OV - - Loop 163

	Loop Id: 163	Module: exec	Source: advec_mom.cpp:72-75 [...]	Coverage: 1.54%

0x237dd0 MOV	%R9,%RAX

0x237dd3 CQTO

0x237dd5 IDIV	%R13

0x237dd8 ADD	%R11D,%R8D

0x237ddb ADD	%R11D,%EAX

0x237dde MOV	-0x58(%RBP),%R11    [11]

0x237de2 LEA	(%R11,%RDI,1),%R9D

0x237de6 ADD	%EDX,%R11D

0x237de9 MOVSXD	%R9D,%RBX

0x237dec MOVSXD	%R11D,%R9

0x237def MOVSXD	%R8D,%R8

0x237df2 CLTQ

0x237df4 MOV	-0x50(%RBP),%R15    [11]

0x237df8 MOV	%R15,%R11

0x237dfb IMUL	%R8,%R11

0x237dff ADD	%RBX,%R11

0x237e02 MOV	-0x80(%RBP),%R10    [11]

0x237e06 VMOVSD	(%R10,%R11,8),%XMM0    [4]

0x237e0c IMUL	%RAX,%R15

0x237e10 ADD	%R9,%R15

0x237e13 VMOVSD	(%R10,%R15,8),%XMM1    [7]

0x237e19 MOV	-0xa8(%RBP),%R10    [11]

0x237e20 MOV	%R10,%R11

0x237e23 IMUL	%R8,%R11

0x237e27 ADD	%RBX,%R11

0x237e2a MOV	-0x78(%RBP),%R15    [11]

0x237e2e VMOVSD	%XMM0,(%R15,%R11,8)    [5]

0x237e34 MOV	%R10,%R11

0x237e37 IMUL	%RAX,%R11

0x237e3b ADD	%R9,%R11

0x237e3e VMOVSD	%XMM1,(%R15,%R11,8)    [2]

0x237e44 MOV	-0x38(%RBP),%R11    [11]

0x237e48 ADD	%R11D,%EDI

0x237e4b ADD	%R11D,%EDX

0x237e4e MOVSXD	%EDI,%RDI

0x237e51 MOVSXD	%EDX,%RDX

0x237e54 MOV	-0x70(%RBP),%R15    [11]

0x237e58 MOV	%R15,%R11

0x237e5b IMUL	%R8,%R11

0x237e5f ADD	%R11,%RDI

0x237e62 VMOVSD	(%R14,%RDI,8),%XMM2    [10]

0x237e68 MOV	%R15,%RDI

0x237e6b IMUL	%RAX,%RDI

0x237e6f ADD	%RDI,%RDX

0x237e72 VMOVHPD	(%R14,%RDX,8),%XMM2,%XMM2    [8]

0x237e78 ADD	%RBX,%R11

0x237e7b VMOVSD	(%R14,%R11,8),%XMM3    [9]

0x237e81 ADD	%R9,%RDI

0x237e84 VMOVHPD	(%R14,%RDI,8),%XMM3,%XMM3    [1]

0x237e8a MOV	-0x68(%RBP),%RDX    [11]

0x237e8e IMUL	%RDX,%R8

0x237e92 ADD	%RBX,%R8

0x237e95 IMUL	%RDX,%RAX

0x237e99 ADD	%R9,%RAX

0x237e9c VUNPCKLPD	%XMM1,%XMM0,%XMM0

0x237ea0 VADDPD	%XMM2,%XMM0,%XMM0

0x237ea4 VSUBPD	%XMM3,%XMM0,%XMM0

0x237ea8 MOV	-0x60(%RBP),%RDX    [11]

0x237eac VMOVLPD	%XMM0,(%RDX,%R8,8)    [3]

0x237eb2 VMOVHPD	%XMM0,(%RDX,%RAX,8)    [6]

0x237eb7 ADD	$0x2,%RSI

0x237ebb ADD	$-0x2,%R12

0x237ebf JE	237f0a

0x237ec1 LEA	0x1(%RSI),%R9

0x237ec5 MOV	%RSI,%RAX

0x237ec8 SHR	$0x20,%RAX

0x237ecc JE	237ee0

0x237ece MOV	%RSI,%RAX

0x237ed1 CQTO

0x237ed3 IDIV	%R13

0x237ed6 MOV	%RDX,%RDI

0x237ed9 MOV	%RAX,%R8

0x237edc JMP	237eec

0x237ee0 MOV	%ESI,%EAX

0x237ee2 XOR	%EDX,%EDX

0x237ee4 DIV	%R13D

0x237ee7 MOV	%EDX,%EDI

0x237ee9 MOV	%EAX,%R8D

0x237eec MOV	-0x2c(%RBP),%R11D    [11]

0x237ef0 MOV	%R9,%RAX

0x237ef3 SHR	$0x20,%RAX

0x237ef7 JNE	237dd0

0x237efd MOV	%R9D,%EAX

0x237f00 XOR	%EDX,%EDX

0x237f02 DIV	%R13D

0x237f05 JMP	237dd8

/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/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);

/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/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]; }

