OV - exec - Loop 35

	Loop Id: 35	Module: exec	Source: PdV_kernel.f90:77-99 [...]	Coverage: 5.52%

0x222f30 VMOVUPD	-0x8(%R9,%RCX,8),%ZMM12    [1]

0x222f3b VMOVUPD	(%R9,%RCX,8),%ZMM13    [1]

0x222f42 VMOVUPD	(%RAX,%RCX,8),%ZMM14    [4]

0x222f49 VMOVUPD	(%R13,%RCX,8),%ZMM15    [2]

0x222f51 VMOVUPD	(%R15,%RCX,8),%ZMM16    [14]

0x222f58 VMOVUPD	(%R14,%RCX,8),%ZMM17    [8]

0x222f5f MOV	0x110(%RSP),%R11    [15]

0x222f67 VADDPD	-0x8(%RSI,%RCX,8),%ZMM12,%ZMM12    [9]

0x222f72 VADDPD	(%RSI,%RCX,8),%ZMM13,%ZMM13    [9]

0x222f79 VDIVPD	%ZMM16,%ZMM17,%ZMM17

0x222f7f VADDPD	%ZMM12,%ZMM12,%ZMM12

0x222f85 VMULPD	-0x8(%RDX,%RCX,8),%ZMM12,%ZMM12    [10]

0x222f90 VADDPD	%ZMM13,%ZMM13,%ZMM13

0x222f96 VMULPD	(%RDX,%RCX,8),%ZMM13,%ZMM13    [10]

0x222f9d VMULPD	%ZMM8,%ZMM12,%ZMM12

0x222fa3 VMULPD	%ZMM8,%ZMM13,%ZMM13

0x222fa9 VMULPD	%ZMM12,%ZMM0,%ZMM12

0x222faf VMULPD	%ZMM13,%ZMM0,%ZMM13

0x222fb5 VMULPD	%ZMM9,%ZMM12,%ZMM12

0x222fbb VMULPD	%ZMM9,%ZMM13,%ZMM13

0x222fc1 VSUBPD	%ZMM12,%ZMM13,%ZMM12

0x222fc7 VADDPD	-0x8(%RAX,%RCX,8),%ZMM14,%ZMM13    [4]

0x222fd2 VMOVUPD	(%R8,%RCX,8),%ZMM14    [7]

0x222fd9 VADDPD	-0x8(%R8,%RCX,8),%ZMM14,%ZMM14    [7]

0x222fe4 VADDPD	%ZMM13,%ZMM13,%ZMM13

0x222fea VMULPD	(%RBP,%RCX,8),%ZMM13,%ZMM13    [5]

0x222ff2 VADDPD	%ZMM14,%ZMM14,%ZMM14

0x222ff8 VMULPD	(%RDI,%RCX,8),%ZMM14,%ZMM14    [3]

0x222fff VMULPD	%ZMM8,%ZMM13,%ZMM13

0x223005 VMULPD	%ZMM8,%ZMM14,%ZMM14

0x22300b VMULPD	%ZMM13,%ZMM0,%ZMM13

0x223011 VMULPD	%ZMM14,%ZMM0,%ZMM14

0x223017 VMULPD	%ZMM9,%ZMM13,%ZMM13

0x22301d VMULPD	%ZMM9,%ZMM14,%ZMM14

0x223023 VADDPD	%ZMM14,%ZMM12,%ZMM12

0x223029 VSUBPD	%ZMM13,%ZMM12,%ZMM12

0x22302f VDIVPD	%ZMM15,%ZMM11,%ZMM13

0x223035 VADDPD	%ZMM12,%ZMM15,%ZMM14

0x22303b VDIVPD	%ZMM14,%ZMM15,%ZMM14

0x223041 VMOVUPD	(%R12,%RCX,8),%ZMM15    [13]

