OV - exec - Loop 308

0x424380 LDR	X8, [X20, #512]    [5]

0x424384 LDR	X9, [X20, #488]    [5]

0x424388 ORR	W1, WZR, W22

0x42438c ORR	X4, XZR, X24

0x424390 ORR	X5, XZR, X23

0x424394 ORR	X6, XZR, X25

0x424398 ORR	X7, XZR, X26

0x42439c ADD	X8, X8, X19

0x4243a0 LDR	X0, [X8, X27,LSL #3]    [3]

0x4243a4 LDR	X8, [X21, #616]    [6]

0x4243a8 LDR	X8, [X8, #24]    [2]

0x4243ac ADD	X8, X8, X27,LSL #2

0x4243b0 LDP	W2, W8, [X8]    [1]

0x4243b4 CMP	W8, W28

0x4243b8 CSEL	W3, W8, W28, #11

0x4243bc LDR	X8, [X20, #464]    [5]

0x4243c0 STP	X8, X9, [SP]    [4]

0x4243c4 BL	422660

0x4243c8 LDR	X8, [X20, #160]    [5]

0x4243cc ADD	X27, X27, #1

0x4243d0 CMP	X8, X27

0x4243d4 B.HI	424380

/home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h: 223 - 223

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

223:     return X[i];

/home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/Jastrow/TwoBodyJastrowRef.h: 269 - 274

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

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

270:   {

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

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

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

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

/usr/lib/gcc/aarch64-amazon-linux/14/../../../../include/c++/14/bits/shared_ptr_base.h: 1667 - 1667

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

1667:       { return _M_ptr; }

/home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/build/miniqmc/src/Particle/ParticleSet.h: 313 - 313

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

313:   inline int first(int igroup) const { return (*group_offsets_)[igroup]; }

/usr/lib/gcc/aarch64-amazon-linux/14/../../../../include/c++/14/bits/stl_vector.h: 1131 - 1263

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

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

[...]

1263:       { return _M_data_ptr(this->_M_impl._M_start); }

Coverage (%)	Name	Source Location	Module
►100.00+	qmcplusplus::WaveFunction::rat[...]	WaveFunction.cpp:201	exec
○	main.omp_outlined.62	miniqmc.cpp:442	exec
○	__kmp_invoke_microtask		libomp.so
○	__kmp_fork_call		libomp.so
○	__kmpc_fork_call		libomp.so
○	main	miniqmc.cpp:409	exec
○	__libc_start_call_main		libc.so.6
○	__libc_start_main		libc.so.6
○	_start	new_allocator.h:172	exec

Coverage (%)	Name	Source Location	Module
►100.00+	qmcplusplus::WaveFunction::rat[...]	WaveFunction.cpp:201	exec
○	main.omp_outlined.62	miniqmc.cpp:442	exec
○	__kmp_invoke_microtask		libomp.so
○	__kmp_invoke_task_func		libomp.so
○	__kmp_fork_call		libomp.so
○	__kmpc_fork_call		libomp.so
○	main	miniqmc.cpp:409	exec
○	__libc_start_call_main		libc.so.6
○	__libc_start_main		libc.so.6
○	_start	new_allocator.h:172	exec

Coverage (%)	Name	Source Location	Module
►100.00+	qmcplusplus::WaveFunction::rat[...]	WaveFunction.cpp:201	exec
○	main.omp_outlined.62	miniqmc.cpp:442	exec
○	__kmp_invoke_microtask		libomp.so
○	__kmp_invoke_task_func		libomp.so
○	__kmp_launch_thread		libomp.so
○	__kmp_launch_worker(void*)		libomp.so
○	start_thread		libc.so.6
○	thread_start		libc.so.6

Coverage (%)	Name	Source Location	Module
►66.67+	qmcplusplus::WaveFunction::rat[...]	WaveFunction.cpp:201	exec
○	main.omp_outlined.62	miniqmc.cpp:442	exec
○	__kmp_invoke_microtask		libomp.so
○	__kmp_invoke_task_func		libomp.so
○	__kmp_launch_thread		libomp.so
○	__kmp_launch_worker(void*)		libomp.so
○	start_thread		libc.so.6
○	thread_start		libc.so.6
►33.33+	qmcplusplus::WaveFunction::rat[...]	WaveFunction.cpp:201	exec
○	main.omp_outlined.62	miniqmc.cpp:442	exec
○	__kmp_invoke_microtask		libomp.so
○	__kmp_invoke_task_func		libomp.so
○	__kmp_fork_call		libomp.so
○	__kmpc_fork_call		libomp.so
○	main	miniqmc.cpp:409	exec
○	__libc_start_call_main		libc.so.6
○	__libc_start_main		libc.so.6
○	_start	new_allocator.h:172	exec

