MOVSB - Latency

Experiment 1

• Instruction: MOVSB
• Chain instructions: MOV R12B, [RDI];MOVSX R12, R12B;MOVSX R12, R12B;MOVSX R12, R12B;MOVSX R12, R12B;MOVSX R12, R12B;MOVSX R12, R12B;XOR RDI, R12; XOR RDI, R12;TEST R15, R15
• Chain latency: ≥9
• Code:

     0:	a4                   	movs   BYTE PTR es:[rdi],BYTE PTR ds:[rsi]
     1:	44 8a 27             	mov    r12b,BYTE PTR [rdi]
     4:	4d 0f be e4          	movsx  r12,r12b
     8:	4d 0f be e4          	movsx  r12,r12b
     c:	4d 0f be e4          	movsx  r12,r12b
    10:	4d 0f be e4          	movsx  r12,r12b
    14:	4d 0f be e4          	movsx  r12,r12b
    18:	4d 0f be e4          	movsx  r12,r12b
    1c:	4c 31 e7             	xor    rdi,r12
    1f:	4c 31 e7             	xor    rdi,r12
    22:	4d 85 ff             	test   r15,r15
  

• Show nanoBench command
• Results:
  • Instructions retired: 11.0
  • Core cycles: 15.06
  • Reference cycles: 15.83
  • UOPS_EXECUTED.THREAD: 16.0

Experiment 2 (with additional nop)

• Instruction: MOVSB
• Chain instructions: MOV R12B, [RDI];MOVSX R12, R12B;MOVSX R12, R12B;MOVSX R12, R12B;MOVSX R12, R12B;MOVSX R12, R12B;MOVSX R12, R12B;XOR RDI, R12; XOR RDI, R12;TEST R15, R15;nop
• Chain latency: ≥9
• Code:

     0:	a4                   	movs   BYTE PTR es:[rdi],BYTE PTR ds:[rsi]
     1:	44 8a 27             	mov    r12b,BYTE PTR [rdi]
     4:	4d 0f be e4          	movsx  r12,r12b
     8:	4d 0f be e4          	movsx  r12,r12b
     c:	4d 0f be e4          	movsx  r12,r12b
    10:	4d 0f be e4          	movsx  r12,r12b
    14:	4d 0f be e4          	movsx  r12,r12b
    18:	4d 0f be e4          	movsx  r12,r12b
    1c:	4c 31 e7             	xor    rdi,r12
    1f:	4c 31 e7             	xor    rdi,r12
    22:	4d 85 ff             	test   r15,r15
    25:	90                   	nop
  

• Show nanoBench command
• Results:
  • Instructions retired: 12.0
  • Core cycles: 15.05
  • Reference cycles: 15.8
  • UOPS_EXECUTED.THREAD: 16.0