VFMADD213PS (YMM, YMM, M256) - Port Usage

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

Code:

   0:	c4 e2 e1 98 d4       	vfmadd132pd xmm2,xmm3,xmm4
   5:	c4 e2 e1 98 ec       	vfmadd132pd xmm5,xmm3,xmm4
   a:	c4 e2 e1 98 f4       	vfmadd132pd xmm6,xmm3,xmm4
   f:	c4 e2 e1 98 fc       	vfmadd132pd xmm7,xmm3,xmm4
  14:	c4 62 e1 98 c4       	vfmadd132pd xmm8,xmm3,xmm4
  19:	c4 62 e1 98 cc       	vfmadd132pd xmm9,xmm3,xmm4
  1e:	c4 62 e1 98 d4       	vfmadd132pd xmm10,xmm3,xmm4
  23:	c4 62 e1 98 dc       	vfmadd132pd xmm11,xmm3,xmm4
  28:	c4 62 e1 98 e4       	vfmadd132pd xmm12,xmm3,xmm4
  2d:	c4 e2 e1 98 d4       	vfmadd132pd xmm2,xmm3,xmm4
  32:	c4 e2 e1 98 ec       	vfmadd132pd xmm5,xmm3,xmm4
  37:	c4 e2 e1 98 f4       	vfmadd132pd xmm6,xmm3,xmm4
  3c:	c4 e2 e1 98 fc       	vfmadd132pd xmm7,xmm3,xmm4
  41:	c4 62 e1 98 c4       	vfmadd132pd xmm8,xmm3,xmm4
  46:	c4 62 e1 98 cc       	vfmadd132pd xmm9,xmm3,xmm4
  4b:	c4 62 e1 98 d4       	vfmadd132pd xmm10,xmm3,xmm4
  50:	c4 62 e1 98 dc       	vfmadd132pd xmm11,xmm3,xmm4
  55:	c4 62 e1 98 e4       	vfmadd132pd xmm12,xmm3,xmm4
  5a:	c4 e2 e1 98 d4       	vfmadd132pd xmm2,xmm3,xmm4
  5f:	c4 e2 e1 98 ec       	vfmadd132pd xmm5,xmm3,xmm4
  64:	c4 c2 75 a8 06       	vfmadd213ps ymm0,ymm1,YMMWORD PTR [r14]

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 XMM2, [R14];
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];
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM0, [R14];
VMOVUPD YMM1, [R14]

Show nanoBench command
Results:
- Instructions retired: 21.0
- Core cycles: 15.12
- Reference cycles: 15.93
- UOPS_PORT_1: 10.55
- UOPS_PORT_0: 10.45

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

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

Code:

   0:	45 8b 46 40          	mov    r8d,DWORD PTR [r14+0x40]
   4:	45 8b 4e 44          	mov    r9d,DWORD PTR [r14+0x44]
   8:	45 8b 56 48          	mov    r10d,DWORD PTR [r14+0x48]
   c:	45 8b 5e 4c          	mov    r11d,DWORD PTR [r14+0x4c]
  10:	45 8b 66 50          	mov    r12d,DWORD PTR [r14+0x50]
  14:	41 8b 5e 54          	mov    ebx,DWORD PTR [r14+0x54]
  18:	41 8b 4e 58          	mov    ecx,DWORD PTR [r14+0x58]
  1c:	41 8b 56 5c          	mov    edx,DWORD PTR [r14+0x5c]
  20:	45 8b 46 40          	mov    r8d,DWORD PTR [r14+0x40]
  24:	45 8b 4e 44          	mov    r9d,DWORD PTR [r14+0x44]
  28:	45 8b 56 48          	mov    r10d,DWORD PTR [r14+0x48]
  2c:	45 8b 5e 4c          	mov    r11d,DWORD PTR [r14+0x4c]
  30:	45 8b 66 50          	mov    r12d,DWORD PTR [r14+0x50]
  34:	41 8b 5e 54          	mov    ebx,DWORD PTR [r14+0x54]
  38:	41 8b 4e 58          	mov    ecx,DWORD PTR [r14+0x58]
  3c:	41 8b 56 5c          	mov    edx,DWORD PTR [r14+0x5c]
  40:	45 8b 46 40          	mov    r8d,DWORD PTR [r14+0x40]
  44:	45 8b 4e 44          	mov    r9d,DWORD PTR [r14+0x44]
  48:	45 8b 56 48          	mov    r10d,DWORD PTR [r14+0x48]
  4c:	45 8b 5e 4c          	mov    r11d,DWORD PTR [r14+0x4c]
  50:	c4 c2 75 a8 06       	vfmadd213ps ymm0,ymm1,YMMWORD PTR [r14]

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;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM0, [R14];
VMOVUPD YMM1, [R14]

Show nanoBench command
Results:
- Instructions retired: 21.0
- Core cycles: 10.78
- Reference cycles: 11.2
- UOPS_PORT_2: 10.5
- UOPS_PORT_3: 10.5

