OV - exec - Function miniqmcreference::DiracDeterminantRef<qmcplusplus::DelayedUpdate<double, double> >::acceptMove(qmcplusplus::ParticleSet&, int)

	Function: miniqmcreference::DiracDeterminantRef<qmcplusplus::DelayedUpdate<double, double> >::accept ...	Module: exec	Source: DiracDeterminantRef.cpp:124-139 [...]	Coverage: 0.19%

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/DiracDeterminantRef.cpp: 124 - 139

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

124: {

125:   const int WorkingIndex = iat - FirstIndex;

126:   PhaseValue += evaluatePhase(curRatio);

127:   LogValue += std::log(std::abs(curRatio));

128:   UpdateTimer->start();

129:   updateEng.acceptRow(psiM, WorkingIndex, psiV);

130:   // invRow becomes invalid after accepting a move

131:   invRow_id = -1;

132:   if (UpdateMode == ORB_PBYP_PARTIAL)

133:   {

134:     simd::copy(dpsiM[WorkingIndex], dpsiV.data(), NumOrbitals);

135:     simd::copy(d2psiM[WorkingIndex], d2psiV.data(), NumOrbitals);

136:   }

137:   UpdateTimer->stop();

138:   curRatio = 1.0;

139: }

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsMatrix.h: 217 - 217

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

217:   inline Type_t* operator[](size_type i) { return X.data() + i * D2; }

/usr/lib/gcc/aarch64-linux-gnu/11/../../../../include/c++/11/bits/std_abs.h: 72 - 72

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

72:   { return __builtin_fabs(__x); }

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h: 248 - 248

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

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

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Platforms/CPU/SIMD/inner_product.hpp: 52 - 52

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

52:   memcpy(target, source, sizeof(T) * n);

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/DeterminantHelper.h: 23 - 23

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

23:   return T((sign_v > 0) ? 0.0 : M_PI);

0x439d80 SUB	SP, SP, #64

0x439d84 STP	X29, X30, [SP, #16]

0x439d88 STR	X21, [SP, #32]

0x439d8c STP	X20, X19, [SP, #48]

0x439d90 ADD	X29, SP, #16

0x439d94 LDRSW	X8, [X0, #1144]

0x439d98 SBFM	X9, X2, #0, #31

0x439d9c LDR	D0, [X0, #1080]

0x439da0 MOVI	D2, #0

0x439da4 ORR	X19, XZR, X0

0x439da8 SUB	X20, X9, X8

0x439dac ADRP	X8,

0x439db0 LDR	D1, [X8, #2368]

0x439db4 FCMP	D0, #0

0x439db8 FABS	D0, D0

0x439dbc FCSEL	D1, D2, D1, #12

0x439dc0 STR	Q1, [SP]

0x439dc4 BL	403eb0

0x439dc8 LDR	Q2, [SP]

0x439dcc LDR	Q1, [X19, #16]

0x439dd0 LDR	X0, [X19, #1088]

0x439dd4 MOV	V0.D[1], V2.D[0]

0x439dd8 FADD	V0.2D, V1.2D, V0.2D

0x439ddc STR	Q0, [X19, #16]

0x439de0 BL	44ec60

0x439de4 ORR	W2, WZR, W20

0x439de8 ADD	X0, X19, #520

0x439dec ADD	X1, X19, #208

0x439df0 ADD	X3, X19, #400

0x439df4 BL	4380a0

0x439df8 MOVN	W8, #0

0x439dfc STR	W8, [X19, #1072]

0x439e00 LDR	W8, [X19, #12]

0x439e04 CMP	W8, #2

0x439e08 B.NE	439eb4

0x439e0c LDR	X9, [X19, #280]

0x439e10 LDR	X8, [X19, #320]

0x439e14 MOVZ	W10, #24

0x439e18 LDR	X1, [X19, #464]

0x439e1c RDVL	X21, #1

0x439e20 CMP	X21, #17

0x439e24 MADD	X9, X9, X20, XZR

0x439e28 MADD	X0, X9, X10, X8

0x439e2c LDRSW	X8, [X19, #1152]

0x439e30 ADD	X8, X8, X8,LSL #1

0x439e34 UBFM	X2, X8, #61, #60

0x439e38 B.CS	439e44

0x439e3c BL	403a40

0x439e40 B	439e60

0x439e44 WHILELO	P0.B, XZR, X2

0x439e48 ORR	X8, XZR, XZR

(1056) 0x439e4c LD1B	{Z0.B}, P0/Z, [X1, X8]

(1056) 0x439e50 ST1B	{Z0.B}, P0, [X0, X8]

(1056) 0x439e54 INCP	X8, P0.B

(1056) 0x439e58 WHILELO	P0.B, X8, X2

(1056) 0x439e5c B.MI	439e4c

0x439e60 LDR	X9, [X19, #344]

