OV - exec - Function qmcplusplus::WaveFunction::ratioGrad(qmcplusplus::ParticleSet&, int, qmcplusplus::TinyVector<double, 3u>&)

	Function: qmcplusplus::WaveFunction::ratioGrad(qmcplusplus::ParticleSet&, int, qmcplusplus::TinyVect ...	Module: exec	Source: WaveFunction.cpp:200-210 [...]	Coverage: 0.02%

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/PETE/OperatorTags.h: 183 - 183

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

183:     return (const_cast<T1&>(a) = b);

/usr/include/c++/11/bits/refwrap.h: 338 - 338

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

338:       { return *_M_data; }

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/WaveFunction.cpp: 200 - 210

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

200: {

201:   grad       = valT(0);

202:   valT ratio = (iat < nelup ? Det_up->ratioGrad(P, iat, grad) : Det_dn->ratioGrad(P, iat, grad));

203:

204:   for (size_t i = 0; i < Jastrows.size(); i++)

205:   {

206:     ScopedTimer local_timer(jastrow_timers[i]);

207:     ratio *= Jastrows[i]->ratioGrad(P, iat, grad);

208:   }

209:   return ratio;

210: }

/usr/include/c++/11/bits/stl_vector.h: 919 - 919

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

919:       { return size_type(this->_M_impl._M_finish - this->_M_impl._M_start); }

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Utilities/NewTimer.h: 242 - 249

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

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

[...]

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

0x40d680 STP	X29, X30, [SP, #944]!

0x40d684 ADD	X29, SP, #0

0x40d688 STP	X21, X22, [SP, #32]

0x40d68c ORR	X22, XZR, X0

0x40d690 LDR	W0, [X0, #52]

0x40d694 STP	X19, X20, [SP, #16]

0x40d698 STP	X23, X24, [SP, #48]

0x40d69c ORR	X23, XZR, X3

0x40d6a0 ORR	W24, WZR, W2

0x40d6a4 STR	X25, [SP, #64]

0x40d6a8 ORR	X25, XZR, X1

0x40d6ac STR	D8, [SP, #72]

0x40d6b0 STR	XZR, [X3]

0x40d6b4 STP	XZR, XZR, [X3, #8]

0x40d6b8 CMP	W0, W2

0x40d6bc B.LE	40d760

0x40d6c0 LDR	X0, [X22]

0x40d6c4 LDR	X6, [X0]

0x40d6c8 LDR	X7, [X6, #32]

0x40d6cc BLR	X7

0x40d6d0 FMOV	D8, D0

(136) 0x40d6d4 LDR	X2, [X22, #16]

(136) 0x40d6d8 ADD	X20, X22, #16

(136) 0x40d6dc MOVZ	X19, #0

(136) 0x40d6e0 ADD	X22, X22, #88

(136) 0x40d6e4 LDR	X1, [X20, #8]

(136) 0x40d6e8 CMP	X1, X2

(136) 0x40d6ec B.EQ	40d740

  (137) 0x40d6f0 LDR	X3, [X22]

  (137) 0x40d6f4 LDR	X21, [X3, X19,LSL #3]

  (137) 0x40d6f8 ORR	X0, XZR, X21

  (137) 0x40d6fc BL	479f20

  (137) 0x40d700 LDR	X8, [X20]

  (137) 0x40d704 ORR	X3, XZR, X23

  (137) 0x40d708 ORR	W2, WZR, W24

  (137) 0x40d70c ORR	X1, XZR, X25

  (137) 0x40d710 LDR	X0, [X8, X19,LSL #3]

  (137) 0x40d714 LDR	X9, [X0]

  (137) 0x40d718 LDR	X10, [X9, #32]

  (137) 0x40d71c BLR	X10

  (137) 0x40d720 FMUL	D8, D8, D0

  (137) 0x40d724 ORR	X0, XZR, X21

  (137) 0x40d728 ADD	X19, X19, #1

  (137) 0x40d72c BL	47a180

  (137) 0x40d730 LDP	X11, X12, [X20]

  (137) 0x40d734 SUB	X13, X12, X11

  (137) 0x40d738 CMP	X19, X13,ASR #3

  (137) 0x40d73c B.CC	40d6f0

(136) 0x40d740 FMOV	D0, D8

(136) 0x40d744 LDP	X19, X20, [SP, #16]

(136) 0x40d748 LDP	X21, X22, [SP, #32]

(136) 0x40d74c LDP	X23, X24, [SP, #48]

(136) 0x40d750 LDR	X25, [SP, #64]

(136) 0x40d754 LDR	D8, [SP, #72]

(136) 0x40d758 LDP	X29, X30, [SP], #80

(136) 0x40d75c RET

(136) 0x40d760 LDR	X0, [X22, #8]

(136) 0x40d764 LDR	X4, [X0]

(136) 0x40d768 LDR	X5, [X4, #32]

(136) 0x40d76c BLR	X5

(136) 0x40d770 FMOV	D8, D0

(136) 0x40d774 B	40d6d4

0x40d778 ORR	X23, XZR, X0

0x40d77c ORR	X0, XZR, X21

0x40d780 BL	47a180

0x40d784 ORR	X0, XZR, X23

0x40d788 BL	403aa0

Coverage (%)	Name	Source Location	Module
►50.00+	main._omp_fn.1	stl_vector.h:1043	exec
○	GOMP_parallel		libomp.so
○	main	miniqmc.cpp:404	exec
○	__libc_start_main		libc-2.31.so
○	_start	miniqmc.cpp:546	exec
►50.00+	main._omp_fn.1	stl_vector.h:1043	exec
○	GOMP_parallel		libomp.so
○	main	miniqmc.cpp:402	exec
○	__libc_start_main		libc-2.31.so
○	_start	miniqmc.cpp:546	exec

Path /

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

Source file and lines	WaveFunction.cpp:200-210
Module	exec

The function is defined in:

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/PETE/OperatorTags.h:183
/usr/include/c++/11/bits/refwrap.h:338
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/WaveFunction.cpp:200-210
/usr/include/c++/11/bits/stl_vector.h:919
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Utilities/NewTimer.h:242,249

Warnings:
Ignoring paths for analysis

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 4.00 to 0.83 cycles (4.80x speedup).

