OV - exec - Loop 258

0x41ce40 MOVPRFX	Z18, Z0

0x41ce44 SDIV	Z18.D, P0/M, Z18.D, Z2.D

0x41ce48 ADD	Z19.D, Z3.D, Z18.D

0x41ce4c MSB	Z18.D, P0/M, Z2.D, Z0.D

0x41ce50 ADD	Z0.D, Z0.D, Z1.D

0x41ce54 SUBS	X12, X12, X10

0x41ce58 SXTW	Z19.D, P0/M, Z19.D

0x41ce5c ADD	Z20.D, Z4.D, Z18.D

0x41ce60 MOVPRFX	Z21, Z20

0x41ce64 SXTW	Z21.D, P0/M, Z20.D

0x41ce68 MOVPRFX	Z22, Z21

0x41ce6c MLA	Z22.D, P0/M, Z5.D, Z19.D

0x41ce70 LD1D	{Z22.D}, P0/Z, [X23, Z22.D,LSL #3]    [7]

0x41ce74 ADD	Z18.D, Z7.D, Z18.D

0x41ce78 MOVPRFX	Z23, Z21

0x41ce7c MLA	Z23.D, P0/M, Z6.D, Z19.D

0x41ce80 ST1D	{Z22.D}, P0, [X21, Z23.D,LSL #3]    [5]

0x41ce84 MOVPRFX	Z23, Z16

0x41ce88 MUL	Z23.D, P0/M, Z23.D, Z19.D

0x41ce8c ADR	Z18.D, [Z23, Z18.D,SXTW]    [1]

0x41ce90 ADR	Z20.D, [Z23, Z20.D,SXTW]    [2]

0x41ce94 MAD	Z19.D, P0/M, Z17.D, Z21.D

0x41ce98 LD1D	{Z18.D}, P0/Z, [X27, Z18.D,LSL #3]    [3]

0x41ce9c LD1D	{Z20.D}, P0/Z, [X27, Z20.D,LSL #3]    [4]

0x41cea0 FADD	Z18.D, Z22.D, Z18.D

0x41cea4 FSUB	Z18.D, Z18.D, Z20.D

0x41cea8 ST1D	{Z18.D}, P0, [X25, Z19.D,LSL #3]    [6]

0x41ceac B.NE	41ce40

/home/hbollore/qaas-runs/170-290-5445/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/advec_mom.cpp: 72 - 75

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

72:     for (int j = (y_min - 2 + 1); j < (y_max + 2 + 2); j++) {

73:       for (int i = (x_min - 2 + 1); i < (x_max + 2 + 2); i++) {

74:         post_vol(i, j) = volume(i, j);

75:         pre_vol(i, j) = post_vol(i, j) + vol_flux_x(i + 1, j + 0) - vol_flux_x(i, j);

/home/hbollore/qaas-runs/170-290-5445/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/context.h: 69 - 69

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

69:   T &operator()(size_t i, size_t j) const { return data[i + j * sizeX]; }

Coverage (%)	Name	Source Location	Module
○100.00	__kmp_invoke_microtask		libomp.so

Path /

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	1.08
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	3.00
Bottlenecks	P6, P7,
Function	.omp_outlined..6#0x41cd00
Source	advec_mom.cpp:72-75,context.h:69-69
Source loop unroll info	unrolled by 4
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	4
CQA cycles	10.50
CQA cycles if no scalar integer	10.50
CQA cycles if FP arith vectorized	10.50
CQA cycles if fully vectorized	9.75
Front-end cycles	3.50
DIV/SQRT cycles	0.50
P0 cycles	0.50
P1 cycles	0.25
P2 cycles	0.25
P3 cycles	0.25
P4 cycles	0.25
P5 cycles	10.50
P6 cycles	10.50
P7 cycles	1.00
P8 cycles	1.00
P9 cycles	2.50
P10 cycles	2.50
P11 cycles	0.00
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	1.00 - 0.50
Inter-iter dependencies cycles	2
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	28.00
Nb uops	28.00
Nb loads	NA
Nb stores	2.00
Nb stack references	0.00
FLOP/cycle	0.76
Nb FLOP add-sub	8.00
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	15.24
Bytes prefetched	0.00
Bytes loaded	96.00
Bytes stored	64.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	70.37
Vectorization ratio load	60.00
Vectorization ratio store	100.00
Vectorization ratio mul	100.00
Vectorization ratio add_sub	72.73
Vectorization ratio fma	100.00
Vectorization ratio div_sqrt	100.00
Vectorization ratio other	28.57
Vector-efficiency ratio all	91.67
Vector-efficiency ratio load	70.00
Vector-efficiency ratio store	100.00
Vector-efficiency ratio mul	100.00
Vector-efficiency ratio add_sub	79.55
Vector-efficiency ratio fma	100.00
Vector-efficiency ratio div_sqrt	100.00
Vector-efficiency ratio other	100.00