Path /

Metric	Value
CQA speedup if no scalar integer	3.48 - 4.77
CQA speedup if FP arith vectorized	1.02
CQA speedup if fully vectorized	5.33 - 4.88
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.41
Bottlenecks
Function	.omp_outlined..6.313
Source	advec_mom.cpp:72-75,context.h:69-69
Source loop unroll info	unrolled by 2
Source loop unroll confidence level	max
Unroll/vectorization loop type	main
Unroll factor	2
CQA cycles	13.33 - 18.29
CQA cycles if no scalar integer	3.83
CQA cycles if FP arith vectorized	13.00 - 18.00
CQA cycles if fully vectorized	2.50 - 3.75
Front-end cycles	13.00
DIV/SQRT cycles	9.56
P0 cycles	9.56
P1 cycles	9.50
P2 cycles	9.50
P3 cycles	3.88
P4 cycles	6.67
P5 cycles	6.67
P6 cycles	6.67
P7 cycles	0.00
P8 cycles	2.00
P9 cycles	2.00
P10 cycles	2.00
P11 cycles	2.00
P12 cycles	2.00
P13 cycles	13.00 - 18.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	77.00
Nb uops	78.00
Nb loads	16.00
Nb stores	4.00
Nb stack references	10.00
FLOP/cycle	0.30 - 0.22
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	8.92 - 11.71
Bytes prefetched	0.00
Bytes loaded	124.00
Bytes stored	32.00
Stride 0	1.00
Stride 1	0.00
Stride n	0.00
Stride unknown	8.00
Stride indirect	0.00
Vectorization ratio all	10.54
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	100.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	13.48
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	25.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	9.38
Vector-efficiency ratio other	12.50

Metric	Value
CQA speedup if no scalar integer	3.43 - 4.70
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	5.27 - 4.80
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.37
Bottlenecks	P8, P9,
Function	.omp_outlined..6.313
Source	advec_mom.cpp:72-75,context.h:69-69
Source loop unroll info	unrolled by 2
Source loop unroll confidence level	max
Unroll/vectorization loop type	main
Unroll factor	2
CQA cycles	13.17 - 18.00
CQA cycles if no scalar integer	3.83
CQA cycles if FP arith vectorized	13.00 - 18.00
CQA cycles if fully vectorized	2.50 - 3.75
Front-end cycles	13.17
DIV/SQRT cycles	9.75
P0 cycles	9.75
P1 cycles	9.50
P2 cycles	9.50
P3 cycles	4.50
P4 cycles	6.67
P5 cycles	6.67
P6 cycles	6.67
P7 cycles	0.00
P8 cycles	2.00
P9 cycles	2.00
P10 cycles	2.00
P11 cycles	2.00
P12 cycles	2.00
P13 cycles	13.00 - 18.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	78.00
Nb uops	79.00
Nb loads	16.00
Nb stores	4.00
Nb stack references	10.00
FLOP/cycle	0.30 - 0.22
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	8.67 - 11.85
Bytes prefetched	0.00
Bytes loaded	124.00
Bytes stored	32.00
Stride 0	1.00
Stride 1	0.00
Stride n	0.00
Stride unknown	8.00
Stride indirect	0.00
Vectorization ratio all	10.53
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	100.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	13.49
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	25.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	9.38
Vector-efficiency ratio other	12.50

Metric	Value
CQA speedup if no scalar integer	3.65 - 6.26
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	1.09 - 1.87
Bottlenecks	P8, P9,
Function	.omp_outlined..6.313
Source	advec_mom.cpp:72-75,context.h:69-69
Source loop unroll info	unrolled by 2
Source loop unroll confidence level	max
Unroll/vectorization loop type	main
Unroll factor	2
CQA cycles	14.00 - 24.00
CQA cycles if no scalar integer	3.83
CQA cycles if FP arith vectorized	14.00 - 24.00
CQA cycles if fully vectorized	3.50 - 6.00
Front-end cycles	12.83
DIV/SQRT cycles	9.50
P0 cycles	9.50
P1 cycles	9.50
P2 cycles	9.50
P3 cycles	4.00
P4 cycles	6.67
P5 cycles	6.67
P6 cycles	6.67
P7 cycles	0.00
P8 cycles	2.00
P9 cycles	2.00
P10 cycles	2.00
P11 cycles	2.00
P12 cycles	2.00
P13 cycles	14.00 - 24.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	77.00
Nb uops	77.00
Nb loads	16.00
Nb stores	4.00
Nb stack references	10.00
FLOP/cycle	0.29 - 0.17
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	6.50 - 11.14
Bytes prefetched	0.00
Bytes loaded	124.00
Bytes stored	32.00
Stride 0	1.00
Stride 1	0.00
Stride n	0.00
Stride unknown	8.00
Stride indirect	0.00
Vectorization ratio all	10.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	100.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	13.75
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	25.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	12.50
Vector-efficiency ratio other	12.50

Metric	Value
CQA speedup if no scalar integer	3.43
CQA speedup if FP arith vectorized	1.10
CQA speedup if fully vectorized	8.78
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	NA
Bottlenecks	micro-operation queue,
Function	.omp_outlined..6.313
Source	advec_mom.cpp:72-75,context.h:69-69
Source loop unroll info	unrolled by 2
Source loop unroll confidence level	max
Unroll/vectorization loop type	main
Unroll factor	2
CQA cycles	13.17
CQA cycles if no scalar integer	3.83
CQA cycles if FP arith vectorized	12.00
CQA cycles if fully vectorized	1.50
Front-end cycles	13.17
DIV/SQRT cycles	9.50
P0 cycles	9.50
P1 cycles	9.50
P2 cycles	9.50
P3 cycles	4.00
P4 cycles	6.67
P5 cycles	6.67
P6 cycles	6.67
P7 cycles	0.00
P8 cycles	2.00
P9 cycles	2.00
P10 cycles	2.00
P11 cycles	2.00
P12 cycles	2.00
P13 cycles	12.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	77.00
Nb uops	79.00
Nb loads	16.00
Nb stores	4.00
Nb stack references	10.00
FLOP/cycle	0.30
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	11.85
Bytes prefetched	0.00
Bytes loaded	124.00
Bytes stored	32.00
Stride 0	1.00
Stride 1	0.00
Stride n	0.00
Stride unknown	8.00
Stride indirect	0.00
Vectorization ratio all	11.11
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	100.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	13.19
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	25.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	6.25
Vector-efficiency ratio other	12.50

