XMM10;
PXOR XMM11, XMM11;
PXOR XMM12, XMM12;
PXOR XMM1, XMM1;
PXOR XMM2, XMM2

Show nanoBench command
Results:
- Instructions retired: 13.0
- Core cycles: 12.01
- Reference cycles: 11.04
- UOPS_PORT_0: 3.82
- UOPS_PORT_1: 5.17
- UOPS_PORT_2: 0.0
- UOPS_PORT_3: 0.0
- UOPS_PORT_4: 0.0
- UOPS_PORT_5: 12.36
- UOPS_RETIRED.ANY: 21.0

With blocking instructions for ports {'0', '5'}:

Code:

   0:	0f 38 1c c1          	pabsb  mm0,mm1
   4:	0f 38 1c d1          	pabsb  mm2,mm1
   8:	0f 38 1c d9          	pabsb  mm3,mm1
   c:	0f 38 1c e1          	pabsb  mm4,mm1
  10:	0f 38 1c e9          	pabsb  mm5,mm1
  14:	0f 38 1c f1          	pabsb  mm6,mm1
  18:	0f 38 1c f9          	pabsb  mm7,mm1
  1c:	0f 38 1c c1          	pabsb  mm0,mm1
  20:	0f 38 1c d1          	pabsb  mm2,mm1
  24:	0f 38 1c d9          	pabsb  mm3,mm1
  28:	0f 38 1c e1          	pabsb  mm4,mm1
  2c:	0f 38 1c e9          	pabsb  mm5,mm1
  30:	0f 38 1c f1          	pabsb  mm6,mm1
  34:	0f 38 1c f9          	pabsb  mm7,mm1
  38:	0f 38 1c c1          	pabsb  mm0,mm1
  3c:	0f 38 1c d1          	pabsb  mm2,mm1
  40:	0f 38 1c d9          	pabsb  mm3,mm1
  44:	0f 38 1c e1          	pabsb  mm4,mm1
  48:	0f 38 1c e9          	pabsb  mm5,mm1
  4c:	0f 38 1c f1          	pabsb  mm6,mm1
  50:	0f 38 1c f9          	pabsb  mm7,mm1
  54:	0f 38 1c c1          	pabsb  mm0,mm1
  58:	0f 38 1c d1          	pabsb  mm2,mm1
  5c:	0f 38 1c d9          	pabsb  mm3,mm1
  60:	66 0f 3a 60 ca 02    	pcmpestrm xmm1,xmm2,0x2

Init:

PXOR MM1, MM1;
PXOR XMM1, XMM1;
PXOR XMM2, XMM2

Show nanoBench command
Results:
- Instructions retired: 25.0
- Core cycles: 13.07
- Reference cycles: 11.8
- UOPS_PORT_0: 12.85
- UOPS_PORT_1: 7.29
- UOPS_PORT_2: 0.0
- UOPS_PORT_3: 0.0
- UOPS_PORT_4: 0.0
- UOPS_PORT_5: 13.27
- UOPS_RETIRED.ANY: 33.0

⇨ One μop that can only use ports {'0', '5'}

With blocking instructions for ports {'0', '1', '5'}:

Code:

   0:	41 b8 00 00 00 01    	mov    r8d,0x1000000
   6:	41 b9 00 00 00 01    	mov    r9d,0x1000000
   c:	41 ba 00 00 00 01    	mov    r10d,0x1000000
  12:	41 bb 00 00 00 01    	mov    r11d,0x1000000
  18:	41 bc 00 00 00 01    	mov    r12d,0x1000000
  1e:	bb 00 00 00 01       	mov    ebx,0x1000000
  23:	b9 00 00 00 01       	mov    ecx,0x1000000
  28:	41 b8 00 00 00 01    	mov    r8d,0x1000000
  2e:	41 b9 00 00 00 01    	mov    r9d,0x1000000
  34:	41 ba 00 00 00 01    	mov    r10d,0x1000000
  3a:	41 bb 00 00 00 01    	mov    r11d,0x1000000
  40:	41 bc 00 00 00 01    	mov    r12d,0x1000000
  46:	bb 00 00 00 01       	mov    ebx,0x1000000
  4b:	b9 00 00 00 01       	mov    ecx,0x1000000
  50:	41 b8 00 00 00 01    	mov    r8d,0x1000000
  56:	41 b9 00 00 00 01    	mov    r9d,0x1000000
  5c:	41 ba 00 00 00 01    	mov    r10d,0x1000000
  62:	41 bb 00 00 00 01    	mov    r11d,0x1000000
  68:	41 bc 00 00 00 01    	mov    r12d,0x1000000
  6e:	bb 00 00 00 01       	mov    ebx,0x1000000
  73:	b9 00 00 00 01       	mov    ecx,0x1000000
  78:	41 b8 00 00 00 01    	mov    r8d,0x1000000
  7e:	41 b9 00 00 00 01    	mov    r9d,0x1000000
  84:	41 ba 00 00 00 01    	mov    r10d,0x1000000
  8a:	41 bb 00 00 00 01    	mov    r11d,0x1000000
  90:	41 bc 00 00 00 01    	mov    r12d,0x1000000
  96:	bb 00 00 00 01       	mov    ebx,0x1000000
  9b:	b9 00 00 00 01       	mov    ecx,0x1000000
  a0:	41 b8 00 00 00 01    	mov    r8d,0x1000000
  a6:	41 b9 00 00 00 01    	mov    r9d,0x1000000
  ac:	41 ba 00 00 00 01    	mov    r10d,0x1000000
  b2:	41 bb 00 00 00 01    	mov    r11d,0x1000000
  b8:	41 bc 00 00 00 01    	mov    r12d,0x1000000
  be:	bb 00 00 00 01       	mov    ebx,0x1000000
  c3:	b9 00 00 00 01       	mov    ecx,0x1000000
  c8:	41 b8 00 00 00 01    	mov    r8d,0x1000000
  ce:	66 0f 3a 60 ca 02    	pcmpestrm xmm1,xmm2,0x2

Init:
```
PXOR XMM1, XMM1;
PXOR XMM2, XMM2
```
Show nanoBench command
Results:
- Instructions retired: 37.0
- Core cycles: 18.52
- Reference cycles: 17.04
- UOPS_PORT_0: 15.74
- UOPS_PORT_1: 12.78
- UOPS_PORT_2: 0.0
- UOPS_PORT_3: 0.0
- UOPS_PORT_4: 0.0
- UOPS_PORT_5: 16.93
- UOPS_RETIRED.ANY: 45.0

⇨ 5 μops that can only use ports {'0', '1', '5'}