OV - exec - Loop 145

	Loop Id: 145	Module: exec	Source: timestep.c:88-94	Coverage: 0.11%

0x40d7ac LDR	W6, [X15, X12,LSL #2]

0x40d7b0 CMP	W6, #0

0x40d7b4 B.LE	40d8ec

0x40d7b8 LDP	X5, X7, [X14, #32]

0x40d7bc SBFM	X4, X11, #0, #31

0x40d7c0 ADD	X10, X11, W6,SXTW

0x40d7c4 LDR	X18, [X5, #32]

0x40d7c8 LDP	X8, X17, [X5, #16]

0x40d7cc ADD	X21, X18, X9

0x40d7d0 ADD	X19, X17, X9

0x40d7d4 TBZ	W6, #0, 40d840

0x40d7d8 LDRSW	X30, [X8, X11,SXTW #2]

0x40d7dc ADD	X4, X4, #1

0x40d7e0 ADD	X21, X21, #24

0x40d7e4 LDR	D23, [X18, X9]

0x40d7e8 LDR	D22, [X17, X9]

0x40d7ec ADD	X20, X7, X30,LSL #4

0x40d7f0 LDR	D29, [X20, #8]

0x40d7f4 FMUL	D0, D31, D23

0x40d7f8 FDIV	D1, D30, D29

0x40d7fc FMADD	D2, D1, D0, D22

0x40d800 STR	D2, [X17, X9]

0x40d804 LDUR	D21, [X21, #496]

0x40d808 LDP	D20, D18, [X19, #8]

0x40d80c ADD	X19, X19, #24

0x40d810 FMUL	D3, D31, D21

0x40d814 FMADD	D4, D1, D3, D20

0x40d818 STUR	D4, [X19, #496]

0x40d81c LDUR	D19, [X21, #504]

0x40d820 FMUL	D5, D31, D19

0x40d824 FMADD	D6, D1, D5, D18

0x40d828 STUR	D6, [X19, #504]

0x40d82c CMP	X10, X4

0x40d830 B.EQ	40d8ec

0x40d834 HINT	#0

0x40d838 HINT	#0

0x40d83c HINT	#0

  (146) 0x40d840 LDRSW	X13, [X8, X4,LSL #2]

  (146) 0x40d844 ADD	X1, X4, #1

  (146) 0x40d848 ADD	X2, X19, #24

  (146) 0x40d84c ORR	X0, XZR, X21

  (146) 0x40d850 ADD	X4, X4, #2

  (146) 0x40d854 LDR	D17, [X21]

  (146) 0x40d858 ADD	X21, X21, #48

  (146) 0x40d85c LDP	D16, D24, [X19]

  (146) 0x40d860 ADD	X3, X7, X13,LSL #4

  (146) 0x40d864 LDRSW	X6, [X8, X1,LSL #2]

  (146) 0x40d868 LDR	D26, [X3, #8]

  (146) 0x40d86c FMUL	D7, D31, D17

  (146) 0x40d870 LDP	D25, D27, [X19, #16]

  (146) 0x40d874 ADD	X19, X19, #48

  (146) 0x40d878 ADD	X5, X7, X6,LSL #4

  (146) 0x40d87c FDIV	D28, D30, D26

  (146) 0x40d880 FMADD	D23, D28, D7, D16

  (146) 0x40d884 STUR	D23, [X19, #464]

  (146) 0x40d888 LDUR	D22, [X21, #472]

  (146) 0x40d88c FMUL	D0, D31, D22

  (146) 0x40d890 FMADD	D29, D28, D0, D24

  (146) 0x40d894 STUR	D29, [X19, #472]

  (146) 0x40d898 LDUR	D1, [X21, #480]

  (146) 0x40d89c FMUL	D2, D31, D1

  (146) 0x40d8a0 FMADD	D20, D28, D2, D25

  (146) 0x40d8a4 STUR	D20, [X19, #480]

  (146) 0x40d8a8 LDR	D3, [X5, #8]

  (146) 0x40d8ac LDUR	D18, [X21, #488]

  (146) 0x40d8b0 FDIV	D4, D30, D3

  (146) 0x40d8b4 FMUL	D21, D31, D18

  (146) 0x40d8b8 FMADD	D19, D4, D21, D27

  (146) 0x40d8bc STUR	D19, [X19, #488]

