OV - exec - Function qmcplusplus::SoaDistanceTableAAOMPTarget<double, 3u, 40>::update(int)

	Function: qmcplusplus::SoaDistanceTableAAOMPTarget<double, 3u, 40>::update(int)	Module: exec	Source: SoaDistanceTableAAOMPTarget.h:429-445 [...]	Coverage (incl. loops): 9.13% \| (excl. loops): 0.00%

/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Utilities/NewTimer.h: 242 - 249

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

242:   ScopeGuard(TIMER& t) : timer(t) { timer.start(); }

[...]

249:   ~ScopeGuard() { timer.stop(); }

/opt/arm/gcc-14.2.0_AmazonLinux-2023/include/c++/14.2.0/bits/stl_vector.h: 1131 - 1131

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

1131: 	return *(this->_M_impl._M_start + __n);

/opt/arm/gcc-14.2.0_AmazonLinux-2023/include/c++/14.2.0/bits/stl_algo.h: 683 - 709

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

683:     { return std::copy(__first, __first + __n, __result); }

[...]

709:       if (__n2 <= 0)

/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h: 223 - 248

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

223:     return X[i];

[...]

248:   inline pointer data() { return X; }

/opt/arm/gcc-14.2.0_AmazonLinux-2023/include/c++/14.2.0/bits/stl_algobase.h: 413 - 452

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

413: 	{ *__to = *__from; }

[...]

451: 	  if (__builtin_expect(_Num > 1, true))

452: 	    __builtin_memmove(__result, __first, sizeof(_Tp) * _Num);

/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/VectorSoAContainer.h: 231 - 271

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

231:   inline const AoSElement_t operator[](size_t i) const { return AoSElement_t(myData + i, nGhosts); }

[...]

244:         *(_base + M * i) = rhs[i];

[...]

263:   inline Accessor operator()(size_t i) { return Accessor(myData + i, nGhosts); }

[...]

271:   inline T* restrict data(size_t i) { return myData + i * nGhosts; }

/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Particle/SoaDistanceTableAAOMPTarget.h: 429 - 445

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

429:   inline void update(IndexType iat) override

430:   {

431:     ScopedTimer local_timer(update_timer_);

[...]

437:     for (int idim = 0; idim < D; ++idim)

438:       std::copy_n(temp_dr_.data(idim), nupdate, displacements_[iat].data(idim));

439:     //copy column

440:     for (size_t i = iat + 1; i < num_targets_; ++i)

441:     {

442:       distances_[i][iat]     = temp_r_[i];

443:       displacements_[i](iat) = -temp_dr_[i];

444:     }

445:   }

/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/TinyVector.h: 182 - 182

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

182:       inverse[d] = -X[d];

0x4e8ee0 STP	X29, X30, [SP, #944]!

0x4e8ee4 ADD	X29, SP, #0

0x4e8ee8 STP	X19, X20, [SP, #16]

0x4e8eec ORR	W19, WZR, W1

0x4e8ef0 SBFM	X20, X1, #61, #31

0x4e8ef4 STP	X21, X22, [SP, #32]

0x4e8ef8 STP	X23, X24, [SP, #48]

0x4e8efc ORR	X23, XZR, X0

0x4e8f00 STP	X25, X26, [SP, #64]

0x4e8f04 MOVZ	W25, #40

0x4e8f08 SMADDL	X25, W1, W25, XZR

0x4e8f0c LDR	X21, [X0, #648]

0x4e8f10 ORR	X0, XZR, X21

0x4e8f14 BL	4be880

0x4e8f18 CMP	W19, #0

0x4e8f1c B.LE	4e8f88

0x4e8f20 LDR	X0, [X23, #72]

0x4e8f24 LDR	X1, [X23, #144]

0x4e8f28 ADD	X2, X0, X25

0x4e8f2c LDR	X0, [X2, #24]

0x4e8f30 CMP	X20, #8

0x4e8f34 B.EQ	4e90d8

0x4e8f38 ORR	X2, XZR, X20

0x4e8f3c ADD	X24, X23, #160

0x4e8f40 ADD	X26, X23, #96

0x4e8f44 MOVZ	X22, #0

0x4e8f48 BL	403d60

(3063) 0x4e8f4c LDR	X13, [X26]

