OV - exec - Loop 67

	Loop Id: 67	Module: exec	Source: haloExchange.c:633-642	Coverage: 0.03%

0x4063a0 LDUR	W3, [X18, #400]    [11]

0x4063a4 SUB	X1, X18, #56

0x4063a8 ADD	X2, X18, #56

0x4063ac ADDS	X13, X13, #4

0x4063b0 FMOV	S0, W3

0x4063b4 ADD	X3, X18, #60

0x4063b8 LD1	{V0.S[1]}, [X1]    [8]

0x4063bc ORR	X1, XZR, X18

0x4063c0 LD1	{V0.S[2]}, [X1], #4    [7]

0x4063c4 LDUR	W4, [X18, #404]    [11]

0x4063c8 LDUR	Q7, [X18, #488]    [11]

0x4063cc LDUR	Q25, [X18, #424]    [11]

0x4063d0 LDUR	Q26, [X18, #472]    [11]

0x4063d4 LDUR	Q23, [X18, #408]    [11]

0x4063d8 LDUR	Q5, [X18, #88]    [11]

0x4063dc LDUR	Q6, [X18, #504]    [11]

0x4063e0 LDUR	Q21, [X18, #24]    [11]

0x4063e4 LDUR	Q22, [X18, #72]    [11]

0x4063e8 LDUR	Q19, [X18, #8]    [11]

0x4063ec LDUR	Q1, [X18, #104]    [11]

0x4063f0 LDUR	Q17, [X18, #440]    [11]

0x4063f4 LDUR	Q3, [X18, #40]    [11]

0x4063f8 EXT	V24.16B, V23.16B, V26.16B, #8

0x4063fc EXT	V20.16B, V19.16B, V22.16B, #8

0x406400 LD1	{V0.S[3]}, [X2]    [10]

0x406404 SUB	X2, X18, #52

0x406408 STR	Q0, [X16], #16    [3]

0x40640c FMOV	S0, W4

0x406410 LD1	{V0.S[1]}, [X2]    [9]

0x406414 LD1	{V0.S[2]}, [X1]    [7]

0x406418 SUB	X1, X18, #48

0x40641c LD1	{V23.D[1]}, [X1]    [6]

0x406420 EXT	V16.16B, V25.16B, V7.16B, #8

0x406424 MOV	V25.D[1], V26.D[1]

0x406428 ADD	X1, X18, #64

0x40642c EXT	V18.16B, V17.16B, V6.16B, #8

0x406430 MOV	V17.D[1], V7.D[1]

0x406434 ADD	X18, X18, #224

0x406438 LD1	{V19.D[1]}, [X1]    [5]

0x40643c EXT	V2.16B, V21.16B, V5.16B, #8

0x406440 MOV	V21.D[1], V22.D[1]

0x406444 ORR	X1, XZR, X17

0x406448 EXT	V4.16B, V3.16B, V1.16B, #8

0x40644c MOV	V3.D[1], V5.D[1]

0x406450 ADD	X17, X17, #96

0x406454 LD1	{V0.S[3]}, [X3]    [12]

0x406458 STR	Q0, [X15], #16    [1]

0x40645c ST3	{V23.2D, V24.2D, V25.2D}, [X1], #48    [4]

0x406460 ST3	{V19.2D, V20.2D, V21.2D}, [X1]    [4]

0x406464 ADD	X1, X0, #96

0x406468 ST3	{V16.2D, V17.2D, V18.2D}, [X0], #48    [2]

0x40646c ST3	{V2.2D, V3.2D, V4.2D}, [X0]    [2]

0x406470 ORR	X0, XZR, X1

0x406474 B.NE	4063a0

/home/hbollore/qaas/qaas-runs/174-108-7908/intel/CoMD/build/CoMD/CoMD/src-openmp/haloExchange.c: 633 - 642

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

633:    for (int ii=begin, iTmp=0; ii<end; ++ii, ++iTmp)

634:    {

635:       atoms->gid[ii]   = tmp[iTmp].gid;

636:       atoms->iSpecies[ii] = tmp[iTmp].type;

637:       atoms->r[ii][0]  = tmp[iTmp].rx;

638:       atoms->r[ii][1]  = tmp[iTmp].ry;

639:       atoms->r[ii][2]  = tmp[iTmp].rz;