0x439e64 LDR	X8, [X19, #384]

0x439e68 LDR	X1, [X19, #504]

0x439e6c CMP	X21, #17

0x439e70 MADD	X9, X9, X20, XZR

0x439e74 ADD	X0, X8, X9,LSL #3

0x439e78 LDRSW	X8, [X19, #1152]

0x439e7c UBFM	X2, X8, #61, #60

0x439e80 B.CS	439e8c

0x439e84 BL	403a40

0x439e88 B	439eb4

0x439e8c WHILELO	P0.B, XZR, X2

0x439e90 ORR	X8, XZR, XZR

0x439e94 HINT	#0

0x439e98 HINT	#0

0x439e9c HINT	#0

(1057) 0x439ea0 LD1B	{Z0.B}, P0/Z, [X1, X8]

(1057) 0x439ea4 ST1B	{Z0.B}, P0, [X0, X8]

(1057) 0x439ea8 INCP	X8, P0.B

(1057) 0x439eac WHILELO	P0.B, X8, X2

(1057) 0x439eb0 B.MI	439ea0

0x439eb4 LDR	X0, [X19, #1088]

0x439eb8 BL	44ed80

0x439ebc MOVZ	X8, #16368

0x439ec0 STR	X8, [X19, #1080]

0x439ec4 LDP	X20, X19, [SP, #48]

0x439ec8 LDR	X21, [SP, #32]

0x439ecc LDP	X29, X30, [SP, #16]

0x439ed0 ADD	SP, SP, #64

0x439ed4 RET

0x439ed8 HINT	#0

0x439edc HINT	#0

Coverage (%)	Name	Source Location	Module
►100.00+	qmcplusplus::WaveFunction::acc[...]	stl_vector.h:919	exec
○	.omp_outlined..62	miniqmc.cpp:445	exec
○	__kmp_invoke_microtask		libomp.so

Path /

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

Source file and lines	DiracDeterminantRef.cpp:124-139
Module	exec

The function is defined in:

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/DiracDeterminantRef.cpp:124-139
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsMatrix.h:217
/usr/lib/gcc/aarch64-linux-gnu/11/../../../../include/c++/11/bits/std_abs.h:72
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h:248
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Platforms/CPU/SIMD/inner_product.hpp:52
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/DeterminantHelper.h:23

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 9.13 to 2.00 cycles (4.56x speedup).

Workaround

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

Vectorization

Your function is poorly vectorized. Only 29% of vector register length is used (average across all VPU instructions). By fully vectorizing your function, you can lower the cost of an iteration from 9.13 to 1.69 cycles (5.41x speedup).

Details

25% of VPU instructions are used in vector version (process two or more data elements in vector registers):

50% of VPU loads are used in vector version.
20% of VPU addition or subtraction instructions are used in vector version.
0% of VPU fused multiply-add instructions are used in vector version.
0% of VPU instructions that are not load, store, addition, subtraction nor multiply instructions are used in vector version.

Since your execution units are vector units, only a fully 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 9.13 to 3.00 cycles (3.04x speedup).

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

The binary function is composed of 2 FP arithmetical operations:

2: addition or subtraction

The binary function is loading 168 bytes. The binary function is storing 84 bytes.

Arithmetic intensity

Arithmetic intensity is 0.01 FP operations per loaded or stored byte.

General properties

nb instructions	78
loop length	312
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

9.13 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	6.00	6.00	8.75	8.75	8.75	8.75	2.00	2.00	2.00	2.00	8.83	8.50	8.67	2.50	2.50
cycles	6.00	6.00	8.75	8.75	8.75	8.75	2.00	2.00	2.00	2.00	8.83	8.50	8.67	2.50	2.50

Cycles executing div or sqrt instructions

Cycles summary

Front-end	9.13
Overall L1	9.13

Vectorization ratios

INT

all	21%
load	50%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	0%
other	0%

all	100%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	25%
load	50%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	20%
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 9.13 cycles. At this rate:

19% of peak load performance is reached (18.41 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
19% of peak store performance is reached (9.21 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the function is: 439d80