  (146) 0x40d8c0 LDUR	D16, [X21, #496]

  (146) 0x40d8c4 LDP	D5, D6, [X2, #8]

  (146) 0x40d8c8 FMUL	D24, D31, D16

  (146) 0x40d8cc FMADD	D25, D4, D24, D5

  (146) 0x40d8d0 STR	D25, [X2, #8]

  (146) 0x40d8d4 LDR	D27, [X0, #40]

  (146) 0x40d8d8 FMUL	D17, D31, D27

  (146) 0x40d8dc FMADD	D7, D4, D17, D6

  (146) 0x40d8e0 STR	D7, [X2, #16]

  (146) 0x40d8e4 CMP	X10, X4

  (146) 0x40d8e8 B.NE	40d840

0x40d8ec ADD	X12, X12, #1

0x40d8f0 ADD	X9, X9, #1536

0x40d8f4 ADD	X11, X11, #64

0x40d8f8 CMP	W16, W12

0x40d8fc B.GT	40d7ac

/home/hbollore/qaas/qaas-runs/174-108-7908/intel/CoMD/build/CoMD/CoMD/src-openmp/timestep.c: 88 - 94

--------------------------------------------------------------------------------

88:       for (int iOff=MAXATOMS*iBox,ii=0; ii<s->boxes->nAtoms[iBox]; ii++,iOff++)

89:       {

90:          int iSpecies = s->atoms->iSpecies[iOff];

91:          real_t invMass = 1.0/s->species[iSpecies].mass;

92:          s->atoms->r[iOff][0] += dt*s->atoms->p[iOff][0]*invMass;

93:          s->atoms->r[iOff][1] += dt*s->atoms->p[iOff][1]*invMass;

94:          s->atoms->r[iOff][2] += dt*s->atoms->p[iOff][2]*invMass;

Coverage (%)	Name	Source Location	Module
►99.19+	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	thread_start		libc.so.6

min	med	avg	max

Percentile Index	10	20	30	40	50	60	70	80	90	100
Value

min	med	avg	max

Percentile Index	10	20	30	40	50	60	70	80	90	100
Value

Path /

Metric	Value
CQA speedup if no scalar integer	1.37
CQA speedup if FP arith vectorized	1.16
CQA speedup if fully vectorized	1.44
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.06
Bottlenecks
Function	advancePosition._omp_fn.0
Source	timestep.c:88-94
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	3.19
CQA cycles if no scalar integer	2.33
CQA cycles if FP arith vectorized	2.75
CQA cycles if fully vectorized	2.21
Front-end cycles	3.19
P0 cycles	1.63
P1 cycles	1.63
P2 cycles	1.77
P3 cycles	1.77
P4 cycles	1.75
P5 cycles	1.75
P6 cycles	1.73
P7 cycles	1.73
P8 cycles	1.25
P9 cycles	1.25
P10 cycles	1.25
P11 cycles	1.25
P12 cycles	2.83
P13 cycles	2.67
P14 cycles	2.75
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	0.88 - 1.75
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	26.25
Nb uops	25.50
Nb loads	NA
Nb stores	1.50
Nb stack references	0.00
FLOP/cycle	1.57
Nb FLOP add-sub	0.00
Nb FLOP mul	1.50
Nb FLOP fma	1.50
Nb FLOP div	0.50
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	19.63
Bytes prefetched	0.00
Bytes loaded	64.00
Bytes stored	12.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	2.17
Vectorization ratio load	6.25
Vectorization ratio store	0.00
Vectorization ratio mul	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	48.67
Vector-efficiency ratio load	45.83
Vector-efficiency ratio store	50.00
Vector-efficiency ratio mul	50.00
Vector-efficiency ratio add_sub	50.00
Vector-efficiency ratio fma	50.00
Vector-efficiency ratio div_sqrt	50.00
Vector-efficiency ratio other	51.04

