OV - exec - Loop 203

	Loop Id: 203	Module: exec	Source: flux_calc_kernel.f90:57-61	Coverage: 4.30%

0x42b840 MOVUPD	(%RDI,%RAX,8),%XMM2    [7]

0x42b845 MULPD	%XMM1,%XMM2

0x42b849 MOVUPD	(%R13,%RAX,8),%XMM3    [6]

0x42b850 MOVUPD	(%RCX,%RAX,8),%XMM4    [9]

0x42b855 ADDPD	%XMM3,%XMM4

0x42b859 MOVUPD	(%R14,%RAX,8),%XMM3    [2]

0x42b85f MOVUPD	(%RSI,%RAX,8),%XMM5    [4]

0x42b864 ADDPD	%XMM3,%XMM5

0x42b868 ADDPD	%XMM4,%XMM5

0x42b86c MULPD	%XMM2,%XMM5

0x42b870 MOVUPD	%XMM5,(%R12,%RAX,8)    [5]

0x42b876 MOVUPD	(%RBX,%RAX,8),%XMM2    [3]

0x42b87b MULPD	%XMM1,%XMM2

0x42b87f MOVUPD	-0x8(%RDX,%RAX,8),%XMM3    [10]

0x42b885 MOVUPD	(%RDX,%RAX,8),%XMM4    [10]

0x42b88a ADDPD	%XMM3,%XMM4

0x42b88e MOVUPD	-0x8(%R15,%RAX,8),%XMM3    [8]

0x42b895 MOVUPD	(%R15,%RAX,8),%XMM5    [8]

0x42b89b ADDPD	%XMM3,%XMM5

0x42b89f ADDPD	%XMM4,%XMM5

0x42b8a3 MULPD	%XMM2,%XMM5

0x42b8a7 MOVUPD	%XMM5,(%R8,%RAX,8)    [1]

0x42b8ad ADD	$0x2,%RAX

0x42b8b1 CMP	%R11,%RAX

0x42b8b4 JL	42b840

/home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90: 57 - 61

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

57:         vol_flux_x(j,k)=0.25_8*dt*xarea(j,k)                  &

58:           *(xvel0(j,k)+xvel0(j,k+1)+xvel1(j,k)+xvel1(j,k+1))

59:         vol_flux_y(j,k)=0.25_8*dt*yarea(j,k)                  &

60:           *(yvel0(j,k)+yvel0(j+1,k)+yvel1(j,k)+yvel1(j+1,k))

61:       ENDDO

Coverage (%)	Name	Source Location	Module
►53.84+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so
○	__kmp_launch_thread		libiomp5.so
○	_INTERNALea951ae6::__kmp_launc[...]		libiomp5.so
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►35.07+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so
○	__kmp_launch_thread		libiomp5.so
○	_INTERNALea951ae6::__kmp_launc[...]		libiomp5.so
►4.23+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so
○	__kmp_launch_thread		libiomp5.so
○	_INTERNALea951ae6::__kmp_launc[...]		libiomp5.so
○	start_thread		libc.so.6
►4.14+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so
○	__kmp_fork_call		libiomp5.so
○	__kmpc_fork_call		libiomp5.so
○	flux_calc_kernel	flux_calc_kernel.f90:67	exec
○	flux_calc	flux_calc.f90:43	exec
○	hydro	hydro.f90:62	exec
○	main		exec
○	__libc_start_call_main		libc.so.6
○	__libc_start_main		libc.so.6
○	_start		exec
►2.49+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so
○	__kmp_launch_thread		libiomp5.so

