OV - exec - Loop 449

	Loop Id: 449	Module: exec	Source: advec_cell_kernel.f90-pp.f90:256-261	Coverage: 4.05%

0x41b724 LD1D	{Z0.D}, P0/Z, [X26, X23,LSL #3]    [1]

0x41b728 LD1D	{Z1.D}, P0/Z, [X19, X23,LSL #3]    [8]

0x41b72c FMUL	Z1.D, Z0.D, Z1.D

0x41b730 LD1D	{Z2.D}, P0/Z, [X30, X23,LSL #3]    [4]

0x41b734 LD1D	{Z3.D}, P0/Z, [X11, X23,LSL #3]    [6]

0x41b738 LD1D	{Z4.D}, P0/Z, [X28, X23,LSL #3]    [9]

0x41b73c FADD	Z2.D, Z1.D, Z2.D

0x41b740 FSUB	Z2.D, Z2.D, Z3.D

0x41b744 LD1D	{Z3.D}, P0/Z, [X24, X23,LSL #3]    [7]

0x41b748 FMAD	Z1.D, P0/M, Z4.D, Z3.D

0x41b74c LD1D	{Z3.D}, P0/Z, [X13, X23,LSL #3]    [2]

0x41b750 FSUB	Z1.D, Z1.D, Z3.D

0x41b754 LD1D	{Z3.D}, P0/Z, [X27, X23,LSL #3]    [3]

0x41b758 FDIV	Z1.D, P0/M, Z1.D, Z2.D

0x41b75c FADD	Z0.D, Z0.D, Z3.D

0x41b760 LD1D	{Z3.D}, P0/Z, [X12, X23,LSL #3]    [5]

0x41b764 FSUB	Z0.D, Z0.D, Z3.D

0x41b768 FDIVR	Z0.D, P0/M, Z0.D, Z2.D

0x41b76c ST1D	{Z1.D}, P0, [X28, X23,LSL #3]    [9]

0x41b770 ST1D	{Z0.D}, P0, [X19, X23,LSL #3]    [8]

0x41b774 ADD	X23, X23, X14

0x41b778 CMP	X29, X23

0x41b77c B.NE	41b724

/home/hbollore/qaas-runs/170-308-3604/intel/CloverLeafFC/build/armclang_1/CMakeFiles/clover_leaf.dir/CloverLeaf_ref/kernels/advec_cell_kernel.f90-pp.f90: 256 - 261

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

256:         DO j=x_min,x_max

257:           pre_mass_s=density1(j,k)*pre_vol(j,k)

258:           post_mass_s=pre_mass_s+mass_flux_y(j,k)-mass_flux_y(j,k+1)

259:           post_ener_s=(energy1(j,k)*pre_mass_s+ener_flux(j,k)-ener_flux(j,k+1))/post_mass_s

260:           advec_vol_s=pre_vol(j,k)+vol_flux_y(j,k)-vol_flux_y(j,k+1)

