OV - exec - Loop 9

	Loop Id: 9	Module: exec	Source: stream.c:348-348	Coverage: 29.77%

0x401740 LDP	Q0, Q1, [X16, #2032]    [2]

0x401744 LDP	Q2, Q3, [X17, #2032]    [3]

0x401748 SUBS	X18, X18, #4

0x40174c ADD	X16, X16, #32

0x401750 ADD	X17, X17, #32

0x401754 FADD	V1.2D, V1.2D, V3.2D

0x401758 FADD	V0.2D, V0.2D, V2.2D

0x40175c STP	Q0, Q1, [X15, #2032]    [1]

0x401760 ADD	X15, X15, #32

0x401764 B.NE	401740

/home/eoseret/Tools/QaaS/qaas_runs/175-492-1446/intel/stream/build/stream/src/stream.c: 348 - 348

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

348: 	    c[j] = a[j]+b[j];

Coverage (%)	Name	Source Location	Module
○100.00	__kmp_invoke_microtask		libomp.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	1.07
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	1.05
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.07
Bottlenecks	micro-operation queue,
Function	main.omp_outlined.31
Source	stream.c:348-348
Source loop unroll info	unrolled by 4
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	4
CQA cycles	1.25
CQA cycles if no scalar integer	1.17
CQA cycles if FP arith vectorized	1.25
CQA cycles if fully vectorized	1.19
Front-end cycles	1.25
P0 cycles	0.50
P1 cycles	0.50
P2 cycles	0.75
P3 cycles	0.75
P4 cycles	0.67
P5 cycles	0.67
P6 cycles	0.58
P7 cycles	0.58
P8 cycles	0.75
P9 cycles	0.75
P10 cycles	0.75
P11 cycles	0.75
P12 cycles	1.17
P13 cycles	0.83
P14 cycles	1.00
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	10.00
Nb uops	10.00
Nb loads	NA
Nb stores	1.00
Nb stack references	0.00
FLOP/cycle	3.20
Nb FLOP add-sub	4.00
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	76.80
Bytes prefetched	0.00
Bytes loaded	64.00
Bytes stored	32.00
Stride 0	0.00
Stride 1	3.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	0.00
Vectorization ratio all	83.33
Vectorization ratio load	100.00
Vectorization ratio store	100.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	66.67
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	NA
Vector-efficiency ratio all	91.67
Vector-efficiency ratio load	100.00
Vector-efficiency ratio store	100.00
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	83.33
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	NA

Metric	Value
CQA speedup if no scalar integer	1.07
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	1.05
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.07
Bottlenecks	micro-operation queue,
Function	main.omp_outlined.31
Source	stream.c:348-348
Source loop unroll info	unrolled by 4
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	4
CQA cycles	1.25
CQA cycles if no scalar integer	1.17
CQA cycles if FP arith vectorized	1.25
CQA cycles if fully vectorized	1.19
Front-end cycles	1.25
P0 cycles	0.50
P1 cycles	0.50
P2 cycles	0.75
P3 cycles	0.75
P4 cycles	0.67
P5 cycles	0.67
P6 cycles	0.58
P7 cycles	0.58
P8 cycles	0.75
P9 cycles	0.75
P10 cycles	0.75
P11 cycles	0.75
P12 cycles	1.17
P13 cycles	0.83
P14 cycles	1.00
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	10.00
Nb uops	10.00
Nb loads	NA
Nb stores	1.00
Nb stack references	0.00
FLOP/cycle	3.20
Nb FLOP add-sub	4.00
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	76.80
Bytes prefetched	0.00
Bytes loaded	64.00
Bytes stored	32.00
Stride 0	0.00
Stride 1	3.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	0.00
Vectorization ratio all	83.33
Vectorization ratio load	100.00
Vectorization ratio store	100.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	66.67
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	NA
Vector-efficiency ratio all	91.67
Vector-efficiency ratio load	100.00
Vector-efficiency ratio store	100.00
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	83.33
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	main.omp_outlined.31
Source file and lines	stream.c:348-348
Module	exec

The loop is defined in /home/eoseret/Tools/QaaS/qaas_runs/175-492-1446/intel/stream/build/stream/src/stream.c:348.

The related source loop is unrolled by 4 (including vectorization).

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 1.25 to 1.17 cycles (1.07x speedup).

Workaround

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

Vectorization

Your loop is partially vectorized. 91% 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 1.25 to 1.19 cycles (1.05x speedup).

Details

83% of VPU instructions are used in vector version (process two or more data elements in vector registers):

66% of VPU addition or subtraction instructions are used in vector version.

Since your execution units are vector units, only a fully vectorized loop can use their full power.

No data for this section

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

The binary loop is composed of 4 FP arithmetical operations:

4: addition or subtraction

The binary loop is loading 64 bytes. The binary loop is storing 32 bytes.

Arithmetic intensity

Arithmetic intensity is 0.04 FP operations per loaded or stored byte.