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	3.20
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.20
Bottlenecks	micro-operation queue, P0, P1,
Function	advancePosition._omp_fn.0
Source	timestep.c:88-94
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	1.00
CQA cycles if no scalar integer	1.00
CQA cycles if FP arith vectorized	1.00
CQA cycles if fully vectorized	0.31
Front-end cycles	1.00
P0 cycles	1.00
P1 cycles	1.00
P2 cycles	0.83
P3 cycles	0.83
P4 cycles	0.83
P5 cycles	0.83
P6 cycles	0.83
P7 cycles	0.83
P8 cycles	0.00
P9 cycles	0.00
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.33
P13 cycles	0.33
P14 cycles	0.33
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	8.00
Nb uops	8.00
Nb loads	NA
Nb stores	0.00
Nb stack references	0.00
FLOP/cycle	0.00
Nb FLOP add-sub	0.00
Nb FLOP mul	0.00
Nb FLOP fma	0.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	4.00
Bytes prefetched	0.00
Bytes loaded	4.00
Bytes stored	0.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	NA
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	37.50
Vector-efficiency ratio load	25.00
Vector-efficiency ratio store	NA
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	50.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	25.00

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	1.33
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.26
Bottlenecks	micro-operation queue,
Function	advancePosition._omp_fn.0
Source	timestep.c:88-94
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	2.00
CQA cycles if no scalar integer	2.00
CQA cycles if FP arith vectorized	2.00
CQA cycles if fully vectorized	1.50
Front-end cycles	2.00
P0 cycles	1.50
P1 cycles	1.50
P2 cycles	1.58
P3 cycles	1.58
P4 cycles	1.50
P5 cycles	1.50
P6 cycles	1.42
P7 cycles	1.42
P8 cycles	0.00
P9 cycles	0.00
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	1.33
P13 cycles	1.33
P14 cycles	1.33
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	16.00
Nb uops	16.00
Nb loads	NA
Nb stores	0.00
Nb stack references	0.00
FLOP/cycle	0.00
Nb FLOP add-sub	0.00
Nb FLOP mul	0.00
Nb FLOP fma	0.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	22.00
Bytes prefetched	0.00
Bytes loaded	44.00
Bytes stored	0.00
Stride 0	0.00
Stride 1	0.00
Stride n	2.00
Stride unknown	7.00
Stride indirect	0.00
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	NA
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	55.00
Vector-efficiency ratio load	58.33
Vector-efficiency ratio store	NA
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	50.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	62.50

Metric	Value
CQA speedup if no scalar integer	1.54
CQA speedup if FP arith vectorized	1.22
CQA speedup if fully vectorized	1.37
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.01
Bottlenecks	micro-operation queue,
Function	advancePosition._omp_fn.0
Source	timestep.c:88-94
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	4.88
CQA cycles if no scalar integer	3.17
CQA cycles if FP arith vectorized	4.00
CQA cycles if fully vectorized	3.57
Front-end cycles	4.88
P0 cycles	2.00
P1 cycles	2.00
P2 cycles	2.33
P3 cycles	2.33
P4 cycles	2.33
P5 cycles	2.33
P6 cycles	2.33
P7 cycles	2.33
P8 cycles	2.50
P9 cycles	2.50
P10 cycles	2.50
P11 cycles	2.50
P12 cycles	4.83
P13 cycles	4.50
P14 cycles	4.67
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	1.75 - 3.50
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	42.00
Nb uops	39.00
Nb loads	NA
Nb stores	3.00
Nb stack references	0.00
FLOP/cycle	2.05
Nb FLOP add-sub	0.00
Nb FLOP mul	3.00
Nb FLOP fma	3.00
Nb FLOP div	1.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	26.26
Bytes prefetched	0.00
Bytes loaded	104.00
Bytes stored	24.00
Stride 0	0.00
Stride 1	0.00
Stride n	2.00
Stride unknown	7.00
Stride indirect	0.00
Vectorization ratio all	4.35
Vectorization ratio load	12.50
Vectorization ratio store	0.00
Vectorization ratio mul	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	51.09
Vector-efficiency ratio load	50.00
Vector-efficiency ratio store	50.00
Vector-efficiency ratio mul	50.00
Vector-efficiency ratio add_sub	50.00
Vector-efficiency ratio fma	50.00
Vector-efficiency ratio div_sqrt	50.00
Vector-efficiency ratio other	58.33

