OV - exec - Loop 959

	Loop Id: 959	Module: exec	Source: update_halo.f90:41-117 [...]	Coverage: 0.02%

0x450af0 MOVZ	X7, #4

0x450af4 MOVZ	X8, #1

0x450af8 MOVN	X6, #0

0x450afc MOVZ	X5, #257

0x450b00 MOVZ	X4, #15

0x450b04 ADD	X22, SP, #240

0x450b08 STP	X7, X8, [SP, #272]    [4]

0x450b0c ORR	X1, XZR, X20

0x450b10 ORR	X0, XZR, X22

0x450b14 STP	X19, X6, [SP, #240]    [4]

0x450b18 STP	X7, X5, [SP, #256]    [4]

0x450b1c STP	X8, X4, [SP, #288]    [4]

0x450b20 BL	410ee8

0x450b24 LDR	X27, [X21, #3896]    [9]

0x450b28 LDR	W9, [X27]    [16]

0x450b2c CBNZ	W9, 450db8

0x450b30 ADRP	X12,

0x450b34 LDR	X13, [X12, #3992]    [11]

0x450b38 LDR	W14, [X13, #4]    [18]

0x450b3c CMN	W14, #1

0x450b40 B.EQ	450b68

0x450b44 LDR	W15, [X13, #8]    [18]

0x450b48 CMN	W15, #1

0x450b4c B.EQ	450b68

0x450b50 LDR	W16, [X13, #12]    [18]

0x450b54 CMN	W16, #1

0x450b58 B.EQ	450b68

0x450b5c LDR	W17, [X13, #16]    [18]

0x450b60 CMN	W17, #1

0x450b64 B.NE	450c8c

0x450b68 ADRP	X18,

0x450b6c LDR	X30, [X18, #3976]    [19]

0x450b70 LDR	W2, [X30]    [2]

0x450b74 CBZ	W2, 450cb8

0x450c8c LDR	X21, [X21, #3896]    [10]

0x450c90 LDR	W19, [X21]    [3]

0x450c94 CBNZ	W19, 450de8

0x450cb8 ADRP	X0,

0x450cbc LDR	X1, [X0, #4032]    [15]

0x450cc0 LDR	W23, [X1]    [1]

0x450cc4 CBZ	W23, 450c8c

0x450db8 ADRP	X10,

0x450dbc LDR	X28, [X10, #3880]    [7]

0x450dc0 LDR	D9, [X28, #112]    [17]

0x450dc4 BL	450480

0x450dc8 FSUB	D2, D0, S8

0x450dcc LDR	W11, [X27]    [16]

0x450dd0 FADD	D3, D2, D9

0x450dd4 STR	D3, [X28, #112]    [17]

0x450dd8 CBZ	W11, 450b30

0x450de8 ADRP	X20,

0x450dec LDR	X22, [X20, #3880]    [8]

0x450df0 LDR	D9, [X22, #104]    [14]

0x450df4 BL	450480

0x450df8 FSUB	D4, D0, S8

0x450dfc LDP	X29, X30, [SP, #128]    [5]

0x450e00 LDP	X19, X20, [SP, #144]    [5]

0x450e04 FADD	D5, D4, D9

0x450e08 LDP	X23, X24, [SP, #176]    [5]

0x450e0c LDP	X27, X28, [SP, #208]    [5]

0x450e10 STR	D5, [X22, #104]    [14]

0x450e14 LDP	X21, X22, [SP, #160]    [5]

0x450e18 LDP	D8, D9, [SP, #224]    [5]

0x450e1c ADD	SP, SP, #304

0x450e20 RET

0x450e24 ADRP	X23,

0x450e28 LDR	X24, [X23, #3880]    [13]

0x450e2c LDR	D9, [X24, #96]    [6]

0x450e30 BL	450480

0x450e34 FSUB	D0, D0, S8

0x450e38 LDR	W3, [X22]    [12]

0x450e3c FADD	D1, D0, D9

0x450e40 STR	D1, [X24, #96]    [6]

0x450e44 CBZ	W3, 450af0

/home/hbollore/qaas/qaas-runs/169-814-4157/intel/CloverLeaf/build/CloverLeaf/CloverLeaf_ref/update_halo.f90: 41 - 117

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

41:     IF(profiler_on) profiler%tile_halo_exchange=profiler%tile_halo_exchange+(timer()-kernel_time)

42:     IF(profiler_on) kernel_time=timer()

43:     CALL clover_exchange(fields,depth)

44:     IF(profiler_on) profiler%mpi_halo_exchange=profiler%mpi_halo_exchange+(timer()-kernel_time)

45:

46:     IF(profiler_on) kernel_time=timer()

47:     IF (  (chunk%chunk_neighbours(CHUNK_LEFT) .EQ. EXTERNAL_FACE) .OR.     &

48:       (chunk%chunk_neighbours(CHUNK_RIGHT) .EQ. EXTERNAL_FACE) .OR.    &

49:       (chunk%chunk_neighbours(CHUNK_BOTTOM) .EQ. EXTERNAL_FACE) .OR.   &

50:       (chunk%chunk_neighbours(CHUNK_TOP) .EQ. EXTERNAL_FACE) ) THEN

51:

52:

53:       IF(use_fortran_kernels) THEN

[...]

