OV - exec - Loop 451

	Loop Id: 451	Module: exec	Source: advec_cell_kernel.f90-pp.f90:204-248 [...]	Coverage: 3.7%

0x41b420 FADD	D4, D17, D16

0x41b424 SUB	W25, W25, #1

0x41b428 CMP	W25, #1

0x41b42c FMUL	D3, D3, D4

0x41b430 STR	D3, [X23, X14,LSL #3]    [5]

0x41b434 ADD	X14, X14, #1

0x41b438 B.LE	41b384

0x41b43c LDR	D3, [X7, X14,LSL #3]    [2]

0x41b440 ORR	W26, WZR, W5

0x41b444 ORR	W28, WZR, W4

0x41b448 FCMP	D3, #0

0x41b44c ORR	W27, WZR, W8

0x41b450 ORR	W19, WZR, W4

0x41b454 B.GT	41b478

0x41b458 LDR	X19, [X20, #32]    [7]

0x41b45c ORR	W28, WZR, W8

0x41b460 ORR	W27, WZR, W4

0x41b464 LDR	W19, [X19]    [12]

0x41b468 ADD	W19, W19, #2

0x41b46c CMP	W1, W19

0x41b470 CSEL	W26, W1, W19, #11

0x41b474 ORR	W19, WZR, W26

0x41b478 ADD	X29, X22, X14

0x41b47c SBFM	X28, X28, #0, #31

0x41b480 SBFM	X19, X19, #0, #31

0x41b484 FABS	D5, D3

0x41b488 SBFM	X26, X26, #0, #31

0x41b48c SBFM	X27, X27, #0, #31

0x41b490 SUB	X19, X19, X0

0x41b494 MADD	X29, X16, X28, X29

0x41b498 LDR	D4, [X17, X29,LSL #3]    [13]

0x41b49c ADD	X29, X24, X14

0x41b4a0 MADD	X26, X15, X26, X29

0x41b4a4 MADD	X27, X15, X27, X29

0x41b4a8 LDR	D16, [X2, X26,LSL #3]    [3]

0x41b4ac FDIV	D17, D5, D4

0x41b4b0 LDR	D5, [X18, X19,LSL #3]    [11]

0x41b4b4 MADD	X19, X15, X28, X29

0x41b4b8 LDR	D7, [X2, X19,LSL #3]    [1]

0x41b4bc FDIV	D5, D2, D5

0x41b4c0 FSUB	D19, D7, S16

0x41b4c4 LDR	D16, [X2, X27,LSL #3]    [4]

0x41b4c8 FSUB	D18, D16, S7

0x41b4cc MOVI	D16, #0

0x41b4d0 FADD	D6, D17, D0

0x41b4d4 FMUL	D20, D19, D18

0x41b4d8 FCMP	D20, #0

0x41b4dc MOVI	D20, #0

0x41b4e0 FMUL	D5, D6, D5

0x41b4e4 FSUB	D6, D1, S17

0x41b4e8 B.LE	41b528

0x41b4ec FABS	D19, D19

0x41b4f0 FABS	D20, D18

0x41b4f4 FSUB	D17, D0, S17

0x41b4f8 FCMP	D19, D20

0x41b4fc FCSEL	D21, D20, D19, #10

0x41b500 FMUL	D19, D5, D19

0x41b504 FMADD	D19, D20, D6, D19

0x41b508 FMOV	D20, X13

0x41b50c FMUL	D19, D19, D20

0x41b510 FNEG	D20, D17

0x41b514 FCMP	D21, D19

0x41b518 FCSEL	D19, D19, D21, #10

0x41b51c FCMP	D18, #0

0x41b520 FCSEL	D17, D17, D20, #8

0x41b524 FMUL	D20, D17, D19

0x41b528 FADD	D17, D7, D20

0x41b52c LDR	D18, [X3, X26,LSL #3]    [9]

0x41b530 FMUL	D3, D3, D17

0x41b534 LDR	D17, [X3, X19,LSL #3]    [8]

0x41b538 STR	D3, [X6, X14,LSL #3]    [6]

0x41b53c FSUB	D19, D17, S18

0x41b540 LDR	D18, [X3, X27,LSL #3]    [10]