Metric	Value
CQA speedup if no scalar integer	1.54
CQA speedup if FP arith vectorized	1.22
CQA speedup if fully vectorized	1.41
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.01
Bottlenecks	micro-operation queue,
Function	advancePosition._omp_fn.0
Source	timestep.c:88-94
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	4.88
CQA cycles if no scalar integer	3.17
CQA cycles if FP arith vectorized	4.00
CQA cycles if fully vectorized	3.47
Front-end cycles	4.88
P0 cycles	2.00
P1 cycles	2.00
P2 cycles	2.33
P3 cycles	2.33
P4 cycles	2.33
P5 cycles	2.33
P6 cycles	2.33
P7 cycles	2.33
P8 cycles	2.50
P9 cycles	2.50
P10 cycles	2.50
P11 cycles	2.50
P12 cycles	4.83
P13 cycles	4.50
P14 cycles	4.67
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	1.75 - 3.50
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	39.00
Nb uops	39.00
Nb loads	NA
Nb stores	3.00
Nb stack references	0.00
FLOP/cycle	2.05
Nb FLOP add-sub	0.00
Nb FLOP mul	3.00
Nb FLOP fma	3.00
Nb FLOP div	1.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	26.26
Bytes prefetched	0.00
Bytes loaded	104.00
Bytes stored	24.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	4.35
Vectorization ratio load	12.50
Vectorization ratio store	0.00
Vectorization ratio mul	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	51.09
Vector-efficiency ratio load	50.00
Vector-efficiency ratio store	50.00
Vector-efficiency ratio mul	50.00
Vector-efficiency ratio add_sub	50.00
Vector-efficiency ratio fma	50.00
Vector-efficiency ratio div_sqrt	50.00
Vector-efficiency ratio other	58.33

Path /

Average path: Display a virtual path defined by average values of all real paths

Function	advancePosition._omp_fn.0
Source file and lines	timestep.c:88-94
Module	exec

The loop is defined in /home/hbollore/qaas/qaas-runs/174-108-7908/intel/CoMD/build/CoMD/CoMD/src-openmp/timestep.c:88-94.

Warnings:
Non-innermost loop: analyzing only self part (ignoring child loops).
This loop has 4 execution paths.

The presence of multiple execution paths is typically the main/first bottleneck.
Try to simplify control inside loop: ideally, try to remove all conditional expressions, for example by (if applicable):

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. Only 48% of vector register length is used (average across all VPU instructions). By fully vectorizing your loop, you can lower the cost of an iteration from 3.19 to 2.21 cycles (1.44x speedup).

Details

2% of VPU instructions are used in vector version (process two or more data elements in vector registers):

6% of VPU loads are used in vector version.
0% of VPU stores are used in vector version.
0% of VPU addition or subtraction instructions are used in vector version.
0% of VPU multiply instructions are used in vector version.
0% of VPU fused multiply-add instructions are used in vector version.
0% of VPU divide and square root instructions are used in vector version.
0% of VPU instructions that are not load, store, addition, subtraction nor multiply instructions are used in vector version.

Since your execution units are vector units, only a fully vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with fassociative-math (included in Ofast or ffast-math) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

FMA

Detected 1.50 FMA (fused multiply-add) operations.

Matching between your loop (in the source code) and the binary loop

The binary loop is composed of 5 FP arithmetical operations:

1.50: addition or subtraction (all inside FMA instructions)
3: multiply (1 inside FMA instructions)
0.50: divide

The binary loop is loading 64 bytes. The binary loop is storing 12 bytes.

Arithmetic intensity

Arithmetic intensity is 0.07 FP operations per loaded or stored byte.

General properties

nb instructions	26.25
loop length	105
nb stack references	0

Front-end

front end

3.19 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	1.63	1.63	1.77	1.77	1.75	1.75	1.73	1.73	1.25	1.25	1.25	1.25	2.83	2.67	2.75	0.00	0.00
cycles	1.63	1.63	1.77	1.77	1.75	1.75	1.73	1.73	1.25	1.25	1.25	1.25	2.83	2.67	2.75	0.00	0.00

Execution ports to units layout:

P0: BRU
P1: BRU
P2: ALU
P3: ALU
P4: ALU
P5: ALU
P6: ALU
P7: ALU
P8 (128 bits): VPU, FP store data, ALU, DIV/SQRT
P9 (128 bits): VPU, ALU, FP store data
P10 (128 bits): VPU, ALU, DIV/SQRT
P11 (128 bits): ALU, VPU
P12 (256 bits): store address, load
P13 (256 bits): store address, load
P14 (256 bits): load
P15 (64 bits): store data
P16 (64 bits): store data

