OV - exec - Loop 374

0x418a68 LDR	X8, [X19, #512]    [5]

0x418a6c LDR	X9, [X19, #488]    [5]

0x418a70 LDUR	X4, [X29, #504]    [6]

0x418a74 ORR	W1, WZR, W21

0x418a78 ORR	X5, XZR, X22

0x418a7c ORR	X6, XZR, X23

0x418a80 ORR	X7, XZR, X24

0x418a84 ADD	X8, X8, X20

0x418a88 LDR	X0, [X8, X27,LSL #3]    [3]

0x418a8c LDR	X8, [X25, #2712]    [4]

0x418a90 ADD	X8, X8, X27,LSL #2

0x418a94 ADD	X27, X27, #1

0x418a98 LDP	W2, W8, [X8]    [2]

0x418a9c CMP	W8, W26

0x418aa0 CSEL	W3, W8, W26, #11

0x418aa4 LDR	X8, [X19, #464]    [5]

0x418aa8 STP	X8, X9, [SP]    [1]

0x418aac BL	4166e0

0x418ab0 LDR	X28, [X19, #160]    [5]

0x418ab4 CMP	X28, X27

0x418ab8 B.HI	418a68

/usr/lib/gcc/aarch64-linux-gnu/11/../../../../include/c++/11/bits/stl_vector.h: 1046 - 1169

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

1046: 	return *(this->_M_impl._M_start + __n);

[...]

1169:       { return _M_data_ptr(this->_M_impl._M_start); }

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h: 229 - 229

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

229:     return X[i];

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/Jastrow/TwoBodyJastrowRef.h: 266 - 271

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

266:   for (int jg = 0; jg < NumGroups; ++jg)

267:   {

268:     const FuncType& f2(*F[igt + jg]);

269:     int iStart = P.first(jg);

270:     int iEnd   = std::min(jelmax, P.last(jg));

271:     f2.evaluateVGL(iat, iStart, iEnd, dist, u, du, d2u, DistCompressed.data(), DistIndice.data());

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Particle/ParticleSet.h: 302 - 305

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

302:   inline int first(int igroup) const { return SubPtcl[igroup]; }

303:

304:   /// return the last index of a group i

305:   inline int last(int igroup) const { return SubPtcl[igroup + 1]; }

Coverage (%)	Name	Source Location	Module
►100.00+	qmcplusplus::WaveFunction::acc[...]	NewTimer.h:249	exec
○	.omp_outlined..62	miniqmc.cpp:445	exec
○	__kmp_invoke_microtask		libomp.so

Path /

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	7.24
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.06
Bottlenecks	P10,
Function	miniqmcreference::TwoBodyJastrowRef >::acceptMove(qmcplusplus::ParticleSet&, int)
Source	stl_vector.h:1046-1046,stl_vector.h:1169-1169,OhmmsVector.h:229-229,TwoBodyJastrowRef.h:266-271,ParticleSet.h:302-305
Source loop unroll info	multi-versionned
Source loop unroll confidence level	high
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	3.17
CQA cycles if no scalar integer	3.17
CQA cycles if FP arith vectorized	3.17
CQA cycles if fully vectorized	0.44
Front-end cycles	2.63
DIV/SQRT cycles	1.00
P0 cycles	1.00
P1 cycles	2.75
P2 cycles	2.75
P3 cycles	2.75
P4 cycles	2.75
P5 cycles	0.00
P6 cycles	0.00
P7 cycles	0.00
P8 cycles	0.00
P9 cycles	3.17
P10 cycles	2.83
P11 cycles	3.00
P12 cycles	0.50
P13 cycles	0.50
P14 cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	21.00
Nb uops	21.00
Nb loads	NA
Nb stores	1.00
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	25.26
Bytes prefetched	0.00
Bytes loaded	64.00
Bytes stored	16.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	NA
Vectorization ratio store	NA
Vectorization ratio mul	NA
Vectorization ratio add_sub	NA
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	20.00
Vector-efficiency ratio load	NA
Vector-efficiency ratio store	NA
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	20.00

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	7.24
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.06
Bottlenecks	P10,
Function	miniqmcreference::TwoBodyJastrowRef >::acceptMove(qmcplusplus::ParticleSet&, int)
Source	stl_vector.h:1046-1046,stl_vector.h:1169-1169,OhmmsVector.h:229-229,TwoBodyJastrowRef.h:266-271,ParticleSet.h:302-305
Source loop unroll info	multi-versionned
Source loop unroll confidence level	high
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	3.17
CQA cycles if no scalar integer	3.17
CQA cycles if FP arith vectorized	3.17
CQA cycles if fully vectorized	0.44
Front-end cycles	2.63
DIV/SQRT cycles	1.00
P0 cycles	1.00
P1 cycles	2.75
P2 cycles	2.75
P3 cycles	2.75
P4 cycles	2.75
P5 cycles	0.00
P6 cycles	0.00
P7 cycles	0.00
P8 cycles	0.00
P9 cycles	3.17
P10 cycles	2.83
P11 cycles	3.00
P12 cycles	0.50
P13 cycles	0.50
P14 cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	21.00
Nb uops	21.00
Nb loads	NA
Nb stores	1.00
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	25.26
Bytes prefetched	0.00
Bytes loaded	64.00
Bytes stored	16.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	NA
Vectorization ratio store	NA
Vectorization ratio mul	NA
Vectorization ratio add_sub	NA
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	20.00
Vector-efficiency ratio load	NA
Vector-efficiency ratio store	NA
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	20.00

Path /

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

Function	miniqmcreference::TwoBodyJastrowRef >::acceptMove(qmcplusplus::ParticleSet&, int)
Source file and lines	TwoBodyJastrowRef.h:266-271
Module	exec

The loop is defined in:

/usr/lib/gcc/aarch64-linux-gnu/11/../../../../include/c++/11/bits/stl_vector.h:1046,1169
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h:229
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/Jastrow/TwoBodyJastrowRef.h:266-271
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Particle/ParticleSet.h:302-305

The related source loop is multi-versionned.