0x41b544 FSUB	D18, D18, S17

0x41b548 FMUL	D20, D19, D18

0x41b54c FCMP	D20, #0

0x41b550 B.LE	41b420

0x41b554 FMUL	D4, D4, D7

0x41b558 FABS	D16, D3

0x41b55c FABS	D7, D19

0x41b560 FMUL	D5, D5, D7

0x41b564 FDIV	D4, D16, D4

0x41b568 FABS	D16, D18

0x41b56c FMADD	D5, D16, D6, D5

0x41b570 FMOV	D6, X13

0x41b574 FCMP	D7, D16

0x41b578 FCSEL	D19, D16, D7, #10

0x41b57c FMUL	D5, D5, D6

0x41b580 FCMP	D19, D5

0x41b584 FCSEL	D5, D5, D19, #10

0x41b588 FCMP	D18, #0

0x41b58c FSUB	D4, D0, S4

0x41b590 FNEG	D6, D4

0x41b594 FCSEL	D4, D4, D6, #8

0x41b598 FMUL	D16, D4, D5

0x41b59c B	41b420

/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: 204 - 248

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

204: # 204 "/home/hbollore/qaas-runs/170-308-3604/intel/CloverLeafFC/build/CloverLeafFC/CloverLeaf_ref/kernels/advec_cell_kernel.f90"

[...]

210:           ELSE

[...]

216: # 216 "/home/hbollore/qaas-runs/170-308-3604/intel/CloverLeafFC/build/CloverLeafFC/CloverLeaf_ref/kernels/advec_cell_kernel.f90"

217:           sigmat=ABS(vol_flux_y(j,k))/pre_vol(j,donor)

218:           sigma3=(1.0_8+sigmat)*(vertexdy(k)/vertexdy(dif))

219:           sigma4=2.0_8-sigmat

220: # 220 "/home/hbollore/qaas-runs/170-308-3604/intel/CloverLeafFC/build/CloverLeafFC/CloverLeaf_ref/kernels/advec_cell_kernel.f90"

221:           sigma=sigmat

222:           sigmav=sigmat

223: # 223 "/home/hbollore/qaas-runs/170-308-3604/intel/CloverLeafFC/build/CloverLeafFC/CloverLeaf_ref/kernels/advec_cell_kernel.f90"

224:           diffuw=density1(j,donor)-density1(j,upwind)

225:           diffdw=density1(j,downwind)-density1(j,donor)

226:           wind=1.0_8

227:           IF(diffdw.LE.0.0) wind=-1.0_8

228:           IF(diffuw*diffdw.GT.0.0)THEN

229:             limiter=(1.0_8-sigmav)*wind*MIN(ABS(diffuw),ABS(diffdw)&

230:               ,one_by_six*(sigma3*ABS(diffuw)+sigma4*ABS(diffdw)))

231:           ELSE

232:             limiter=0.0

233:           ENDIF

234:           mass_flux_y(j,k)=vol_flux_y(j,k)*(density1(j,donor)+limiter)

235: # 235 "/home/hbollore/qaas-runs/170-308-3604/intel/CloverLeafFC/build/CloverLeafFC/CloverLeaf_ref/kernels/advec_cell_kernel.f90"

236:           sigmam=ABS(mass_flux_y(j,k))/(density1(j,donor)*pre_vol(j,donor))

237:           diffuw=energy1(j,donor)-energy1(j,upwind)

238:           diffdw=energy1(j,downwind)-energy1(j,donor)

239:           wind=1.0_8

240:           IF(diffdw.LE.0.0) wind=-1.0_8

241:           IF(diffuw*diffdw.GT.0.0)THEN

242:             limiter=(1.0_8-sigmam)*wind*MIN(ABS(diffuw),ABS(diffdw)&

243:               ,one_by_six*(sigma3*ABS(diffuw)+sigma4*ABS(diffdw)))

244:           ELSE

245:             limiter=0.0

246:           ENDIF

247:           ener_flux(j,k)=mass_flux_y(j,k)*(energy1(j,donor)+limiter)

248: # 248 "/home/hbollore/qaas-runs/170-308-3604/intel/CloverLeafFC/build/CloverLeafFC/CloverLeaf_ref/kernels/advec_cell_kernel.f90"