Workaround

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

Vectorization

Your function is not vectorized. Only 22% of vector register length is used (average across all VPU instructions). By vectorizing your function, you can lower the cost of an iteration from 4.00 to 0.56 cycles (7.11x 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 4.00 to 1.50 cycles (2.67x 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 28 bytes. The binary function is storing 104 bytes.

General properties

nb instructions	26
loop length	104
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

3.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	2.00	2.00	3.25	3.25	3.25	3.25	0.50	0.50	0.50	0.50	4.00	4.00	4.00	3.50	3.50
cycles	2.00	2.00	3.25	3.25	3.25	3.25	0.50	0.50	0.50	0.50	4.00	4.00	4.00	3.50	3.50

Cycles executing div or sqrt instructions

Cycles summary

Front-end	3.25
Overall L1	4.00

Vectorization ratios

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 4.00 cycles. At this rate:

7% of peak load performance is reached (7.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
54% of peak store performance is reached (26.00 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the function is: 40d680

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput
STP X29, X30, [SP, #944]!	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
ADD X29, SP, #0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STP X21, X22, [SP, #32]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
ORR X22, XZR, X0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR W0, [X0, #52]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
STP X19, X20, [SP, #16]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
STP X23, X24, [SP, #48]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
ORR X23, XZR, X3	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ORR W24, WZR, W2	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STR X25, [SP, #64]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
ORR X25, XZR, X1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STR D8, [SP, #72]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
STR XZR, [X3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
STP XZR, XZR, [X3, #8]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
CMP W0, W2	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
B.LE 40d760 <_ZN11qmcplusplus12WaveFunction9ratioGradERNS_11ParticleSetEiRNS_10TinyVectorIdLj3EEE+0xe0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X0, [X22]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X6, [X0]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X7, [X6, #32]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
BLR X7	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FMOV D8, D0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
ORR X23, XZR, X0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ORR X0, XZR, X21	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
BL 47a180 <_ZN11qmcplusplus9TimerTypeINS_8CPUClockEE4stopEv>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ORR X0, XZR, X23	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
BL 403aa0 <@plt_start@+0x420>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50

Source file and lines	WaveFunction.cpp:200-210
Module	exec

The function is defined in:

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/PETE/OperatorTags.h:183
/usr/include/c++/11/bits/refwrap.h:338
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/WaveFunction.cpp:200-210
/usr/include/c++/11/bits/stl_vector.h:919
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Utilities/NewTimer.h:242,249

Warnings:
Ignoring paths for analysis

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 4.00 to 0.83 cycles (4.80x speedup).

Workaround

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

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 4.00 to 1.50 cycles (2.67x 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 28 bytes. The binary function is storing 104 bytes.

General properties

nb instructions	26
loop length	104
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

3.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	2.00	2.00	3.25	3.25	3.25	3.25	0.50	0.50	0.50	0.50	4.00	4.00	4.00	3.50	3.50
cycles	2.00	2.00	3.25	3.25	3.25	3.25	0.50	0.50	0.50	0.50	4.00	4.00	4.00	3.50	3.50

Cycles executing div or sqrt instructions

Cycles summary

Front-end	3.25
Overall L1	4.00

Vectorization ratios

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 4.00 cycles. At this rate:

7% of peak load performance is reached (7.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
54% of peak store performance is reached (26.00 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the function is: 40d680

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput
STP X29, X30, [SP, #944]!	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
ADD X29, SP, #0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STP X21, X22, [SP, #32]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
ORR X22, XZR, X0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR W0, [X0, #52]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
STP X19, X20, [SP, #16]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
STP X23, X24, [SP, #48]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
ORR X23, XZR, X3	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ORR W24, WZR, W2	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STR X25, [SP, #64]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
ORR X25, XZR, X1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STR D8, [SP, #72]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
STR XZR, [X3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
STP XZR, XZR, [X3, #8]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
CMP W0, W2	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
B.LE 40d760 <_ZN11qmcplusplus12WaveFunction9ratioGradERNS_11ParticleSetEiRNS_10TinyVectorIdLj3EEE+0xe0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X0, [X22]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X6, [X0]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X7, [X6, #32]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
BLR X7	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FMOV D8, D0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
ORR X23, XZR, X0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ORR X0, XZR, X21	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
BL 47a180 <_ZN11qmcplusplus9TimerTypeINS_8CPUClockEE4stopEv>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ORR X0, XZR, X23	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
BL 403aa0 <@plt_start@+0x420>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50

Name	Coverage (%)	Time (s)
▼qmcplusplus::WaveFunction::ratioGrad(qmcplusplus::ParticleSet&, int, qmcplusplus::TinyVector&)–	0.02	0.01
▼Loop 136 - WaveFunction.cpp:202-210 - exec–	0	0
○Loop 137 - WaveFunction.cpp:204-207 - exec	0.02	0.01

Report Configuration

Code clean check

Workaround

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

Code clean check

Workaround

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