Metric	Value
CQA speedup if no scalar integer	3.39 - 4.70
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	5.20 - 4.80
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.01 - 1.40
Bottlenecks	P8, P9,
Function	.omp_outlined..6.313
Source	advec_mom.cpp:72-75,context.h:69-69
Source loop unroll info	unrolled by 2
Source loop unroll confidence level	max
Unroll/vectorization loop type	main
Unroll factor	2
CQA cycles	13.00 - 18.00
CQA cycles if no scalar integer	3.83
CQA cycles if FP arith vectorized	13.00 - 18.00
CQA cycles if fully vectorized	2.50 - 3.75
Front-end cycles	12.83
DIV/SQRT cycles	9.50
P0 cycles	9.50
P1 cycles	9.50
P2 cycles	9.50
P3 cycles	3.00
P4 cycles	6.67
P5 cycles	6.67
P6 cycles	6.67
P7 cycles	0.00
P8 cycles	2.00
P9 cycles	2.00
P10 cycles	2.00
P11 cycles	2.00
P12 cycles	2.00
P13 cycles	13.00 - 18.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	76.00
Nb uops	77.00
Nb loads	16.00
Nb stores	4.00
Nb stack references	10.00
FLOP/cycle	0.31 - 0.22
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	8.67 - 12.00
Bytes prefetched	0.00
Bytes loaded	124.00
Bytes stored	32.00
Stride 0	1.00
Stride 1	0.00
Stride n	0.00
Stride unknown	8.00
Stride indirect	0.00
Vectorization ratio all	10.53
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	100.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	13.49
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	25.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	9.38
Vector-efficiency ratio other	12.50

Path /

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

Function	.omp_outlined..6.313
Source file and lines	advec_mom.cpp:72-75
Module	exec

The loop is defined in:

/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/advec_mom.cpp:72-75
/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/context.h:69

It is main loop of related source loop which is unrolled by 2 (including vectorization).
This loop has 4 execution paths.

The presence of multiple execution paths is typically the main/first bottleneck.
Try to simplify control inside loop: ideally, try to remove all conditional expressions, for example by (if applicable):

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. Only 13% of vector register length is used (average across all SSE/AVX instructions). By fully vectorizing your loop, you can lower the cost of an iteration from 18.29 to 3.75 cycles (4.88x speedup).

Details

10% of SSE/AVX instructions are used in vector version (process two or more data elements in vector registers):

0% of SSE/AVX loads are used in vector version.
0% of SSE/AVX stores are used in vector version.
0% of SSE/AVX divide and square root instructions are used in vector version.
0% of SSE/AVX 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:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
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)

Expensive FP math instructions/calls

Detected performance impact from expensive FP math instructions/calls. By removing/reexpressing them, you can lower the cost of an iteration from 18.29 to 12.67 cycles (1.44x speedup).

Complex instructions

Detected COMPLEX INSTRUCTIONS.

Details

These instructions generate more than one micro-operation and only one of them can be decoded during a cycle and the extra micro-operations increase pressure on execution units.

DIV: 1.33 occurrences
IDIV: 1.33 occurrences
VMOVHPD: 1 occurrences

Conversion instructions

Detected expensive conversion instructions.

Details

CLTQ: 1 occurrences
CQTO: 1.33 occurrences

Workaround

Avoid mixing data with different types. In particular, check if the type of constants is the same as array elements.

Type of elements and instruction set

2 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 4 FP arithmetical operations:

4: addition or subtraction

The binary loop is loading 124 bytes (15 double precision FP elements). The binary loop is storing 32 bytes (4 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.03 FP operations per loaded or stored byte.

General properties

nb instructions	77
nb uops	78
loop length	287.50
used x86 registers	14
used mmx registers	0
used xmm registers	4
used ymm registers	0
used zmm registers	0
nb stack references	10

Front-end

MACRO FUSION NOT POSSIBLE

micro-operation queue	13.00 cycles
front end	13.00 cycles

Back-end

	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5
uops	9.56	9.56	9.50	9.50	3.88	6.67	6.67	6.67	0.00	2.00	2.00	2.00	2.00	2.00
cycles	9.56	9.56	9.50	9.50	3.88	6.67	6.67	6.67	0.00	2.00	2.00	2.00	2.00	2.00

Execution ports to units layout:

ALU0/BRU0: ALU
ALU1: ALU
ALU2: ALU
ALU3: ALU
BRU1:
AGU0 (256 bits): store address, load
AGU1 (256 bits): store address, load
AGU2 (256 bits): store address, load
FP0 (256 bits): VPU, DIV/SQRT
FP1 (256 bits): VPU, DIV/SQRT
FP2 (256 bits): VPU
FP3 (256 bits): VPU
FP4 (256 bits): FP store data
FP5 (256 bits): FP store data

Cycles executing div or sqrt instructions	13.00-18.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	13.00
Dispatch	9.56
DIV/SQRT	13.00-18.00
Data deps.	1.00
Overall L1	13.33-18.29

Vectorization ratios

INT

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

all	15%
load	0%
store	0%
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	0%

INT+FP

all	10%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	0%
other	0%

Vector efficiency ratios

INT

all	11%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
other	11%

all	14%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

INT+FP

all	13%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	9%
other	12%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 18.29 cycles. At this rate:

7% of peak load performance is reached (7.09 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (1.83 out of 64.00 bytes stored per cycle (GB/s @ 1GHz))

Function	.omp_outlined..6.313
Source file and lines	advec_mom.cpp:72-75
Module	exec

The loop is defined in:

/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/advec_mom.cpp:72-75
/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/context.h:69

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

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 18.00 to 3.83 cycles (4.70x speedup).