(3063) 0x4e8f50 ORR	X2, XZR, X20

(3063) 0x4e8f54 LDR	X10, [X24, #8]

(3063) 0x4e8f58 LDR	X11, [X24, #24]

(3063) 0x4e8f5c ADD	X14, X13, X25

(3063) 0x4e8f60 LDR	X15, [X14, #8]

(3063) 0x4e8f64 MADD	X12, X22, X10, XZR

(3063) 0x4e8f68 LDR	X16, [X14, #24]

(3063) 0x4e8f6c ADD	X1, X11, X12,LSL #3

(3063) 0x4e8f70 MADD	X17, X22, X15, XZR

(3063) 0x4e8f74 ADD	X22, X22, #1

(3063) 0x4e8f78 ADD	X0, X16, X17,LSL #3

(3063) 0x4e8f7c BL	403d60

(3063) 0x4e8f80 CMP	X22, #3

(3063) 0x4e8f84 B.NE	4e8f4c

(3061) 0x4e8f88 LDR	X30, [X23, #24]

(3061) 0x4e8f8c ADD	W18, W19, #1

(3061) 0x4e8f90 SBFM	X1, X18, #0, #31

(3061) 0x4e8f94 CMP	X1, X30

(3061) 0x4e8f98 B.CS	4e90b8

(3061) 0x4e8f9c LDR	X24, [X23, #72]

(3061) 0x4e8fa0 SUB	X19, X30, X1

(3061) 0x4e8fa4 ADD	X9, X25, #64

(3061) 0x4e8fa8 LDR	X26, [X23, #96]

(3061) 0x4e8fac LDR	X11, [X23, #144]

(3061) 0x4e8fb0 LDRSW	X0, [X23, #168]

(3061) 0x4e8fb4 SUB	X10, X26, #16

(3061) 0x4e8fb8 LDR	X23, [X23, #184]

(3061) 0x4e8fbc ADD	X12, X23, X0,LSL #3

(3061) 0x4e8fc0 ADD	X6, X23, X0,LSL #4

(3061) 0x4e8fc4 TBZ	W19, #0, 4e9020

(3061) 0x4e8fc8 LDR	D3, [X11, X1,LSL #3]

(3061) 0x4e8fcc LDR	X3, [X24, X9]

(3061) 0x4e8fd0 LDR	X5, [X10, X9]

(3061) 0x4e8fd4 LDR	X2, [X26, X9]

(3061) 0x4e8fd8 ADD	X9, X25, #104

(3061) 0x4e8fdc STR	D3, [X3, X20]

(3061) 0x4e8fe0 LDR	D29, [X23, X1,LSL #3]

(3061) 0x4e8fe4 ADD	X25, X20, X5,LSL #3

(3061) 0x4e8fe8 LDR	D4, [X12, X1,LSL #3]

(3061) 0x4e8fec ADD	X4, X2, X5,LSL #3

(3061) 0x4e8ff0 LDR	D30, [X6, X1,LSL #3]

(3061) 0x4e8ff4 ADD	X1, X1, #1

(3061) 0x4e8ff8 FNEG	D5, D29

(3061) 0x4e8ffc FNEG	D6, D4

(3061) 0x4e9000 FNEG	D7, D30

(3061) 0x4e9004 STR	D5, [X2, X20]

(3061) 0x4e9008 STR	D6, [X2, X25]

(3061) 0x4e900c STR	D7, [X4, X25]

(3061) 0x4e9010 CMP	X1, X30

(3061) 0x4e9014 B.EQ	4e90b8

(3061) 0x4e9018 HINT	#0

(3061) 0x4e901c HINT	#0

  (3062) 0x4e9020 LDR	D16, [X11, X1,LSL #3]

  (3062) 0x4e9024 ADD	X14, X9, #40

  (3062) 0x4e9028 ADD	X13, X1, #1

  (3062) 0x4e902c LDR	X7, [X24, X9]

  (3062) 0x4e9030 LDR	X8, [X10, X9]

  (3062) 0x4e9034 LDR	X15, [X26, X9]