0x223048 VDIVPD	%ZMM16,%ZMM15,%ZMM15

0x22304e VADDPD	%ZMM17,%ZMM15,%ZMM15

0x223054 VMULPD	%ZMM15,%ZMM12,%ZMM12

0x22305a VMULPD	%ZMM12,%ZMM13,%ZMM12

0x223060 VADDPD	(%R11,%RCX,8),%ZMM12,%ZMM12    [12]

0x223067 MOV	0x10(%RSP),%R11    [15]

0x22306c VMOVUPD	%ZMM12,(%R11,%RCX,8)    [11]

0x223073 MOV	0x80(%RSP),%R11    [15]

0x22307b VMULPD	%ZMM14,%ZMM16,%ZMM12

0x223081 VMOVUPD	%ZMM12,(%R11,%RCX,8)    [6]

0x223088 ADD	$0x8,%RCX

0x22308c CMP	%RCX,%R10

0x22308f JNE	222f30

/home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90: 77 - 99

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

77:           left_flux=  (xarea(j  ,k  )*(xvel0(j  ,k  )+xvel0(j  ,k+1)                     &

78:             +xvel0(j  ,k  )+xvel0(j  ,k+1)))*0.25_8*dt*0.5

79:           right_flux= (xarea(j+1,k  )*(xvel0(j+1,k  )+xvel0(j+1,k+1)                     &

80:             +xvel0(j+1,k  )+xvel0(j+1,k+1)))*0.25_8*dt*0.5

81:           bottom_flux=(yarea(j  ,k  )*(yvel0(j  ,k  )+yvel0(j+1,k  )                     &

82:             +yvel0(j  ,k  )+yvel0(j+1,k  )))*0.25_8*dt*0.5

83:           top_flux=   (yarea(j  ,k+1)*(yvel0(j  ,k+1)+yvel0(j+1,k+1)                     &

84:             +yvel0(j  ,k+1)+yvel0(j+1,k+1)))*0.25_8*dt*0.5

85:           total_flux=right_flux-left_flux+top_flux-bottom_flux

86:

87:           volume_change_s=volume(j,k)/(volume(j,k)+total_flux)

[...]

93:           recip_volume=1.0/volume(j,k)

94:

95:           energy_change=(pressure(j,k)/density0(j,k)+viscosity(j,k)/density0(j,k))*total_flux*recip_volume

96:

97:           energy1(j,k)=energy0(j,k)-energy_change

98:

99:           density1(j,k)=density0(j,k)*volume_change_s

Coverage (%)	Name	Source Location	Module

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	1.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.39
Bottlenecks	P10, P11,
Function	__nv_pdv_kernel_module_pdv_kernel__PARALLEL_F2L67_1
Source	PdV_kernel.f90:77-87,PdV_kernel.f90:93-99
Source loop unroll info	unrolled by 8
Source loop unroll confidence level	max
Unroll/vectorization loop type	main
Unroll factor	8
CQA cycles	16.00
CQA cycles if no scalar integer	16.00
CQA cycles if FP arith vectorized	16.00
CQA cycles if fully vectorized	16.00
Front-end cycles	6.50
P0 cycles	0.33
P1 cycles	0.33
P2 cycles	0.33
P3 cycles	0.33
P4 cycles	0.33
P5 cycles	0.33
P6 cycles	5.50
P7 cycles	5.50
P8 cycles	5.50
P9 cycles	5.50
P10 cycles	11.50
P11 cycles	11.50
P12 cycles	7.00
P13 cycles	7.00
P14 cycles	1.00
P15 cycles	1.00
DIV/SQRT cycles	16.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	53.00
Nb uops	52.00
Nb loads	20.00
Nb stores	2.00
Nb stack references	3.00
FLOP/cycle	18.50
Nb FLOP add-sub	112.00
Nb FLOP mul	152.00
Nb FLOP fma	0.00
Nb FLOP div	32.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	77.50
Bytes prefetched	0.00
Bytes loaded	1112.00
Bytes stored	128.00
Stride 0	1.00
Stride 1	9.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	0.00
Vectorization ratio all	100.00
Vectorization ratio load	100.00
Vectorization ratio store	100.00
Vectorization ratio mul	100.00
Vectorization ratio add_sub	100.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	100.00
Vectorization ratio other	NA
Vector-efficiency ratio all	100.00
Vector-efficiency ratio load	100.00
Vector-efficiency ratio store	100.00
Vector-efficiency ratio mul	100.00
Vector-efficiency ratio add_sub	100.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	100.00
Vector-efficiency ratio other	NA