General properties

nb instructions	10
nb uops	10
loop length	40
used w registers	0
used x registers	4
used b registers	0
used h registers	0
used s registers	0
used d registers	0
used q registers	4
used v registers	4
used z registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	1.25 cycles
front end	1.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	0.50	0.50	0.75	0.75	0.67	0.67	0.58	0.58	0.75	0.75	0.75	0.75	1.17	0.83	1.00	0.00	0.00
cycles	0.50	0.50	0.75	0.75	0.67	0.67	0.58	0.58	0.75	0.75	0.75	0.75	1.17	0.83	1.00	0.00	0.00

Execution ports to units layout:

P0: BRU
P1: BRU
P2: ALU
P3: ALU
P4: ALU
P5: ALU
P6: ALU
P7: ALU
P8 (128 bits): VPU, FP store data, ALU, DIV/SQRT
P9 (128 bits): VPU, ALU, FP store data
P10 (128 bits): VPU, ALU, DIV/SQRT
P11 (128 bits): ALU, VPU
P12 (256 bits): store address, load
P13 (256 bits): store address, load
P14 (256 bits): load
P15 (64 bits): store data
P16 (64 bits): store data

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

Cycles summary

Front-end	1.25
Dispatch	1.17
Data deps.	1.00
Overall L1	1.25

Vectorization ratios

INT

all	75%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

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

INT+FP

all	83%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	66%
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

INT

all	87%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	50%
fma	NA (no fma vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

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

INT+FP

all	91%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	83%
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 1.25 cycles. At this rate:

53% of peak load performance is reached (51.20 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
53% of peak store performance is reached (25.60 out of 48.00 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 1.25 to 1.17 cycles (1.07x speedup).

ASM code

In the binary file, the address of the loop is: 401740

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDP Q0, Q1, [X16, #2032]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.67	vect (100.0%)
LDP Q2, Q3, [X17, #2032]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.67	vect (100.0%)
SUBS X18, X18, #4	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	1	0.33	scal (50.0%)
ADD X16, X16, #32	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
ADD X17, X17, #32	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
FADD V1.2D, V1.2D, V3.2D	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
FADD V0.2D, V0.2D, V2.2D	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
STP Q0, Q1, [X15, #2032]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	1	vect (100.0%)
ADD X15, X15, #32	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
B.NE 401740 <main.omp_outlined.31+0xb0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Function	main.omp_outlined.31
Source file and lines	stream.c:348-348
Module	exec

The loop is defined in /home/eoseret/Tools/QaaS/qaas_runs/175-492-1446/intel/stream/build/stream/src/stream.c:348.

The related source loop is unrolled by 4 (including vectorization).

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 1.25 to 1.17 cycles (1.07x speedup).

Workaround

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

Vectorization

Details

83% of VPU instructions are used in vector version (process two or more data elements in vector registers):

66% of VPU addition or subtraction instructions are used in vector version.

Since your execution units are vector units, only a fully vectorized loop can use their full power.

No data for this section

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

The binary loop is composed of 4 FP arithmetical operations:

4: addition or subtraction

The binary loop is loading 64 bytes. The binary loop is storing 32 bytes.

Arithmetic intensity

Arithmetic intensity is 0.04 FP operations per loaded or stored byte.

General properties

nb instructions	10
nb uops	10
loop length	40
used w registers	0
used x registers	4
used b registers	0
used h registers	0
used s registers	0
used d registers	0
used q registers	4
used v registers	4
used z registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	1.25 cycles
front end	1.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	0.50	0.50	0.75	0.75	0.67	0.67	0.58	0.58	0.75	0.75	0.75	0.75	1.17	0.83	1.00	0.00	0.00
cycles	0.50	0.50	0.75	0.75	0.67	0.67	0.58	0.58	0.75	0.75	0.75	0.75	1.17	0.83	1.00	0.00	0.00

Execution ports to units layout:

P0: BRU
P1: BRU
P2: ALU
P3: ALU
P4: ALU
P5: ALU
P6: ALU
P7: ALU
P8 (128 bits): VPU, FP store data, ALU, DIV/SQRT
P9 (128 bits): VPU, ALU, FP store data
P10 (128 bits): VPU, ALU, DIV/SQRT
P11 (128 bits): ALU, VPU
P12 (256 bits): store address, load
P13 (256 bits): store address, load
P14 (256 bits): load
P15 (64 bits): store data
P16 (64 bits): store data

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

Cycles summary

Front-end	1.25
Dispatch	1.17
Data deps.	1.00
Overall L1	1.25

Vectorization ratios

INT

all	75%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

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

INT+FP

all	83%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	66%
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

INT

all	87%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	50%
fma	NA (no fma vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

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

INT+FP

all	91%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	83%
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 1.25 cycles. At this rate:

53% of peak load performance is reached (51.20 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
53% of peak store performance is reached (25.60 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

ASM code

In the binary file, the address of the loop is: 401740

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDP Q0, Q1, [X16, #2032]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.67	vect (100.0%)
LDP Q2, Q3, [X17, #2032]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.67	vect (100.0%)
SUBS X18, X18, #4	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	1	0.33	scal (50.0%)
ADD X16, X16, #32	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
ADD X17, X17, #32	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
FADD V1.2D, V1.2D, V3.2D	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
FADD V0.2D, V0.2D, V2.2D	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	vect (100.0%)
STP Q0, Q1, [X15, #2032]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	1	vect (100.0%)
ADD X15, X15, #32	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
B.NE 401740 <main.omp_outlined.31+0xb0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Report Configuration

Code clean check

Workaround

Vectorization

Details

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

Code clean check

Workaround

Vectorization

Details

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