VSCATTERQPD (VSIB_ZMM, K, ZMM) - Throughput and Uops

With unroll_count=500 and no inner loop

• Code:

     0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
  

• Init:

  MOV R15, 10000;
  L: VADDPS ZMM0, ZMM1, ZMM1;
  VADDPS ZMM0, ZMM1, ZMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x4000000040000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  MOV [R14+16], RAX;
  MOV [R14+24], RAX;
  MOV [R14+32], RAX;
  MOV [R14+40], RAX;
  MOV [R14+48], RAX;
  MOV [R14+56], RAX;
  VMOVUPD ZMM0, [R14];
  VXORPS ZMM14, ZMM14, ZMM14

• Show nanoBench command
• Results:
  • Instructions retired: 1.0
  • Core cycles: 7.0
  • Reference cycles: 4.33
  • UOPS_EXECUTED.THREAD: 19.0
  • RETIRE_SLOTS: 20.0
  • UOPS_MITE: 4.0
  • UOPS_MS: 16.0
  • UOPS_PORT_0: 2.0
  • UOPS_PORT_1: 0.0
  • UOPS_PORT_5: 1.0
  • UOPS_PORT_6: 0.0
  • UOPS_PORT_23: 0.0
  • UOPS_PORT_49: 8.0
  • UOPS_PORT_78: 8.0
  • DIV_CYCLES: 0.0
  • ILD_STALL.LCP: 0.0
  • UOPS_MITE>=1: 1.0

With unroll_count=500, no inner loop, and 1 NOP

• Code:

     0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
     7:	90                   	nop
  

• Init:

  MOV R15, 10000;
  L: VADDPS ZMM0, ZMM1, ZMM1;
  VADDPS ZMM0, ZMM1, ZMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x4000000040000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  MOV [R14+16], RAX;
  MOV [R14+24], RAX;
  MOV [R14+32], RAX;
  MOV [R14+40], RAX;
  MOV [R14+48], RAX;
  MOV [R14+56], RAX;
  VMOVUPD ZMM0, [R14];
  VXORPS ZMM14, ZMM14, ZMM14

• Show nanoBench command
• Results:
  • Instructions retired: 2.0
  • Core cycles: 8.0
  • Reference cycles: 4.89
  • UOPS_EXECUTED.THREAD: 19.0
  • RETIRE_SLOTS: 21.0
  • UOPS_MITE: 5.0
  • UOPS_MS: 16.0
  • UOPS_PORT_0: 2.0
  • UOPS_PORT_1: 0.0
  • UOPS_PORT_5: 1.0
  • UOPS_PORT_6: 0.0
  • UOPS_PORT_23: 0.0
  • UOPS_PORT_49: 8.0
  • UOPS_PORT_78: 8.0
  • DIV_CYCLES: 0.0
  • ILD_STALL.LCP: 0.0
  • UOPS_MITE>=1: 2.0

With loop_count=1000 and unroll_count=10

• Code:

     0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
  

• Init:

  MOV R15, 10000;
  L: VADDPS ZMM0, ZMM1, ZMM1;
  VADDPS ZMM0, ZMM1, ZMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x4000000040000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  MOV [R14+16], RAX;
  MOV [R14+24], RAX;
  MOV [R14+32], RAX;
  MOV [R14+40], RAX;
  MOV [R14+48], RAX;
  MOV [R14+56], RAX;
  VMOVUPD ZMM0, [R14];
  VXORPS ZMM14, ZMM14, ZMM14

• Show nanoBench command
• Results:
  • Instructions retired: 1.2
  • Core cycles: 7.1
  • Reference cycles: 4.34
  • UOPS_EXECUTED.THREAD: 19.1
  • RETIRE_SLOTS: 20.1
  • UOPS_MITE: 4.1
  • UOPS_MS: 16.0
  • UOPS_PORT_0: 2.0
  • UOPS_PORT_1: 0.0
  • UOPS_PORT_5: 1.0
  • UOPS_PORT_6: 0.1
  • UOPS_PORT_23: 0.0
  • UOPS_PORT_49: 8.0
  • UOPS_PORT_78: 8.0
  • DIV_CYCLES: 0.0
  • ILD_STALL.LCP: 0.0
  • UOPS_MITE>=1: 1.1