Path /

Metric	Value
CQA speedup if no scalar integer	3.82
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	7.72
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.17
Bottlenecks	P6, P7, P8, P9,
Function	__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_
Source	advec_cell_kernel.f90-pp.f90:204-204,advec_cell_kernel.f90-pp.f90:210-210,advec_cell_kernel.f90-pp.f90:216-248
Source loop unroll info	multi-versionned
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	14.00
CQA cycles if no scalar integer	3.67
CQA cycles if FP arith vectorized	14.00
CQA cycles if fully vectorized	1.81
Front-end cycles	12.00
DIV/SQRT cycles	2.50
P0 cycles	2.50
P1 cycles	6.00
P2 cycles	6.00
P3 cycles	6.00
P4 cycles	6.00
P5 cycles	14.00
P6 cycles	14.00
P7 cycles	14.00
P8 cycles	14.00
P9 cycles	4.33
P10 cycles	4.33
P11 cycles	4.33
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	3.00 - 1.50
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	96.00
Nb uops	96.00
Nb loads	NA
Nb stores	2.00
Nb stack references	0.00
FLOP/cycle	2.07
Nb FLOP add-sub	10.00
Nb FLOP mul	12.00
Nb FLOP fma	2.00
Nb FLOP div	3.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	7.14
Bytes prefetched	0.00
Bytes loaded	84.00
Bytes stored	16.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	21.53
Vector-efficiency ratio load	25.00
Vector-efficiency ratio store	25.00
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	12.50
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	16.67

Metric	Value
CQA speedup if no scalar integer	3.82
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	7.72
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.17
Bottlenecks	P6, P7, P8, P9,
Function	__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_
Source	advec_cell_kernel.f90-pp.f90:204-204,advec_cell_kernel.f90-pp.f90:210-210,advec_cell_kernel.f90-pp.f90:216-248
Source loop unroll info	multi-versionned
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	14.00
CQA cycles if no scalar integer	3.67
CQA cycles if FP arith vectorized	14.00
CQA cycles if fully vectorized	1.81
Front-end cycles	12.00
DIV/SQRT cycles	2.50
P0 cycles	2.50
P1 cycles	6.00
P2 cycles	6.00
P3 cycles	6.00
P4 cycles	6.00
P5 cycles	14.00
P6 cycles	14.00
P7 cycles	14.00
P8 cycles	14.00
P9 cycles	4.33
P10 cycles	4.33
P11 cycles	4.33
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	3.00 - 1.50
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	96.00
Nb uops	96.00
Nb loads	NA
Nb stores	2.00
Nb stack references	0.00
FLOP/cycle	2.07
Nb FLOP add-sub	10.00
Nb FLOP mul	12.00
Nb FLOP fma	2.00
Nb FLOP div	3.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	7.14
Bytes prefetched	0.00
Bytes loaded	84.00
Bytes stored	16.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	21.53
Vector-efficiency ratio load	25.00
Vector-efficiency ratio store	25.00
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	12.50
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	16.67

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.f90:204-248
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.f90:204,210,216-248.

The related source loop is multi-versionned.
Warnings:

Ignoring paths for analysis
Failed to get the number of paths
RecMII not computed since number of paths is unknown or > max_paths

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 14.00 to 3.67 cycles (3.82x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)
To reference allocatable arrays, use "allocatable" instead of "pointer" pointers or qualify them with the "contiguous" attribute (Fortran 2008)
For structures, limit to one indirection. For example, use a_b%c instead of a%b%c with a_b set to a%b before this loop

Vectorization

Your loop is not vectorized. Only 21% of vector register length is used (average across all VPU instructions). By vectorizing your loop, you can lower the cost of an iteration from 14.00 to 1.81 cycles (7.72x speedup).

Details

All VPU instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a 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: Fortran storage order is column-major: do i do j a(i,j) = b(i,j) (slow, non stride 1) => do i do j a(j,i) = b(i,j) (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): do i a(i)%x = b(i)%x (slow, non stride 1) => do i a%x(i) = b%x(i) (fast, stride 1)

FMA

Detected 2 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 29 FP arithmetical operations:

12: addition or subtraction (2 inside FMA instructions)
14: multiply (2 inside FMA instructions)
3: divide

The binary loop is loading 84 bytes. The binary loop is storing 16 bytes.

Arithmetic intensity

Arithmetic intensity is 0.29 FP operations per loaded or stored byte.