Coverage (%)	Name	Source Location	Module
►100.00+	qmcplusplus::WaveFunction::rat[...]	WaveFunction.cpp:201	exec
○	main.omp_outlined.62	miniqmc.cpp:442	exec
○	__kmp_invoke_microtask		libomp.so
○	__kmp_invoke_task_func		libomp.so
○	__kmp_launch_thread		libomp.so
○	__kmp_launch_worker(void*)		libomp.so
○	start_thread		libc.so.6
○	thread_start		libc.so.6

Coverage (%)	Name	Source Location	Module
►100.00+	qmcplusplus::WaveFunction::rat[...]	WaveFunction.cpp:201	exec
○	main.omp_outlined.62	miniqmc.cpp:442	exec
○	__kmp_invoke_microtask		libomp.so
○	__kmp_invoke_task_func		libomp.so
○	__kmp_launch_thread		libomp.so
○	__kmp_launch_worker(void*)		libomp.so
○	start_thread		libc.so.6
○	thread_start		libc.so.6

Coverage (%)	Name	Source Location	Module
►100.00+	qmcplusplus::WaveFunction::rat[...]	WaveFunction.cpp:201	exec
○	main.omp_outlined.62	miniqmc.cpp:442	exec
○	__kmp_invoke_microtask		libomp.so
○	__kmp_invoke_task_func		libomp.so
○	__kmp_launch_thread		libomp.so
○	__kmp_launch_worker(void*)		libomp.so
○	start_thread		libc.so.6
○	thread_start		libc.so.6

Coverage (%)	Name	Source Location	Module
►100.00+	qmcplusplus::WaveFunction::rat[...]	WaveFunction.cpp:201	exec
○	main.omp_outlined.62	miniqmc.cpp:442	exec
○	__kmp_invoke_microtask		libomp.so
○	__kmp_invoke_task_func		libomp.so
○	__kmp_launch_thread		libomp.so
○	__kmp_launch_worker(void*)		libomp.so
○	start_thread		libc.so.6
○	thread_start		libc.so.6

Coverage (%)	Name	Source Location	Module
►100.00+	qmcplusplus::WaveFunction::rat[...]	WaveFunction.cpp:201	exec
○	main.omp_outlined.62	miniqmc.cpp:442	exec
○	__kmp_invoke_microtask		libomp.so
○	__kmp_invoke_task_func		libomp.so
○	__kmp_launch_thread		libomp.so
○	__kmp_launch_worker(void*)		libomp.so
○	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	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	5.43
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.06
Bottlenecks	P10,
Function	miniqmcreference::TwoBodyJastrowRef >::ratioGrad(qmcplusplus::ParticleSet&, int, qmcplusplus::TinyVector&)
Source	OhmmsVector.h:223-223,TwoBodyJastrowRef.h:269-274,shared_ptr_base.h:1667-1667,ParticleSet.h:313-313,stl_vector.h:1131-1131,stl_vector.h:1263-1263
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	NA
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	0.58
Front-end cycles	2.75
P0 cycles	1.00
P1 cycles	1.00
P2 cycles	3.00
P3 cycles	3.00
P4 cycles	3.00
P5 cycles	3.00
P6 cycles	0.00
P7 cycles	0.00
P8 cycles	0.00
P9 cycles	0.00
P10 cycles	3.17
P11 cycles	2.83
P12 cycles	3.00
P13 cycles	0.50
P14 cycles	0.50
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	22.00
Nb uops	22.00
Nb loads	NA
Nb stores	1.00
Nb stack references	1.00
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	0.00
Bytes prefetched	0.00
Bytes loaded	0.00
Bytes stored	0.00
Stride 0	3.00
Stride 1	0.00
Stride n	1.00
Stride unknown	2.00
Stride indirect	0.00
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
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	23.75
Vector-efficiency ratio load	25.00
Vector-efficiency ratio store	50.00
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	19.64