Cycles executing div or sqrt instructions	0.88-1.75
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	3.19
Data deps.	1.00
Overall L1	3.19

Vectorization ratios

INT

all	3%
load	6%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	0%

all	0%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	0%
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	0%
div/sqrt	0%
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	2%
load	6%
store	0%
mul	0%
add-sub	0%
fma	0%
div/sqrt	0%
other	0%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 3.19 cycles. At this rate:

17% of peak load performance is reached (17.17 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
5% of peak store performance is reached (2.46 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

Function	advancePosition._omp_fn.0
Source file and lines	timestep.c:88-94
Module	exec

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. Only 37% of vector register length is used (average across all VPU instructions). By vectorizing your loop, you can lower the cost of an iteration from 1.00 to 0.31 cycles (3.20x speedup).

Details

All VPU instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with fassociative-math (included in Ofast or ffast-math) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

No data for this section

Matching between your loop (in the source code) and the binary loop

The binary loop does not contain any FP arithmetical operations. The binary loop is loading 4 bytes.

General properties

nb instructions	8
loop length	32
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

1.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	1.00	1.00	0.83	0.83	0.83	0.83	0.83	0.83	0.00	0.00	0.00	0.00	0.33	0.33	0.33	0.00	0.00
cycles	1.00	1.00	0.83	0.83	0.83	0.83	0.83	0.83	0.00	0.00	0.00	0.00	0.33	0.33	0.33	0.00	0.00

Execution ports to units layout:

P0: BRU
P1: BRU
P2: ALU
P3: ALU
P4: ALU
P5: ALU
P6: ALU
P7: ALU
P8 (128 bits): VPU, FP store data, ALU, DIV/SQRT
P9 (128 bits): VPU, ALU, FP store data
P10 (128 bits): VPU, ALU, DIV/SQRT
P11 (128 bits): ALU, VPU
P12 (256 bits): store address, load
P13 (256 bits): store address, load
P14 (256 bits): load
P15 (64 bits): store data
P16 (64 bits): store data

Cycles executing div or sqrt instructions	NA
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	1.00
Data deps.	1.00
Overall L1	1.00

Vectorization ratios

all	0%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	0%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 1.00 cycles. At this rate:

4% of peak load performance is reached (4.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 40d7ac

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDR W6, [X15, X12,LSL #2]	1	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	scal (25.0%)
CMP W6, #0	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	1	0.33	scal (25.0%)
B.LE 40d8ec <advancePosition._omp_fn.0+0x1ac>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50	N/A
ADD X12, X12, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	1	0.17	N/A
ADD X9, X9, #1536	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	1	0.17	scal (50.0%)
ADD X11, X11, #64	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	1	0.17	scal (50.0%)
CMP W16, W12	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	1	0.33	N/A
B.GT 40d7ac <advancePosition._omp_fn.0+0x6c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Function	advancePosition._omp_fn.0
Source file and lines	timestep.c:88-94
Module	exec

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. Only 55% of vector register length is used (average across all VPU instructions). By vectorizing your loop, you can lower the cost of an iteration from 2.00 to 1.50 cycles (1.33x speedup).

Details

All VPU instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with fassociative-math (included in Ofast or ffast-math) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

No data for this section

Matching between your loop (in the source code) and the binary loop

The binary loop does not contain any FP arithmetical operations. The binary loop is loading 44 bytes.

General properties

nb instructions	16
loop length	64
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

2.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	1.50	1.50	1.58	1.58	1.50	1.50	1.42	1.42	0.00	0.00	0.00	0.00	1.33	1.33	1.33	0.00	0.00
cycles	1.50	1.50	1.58	1.58	1.50	1.50	1.42	1.42	0.00	0.00	0.00	0.00	1.33	1.33	1.33	0.00	0.00

Execution ports to units layout:

P0: BRU
P1: BRU
P2: ALU
P3: ALU
P4: ALU
P5: ALU
P6: ALU
P7: ALU
P8 (128 bits): VPU, FP store data, ALU, DIV/SQRT
P9 (128 bits): VPU, ALU, FP store data
P10 (128 bits): VPU, ALU, DIV/SQRT
P11 (128 bits): ALU, VPU
P12 (256 bits): store address, load
P13 (256 bits): store address, load
P14 (256 bits): load
P15 (64 bits): store data
P16 (64 bits): store data

