VPHADDD (YMM, YMM, YMM) - Port Usage

With blocking instructions for port '1':

Code:

   0:	47 8d 84 6e 80 00 00 	lea    r8d,[r14+r13*2+0x80]
   7:	00 
   8:	47 8d 8c 6e 80 00 00 	lea    r9d,[r14+r13*2+0x80]
   f:	00 
  10:	47 8d 94 6e 80 00 00 	lea    r10d,[r14+r13*2+0x80]
  17:	00 
  18:	47 8d 9c 6e 80 00 00 	lea    r11d,[r14+r13*2+0x80]
  1f:	00 
  20:	47 8d a4 6e 80 00 00 	lea    r12d,[r14+r13*2+0x80]
  27:	00 
  28:	43 8d 9c 6e 80 00 00 	lea    ebx,[r14+r13*2+0x80]
  2f:	00 
  30:	43 8d 8c 6e 80 00 00 	lea    ecx,[r14+r13*2+0x80]
  37:	00 
  38:	43 8d 94 6e 80 00 00 	lea    edx,[r14+r13*2+0x80]
  3f:	00 
  40:	47 8d 84 6e 80 00 00 	lea    r8d,[r14+r13*2+0x80]
  47:	00 
  48:	47 8d 8c 6e 80 00 00 	lea    r9d,[r14+r13*2+0x80]
  4f:	00 
  50:	c4 e2 75 02 c2       	vphaddd ymm0,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
VXORPS YMM1, YMM1, YMM1;
VXORPS YMM2, YMM2, YMM2

Show nanoBench command
Results:
- Instructions retired: 11.0
- Core cycles: 10.07
- Reference cycles: 10.6
- UOPS_PORT_1: 10.0

With blocking instructions for port '5':

Code:

   0:	41 0f 6e c0          	movd   mm0,r8d
   4:	41 0f 6e c8          	movd   mm1,r8d
   8:	41 0f 6e d0          	movd   mm2,r8d
   c:	41 0f 6e d8          	movd   mm3,r8d
  10:	41 0f 6e e0          	movd   mm4,r8d
  14:	41 0f 6e e8          	movd   mm5,r8d
  18:	41 0f 6e f0          	movd   mm6,r8d
  1c:	41 0f 6e f8          	movd   mm7,r8d
  20:	41 0f 6e c0          	movd   mm0,r8d
  24:	41 0f 6e c8          	movd   mm1,r8d
  28:	c4 e2 75 02 c2       	vphaddd ymm0,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
VXORPS YMM1, YMM1, YMM1;
VXORPS YMM2, YMM2, YMM2

Show nanoBench command
Results:
- Instructions retired: 11.0
- Core cycles: 12.05
- Reference cycles: 12.6
- UOPS_PORT_5: 11.99

⇨ 2 μops that can only use port '5'

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

Code:

   0:	c4 e2 d9 98 dd       	vfmadd132pd xmm3,xmm4,xmm5
   5:	c4 e2 d9 98 f5       	vfmadd132pd xmm6,xmm4,xmm5
   a:	c4 e2 d9 98 fd       	vfmadd132pd xmm7,xmm4,xmm5
   f:	c4 62 d9 98 c5       	vfmadd132pd xmm8,xmm4,xmm5
  14:	c4 62 d9 98 cd       	vfmadd132pd xmm9,xmm4,xmm5
  19:	c4 62 d9 98 d5       	vfmadd132pd xmm10,xmm4,xmm5
  1e:	c4 62 d9 98 dd       	vfmadd132pd xmm11,xmm4,xmm5
  23:	c4 62 d9 98 e5       	vfmadd132pd xmm12,xmm4,xmm5
  28:	c4 e2 d9 98 dd       	vfmadd132pd xmm3,xmm4,xmm5
  2d:	c4 e2 d9 98 f5       	vfmadd132pd xmm6,xmm4,xmm5
  32:	c4 e2 75 02 c2       	vphaddd ymm0,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
VMOVUPD XMM3, [R14];
VMOVUPD XMM4, [R14];
VMOVUPD XMM5, [R14];
VMOVUPD XMM6, [R14];
VMOVUPD XMM7, [R14];
VMOVUPD XMM8, [R14];
VMOVUPD XMM9, [R14];
VMOVUPD XMM10, [R14];
VMOVUPD XMM11, [R14];
VMOVUPD XMM12, [R14];
VXORPS YMM1, YMM1, YMM1;
VXORPS YMM2, YMM2, YMM2

Show nanoBench command
Results:
- Instructions retired: 11.0
- Core cycles: 10.17
- Reference cycles: 10.77
- UOPS_PORT_1: 5.0
- UOPS_PORT_0: 5.0

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

Code:

   0:	45 8d 06             	lea    r8d,[r14]
   3:	45 8d 0e             	lea    r9d,[r14]
   6:	45 8d 16             	lea    r10d,[r14]
   9:	45 8d 1e             	lea    r11d,[r14]
   c:	45 8d 26             	lea    r12d,[r14]
   f:	41 8d 1e             	lea    ebx,[r14]
  12:	41 8d 0e             	lea    ecx,[r14]
  15:	41 8d 16             	lea    edx,[r14]
  18:	45 8d 06             	lea    r8d,[r14]
  1b:	45 8d 0e             	lea    r9d,[r14]
  1e:	c4 e2 75 02 c2       	vphaddd ymm0,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
VXORPS YMM1, YMM1, YMM1;
VXORPS YMM2, YMM2, YMM2

Show nanoBench command
Results:
- Instructions retired: 11.0
- Core cycles: 6.56
- Reference cycles: 6.93
- UOPS_PORT_1: 6.49
- UOPS_PORT_5: 6.5

⇨ One μop that can only use ports {'1', '5'}