83:       ELSEIF(use_C_kernels) THEN

[...]

115:     IF(profiler_on) profiler%self_halo_exchange=profiler%self_halo_exchange+(timer()-kernel_time)

116:

117:   END SUBROUTINE update_halo

Coverage (%)	Name	Source Location	Module
►100.00+	advection	advection.f90:103	exec
○	hydro	hydro.f90:64	exec
○	main	clover_leaf.f90:41	exec
○	__libc_start_main		libc-2.31.so
○	_start	clover_leaf.f90:41	exec

Path /

Metric	Value
CQA speedup if no scalar integer	3.83
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	6.42
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.01
Bottlenecks	P10,
Function	update_halo
Source	update_halo.f90:41-53,update_halo.f90:83-83,update_halo.f90:115-117
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	11.50
CQA cycles if no scalar integer	3.00
CQA cycles if FP arith vectorized	11.50
CQA cycles if fully vectorized	1.79
Front-end cycles	9.25
DIV/SQRT cycles	7.50
P0 cycles	7.50
P1 cycles	5.75
P2 cycles	5.75
P3 cycles	5.75
P4 cycles	5.75
P5 cycles	2.25
P6 cycles	2.25
P7 cycles	2.25
P8 cycles	2.25
P9 cycles	11.50
P10 cycles	11.17
P11 cycles	11.33
P12 cycles	2.00
P13 cycles	2.00
P14 cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	74.00
Nb uops	74.00
Nb loads	NA
Nb stores	7.00
Nb stack references	0.00
FLOP/cycle	0.52
Nb FLOP add-sub	6.00
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	27.13
Bytes prefetched	0.00
Bytes loaded	224.00
Bytes stored	88.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	6.25
Vectorization ratio load	25.00
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	23.44
Vector-efficiency ratio load	31.25
Vector-efficiency ratio store	25.00
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	19.44

Metric	Value
CQA speedup if no scalar integer	3.83
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	6.42
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.01
Bottlenecks	P10,
Function	update_halo
Source	update_halo.f90:41-53,update_halo.f90:83-83,update_halo.f90:115-117
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	11.50
CQA cycles if no scalar integer	3.00
CQA cycles if FP arith vectorized	11.50
CQA cycles if fully vectorized	1.79
Front-end cycles	9.25
DIV/SQRT cycles	7.50
P0 cycles	7.50
P1 cycles	5.75
P2 cycles	5.75
P3 cycles	5.75
P4 cycles	5.75
P5 cycles	2.25
P6 cycles	2.25
P7 cycles	2.25
P8 cycles	2.25
P9 cycles	11.50
P10 cycles	11.17
P11 cycles	11.33
P12 cycles	2.00
P13 cycles	2.00
P14 cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	74.00
Nb uops	74.00
Nb loads	NA
Nb stores	7.00
Nb stack references	0.00
FLOP/cycle	0.52
Nb FLOP add-sub	6.00
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	27.13
Bytes prefetched	0.00
Bytes loaded	224.00
Bytes stored	88.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	6.25
Vectorization ratio load	25.00
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	23.44
Vector-efficiency ratio load	31.25
Vector-efficiency ratio store	25.00
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	19.44

Path /

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

Function	update_halo
Source file and lines	update_halo.f90:41-117
Module	exec

The loop is defined in /home/hbollore/qaas/qaas-runs/169-814-4157/intel/CloverLeaf/build/CloverLeaf/CloverLeaf_ref/update_halo.f90:41-53,83,115-117.

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

Ignoring paths for analysis
Failed to get the number of paths
RecMII not computed since number of paths is unknown or > max_paths

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 11.50 to 3.00 cycles (3.83x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)
To reference allocatable arrays, use "allocatable" instead of "pointer" pointers or qualify them with the "contiguous" attribute (Fortran 2008)
For structures, limit to one indirection. For example, use a_b%c instead of a%b%c with a_b set to a%b before this loop

Vectorization

Your loop is not vectorized. Only 23% of vector register length is used (average across all VPU instructions). By fully vectorizing your loop, you can lower the cost of an iteration from 11.50 to 1.79 cycles (6.42x speedup).

