VDPBF16PS (ZMM, ZMM, ZMM)

Experiment 1

Instruction: VDPBF16PS ZMM0, ZMM1, ZMM2

Code:

   0:	62 f2 76 48 52 c2    	vdpbf16ps zmm0,zmm1,zmm2

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 ZMM1, ZMM1, ZMM1;
VXORPS ZMM2, ZMM2, ZMM2

Show nanoBench command
Results:
- MPERF: 4.18
- APERF: 6.08
- UOPS: 1.0

Experiment 2

Instruction: VDPBF16PS ZMM0, ZMM1, ZMM2
Chain instructions: VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 76 48 52 c2    	vdpbf16ps zmm0,zmm1,zmm2
   6:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
   c:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  12:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  18:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  1e:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  24:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  2a:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  30:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  36:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,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 ZMM1, ZMM1, ZMM1;
VXORPS ZMM2, ZMM2, ZMM2

Show nanoBench command
Results:
- MPERF: 16.34
- APERF: 23.67
- UOPS: 10.0

Experiment 3

Instruction: VDPBF16PS ZMM0, ZMM1, ZMM2
Chain instructions: VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 76 48 52 c2    	vdpbf16ps zmm0,zmm1,zmm2
   6:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
   c:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  12:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  18:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  1e:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  24:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  2a:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  30:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  36:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,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 ZMM1, ZMM1, ZMM1;
VXORPS ZMM2, ZMM2, ZMM2

Show nanoBench command
Results:
- MPERF: 16.34
- APERF: 23.68
- UOPS: 10.0

Experiment 4 (source registers initialized by an instruction of the same kind)

Instruction: VDPBF16PS ZMM0, ZMM1, ZMM2

Code:

   0:	62 f2 76 48 52 c2    	vdpbf16ps zmm0,zmm1,zmm2

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 ZMM1, ZMM1, ZMM1;
VXORPS ZMM2, ZMM2, ZMM2;
VDPBF16PS ZMM0, ZMM1, ZMM2;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1

Show nanoBench command
Results:
- MPERF: 4.18
- APERF: 5.99
- UOPS: 1.0

Experiment 5 (source registers initialized by an instruction of the same kind)

Instruction: VDPBF16PS ZMM0, ZMM1, ZMM2
Chain instructions: VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 76 48 52 c2    	vdpbf16ps zmm0,zmm1,zmm2
   6:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
   c:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  12:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  18:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  1e:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  24:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  2a:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  30:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  36:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,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 ZMM1, ZMM1, ZMM1;
VXORPS ZMM2, ZMM2, ZMM2;
VDPBF16PS ZMM0, ZMM1, ZMM2;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1

Show nanoBench command
Results:
- MPERF: 16.47
- APERF: 24.03
- UOPS: 10.0

Experiment 6 (source registers initialized by an instruction of the same kind)

Instruction: VDPBF16PS ZMM0, ZMM1, ZMM2
Chain instructions: VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 76 48 52 c2    	vdpbf16ps zmm0,zmm1,zmm2
   6:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
   c:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  12:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  18:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  1e:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  24:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  2a:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  30:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  36:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,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 ZMM1, ZMM1, ZMM1;
VXORPS ZMM2, ZMM2, ZMM2;
VDPBF16PS ZMM0, ZMM1, ZMM2;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1

Show nanoBench command
Results:
- MPERF: 16.53
- APERF: 24.02
- UOPS: 10.0

Latency operand 2 → 1: 5

Experiment 1

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instruction: VDPBF16PS ZMM0, ZMM1, ZMM2

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f2 76 48 52 c2    	vdpbf16ps zmm0,zmm1,zmm2

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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2

Show nanoBench command
Results:
- MPERF: 8.17
- APERF: 12.0
- UOPS: 2.0

Experiment 2 (with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instructions: VDPBF16PS ZMM0, ZMM1, ZMM2;VMOVUPD ZMM1, ZMM15

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f2 76 48 52 c2    	vdpbf16ps zmm0,zmm1,zmm2
   c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2

