OV - exec - Loop 3899

	Loop Id: 3899	Module: exec	Source: SoaDistanceTableABOMPTarget.h:228-228 [...]	Coverage: 5.75%

0x520c84 LDR	D1, [X8, X0,LSL #3]    [3]

0x520c88 LDR	D2, [X19, X0,LSL #3]    [4]

0x520c8c LDR	D0, [X19, X2]    [4]

0x520c90 FSUB	D1, D1, S30

0x520c94 FSUB	D2, D2, S31

0x520c98 FSUB	D0, D0, S29

0x520c9c FMUL	D14, D1, D24

0x520ca0 FMUL	D15, D1, D27

0x520ca4 FMUL	D1, D1, D21

0x520ca8 FMADD	D14, D2, D25, D14

0x520cac FMADD	D15, D2, D28, D15

0x520cb0 FMADD	D1, D2, D22, D1

0x520cb4 FMADD	D14, D0, D23, D14

0x520cb8 FMADD	D15, D0, D26, D15

0x520cbc FRINTA	D2, D15

0x520cc0 FMADD	D1, D0, D20, D1

0x520cc4 FRINTA	D0, D14

0x520cc8 FSUB	D0, D14, S0

0x520ccc FSUB	D2, D15, S2

0x520cd0 FRINTA	D14, D1

0x520cd4 FSUB	D15, D1, S14

0x520cd8 FMUL	D14, D0, D18

0x520cdc FMUL	D1, D0, D7

0x520ce0 FMUL	D0, D0, D4

0x520ce4 FMADD	D14, D2, D19, D14

0x520ce8 FMADD	D1, D2, D16, D1

0x520cec FMADD	D2, D2, D5, D0

0x520cf0 FMADD	D14, D15, D17, D14

0x520cf4 FMADD	D1, D15, D6, D1

0x520cf8 FMADD	D15, D15, D3, D2

0x520cfc STR	D14, [X3, X0,LSL #3]    [5]

0x520d00 STR	D1, [X5, X0,LSL #3]    [1]

0x520d04 STR	D15, [X3, X2]    [5]

0x520d08 ADD	X2, X2, #8

0x520d0c LDR	D0, [X5, X0,LSL #3]    [1]

0x520d10 LDR	D2, [X3, X0,LSL #3]    [5]

0x520d14 FMUL	D14, D0, D0

0x520d18 FMADD	D1, D2, D2, D14

0x520d1c FMADD	D15, D15, D15, D1

0x520d20 FSQRT	D0, D15

0x520d24 STR	D0, [X7, X0,LSL #3]    [2]

0x520d28 ADD	X0, X0, #1

0x520d2c CMP	W12, W0

0x520d30 B.GT	520c84

/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Particle/Lattice/ParticleBConds3DSoa.h: 280 - 298

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

280:     T displ_0 = px[iat] - x0;

281:     T displ_1 = py[iat] - y0;

282:     T displ_2 = pz[iat] - z0;

283:

284:     T ar_0 = displ_0 * g00 + displ_1 * g10 + displ_2 * g20;

285:     T ar_1 = displ_0 * g01 + displ_1 * g11 + displ_2 * g21;

286:     T ar_2 = displ_0 * g02 + displ_1 * g12 + displ_2 * g22;

287:

288:     //put them in the box

289:     ar_0 -= round(ar_0);

290:     ar_1 -= round(ar_1);

291:     ar_2 -= round(ar_2);

292:

293:     //unit2cart

294:     dx[iat] = ar_0 * r00 + ar_1 * r10 + ar_2 * r20;

295:     dy[iat] = ar_0 * r01 + ar_1 * r11 + ar_2 * r21;

296:     dz[iat] = ar_0 * r02 + ar_1 * r12 + ar_2 * r22;

297:

298:     temp_r[iat] = std::sqrt(dx[iat] * dx[iat] + dy[iat] * dy[iat] + dz[iat] * dz[iat]);

/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Particle/SoaDistanceTableABOMPTarget.h: 228 - 228

