OV - exec - Loop 1373

0x44ef10 LDR	X11, [X10, #32]    [1]

0x44ef14 ORR	X12, XZR, X8

0x44ef18 CMP	X11, X8

0x44ef1c B.NE	44ef30

0x44ef20 LDR	X11, [X10, #40]    [1]

0x44ef24 ORR	X12, XZR, X9

0x44ef28 CMP	X11, X9

0x44ef2c B.EQ	44ef00

0x44ef30 CMP	X11, X12

0x44ef34 B.GE	44ef00

0x44ef38 ADD	X11, X10, #24

0x44ef3c LDR	X10, [X11]    [2]

0x44ef40 CBNZ	X10, 44ef10

/usr/lib/gcc/aarch64-linux-gnu/11/../../../../include/c++/11/bits/stl_tree.h: 788 - 1903

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

788:       { return static_cast<_Link_type>(__x->_M_right); }

[...]

1902:       while (__x != 0)

1903: 	if (!_M_impl._M_key_compare(_S_key(__x), __k))

/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Utilities/NewTimer.h: 119 - 119

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

119:       if (!(this->long_buckets[j] == rhs.long_buckets[j]))

Coverage (%)	Name	Source Location	Module
►40.00+	qmcplusplus::WaveFunction::rat[...]	stl_vector.h:919	exec
○	.omp_outlined..62	miniqmc.cpp:437	exec
○	__kmp_invoke_microtask		libomp.so
►40.00+	qmcplusplus::WaveFunction::rat[...]	stl_vector.h:919	exec
○	.omp_outlined..62	miniqmc.cpp:485	exec
○	__kmp_invoke_microtask		libomp.so
►20.00+	qmcplusplus::ParticleSet::make[...]	ParticleSet.cpp:292	exec
○	.omp_outlined..62	miniqmc.cpp:482	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	4.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	2.86
Bottlenecks
Function	qmcplusplus::TimerType::stop()
Source	stl_tree.h:788-788,stl_tree.h:1902-1903,NewTimer.h:119-119
Source loop unroll info	multi-versionned
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	5.00
CQA cycles if no scalar integer	5.00
CQA cycles if FP arith vectorized	5.00
CQA cycles if fully vectorized	1.25
Front-end cycles	1.38
DIV/SQRT cycles	1.75
P0 cycles	1.75
P1 cycles	1.25
P2 cycles	1.25
P3 cycles	1.25
P4 cycles	1.25
P5 cycles	0.00
P6 cycles	0.00
P7 cycles	0.00
P8 cycles	0.00
P9 cycles	0.83
P10 cycles	0.83
P11 cycles	0.83
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	0.00
Inter-iter dependencies cycles	5
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	11.00
Nb uops	11.00
Nb loads	NA
Nb stores	0.00
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	4.00
Bytes prefetched	0.00
Bytes loaded	20.00
Bytes stored	0.00
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	25.00
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	25.00

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	4.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	2.50
Bottlenecks
Function	qmcplusplus::TimerType::stop()
Source	stl_tree.h:788-788,stl_tree.h:1902-1903,NewTimer.h:119-119
Source loop unroll info	multi-versionned
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	5.00
CQA cycles if no scalar integer	5.00
CQA cycles if FP arith vectorized	5.00
CQA cycles if fully vectorized	1.25
Front-end cycles	1.63
DIV/SQRT cycles	2.00
P0 cycles	2.00
P1 cycles	1.50
P2 cycles	1.50
P3 cycles	1.50
P4 cycles	1.50
P5 cycles	0.00
P6 cycles	0.00
P7 cycles	0.00
P8 cycles	0.00
P9 cycles	1.00
P10 cycles	1.00
P11 cycles	1.00
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	0.00
Inter-iter dependencies cycles	5
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	13.00
Nb uops	13.00
Nb loads	NA
Nb stores	0.00
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	4.80
Bytes prefetched	0.00
Bytes loaded	24.00
Bytes stored	0.00
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	25.00
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	25.00

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	4.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	3.33
Bottlenecks
Function	qmcplusplus::TimerType::stop()
Source	stl_tree.h:788-788,stl_tree.h:1902-1903,NewTimer.h:119-119
Source loop unroll info	multi-versionned
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	5.00
CQA cycles if no scalar integer	5.00
CQA cycles if FP arith vectorized	5.00
CQA cycles if fully vectorized	1.25
Front-end cycles	1.13
DIV/SQRT cycles	1.50
P0 cycles	1.50
P1 cycles	1.00
P2 cycles	1.00
P3 cycles	1.00
P4 cycles	1.00
P5 cycles	0.00
P6 cycles	0.00
P7 cycles	0.00
P8 cycles	0.00
P9 cycles	0.67
P10 cycles	0.67
P11 cycles	0.67
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	0.00
Inter-iter dependencies cycles	5
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	9.00
Nb uops	9.00
Nb loads	NA
Nb stores	0.00
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	3.20
Bytes prefetched	0.00
Bytes loaded	16.00
Bytes stored	0.00
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	25.00
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	25.00

Path /

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

Function	qmcplusplus::TimerType::stop()
Source file and lines	stl_tree.h:788-1903
Module	exec

The loop is defined in:

/usr/lib/gcc/aarch64-linux-gnu/11/../../../../include/c++/11/bits/stl_tree.h:788,1902-1903
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Utilities/NewTimer.h:119

The related source loop is multi-versionned.
The structure of this loop is probably <if then [else] end>.

The presence of multiple execution paths is typically the main/first bottleneck.
Try to simplify control inside loop: ideally, try to remove all conditional expressions, for example by (if applicable):

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. 4 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 5.00 to 1.25 cycles (4.00x 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

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 20 bytes.