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	1.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.39
Bottlenecks	P10, P11,
Function	__nv_pdv_kernel_module_pdv_kernel__PARALLEL_F2L67_1
Source	PdV_kernel.f90:77-87,PdV_kernel.f90:93-99
Source loop unroll info	unrolled by 8
Source loop unroll confidence level	max
Unroll/vectorization loop type	main
Unroll factor	8
CQA cycles	16.00
CQA cycles if no scalar integer	16.00
CQA cycles if FP arith vectorized	16.00
CQA cycles if fully vectorized	16.00
Front-end cycles	6.50
P0 cycles	0.33
P1 cycles	0.33
P2 cycles	0.33
P3 cycles	0.33
P4 cycles	0.33
P5 cycles	0.33
P6 cycles	5.50
P7 cycles	5.50
P8 cycles	5.50
P9 cycles	5.50
P10 cycles	11.50
P11 cycles	11.50
P12 cycles	7.00
P13 cycles	7.00
P14 cycles	1.00
P15 cycles	1.00
DIV/SQRT cycles	16.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	53.00
Nb uops	52.00
Nb loads	20.00
Nb stores	2.00
Nb stack references	3.00
FLOP/cycle	18.50
Nb FLOP add-sub	112.00
Nb FLOP mul	152.00
Nb FLOP fma	0.00
Nb FLOP div	32.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	77.50
Bytes prefetched	0.00
Bytes loaded	1112.00
Bytes stored	128.00
Stride 0	1.00
Stride 1	9.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	0.00
Vectorization ratio all	100.00
Vectorization ratio load	100.00
Vectorization ratio store	100.00
Vectorization ratio mul	100.00
Vectorization ratio add_sub	100.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	100.00
Vectorization ratio other	NA
Vector-efficiency ratio all	100.00
Vector-efficiency ratio load	100.00
Vector-efficiency ratio store	100.00
Vector-efficiency ratio mul	100.00
Vector-efficiency ratio add_sub	100.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	100.00
Vector-efficiency ratio other	NA

Path /

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

Function	__nv_pdv_kernel_module_pdv_kernel__PARALLEL_F2L67_1
Source file and lines	PdV_kernel.f90:77-99
Module	exec

The loop is defined in /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:77-87,93-99.

It is main loop of related source loop which is unrolled by 8 (including vectorization).

gain
potential
hint
expert

Vectorization

Your loop is fully vectorized, using full register length.

Details

All SSE/AVX instructions are used in vector version (process two or more data elements in vector registers).

Execution units bottlenecks

Performance is limited by execution of divide and square root operations (the divide/square root unit is a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 16.00 to 11.50 cycles (1.39x speedup).

Workaround

Reduce the number of division or square root instructions:
- If denominator is constant over iterations, use reciprocal (replace x/y with x*(1/y)). Check precision impact.
Check whether you really need double precision. If not, switch to single precision to speedup execution

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 16.00 to 9.50 cycles (1.68x speedup).

FMA

Presence of both ADD/SUB and MUL operations.

Workaround

Pass to your compiler a micro-architecture specialization option:
- Please look into your compiler manual for march=native or equivalent
Try to change order in which elements are evaluated (using parentheses) in arithmetic expressions containing both ADD/SUB and MUL operations to enable your compiler to generate FMA instructions wherever possible. For instance a + b*c is a valid FMA (MUL then ADD). However (a+b)* c cannot be translated into an FMA (ADD then MUL).

Vector unaligned load/store instructions

Detected 10 optimal vector unaligned load/store instructions.