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

With an indexed addressing mode

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

Code:

   0:	c4 e2 e1 98 d4       	vfmadd132pd xmm2,xmm3,xmm4
   5:	c4 e2 e1 98 ec       	vfmadd132pd xmm5,xmm3,xmm4
   a:	c4 e2 e1 98 f4       	vfmadd132pd xmm6,xmm3,xmm4
   f:	c4 e2 e1 98 fc       	vfmadd132pd xmm7,xmm3,xmm4
  14:	c4 62 e1 98 c4       	vfmadd132pd xmm8,xmm3,xmm4
  19:	c4 62 e1 98 cc       	vfmadd132pd xmm9,xmm3,xmm4
  1e:	c4 62 e1 98 d4       	vfmadd132pd xmm10,xmm3,xmm4
  23:	c4 62 e1 98 dc       	vfmadd132pd xmm11,xmm3,xmm4
  28:	c4 62 e1 98 e4       	vfmadd132pd xmm12,xmm3,xmm4
  2d:	c4 e2 e1 98 d4       	vfmadd132pd xmm2,xmm3,xmm4
  32:	c4 e2 e1 98 ec       	vfmadd132pd xmm5,xmm3,xmm4
  37:	c4 e2 e1 98 f4       	vfmadd132pd xmm6,xmm3,xmm4
  3c:	c4 e2 e1 98 fc       	vfmadd132pd xmm7,xmm3,xmm4
  41:	c4 62 e1 98 c4       	vfmadd132pd xmm8,xmm3,xmm4
  46:	c4 62 e1 98 cc       	vfmadd132pd xmm9,xmm3,xmm4
  4b:	c4 62 e1 98 d4       	vfmadd132pd xmm10,xmm3,xmm4
  50:	c4 62 e1 98 dc       	vfmadd132pd xmm11,xmm3,xmm4
  55:	c4 62 e1 98 e4       	vfmadd132pd xmm12,xmm3,xmm4
  5a:	c4 e2 e1 98 d4       	vfmadd132pd xmm2,xmm3,xmm4
  5f:	c4 e2 e1 98 ec       	vfmadd132pd xmm5,xmm3,xmm4
  64:	c4 82 75 a8 04 2e    	vfmadd213ps ymm0,ymm1,YMMWORD PTR [r14+r13*1]

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 XMM2, [R14];
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];
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM0, [R14];
VMOVUPD YMM1, [R14];
XOR R13, R13

Show nanoBench command
Results:
- Instructions retired: 21.0
- Core cycles: 15.1
- Reference cycles: 15.93
- UOPS_PORT_1: 10.22
- UOPS_PORT_0: 10.76

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

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

Code:

   0:	45 8b 46 40          	mov    r8d,DWORD PTR [r14+0x40]
   4:	45 8b 4e 44          	mov    r9d,DWORD PTR [r14+0x44]
   8:	45 8b 56 48          	mov    r10d,DWORD PTR [r14+0x48]
   c:	45 8b 5e 4c          	mov    r11d,DWORD PTR [r14+0x4c]
  10:	45 8b 66 50          	mov    r12d,DWORD PTR [r14+0x50]
  14:	41 8b 5e 54          	mov    ebx,DWORD PTR [r14+0x54]
  18:	41 8b 4e 58          	mov    ecx,DWORD PTR [r14+0x58]
  1c:	41 8b 56 5c          	mov    edx,DWORD PTR [r14+0x5c]
  20:	45 8b 46 40          	mov    r8d,DWORD PTR [r14+0x40]
  24:	45 8b 4e 44          	mov    r9d,DWORD PTR [r14+0x44]
  28:	45 8b 56 48          	mov    r10d,DWORD PTR [r14+0x48]
  2c:	45 8b 5e 4c          	mov    r11d,DWORD PTR [r14+0x4c]
  30:	45 8b 66 50          	mov    r12d,DWORD PTR [r14+0x50]
  34:	41 8b 5e 54          	mov    ebx,DWORD PTR [r14+0x54]
  38:	41 8b 4e 58          	mov    ecx,DWORD PTR [r14+0x58]
  3c:	41 8b 56 5c          	mov    edx,DWORD PTR [r14+0x5c]
  40:	45 8b 46 40          	mov    r8d,DWORD PTR [r14+0x40]
  44:	45 8b 4e 44          	mov    r9d,DWORD PTR [r14+0x44]
  48:	45 8b 56 48          	mov    r10d,DWORD PTR [r14+0x48]
  4c:	45 8b 5e 4c          	mov    r11d,DWORD PTR [r14+0x4c]
  50:	c4 82 75 a8 04 2e    	vfmadd213ps ymm0,ymm1,YMMWORD PTR [r14+r13*1]

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;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM0, [R14];
VMOVUPD YMM1, [R14];
XOR R13, R13

Show nanoBench command
Results:
- Instructions retired: 21.0
- Core cycles: 10.78
- Reference cycles: 11.6
- UOPS_PORT_2: 10.5
- UOPS_PORT_3: 10.5

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