ADDSD (XMM, M64) - Port Usage

With blocking instructions for port '23A':

Code:

   0:	f2 41 0f f0 4e 40    	lddqu  xmm1,[r14+0x40]
   6:	f2 41 0f f0 56 50    	lddqu  xmm2,[r14+0x50]
   c:	f2 41 0f f0 5e 60    	lddqu  xmm3,[r14+0x60]
  12:	f2 41 0f f0 66 70    	lddqu  xmm4,[r14+0x70]
  18:	f2 41 0f f0 ae 80 00 	lddqu  xmm5,[r14+0x80]
  1f:	00 00 
  21:	f2 41 0f f0 b6 90 00 	lddqu  xmm6,[r14+0x90]
  28:	00 00 
  2a:	f2 41 0f f0 be a0 00 	lddqu  xmm7,[r14+0xa0]
  31:	00 00 
  33:	f2 45 0f f0 86 b0 00 	lddqu  xmm8,[r14+0xb0]
  3a:	00 00 
  3c:	f2 45 0f f0 8e c0 00 	lddqu  xmm9,[r14+0xc0]
  43:	00 00 
  45:	f2 45 0f f0 96 d0 00 	lddqu  xmm10,[r14+0xd0]
  4c:	00 00 
  4e:	f2 45 0f f0 9e e0 00 	lddqu  xmm11,[r14+0xe0]
  55:	00 00 
  57:	f2 45 0f f0 a6 f0 00 	lddqu  xmm12,[r14+0xf0]
  5e:	00 00 
  60:	f2 41 0f 58 06       	addsd  xmm0,QWORD PTR [r14]

Init:

VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOVUPD XMM0, [R14]

Show nanoBench command
Results:
- Instructions retired: 13.0
- Core cycles: 4.48
- Reference cycles: 3.28
- UOPS_PORT_23A: 13.0

⇨ One μop that can only use port '23A'

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

Code:

   0:	66 0f 58 ca          	addpd  xmm1,xmm2
   4:	66 0f 58 da          	addpd  xmm3,xmm2
   8:	66 0f 58 e2          	addpd  xmm4,xmm2
   c:	66 0f 58 ea          	addpd  xmm5,xmm2
  10:	66 0f 58 f2          	addpd  xmm6,xmm2
  14:	66 0f 58 fa          	addpd  xmm7,xmm2
  18:	66 44 0f 58 c2       	addpd  xmm8,xmm2
  1d:	66 44 0f 58 ca       	addpd  xmm9,xmm2
  22:	66 44 0f 58 d2       	addpd  xmm10,xmm2
  27:	66 44 0f 58 da       	addpd  xmm11,xmm2
  2c:	f2 41 0f 58 06       	addsd  xmm0,QWORD PTR [r14]

Init:

VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOVUPD XMM1, [R14];
MOVUPD XMM2, [R14];
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];
MOVUPD XMM0, [R14]

Show nanoBench command
Results:
- Instructions retired: 11.0
- Core cycles: 5.55
- Reference cycles: 4.16
- UOPS_PORT_1: 5.49
- UOPS_PORT_5B: 5.51
- UOPS_PORT_5B>=2: 0.0

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

With an indexed addressing mode

With blocking instructions for port '23A':

Code:

   0:	f2 41 0f f0 4e 40    	lddqu  xmm1,[r14+0x40]
   6:	f2 41 0f f0 56 50    	lddqu  xmm2,[r14+0x50]
   c:	f2 41 0f f0 5e 60    	lddqu  xmm3,[r14+0x60]
  12:	f2 41 0f f0 66 70    	lddqu  xmm4,[r14+0x70]
  18:	f2 41 0f f0 ae 80 00 	lddqu  xmm5,[r14+0x80]
  1f:	00 00 
  21:	f2 41 0f f0 b6 90 00 	lddqu  xmm6,[r14+0x90]
  28:	00 00 
  2a:	f2 41 0f f0 be a0 00 	lddqu  xmm7,[r14+0xa0]
  31:	00 00 
  33:	f2 45 0f f0 86 b0 00 	lddqu  xmm8,[r14+0xb0]
  3a:	00 00 
  3c:	f2 45 0f f0 8e c0 00 	lddqu  xmm9,[r14+0xc0]
  43:	00 00 
  45:	f2 45 0f f0 96 d0 00 	lddqu  xmm10,[r14+0xd0]
  4c:	00 00 
  4e:	f2 45 0f f0 9e e0 00 	lddqu  xmm11,[r14+0xe0]
  55:	00 00 
  57:	f2 45 0f f0 a6 f0 00 	lddqu  xmm12,[r14+0xf0]
  5e:	00 00 
  60:	f2 43 0f 58 04 2e    	addsd  xmm0,QWORD PTR [r14+r13*1]

Init:

VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOVUPD XMM0, [R14];
XOR R13, R13

Show nanoBench command
Results:
- Instructions retired: 13.0
- Core cycles: 4.46
- Reference cycles: 3.47
- UOPS_PORT_23A: 13.0

⇨ One μop that can only use port '23A'