  (3062) 0x4e9038 ADD	X9, X9, #80

  (3062) 0x4e903c STR	D16, [X7, X20]

  (3062) 0x4e9040 LDR	D23, [X23, X1,LSL #3]

  (3062) 0x4e9044 ADD	X16, X20, X8,LSL #3

  (3062) 0x4e9048 LDR	D25, [X12, X1,LSL #3]

  (3062) 0x4e904c ADD	X17, X15, X8,LSL #3

  (3062) 0x4e9050 LDR	D24, [X6, X1,LSL #3]

  (3062) 0x4e9054 ADD	X1, X1, #2

  (3062) 0x4e9058 FNEG	D17, D23

  (3062) 0x4e905c LDR	X18, [X10, X14]

  (3062) 0x4e9060 FNEG	D18, D25

  (3062) 0x4e9064 LDR	X22, [X24, X14]

  (3062) 0x4e9068 FNEG	D19, D24

  (3062) 0x4e906c STR	D17, [X15, X20]

  (3062) 0x4e9070 LDR	X19, [X26, X14]

  (3062) 0x4e9074 ADD	X0, X20, X18,LSL #3

  (3062) 0x4e9078 STR	D18, [X15, X16]

  (3062) 0x4e907c STR	D19, [X17, X16]

  (3062) 0x4e9080 LDR	D20, [X11, X13,LSL #3]

  (3062) 0x4e9084 ADD	X3, X19, X18,LSL #3

  (3062) 0x4e9088 STR	D20, [X22, X20]

  (3062) 0x4e908c LDR	D26, [X23, X13,LSL #3]

  (3062) 0x4e9090 LDR	D28, [X12, X13,LSL #3]

  (3062) 0x4e9094 LDR	D27, [X6, X13,LSL #3]

  (3062) 0x4e9098 FNEG	D21, D26

  (3062) 0x4e909c FNEG	D22, D28

  (3062) 0x4e90a0 FNEG	D31, D27

  (3062) 0x4e90a4 STR	D21, [X19, X20]

  (3062) 0x4e90a8 STR	D22, [X19, X0]

  (3062) 0x4e90ac STR	D31, [X3, X0]

  (3062) 0x4e90b0 CMP	X1, X30

  (3062) 0x4e90b4 B.NE	4e9020

(3061) 0x4e90b8 ORR	X0, XZR, X21

(3061) 0x4e90bc BL	4beb20

(3061) 0x4e90c0 LDP	X19, X20, [SP, #16]

(3061) 0x4e90c4 LDP	X21, X22, [SP, #32]

(3061) 0x4e90c8 LDP	X23, X24, [SP, #48]

(3061) 0x4e90cc LDP	X25, X26, [SP, #64]

(3061) 0x4e90d0 LDP	X29, X30, [SP], #80

(3061) 0x4e90d4 RET

(3061) 0x4e90d8 LDR	D31, [X1]

(3061) 0x4e90dc LDR	X1, [X23, #96]

(3061) 0x4e90e0 LDR	X5, [X23, #184]

(3061) 0x4e90e4 LDR	X3, [X23, #168]

(3061) 0x4e90e8 ADD	X6, X1, X25

(3061) 0x4e90ec STR	D31, [X0]

(3061) 0x4e90f0 LDR	X7, [X6, #24]

(3061) 0x4e90f4 LDR	D0, [X5]

(3061) 0x4e90f8 UBFM	X4, X3, #60, #59

(3061) 0x4e90fc LDR	X8, [X6, #8]

(3061) 0x4e9100 STR	D0, [X7]

(3061) 0x4e9104 LDR	D1, [X5, X3,LSL #3]

(3061) 0x4e9108 UBFM	X9, X8, #60, #59

(3061) 0x4e910c STR	D1, [X7, X8,LSL #3]

(3061) 0x4e9110 LDR	D2, [X5, X4]

(3061) 0x4e9114 STR	D2, [X7, X9]

(3061) 0x4e9118 B	4e8f88

0x4e911c HINT	#0

Coverage (%)	Name	Source Location	Module
►98.95+	qmcplusplus::ParticleSet::acce[...]	stl_vector.h:993	exec
○	main._omp_fn.1	miniqmc.cpp:453	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 /