Show nanoBench command
Results:
- MPERF: 8.11
- APERF: 11.99
- UOPS: 3.0

Experiment 3

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instructions: VANDPD ZMM0, ZMM1, ZMM1;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f1 f5 48 54 c1    	vandpd zmm0,zmm1,zmm1
   c:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  12:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  18:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  1e:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  24:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  2a:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  30:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  36:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2

Show nanoBench command
Results:
- MPERF: 16.4
- APERF: 24.01
- UOPS: 10.0

Experiment 4 (with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instructions: VANDPD ZMM0, ZMM1, ZMM1;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;;VMOVUPD ZMM1, ZMM15
Chain latency: 18

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f1 f5 48 54 c1    	vandpd zmm0,zmm1,zmm1
   c:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  12:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  18:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  1e:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  24:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  2a:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  30:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  36:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  3c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2

Show nanoBench command
Results:
- MPERF: 16.21
- APERF: 23.7
- UOPS: 11.0

Experiment 5

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instructions: VPANDD ZMM0, ZMM1, ZMM1;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f1 75 48 db c1    	vpandd zmm0,zmm1,zmm1
   c:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  12:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  18:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  1e:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  24:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  2a:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  30:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  36:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2

Show nanoBench command
Results:
- MPERF: 16.4
- APERF: 24.02
- UOPS: 10.0

Experiment 6 (with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instructions: VPANDD ZMM0, ZMM1, ZMM1;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;;VMOVUPD ZMM1, ZMM15
Chain latency: 18

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f1 75 48 db c1    	vpandd zmm0,zmm1,zmm1
   c:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  12:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  18:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  1e:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  24:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  2a:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  30:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  36:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  3c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2

Show nanoBench command
Results:
- MPERF: 16.47
- APERF: 24.0
- UOPS: 11.0

Experiment 7 (source registers initialized by an instruction of the same kind)

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instruction: VDPBF16PS ZMM0, ZMM1, ZMM2

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f2 76 48 52 c2    	vdpbf16ps zmm0,zmm1,zmm2

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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM0, ZMM1, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1

Show nanoBench command
Results:
- MPERF: 8.17
- APERF: 12.07
- UOPS: 2.0

Experiment 8 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instructions: VDPBF16PS ZMM0, ZMM1, ZMM2;VMOVUPD ZMM1, ZMM15

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f2 76 48 52 c2    	vdpbf16ps zmm0,zmm1,zmm2
   c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM0, ZMM1, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1

Show nanoBench command
Results:
- MPERF: 8.36
- APERF: 12.01
- UOPS: 3.0

Experiment 9 (source registers initialized by an instruction of the same kind)

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instructions: VANDPD ZMM0, ZMM1, ZMM1;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f1 f5 48 54 c1    	vandpd zmm0,zmm1,zmm1
   c:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  12:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  18:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  1e:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  24:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  2a:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  30:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  36:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM0, ZMM1, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1

Show nanoBench command
Results:
- MPERF: 16.47
- APERF: 24.02
- UOPS: 10.0

Experiment 10 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instructions: VANDPD ZMM0, ZMM1, ZMM1;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;;VMOVUPD ZMM1, ZMM15
Chain latency: 18

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f1 f5 48 54 c1    	vandpd zmm0,zmm1,zmm1
   c:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  12:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  18:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  1e:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  24:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  2a:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  30:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  36:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  3c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM0, ZMM1, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1

Show nanoBench command
Results:
- MPERF: 16.21
- APERF: 23.71
- UOPS: 11.0

Experiment 11 (source registers initialized by an instruction of the same kind)

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instructions: VPANDD ZMM0, ZMM1, ZMM1;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f1 75 48 db c1    	vpandd zmm0,zmm1,zmm1
   c:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  12:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  18:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  1e:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  24:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  2a:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  30:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  36:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM0, ZMM1, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1

Show nanoBench command
Results:
- MPERF: 16.34
- APERF: 23.69
- UOPS: 10.0

Experiment 12 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM0, ZMM2
Chain instructions: VPANDD ZMM0, ZMM1, ZMM1;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;;VMOVUPD ZMM1, ZMM15
Chain latency: 18

Code:

   0:	62 f2 7e 48 52 ca    	vdpbf16ps zmm1,zmm0,zmm2
   6:	62 f1 75 48 db c1    	vpandd zmm0,zmm1,zmm1
   c:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  12:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  18:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  1e:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  24:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  2a:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  30:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  36:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  3c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM0, ZMM0, ZMM0;
VXORPS ZMM2, ZMM2, ZMM2;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM0, ZMM1, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1

Show nanoBench command
Results:
- MPERF: 16.4
- APERF: 24.0
- UOPS: 11.0

Latency operand 3 → 1: 5

Experiment 1

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instruction: VDPBF16PS ZMM0, ZMM2, ZMM1

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f2 6e 48 52 c1    	vdpbf16ps zmm0,zmm2,zmm1

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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0

Show nanoBench command
Results:
- MPERF: 8.17
- APERF: 12.0
- UOPS: 2.0

Experiment 2 (with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instructions: VDPBF16PS ZMM0, ZMM2, ZMM1;VMOVUPD ZMM1, ZMM15

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f2 6e 48 52 c1    	vdpbf16ps zmm0,zmm2,zmm1
   c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0

Show nanoBench command
Results:
- MPERF: 8.23
- APERF: 12.0
- UOPS: 3.0

Experiment 3

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instructions: VANDPD ZMM0, ZMM1, ZMM1;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f1 f5 48 54 c1    	vandpd zmm0,zmm1,zmm1
   c:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  12:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  18:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  1e:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  24:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  2a:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  30:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  36:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0

Show nanoBench command
Results:
- MPERF: 16.34
- APERF: 24.02
- UOPS: 10.0

Experiment 4 (with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instructions: VANDPD ZMM0, ZMM1, ZMM1;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;;VMOVUPD ZMM1, ZMM15
Chain latency: 18

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f1 f5 48 54 c1    	vandpd zmm0,zmm1,zmm1
   c:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  12:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  18:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  1e:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  24:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  2a:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  30:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  36:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  3c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0

Show nanoBench command
Results:
- MPERF: 16.47
- APERF: 24.0
- UOPS: 11.0

Experiment 5

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instructions: VPANDD ZMM0, ZMM1, ZMM1;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f1 75 48 db c1    	vpandd zmm0,zmm1,zmm1
   c:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  12:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  18:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  1e:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  24:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  2a:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  30:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  36:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0

Show nanoBench command
Results:
- MPERF: 16.34
- APERF: 24.02
- UOPS: 10.0

Experiment 6 (with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instructions: VPANDD ZMM0, ZMM1, ZMM1;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;;VMOVUPD ZMM1, ZMM15
Chain latency: 18

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f1 75 48 db c1    	vpandd zmm0,zmm1,zmm1
   c:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  12:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  18:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  1e:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  24:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  2a:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  30:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  36:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  3c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0

Show nanoBench command
Results:
- MPERF: 16.34
- APERF: 23.99
- UOPS: 11.0

Experiment 7 (source registers initialized by an instruction of the same kind)

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instruction: VDPBF16PS ZMM0, ZMM2, ZMM1

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f2 6e 48 52 c1    	vdpbf16ps zmm0,zmm2,zmm1

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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1;
VDPBF16PS ZMM0, ZMM1, ZMM2

Show nanoBench command
Results:
- MPERF: 8.36
- APERF: 12.07
- UOPS: 2.0

Experiment 8 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instructions: VDPBF16PS ZMM0, ZMM2, ZMM1;VMOVUPD ZMM1, ZMM15

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f2 6e 48 52 c1    	vdpbf16ps zmm0,zmm2,zmm1
   c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1;
VDPBF16PS ZMM0, ZMM1, ZMM2

Show nanoBench command
Results:
- MPERF: 8.17
- APERF: 12.0
- UOPS: 3.0

Experiment 9 (source registers initialized by an instruction of the same kind)

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instructions: VANDPD ZMM0, ZMM1, ZMM1;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f1 f5 48 54 c1    	vandpd zmm0,zmm1,zmm1
   c:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  12:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  18:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  1e:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  24:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  2a:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  30:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  36:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1;
VDPBF16PS ZMM0, ZMM1, ZMM2