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	1.08
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	3.00
Bottlenecks	P6, P7,
Function	.omp_outlined..6#0x41cd00
Source	advec_mom.cpp:72-75,context.h:69-69
Source loop unroll info	unrolled by 4
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	4
CQA cycles	10.50
CQA cycles if no scalar integer	10.50
CQA cycles if FP arith vectorized	10.50
CQA cycles if fully vectorized	9.75
Front-end cycles	3.50
DIV/SQRT cycles	0.50
P0 cycles	0.50
P1 cycles	0.25
P2 cycles	0.25
P3 cycles	0.25
P4 cycles	0.25
P5 cycles	10.50
P6 cycles	10.50
P7 cycles	1.00
P8 cycles	1.00
P9 cycles	2.50
P10 cycles	2.50
P11 cycles	0.00
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	1.00 - 0.50
Inter-iter dependencies cycles	2
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	28.00
Nb uops	28.00
Nb loads	NA
Nb stores	2.00
Nb stack references	0.00
FLOP/cycle	0.76
Nb FLOP add-sub	8.00
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	15.24
Bytes prefetched	0.00
Bytes loaded	96.00
Bytes stored	64.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	70.37
Vectorization ratio load	60.00
Vectorization ratio store	100.00
Vectorization ratio mul	100.00
Vectorization ratio add_sub	72.73
Vectorization ratio fma	100.00
Vectorization ratio div_sqrt	100.00
Vectorization ratio other	28.57
Vector-efficiency ratio all	91.67
Vector-efficiency ratio load	70.00
Vector-efficiency ratio store	100.00
Vector-efficiency ratio mul	100.00
Vector-efficiency ratio add_sub	79.55
Vector-efficiency ratio fma	100.00
Vector-efficiency ratio div_sqrt	100.00
Vector-efficiency ratio other	100.00

Path /

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

Function	.omp_outlined..6#0x41cd00
Source file and lines	advec_mom.cpp:72-75
Module	exec

The loop is defined in:

/home/hbollore/qaas-runs/170-290-5445/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/advec_mom.cpp:72-75
/home/hbollore/qaas-runs/170-290-5445/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/context.h:69

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

gain
potential
hint
expert

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 10.50 to 9.75 cycles (1.08x speedup).

Details

70% of VPU instructions are used in vector version (process two or more data elements in vector registers):

60% of VPU loads are used in vector version.
72% of VPU addition or subtraction instructions are used in vector version.
28% 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
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

No data for this section

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

The binary loop is composed of 8 FP arithmetical operations:

8: addition or subtraction

The binary loop is loading 96 bytes. The binary loop is storing 64 bytes.

Arithmetic intensity

Arithmetic intensity is 0.05 FP operations per loaded or stored byte.

General properties

nb instructions	28
loop length	112
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

3.50 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	0.50	0.50	0.25	0.25	0.25	0.25	10.50	10.50	1.00	1.00	2.50	2.50	0.00	0.00	0.00
cycles	0.50	0.50	0.25	0.25	0.25	0.25	10.50	10.50	1.00	1.00	2.50	2.50	0.00	0.00	0.00

Cycles executing div or sqrt instructions	1.00-0.50
Longest recurrence chain latency (RecMII)	2.00

Cycles summary

Front-end	3.50
Data deps.	2.00
Overall L1	10.50

Vectorization ratios

INT

all	68%
load	60%
store	100%
mul	100%
add-sub	66%
fma	100%
other	37%

FP

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	70%
load	60%
store	100%
mul	100%
add-sub	72%
fma	100%
div/sqrt	100%
other	28%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 10.50 cycles. At this rate:

9% of peak load performance is reached (9.14 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
12% of peak store performance is reached (6.10 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 41ce40

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput
MOVPRFX Z18, Z0	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
SDIV Z18.D, P0/M, Z18.D, Z2.D	1	0	0	0	0	0	0	1	0	0	0	0	0	7-20	1-0.50
ADD Z19.D, Z3.D, Z18.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MSB Z18.D, P0/M, Z2.D, Z0.D	1	0	0	0	0	0	0	1	0	0	0	0	0	5	2
ADD Z0.D, Z0.D, Z1.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
SUBS X12, X12, X10	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.33
SXTW Z19.D, P0/M, Z19.D	1	0	0	0	0	0	0	0	1	0	0	0	0	2	1
ADD Z20.D, Z4.D, Z18.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MOVPRFX Z21, Z20	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
SXTW Z21.D, P0/M, Z20.D	1	0	0	0	0	0	0	0	1	0	0	0	0	2	1
MOVPRFX Z22, Z21	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MLA Z22.D, P0/M, Z5.D, Z19.D	1	0	0	0	0	0	0	1	0	0	0	0	0	5	2
LD1D {Z22.D}, P0/Z, [X23, Z22.D,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
ADD Z18.D, Z7.D, Z18.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MOVPRFX Z23, Z21	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MLA Z23.D, P0/M, Z6.D, Z19.D	1	0	0	0	0	0	0	1	0	0	0	0	0	5	2
ST1D {Z22.D}, P0, [X21, Z23.D,LSL #3]	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	2	0.50
MOVPRFX Z23, Z16	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MUL Z23.D, P0/M, Z23.D, Z19.D	1	0	0	0	0	0	0	1	0	0	0	0	0	5	2
ADR Z18.D, [Z23, Z18.D,SXTW]	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
ADR Z20.D, [Z23, Z20.D,SXTW]	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MAD Z19.D, P0/M, Z17.D, Z21.D	1	0	0	0	0	0	0	1	0	0	0	0	0	5	2
LD1D {Z18.D}, P0/Z, [X27, Z18.D,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
LD1D {Z20.D}, P0/Z, [X27, Z20.D,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FADD Z18.D, Z22.D, Z18.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
FSUB Z18.D, Z18.D, Z20.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
ST1D {Z18.D}, P0, [X25, Z19.D,LSL #3]	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	2	0.50
B.NE 41ce40 <.omp_outlined..6+0x140>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	1	0.50