Cycles executing div or sqrt instructions	NA
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	2.00
Data deps.	1.00
Overall L1	2.00

Vectorization ratios

all	0%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	0%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 2.00 cycles. At this rate:

22% of peak load performance is reached (22.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 40d7ac

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDR W6, [X15, X12,LSL #2]	1	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	scal (25.0%)
CMP W6, #0	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	1	0.33	scal (25.0%)
B.LE 40d8ec <advancePosition._omp_fn.0+0x1ac>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDP X5, X7, [X14, #32]	1	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.50	N/A
SBFM X4, X11, #0, #31	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	1	0.17	scal (100.0%)
ADD X10, X11, W6,SXTW	1	0	0	0	0	0	0	0.50	0.50	0	0	0	2	0.50	scal (50.0%)
LDR X18, [X5, #32]	1	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	scal (50.0%)
LDP X8, X17, [X5, #16]	1	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.50	scal (100.0%)
ADD X21, X18, X9	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	1	0.17	scal (50.0%)
ADD X19, X17, X9	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	1	0.17	scal (50.0%)
TBZ W6, #0, 40d840 <advancePosition._omp_fn.0+0x100>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50	N/A
ADD X12, X12, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	1	0.17	N/A
ADD X9, X9, #1536	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	1	0.17	scal (50.0%)
ADD X11, X11, #64	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	1	0.17	scal (50.0%)
CMP W16, W12	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	1	0.33	N/A
B.GT 40d7ac <advancePosition._omp_fn.0+0x6c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Function	advancePosition._omp_fn.0
Source file and lines	timestep.c:88-94
Module	exec

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Code clean check

Detected a slowdown caused by scalar integer instructions (typically used for address computation). By removing them, you can lower the cost of an iteration from 4.88 to 3.17 cycles (1.54x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)

Vectorization

Your loop is not vectorized. Only 51% of vector register length is used (average across all VPU instructions). By fully vectorizing your loop, you can lower the cost of an iteration from 4.88 to 3.57 cycles (1.37x speedup).

Details

4% of VPU instructions are used in vector version (process two or more data elements in vector registers):

12% of VPU loads are used in vector version.
0% of VPU stores are used in vector version.
0% of VPU addition or subtraction instructions are used in vector version.
0% of VPU multiply instructions are used in vector version.
0% of VPU fused multiply-add instructions are used in vector version.
0% of VPU divide and square root instructions are used in vector version.
0% of VPU instructions that are not load, store, addition, subtraction nor multiply instructions are used in vector version.

Since your execution units are vector units, only a fully vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with fassociative-math (included in Ofast or ffast-math) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

FMA

Detected 3 FMA (fused multiply-add) operations.

Matching between your loop (in the source code) and the binary loop

The binary loop is composed of 10 FP arithmetical operations:

3: addition or subtraction (all inside FMA instructions)
6: multiply (3 inside FMA instructions)
1: divide

The binary loop is loading 104 bytes. The binary loop is storing 24 bytes.

Arithmetic intensity

Arithmetic intensity is 0.08 FP operations per loaded or stored byte.

General properties

nb instructions	42
loop length	168
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

4.88 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	2.00	2.00	2.33	2.33	2.33	2.33	2.33	2.33	2.50	2.50	2.50	2.50	4.83	4.50	4.67	0.00	0.00
cycles	2.00	2.00	2.33	2.33	2.33	2.33	2.33	2.33	2.50	2.50	2.50	2.50	4.83	4.50	4.67	0.00	0.00

Execution ports to units layout:

P0: BRU
P1: BRU
P2: ALU
P3: ALU
P4: ALU
P5: ALU
P6: ALU
P7: ALU
P8 (128 bits): VPU, FP store data, ALU, DIV/SQRT
P9 (128 bits): VPU, ALU, FP store data
P10 (128 bits): VPU, ALU, DIV/SQRT
P11 (128 bits): ALU, VPU
P12 (256 bits): store address, load
P13 (256 bits): store address, load
P14 (256 bits): load
P15 (64 bits): store data
P16 (64 bits): store data