With loop_count=100 and unroll_count=100

• Code:

     0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
  

• Init:

  MOV R15, 10000;
  L: VADDPS ZMM0, ZMM1, ZMM1;
  VADDPS ZMM0, ZMM1, ZMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x4000000040000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  MOV [R14+16], RAX;
  MOV [R14+24], RAX;
  MOV [R14+32], RAX;
  MOV [R14+40], RAX;
  MOV [R14+48], RAX;
  MOV [R14+56], RAX;
  VMOVUPD ZMM0, [R14];
  VXORPS ZMM14, ZMM14, ZMM14

• Show nanoBench command
• Results:
  • Instructions retired: 1.02
  • Core cycles: 7.01
  • Reference cycles: 4.28
  • UOPS_EXECUTED.THREAD: 19.01
  • RETIRE_SLOTS: 20.01
  • UOPS_MITE: 4.01
  • UOPS_MS: 16.0
  • UOPS_PORT_0: 2.0
  • UOPS_PORT_1: 0.0
  • UOPS_PORT_5: 1.0
  • UOPS_PORT_6: 0.01
  • UOPS_PORT_23: 0.0
  • UOPS_PORT_49: 8.0
  • UOPS_PORT_78: 8.0
  • DIV_CYCLES: 0.0
  • ILD_STALL.LCP: 0.0
  • UOPS_MITE>=1: 1.01

With unroll_count=200 and no inner loop

• Code:

     0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
     7:	62 92 fd 4a a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x8]{k2},zmm0
     e:	01 
     f:	62 92 fd 4b a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x10]{k3},zmm0
    16:	02 
    17:	62 92 fd 4c a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x18]{k4},zmm0
    1e:	03 
  

• Init:

  MOV R15, 10000;
  L: VADDPS ZMM0, ZMM1, ZMM1;
  VADDPS ZMM0, ZMM1, ZMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x4000000040000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  MOV [R14+16], RAX;
  MOV [R14+24], RAX;
  MOV [R14+32], RAX;
  MOV [R14+40], RAX;
  MOV [R14+48], RAX;
  MOV [R14+56], RAX;
  VMOVUPD ZMM0, [R14];
  VXORPS ZMM14, ZMM14, ZMM14

• Show nanoBench command
• Results:
  • Instructions retired: 4.0
  • Core cycles: 28.0
  • Reference cycles: 17.09
  • UOPS_EXECUTED.THREAD: 76.0
  • RETIRE_SLOTS: 80.0
  • UOPS_MITE: 16.0
  • UOPS_MS: 64.0
  • UOPS_PORT_0: 8.0
  • UOPS_PORT_1: 0.0
  • UOPS_PORT_5: 4.0
  • UOPS_PORT_6: 0.0
  • UOPS_PORT_23: 0.0
  • UOPS_PORT_49: 32.0
  • UOPS_PORT_78: 32.0
  • DIV_CYCLES: 0.0
  • ILD_STALL.LCP: 0.0
  • UOPS_MITE>=1: 4.0

With loop_count=1000 and unroll_count=2

• Code:

     0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
     7:	62 92 fd 4a a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x8]{k2},zmm0
     e:	01 
     f:	62 92 fd 4b a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x10]{k3},zmm0
    16:	02 
    17:	62 92 fd 4c a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x18]{k4},zmm0
    1e:	03 
  

• Init:

  MOV R15, 10000;
  L: VADDPS ZMM0, ZMM1, ZMM1;
  VADDPS ZMM0, ZMM1, ZMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x4000000040000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  MOV [R14+16], RAX;
  MOV [R14+24], RAX;
  MOV [R14+32], RAX;
  MOV [R14+40], RAX;
  MOV [R14+48], RAX;
  MOV [R14+56], RAX;
  VMOVUPD ZMM0, [R14];
  VXORPS ZMM14, ZMM14, ZMM14