Workaround

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

Vectorization

Details

10% of SSE/AVX instructions are used in vector version (process two or more data elements in vector registers):

0% of SSE/AVX loads are used in vector version.
0% of SSE/AVX stores are used in vector version.
0% of SSE/AVX divide and square root instructions are used in vector version.
0% of SSE/AVX 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:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
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)

Execution units bottlenecks

Performance is limited by execution of divide and square root operations (the divide/square root unit is a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 18.00 to 13.17 cycles (1.37x speedup).

Workaround

Reduce the number of division or square root instructions:
- If denominator is constant over iterations, use reciprocal (replace x/y with x*(1/y)). Check precision impact. This will be done by your compiler with ffast-math or Ofast
Check whether you really need double precision. If not, switch to single precision to speedup execution

Expensive FP math instructions/calls

Detected performance impact from expensive FP math instructions/calls. By removing/reexpressing them, you can lower the cost of an iteration from 18.00 to 12.83 cycles (1.40x speedup).

Complex instructions

Detected COMPLEX INSTRUCTIONS.

Details

These instructions generate more than one micro-operation and only one of them can be decoded during a cycle and the extra micro-operations increase pressure on execution units.

DIV: 1 occurrences
IDIV: 1 occurrences
VMOVHPD: 1 occurrences

Slow data structures access

Detected data structures (typically arrays) that cannot be efficiently read/written

Details

Constant unknown stride: 8 occurrence(s)

Non-unit stride (uncontiguous) accesses are not efficiently using data caches

Workaround

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

Conversion instructions

Detected expensive conversion instructions.

Details

CLTQ: 1 occurrences
CQTO: 1 occurrences

Workaround

Avoid mixing data with different types. In particular, check if the type of constants is the same as array elements.

Type of elements and instruction set

2 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 4 FP arithmetical operations:

4: addition or subtraction

The binary loop is loading 124 bytes (15 double precision FP elements). The binary loop is storing 32 bytes (4 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.03 FP operations per loaded or stored byte.

General properties

nb instructions	78
nb uops	79
loop length	292
used x86 registers	14
used mmx registers	0
used xmm registers	4
used ymm registers	0
used zmm registers	0
nb stack references	10

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	13.17 cycles
front end	13.17 cycles

Back-end

	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5
uops	9.75	9.75	9.50	9.50	4.50	6.67	6.67	6.67	0.00	2.00	2.00	2.00	2.00	2.00
cycles	9.75	9.75	9.50	9.50	4.50	6.67	6.67	6.67	0.00	2.00	2.00	2.00	2.00	2.00

Execution ports to units layout:

ALU0/BRU0: ALU
ALU1: ALU
ALU2: ALU
ALU3: ALU
BRU1:
AGU0 (256 bits): store address, load
AGU1 (256 bits): store address, load
AGU2 (256 bits): store address, load
FP0 (256 bits): VPU, DIV/SQRT
FP1 (256 bits): VPU, DIV/SQRT
FP2 (256 bits): VPU
FP3 (256 bits): VPU
FP4 (256 bits): FP store data
FP5 (256 bits): FP store data

Cycles executing div or sqrt instructions	13.00-18.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	13.17
Dispatch	9.75
DIV/SQRT	13.00-18.00
Data deps.	1.00
Overall L1	13.17-18.00

Vectorization ratios

INT

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

all	15%
load	0%
store	0%
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	0%

INT+FP

all	10%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	0%
other	0%

Vector efficiency ratios

INT

all	11%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
other	11%

all	14%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

INT+FP

all	13%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	9%
other	12%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 18.00 cycles. At this rate:

7% of peak load performance is reached (6.89 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (1.78 out of 64.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 237ec1

Instruction	Nb FU	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5	Latency	Recip. throughput
ADD %R11D,%R8D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %R11D,%EAX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x58(%RBP),%R11	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
LEA (%R11,%RDI,1),%R9D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %EDX,%R11D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R9D,%RBX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R11D,%R9	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R8D,%R8	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
CLTQ
MOV -0x50(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R15,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x80(%RBP),%R10	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVSD (%R10,%R11,8),%XMM0	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
IMUL %RAX,%R15	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%R15	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R10,%R15,8),%XMM1	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
MOV -0xa8(%RBP),%R10	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R10,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x78(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVSD %XMM0,(%R15,%R11,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
MOV %R10,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %RAX,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD %XMM1,(%R15,%R11,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
MOV -0x38(%RBP),%R11	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
ADD %R11D,%EDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %R11D,%EDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %EDI,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %EDX,%RDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x70(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R15,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R11,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R14,%RDI,8),%XMM2	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
MOV %R15,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %RAX,%RDI	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RDI,%RDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVHPD (%R14,%RDX,8),%XMM2,%XMM2	1	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0	0	9	0.50
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R14,%R11,8),%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
ADD %R9,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVHPD (%R14,%RDI,8),%XMM3,%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0	0	9	0.50
MOV -0x68(%RBP),%RDX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
IMUL %RDX,%R8	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R8	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
IMUL %RDX,%RAX	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%RAX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VUNPCKLPD %XMM1,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	1	0.33
VADDPD %XMM2,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	3	0.50
VSUBPD %XMM3,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	3	0.50
MOV -0x60(%RBP),%RDX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVLPD %XMM0,(%RDX,%R8,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
VMOVHPD %XMM0,(%RDX,%RAX,8)	2	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0.50	0.50	9-11	1
ADD $0x2,%RSI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD $-0x2,%R12	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
JE 237f0a <.omp_outlined..6.313+0x29a>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
LEA 0x1(%RSI),%R9	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV %RSI,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
SHR $0x20,%RAX	1	0	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
JE 237ee0 <.omp_outlined..6.313+0x270>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
MOV %RSI,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
CQTO
IDIV %R13	2	2	0	0	0	0	0	0	0	0	0	0	0	0	0	9-19	7-12
MOV %RDX,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
MOV %RAX,%R8	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
JMP 237eec <.omp_outlined..6.313+0x27c>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	0	1
MOV -0x2c(%RBP),%R11D	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R9,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
SHR $0x20,%RAX	1	0	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
JNE 237dd0 <.omp_outlined..6.313+0x160>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
MOV %R9D,%EAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
XOR %EDX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.25
DIV %R13D	2	2	0	0	0	0	0	0	0	0	0	0	0	0	0	10-17	6
JMP 237dd8 <.omp_outlined..6.313+0x168>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	0	1