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	4.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.50
Bottlenecks	micro-operation queue, P6, P7, P8, P9, P12, P13,
Function	flux_calc_kernel_module_mp_flux_calc_kernel_.DIR.OMP.PARALLEL.2
Source	flux_calc_kernel.f90:57-61
Source loop unroll info	unrolled by 2
Source loop unroll confidence level	max
Unroll/vectorization loop type	main
Unroll factor	2
CQA cycles	3.00
CQA cycles if no scalar integer	3.00
CQA cycles if FP arith vectorized	3.00
CQA cycles if fully vectorized	0.75
Front-end cycles	3.00
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	3.00
P7 cycles	3.00
P8 cycles	3.00
P9 cycles	3.00
P10 cycles	2.00
P11 cycles	2.00
P12 cycles	3.00
P13 cycles	3.00
P14 cycles	1.00
P15 cycles	1.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	25.00
Nb uops	24.00
Nb loads	10.00
Nb stores	2.00
Nb stack references	0.00
FLOP/cycle	6.67
Nb FLOP add-sub	12.00
Nb FLOP mul	8.00
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	64.00
Bytes prefetched	0.00
Bytes loaded	160.00
Bytes stored	32.00
Stride 0	0.00
Stride 1	8.00
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	NA
Vectorization ratio other	NA
Vector-efficiency ratio all	25.00
Vector-efficiency ratio load	25.00
Vector-efficiency ratio store	25.00
Vector-efficiency ratio mul	25.00
Vector-efficiency ratio add_sub	25.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
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	4.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.50
Bottlenecks	micro-operation queue, P6, P7, P8, P9, P12, P13,
Function	flux_calc_kernel_module_mp_flux_calc_kernel_.DIR.OMP.PARALLEL.2
Source	flux_calc_kernel.f90:57-61
Source loop unroll info	unrolled by 2
Source loop unroll confidence level	max
Unroll/vectorization loop type	main
Unroll factor	2
CQA cycles	3.00
CQA cycles if no scalar integer	3.00
CQA cycles if FP arith vectorized	3.00
CQA cycles if fully vectorized	0.75
Front-end cycles	3.00
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	3.00
P7 cycles	3.00
P8 cycles	3.00
P9 cycles	3.00
P10 cycles	2.00
P11 cycles	2.00
P12 cycles	3.00
P13 cycles	3.00
P14 cycles	1.00
P15 cycles	1.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	25.00
Nb uops	24.00
Nb loads	10.00
Nb stores	2.00
Nb stack references	0.00
FLOP/cycle	6.67
Nb FLOP add-sub	12.00
Nb FLOP mul	8.00
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	64.00
Bytes prefetched	0.00
Bytes loaded	160.00
Bytes stored	32.00
Stride 0	0.00
Stride 1	8.00
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	NA
Vectorization ratio other	NA
Vector-efficiency ratio all	25.00
Vector-efficiency ratio load	25.00
Vector-efficiency ratio store	25.00
Vector-efficiency ratio mul	25.00
Vector-efficiency ratio add_sub	25.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	NA

Path /

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

Function	flux_calc_kernel_module_mp_flux_calc_kernel_.DIR.OMP.PARALLEL.2
Source file and lines	flux_calc_kernel.f90:57-61
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/flux_calc_kernel.f90:57-61.

It is main loop of related source loop which is unrolled by 2 (including vectorization).

gain
potential
hint
expert

Vectorization

Your loop is vectorized, but using only 128 out of 512 bits (SSE/AVX-128 instructions on AVX-512 processors).

By fully vectorizing your loop, you can lower the cost of an iteration from 3.00 to 0.75 cycles (4.00x speedup).

Details

All SSE/AVX instructions are used in vector version (process two or more data elements in vector registers). Since your execution units are vector units, only a fully vectorized loop can use their full power.

Workaround

Pass to your compiler a micro-architecture specialization option:
- use axHost or xHost
Use vector aligned instructions:
1. align your arrays on 64 bytes boundaries: compile with -align array64byte (remark: not affects arrays in COMMON blocks).
2. inform your compiler that your arrays are vector aligned: Append !DIR$ VECTOR ALIGNED to the loop if all accessed arrays are aligned, or !DIR$ ASSUME_ALIGNED FOO: 64 if only FOO is aligned.
Use the LOOP COUNT directive

Execution units bottlenecks

Performance is limited by:

execution of FP add operations (the FP add unit is a bottleneck)
reading data from caches/RAM (load units are a bottleneck)
writing data to caches/RAM (the store unit is a bottleneck)

By removing all these bottlenecks, you can lower the cost of an iteration from 3.00 to 2.00 cycles (1.50x speedup).

Workaround

Reduce the number of FP add instructions
Read less array elements
Write less array elements
Provide more information to your compiler:
- hardcode the bounds of the corresponding 'for' loop

FMA

Presence of both ADD/SUB and MUL operations.

Workaround

Pass to your compiler a micro-architecture specialization option:
- use axHost or xHost
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 12 suboptimal vector unaligned load/store instructions.

Details

MOVUPD: 12 occurrences►
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:57
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:57
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:57
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:58
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:58
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:57
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:59
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:59
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:59
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:60
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:60
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:59

Workaround

Pass to your compiler a micro-architecture specialization option:
- use axHost or xHost
Use vector aligned instructions:
1. align your arrays on 64 bytes boundaries: compile with -align array64byte (remark: not affects arrays in COMMON blocks).
2. inform your compiler that your arrays are vector aligned: Append !DIR$ VECTOR ALIGNED to the loop if all accessed arrays are aligned, or !DIR$ ASSUME_ALIGNED FOO: 64 if only FOO is aligned.