Details

VMOVUPD: 10 occurrences►
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:77
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:79
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:81
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:87
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:95
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:95
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:83
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:95
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:97
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:99

Workaround

Use vector aligned instructions:

align your arrays on 64 bytes boundaries
inform your compiler that your arrays are vector aligned: read your compiler manual.

Type of elements and instruction set

37 AVX-512 instructions are processing arithmetic or math operations on double precision FP elements in vector mode (eight at a time).

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

The binary loop is composed of 296 FP arithmetical operations:

112: addition or subtraction
152: multiply
32: divide

The binary loop is loading 1112 bytes (139 double precision FP elements). The binary loop is storing 128 bytes (16 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.24 FP operations per loaded or stored byte.

General properties

nb instructions	53
nb uops	52
loop length	357
used x86 registers	15
used mmx registers	0
used xmm registers	0
used ymm registers	0
used zmm registers	10
nb stack references	3
ADD-SUB / MUL ratio	0.74

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	6.50 cycles
front end	6.50 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	0.33	0.33	0.33	0.33	0.33	0.33	5.50	5.50	5.50	5.50	11.50	11.50	7.00	7.00	1.00	1.00
cycles	0.33	0.33	0.33	0.33	0.33	0.33	5.50	5.50	5.50	5.50	11.50	11.50	7.00	7.00	1.00	1.00

Execution ports to units layout:

P0: ALU
P1: ALU
P2: ALU
P3: ALU, BRU
P4: ALU, BRU
P5: ALU, BRU
P6 (512 bits): store address, load
P7 (512 bits): store address, load
P8 (512 bits): store address, load
P9 (512 bits): store address, load
P10 (512 bits): VPU, DIV/SQRT
P11 (512 bits): VPU, DIV/SQRT
P12 (512 bits): VPU
P13 (512 bits): VPU
P14 (512 bits): FP store data
P15 (512 bits): FP store data

Cycles executing div or sqrt instructions	16.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	6.50
Dispatch	11.50
DIV/SQRT	16.00
Data deps.	1.00
Overall L1	16.00

Vectorization ratios

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

Vector efficiency ratios

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 16.00 cycles. At this rate:

27% of peak load performance is reached (69.50 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
6% of peak store performance is reached (8.00 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 222f30

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
VMOVUPD -0x8(%R9,%RCX,8),%ZMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD (%R9,%RCX,8),%ZMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD (%RAX,%RCX,8),%ZMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD (%R13,%RCX,8),%ZMM15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD (%R15,%RCX,8),%ZMM16	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD (%R14,%RCX,8),%ZMM17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
MOV 0x110(%RSP),%R11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.25	N/A
VADDPD -0x8(%RSI,%RCX,8),%ZMM12,%ZMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VADDPD (%RSI,%RCX,8),%ZMM13,%ZMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VDIVPD %ZMM16,%ZMM17,%ZMM17	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	vect (100.0%)
VADDPD %ZMM12,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMULPD -0x8(%RDX,%RCX,8),%ZMM12,%ZMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VADDPD %ZMM13,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMULPD (%RDX,%RCX,8),%ZMM13,%ZMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM8,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM8,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM12,%ZMM0,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM13,%ZMM0,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM9,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM9,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VSUBPD %ZMM12,%ZMM13,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VADDPD -0x8(%RAX,%RCX,8),%ZMM14,%ZMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMOVUPD (%R8,%RCX,8),%ZMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VADDPD -0x8(%R8,%RCX,8),%ZMM14,%ZMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VADDPD %ZMM13,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMULPD (%RBP,%RCX,8),%ZMM13,%ZMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VADDPD %ZMM14,%ZMM14,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMULPD (%RDI,%RCX,8),%ZMM14,%ZMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM8,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM8,%ZMM14,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM13,%ZMM0,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM14,%ZMM0,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM9,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM9,%ZMM14,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VADDPD %ZMM14,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VSUBPD %ZMM13,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VDIVPD %ZMM15,%ZMM11,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	vect (100.0%)
VADDPD %ZMM12,%ZMM15,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VDIVPD %ZMM14,%ZMM15,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	vect (100.0%)
VMOVUPD (%R12,%RCX,8),%ZMM15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VDIVPD %ZMM16,%ZMM15,%ZMM15	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	vect (100.0%)
VADDPD %ZMM17,%ZMM15,%ZMM15	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMULPD %ZMM15,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM12,%ZMM13,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VADDPD (%R11,%RCX,8),%ZMM12,%ZMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
MOV 0x10(%RSP),%R11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.25	N/A
VMOVUPD %ZMM12,(%R11,%RCX,8)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	1	vect (100.0%)
MOV 0x80(%RSP),%R11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.25	N/A
VMULPD %ZMM14,%ZMM16,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD %ZMM12,(%R11,%RCX,8)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	1	vect (100.0%)
ADD $0x8,%RCX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
CMP %RCX,%R10	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
JNE 222f30 <__nv_pdv_kernel_module_pdv_kernel__PARALLEL_F2L67_1_+0x1380>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	0	1	0.50-1	N/A