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. Only 20% of vector register length is used (average across all VPU instructions). By vectorizing your loop, you can lower the cost of an iteration from 3.17 to 0.44 cycles (7.24x 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)

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 3.17 to 1.00 cycles (3.17x speedup).

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 64 bytes. The binary loop is storing 16 bytes.

General properties

nb instructions	21
loop length	84
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

2.63 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	1.00	1.00	2.75	2.75	2.75	2.75	0.00	0.00	0.00	0.00	3.17	2.83	3.00	0.50	0.50
cycles	1.00	1.00	2.75	2.75	2.75	2.75	0.00	0.00	0.00	0.00	3.17	2.83	3.00	0.50	0.50

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

Cycles summary

Front-end	2.63
Data deps.	1.00
Overall L1	3.17

Vectorization ratios

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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 3.17 cycles. At this rate:

21% of peak load performance is reached (20.21 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
10% of peak store performance is reached (5.05 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 418a68

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P10	P11	P12	P13	P14	Latency	Recip. throughput
LDR X8, [X19, #512]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X9, [X19, #488]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDUR X4, [X29, #504]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
ORR W1, WZR, W21	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
ORR X5, XZR, X22	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
ORR X6, XZR, X23	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
ORR X7, XZR, X24	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
ADD X8, X8, X20	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
LDR X0, [X8, X27,LSL #3]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X8, [X25, #2712]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
ADD X8, X8, X27,LSL #2	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
ADD X27, X27, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
LDP W2, W8, [X8]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMP W8, W26	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.33
CSEL W3, W8, W26, #11	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
LDR X8, [X19, #464]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
STP X8, X9, [SP]	1	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
BL 4166e0 <_ZNK11qmcplusplus14BsplineFunctorIdE11evaluateVGLEiiiPKdPdS4_S4_S4_Pi>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	1	0.50
LDR X28, [X19, #160]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMP X28, X27	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.33
B.HI 418a68 <_ZN16miniqmcreference17TwoBodyJastrowRefIN11qmcplusplus14BsplineFunctorIdEEE10acceptMoveERNS1_11ParticleSetEi+0x168>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50

Function	miniqmcreference::TwoBodyJastrowRef >::acceptMove(qmcplusplus::ParticleSet&, int)
Source file and lines	TwoBodyJastrowRef.h:266-271
Module	exec

The loop is defined in:

/usr/lib/gcc/aarch64-linux-gnu/11/../../../../include/c++/11/bits/stl_vector.h:1046,1169
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h:229
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/Jastrow/TwoBodyJastrowRef.h:266-271
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Particle/ParticleSet.h:302-305

The related source loop is multi-versionned.

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. Only 20% of vector register length is used (average across all VPU instructions). By vectorizing your loop, you can lower the cost of an iteration from 3.17 to 0.44 cycles (7.24x 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)

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 3.17 to 1.00 cycles (3.17x speedup).

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 64 bytes. The binary loop is storing 16 bytes.

General properties

nb instructions	21
loop length	84
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

2.63 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	1.00	1.00	2.75	2.75	2.75	2.75	0.00	0.00	0.00	0.00	3.17	2.83	3.00	0.50	0.50
cycles	1.00	1.00	2.75	2.75	2.75	2.75	0.00	0.00	0.00	0.00	3.17	2.83	3.00	0.50	0.50

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

Cycles summary

Front-end	2.63
Data deps.	1.00
Overall L1	3.17

Vectorization ratios

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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 3.17 cycles. At this rate:

21% of peak load performance is reached (20.21 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
10% of peak store performance is reached (5.05 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 418a68

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P10	P11	P12	P13	P14	Latency	Recip. throughput
LDR X8, [X19, #512]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X9, [X19, #488]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDUR X4, [X29, #504]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
ORR W1, WZR, W21	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
ORR X5, XZR, X22	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
ORR X6, XZR, X23	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
ORR X7, XZR, X24	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
ADD X8, X8, X20	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
LDR X0, [X8, X27,LSL #3]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X8, [X25, #2712]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
ADD X8, X8, X27,LSL #2	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
ADD X27, X27, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
LDP W2, W8, [X8]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMP W8, W26	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.33
CSEL W3, W8, W26, #11	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25
LDR X8, [X19, #464]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
STP X8, X9, [SP]	1	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
BL 4166e0 <_ZNK11qmcplusplus14BsplineFunctorIdE11evaluateVGLEiiiPKdPdS4_S4_S4_Pi>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	1	0.50
LDR X28, [X19, #160]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMP X28, X27	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.33
B.HI 418a68 <_ZN16miniqmcreference17TwoBodyJastrowRefIN11qmcplusplus14BsplineFunctorIdEEE10acceptMoveERNS1_11ParticleSetEi+0x168>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50

Report Configuration

Vectorization

Details

Workaround

Expensive FP math instructions/calls

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

Expensive FP math instructions/calls

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