VSCATTERQPD (VSIB_ZMM, K, ZMM)

Experiment 1

Instruction: VSCATTERQPD [R14+ZMM14] {K1}, ZMM0
Chain instructions: MOV R12, [R14];MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;XOR R14, R12; XOR R14, R12;VPCMPD K1, XMM15, XMM15, 7
Chain latency: ≥14

Code:

   0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
   7:	4d 8b 26             	mov    r12,QWORD PTR [r14]
   a:	4d 63 e4             	movsxd r12,r12d
   d:	4d 63 e4             	movsxd r12,r12d
  10:	4d 63 e4             	movsxd r12,r12d
  13:	4d 63 e4             	movsxd r12,r12d
  16:	4d 63 e4             	movsxd r12,r12d
  19:	4d 63 e4             	movsxd r12,r12d
  1c:	4d 63 e4             	movsxd r12,r12d
  1f:	4d 63 e4             	movsxd r12,r12d
  22:	4d 63 e4             	movsxd r12,r12d
  25:	4d 63 e4             	movsxd r12,r12d
  28:	4d 63 e4             	movsxd r12,r12d
  2b:	4d 63 e4             	movsxd r12,r12d
  2e:	4d 31 e6             	xor    r14,r12
  31:	4d 31 e6             	xor    r14,r12
  34:	62 d3 05 08 1f cf 07 	vpcmpd k1,xmm15,xmm15,0x7

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:
- MPERF: 55.1
- APERF: 80.53
- UOPS: 64.0

Experiment 2 (with additional nop)

Instruction: VSCATTERQPD [R14+ZMM14] {K1}, ZMM0
Chain instructions: MOV R12, [R14];MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;MOVSX R12, R12D;XOR R14, R12; XOR R14, R12;nop;VPCMPD K1, XMM15, XMM15, 7
Chain latency: ≥14

Code:

   0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
   7:	4d 8b 26             	mov    r12,QWORD PTR [r14]
   a:	4d 63 e4             	movsxd r12,r12d
   d:	4d 63 e4             	movsxd r12,r12d
  10:	4d 63 e4             	movsxd r12,r12d
  13:	4d 63 e4             	movsxd r12,r12d
  16:	4d 63 e4             	movsxd r12,r12d
  19:	4d 63 e4             	movsxd r12,r12d
  1c:	4d 63 e4             	movsxd r12,r12d
  1f:	4d 63 e4             	movsxd r12,r12d
  22:	4d 63 e4             	movsxd r12,r12d
  25:	4d 63 e4             	movsxd r12,r12d
  28:	4d 63 e4             	movsxd r12,r12d
  2b:	4d 63 e4             	movsxd r12,r12d
  2e:	4d 31 e6             	xor    r14,r12
  31:	4d 31 e6             	xor    r14,r12
  34:	90                   	nop
  35:	62 d3 05 08 1f cf 07 	vpcmpd k1,xmm15,xmm15,0x7

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:
- MPERF: 55.16
- APERF: 80.73
- UOPS: 65.0

Latency operand 2 → 2: 0

Experiment 1

Instruction: VSCATTERQPD [R14+ZMM14] {K1}, ZMM0

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:
- MPERF: 6.08
- APERF: 9.0
- UOPS: 48.0

Experiment 2

Instruction: VSCATTERQPD [R14+ZMM14] {K1}, ZMM0
Chain instructions: KMOVQ K1, K1;KMOVQ K1, K1;KMOVQ K1, K1;KMOVQ K1, K1;KMOVQ K1, K1;KMOVQ K1, K1;KMOVQ K1, K1;KMOVQ K1, K1;KMOVQ K1, K1;KMOVQ K1, K1;KMOVQ K1, K1;KMOVQ K1, K1;KMOVQ K1, K1
Chain latency: 13

Code:

   0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
   7:	c4 e1 f8 90 c9       	kmovq  k1,k1
   c:	c4 e1 f8 90 c9       	kmovq  k1,k1
  11:	c4 e1 f8 90 c9       	kmovq  k1,k1
  16:	c4 e1 f8 90 c9       	kmovq  k1,k1
  1b:	c4 e1 f8 90 c9       	kmovq  k1,k1
  20:	c4 e1 f8 90 c9       	kmovq  k1,k1
  25:	c4 e1 f8 90 c9       	kmovq  k1,k1
  2a:	c4 e1 f8 90 c9       	kmovq  k1,k1
  2f:	c4 e1 f8 90 c9       	kmovq  k1,k1
  34:	c4 e1 f8 90 c9       	kmovq  k1,k1
  39:	c4 e1 f8 90 c9       	kmovq  k1,k1
  3e:	c4 e1 f8 90 c9       	kmovq  k1,k1
  43:	c4 e1 f8 90 c9       	kmovq  k1,k1