General properties

nb instructions	11
loop length	44
nb stack references	0

Front-end

front end

1.38 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	1.75	1.75	1.25	1.25	1.25	1.25	0.00	0.00	0.00	0.00	0.83	0.83	0.83	0.00	0.00
cycles	1.75	1.75	1.25	1.25	1.25	1.25	0.00	0.00	0.00	0.00	0.83	0.83	0.83	0.00	0.00

Cycles executing div or sqrt instructions	NA
Longest recurrence chain latency (RecMII)	5.00

Cycles summary

Front-end	1.38
Data deps.	5.00
Overall L1	5.00

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 5.00 cycles. At this rate:

4% of peak load performance is reached (4.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))

Function	qmcplusplus::TimerType::stop()
Source file and lines	stl_tree.h:788-1903
Module	exec

The loop is defined in:

/usr/lib/gcc/aarch64-linux-gnu/11/../../../../include/c++/11/bits/stl_tree.h:788,1902-1903
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Utilities/NewTimer.h:119

The related source loop is multi-versionned.
The structure of this loop is probably <if then [else] end>.

The presence of multiple execution paths is typically the main/first bottleneck.
Try to simplify control inside loop: ideally, try to remove all conditional expressions, for example by (if applicable):

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. 4 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 5.00 to 1.25 cycles (4.00x 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

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 24 bytes.

General properties

nb instructions	13
loop length	52
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

1.63 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	2.00	2.00	1.50	1.50	1.50	1.50	0.00	0.00	0.00	0.00	1.00	1.00	1.00	0.00	0.00
cycles	2.00	2.00	1.50	1.50	1.50	1.50	0.00	0.00	0.00	0.00	1.00	1.00	1.00	0.00	0.00

Cycles executing div or sqrt instructions	NA
Longest recurrence chain latency (RecMII)	5.00

Cycles summary

Front-end	1.63
Data deps.	5.00
Overall L1	5.00

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 5.00 cycles. At this rate:

5% of peak load performance is reached (4.80 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 44ef10

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P10	P11	P12	Latency	Recip. throughput
LDR X11, [X10, #32]	1	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
ORR X12, XZR, X8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
CMP X11, X8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.33
B.NE 44ef30 <_ZN11qmcplusplus9TimerTypeINS_8CPUClockEE4stopEv+0xd0>	1	0.50	0.50	0	0	0	0	0	0	0	1	0.50
LDR X11, [X10, #40]	1	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
ORR X12, XZR, X9	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
CMP X11, X9	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.33
B.EQ 44ef00 <_ZN11qmcplusplus9TimerTypeINS_8CPUClockEE4stopEv+0xa0>	1	0.50	0.50	0	0	0	0	0	0	0	1	0.50
CMP X11, X12	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.33
B.GE 44ef00 <_ZN11qmcplusplus9TimerTypeINS_8CPUClockEE4stopEv+0xa0>	1	0.50	0.50	0	0	0	0	0	0	0	1	0.50
ADD X11, X10, #24	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
LDR X10, [X11]	1	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
CBNZ X10, 44ef10 <_ZN11qmcplusplus9TimerTypeINS_8CPUClockEE4stopEv+0xb0>	1	0.50	0.50	0	0	0	0	0	0	0	1	0.50

Function	qmcplusplus::TimerType::stop()
Source file and lines	stl_tree.h:788-1903
Module	exec

The loop is defined in:

/usr/lib/gcc/aarch64-linux-gnu/11/../../../../include/c++/11/bits/stl_tree.h:788,1902-1903
/home/hbollore/qaas/qaas-runs/169-816-8620/intel/miniqmc/build/miniqmc/src/Utilities/NewTimer.h:119

The related source loop is multi-versionned.
The structure of this loop is probably <if then [else] end>.

The presence of multiple execution paths is typically the main/first bottleneck.
Try to simplify control inside loop: ideally, try to remove all conditional expressions, for example by (if applicable):

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. 4 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 5.00 to 1.25 cycles (4.00x 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

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 16 bytes.

General properties

nb instructions	9
loop length	36
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

1.13 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	1.50	1.50	1.00	1.00	1.00	1.00	0.00	0.00	0.00	0.00	0.67	0.67	0.67	0.00	0.00
cycles	1.50	1.50	1.00	1.00	1.00	1.00	0.00	0.00	0.00	0.00	0.67	0.67	0.67	0.00	0.00

Cycles executing div or sqrt instructions	NA
Longest recurrence chain latency (RecMII)	5.00

Cycles summary

Front-end	1.13
Data deps.	5.00
Overall L1	5.00

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 5.00 cycles. At this rate:

3% of peak load performance is reached (3.20 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 44ef10

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P10	P11	P12	Latency	Recip. throughput
LDR X11, [X10, #32]	1	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
ORR X12, XZR, X8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
CMP X11, X8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.33
B.NE 44ef30 <_ZN11qmcplusplus9TimerTypeINS_8CPUClockEE4stopEv+0xd0>	1	0.50	0.50	0	0	0	0	0	0	0	1	0.50
CMP X11, X12	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.33
B.GE 44ef00 <_ZN11qmcplusplus9TimerTypeINS_8CPUClockEE4stopEv+0xa0>	1	0.50	0.50	0	0	0	0	0	0	0	1	0.50
ADD X11, X10, #24	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
LDR X10, [X11]	1	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
CBNZ X10, 44ef10 <_ZN11qmcplusplus9TimerTypeINS_8CPUClockEE4stopEv+0xb0>	1	0.50	0.50	0	0	0	0	0	0	0	1	0.50

Report Configuration

Vectorization

Details

Workaround

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

Vectorization

Details

Workaround

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

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