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	5.43
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.06
Bottlenecks	P10,
Function	miniqmcreference::TwoBodyJastrowRef >::ratioGrad(qmcplusplus::ParticleSet&, int, qmcplusplus::TinyVector&)
Source	OhmmsVector.h:223-223,TwoBodyJastrowRef.h:269-274,shared_ptr_base.h:1667-1667,ParticleSet.h:313-313,stl_vector.h:1131-1131,stl_vector.h:1263-1263
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	NA
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	0.58
Front-end cycles	2.75
P0 cycles	1.00
P1 cycles	1.00
P2 cycles	3.00
P3 cycles	3.00
P4 cycles	3.00
P5 cycles	3.00
P6 cycles	0.00
P7 cycles	0.00
P8 cycles	0.00
P9 cycles	0.00
P10 cycles	3.17
P11 cycles	2.83
P12 cycles	3.00
P13 cycles	0.50
P14 cycles	0.50
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	22.00
Nb uops	22.00
Nb loads	NA
Nb stores	1.00
Nb stack references	1.00
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	0.00
Bytes prefetched	0.00
Bytes loaded	0.00
Bytes stored	0.00
Stride 0	3.00
Stride 1	0.00
Stride n	1.00
Stride unknown	2.00
Stride indirect	0.00
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
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	23.75
Vector-efficiency ratio load	25.00
Vector-efficiency ratio store	50.00
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	19.64

Path /

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

Function	miniqmcreference::TwoBodyJastrowRef >::ratioGrad(qmcplusplus::ParticleSet&, int, qmcplusplus::TinyVector&)
Source file and lines	TwoBodyJastrowRef.h:269-274
Module	exec

The loop is defined in:

/home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h:223
/home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/Jastrow/TwoBodyJastrowRef.h:269-274
/usr/lib/gcc/aarch64-amazon-linux/14/../../../../include/c++/14/bits/shared_ptr_base.h:1667
/home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/build/miniqmc/src/Particle/ParticleSet.h:313
/usr/lib/gcc/aarch64-amazon-linux/14/../../../../include/c++/14/bits/stl_vector.h:1131,1263

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

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. Only 23% 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 0.58 cycles (5.43x 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)

No data for this section

CALL instructions

Detected function call instructions.

Details

Calling (and then returning from) a function prevents many compiler optimizations (like vectorization), breaks control flow (which reduces pipeline performance) and executes extra instructions to save/restore the registers used inside it, which is very expensive (dozens of cycles). Consider to inline small functions.

qmcplusplus::BsplineFunctor
::evaluateVGL(int, int, int, double const*, double*, double*, double*, double*, int*) const: 1 occurrences►
- /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/Jastrow/TwoBodyJastrowRef.h:274

Slow data structures access

Detected data structures (typically arrays) that cannot be efficiently read/written

Details

Constant unknown stride: 2 occurrence(s)
Constant non-unit stride: 1 occurrence(s)

Non-unit stride (uncontiguous) accesses are not efficiently using data caches

Workaround

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

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 does not load or store any data.

General properties

nb instructions	22
nb uops	22
loop length	88
used w registers	7
used x registers	16
used b registers	0
used h registers	0
used s registers	1
used d registers	0
used q registers	0
used v registers	0
used z registers	0
nb stack references	1

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	2.75 cycles
front end	2.75 cycles

Back-end

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

Execution ports to units layout:

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

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

Cycles summary

Front-end	2.75
Dispatch	3.17
Data deps.	1.00
Overall L1	3.17

Vectorization ratios

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)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	0%

Vector efficiency ratios

all	23%
load	25%
store	50%
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	19%

Cycles and memory resources usage

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

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 424380

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDR X8, [X20, #512]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
LDR X9, [X20, #488]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (25.0%)
ORR W1, WZR, W22	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (12.5%)
ORR X4, XZR, X24	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (25.0%)
ORR X5, XZR, X23	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (25.0%)
ORR X6, XZR, X25	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (25.0%)
ORR X7, XZR, X26	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (25.0%)
ADD X8, X8, X19	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	N/A
LDR X0, [X8, X27,LSL #3]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (25.0%)
LDR X8, [X21, #616]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
LDR X8, [X8, #24]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
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	N/A
LDP W2, W8, [X8]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
CMP W8, W28	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.33	scal (12.5%)
CSEL W3, W8, W28, #11	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (12.5%)
LDR X8, [X20, #464]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
STP X8, X9, [SP]	1	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (50.0%)
BL 422660 <_ZNK11qmcplusplus14BsplineFunctorIdE11evaluateVGLEiiiPKdPdS4_S4_S4_Pi>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	1	0.50	N/A
LDR X8, [X20, #160]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
ADD X27, X27, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	N/A
CMP X8, X27	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.33	N/A
B.HI 424380 <_ZN16miniqmcreference17TwoBodyJastrowRefIN11qmcplusplus14BsplineFunctorIdEEE9ratioGradERNS1_11ParticleSetEiRNS1_10TinyVectorIdLj3EEE+0xc0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Function	miniqmcreference::TwoBodyJastrowRef >::ratioGrad(qmcplusplus::ParticleSet&, int, qmcplusplus::TinyVector&)
Source file and lines	TwoBodyJastrowRef.h:269-274
Module	exec