Function	.omp_outlined..6.313
Source file and lines	advec_mom.cpp:72-75
Module	exec

The loop is defined in:

/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/advec_mom.cpp:72-75
/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/context.h:69

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

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 24.00 to 3.83 cycles (6.26x speedup).

Workaround

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

Vectorization

Details

10% of SSE/AVX instructions are used in vector version (process two or more data elements in vector registers):

0% of SSE/AVX loads are used in vector version.
0% of SSE/AVX stores are used in vector version.
0% of SSE/AVX divide and square root instructions are used in vector version.
0% of SSE/AVX 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:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
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)

Execution units bottlenecks

Workaround

Reduce the number of division or square root instructions:
- If denominator is constant over iterations, use reciprocal (replace x/y with x*(1/y)). Check precision impact. This will be done by your compiler with ffast-math or Ofast
Check whether you really need double precision. If not, switch to single precision to speedup execution

Expensive FP math instructions/calls

Detected performance impact from expensive FP math instructions/calls. By removing/reexpressing them, you can lower the cost of an iteration from 24.00 to 12.50 cycles (1.92x speedup).

Complex instructions

Detected COMPLEX INSTRUCTIONS.

Details

These instructions generate more than one micro-operation and only one of them can be decoded during a cycle and the extra micro-operations increase pressure on execution units.

IDIV: 2 occurrences
VMOVHPD: 1 occurrences

Slow data structures access

Detected data structures (typically arrays) that cannot be efficiently read/written

Details

Constant unknown stride: 8 occurrence(s)

Non-unit stride (uncontiguous) accesses are not efficiently using data caches

Workaround

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

Conversion instructions

Detected expensive conversion instructions.

Details

CLTQ: 1 occurrences
CQTO: 2 occurrences

Workaround

Avoid mixing data with different types. In particular, check if the type of constants is the same as array elements.

Type of elements and instruction set

2 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 4 FP arithmetical operations:

4: addition or subtraction

The binary loop is loading 124 bytes (15 double precision FP elements). The binary loop is storing 32 bytes (4 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.03 FP operations per loaded or stored byte.

General properties

nb instructions	77
nb uops	77
loop length	287
used x86 registers	14
used mmx registers	0
used xmm registers	4
used ymm registers	0
used zmm registers	0
nb stack references	10

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	12.83 cycles
front end	12.83 cycles

Back-end

	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5
uops	9.50	9.50	9.50	9.50	4.00	6.67	6.67	6.67	0.00	2.00	2.00	2.00	2.00	2.00
cycles	9.50	9.50	9.50	9.50	4.00	6.67	6.67	6.67	0.00	2.00	2.00	2.00	2.00	2.00

Execution ports to units layout:

ALU0/BRU0: ALU
ALU1: ALU
ALU2: ALU
ALU3: ALU
BRU1:
AGU0 (256 bits): store address, load
AGU1 (256 bits): store address, load
AGU2 (256 bits): store address, load
FP0 (256 bits): VPU, DIV/SQRT
FP1 (256 bits): VPU, DIV/SQRT
FP2 (256 bits): VPU
FP3 (256 bits): VPU
FP4 (256 bits): FP store data
FP5 (256 bits): FP store data

Cycles executing div or sqrt instructions	14.00-24.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	12.83
Dispatch	9.50
DIV/SQRT	14.00-24.00
Data deps.	1.00
Overall L1	14.00-24.00

Vectorization ratios

INT

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

all	15%
load	0%
store	0%
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	0%

INT+FP

all	10%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	0%
other	0%

Vector efficiency ratios

INT

all	12%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
other	12%

all	14%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

INT+FP

all	13%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	12%
other	12%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 24.00 cycles. At this rate:

5% of peak load performance is reached (5.17 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (1.33 out of 64.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 237ec1