Function	__nv_pdv_kernel_module_pdv_kernel__PARALLEL_F2L67_1
Source file and lines	PdV_kernel.f90:77-99
Module	exec

gain
potential
hint
expert

Vectorization

Your loop is fully vectorized, using full register length.

Details

All SSE/AVX instructions are used in vector version (process two or more data elements in vector registers).

Execution units bottlenecks

Workaround

Reduce the number of division or square root instructions:
- If denominator is constant over iterations, use reciprocal (replace x/y with x*(1/y)). Check precision impact.
Check whether you really need double precision. If not, switch to single precision to speedup execution

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 16.00 to 9.50 cycles (1.68x speedup).

FMA

Presence of both ADD/SUB and MUL operations.

Workaround

Pass to your compiler a micro-architecture specialization option:
- Please look into your compiler manual for march=native or equivalent
Try to change order in which elements are evaluated (using parentheses) in arithmetic expressions containing both ADD/SUB and MUL operations to enable your compiler to generate FMA instructions wherever possible. For instance a + b*c is a valid FMA (MUL then ADD). However (a+b)* c cannot be translated into an FMA (ADD then MUL).

Vector unaligned load/store instructions

Detected 10 optimal vector unaligned load/store instructions.

Details

VMOVUPD: 10 occurrences►
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:77
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:79
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:81
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:87
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:95
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:95
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:83
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:95
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:97
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/PdV_kernel.f90:99

Workaround

Use vector aligned instructions:

align your arrays on 64 bytes boundaries
inform your compiler that your arrays are vector aligned: read your compiler manual.

Type of elements and instruction set

37 AVX-512 instructions are processing arithmetic or math operations on double precision FP elements in vector mode (eight at a time).

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

The binary loop is composed of 296 FP arithmetical operations:

112: addition or subtraction
152: multiply
32: divide

The binary loop is loading 1112 bytes (139 double precision FP elements). The binary loop is storing 128 bytes (16 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.24 FP operations per loaded or stored byte.

General properties

nb instructions	53
nb uops	52
loop length	357
used x86 registers	15
used mmx registers	0
used xmm registers	0
used ymm registers	0
used zmm registers	10
nb stack references	3
ADD-SUB / MUL ratio	0.74

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	6.50 cycles
front end	6.50 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	0.33	0.33	0.33	0.33	0.33	0.33	5.50	5.50	5.50	5.50	11.50	11.50	7.00	7.00	1.00	1.00
cycles	0.33	0.33	0.33	0.33	0.33	0.33	5.50	5.50	5.50	5.50	11.50	11.50	7.00	7.00	1.00	1.00

Execution ports to units layout:

P0: ALU
P1: ALU
P2: ALU
P3: ALU, BRU
P4: ALU, BRU
P5: ALU, BRU
P6 (512 bits): store address, load
P7 (512 bits): store address, load
P8 (512 bits): store address, load
P9 (512 bits): store address, load
P10 (512 bits): VPU, DIV/SQRT
P11 (512 bits): VPU, DIV/SQRT
P12 (512 bits): VPU
P13 (512 bits): VPU
P14 (512 bits): FP store data
P15 (512 bits): FP store data