Function	.omp_outlined..6#0x41cd00
Source file and lines	advec_mom.cpp:72-75
Module	exec

The loop is defined in:

/home/hbollore/qaas-runs/170-290-5445/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/advec_mom.cpp:72-75
/home/hbollore/qaas-runs/170-290-5445/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/context.h:69

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

gain
potential
hint
expert

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 10.50 to 9.75 cycles (1.08x speedup).

Details

70% of VPU instructions are used in vector version (process two or more data elements in vector registers):

60% of VPU loads are used in vector version.
72% of VPU addition or subtraction instructions are used in vector version.
28% 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
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

No data for this section

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

The binary loop is composed of 8 FP arithmetical operations:

8: addition or subtraction

The binary loop is loading 96 bytes. The binary loop is storing 64 bytes.

Arithmetic intensity

Arithmetic intensity is 0.05 FP operations per loaded or stored byte.

General properties

nb instructions	28
loop length	112
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

3.50 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	0.50	0.50	0.25	0.25	0.25	0.25	10.50	10.50	1.00	1.00	2.50	2.50	0.00	0.00	0.00
cycles	0.50	0.50	0.25	0.25	0.25	0.25	10.50	10.50	1.00	1.00	2.50	2.50	0.00	0.00	0.00

Cycles executing div or sqrt instructions	1.00-0.50
Longest recurrence chain latency (RecMII)	2.00

Cycles summary

Front-end	3.50
Data deps.	2.00
Overall L1	10.50

Vectorization ratios

INT

all	68%
load	60%
store	100%
mul	100%
add-sub	66%
fma	100%
other	37%

FP

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	70%
load	60%
store	100%
mul	100%
add-sub	72%
fma	100%
div/sqrt	100%
other	28%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 10.50 cycles. At this rate:

9% of peak load performance is reached (9.14 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
12% of peak store performance is reached (6.10 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 41ce40

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput
MOVPRFX Z18, Z0	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
SDIV Z18.D, P0/M, Z18.D, Z2.D	1	0	0	0	0	0	0	1	0	0	0	0	0	7-20	1-0.50
ADD Z19.D, Z3.D, Z18.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MSB Z18.D, P0/M, Z2.D, Z0.D	1	0	0	0	0	0	0	1	0	0	0	0	0	5	2
ADD Z0.D, Z0.D, Z1.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
SUBS X12, X12, X10	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.33
SXTW Z19.D, P0/M, Z19.D	1	0	0	0	0	0	0	0	1	0	0	0	0	2	1
ADD Z20.D, Z4.D, Z18.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MOVPRFX Z21, Z20	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
SXTW Z21.D, P0/M, Z20.D	1	0	0	0	0	0	0	0	1	0	0	0	0	2	1
MOVPRFX Z22, Z21	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MLA Z22.D, P0/M, Z5.D, Z19.D	1	0	0	0	0	0	0	1	0	0	0	0	0	5	2
LD1D {Z22.D}, P0/Z, [X23, Z22.D,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
ADD Z18.D, Z7.D, Z18.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MOVPRFX Z23, Z21	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MLA Z23.D, P0/M, Z6.D, Z19.D	1	0	0	0	0	0	0	1	0	0	0	0	0	5	2
ST1D {Z22.D}, P0, [X21, Z23.D,LSL #3]	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	2	0.50
MOVPRFX Z23, Z16	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MUL Z23.D, P0/M, Z23.D, Z19.D	1	0	0	0	0	0	0	1	0	0	0	0	0	5	2
ADR Z18.D, [Z23, Z18.D,SXTW]	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
ADR Z20.D, [Z23, Z20.D,SXTW]	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
MAD Z19.D, P0/M, Z17.D, Z21.D	1	0	0	0	0	0	0	1	0	0	0	0	0	5	2
LD1D {Z18.D}, P0/Z, [X27, Z18.D,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
LD1D {Z20.D}, P0/Z, [X27, Z20.D,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FADD Z18.D, Z22.D, Z18.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
FSUB Z18.D, Z18.D, Z20.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
ST1D {Z18.D}, P0, [X25, Z19.D,LSL #3]	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	2	0.50
B.NE 41ce40 <.omp_outlined..6+0x140>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	1	0.50

Report Configuration

Vectorization

Details

Workaround

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code

Vectorization

Details

Workaround

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code