The loop is defined in:

/home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h:223
/home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/Jastrow/TwoBodyJastrowRef.h:269-274
/usr/lib/gcc/aarch64-amazon-linux/14/../../../../include/c++/14/bits/shared_ptr_base.h:1667
/home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/build/miniqmc/src/Particle/ParticleSet.h:313
/usr/lib/gcc/aarch64-amazon-linux/14/../../../../include/c++/14/bits/stl_vector.h:1131,1263

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

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. Only 23% 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 0.58 cycles (5.43x 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)

No data for this section

CALL instructions

Detected function call instructions.

Details

Calling (and then returning from) a function prevents many compiler optimizations (like vectorization), breaks control flow (which reduces pipeline performance) and executes extra instructions to save/restore the registers used inside it, which is very expensive (dozens of cycles). Consider to inline small functions.

qmcplusplus::BsplineFunctor
::evaluateVGL(int, int, int, double const*, double*, double*, double*, double*, int*) const: 1 occurrences►
- /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/Jastrow/TwoBodyJastrowRef.h:274

Slow data structures access

Detected data structures (typically arrays) that cannot be efficiently read/written

Details

Constant unknown stride: 2 occurrence(s)
Constant non-unit stride: 1 occurrence(s)

Non-unit stride (uncontiguous) accesses are not efficiently using data caches

Workaround

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

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 does not load or store any data.

General properties

nb instructions	22
nb uops	22
loop length	88
used w registers	7
used x registers	16
used b registers	0
used h registers	0
used s registers	1
used d registers	0
used q registers	0
used v registers	0
used z registers	0
nb stack references	1

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	2.75 cycles
front end	2.75 cycles

Back-end

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

Execution ports to units layout:

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

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

Cycles summary

Front-end	2.75
Dispatch	3.17
Data deps.	1.00
Overall L1	3.17

Vectorization ratios

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)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	0%

Vector efficiency ratios

all	23%
load	25%
store	50%
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	19%

Cycles and memory resources usage

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

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 424380

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDR X8, [X20, #512]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
LDR X9, [X20, #488]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (25.0%)
ORR W1, WZR, W22	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (12.5%)
ORR X4, XZR, X24	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (25.0%)
ORR X5, XZR, X23	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (25.0%)
ORR X6, XZR, X25	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (25.0%)
ORR X7, XZR, X26	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (25.0%)
ADD X8, X8, X19	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	N/A
LDR X0, [X8, X27,LSL #3]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (25.0%)
LDR X8, [X21, #616]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
LDR X8, [X8, #24]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
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	N/A
LDP W2, W8, [X8]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
CMP W8, W28	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.33	scal (12.5%)
CSEL W3, W8, W28, #11	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	scal (12.5%)
LDR X8, [X20, #464]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
STP X8, X9, [SP]	1	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (50.0%)
BL 422660 <_ZNK11qmcplusplus14BsplineFunctorIdE11evaluateVGLEiiiPKdPdS4_S4_S4_Pi>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	1	0.50	N/A
LDR X8, [X20, #160]	1	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
ADD X27, X27, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.25	N/A
CMP X8, X27	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	1	0.33	N/A
B.HI 424380 <_ZN16miniqmcreference17TwoBodyJastrowRefIN11qmcplusplus14BsplineFunctorIdEEE9ratioGradERNS1_11ParticleSetEiRNS1_10TinyVectorIdLj3EEE+0xc0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50	N/A