Details

6% of VPU instructions are used in vector version (process two or more data elements in vector registers):

25% of VPU loads are used in vector version.
0% of VPU stores 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 loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with fassociative-math (included in Ofast or ffast-math) to extend loop vectorization to FP reductions.
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: Fortran storage order is column-major: do i do j a(i,j) = b(i,j) (slow, non stride 1) => do i do j a(j,i) = b(i,j) (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): do i a(i)%x = b(i)%x (slow, non stride 1) => do 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 11.50 to 2.00 cycles (5.75x speedup).

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

The binary loop is composed of 6 FP arithmetical operations:

6: addition or subtraction

The binary loop is loading 224 bytes. The binary loop is storing 88 bytes.

Arithmetic intensity

Arithmetic intensity is 0.02 FP operations per loaded or stored byte.

General properties

nb instructions	74
loop length	296
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

9.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	7.50	7.50	5.75	5.75	5.75	5.75	2.25	2.25	2.25	2.25	11.50	11.17	11.33	2.00	2.00
cycles	7.50	7.50	5.75	5.75	5.75	5.75	2.25	2.25	2.25	2.25	11.50	11.17	11.33	2.00	2.00

Cycles executing div or sqrt instructions

Cycles summary

Front-end	9.25
Overall L1	11.50

Vectorization ratios

all	6%
load	25%
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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 11.50 cycles. At this rate:

20% of peak load performance is reached (19.48 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
15% of peak store performance is reached (7.65 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 450af0

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput
MOVZ X7, #4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
MOVZ X8, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
MOVN X6, #0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
MOVZ X5, #257	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
MOVZ X4, #15	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ADD X22, SP, #240	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STP X7, X8, [SP, #272]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
ORR X1, XZR, X20	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, X22	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STP X19, X6, [SP, #240]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
STP X7, X5, [SP, #256]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
STP X8, X4, [SP, #288]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
BL 410ee8 <__clover_module_MOD_clover_exchange>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X27, [X21, #3896]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR W9, [X27]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CBNZ W9, 450db8 <__update_halo_module_MOD_update_halo+0x318>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADRP X12, <46db30>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X13, [X12, #3992]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR W14, [X13, #4]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMN W14, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
B.EQ 450b68 <__update_halo_module_MOD_update_halo+0xc8>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR W15, [X13, #8]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMN W15, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
B.EQ 450b68 <__update_halo_module_MOD_update_halo+0xc8>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR W16, [X13, #12]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMN W16, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
B.EQ 450b68 <__update_halo_module_MOD_update_halo+0xc8>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR W17, [X13, #16]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMN W17, #1	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 450c8c <__update_halo_module_MOD_update_halo+0x1ec>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADRP X18, <46db68>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X30, [X18, #3976]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR W2, [X30]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CBZ W2, 450cb8 <__update_halo_module_MOD_update_halo+0x218>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X21, [X21, #3896]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR W19, [X21]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CBNZ W19, 450de8 <__update_halo_module_MOD_update_halo+0x348>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADRP X0, <46dcb8>	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, [X0, #4032]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR W23, [X1]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CBZ W23, 450c8c <__update_halo_module_MOD_update_halo+0x1ec>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADRP X10, <46ddb8>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X28, [X10, #3880]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR D9, [X28, #112]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
BL 450480 <timer_>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FSUB D2, D0, S8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
LDR W11, [X27]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
FADD D3, D2, D9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
STR D3, [X28, #112]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
CBZ W11, 450b30 <__update_halo_module_MOD_update_halo+0x90>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADRP X20, <46dde8>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X22, [X20, #3880]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR D9, [X22, #104]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
BL 450480 <timer_>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FSUB D4, D0, S8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
LDP X29, X30, [SP, #128]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
LDP X19, X20, [SP, #144]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
FADD D5, D4, D9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
LDP X23, X24, [SP, #176]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
LDP X27, X28, [SP, #208]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
STR D5, [X22, #104]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
LDP X21, X22, [SP, #160]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
LDP D8, D9, [SP, #224]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
ADD SP, SP, #304	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
ADRP X23, <46de24>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X24, [X23, #3880]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR D9, [X24, #96]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
BL 450480 <timer_>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FSUB D0, D0, S8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
LDR W3, [X22]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
FADD D1, D0, D9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
STR D1, [X24, #96]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
CBZ W3, 450af0 <__update_halo_module_MOD_update_halo+0x50>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50

Function	update_halo
Source file and lines	update_halo.f90:41-117
Module	exec

Ignoring paths for analysis
Failed to get the number of paths
RecMII not computed since number of paths is unknown or > max_paths

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 11.50 to 3.00 cycles (3.83x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)
To reference allocatable arrays, use "allocatable" instead of "pointer" pointers or qualify them with the "contiguous" attribute (Fortran 2008)
For structures, limit to one indirection. For example, use a_b%c instead of a%b%c with a_b set to a%b before this loop

Vectorization

Details

6% of VPU instructions are used in vector version (process two or more data elements in vector registers):

25% of VPU loads are used in vector version.
0% of VPU stores 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 loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with fassociative-math (included in Ofast or ffast-math) to extend loop vectorization to FP reductions.
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: Fortran storage order is column-major: do i do j a(i,j) = b(i,j) (slow, non stride 1) => do i do j a(j,i) = b(i,j) (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): do i a(i)%x = b(i)%x (slow, non stride 1) => do 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 11.50 to 2.00 cycles (5.75x speedup).

