VADDPS (YMM, YMM, YMM)

With unroll_count=500 and no inner loop

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 1.0
- Core cycles: 1.0
- Reference cycles: 0.99
- UOPS_RETIRED.ALL: 2.0
- UOPS_MS: 0.0

With loop_count=1000 and unroll_count=10

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 1.2
- Core cycles: 1.0
- Reference cycles: 1.03
- UOPS_RETIRED.ALL: 2.2
- UOPS_MS: 0.0

With loop_count=100 and unroll_count=100

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 1.02
- Core cycles: 1.0
- Reference cycles: 1.03
- UOPS_RETIRED.ALL: 2.02
- UOPS_MS: 0.0

With 4 independent instructions

With unroll_count=200 and no inner loop

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2
   4:	c5 f4 58 da          	vaddps ymm3,ymm1,ymm2
   8:	c5 f4 58 e2          	vaddps ymm4,ymm1,ymm2
   c:	c5 f4 58 ea          	vaddps ymm5,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 4.0
- Core cycles: 4.0
- Reference cycles: 4.0
- UOPS_RETIRED.ALL: 8.0
- UOPS_MS: 0.0

With loop_count=1000 and unroll_count=2

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2
   4:	c5 f4 58 da          	vaddps ymm3,ymm1,ymm2
   8:	c5 f4 58 e2          	vaddps ymm4,ymm1,ymm2
   c:	c5 f4 58 ea          	vaddps ymm5,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 5.0
- Core cycles: 4.01
- Reference cycles: 4.1
- UOPS_RETIRED.ALL: 9.0
- UOPS_MS: 0.0

With loop_count=100 and unroll_count=20

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2
   4:	c5 f4 58 da          	vaddps ymm3,ymm1,ymm2
   8:	c5 f4 58 e2          	vaddps ymm4,ymm1,ymm2
   c:	c5 f4 58 ea          	vaddps ymm5,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 4.1
- Core cycles: 4.01
- Reference cycles: 4.1
- UOPS_RETIRED.ALL: 8.1
- UOPS_MS: 0.0

With 8 independent instructions

With unroll_count=100 and no inner loop

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2
   4:	c5 f4 58 da          	vaddps ymm3,ymm1,ymm2
   8:	c5 f4 58 e2          	vaddps ymm4,ymm1,ymm2
   c:	c5 f4 58 ea          	vaddps ymm5,ymm1,ymm2
  10:	c5 f4 58 f2          	vaddps ymm6,ymm1,ymm2
  14:	c5 f4 58 fa          	vaddps ymm7,ymm1,ymm2
  18:	c5 74 58 c2          	vaddps ymm8,ymm1,ymm2
  1c:	c5 74 58 ca          	vaddps ymm9,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 8.0
- Core cycles: 8.0
- Reference cycles: 8.32
- UOPS_RETIRED.ALL: 16.0
- UOPS_MS: 0.0

With loop_count=1000 and unroll_count=1

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2
   4:	c5 f4 58 da          	vaddps ymm3,ymm1,ymm2
   8:	c5 f4 58 e2          	vaddps ymm4,ymm1,ymm2
   c:	c5 f4 58 ea          	vaddps ymm5,ymm1,ymm2
  10:	c5 f4 58 f2          	vaddps ymm6,ymm1,ymm2
  14:	c5 f4 58 fa          	vaddps ymm7,ymm1,ymm2
  18:	c5 74 58 c2          	vaddps ymm8,ymm1,ymm2
  1c:	c5 74 58 ca          	vaddps ymm9,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 10.0
- Core cycles: 8.02
- Reference cycles: 8.21
- UOPS_RETIRED.ALL: 18.0
- UOPS_MS: 0.0

With loop_count=100 and unroll_count=10

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2
   4:	c5 f4 58 da          	vaddps ymm3,ymm1,ymm2
   8:	c5 f4 58 e2          	vaddps ymm4,ymm1,ymm2
   c:	c5 f4 58 ea          	vaddps ymm5,ymm1,ymm2
  10:	c5 f4 58 f2          	vaddps ymm6,ymm1,ymm2
  14:	c5 f4 58 fa          	vaddps ymm7,ymm1,ymm2
  18:	c5 74 58 c2          	vaddps ymm8,ymm1,ymm2
  1c:	c5 74 58 ca          	vaddps ymm9,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 8.2
- Core cycles: 8.02
- Reference cycles: 8.24
- UOPS_RETIRED.ALL: 16.2
- UOPS_MS: 0.0

With 11 independent instructions

With unroll_count=100 and no inner loop

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2
   4:	c5 f4 58 da          	vaddps ymm3,ymm1,ymm2
   8:	c5 f4 58 e2          	vaddps ymm4,ymm1,ymm2
   c:	c5 f4 58 ea          	vaddps ymm5,ymm1,ymm2
  10:	c5 f4 58 f2          	vaddps ymm6,ymm1,ymm2
  14:	c5 f4 58 fa          	vaddps ymm7,ymm1,ymm2
  18:	c5 74 58 c2          	vaddps ymm8,ymm1,ymm2
  1c:	c5 74 58 ca          	vaddps ymm9,ymm1,ymm2
  20:	c5 74 58 d2          	vaddps ymm10,ymm1,ymm2
  24:	c5 74 58 da          	vaddps ymm11,ymm1,ymm2
  28:	c5 74 58 e2          	vaddps ymm12,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 11.0
- Core cycles: 11.0
- Reference cycles: 11.04
- UOPS_RETIRED.ALL: 22.0
- UOPS_MS: 0.0