Cycles executing div or sqrt instructions	1.75-3.50
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	4.88
Data deps.	1.00
Overall L1	4.88

Vectorization ratios

INT

all	6%
load	12%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	0%

all	0%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	0%
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	0%
div/sqrt	0%
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	4%
load	12%
store	0%
mul	0%
add-sub	0%
fma	0%
div/sqrt	0%
other	0%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 4.88 cycles. At this rate:

22% of peak load performance is reached (21.33 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
10% of peak store performance is reached (4.92 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 40d7ac

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16	Latency	Recip. throughput	Vectorization
LDR W6, [X15, X12,LSL #2]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (25.0%)
CMP W6, #0	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33	scal (25.0%)
B.LE 40d8ec <advancePosition._omp_fn.0+0x1ac>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDP X5, X7, [X14, #32]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.50	N/A
SBFM X4, X11, #0, #31	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (100.0%)
ADD X10, X11, W6,SXTW	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	0	0	0	0	2	0.50	scal (50.0%)
LDR X18, [X5, #32]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
LDP X8, X17, [X5, #16]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.50	N/A
ADD X21, X18, X9	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD X19, X17, X9	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
TBZ W6, #0, 40d840 <advancePosition._omp_fn.0+0x100>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDRSW X30, [X8, X11,SXTW #2]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (25.0%)
ADD X4, X4, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (50.0%)
ADD X21, X21, #24	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
LDR D23, [X18, X9]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
LDR D22, [X17, X9]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
ADD X20, X7, X30,LSL #4	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
LDR D29, [X20, #8]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
FMUL D0, D31, D23	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	3	0.25	scal (50.0%)
FDIV D1, D30, D29	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	0	0	7-15	1.75-3.50	scal (50.0%)
FMADD D2, D1, D0, D22	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	4	0.25	scal (50.0%)
STR D2, [X17, X9]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50	scal (50.0%)
LDUR D21, [X21, #496]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
LDP D20, D18, [X19, #8]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	vect (100.0%)
ADD X19, X19, #24	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
FMUL D3, D31, D21	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	3	0.25	scal (50.0%)
FMADD D4, D1, D3, D20	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	4	0.25	scal (50.0%)
STUR D4, [X19, #496]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50	scal (50.0%)
LDUR D19, [X21, #504]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
FMUL D5, D31, D19	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	3	0.25	scal (50.0%)
FMADD D6, D1, D5, D18	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	4	0.25	scal (50.0%)
STUR D6, [X19, #504]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50	scal (50.0%)
CMP X10, X4	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33	scal (50.0%)
B.EQ 40d8ec <advancePosition._omp_fn.0+0x1ac>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
HINT #0																					N/A
HINT #0																					N/A
HINT #0																					N/A
ADD X12, X12, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD X9, X9, #1536	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD X11, X11, #64	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
CMP W16, W12	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33	N/A
B.GT 40d7ac <advancePosition._omp_fn.0+0x6c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Function	advancePosition._omp_fn.0
Source file and lines	timestep.c:88-94
Module	exec

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Code clean check

Detected a slowdown caused by scalar integer instructions (typically used for address computation). By removing them, you can lower the cost of an iteration from 4.88 to 3.17 cycles (1.54x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)

Vectorization

Details

4% of VPU instructions are used in vector version (process two or more data elements in vector registers):

12% of VPU loads are used in vector version.
0% of VPU stores are used in vector version.
0% of VPU addition or subtraction instructions are used in vector version.
0% of VPU multiply instructions are used in vector version.
0% of VPU fused multiply-add instructions are used in vector version.
0% of VPU divide and square root instructions are used in vector version.
0% of VPU instructions that are not load, store, addition, subtraction nor multiply instructions are used in vector version.

Since your execution units are vector units, only a fully vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with fassociative-math (included in Ofast or ffast-math) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

FMA

Detected 3 FMA (fused multiply-add) operations.

Matching between your loop (in the source code) and the binary loop

The binary loop is composed of 10 FP arithmetical operations:

3: addition or subtraction (all inside FMA instructions)
6: multiply (3 inside FMA instructions)
1: divide

The binary loop is loading 104 bytes. The binary loop is storing 24 bytes.

Arithmetic intensity

Arithmetic intensity is 0.08 FP operations per loaded or stored byte.