Type of elements and instruction set

10 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in vector mode (two at a time).

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

The binary loop is composed of 20 FP arithmetical operations:

12: addition or subtraction
8: multiply

The binary loop is loading 160 bytes (20 double precision FP elements). The binary loop is storing 32 bytes (4 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.10 FP operations per loaded or stored byte.

General properties

nb instructions	25
nb uops	24
loop length	118
used x86 registers	12
used mmx registers	0
used xmm registers	5
used ymm registers	0
used zmm registers	0
nb stack references	0
ADD-SUB / MUL ratio	1.50

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	3.00 cycles
front end	3.00 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	3.00	3.00	3.00	3.00	2.00	2.00	3.00	3.00	1.00	1.00
cycles	0.33	0.33	0.33	0.33	0.33	0.33	3.00	3.00	3.00	3.00	2.00	2.00	3.00	3.00	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	NA
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	3.00
Dispatch	3.00
Data deps.	1.00
Overall L1	3.00

Vectorization ratios

all	100%
load	100%
store	100%
mul	100%
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)

Vector efficiency ratios

all	25%
load	25%
store	25%
mul	25%
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
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 3.00 cycles. At this rate:

20% of peak load performance is reached (53.33 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
8% of peak store performance is reached (10.67 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Performance is limited by instruction throughput (loading/decoding program instructions to execution core) (front-end is a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 3.00 to 2.00 cycles (1.50x speedup).

ASM code

In the binary file, the address of the loop is: 42b840

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
MOVUPD (%RDI,%RAX,8),%XMM2	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 (25.0%)
MULPD %XMM1,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (25.0%)
MOVUPD (%R13,%RAX,8),%XMM3	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 (25.0%)
MOVUPD (%RCX,%RAX,8),%XMM4	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 (25.0%)
ADDPD %XMM3,%XMM4	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
MOVUPD (%R14,%RAX,8),%XMM3	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 (25.0%)
MOVUPD (%RSI,%RAX,8),%XMM5	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 (25.0%)
ADDPD %XMM3,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
ADDPD %XMM4,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
MULPD %XMM2,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (25.0%)
MOVUPD %XMM5,(%R12,%RAX,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	0.50	vect (25.0%)
MOVUPD (%RBX,%RAX,8),%XMM2	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 (25.0%)
MULPD %XMM1,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (25.0%)
MOVUPD -0x8(%RDX,%RAX,8),%XMM3	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 (25.0%)
MOVUPD (%RDX,%RAX,8),%XMM4	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 (25.0%)
ADDPD %XMM3,%XMM4	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
MOVUPD -0x8(%R15,%RAX,8),%XMM3	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 (25.0%)
MOVUPD (%R15,%RAX,8),%XMM5	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 (25.0%)
ADDPD %XMM3,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
ADDPD %XMM4,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
MULPD %XMM2,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (25.0%)
MOVUPD %XMM5,(%R8,%RAX,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	0.50	vect (25.0%)
ADD $0x2,%RAX	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 %R11,%RAX	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
JL 42b840 <flux_calc_kernel_module_mp_flux_calc_kernel_.DIR.OMP.PARALLEL.2+0x500>	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	flux_calc_kernel_module_mp_flux_calc_kernel_.DIR.OMP.PARALLEL.2
Source file and lines	flux_calc_kernel.f90:57-61
Module	exec

gain
potential
hint
expert

Vectorization

Your loop is vectorized, but using only 128 out of 512 bits (SSE/AVX-128 instructions on AVX-512 processors).

By fully vectorizing your loop, you can lower the cost of an iteration from 3.00 to 0.75 cycles (4.00x speedup).

Details

Workaround

Pass to your compiler a micro-architecture specialization option:
- use axHost or xHost
Use vector aligned instructions:
1. align your arrays on 64 bytes boundaries: compile with -align array64byte (remark: not affects arrays in COMMON blocks).
2. inform your compiler that your arrays are vector aligned: Append !DIR$ VECTOR ALIGNED to the loop if all accessed arrays are aligned, or !DIR$ ASSUME_ALIGNED FOO: 64 if only FOO is aligned.
Use the LOOP COUNT directive

Execution units bottlenecks

Performance is limited by:

execution of FP add operations (the FP add unit is a bottleneck)
reading data from caches/RAM (load units are a bottleneck)
writing data to caches/RAM (the store unit is a bottleneck)

By removing all these bottlenecks, you can lower the cost of an iteration from 3.00 to 2.00 cycles (1.50x speedup).