General properties

nb instructions	96
loop length	384
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

12.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	2.50	2.50	6.00	6.00	6.00	6.00	14.00	14.00	14.00	14.00	4.33	4.33	4.33	0.00	0.00
cycles	2.50	2.50	6.00	6.00	6.00	6.00	14.00	14.00	14.00	14.00	4.33	4.33	4.33	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

3.00-1.50

Cycles summary

Front-end	12.00
Overall L1	14.00

Vectorization ratios

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 14.00 cycles. At this rate:

6% of peak load performance is reached (6.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (1.14 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 41b420

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	Latency	Recip. throughput
FADD D4, D17, D16	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
SUB W25, W25, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
CMP W25, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
FMUL D3, D3, D4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
STR D3, [X23, X14,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
ADD X14, X14, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
B.LE 41b384 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0xe74>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR D3, [X7, X14,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
ORR W26, WZR, W5	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ORR W28, WZR, W4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
FCMP D3, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
ORR W27, WZR, W8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ORR W19, WZR, W4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
B.GT 41b478 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0xf68>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X19, [X20, #32]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
ORR W28, WZR, W8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ORR W27, WZR, W4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
LDR W19, [X19]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
ADD W19, W19, #2	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
CMP W1, W19	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
CSEL W26, W1, W19, #11	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ORR W19, WZR, W26	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ADD X29, X22, X14	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
SBFM X28, X28, #0, #31	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
SBFM X19, X19, #0, #31	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
FABS D5, D3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
SBFM X26, X26, #0, #31	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
SBFM X27, X27, #0, #31	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
SUB X19, X19, X0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
MADD X29, X16, X28, X29	1	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1
LDR D4, [X17, X29,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
ADD X29, X24, X14	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
MADD X26, X15, X26, X29	1	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1
MADD X27, X15, X27, X29	1	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1
LDR D16, [X2, X26,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FDIV D17, D5, D4	1	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	7-15	1-0.50
LDR D5, [X18, X19,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
MADD X19, X15, X28, X29	1	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1
LDR D7, [X2, X19,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FDIV D5, D2, D5	1	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	7-15	1-0.50
FSUB D19, D7, S16	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
LDR D16, [X2, X27,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FSUB D18, D16, S7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
MOVI D16, #0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FADD D6, D17, D0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMUL D20, D19, D18	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FCMP D20, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
MOVI D20, #0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMUL D5, D6, D5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FSUB D6, D1, S17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
B.LE 41b528 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0x1018>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FABS D19, D19	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FABS D20, D18	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FSUB D17, D0, S17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FCMP D19, D20	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FCSEL D21, D20, D19, #10	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FMUL D19, D5, D19	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FMADD D19, D20, D6, D19	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	4	0.25
FMOV D20, X13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMUL D19, D19, D20	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FNEG D20, D17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FCMP D21, D19	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FCSEL D19, D19, D21, #10	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FCMP D18, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FCSEL D17, D17, D20, #8	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FMUL D20, D17, D19	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FADD D17, D7, D20	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
LDR D18, [X3, X26,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FMUL D3, D3, D17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
LDR D17, [X3, X19,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
STR D3, [X6, X14,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
FSUB D19, D17, S18	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
LDR D18, [X3, X27,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FSUB D18, D18, S17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMUL D20, D19, D18	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FCMP D20, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
B.LE 41b420 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0xf10>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FMUL D4, D4, D7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FABS D16, D3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FABS D7, D19	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMUL D5, D5, D7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FDIV D4, D16, D4	1	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	7-15	1-0.50
FABS D16, D18	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMADD D5, D16, D6, D5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	4	0.25
FMOV D6, X13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FCMP D7, D16	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FCSEL D19, D16, D7, #10	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FMUL D5, D5, D6	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FCMP D19, D5	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FCSEL D5, D5, D19, #10	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FCMP D18, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FSUB D4, D0, S4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FNEG D6, D4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FCSEL D4, D4, D6, #8	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FMUL D16, D4, D5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
B 41b420 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0xf10>	1	0.50	0.50	0	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.f90:204-248
Module	exec

Ignoring paths for analysis
Failed to get the number of paths
RecMII not computed since number of paths is unknown or > max_paths

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 14.00 to 3.67 cycles (3.82x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)
To reference allocatable arrays, use "allocatable" instead of "pointer" pointers or qualify them with the "contiguous" attribute (Fortran 2008)
For structures, limit to one indirection. For example, use a_b%c instead of a%b%c with a_b set to a%b before this loop

Vectorization

Details

All VPU instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a 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: Fortran storage order is column-major: do i do j a(i,j) = b(i,j) (slow, non stride 1) => do i do j a(j,i) = b(i,j) (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): do i a(i)%x = b(i)%x (slow, non stride 1) => do i a%x(i) = b%x(i) (fast, stride 1)

FMA

Detected 2 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 29 FP arithmetical operations:

12: addition or subtraction (2 inside FMA instructions)
14: multiply (2 inside FMA instructions)
3: divide

The binary loop is loading 84 bytes. The binary loop is storing 16 bytes.