General properties

nb instructions	39
loop length	156
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

4.88 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	2.00	2.00	2.33	2.33	2.33	2.33	2.33	2.33	2.50	2.50	2.50	2.50	4.83	4.50	4.67	0.00	0.00
cycles	2.00	2.00	2.33	2.33	2.33	2.33	2.33	2.33	2.50	2.50	2.50	2.50	4.83	4.50	4.67	0.00	0.00

Execution ports to units layout:

P0: BRU
P1: BRU
P2: ALU
P3: ALU
P4: ALU
P5: ALU
P6: ALU
P7: ALU
P8 (128 bits): VPU, FP store data, ALU, DIV/SQRT
P9 (128 bits): VPU, ALU, FP store data
P10 (128 bits): VPU, ALU, DIV/SQRT
P11 (128 bits): ALU, VPU
P12 (256 bits): store address, load
P13 (256 bits): store address, load
P14 (256 bits): load
P15 (64 bits): store data
P16 (64 bits): store data

Cycles executing div or sqrt instructions	1.75-3.50
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	4.88
Data deps.	1.00
Overall L1	4.88

Vectorization ratios

INT

all	6%
load	12%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	0%

all	0%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	0%
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	0%
div/sqrt	0%
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	4%
load	12%
store	0%
mul	0%
add-sub	0%
fma	0%
div/sqrt	0%
other	0%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 4.88 cycles. At this rate:

22% of peak load performance is reached (21.33 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
10% of peak store performance is reached (4.92 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 40d7ac

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDR W6, [X15, X12,LSL #2]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	scal (25.0%)
CMP W6, #0	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	1	0.33	scal (25.0%)
B.LE 40d8ec <advancePosition._omp_fn.0+0x1ac>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDP X5, X7, [X14, #32]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.50	N/A
SBFM X4, X11, #0, #31	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	scal (100.0%)
ADD X10, X11, W6,SXTW	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	0	0	2	0.50	scal (50.0%)
LDR X18, [X5, #32]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDP X8, X17, [X5, #16]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.50	N/A
ADD X21, X18, X9	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
ADD X19, X17, X9	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
TBZ W6, #0, 40d840 <advancePosition._omp_fn.0+0x100>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDRSW X30, [X8, X11,SXTW #2]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	scal (25.0%)
ADD X4, X4, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	scal (50.0%)
ADD X21, X21, #24	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR D23, [X18, X9]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
LDR D22, [X17, X9]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
ADD X20, X7, X30,LSL #4	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR D29, [X20, #8]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FMUL D0, D31, D23	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25	scal (50.0%)
FDIV D1, D30, D29	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	7-15	1.75-3.50	scal (50.0%)
FMADD D2, D1, D0, D22	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	4	0.25	scal (50.0%)
STR D2, [X17, X9]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
LDUR D21, [X21, #496]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
LDP D20, D18, [X19, #8]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	vect (100.0%)
ADD X19, X19, #24	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
FMUL D3, D31, D21	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25	scal (50.0%)
FMADD D4, D1, D3, D20	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	4	0.25	scal (50.0%)
STUR D4, [X19, #496]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
LDUR D19, [X21, #504]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FMUL D5, D31, D19	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25	scal (50.0%)
FMADD D6, D1, D5, D18	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	4	0.25	scal (50.0%)
STUR D6, [X19, #504]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
CMP X10, X4	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	1	0.33	scal (50.0%)
B.EQ 40d8ec <advancePosition._omp_fn.0+0x1ac>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
ADD X12, X12, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
ADD X9, X9, #1536	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
ADD X11, X11, #64	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
CMP W16, W12	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	1	0.33	N/A
B.GT 40d7ac <advancePosition._omp_fn.0+0x6c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Report Configuration

Vectorization

Details

Workaround

FMA

Matching between your loop (in the source code) and the binary loop

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

Vectorization

Details

Workaround

Matching between your loop (in the source code) and the binary loop

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code

Vectorization

Details

Workaround

Matching between your loop (in the source code) and the binary loop

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code

Code clean check

Workaround

Vectorization

Details

Workaround

FMA

Matching between your loop (in the source code) and the binary loop

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code

Code clean check

Workaround

Vectorization

Details

Workaround

FMA

Matching between your loop (in the source code) and the binary loop

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code