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

228:           for (int iel = first; iel < last; iel++)

Coverage (%)	Name	Source Location	Module
►98.75+	qmcplusplus::ParticleSet::upda[...]	stl_vector.h:993	exec
○	main._omp_fn.1	NonLocalPP.hpp:135	exec
○	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	2.92
CQA speedup if FP arith vectorized	1.67
CQA speedup if fully vectorized	2.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.59
Bottlenecks	P8, P9, P10, P11,
Function	qmcplusplus::SoaDistanceTableABOMPTarget::evaluate(qmcplusplus::ParticleSet&)
Source	ParticleBConds3DSoa.h:280-298,SoaDistanceTableABOMPTarget.h:228-228
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	8.75
CQA cycles if no scalar integer	3.00
CQA cycles if FP arith vectorized	5.25
CQA cycles if fully vectorized	4.38
Front-end cycles	5.50
P0 cycles	0.50
P1 cycles	0.50
P2 cycles	0.58
P3 cycles	0.58
P4 cycles	0.50
P5 cycles	0.50
P6 cycles	0.42
P7 cycles	0.42
P8 cycles	8.75
P9 cycles	8.75
P10 cycles	8.75
P11 cycles	8.75
P12 cycles	3.00
P13 cycles	3.00
P14 cycles	3.00
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	1.75 - 3.75
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	44.00
Nb uops	44.00
Nb loads	NA
Nb stores	4.00
Nb stack references	0.00
FLOP/cycle	4.80
Nb FLOP add-sub	6.00
Nb FLOP mul	7.00
Nb FLOP fma	14.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	1.00
Nb FLOP rsqrt	0.00
Bytes/cycle	8.23
Bytes prefetched	0.00
Bytes loaded	40.00
Bytes stored	32.00
Stride 0	0.00
Stride 1	0.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	0.00
Vectorization ratio all	0.00
Vectorization ratio load	0.00
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	50.00
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	50.00

Metric	Value
CQA speedup if no scalar integer	2.92
CQA speedup if FP arith vectorized	1.67
CQA speedup if fully vectorized	2.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.59
Bottlenecks	P8, P9, P10, P11,
Function	qmcplusplus::SoaDistanceTableABOMPTarget::evaluate(qmcplusplus::ParticleSet&)
Source	ParticleBConds3DSoa.h:280-298,SoaDistanceTableABOMPTarget.h:228-228
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	8.75
CQA cycles if no scalar integer	3.00
CQA cycles if FP arith vectorized	5.25
CQA cycles if fully vectorized	4.38
Front-end cycles	5.50
P0 cycles	0.50
P1 cycles	0.50
P2 cycles	0.58
P3 cycles	0.58
P4 cycles	0.50
P5 cycles	0.50
P6 cycles	0.42
P7 cycles	0.42
P8 cycles	8.75
P9 cycles	8.75
P10 cycles	8.75
P11 cycles	8.75
P12 cycles	3.00
P13 cycles	3.00
P14 cycles	3.00
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	1.75 - 3.75
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	44.00
Nb uops	44.00
Nb loads	NA
Nb stores	4.00
Nb stack references	0.00
FLOP/cycle	4.80
Nb FLOP add-sub	6.00
Nb FLOP mul	7.00
Nb FLOP fma	14.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	1.00
Nb FLOP rsqrt	0.00
Bytes/cycle	8.23
Bytes prefetched	0.00
Bytes loaded	40.00
Bytes stored	32.00
Stride 0	0.00
Stride 1	0.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	0.00
Vectorization ratio all	0.00
Vectorization ratio load	0.00
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	50.00
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	50.00

Path /

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

Function	qmcplusplus::SoaDistanceTableABOMPTarget::evaluate(qmcplusplus::ParticleSet&)
Source file and lines	SoaDistanceTableABOMPTarget.h:228-228
Module	exec

The loop is defined in:

/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Particle/Lattice/ParticleBConds3DSoa.h:280-298
/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Particle/SoaDistanceTableABOMPTarget.h:228

The related source loop is not unrolled or unrolled with no peel/tail loop.