▶Scalability Runs Description

Run 1x1	Number processes: 1Number nodes: NARun Command: <executable> -g "4 2 2" -bMPI Command: mpirun -n <number_processes> --bind-to core --map-by package:PE=64 --rank-by fill --report-bindings Dataset: Run Directory: /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/run/oneview_runs/multicore/armclang/oneview_run_1782144418OMP_PROC_BIND: spreadOMP_DISPLAY_AFFINITY: TRUEOMP_AFFINITY_FORMAT: 'OMP: pid %P tid %i thread %n bound to OS proc set {%A}'OMP_DISPLAY_ENV: TRUEOMP_NUM_THREADS: 1OMP_PLACES: threads
Run 1x2	Number processes: 1Run Command: <executable> -g "4 2 2" -bMPI Command: mpirun -n <number_processes> --bind-to core --map-by package:PE=64 --rank-by fill --report-bindings Dataset: Run Directory: /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/run/oneview_runs/multicore/armclang/oneview_run_1782144418OMP_NUM_THREADS: 2OMP_PROC_BIND: spreadOMP_DISPLAY_AFFINITY: TRUEOMP_AFFINITY_FORMAT: 'OMP: pid %P tid %i thread %n bound to OS proc set {%A}'OMP_DISPLAY_ENV: TRUEOMP_PLACES: threads
Run 1x4	Number processes: 1Run Command: <executable> -g "4 2 2" -bMPI Command: mpirun -n <number_processes> --bind-to core --map-by package:PE=64 --rank-by fill --report-bindings Dataset: Run Directory: /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/run/oneview_runs/multicore/armclang/oneview_run_1782144418OMP_NUM_THREADS: 4OMP_PROC_BIND: spreadOMP_DISPLAY_AFFINITY: TRUEOMP_AFFINITY_FORMAT: 'OMP: pid %P tid %i thread %n bound to OS proc set {%A}'OMP_DISPLAY_ENV: TRUEOMP_PLACES: threads
Run 1x8	Number processes: 1Run Command: <executable> -g "4 2 2" -bMPI Command: mpirun -n <number_processes> --bind-to core --map-by package:PE=64 --rank-by fill --report-bindings Dataset: Run Directory: /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/run/oneview_runs/multicore/armclang/oneview_run_1782144418OMP_NUM_THREADS: 8OMP_PROC_BIND: spreadOMP_DISPLAY_AFFINITY: TRUEOMP_AFFINITY_FORMAT: 'OMP: pid %P tid %i thread %n bound to OS proc set {%A}'OMP_DISPLAY_ENV: TRUEOMP_PLACES: threads
Run 1x16	Number processes: 1Run Command: <executable> -g "4 2 2" -bMPI Command: mpirun -n <number_processes> --bind-to core --map-by package:PE=64 --rank-by fill --report-bindings Dataset: Run Directory: /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/run/oneview_runs/multicore/armclang/oneview_run_1782144418OMP_NUM_THREADS: 16OMP_PROC_BIND: spreadOMP_DISPLAY_AFFINITY: TRUEOMP_AFFINITY_FORMAT: 'OMP: pid %P tid %i thread %n bound to OS proc set {%A}'OMP_DISPLAY_ENV: TRUEOMP_PLACES: threads
Run 1x24	Number processes: 1Run Command: <executable> -g "4 2 2" -bMPI Command: mpirun -n <number_processes> --bind-to core --map-by package:PE=64 --rank-by fill --report-bindings Dataset: Run Directory: /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/run/oneview_runs/multicore/armclang/oneview_run_1782144418OMP_NUM_THREADS: 24OMP_PROC_BIND: spreadOMP_DISPLAY_AFFINITY: TRUEOMP_AFFINITY_FORMAT: 'OMP: pid %P tid %i thread %n bound to OS proc set {%A}'OMP_DISPLAY_ENV: TRUEOMP_PLACES: threads
Run 1x32	Number processes: 1Run Command: <executable> -g "4 2 2" -bMPI Command: mpirun -n <number_processes> --bind-to core --map-by package:PE=64 --rank-by fill --report-bindings Dataset: Run Directory: /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/run/oneview_runs/multicore/armclang/oneview_run_1782144418OMP_NUM_THREADS: 32OMP_PROC_BIND: spreadOMP_DISPLAY_AFFINITY: TRUEOMP_AFFINITY_FORMAT: 'OMP: pid %P tid %i thread %n bound to OS proc set {%A}'OMP_DISPLAY_ENV: TRUEOMP_PLACES: threads
Run 1x40	Number processes: 1Run Command: <executable> -g "4 2 2" -bMPI Command: mpirun -n <number_processes> --bind-to core --map-by package:PE=64 --rank-by fill --report-bindings Dataset: Run Directory: /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/run/oneview_runs/multicore/armclang/oneview_run_1782144418OMP_NUM_THREADS: 40OMP_PROC_BIND: spreadOMP_DISPLAY_AFFINITY: TRUEOMP_AFFINITY_FORMAT: 'OMP: pid %P tid %i thread %n bound to OS proc set {%A}'OMP_DISPLAY_ENV: TRUEOMP_PLACES: threads
Run 1x48	Number processes: 1Run Command: <executable> -g "4 2 2" -bMPI Command: mpirun -n <number_processes> --bind-to core --map-by package:PE=64 --rank-by fill --report-bindings Dataset: Run Directory: /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/run/oneview_runs/multicore/armclang/oneview_run_1782144418OMP_NUM_THREADS: 48OMP_PROC_BIND: spreadOMP_DISPLAY_AFFINITY: TRUEOMP_AFFINITY_FORMAT: 'OMP: pid %P tid %i thread %n bound to OS proc set {%A}'OMP_DISPLAY_ENV: TRUEOMP_PLACES: threads
Run 1x56	Number processes: 1Run Command: <executable> -g "4 2 2" -bMPI Command: mpirun -n <number_processes> --bind-to core --map-by package:PE=64 --rank-by fill --report-bindings Dataset: Run Directory: /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/run/oneview_runs/multicore/armclang/oneview_run_1782144418OMP_NUM_THREADS: 56OMP_PROC_BIND: spreadOMP_DISPLAY_AFFINITY: TRUEOMP_AFFINITY_FORMAT: 'OMP: pid %P tid %i thread %n bound to OS proc set {%A}'OMP_DISPLAY_ENV: TRUEOMP_PLACES: threads
Run 1x64	Number processes: 1Run Command: <executable> -g "4 2 2" -bMPI Command: mpirun -n <number_processes> --bind-to core --map-by package:PE=64 --rank-by fill --report-bindings Dataset: Run Directory: /home/eoseret/qaas/qaas_runs/178-212-9071/intel/miniqmc/run/oneview_runs/multicore/armclang/oneview_run_1782144418OMP_NUM_THREADS: 64OMP_PROC_BIND: spreadOMP_DISPLAY_AFFINITY: TRUEOMP_AFFINITY_FORMAT: 'OMP: pid %P tid %i thread %n bound to OS proc set {%A}'OMP_DISPLAY_ENV: TRUEOMP_PLACES: threads