With loop_count=1000 and unroll_count=1

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2
   4:	c5 f4 58 da          	vaddps ymm3,ymm1,ymm2
   8:	c5 f4 58 e2          	vaddps ymm4,ymm1,ymm2
   c:	c5 f4 58 ea          	vaddps ymm5,ymm1,ymm2
  10:	c5 f4 58 f2          	vaddps ymm6,ymm1,ymm2
  14:	c5 f4 58 fa          	vaddps ymm7,ymm1,ymm2
  18:	c5 74 58 c2          	vaddps ymm8,ymm1,ymm2
  1c:	c5 74 58 ca          	vaddps ymm9,ymm1,ymm2
  20:	c5 74 58 d2          	vaddps ymm10,ymm1,ymm2
  24:	c5 74 58 da          	vaddps ymm11,ymm1,ymm2
  28:	c5 74 58 e2          	vaddps ymm12,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 13.0
- Core cycles: 11.02
- Reference cycles: 11.28
- UOPS_RETIRED.ALL: 24.0
- UOPS_MS: 0.0

With loop_count=100 and unroll_count=10

Code:

   0:	c5 f4 58 c2          	vaddps ymm0,ymm1,ymm2
   4:	c5 f4 58 da          	vaddps ymm3,ymm1,ymm2
   8:	c5 f4 58 e2          	vaddps ymm4,ymm1,ymm2
   c:	c5 f4 58 ea          	vaddps ymm5,ymm1,ymm2
  10:	c5 f4 58 f2          	vaddps ymm6,ymm1,ymm2
  14:	c5 f4 58 fa          	vaddps ymm7,ymm1,ymm2
  18:	c5 74 58 c2          	vaddps ymm8,ymm1,ymm2
  1c:	c5 74 58 ca          	vaddps ymm9,ymm1,ymm2
  20:	c5 74 58 d2          	vaddps ymm10,ymm1,ymm2
  24:	c5 74 58 da          	vaddps ymm11,ymm1,ymm2
  28:	c5 74 58 e2          	vaddps ymm12,ymm1,ymm2

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM1, [R14];
VMOVUPD YMM2, [R14]

Show nanoBench command
Results:
- Instructions retired: 11.2
- Core cycles: 11.02
- Reference cycles: 11.25
- UOPS_RETIRED.ALL: 22.2
- UOPS_MS: 0.0

With the same register for for different operands

With unroll_count=500 and no inner loop

Code:

   0:	c5 fc 58 c0          	vaddps ymm0,ymm0,ymm0

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM0, [R14]

Show nanoBench command
Results:
- Instructions retired: 1.0
- Core cycles: 3.0
- Reference cycles: 3.07
- UOPS_RETIRED.ALL: 2.0
- UOPS_MS: 0.0

With loop_count=1000 and unroll_count=10

Code:

   0:	c5 fc 58 c0          	vaddps ymm0,ymm0,ymm0

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM0, [R14]

Show nanoBench command
Results:
- Instructions retired: 1.2
- Core cycles: 3.0
- Reference cycles: 3.08
- UOPS_RETIRED.ALL: 2.2
- UOPS_MS: 0.0

With loop_count=100 and unroll_count=100

Code:

   0:	c5 fc 58 c0          	vaddps ymm0,ymm0,ymm0

Init:

MOV R15, 10000;
L: VADDPS YMM0, YMM1, YMM1;
VADDPS YMM0, YMM1, YMM1;
DEC R15;
JNZ L;
VZEROALL;
MOV RAX, 0x4000000040000000;
MOV [R14+0], RAX;
MOV [R14+8], RAX;
MOV [R14+16], RAX;
MOV [R14+24], RAX;
VMOVUPD YMM0, [R14]

Show nanoBench command
Results:
- Instructions retired: 1.02
- Core cycles: 3.0
- Reference cycles: 3.08
- UOPS_RETIRED.ALL: 2.02
- UOPS_MS: 0.0

VADDPS (YMM, YMM, YMM) - Throughput and Uops

With different registers for different operands

With 1 independent instruction

With unroll_count=500 and no inner loop

With loop_count=1000 and unroll_count=10

With loop_count=100 and unroll_count=100

With 4 independent instructions

With unroll_count=200 and no inner loop

With loop_count=1000 and unroll_count=2

With loop_count=100 and unroll_count=20

With 8 independent instructions

With unroll_count=100 and no inner loop

With loop_count=1000 and unroll_count=1

With loop_count=100 and unroll_count=10

With 11 independent instructions

With unroll_count=100 and no inner loop

With loop_count=1000 and unroll_count=1

With loop_count=100 and unroll_count=10

With the same register for for different operands

With unroll_count=500 and no inner loop

With loop_count=1000 and unroll_count=10

With loop_count=100 and unroll_count=100