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:
- MPERF: 7.6
- APERF: 11.38
- UOPS: 61.0

Latency operand 3 → 1: ≤38

Experiment 1

Instruction: VSCATTERQPD [R14+ZMM14] {K1}, ZMM0
Chain instructions: VMOVSD XMM13, [R14];VANDPD XMM0, XMM13, XMM13;VANDPD XMM0, XMM0, XMM0;VANDPD XMM0, XMM0, XMM0;VANDPD XMM0, XMM0, XMM0;VANDPD XMM0, XMM0, XMM0;VANDPD XMM0, XMM0, XMM0;VANDPD XMM0, XMM0, XMM0;VANDPD XMM0, XMM0, XMM0;VANDPD XMM0, XMM0, XMM0;VANDPD XMM0, XMM0, XMM0;VANDPD XMM0, XMM0, XMM0;VANDPD XMM0, XMM0, XMM0;VANDPD XMM0, XMM0, XMM0;VPCMPD K1, XMM15, XMM15, 7
Chain latency: ≥27

Code:

   0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
   7:	c4 41 7b 10 2e       	vmovsd xmm13,QWORD PTR [r14]
   c:	c4 c1 11 54 c5       	vandpd xmm0,xmm13,xmm13
  11:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  15:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  19:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  1d:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  21:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  25:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  29:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  2d:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  31:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  35:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  39:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  3d:	c5 f9 54 c0          	vandpd xmm0,xmm0,xmm0
  41:	62 d3 05 08 1f cf 07 	vpcmpd k1,xmm15,xmm15,0x7

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:
- MPERF: 44.97
- APERF: 65.75
- UOPS: 63.0

Experiment 2

Instruction: VSCATTERQPD [R14+ZMM14] {K1}, ZMM0
Chain instructions: VMOVQ XMM13, [R14];VPAND XMM0, XMM13, XMM13;VPAND XMM0, XMM0, XMM0;VPAND XMM0, XMM0, XMM0;VPAND XMM0, XMM0, XMM0;VPAND XMM0, XMM0, XMM0;VPAND XMM0, XMM0, XMM0;VPAND XMM0, XMM0, XMM0;VPAND XMM0, XMM0, XMM0;VPAND XMM0, XMM0, XMM0;VPAND XMM0, XMM0, XMM0;VPAND XMM0, XMM0, XMM0;VPAND XMM0, XMM0, XMM0;VPAND XMM0, XMM0, XMM0;VPCMPD K1, XMM15, XMM15, 7
Chain latency: ≥27

Code:

   0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
   7:	c4 41 7a 7e 2e       	vmovq  xmm13,QWORD PTR [r14]
   c:	c4 c1 11 db c5       	vpand  xmm0,xmm13,xmm13
  11:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  15:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  19:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  1d:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  21:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  25:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  29:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  2d:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  31:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  35:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  39:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  3d:	c5 f9 db c0          	vpand  xmm0,xmm0,xmm0
  41:	62 d3 05 08 1f cf 07 	vpcmpd k1,xmm15,xmm15,0x7

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:
- MPERF: 44.84
- APERF: 65.75
- UOPS: 63.0

Latency operand 3 → 2: 7

Experiment 1

Instruction: VSCATTERQPD [R14+ZMM14] {K1}, ZMM0
Chain instructions: VMOVUPS ZMM0 {K1}, ZMM14;VPCMPD K1, XMM15, XMM15, 7
Chain latency: 2

Code:

   0:	62 92 fd 49 a3 04 36 	vscatterqpd QWORD PTR [r14+zmm14*1]{k1},zmm0
   7:	62 d1 7c 49 10 c6    	vmovups zmm0{k1},zmm14
   d:	62 d3 05 08 1f cf 07 	vpcmpd k1,xmm15,xmm15,0x7

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:
- MPERF: 6.14
- APERF: 9.21
- UOPS: 50.0

VSCATTERQPD (VSIB_ZMM, K, ZMM) - Latency

Operands

Latency operand 1 → 1 (address, base register): ≤66

Latency operand 2 → 2: 0

Latency operand 3 → 1: ≤38

Latency operand 3 → 2: 7

Latency operand 1 → 1 (address, base register): ≤66

Experiment 1

Experiment 2 (with additional nop)

Latency operand 2 → 2: 0

Experiment 1

Experiment 2

Latency operand 3 → 1: ≤38

Experiment 1

Experiment 2

Latency operand 3 → 2: 7

Experiment 1