Cycles executing div or sqrt instructions	16.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	6.50
Dispatch	11.50
DIV/SQRT	16.00
Data deps.	1.00
Overall L1	16.00

Vectorization ratios

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

Vector efficiency ratios

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 16.00 cycles. At this rate:

27% of peak load performance is reached (69.50 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
6% of peak store performance is reached (8.00 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 222f30

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
VMOVUPD -0x8(%R9,%RCX,8),%ZMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD (%R9,%RCX,8),%ZMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD (%RAX,%RCX,8),%ZMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD (%R13,%RCX,8),%ZMM15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD (%R15,%RCX,8),%ZMM16	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD (%R14,%RCX,8),%ZMM17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
MOV 0x110(%RSP),%R11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.25	N/A
VADDPD -0x8(%RSI,%RCX,8),%ZMM12,%ZMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VADDPD (%RSI,%RCX,8),%ZMM13,%ZMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VDIVPD %ZMM16,%ZMM17,%ZMM17	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	vect (100.0%)
VADDPD %ZMM12,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMULPD -0x8(%RDX,%RCX,8),%ZMM12,%ZMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VADDPD %ZMM13,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMULPD (%RDX,%RCX,8),%ZMM13,%ZMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM8,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM8,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM12,%ZMM0,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM13,%ZMM0,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM9,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM9,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VSUBPD %ZMM12,%ZMM13,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VADDPD -0x8(%RAX,%RCX,8),%ZMM14,%ZMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMOVUPD (%R8,%RCX,8),%ZMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VADDPD -0x8(%R8,%RCX,8),%ZMM14,%ZMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VADDPD %ZMM13,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMULPD (%RBP,%RCX,8),%ZMM13,%ZMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VADDPD %ZMM14,%ZMM14,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMULPD (%RDI,%RCX,8),%ZMM14,%ZMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM8,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM8,%ZMM14,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM13,%ZMM0,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM14,%ZMM0,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM9,%ZMM13,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM9,%ZMM14,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VADDPD %ZMM14,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VSUBPD %ZMM13,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VDIVPD %ZMM15,%ZMM11,%ZMM13	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	vect (100.0%)
VADDPD %ZMM12,%ZMM15,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VDIVPD %ZMM14,%ZMM15,%ZMM14	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	vect (100.0%)
VMOVUPD (%R12,%RCX,8),%ZMM15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (100.0%)
VDIVPD %ZMM16,%ZMM15,%ZMM15	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	vect (100.0%)
VADDPD %ZMM17,%ZMM15,%ZMM15	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
VMULPD %ZMM15,%ZMM12,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMULPD %ZMM12,%ZMM13,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VADDPD (%R11,%RCX,8),%ZMM12,%ZMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (100.0%)
MOV 0x10(%RSP),%R11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.25	N/A
VMOVUPD %ZMM12,(%R11,%RCX,8)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	1	vect (100.0%)
MOV 0x80(%RSP),%R11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.25	N/A
VMULPD %ZMM14,%ZMM16,%ZMM12	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD %ZMM12,(%R11,%RCX,8)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	1	vect (100.0%)
ADD $0x8,%RCX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
CMP %RCX,%R10	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
JNE 222f30 <__nv_pdv_kernel_module_pdv_kernel__PARALLEL_F2L67_1_+0x1380>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	0	1	0.50-1	N/A

Report Configuration

Vectorization

Details

Execution units bottlenecks

Workaround

Expensive FP math instructions/calls

FMA

Workaround

Vector unaligned load/store instructions

Details

Workaround

Type of elements and instruction set

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

Arithmetic intensity

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

Execution units bottlenecks

Workaround

Expensive FP math instructions/calls

FMA

Workaround

Vector unaligned load/store instructions

Details

Workaround

Type of elements and instruction set

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code