• Show nanoBench command
• Results:
  • Instructions retired: 5.0
  • Core cycles: 28.51
  • Reference cycles: 17.4
  • UOPS_EXECUTED.THREAD: 76.5
  • RETIRE_SLOTS: 80.5
  • UOPS_MITE: 16.51
  • UOPS_MS: 64.0
  • UOPS_PORT_0: 8.0
  • UOPS_PORT_1: 0.0
  • UOPS_PORT_5: 4.0
  • UOPS_PORT_6: 0.5
  • UOPS_PORT_23: 0.0
  • UOPS_PORT_49: 32.0
  • UOPS_PORT_78: 32.0
  • DIV_CYCLES: 0.0
  • ILD_STALL.LCP: 0.0
  • UOPS_MITE>=1: 4.5

With loop_count=100 and unroll_count=20

• Code:

     0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
     7:	62 92 fd 4a a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x8]{k2},zmm0
     e:	01 
     f:	62 92 fd 4b a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x10]{k3},zmm0
    16:	02 
    17:	62 92 fd 4c a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x18]{k4},zmm0
    1e:	03 
  

• Init:

  MOV R15, 10000;
  L: VADDPS ZMM0, ZMM1, ZMM1;
  VADDPS ZMM0, ZMM1, ZMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x4000000040000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  MOV [R14+16], RAX;
  MOV [R14+24], RAX;
  MOV [R14+32], RAX;
  MOV [R14+40], RAX;
  MOV [R14+48], RAX;
  MOV [R14+56], RAX;
  VMOVUPD ZMM0, [R14];
  VXORPS ZMM14, ZMM14, ZMM14

• Show nanoBench command
• Results:
  • Instructions retired: 4.1
  • Core cycles: 28.05
  • Reference cycles: 17.13
  • UOPS_EXECUTED.THREAD: 76.05
  • RETIRE_SLOTS: 80.05
  • UOPS_MITE: 16.05
  • UOPS_MS: 64.0
  • UOPS_PORT_0: 8.0
  • UOPS_PORT_1: 0.0
  • UOPS_PORT_5: 4.0
  • UOPS_PORT_6: 0.05
  • UOPS_PORT_23: 0.0
  • UOPS_PORT_49: 32.0
  • UOPS_PORT_78: 32.0
  • DIV_CYCLES: 0.0
  • ILD_STALL.LCP: 0.0
  • UOPS_MITE>=1: 4.05

With unroll_count=100 and no inner loop

• Code:

     0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
     7:	62 92 fd 4a a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x8]{k2},zmm0
     e:	01 
     f:	62 92 fd 4b a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x10]{k3},zmm0
    16:	02 
    17:	62 92 fd 4c a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x18]{k4},zmm0
    1e:	03 
    1f:	62 92 fd 4d a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x20]{k5},zmm0
    26:	04 
    27:	62 92 fd 4e a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x28]{k6},zmm0
    2e:	05 
    2f:	62 92 fd 4f a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x30]{k7},zmm0
    36:	06 
  

• Init:

  MOV R15, 10000;
  L: VADDPS ZMM0, ZMM1, ZMM1;
  VADDPS ZMM0, ZMM1, ZMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x4000000040000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  MOV [R14+16], RAX;
  MOV [R14+24], RAX;
  MOV [R14+32], RAX;
  MOV [R14+40], RAX;
  MOV [R14+48], RAX;
  MOV [R14+56], RAX;
  VMOVUPD ZMM0, [R14];
  VXORPS ZMM14, ZMM14, ZMM14

• Show nanoBench command
• Results:
  • Instructions retired: 7.0
  • Core cycles: 49.0
  • Reference cycles: 29.93
  • UOPS_EXECUTED.THREAD: 133.0
  • RETIRE_SLOTS: 140.0
  • UOPS_MITE: 28.0
  • UOPS_MS: 112.0
  • UOPS_PORT_0: 14.0
  • UOPS_PORT_1: 0.0
  • UOPS_PORT_5: 7.0
  • UOPS_PORT_6: 0.0
  • UOPS_PORT_23: 0.0
  • UOPS_PORT_49: 56.0
  • UOPS_PORT_78: 56.0
  • DIV_CYCLES: 0.0
  • ILD_STALL.LCP: 0.0
  • UOPS_MITE>=1: 7.0