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

The binary loop is composed of 6 FP arithmetical operations:

6: addition or subtraction

The binary loop is loading 224 bytes. The binary loop is storing 88 bytes.

Arithmetic intensity

Arithmetic intensity is 0.02 FP operations per loaded or stored byte.

General properties

nb instructions	74
loop length	296
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

9.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	7.50	7.50	5.75	5.75	5.75	5.75	2.25	2.25	2.25	2.25	11.50	11.17	11.33	2.00	2.00
cycles	7.50	7.50	5.75	5.75	5.75	5.75	2.25	2.25	2.25	2.25	11.50	11.17	11.33	2.00	2.00

Cycles executing div or sqrt instructions

Cycles summary

Front-end	9.25
Overall L1	11.50

Vectorization ratios

all	6%
load	25%
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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 11.50 cycles. At this rate:

20% of peak load performance is reached (19.48 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
15% of peak store performance is reached (7.65 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 450af0

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput
MOVZ X7, #4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
MOVZ X8, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
MOVN X6, #0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
MOVZ X5, #257	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
MOVZ X4, #15	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
ADD X22, SP, #240	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STP X7, X8, [SP, #272]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
ORR X1, XZR, X20	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, X22	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
STP X19, X6, [SP, #240]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
STP X7, X5, [SP, #256]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
STP X8, X4, [SP, #288]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50
BL 410ee8 <__clover_module_MOD_clover_exchange>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X27, [X21, #3896]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR W9, [X27]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CBNZ W9, 450db8 <__update_halo_module_MOD_update_halo+0x318>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADRP X12, <46db30>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X13, [X12, #3992]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR W14, [X13, #4]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMN W14, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
B.EQ 450b68 <__update_halo_module_MOD_update_halo+0xc8>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR W15, [X13, #8]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMN W15, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
B.EQ 450b68 <__update_halo_module_MOD_update_halo+0xc8>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR W16, [X13, #12]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMN W16, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33
B.EQ 450b68 <__update_halo_module_MOD_update_halo+0xc8>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR W17, [X13, #16]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CMN W17, #1	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 450c8c <__update_halo_module_MOD_update_halo+0x1ec>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADRP X18, <46db68>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X30, [X18, #3976]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR W2, [X30]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CBZ W2, 450cb8 <__update_halo_module_MOD_update_halo+0x218>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X21, [X21, #3896]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR W19, [X21]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CBNZ W19, 450de8 <__update_halo_module_MOD_update_halo+0x348>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADRP X0, <46dcb8>	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, [X0, #4032]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR W23, [X1]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
CBZ W23, 450c8c <__update_halo_module_MOD_update_halo+0x1ec>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADRP X10, <46ddb8>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X28, [X10, #3880]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR D9, [X28, #112]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
BL 450480 <timer_>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FSUB D2, D0, S8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
LDR W11, [X27]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
FADD D3, D2, D9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
STR D3, [X28, #112]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
CBZ W11, 450b30 <__update_halo_module_MOD_update_halo+0x90>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADRP X20, <46dde8>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X22, [X20, #3880]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR D9, [X22, #104]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
BL 450480 <timer_>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FSUB D4, D0, S8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
LDP X29, X30, [SP, #128]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
LDP X19, X20, [SP, #144]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
FADD D5, D4, D9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
LDP X23, X24, [SP, #176]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
LDP X27, X28, [SP, #208]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
STR D5, [X22, #104]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
LDP X21, X22, [SP, #160]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	1
LDP D8, D9, [SP, #224]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
ADD SP, SP, #304	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
ADRP X23, <46de24>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.25
LDR X24, [X23, #3880]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
LDR D9, [X24, #96]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33
BL 450480 <timer_>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FSUB D0, D0, S8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
LDR W3, [X22]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33
FADD D1, D0, D9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25
STR D1, [X24, #96]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50
CBZ W3, 450af0 <__update_halo_module_MOD_update_halo+0x50>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

Expensive FP math instructions/calls

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

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 loop (in the source code) and the binary loop

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code