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 8.75 to 3.00 cycles (2.92x 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. 2 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 8.75 to 4.37 cycles (2.00x 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
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 14 FMA (fused multiply-add) operations. Presence of both ADD/SUB and MUL operations.

Workaround

Try to change order in which elements are evaluated (using parentheses) in arithmetic expressions containing both ADD/SUB and MUL operations to enable your compiler to generate FMA instructions wherever possible. For instance a + b*c is a valid FMA (MUL then ADD). However (a+b)* c cannot be translated into an FMA (ADD then MUL).

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

The binary loop is composed of 42 FP arithmetical operations:

20: addition or subtraction (14 inside FMA instructions)
21: multiply (14 inside FMA instructions)
1: square root

The binary loop is loading 40 bytes. The binary loop is storing 32 bytes.

Arithmetic intensity

Arithmetic intensity is 0.58 FP operations per loaded or stored byte.

General properties

nb instructions	44
loop length	176
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

5.50 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	0.58	0.58	0.50	0.50	0.42	0.42	8.75	8.75	8.75	8.75	3.00	3.00	3.00	0.00	0.00
cycles	0.50	0.50	0.58	0.58	0.50	0.50	0.42	0.42	8.75	8.75	8.75	8.75	3.00	3.00	3.00	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	1.75-3.75
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	5.50
Data deps.	1.00
Overall L1	8.75

Vectorization ratios

INT

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

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

INT+FP

all	0%
load	0%
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 8.75 cycles. At this rate:

4% of peak load performance is reached (4.57 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
7% of peak store performance is reached (3.66 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 520c84

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDR D1, [X8, X0,LSL #3]	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 D2, [X19, X0,LSL #3]	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 D0, [X19, X2]	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%)
FSUB D1, D1, S30	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%)
FSUB D2, D2, S31	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%)
FSUB D0, D0, S29	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%)
FMUL D14, D1, D24	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%)
FMUL D15, D1, D27	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%)
FMUL D1, D1, 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 D14, D2, D25, D14	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%)
FMADD D15, D2, D28, D15	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%)
FMADD D1, D2, D22, D1	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%)
FMADD D14, D0, D23, D14	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%)
FMADD D15, D0, D26, D15	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%)
FRINTA D2, D15	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	3	0.50	scal (50.0%)
FMADD D1, D0, D20, D1	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%)
FRINTA D0, D14	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	3	0.50	scal (50.0%)
FSUB D0, D14, S0	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%)
FSUB D2, D15, S2	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%)
FRINTA D14, D1	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	3	0.50	scal (50.0%)
FSUB D15, D1, S14	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%)
FMUL D14, D0, D18	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%)
FMUL D1, D0, D7	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%)
FMUL D0, D0, D4	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 D14, D2, D19, D14	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%)
FMADD D1, D2, D16, D1	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%)
FMADD D2, D2, D5, D0	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%)
FMADD D14, D15, D17, D14	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%)
FMADD D1, D15, D6, D1	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%)
FMADD D15, D15, D3, D2	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 D14, [X3, X0,LSL #3]	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%)
STR D1, [X5, X0,LSL #3]	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%)
STR D15, [X3, X2]	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%)
ADD X2, X2, #8	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 D0, [X5, X0,LSL #3]	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 D2, [X3, X0,LSL #3]	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 D14, D0, D0	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 D1, D2, D2, D14	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%)
FMADD D15, D15, D15, D1	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%)
FSQRT D0, D15	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	7-16	1.75-3.75	scal (50.0%)
STR D0, [X7, X0,LSL #3]	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%)
ADD X0, X0, #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
CMP W12, W0	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 520c84 <_ZN11qmcplusplus27SoaDistanceTableABOMPTargetIdLj3ELi40EE8evaluateERNS_11ParticleSetE+0x3a4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Function	qmcplusplus::SoaDistanceTableABOMPTarget::evaluate(qmcplusplus::ParticleSet&)
Source file and lines	SoaDistanceTableABOMPTarget.h:228-228
Module	exec

The loop is defined in:

/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Particle/Lattice/ParticleBConds3DSoa.h:280-298
/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Particle/SoaDistanceTableABOMPTarget.h:228

The related source loop is not unrolled or unrolled with no peel/tail loop.