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

The code analyzed by CQA in that panel excludes loops and represents 0.00% of application time for run gcc_4

Source file and lines	SoaDistanceTableAAOMPTarget.h:429-445
Module	exec

The function is defined in:

/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Utilities/NewTimer.h:242,249
/opt/arm/gcc-14.2.0_AmazonLinux-2023/include/c++/14.2.0/bits/stl_vector.h:1131
/opt/arm/gcc-14.2.0_AmazonLinux-2023/include/c++/14.2.0/bits/stl_algo.h:683,709
/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h:223,248
/opt/arm/gcc-14.2.0_AmazonLinux-2023/include/c++/14.2.0/bits/stl_algobase.h:413,451-452
/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/VectorSoAContainer.h:231,244,263,271
/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Particle/SoaDistanceTableAAOMPTarget.h:429-431,437-445
/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/TinyVector.h:182

Warnings:
Ignoring paths for analysis

gain
potential
hint
expert

Vectorization

Your function is not vectorized. Only 62% of vector register length is used (average across all VPU instructions). By vectorizing your function, you can lower the cost of an iteration from 3.38 to 2.83 cycles (1.19x 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 function can use their full power.

Workaround

Try another compiler or update/tune your current one
Make array accesses unit-stride:
- If your function 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)

Expensive FP math instructions/calls

Detected performance impact from expensive FP math instructions/calls. By removing/reexpressing them, you can lower the cost of an iteration from 3.38 to 1.00 cycles (3.37x speedup).

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

The binary function does not contain any FP arithmetical operations. The binary function is loading 32 bytes. The binary function is storing 80 bytes.

General properties

nb instructions	28
loop length	112
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

3.38 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	2.92	2.92	2.83	2.83	2.75	2.75	0.00	0.00	0.00	0.00	3.17	2.83	3.00	2.50	2.50
cycles	2.00	2.00	2.92	2.92	2.83	2.83	2.75	2.75	0.00	0.00	0.00	0.00	3.17	2.83	3.00	2.50	2.50

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

Cycles summary

Front-end	3.38
Overall L1	3.38

Vectorization ratios

all	0%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	0%
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 call to the function takes 3.38 cycles. At this rate:

9% of peak load performance is reached (9.48 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
49% of peak store performance is reached (23.70 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the function is: 4e8ee0

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16	Latency	Recip. throughput	Vectorization
STP X29, X30, [SP, #944]!	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
ADD X29, SP, #0	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%)
STP X19, X20, [SP, #16]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
ORR W19, WZR, W1	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 (25.0%)
SBFM X20, X1, #61, #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%)
STP X21, X22, [SP, #32]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
STP X23, X24, [SP, #48]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
ORR X23, XZR, X0	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
STP X25, X26, [SP, #64]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
MOVZ W25, #40	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 (25.0%)
SMADDL X25, W1, W25, XZR	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1	scal (25.0%)
LDR X21, [X0, #648]	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 (50.0%)
ORR X0, XZR, X21	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
BL 4be880 <_ZN11qmcplusplus9TimerTypeINSt6chrono3_V212system_clockEE5startEv>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
CMP W19, #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 4e8f88 <_ZN11qmcplusplus27SoaDistanceTableAAOMPTargetIdLj3ELi40EE6updateEi+0xa8>	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
LDR X0, [X23, #72]	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
LDR X1, [X23, #144]	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 (50.0%)
ADD X2, X0, X25	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 X0, [X2, #24]	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
CMP X20, #8	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 4e90d8 <_ZN11qmcplusplus27SoaDistanceTableAAOMPTargetIdLj3ELi40EE6updateEi+0x1f8>	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
ORR X2, XZR, X20	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 X24, X23, #160	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 X26, X23, #96	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
MOVZ X22, #0	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%)
BL 403d60 <@plt_start@+0x80>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
HINT #0																					N/A

The code analyzed by CQA in that panel excludes loops and represents 0.00% of application time for run gcc_4

Source file and lines	SoaDistanceTableAAOMPTarget.h:429-445
Module	exec

The function is defined in:

/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Utilities/NewTimer.h:242,249
/opt/arm/gcc-14.2.0_AmazonLinux-2023/include/c++/14.2.0/bits/stl_vector.h:1131
/opt/arm/gcc-14.2.0_AmazonLinux-2023/include/c++/14.2.0/bits/stl_algo.h:683,709
/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h:223,248
/opt/arm/gcc-14.2.0_AmazonLinux-2023/include/c++/14.2.0/bits/stl_algobase.h:413,451-452
/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/VectorSoAContainer.h:231,244,263,271
/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Particle/SoaDistanceTableAAOMPTarget.h:429-431,437-445
/home/hbollore/qaas/qaas-runs/174-135-6342/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/TinyVector.h:182

Warnings:
Ignoring paths for analysis

gain
potential
hint
expert

Vectorization

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 function can use their full power.

Workaround

Try another compiler or update/tune your current one
Make array accesses unit-stride:
- If your function 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)

Expensive FP math instructions/calls

Detected performance impact from expensive FP math instructions/calls. By removing/reexpressing them, you can lower the cost of an iteration from 3.38 to 1.00 cycles (3.37x speedup).

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

The binary function does not contain any FP arithmetical operations. The binary function is loading 32 bytes. The binary function is storing 80 bytes.

General properties

nb instructions	28
loop length	112
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

3.38 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	2.92	2.92	2.83	2.83	2.75	2.75	0.00	0.00	0.00	0.00	3.17	2.83	3.00	2.50	2.50
cycles	2.00	2.00	2.92	2.92	2.83	2.83	2.75	2.75	0.00	0.00	0.00	0.00	3.17	2.83	3.00	2.50	2.50

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

Cycles summary

Front-end	3.38
Overall L1	3.38

Vectorization ratios

all	0%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	0%
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 call to the function takes 3.38 cycles. At this rate:

9% of peak load performance is reached (9.48 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
49% of peak store performance is reached (23.70 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the function is: 4e8ee0

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16	Latency	Recip. throughput	Vectorization
STP X29, X30, [SP, #944]!	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
ADD X29, SP, #0	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%)
STP X19, X20, [SP, #16]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
ORR W19, WZR, W1	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 (25.0%)
SBFM X20, X1, #61, #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%)
STP X21, X22, [SP, #32]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
STP X23, X24, [SP, #48]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
ORR X23, XZR, X0	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
STP X25, X26, [SP, #64]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
MOVZ W25, #40	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 (25.0%)
SMADDL X25, W1, W25, XZR	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1	scal (25.0%)
LDR X21, [X0, #648]	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 (50.0%)
ORR X0, XZR, X21	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
BL 4be880 <_ZN11qmcplusplus9TimerTypeINSt6chrono3_V212system_clockEE5startEv>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
CMP W19, #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 4e8f88 <_ZN11qmcplusplus27SoaDistanceTableAAOMPTargetIdLj3ELi40EE6updateEi+0xa8>	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
LDR X0, [X23, #72]	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
LDR X1, [X23, #144]	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 (50.0%)
ADD X2, X0, X25	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 X0, [X2, #24]	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
CMP X20, #8	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 4e90d8 <_ZN11qmcplusplus27SoaDistanceTableAAOMPTargetIdLj3ELi40EE6updateEi+0x1f8>	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
ORR X2, XZR, X20	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 X24, X23, #160	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 X26, X23, #96	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
MOVZ X22, #0	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%)
BL 403d60 <@plt_start@+0x80>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
HINT #0																					N/A

Name	Coverage (%)	Time (s)
▼qmcplusplus::SoaDistanceTableAAOMPTarget::update(int)–	9.13	12.60
▼Loop 3061 - stl_algobase.h:413-413 - exec–	0.00	0.01
○Loop 3062 - SoaDistanceTableAAOMPTarget.h:440-442 - exec	9.13	12.47
○Loop 3063 - SoaDistanceTableAAOMPTarget.h:437-437 - exec	0.00	0.00

Report Configuration

Vectorization

Details

Workaround

Expensive FP math instructions/calls

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code

Vectorization

Details

Workaround

Expensive FP math instructions/calls

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code