261:           density1(j,k)=post_mass_s/advec_vol_s

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	1.22
Bottlenecks	P10, P11,
Function	__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_
Source	advec_cell_kernel.f90-pp.f90:256-261
Source loop unroll info	unrolled by 4
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	4
CQA cycles	5.50
CQA cycles if no scalar integer	5.50
CQA cycles if FP arith vectorized	5.50
CQA cycles if fully vectorized	5.50
Front-end cycles	2.88
DIV/SQRT cycles	0.50
P0 cycles	0.50
P1 cycles	0.50
P2 cycles	0.50
P3 cycles	0.50
P4 cycles	0.50
P5 cycles	4.50
P6 cycles	4.50
P7 cycles	3.50
P8 cycles	3.50
P9 cycles	5.50
P10 cycles	5.50
P11 cycles	0.00
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	2.00 - 1.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	23.00
Nb uops	23.00
Nb loads	NA
Nb stores	2.00
Nb stack references	0.00
FLOP/cycle	7.27
Nb FLOP add-sub	20.00
Nb FLOP mul	4.00
Nb FLOP fma	4.00
Nb FLOP div	8.00
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	288.00
Bytes stored	64.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
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	100.00
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	100.00
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	1.22
Bottlenecks	P10, P11,
Function	__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_
Source	advec_cell_kernel.f90-pp.f90:256-261
Source loop unroll info	unrolled by 4
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	4
CQA cycles	5.50
CQA cycles if no scalar integer	5.50
CQA cycles if FP arith vectorized	5.50
CQA cycles if fully vectorized	5.50
Front-end cycles	2.88
DIV/SQRT cycles	0.50
P0 cycles	0.50
P1 cycles	0.50
P2 cycles	0.50
P3 cycles	0.50
P4 cycles	0.50
P5 cycles	4.50
P6 cycles	4.50
P7 cycles	3.50
P8 cycles	3.50
P9 cycles	5.50
P10 cycles	5.50
P11 cycles	0.00
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	2.00 - 1.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	23.00
Nb uops	23.00
Nb loads	NA
Nb stores	2.00
Nb stack references	0.00
FLOP/cycle	7.27
Nb FLOP add-sub	20.00
Nb FLOP mul	4.00
Nb FLOP fma	4.00
Nb FLOP div	8.00
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	288.00
Bytes stored	64.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
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	100.00
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	100.00
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_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_
Source file and lines	advec_cell_kernel.f90-pp.f90:256-261
Module	exec

The loop is defined in /home/hbollore/qaas-runs/170-308-3604/intel/CloverLeafFC/build/armclang_1/CMakeFiles/clover_leaf.dir/CloverLeaf_ref/kernels/advec_cell_kernel.f90-pp.f90:256-261.

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

gain
potential
hint
expert

Vectorization

Your loop is fully vectorized, using full register length.

Details

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

FMA

Detected 4 FMA (fused multiply-add) operations. Presence of both ADD/SUB and MUL operations.

Workaround

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).

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

The binary loop is composed of 40 FP arithmetical operations:

24: addition or subtraction (4 inside FMA instructions)
8: multiply (4 inside FMA instructions)
8: divide

The binary loop is loading 288 bytes. The binary loop is storing 64 bytes.

Arithmetic intensity

Arithmetic intensity is 0.11 FP operations per loaded or stored byte.

General properties

nb instructions	23
loop length	92
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

2.88 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	0.50	0.50	0.50	0.50	0.50	0.50	4.50	4.50	1.00	1.00	5.50	5.50	0.00	0.00	0.00
cycles	0.50	0.50	0.50	0.50	0.50	0.50	4.50	4.50	3.50	3.50	5.50	5.50	0.00	0.00	0.00

Execution ports to units layout:

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

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

Cycles summary

Front-end	2.88
Data deps.	1.00
Overall L1	5.50

Vectorization ratios

INT

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

INT+FP

all	100%
load	100%
store	100%
mul	100%
add-sub	100%
fma	100%
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 5.50 cycles. At this rate:

54% of peak load performance is reached (52.36 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
24% of peak store performance is reached (11.64 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 41b724

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput
LD1D {Z0.D}, P0/Z, [X26, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
LD1D {Z1.D}, P0/Z, [X19, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FMUL Z1.D, Z0.D, Z1.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50
LD1D {Z2.D}, P0/Z, [X30, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
LD1D {Z3.D}, P0/Z, [X11, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
LD1D {Z4.D}, P0/Z, [X28, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FADD Z2.D, Z1.D, Z2.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
FSUB Z2.D, Z2.D, Z3.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
LD1D {Z3.D}, P0/Z, [X24, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FMAD Z1.D, P0/M, Z4.D, Z3.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	4	0.50
LD1D {Z3.D}, P0/Z, [X13, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FSUB Z1.D, Z1.D, Z3.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
LD1D {Z3.D}, P0/Z, [X27, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FDIV Z1.D, P0/M, Z1.D, Z2.D	1	0	0	0	0	0	0	1	0	0	0	0	0	7-15	1-0.50
FADD Z0.D, Z0.D, Z3.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
LD1D {Z3.D}, P0/Z, [X12, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FSUB Z0.D, Z0.D, Z3.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
FDIVR Z0.D, P0/M, Z0.D, Z2.D	1	0	0	0	0	0	0	1	0	0	0	0	0	7-15	1-0.50
ST1D {Z1.D}, P0, [X28, X23,LSL #3]	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	2	0.50
ST1D {Z0.D}, P0, [X19, X23,LSL #3]	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	2	0.50
ADD X23, X23, X14	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.25
CMP X29, X23	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.33
B.NE 41b724 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0x1214>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	1	0.50