Arithmetic intensity

Arithmetic intensity is 0.29 FP operations per loaded or stored byte.

General properties

nb instructions	96
loop length	384
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

12.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	2.50	2.50	6.00	6.00	6.00	6.00	14.00	14.00	14.00	14.00	4.33	4.33	4.33	0.00	0.00
cycles	2.50	2.50	6.00	6.00	6.00	6.00	14.00	14.00	14.00	14.00	4.33	4.33	4.33	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

3.00-1.50

Cycles summary

Front-end	12.00
Overall L1	14.00

Vectorization ratios

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 14.00 cycles. At this rate:

6% of peak load performance is reached (6.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (1.14 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 41b420

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	Latency	Recip. throughput
FADD D4, D17, D16	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
SUB W25, W25, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
CMP W25, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
FMUL D3, D3, D4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
STR D3, [X23, X14,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
ADD X14, X14, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
B.LE 41b384 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0xe74>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR D3, [X7, X14,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
ORR W26, WZR, W5	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ORR W28, WZR, W4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
FCMP D3, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
ORR W27, WZR, W8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ORR W19, WZR, W4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
B.GT 41b478 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0xf68>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X19, [X20, #32]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
ORR W28, WZR, W8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ORR W27, WZR, W4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
LDR W19, [X19]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
ADD W19, W19, #2	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
CMP W1, W19	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
CSEL W26, W1, W19, #11	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ORR W19, WZR, W26	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ADD X29, X22, X14	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
SBFM X28, X28, #0, #31	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
SBFM X19, X19, #0, #31	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
FABS D5, D3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
SBFM X26, X26, #0, #31	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
SBFM X27, X27, #0, #31	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
SUB X19, X19, X0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
MADD X29, X16, X28, X29	1	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1
LDR D4, [X17, X29,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
ADD X29, X24, X14	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
MADD X26, X15, X26, X29	1	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1
MADD X27, X15, X27, X29	1	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1
LDR D16, [X2, X26,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FDIV D17, D5, D4	1	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	7-15	1-0.50
LDR D5, [X18, X19,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
MADD X19, X15, X28, X29	1	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1
LDR D7, [X2, X19,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FDIV D5, D2, D5	1	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	7-15	1-0.50
FSUB D19, D7, S16	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
LDR D16, [X2, X27,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FSUB D18, D16, S7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
MOVI D16, #0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FADD D6, D17, D0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMUL D20, D19, D18	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FCMP D20, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
MOVI D20, #0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMUL D5, D6, D5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FSUB D6, D1, S17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
B.LE 41b528 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0x1018>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FABS D19, D19	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FABS D20, D18	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FSUB D17, D0, S17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FCMP D19, D20	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FCSEL D21, D20, D19, #10	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FMUL D19, D5, D19	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FMADD D19, D20, D6, D19	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	4	0.25
FMOV D20, X13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMUL D19, D19, D20	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FNEG D20, D17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FCMP D21, D19	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FCSEL D19, D19, D21, #10	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FCMP D18, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FCSEL D17, D17, D20, #8	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FMUL D20, D17, D19	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FADD D17, D7, D20	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
LDR D18, [X3, X26,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FMUL D3, D3, D17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
LDR D17, [X3, X19,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
STR D3, [X6, X14,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
FSUB D19, D17, S18	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
LDR D18, [X3, X27,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FSUB D18, D18, S17	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMUL D20, D19, D18	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FCMP D20, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
B.LE 41b420 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0xf10>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FMUL D4, D4, D7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FABS D16, D3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FABS D7, D19	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMUL D5, D5, D7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FDIV D4, D16, D4	1	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	7-15	1-0.50
FABS D16, D18	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FMADD D5, D16, D6, D5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	4	0.25
FMOV D6, X13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FCMP D7, D16	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FCSEL D19, D16, D7, #10	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FMUL D5, D5, D6	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FCMP D19, D5	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FCSEL D5, D5, D19, #10	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FCMP D18, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
FSUB D4, D0, S4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FNEG D6, D4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
FCSEL D4, D4, D6, #8	1	0	0	0	0	0	0	0.50	0.50	0	0	0	0	0	2	0.50
FMUL D16, D4, D5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
B 41b420 <__nv_advec_cell_kernel_module_advec_cell_kernel__F1L81_1_+0xf10>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

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

Code clean check

Workaround

Vectorization

Details

Workaround

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