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 8.75 to 3.00 cycles (2.92x 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. 2 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 8.75 to 4.37 cycles (2.00x 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
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 14 FMA (fused multiply-add) operations. Presence of both ADD/SUB and MUL operations.

Workaround

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

The binary loop is composed of 42 FP arithmetical operations:

20: addition or subtraction (14 inside FMA instructions)
21: multiply (14 inside FMA instructions)
1: square root

The binary loop is loading 40 bytes. The binary loop is storing 32 bytes.

Arithmetic intensity

Arithmetic intensity is 0.58 FP operations per loaded or stored byte.

General properties

nb instructions	44
loop length	176
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

5.50 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	0.58	0.58	0.50	0.50	0.42	0.42	8.75	8.75	8.75	8.75	3.00	3.00	3.00	0.00	0.00
cycles	0.50	0.50	0.58	0.58	0.50	0.50	0.42	0.42	8.75	8.75	8.75	8.75	3.00	3.00	3.00	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	1.75-3.75
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	5.50
Data deps.	1.00
Overall L1	8.75

Vectorization ratios

INT

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

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

INT+FP

all	0%
load	0%
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 8.75 cycles. At this rate:

4% of peak load performance is reached (4.57 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
7% of peak store performance is reached (3.66 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 520c84

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDR D1, [X8, X0,LSL #3]	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 D2, [X19, X0,LSL #3]	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 D0, [X19, X2]	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%)
FSUB D1, D1, S30	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%)
FSUB D2, D2, S31	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%)
FSUB D0, D0, S29	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%)
FMUL D14, D1, D24	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%)
FMUL D15, D1, D27	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%)
FMUL D1, D1, 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 D14, D2, D25, D14	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%)
FMADD D15, D2, D28, D15	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%)
FMADD D1, D2, D22, D1	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%)
FMADD D14, D0, D23, D14	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%)
FMADD D15, D0, D26, D15	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%)
FRINTA D2, D15	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	3	0.50	scal (50.0%)
FMADD D1, D0, D20, D1	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%)
FRINTA D0, D14	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	3	0.50	scal (50.0%)
FSUB D0, D14, S0	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%)
FSUB D2, D15, S2	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%)
FRINTA D14, D1	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	3	0.50	scal (50.0%)
FSUB D15, D1, S14	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%)
FMUL D14, D0, D18	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%)
FMUL D1, D0, D7	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%)
FMUL D0, D0, D4	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 D14, D2, D19, D14	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%)
FMADD D1, D2, D16, D1	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%)
FMADD D2, D2, D5, D0	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%)
FMADD D14, D15, D17, D14	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%)
FMADD D1, D15, D6, D1	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%)
FMADD D15, D15, D3, D2	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 D14, [X3, X0,LSL #3]	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%)
STR D1, [X5, X0,LSL #3]	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%)
STR D15, [X3, X2]	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%)
ADD X2, X2, #8	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 D0, [X5, X0,LSL #3]	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 D2, [X3, X0,LSL #3]	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 D14, D0, D0	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 D1, D2, D2, D14	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%)
FMADD D15, D15, D15, D1	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%)
FSQRT D0, D15	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	7-16	1.75-3.75	scal (50.0%)
STR D0, [X7, X0,LSL #3]	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%)
ADD X0, X0, #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
CMP W12, W0	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 520c84 <_ZN11qmcplusplus27SoaDistanceTableABOMPTargetIdLj3ELi40EE8evaluateERNS_11ParticleSetE+0x3a4>	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

FMA

Workaround

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

Workaround

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