Function	__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_
Source file and lines	advec_cell_kernel.f90-pp.f90:256-261
Module	exec

gain
potential
hint
expert

Vectorization

Your loop is fully vectorized, using full register length.

Details

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

FMA

Detected 4 FMA (fused multiply-add) operations. Presence of both ADD/SUB and MUL operations.

Workaround

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

The binary loop is composed of 40 FP arithmetical operations:

24: addition or subtraction (4 inside FMA instructions)
8: multiply (4 inside FMA instructions)
8: divide

The binary loop is loading 288 bytes. The binary loop is storing 64 bytes.

Arithmetic intensity

Arithmetic intensity is 0.11 FP operations per loaded or stored byte.

General properties

nb instructions	23
loop length	92
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

2.88 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	0.50	0.50	0.50	0.50	0.50	0.50	4.50	4.50	1.00	1.00	5.50	5.50	0.00	0.00	0.00
cycles	0.50	0.50	0.50	0.50	0.50	0.50	4.50	4.50	3.50	3.50	5.50	5.50	0.00	0.00	0.00

Execution ports to units layout:

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

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

Cycles summary

Front-end	2.88
Data deps.	1.00
Overall L1	5.50

Vectorization ratios

INT

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

INT+FP

all	100%
load	100%
store	100%
mul	100%
add-sub	100%
fma	100%
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 5.50 cycles. At this rate:

54% of peak load performance is reached (52.36 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
24% of peak store performance is reached (11.64 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 41b724

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput
LD1D {Z0.D}, P0/Z, [X26, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
LD1D {Z1.D}, P0/Z, [X19, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FMUL Z1.D, Z0.D, Z1.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50
LD1D {Z2.D}, P0/Z, [X30, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
LD1D {Z3.D}, P0/Z, [X11, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
LD1D {Z4.D}, P0/Z, [X28, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FADD Z2.D, Z1.D, Z2.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
FSUB Z2.D, Z2.D, Z3.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
LD1D {Z3.D}, P0/Z, [X24, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FMAD Z1.D, P0/M, Z4.D, Z3.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	4	0.50
LD1D {Z3.D}, P0/Z, [X13, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FSUB Z1.D, Z1.D, Z3.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
LD1D {Z3.D}, P0/Z, [X27, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FDIV Z1.D, P0/M, Z1.D, Z2.D	1	0	0	0	0	0	0	1	0	0	0	0	0	7-15	1-0.50
FADD Z0.D, Z0.D, Z3.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
LD1D {Z3.D}, P0/Z, [X12, X23,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	6	0.50
FSUB Z0.D, Z0.D, Z3.D	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	2	0.50
FDIVR Z0.D, P0/M, Z0.D, Z2.D	1	0	0	0	0	0	0	1	0	0	0	0	0	7-15	1-0.50
ST1D {Z1.D}, P0, [X28, X23,LSL #3]	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	2	0.50
ST1D {Z0.D}, P0, [X19, X23,LSL #3]	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	2	0.50
ADD X23, X23, X14	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.25
CMP X29, X23	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.33
B.NE 41b724 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0x1214>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	1	0.50

Report Configuration

Vectorization

Details

FMA

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

FMA

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