Workaround

Reduce the number of FP add instructions
Read less array elements
Write less array elements
Provide more information to your compiler:
- hardcode the bounds of the corresponding 'for' loop

FMA

Presence of both ADD/SUB and MUL operations.

Workaround

Pass to your compiler a micro-architecture specialization option:
- use axHost or xHost
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 12 suboptimal vector unaligned load/store instructions.

Details

MOVUPD: 12 occurrences►
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:57
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:57
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:57
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:58
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:58
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:57
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:59
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:59
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:59
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:60
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:60
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/flux_calc_kernel.f90:59

Workaround

Pass to your compiler a micro-architecture specialization option:
- use axHost or xHost
Use vector aligned instructions:
1. align your arrays on 64 bytes boundaries: compile with -align array64byte (remark: not affects arrays in COMMON blocks).
2. inform your compiler that your arrays are vector aligned: Append !DIR$ VECTOR ALIGNED to the loop if all accessed arrays are aligned, or !DIR$ ASSUME_ALIGNED FOO: 64 if only FOO is aligned.

Type of elements and instruction set

10 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in vector mode (two at a time).

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

The binary loop is composed of 20 FP arithmetical operations:

12: addition or subtraction
8: multiply

The binary loop is loading 160 bytes (20 double precision FP elements). The binary loop is storing 32 bytes (4 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.10 FP operations per loaded or stored byte.

General properties

nb instructions	25
nb uops	24
loop length	118
used x86 registers	12
used mmx registers	0
used xmm registers	5
used ymm registers	0
used zmm registers	0
nb stack references	0
ADD-SUB / MUL ratio	1.50

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	3.00 cycles
front end	3.00 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	3.00	3.00	3.00	3.00	2.00	2.00	3.00	3.00	1.00	1.00
cycles	0.33	0.33	0.33	0.33	0.33	0.33	3.00	3.00	3.00	3.00	2.00	2.00	3.00	3.00	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	NA
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	3.00
Dispatch	3.00
Data deps.	1.00
Overall L1	3.00

Vectorization ratios

all	100%
load	100%
store	100%
mul	100%
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)

Vector efficiency ratios

all	25%
load	25%
store	25%
mul	25%
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
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 3.00 cycles. At this rate:

20% of peak load performance is reached (53.33 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
8% of peak store performance is reached (10.67 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

ASM code

In the binary file, the address of the loop is: 42b840

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
MOVUPD (%RDI,%RAX,8),%XMM2	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 (25.0%)
MULPD %XMM1,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (25.0%)
MOVUPD (%R13,%RAX,8),%XMM3	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 (25.0%)
MOVUPD (%RCX,%RAX,8),%XMM4	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 (25.0%)
ADDPD %XMM3,%XMM4	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
MOVUPD (%R14,%RAX,8),%XMM3	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 (25.0%)
MOVUPD (%RSI,%RAX,8),%XMM5	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 (25.0%)
ADDPD %XMM3,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
ADDPD %XMM4,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
MULPD %XMM2,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (25.0%)
MOVUPD %XMM5,(%R12,%RAX,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	0.50	vect (25.0%)
MOVUPD (%RBX,%RAX,8),%XMM2	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 (25.0%)
MULPD %XMM1,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (25.0%)
MOVUPD -0x8(%RDX,%RAX,8),%XMM3	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 (25.0%)
MOVUPD (%RDX,%RAX,8),%XMM4	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 (25.0%)
ADDPD %XMM3,%XMM4	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
MOVUPD -0x8(%R15,%RAX,8),%XMM3	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 (25.0%)
MOVUPD (%R15,%RAX,8),%XMM5	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 (25.0%)
ADDPD %XMM3,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
ADDPD %XMM4,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	vect (25.0%)
MULPD %XMM2,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	vect (25.0%)
MOVUPD %XMM5,(%R8,%RAX,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	0.50	vect (25.0%)
ADD $0x2,%RAX	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 %R11,%RAX	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
JL 42b840 <flux_calc_kernel_module_mp_flux_calc_kernel_.DIR.OMP.PARALLEL.2+0x500>	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

Workaround

Execution units bottlenecks

Workaround

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

Workaround

Execution units bottlenecks

Workaround

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