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput
SUB SP, SP, #64	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STP X29, X30, [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
STR X21, [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
STP X20, X19, [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
ADD X29, SP, #16	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDRSW X8, [X0, #1144]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
SBFM X9, X2, #0, #31	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR D0, [X0, #1080]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
MOVI D2, #0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
ORR X19, 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
SUB X20, X9, X8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ADRP X8, <465dac>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR D1, [X8, #2368]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
FCMP D0, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1
FABS D0, D0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
FCSEL D1, D2, D1, #12	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	0	0	2	0.50
STR Q1, [SP]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
BL 403eb0 <@plt_start@+0x580>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR Q2, [SP]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
LDR Q1, [X19, #16]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
LDR X0, [X19, #1088]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
MOV V0.D[1], V2.D[0]	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
FADD V0.2D, V1.2D, V0.2D	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
STR Q0, [X19, #16]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
BL 44ec60 <_ZN11qmcplusplus9TimerTypeINS_8CPUClockEE5startEv>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ORR W2, WZR, W20	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ADD X0, X19, #520	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ADD X1, X19, #208	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ADD X3, X19, #400	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
BL 4380a0 <_ZN11qmcplusplus13DelayedUpdateIddE9acceptRowINS_6VectorIdSaIdEEEEEvRNS_6MatrixIdS4_EEiRKT_>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
MOVN W8, #0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STR W8, [X19, #1072]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
LDR W8, [X19, #12]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMP W8, #2	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
B.NE 439eb4 <_ZN16miniqmcreference19DiracDeterminantRefIN11qmcplusplus13DelayedUpdateIddEEE10acceptMoveERNS1_11ParticleSetEi+0x134>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X9, [X19, #280]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X8, [X19, #320]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
MOVZ W10, #24	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X1, [X19, #464]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
RDVL X21, #1	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1
CMP X21, #17	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
MADD X9, X9, X20, XZR	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1
MADD X0, X9, X10, X8	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1
LDRSW X8, [X19, #1152]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
ADD X8, X8, X8,LSL #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
UBFM X2, X8, #61, #60	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
B.CS 439e44 <_ZN16miniqmcreference19DiracDeterminantRefIN11qmcplusplus13DelayedUpdateIddEEE10acceptMoveERNS1_11ParticleSetEi+0xc4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
BL 403a40 <@plt_start@+0x110>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
B 439e60 <_ZN16miniqmcreference19DiracDeterminantRefIN11qmcplusplus13DelayedUpdateIddEEE10acceptMoveERNS1_11ParticleSetEi+0xe0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
WHILELO P0.B, XZR, X2	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	3	2
ORR X8, XZR, XZR	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X9, [X19, #344]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X8, [X19, #384]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X1, [X19, #504]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMP X21, #17	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
MADD X9, X9, X20, XZR	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1
ADD X0, X8, X9,LSL #3	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDRSW X8, [X19, #1152]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
UBFM X2, X8, #61, #60	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
B.CS 439e8c <_ZN16miniqmcreference19DiracDeterminantRefIN11qmcplusplus13DelayedUpdateIddEEE10acceptMoveERNS1_11ParticleSetEi+0x10c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
BL 403a40 <@plt_start@+0x110>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
B 439eb4 <_ZN16miniqmcreference19DiracDeterminantRefIN11qmcplusplus13DelayedUpdateIddEEE10acceptMoveERNS1_11ParticleSetEi+0x134>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
WHILELO P0.B, XZR, X2	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	3	2
ORR X8, XZR, XZR	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
HINT #0
HINT #0
HINT #0
LDR X0, [X19, #1088]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
BL 44ed80 <_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
MOVZ X8, #16368	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STR X8, [X19, #1080]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
LDP X20, X19, [SP, #48]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
LDR X21, [SP, #32]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDP X29, X30, [SP, #16]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
ADD SP, SP, #64	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
RET	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
HINT #0
HINT #0

Source file and lines	DiracDeterminantRef.cpp:124-139
Module	exec

The function is defined in:

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/DiracDeterminantRef.cpp:124-139
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsMatrix.h:217
/usr/lib/gcc/aarch64-linux-gnu/11/../../../../include/c++/11/bits/std_abs.h:72
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Numerics/OhmmsPETE/OhmmsVector.h:248
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Platforms/CPU/SIMD/inner_product.hpp:52
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/QMCWaveFunctions/DeterminantHelper.h:23

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 9.13 to 2.00 cycles (4.56x speedup).

Workaround

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

Vectorization

Details

25% of VPU instructions are used in vector version (process two or more data elements in vector registers):

50% of VPU loads are used in vector version.
20% of VPU addition or subtraction instructions are used in vector version.
0% of VPU fused multiply-add instructions are used in vector version.
0% of VPU instructions that are not load, store, addition, subtraction nor multiply instructions are used in vector version.

Since your execution units are vector units, only a fully 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 9.13 to 3.00 cycles (3.04x speedup).

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

The binary function is composed of 2 FP arithmetical operations:

2: addition or subtraction

The binary function is loading 168 bytes. The binary function is storing 84 bytes.

Arithmetic intensity

Arithmetic intensity is 0.01 FP operations per loaded or stored byte.