640:       atoms->p[ii][0]  = tmp[iTmp].px;

641:       atoms->p[ii][1]  = tmp[iTmp].py;

642:       atoms->p[ii][2]  = tmp[iTmp].pz;

Coverage (%)	Name	Source Location	Module
►100.00+	redistributeAtoms.omp_outlined	timestep.c:153	exec
○	__kmp_invoke_microtask		libomp.so

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	1.08
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	1.29
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.38
Bottlenecks	P12, P13, P14,
Function	sortAtomsInCell
Source	haloExchange.c:633-642
Source loop unroll info	unrolled by 4
Source loop unroll confidence level	high
Unroll/vectorization loop type	NA
Unroll factor	4
CQA cycles	9.33
CQA cycles if no scalar integer	8.67
CQA cycles if FP arith vectorized	9.33
CQA cycles if fully vectorized	7.25
Front-end cycles	6.75
P0 cycles	0.50
P1 cycles	0.50
P2 cycles	2.92
P3 cycles	2.92
P4 cycles	2.83
P5 cycles	2.83
P6 cycles	2.75
P7 cycles	2.75
P8 cycles	6.50
P9 cycles	6.50
P10 cycles	6.50
P11 cycles	6.50
P12 cycles	9.33
P13 cycles	9.33
P14 cycles	9.33
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	2
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	54.00
Nb uops	54.00
Nb loads	NA
Nb stores	6.00
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	49.71
Bytes prefetched	0.00
Bytes loaded	240.00
Bytes stored	224.00
Stride 0	2.00
Stride 1	0.00
Stride n	0.00
Stride unknown	9.00
Stride indirect	0.00
Vectorization ratio all	58.54
Vectorization ratio load	57.14
Vectorization ratio store	100.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	50.00
Vector-efficiency ratio all	73.78
Vector-efficiency ratio load	70.24
Vector-efficiency ratio store	100.00
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	70.83

Metric	Value
CQA speedup if no scalar integer	1.08
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	1.29
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.38
Bottlenecks	P12, P13, P14,
Function	sortAtomsInCell
Source	haloExchange.c:633-642
Source loop unroll info	unrolled by 4
Source loop unroll confidence level	high
Unroll/vectorization loop type	NA
Unroll factor	4
CQA cycles	9.33
CQA cycles if no scalar integer	8.67
CQA cycles if FP arith vectorized	9.33
CQA cycles if fully vectorized	7.25
Front-end cycles	6.75
P0 cycles	0.50
P1 cycles	0.50
P2 cycles	2.92
P3 cycles	2.92
P4 cycles	2.83
P5 cycles	2.83
P6 cycles	2.75
P7 cycles	2.75
P8 cycles	6.50
P9 cycles	6.50
P10 cycles	6.50
P11 cycles	6.50
P12 cycles	9.33
P13 cycles	9.33
P14 cycles	9.33
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	2
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	54.00
Nb uops	54.00
Nb loads	NA
Nb stores	6.00
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	49.71
Bytes prefetched	0.00
Bytes loaded	240.00
Bytes stored	224.00
Stride 0	2.00
Stride 1	0.00
Stride n	0.00
Stride unknown	9.00
Stride indirect	0.00
Vectorization ratio all	58.54
Vectorization ratio load	57.14
Vectorization ratio store	100.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	50.00
Vector-efficiency ratio all	73.78
Vector-efficiency ratio load	70.24
Vector-efficiency ratio store	100.00
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	70.83

Path /

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

Function	sortAtomsInCell
Source file and lines	haloExchange.c:633-642
Module	exec

The loop is defined in /home/hbollore/qaas/qaas-runs/174-108-7908/intel/CoMD/build/CoMD/CoMD/src-openmp/haloExchange.c:633-642.

The related source loop is unrolled by 4 (including vectorization).

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 9.33 to 8.67 cycles (1.08x speedup).

Workaround

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

Vectorization

Your loop is partially vectorized. Only 73% 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 9.33 to 7.25 cycles (1.29x speedup).

Details

58% of VPU instructions are used in vector version (process two or more data elements in vector registers):

57% of VPU loads are used in vector version.
0% of VPU addition or subtraction instructions are used in vector version.
50% 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
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 240 bytes. The binary loop is storing 224 bytes.