With loop_count=1000 and unroll_count=1

• Code:

     0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
     7:	62 92 fd 4a a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x8]{k2},zmm0
     e:	01 
     f:	62 92 fd 4b a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x10]{k3},zmm0
    16:	02 
    17:	62 92 fd 4c a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x18]{k4},zmm0
    1e:	03 
    1f:	62 92 fd 4d a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x20]{k5},zmm0
    26:	04 
    27:	62 92 fd 4e a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x28]{k6},zmm0
    2e:	05 
    2f:	62 92 fd 4f a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x30]{k7},zmm0
    36:	06 
  

• Init:

  MOV R15, 10000;
  L: VADDPS ZMM0, ZMM1, ZMM1;
  VADDPS ZMM0, ZMM1, ZMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x4000000040000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  MOV [R14+16], RAX;
  MOV [R14+24], RAX;
  MOV [R14+32], RAX;
  MOV [R14+40], RAX;
  MOV [R14+48], RAX;
  MOV [R14+56], RAX;
  VMOVUPD ZMM0, [R14];
  VXORPS ZMM14, ZMM14, ZMM14

• Show nanoBench command
• Results:
  • Instructions retired: 9.0
  • Core cycles: 50.01
  • Reference cycles: 30.55
  • UOPS_EXECUTED.THREAD: 134.01
  • RETIRE_SLOTS: 141.0
  • UOPS_MITE: 29.02
  • UOPS_MS: 112.01
  • UOPS_PORT_0: 14.0
  • UOPS_PORT_1: 0.0
  • UOPS_PORT_5: 7.0
  • UOPS_PORT_6: 1.0
  • UOPS_PORT_23: 0.0
  • UOPS_PORT_49: 56.0
  • UOPS_PORT_78: 56.0
  • DIV_CYCLES: 0.0
  • ILD_STALL.LCP: 0.0
  • UOPS_MITE>=1: 8.0

With loop_count=100 and unroll_count=10

• Code:

     0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
     7:	62 92 fd 4a a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x8]{k2},zmm0
     e:	01 
     f:	62 92 fd 4b a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x10]{k3},zmm0
    16:	02 
    17:	62 92 fd 4c a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x18]{k4},zmm0
    1e:	03 
    1f:	62 92 fd 4d a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x20]{k5},zmm0
    26:	04 
    27:	62 92 fd 4e a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x28]{k6},zmm0
    2e:	05 
    2f:	62 92 fd 4f a3 44 36 	vscatterqpd QWORD PTR [r14+zmm14*1+0x30]{k7},zmm0
    36:	06 
  

• Init:

  MOV R15, 10000;
  L: VADDPS ZMM0, ZMM1, ZMM1;
  VADDPS ZMM0, ZMM1, ZMM1;
  DEC R15;
  JNZ L;
  VZEROALL;
  MOV RAX, 0x4000000040000000;
  MOV [R14+0], RAX;
  MOV [R14+8], RAX;
  MOV [R14+16], RAX;
  MOV [R14+24], RAX;
  MOV [R14+32], RAX;
  MOV [R14+40], RAX;
  MOV [R14+48], RAX;
  MOV [R14+56], RAX;
  VMOVUPD ZMM0, [R14];
  VXORPS ZMM14, ZMM14, ZMM14

• Show nanoBench command
• Results:
  • Instructions retired: 7.2
  • Core cycles: 49.1
  • Reference cycles: 29.98
  • UOPS_EXECUTED.THREAD: 133.1
  • RETIRE_SLOTS: 140.1
  • UOPS_MITE: 28.1
  • UOPS_MS: 111.99
  • UOPS_PORT_0: 14.0
  • UOPS_PORT_1: 0.0
  • UOPS_PORT_5: 7.0
  • UOPS_PORT_6: 0.1
  • UOPS_PORT_23: 0.0
  • UOPS_PORT_49: 56.0
  • UOPS_PORT_78: 56.0
  • DIV_CYCLES: 0.0
  • ILD_STALL.LCP: 0.0
  • UOPS_MITE>=1: 7.1