VSHUFPS (YMM, YMM, M256, I8) - Port Usage

With blocking instructions for port '23':

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:	c4 c1 74 c6 06 02    	vshufps ymm0,ymm1,YMMWORD PTR [r14],0x2

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 YMM1, [R14]

Show nanoBench command
Results:
- Instructions retired: 11.0
- Core cycles: 5.65
- Reference cycles: 3.12
- UOPS_PORT_23: 11.0

⇨ One μop that can only use port '23'

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

Code:

   0:	47 8d 04 6e          	lea    r8d,[r14+r13*2]
   4:	47 8d 0c 6e          	lea    r9d,[r14+r13*2]
   8:	47 8d 14 6e          	lea    r10d,[r14+r13*2]
   c:	47 8d 1c 6e          	lea    r11d,[r14+r13*2]
  10:	47 8d 24 6e          	lea    r12d,[r14+r13*2]
  14:	43 8d 1c 6e          	lea    ebx,[r14+r13*2]
  18:	43 8d 0c 6e          	lea    ecx,[r14+r13*2]
  1c:	43 8d 14 6e          	lea    edx,[r14+r13*2]
  20:	47 8d 04 6e          	lea    r8d,[r14+r13*2]
  24:	47 8d 0c 6e          	lea    r9d,[r14+r13*2]
  28:	c4 c1 74 c6 06 02    	vshufps ymm0,ymm1,YMMWORD PTR [r14],0x2

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 YMM1, [R14]

Show nanoBench command
Results:
- Instructions retired: 11.0
- Core cycles: 5.58
- Reference cycles: 2.77
- UOPS_PORT_5: 5.5
- UOPS_PORT_1: 5.5

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

With an indexed addressing mode

With blocking instructions for port '23':

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:	c4 81 74 c6 04 2e 02 	vshufps ymm0,ymm1,YMMWORD PTR [r14+r13*1],0x2

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 YMM1, [R14];
XOR R13, R13

Show nanoBench command
Results:
- Instructions retired: 11.0
- Core cycles: 5.65
- Reference cycles: 2.6
- UOPS_PORT_23: 11.0

⇨ One μop that can only use port '23'

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

Code:

   0:	47 8d 04 6e          	lea    r8d,[r14+r13*2]
   4:	47 8d 0c 6e          	lea    r9d,[r14+r13*2]
   8:	47 8d 14 6e          	lea    r10d,[r14+r13*2]
   c:	47 8d 1c 6e          	lea    r11d,[r14+r13*2]
  10:	47 8d 24 6e          	lea    r12d,[r14+r13*2]
  14:	43 8d 1c 6e          	lea    ebx,[r14+r13*2]
  18:	43 8d 0c 6e          	lea    ecx,[r14+r13*2]
  1c:	43 8d 14 6e          	lea    edx,[r14+r13*2]
  20:	47 8d 04 6e          	lea    r8d,[r14+r13*2]
  24:	47 8d 0c 6e          	lea    r9d,[r14+r13*2]
  28:	c4 81 74 c6 04 2e 02 	vshufps ymm0,ymm1,YMMWORD PTR [r14+r13*1],0x2

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 YMM1, [R14];
XOR R13, R13

Show nanoBench command
Results:
- Instructions retired: 11.0
- Core cycles: 5.58
- Reference cycles: 2.77
- UOPS_PORT_5: 5.5
- UOPS_PORT_1: 5.5

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