Instruction	Nb FU	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5	Latency	Recip. throughput
MOV %R9,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
CQTO
IDIV %R13	2	2	0	0	0	0	0	0	0	0	0	0	0	0	0	9-19	7-12
ADD %R11D,%R8D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %R11D,%EAX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x58(%RBP),%R11	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
LEA (%R11,%RDI,1),%R9D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %EDX,%R11D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R9D,%RBX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R11D,%R9	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R8D,%R8	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
CLTQ
MOV -0x50(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R15,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x80(%RBP),%R10	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVSD (%R10,%R11,8),%XMM0	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
IMUL %RAX,%R15	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%R15	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R10,%R15,8),%XMM1	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
MOV -0xa8(%RBP),%R10	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R10,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x78(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVSD %XMM0,(%R15,%R11,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
MOV %R10,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %RAX,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD %XMM1,(%R15,%R11,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
MOV -0x38(%RBP),%R11	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
ADD %R11D,%EDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %R11D,%EDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %EDI,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %EDX,%RDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x70(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R15,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R11,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R14,%RDI,8),%XMM2	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
MOV %R15,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %RAX,%RDI	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RDI,%RDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVHPD (%R14,%RDX,8),%XMM2,%XMM2	1	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0	0	9	0.50
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R14,%R11,8),%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
ADD %R9,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVHPD (%R14,%RDI,8),%XMM3,%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0	0	9	0.50
MOV -0x68(%RBP),%RDX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
IMUL %RDX,%R8	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R8	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
IMUL %RDX,%RAX	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%RAX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VUNPCKLPD %XMM1,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	1	0.33
VADDPD %XMM2,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	3	0.50
VSUBPD %XMM3,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	3	0.50
MOV -0x60(%RBP),%RDX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVLPD %XMM0,(%RDX,%R8,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
VMOVHPD %XMM0,(%RDX,%RAX,8)	2	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0.50	0.50	9-11	1
ADD $0x2,%RSI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD $-0x2,%R12	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
JE 237f0a <.omp_outlined..6.313+0x29a>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
LEA 0x1(%RSI),%R9	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV %RSI,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
SHR $0x20,%RAX	1	0	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
JE 237ee0 <.omp_outlined..6.313+0x270>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
MOV %RSI,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
CQTO
IDIV %R13	2	2	0	0	0	0	0	0	0	0	0	0	0	0	0	9-19	7-12
MOV %RDX,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
MOV %RAX,%R8	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
JMP 237eec <.omp_outlined..6.313+0x27c>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	0	1
MOV -0x2c(%RBP),%R11D	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R9,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
SHR $0x20,%RAX	1	0	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
JNE 237dd0 <.omp_outlined..6.313+0x160>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1

Function	.omp_outlined..6.313
Source file and lines	advec_mom.cpp:72-75
Module	exec

The loop is defined in:

/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/advec_mom.cpp:72-75
/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/context.h:69

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

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 13.17 to 3.83 cycles (3.43x speedup).

Workaround

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

Vectorization

Details

11% of SSE/AVX instructions are used in vector version (process two or more data elements in vector registers):

0% of SSE/AVX loads are used in vector version.
0% of SSE/AVX stores are used in vector version.
0% of SSE/AVX divide and square root instructions are used in vector version.
0% of SSE/AVX 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:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
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)

Execution units bottlenecks

Found no such bottlenecks but see expert reports for more complex bottlenecks.

No data for this section

Complex instructions

Detected COMPLEX INSTRUCTIONS.

Details

These instructions generate more than one micro-operation and only one of them can be decoded during a cycle and the extra micro-operations increase pressure on execution units.

DIV: 2 occurrences
VMOVHPD: 1 occurrences

Slow data structures access

Detected data structures (typically arrays) that cannot be efficiently read/written

Details

Constant unknown stride: 8 occurrence(s)

Non-unit stride (uncontiguous) accesses are not efficiently using data caches

Workaround

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

Conversion instructions

Detected expensive conversion instructions.

Details

CLTQ: 1 occurrences

Workaround

Avoid mixing data with different types. In particular, check if the type of constants is the same as array elements.

Type of elements and instruction set

2 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 4 FP arithmetical operations:

4: addition or subtraction

The binary loop is loading 124 bytes (15 double precision FP elements). The binary loop is storing 32 bytes (4 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.03 FP operations per loaded or stored byte.

General properties

nb instructions	77
nb uops	79
loop length	288
used x86 registers	14
used mmx registers	0
used xmm registers	4
used ymm registers	0
used zmm registers	0
nb stack references	10

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	13.17 cycles
front end	13.17 cycles

Back-end

	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5
uops	9.50	9.50	9.50	9.50	4.00	6.67	6.67	6.67	0.00	2.00	2.00	2.00	2.00	2.00
cycles	9.50	9.50	9.50	9.50	4.00	6.67	6.67	6.67	0.00	2.00	2.00	2.00	2.00	2.00

Execution ports to units layout:

ALU0/BRU0: ALU
ALU1: ALU
ALU2: ALU
ALU3: ALU
BRU1:
AGU0 (256 bits): store address, load
AGU1 (256 bits): store address, load
AGU2 (256 bits): store address, load
FP0 (256 bits): VPU, DIV/SQRT
FP1 (256 bits): VPU, DIV/SQRT
FP2 (256 bits): VPU
FP3 (256 bits): VPU
FP4 (256 bits): FP store data
FP5 (256 bits): FP store data

Cycles executing div or sqrt instructions	12.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	13.17
Dispatch	9.50
DIV/SQRT	12.00
Data deps.	1.00
Overall L1	13.17

Vectorization ratios

INT

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

all	15%
load	0%
store	0%
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	0%

INT+FP

all	11%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	0%
other	0%

Vector efficiency ratios

INT

all	10%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
other	10%

all	14%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

INT+FP

all	13%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	6%
other	12%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 13.17 cycles. At this rate:

9% of peak load performance is reached (9.42 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
3% of peak store performance is reached (2.43 out of 64.00 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).