General properties

nb instructions	54
loop length	216
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

6.75 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	0.50	0.50	2.92	2.92	2.83	2.83	2.75	2.75	6.50	6.50	6.50	6.50	9.33	9.33	9.33	0.00	0.00
cycles	0.50	0.50	2.92	2.92	2.83	2.83	2.75	2.75	6.50	6.50	6.50	6.50	9.33	9.33	9.33	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)	2.00

Cycles summary

Front-end	6.75
Data deps.	2.00
Overall L1	9.33

Vectorization ratios

INT

all	61%
load	57%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	60%

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

INT+FP

all	58%
load	57%
store	100%
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	50%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 9.33 cycles. At this rate:

26% of peak load performance is reached (25.71 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
50% of peak store performance is reached (24.00 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 4063a0

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDUR W3, [X18, #400]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
SUB X1, X18, #56	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 X2, X18, #56	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
ADDS X13, X13, #4	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%)
FMOV S0, W3	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (25.0%)
ADD X3, X18, #60	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
LD1 {V0.S[1]}, [X1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
ORR X1, XZR, X18	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
LD1 {V0.S[2]}, [X1], #4	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
LDUR W4, [X18, #404]	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%)
LDUR Q7, [X18, #488]	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%)
LDUR Q25, [X18, #424]	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%)
LDUR Q26, [X18, #472]	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%)
LDUR Q23, [X18, #408]	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%)
LDUR Q5, [X18, #88]	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%)
LDUR Q6, [X18, #504]	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%)
LDUR Q21, [X18, #24]	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%)
LDUR Q22, [X18, #72]	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%)
LDUR Q19, [X18, #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%)
LDUR Q1, [X18, #104]	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%)
LDUR Q17, [X18, #440]	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%)
LDUR Q3, [X18, #40]	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%)
EXT V24.16B, V23.16B, V26.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
EXT V20.16B, V19.16B, V22.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
LD1 {V0.S[3]}, [X2]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
SUB X2, X18, #52	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
STR Q0, [X16], #16	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0.50	0.50	0	0	0.50	0.50	0	2	0.50	vect (100.0%)
FMOV S0, W4	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (25.0%)
LD1 {V0.S[1]}, [X2]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
LD1 {V0.S[2]}, [X1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
SUB X1, X18, #48	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
LD1 {V23.D[1]}, [X1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (50.0%)
EXT V16.16B, V25.16B, V7.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
MOV V25.D[1], V26.D[1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
ADD X1, X18, #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
EXT V18.16B, V17.16B, V6.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
MOV V17.D[1], V7.D[1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
ADD X18, X18, #224	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
LD1 {V19.D[1]}, [X1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (50.0%)
EXT V2.16B, V21.16B, V5.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
MOV V21.D[1], V22.D[1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
ORR X1, XZR, X17	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
EXT V4.16B, V3.16B, V1.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
MOV V3.D[1], V5.D[1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
ADD X17, X17, #96	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%)
LD1 {V0.S[3]}, [X3]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
STR Q0, [X15], #16	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0.50	0.50	0	0	0.50	0.50	0	2	0.50	vect (100.0%)
ST3 {V23.2D, V24.2D, V25.2D}, [X1], #48	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0.50	0.50	0	0	0.50	0.50	0	6	3.00	vect (100.0%)
ST3 {V19.2D, V20.2D, V21.2D}, [X1]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	6	3.00	vect (100.0%)
ADD X1, X0, #96	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
ST3 {V16.2D, V17.2D, V18.2D}, [X0], #48	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0.50	0.50	0	0	0.50	0.50	0	6	3.00	vect (100.0%)
ST3 {V2.2D, V3.2D, V4.2D}, [X0]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	6	3.00	vect (100.0%)
ORR X0, XZR, X1	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
B.NE 4063a0 <sortAtomsInCell+0x220>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Function	sortAtomsInCell
Source file and lines	haloExchange.c:633-642
Module	exec

The loop is defined in /home/hbollore/qaas/qaas-runs/174-108-7908/intel/CoMD/build/CoMD/CoMD/src-openmp/haloExchange.c:633-642.

The related source loop is unrolled by 4 (including vectorization).

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 9.33 to 8.67 cycles (1.08x speedup).