General properties

nb instructions	78
loop length	312
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

9.13 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	6.00	6.00	8.75	8.75	8.75	8.75	2.00	2.00	2.00	2.00	8.83	8.50	8.67	2.50	2.50
cycles	6.00	6.00	8.75	8.75	8.75	8.75	2.00	2.00	2.00	2.00	8.83	8.50	8.67	2.50	2.50

Cycles executing div or sqrt instructions

Cycles summary

Front-end	9.13
Overall L1	9.13

Vectorization ratios

INT

all	21%
load	50%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	0%
other	0%

all	100%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	25%
load	50%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	20%
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 9.13 cycles. At this rate:

19% of peak load performance is reached (18.41 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
19% of peak store performance is reached (9.21 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the function is: 439d80

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput
SUB SP, SP, #64	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STP X29, X30, [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
STR X21, [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
STP X20, X19, [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
ADD X29, SP, #16	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDRSW X8, [X0, #1144]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
SBFM X9, X2, #0, #31	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR D0, [X0, #1080]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
MOVI D2, #0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
ORR X19, 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
SUB X20, X9, X8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ADRP X8, <465dac>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR D1, [X8, #2368]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
FCMP D0, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1
FABS D0, D0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
FCSEL D1, D2, D1, #12	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	0	0	2	0.50
STR Q1, [SP]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
BL 403eb0 <@plt_start@+0x580>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR Q2, [SP]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
LDR Q1, [X19, #16]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
LDR X0, [X19, #1088]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
MOV V0.D[1], V2.D[0]	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
FADD V0.2D, V1.2D, V0.2D	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
STR Q0, [X19, #16]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
BL 44ec60 <_ZN11qmcplusplus9TimerTypeINS_8CPUClockEE5startEv>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ORR W2, WZR, W20	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ADD X0, X19, #520	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ADD X1, X19, #208	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ADD X3, X19, #400	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
BL 4380a0 <_ZN11qmcplusplus13DelayedUpdateIddE9acceptRowINS_6VectorIdSaIdEEEEEvRNS_6MatrixIdS4_EEiRKT_>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
MOVN W8, #0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STR W8, [X19, #1072]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
LDR W8, [X19, #12]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMP W8, #2	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
B.NE 439eb4 <_ZN16miniqmcreference19DiracDeterminantRefIN11qmcplusplus13DelayedUpdateIddEEE10acceptMoveERNS1_11ParticleSetEi+0x134>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X9, [X19, #280]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X8, [X19, #320]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
MOVZ W10, #24	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X1, [X19, #464]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
RDVL X21, #1	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1
CMP X21, #17	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
MADD X9, X9, X20, XZR	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1
MADD X0, X9, X10, X8	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1
LDRSW X8, [X19, #1152]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
ADD X8, X8, X8,LSL #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
UBFM X2, X8, #61, #60	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
B.CS 439e44 <_ZN16miniqmcreference19DiracDeterminantRefIN11qmcplusplus13DelayedUpdateIddEEE10acceptMoveERNS1_11ParticleSetEi+0xc4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
BL 403a40 <@plt_start@+0x110>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
B 439e60 <_ZN16miniqmcreference19DiracDeterminantRefIN11qmcplusplus13DelayedUpdateIddEEE10acceptMoveERNS1_11ParticleSetEi+0xe0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
WHILELO P0.B, XZR, X2	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	3	2
ORR X8, XZR, XZR	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X9, [X19, #344]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X8, [X19, #384]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR X1, [X19, #504]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMP X21, #17	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
MADD X9, X9, X20, XZR	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1
ADD X0, X8, X9,LSL #3	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDRSW X8, [X19, #1152]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
UBFM X2, X8, #61, #60	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
B.CS 439e8c <_ZN16miniqmcreference19DiracDeterminantRefIN11qmcplusplus13DelayedUpdateIddEEE10acceptMoveERNS1_11ParticleSetEi+0x10c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
BL 403a40 <@plt_start@+0x110>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
B 439eb4 <_ZN16miniqmcreference19DiracDeterminantRefIN11qmcplusplus13DelayedUpdateIddEEE10acceptMoveERNS1_11ParticleSetEi+0x134>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
WHILELO P0.B, XZR, X2	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	3	2
ORR X8, XZR, XZR	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
HINT #0
HINT #0
HINT #0
LDR X0, [X19, #1088]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
BL 44ed80 <_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
MOVZ X8, #16368	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STR X8, [X19, #1080]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
LDP X20, X19, [SP, #48]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
LDR X21, [SP, #32]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDP X29, X30, [SP, #16]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
ADD SP, SP, #64	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
RET	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
HINT #0
HINT #0

Name	Coverage (%)	Time (s)
▼miniqmcreference::DiracDeterminantRef >::acceptMove(qmcplusplus::ParticleSet&, int)–	0.19	0.09
○Loop 1056 - inner_product.hpp:52-52 - exec	0.17	0.08
○Loop 1057 - inner_product.hpp:52-52 - 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

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

Expensive FP math instructions/calls

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code