ASM code

In the binary file, the address of the loop is: 237ec1

Instruction	Nb FU	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5	Latency	Recip. throughput
ADD %R11D,%R8D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %R11D,%EAX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x58(%RBP),%R11	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
LEA (%R11,%RDI,1),%R9D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %EDX,%R11D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R9D,%RBX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R11D,%R9	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R8D,%R8	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
CLTQ
MOV -0x50(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R15,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x80(%RBP),%R10	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVSD (%R10,%R11,8),%XMM0	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
IMUL %RAX,%R15	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%R15	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R10,%R15,8),%XMM1	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
MOV -0xa8(%RBP),%R10	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R10,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x78(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVSD %XMM0,(%R15,%R11,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
MOV %R10,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %RAX,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD %XMM1,(%R15,%R11,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
MOV -0x38(%RBP),%R11	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
ADD %R11D,%EDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %R11D,%EDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %EDI,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %EDX,%RDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x70(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R15,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R11,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R14,%RDI,8),%XMM2	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
MOV %R15,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %RAX,%RDI	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RDI,%RDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVHPD (%R14,%RDX,8),%XMM2,%XMM2	1	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0	0	9	0.50
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R14,%R11,8),%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
ADD %R9,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVHPD (%R14,%RDI,8),%XMM3,%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0	0	9	0.50
MOV -0x68(%RBP),%RDX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
IMUL %RDX,%R8	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R8	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
IMUL %RDX,%RAX	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%RAX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VUNPCKLPD %XMM1,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	1	0.33
VADDPD %XMM2,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	3	0.50
VSUBPD %XMM3,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	3	0.50
MOV -0x60(%RBP),%RDX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVLPD %XMM0,(%RDX,%R8,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
VMOVHPD %XMM0,(%RDX,%RAX,8)	2	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0.50	0.50	9-11	1
ADD $0x2,%RSI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD $-0x2,%R12	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
JE 237f0a <.omp_outlined..6.313+0x29a>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
LEA 0x1(%RSI),%R9	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV %RSI,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
SHR $0x20,%RAX	1	0	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
JE 237ee0 <.omp_outlined..6.313+0x270>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
MOV %ESI,%EAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
XOR %EDX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.25
DIV %R13D	2	2	0	0	0	0	0	0	0	0	0	0	0	0	0	10-17	6
MOV %EDX,%EDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
MOV %EAX,%R8D	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
MOV -0x2c(%RBP),%R11D	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R9,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
SHR $0x20,%RAX	1	0	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
JNE 237dd0 <.omp_outlined..6.313+0x160>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
MOV %R9D,%EAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
XOR %EDX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.25
DIV %R13D	2	2	0	0	0	0	0	0	0	0	0	0	0	0	0	10-17	6
JMP 237dd8 <.omp_outlined..6.313+0x168>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	0	1

Function	.omp_outlined..6.313
Source file and lines	advec_mom.cpp:72-75
Module	exec

The loop is defined in:

/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/advec_mom.cpp:72-75
/beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/intel/CloverLeafCXX/build/CloverLeafCXX/src/omp/context.h:69

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

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 18.00 to 3.83 cycles (4.70x speedup).

Workaround

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

Vectorization

Details

10% of SSE/AVX instructions are used in vector version (process two or more data elements in vector registers):

0% of SSE/AVX loads are used in vector version.
0% of SSE/AVX stores are used in vector version.
0% of SSE/AVX divide and square root instructions are used in vector version.
0% of SSE/AVX 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:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
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)

Execution units bottlenecks

Workaround

Reduce the number of division or square root instructions:
- If denominator is constant over iterations, use reciprocal (replace x/y with x*(1/y)). Check precision impact. This will be done by your compiler with ffast-math or Ofast
Check whether you really need double precision. If not, switch to single precision to speedup execution

Expensive FP math instructions/calls

Detected performance impact from expensive FP math instructions/calls. By removing/reexpressing them, you can lower the cost of an iteration from 18.00 to 12.50 cycles (1.44x speedup).

Complex instructions

Detected COMPLEX INSTRUCTIONS.

Details

These instructions generate more than one micro-operation and only one of them can be decoded during a cycle and the extra micro-operations increase pressure on execution units.

DIV: 1 occurrences
IDIV: 1 occurrences
VMOVHPD: 1 occurrences

Slow data structures access

Detected data structures (typically arrays) that cannot be efficiently read/written

Details

Constant unknown stride: 8 occurrence(s)

Non-unit stride (uncontiguous) accesses are not efficiently using data caches

Workaround

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

Conversion instructions

Detected expensive conversion instructions.

Details

CLTQ: 1 occurrences
CQTO: 1 occurrences

Workaround

Avoid mixing data with different types. In particular, check if the type of constants is the same as array elements.

Type of elements and instruction set

2 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 4 FP arithmetical operations:

4: addition or subtraction

The binary loop is loading 124 bytes (15 double precision FP elements). The binary loop is storing 32 bytes (4 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.03 FP operations per loaded or stored byte.

General properties

nb instructions	76
nb uops	77
loop length	283
used x86 registers	14
used mmx registers	0
used xmm registers	4
used ymm registers	0
used zmm registers	0
nb stack references	10

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	12.83 cycles
front end	12.83 cycles

Back-end

	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5
uops	9.50	9.50	9.50	9.50	3.00	6.67	6.67	6.67	0.00	2.00	2.00	2.00	2.00	2.00
cycles	9.50	9.50	9.50	9.50	3.00	6.67	6.67	6.67	0.00	2.00	2.00	2.00	2.00	2.00

Execution ports to units layout:

ALU0/BRU0: ALU
ALU1: ALU
ALU2: ALU
ALU3: ALU
BRU1:
AGU0 (256 bits): store address, load
AGU1 (256 bits): store address, load
AGU2 (256 bits): store address, load
FP0 (256 bits): VPU, DIV/SQRT
FP1 (256 bits): VPU, DIV/SQRT
FP2 (256 bits): VPU
FP3 (256 bits): VPU
FP4 (256 bits): FP store data
FP5 (256 bits): FP store data

Cycles executing div or sqrt instructions	13.00-18.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	12.83
Dispatch	9.50
DIV/SQRT	13.00-18.00
Data deps.	1.00
Overall L1	13.00-18.00

Vectorization ratios

INT

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

all	15%
load	0%
store	0%
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	0%

INT+FP

all	10%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	0%
other	0%

Vector efficiency ratios

INT

all	11%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
other	11%

all	14%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

INT+FP

all	13%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	25%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	9%
other	12%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 18.00 cycles. At this rate:

7% of peak load performance is reached (6.89 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (1.78 out of 64.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 237ec1

Instruction	Nb FU	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5	Latency	Recip. throughput
MOV %R9,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
CQTO
IDIV %R13	2	2	0	0	0	0	0	0	0	0	0	0	0	0	0	9-19	7-12
ADD %R11D,%R8D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %R11D,%EAX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x58(%RBP),%R11	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
LEA (%R11,%RDI,1),%R9D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %EDX,%R11D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R9D,%RBX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R11D,%R9	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %R8D,%R8	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
CLTQ
MOV -0x50(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R15,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x80(%RBP),%R10	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVSD (%R10,%R11,8),%XMM0	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
IMUL %RAX,%R15	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%R15	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R10,%R15,8),%XMM1	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
MOV -0xa8(%RBP),%R10	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R10,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x78(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVSD %XMM0,(%R15,%R11,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
MOV %R10,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %RAX,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD %XMM1,(%R15,%R11,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
MOV -0x38(%RBP),%R11	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
ADD %R11D,%EDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD %R11D,%EDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %EDI,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOVSXD %EDX,%RDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV -0x70(%RBP),%R15	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R15,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %R8,%R11	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R11,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R14,%RDI,8),%XMM2	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
MOV %R15,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
IMUL %RAX,%RDI	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RDI,%RDX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVHPD (%R14,%RDX,8),%XMM2,%XMM2	1	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0	0	9	0.50
ADD %RBX,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVSD (%R14,%R11,8),%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50
ADD %R9,%RDI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VMOVHPD (%R14,%RDI,8),%XMM3,%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0	0	9	0.50
MOV -0x68(%RBP),%RDX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
IMUL %RDX,%R8	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %RBX,%R8	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
IMUL %RDX,%RAX	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	3	1
ADD %R9,%RAX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
VUNPCKLPD %XMM1,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	1	0.33
VADDPD %XMM2,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	3	0.50
VSUBPD %XMM3,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	3	0.50
MOV -0x60(%RBP),%RDX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
VMOVLPD %XMM0,(%RDX,%R8,8)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0.50	0.50	1	1
VMOVHPD %XMM0,(%RDX,%RAX,8)	2	0	0	0	0	0	0.33	0.33	0.33	0	0.33	0.33	0.33	0.50	0.50	9-11	1
ADD $0x2,%RSI	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
ADD $-0x2,%R12	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
JE 237f0a <.omp_outlined..6.313+0x29a>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
LEA 0x1(%RSI),%R9	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
MOV %RSI,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
SHR $0x20,%RAX	1	0	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
JE 237ee0 <.omp_outlined..6.313+0x270>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
MOV %ESI,%EAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
XOR %EDX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.25
DIV %R13D	2	2	0	0	0	0	0	0	0	0	0	0	0	0	0	10-17	6
MOV %EDX,%EDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
MOV %EAX,%R8D	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
MOV -0x2c(%RBP),%R11D	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV %R9,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
SHR $0x20,%RAX	1	0	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
JNE 237dd0 <.omp_outlined..6.313+0x160>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1

▶Scalability Runs Description

Run 2x1	Number processes: 2Number nodes: 1Number processes per node: 2Run Command: <executable> MPI Command: mpirun -np <number_processes>Dataset: Run Directory: /beegfs/hackathon/users/eoseret/qaas_runs/170-854-8685/intel/CloverLeafCXX/run/oneview_runs/compilers/aocc_15/oneview_run_1708555871I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spreadOMP_NUM_THREADS: 1
Run 2x2	OMP_NUM_THREADS: 2I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x4	OMP_NUM_THREADS: 4I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x8	OMP_NUM_THREADS: 8I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x16	OMP_NUM_THREADS: 16I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x32	OMP_NUM_THREADS: 32I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x64	OMP_NUM_THREADS: 64I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x96	OMP_NUM_THREADS: 96I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread

(2x1) Efficiency	(2x1) Potential Speed-Up (%)	(2x2) Efficiency	(2x2) Potential Speed-Up (%)	(2x4) Efficiency	(2x4) Potential Speed-Up (%)	(2x8) Efficiency	(2x8) Potential Speed-Up (%)	(2x16) Efficiency	(2x16) Potential Speed-Up (%)	(2x32) Efficiency	(2x32) Potential Speed-Up (%)	(2x64) Efficiency	(2x64) Potential Speed-Up (%)	(2x96) Efficiency	(2x96) Potential Speed-Up (%)
1	0	0.99	0.01	1	0	1	0	0.92	0.12	0.61	0.78	0.31	1.56	0.21	2.02

Speed-Up and Efficiency

▶Detailed Speed-Up and Efficiency

Run	Number of threads	Efficiency (ideal is 1)	Speedup	Ideal Speedup	Time (s)	Coverage (%)
2x1	2	1	1	1	26.46	1.54
2x2	4	0.99	1.99	2	13.38	1.55
2x4	8	1	3.99	4	6.65	1.54
2x8	16	1	7.99	8	3.34	1.53
2x16	32	0.92	14.78	16	2.05	1.57
2x32	64	0.61	19.59	32	1.56	2.02
2x64	128	0.31	19.74	64	1.53	2.25
2x96	192	0.21	19.74	96	1.37	2.54

Report Configuration

Vectorization

Details

Workaround

Expensive FP math instructions/calls

Complex instructions

Details

Conversion 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

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

Workaround

Expensive FP math instructions/calls

Complex instructions

Details

Slow data structures access

Details

Workaround

Conversion 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

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

Workaround

Expensive FP math instructions/calls

Complex instructions

Details

Slow data structures access

Details

Workaround

Conversion 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

Code clean check

Workaround

Vectorization

Details