Workaround

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

Vectorization

Details

58% of VPU instructions are used in vector version (process two or more data elements in vector registers):

57% of VPU loads are used in vector version.
0% of VPU addition or subtraction instructions are used in vector version.
50% 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
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 240 bytes. The binary loop is storing 224 bytes.

General properties

nb instructions	54
loop length	216
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

6.75 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	0.50	0.50	2.92	2.92	2.83	2.83	2.75	2.75	6.50	6.50	6.50	6.50	9.33	9.33	9.33	0.00	0.00
cycles	0.50	0.50	2.92	2.92	2.83	2.83	2.75	2.75	6.50	6.50	6.50	6.50	9.33	9.33	9.33	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)	2.00

Cycles summary

Front-end	6.75
Data deps.	2.00
Overall L1	9.33

Vectorization ratios

INT

all	61%
load	57%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	60%

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

INT+FP

all	58%
load	57%
store	100%
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	50%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 9.33 cycles. At this rate:

26% of peak load performance is reached (25.71 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
50% of peak store performance is reached (24.00 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 4063a0

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDUR W3, [X18, #400]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
SUB X1, X18, #56	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 X2, X18, #56	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
ADDS X13, X13, #4	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%)
FMOV S0, W3	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (25.0%)
ADD X3, X18, #60	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
LD1 {V0.S[1]}, [X1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
ORR X1, XZR, X18	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
LD1 {V0.S[2]}, [X1], #4	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
LDUR W4, [X18, #404]	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%)
LDUR Q7, [X18, #488]	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%)
LDUR Q25, [X18, #424]	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%)
LDUR Q26, [X18, #472]	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%)
LDUR Q23, [X18, #408]	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%)
LDUR Q5, [X18, #88]	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%)
LDUR Q6, [X18, #504]	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%)
LDUR Q21, [X18, #24]	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%)
LDUR Q22, [X18, #72]	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%)
LDUR Q19, [X18, #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%)
LDUR Q1, [X18, #104]	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%)
LDUR Q17, [X18, #440]	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%)
LDUR Q3, [X18, #40]	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%)
EXT V24.16B, V23.16B, V26.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
EXT V20.16B, V19.16B, V22.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
LD1 {V0.S[3]}, [X2]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
SUB X2, X18, #52	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
STR Q0, [X16], #16	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0.50	0.50	0	0	0.50	0.50	0	2	0.50	vect (100.0%)
FMOV S0, W4	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (25.0%)
LD1 {V0.S[1]}, [X2]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
LD1 {V0.S[2]}, [X1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
SUB X1, X18, #48	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
LD1 {V23.D[1]}, [X1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (50.0%)
EXT V16.16B, V25.16B, V7.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
MOV V25.D[1], V26.D[1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
ADD X1, X18, #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
EXT V18.16B, V17.16B, V6.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
MOV V17.D[1], V7.D[1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
ADD X18, X18, #224	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
LD1 {V19.D[1]}, [X1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (50.0%)
EXT V2.16B, V21.16B, V5.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
MOV V21.D[1], V22.D[1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
ORR X1, XZR, X17	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
EXT V4.16B, V3.16B, V1.16B, #8	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
MOV V3.D[1], V5.D[1]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
ADD X17, X17, #96	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%)
LD1 {V0.S[3]}, [X3]	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.33	0.33	0.33	8	0.33	scal (25.0%)
STR Q0, [X15], #16	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0.50	0.50	0	0	0.50	0.50	0	2	0.50	vect (100.0%)
ST3 {V23.2D, V24.2D, V25.2D}, [X1], #48	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0.50	0.50	0	0	0.50	0.50	0	6	3.00	vect (100.0%)
ST3 {V19.2D, V20.2D, V21.2D}, [X1]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	6	3.00	vect (100.0%)
ADD X1, X0, #96	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
ST3 {V16.2D, V17.2D, V18.2D}, [X0], #48	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0.50	0.50	0	0	0.50	0.50	0	6	3.00	vect (100.0%)
ST3 {V2.2D, V3.2D, V4.2D}, [X0]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	6	3.00	vect (100.0%)
ORR X0, XZR, X1	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
B.NE 4063a0 <sortAtomsInCell+0x220>	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

Code clean check

Workaround

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

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