(1x1) Efficiency	(1x1) Potential Speed-Up (%)	(1x2) Efficiency	(1x2) Potential Speed-Up (%)	(1x4) Efficiency	(1x4) Potential Speed-Up (%)	(1x8) Efficiency	(1x8) Potential Speed-Up (%)	(1x16) Efficiency	(1x16) Potential Speed-Up (%)	(1x24) Efficiency	(1x24) Potential Speed-Up (%)	(1x32) Efficiency	(1x32) Potential Speed-Up (%)	(1x40) Efficiency	(1x40) Potential Speed-Up (%)	(1x48) Efficiency	(1x48) Potential Speed-Up (%)	(1x56) Efficiency	(1x56) Potential Speed-Up (%)	(1x64) Efficiency	(1x64) Potential Speed-Up (%)
1	0	2	0	3.99	0	2.66	0			11.96	0	3.19	0	7.95	0	5.3	0	13.9	0	15.89	0

Speed-Up and Efficiency

▶Detailed Speed-Up and Efficiency

Run	Number of threads	Efficiency (ideal is 1)	Speedup	Ideal Speedup	Time (s)	Coverage (%)
1x1	1	1	1	1	0.0049999998882413	0.011381772346795
1x2	1	2	2	2	0.00499999942258	0.0056465421803296
1x4	1	3.99	3.99	4	0.0049999998882413	0.0027647295501083
1x8	3	2.66	2.66	8	0.0050000003539026	0.0038556014187634
1x16
1x24	1	11.96	11.96	24	0.0099999997764826	0.00053075782489032
1x32	5	3.19	3.19	32	0.0099999997764826	0.0016972656594589
1x40	1	7.95	7.95	40	0.024999998509884	0.00058711809106171
1x48	3	5.3	5.3	48	0.015000000596046	0.0007708488847129
1x56	1	13.9	13.9	56	0.019999999552965	0.00026210828218609
1x64	1	15.89	15.89	64	0.02000000141561	0.00020718551240861

Report Configuration

Vectorization

Details

Workaround

CALL instructions

Details

Slow data structures access

Details

Workaround

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code

Vectorization

Details

Workaround

CALL instructions

Details

Slow data structures access

Details

Workaround

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code

▶Scalability Runs Description

▼Columns Filter

Speed-Up and Efficiency

▶Detailed Speed-Up and Efficiency