Show nanoBench command
Results:
- MPERF: 16.34
- APERF: 24.02
- UOPS: 10.0

Experiment 10 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instructions: VANDPD ZMM0, ZMM1, ZMM1;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;VANDPD ZMM0, ZMM0, ZMM0;;VMOVUPD ZMM1, ZMM15
Chain latency: 18

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f1 f5 48 54 c1    	vandpd zmm0,zmm1,zmm1
   c:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  12:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  18:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  1e:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  24:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  2a:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  30:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  36:	62 f1 fd 48 54 c0    	vandpd zmm0,zmm0,zmm0
  3c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1;
VDPBF16PS ZMM0, ZMM1, ZMM2

Show nanoBench command
Results:
- MPERF: 16.34
- APERF: 24.0
- UOPS: 11.0

Experiment 11 (source registers initialized by an instruction of the same kind)

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instructions: VPANDD ZMM0, ZMM1, ZMM1;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0
Chain latency: 18

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f1 75 48 db c1    	vpandd zmm0,zmm1,zmm1
   c:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  12:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  18:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  1e:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  24:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  2a:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  30:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  36:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1;
VDPBF16PS ZMM0, ZMM1, ZMM2

Show nanoBench command
Results:
- MPERF: 16.34
- APERF: 23.67
- UOPS: 10.0

Experiment 12 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)

Instruction: VDPBF16PS ZMM1, ZMM2, ZMM0
Chain instructions: VPANDD ZMM0, ZMM1, ZMM1;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;VPANDD ZMM0, ZMM0, ZMM0;;VMOVUPD ZMM1, ZMM15
Chain latency: 18

Code:

   0:	62 f2 6e 48 52 c8    	vdpbf16ps zmm1,zmm2,zmm0
   6:	62 f1 75 48 db c1    	vpandd zmm0,zmm1,zmm1
   c:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  12:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  18:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  1e:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  24:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  2a:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  30:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  36:	62 f1 7d 48 db c0    	vpandd zmm0,zmm0,zmm0
  3c:	62 d1 fd 48 10 cf    	vmovupd zmm1,zmm15

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 ZMM1, [R14];
VXORPS ZMM2, ZMM2, ZMM2;
VXORPS ZMM0, ZMM0, ZMM0;
VDPBF16PS ZMM1, ZMM0, ZMM2;
VDPBF16PS ZMM2, ZMM0, ZMM1;
VDPBF16PS ZMM0, ZMM1, ZMM2

Show nanoBench command
Results:
- MPERF: 16.4
- APERF: 23.99
- UOPS: 11.0

VDPBF16PS (ZMM, ZMM, ZMM) - Latency

Operands

Latency operand 1 → 1: 5

Latency operand 2 → 1: 5

Latency operand 3 → 1: 5

Latency operand 1 → 1: 5

Experiment 1

Experiment 2

Experiment 3

Experiment 4 (source registers initialized by an instruction of the same kind)

Experiment 5 (source registers initialized by an instruction of the same kind)

Experiment 6 (source registers initialized by an instruction of the same kind)

Latency operand 2 → 1: 5

Experiment 1

Experiment 2 (with dependency-breaking instructions)

Experiment 3

Experiment 4 (with dependency-breaking instructions)

Experiment 5

Experiment 6 (with dependency-breaking instructions)

Experiment 7 (source registers initialized by an instruction of the same kind)

Experiment 8 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)

Experiment 9 (source registers initialized by an instruction of the same kind)

Experiment 10 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)

Experiment 11 (source registers initialized by an instruction of the same kind)

Experiment 12 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)

Latency operand 3 → 1: 5

Experiment 1

Experiment 2 (with dependency-breaking instructions)

Experiment 3

Experiment 4 (with dependency-breaking instructions)

Experiment 5

Experiment 6 (with dependency-breaking instructions)

Experiment 7 (source registers initialized by an instruction of the same kind)

Experiment 8 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)

Experiment 9 (source registers initialized by an instruction of the same kind)

Experiment 10 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)

Experiment 11 (source registers initialized by an instruction of the same kind)

Experiment 12 (source registers initialized by an instruction of the same kind, with dependency-breaking instructions)