OV - exec - Loop 115

0x2121b0 MOVSXD	(%R9,%R15,4),%R12    [2]

0x2121b4 VMOVSD	-0x8(%R11),%XMM4    [3]

0x2121ba VMOVSD	-0x10(%R11),%XMM3    [3]

0x2121c0 INC	%R15

0x2121c3 SAL	$0x4,%R12

0x2121c7 VDIVSD	0x8(%R10,%R12,1),%XMM1,%XMM2    [1]

0x2121ce VMULSD	%XMM4,%XMM4,%XMM4

0x2121d2 VFMADD231SD	%XMM3,%XMM3,%XMM4

0x2121d7 VMOVSD	(%R11),%XMM3    [3]

0x2121dc ADD	$0x18,%R11

0x2121e0 VFMADD213SD	%XMM4,%XMM3,%XMM3

0x2121e5 VFMADD231SD	%XMM3,%XMM2,%XMM0

0x2121ea VMOVSD	%XMM0,-0x30(%RBP)    [4]

0x2121ef CMP	%R15D,%R8D

0x2121f2 JNE	2121b0

/beegfs/hackathon/users/eoseret/qaas_runs/170-850-7424/intel/CoMD/build/CoMD/CoMD/src-openmp/timestep.c: 110 - 116

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

110:       for (int iOff=MAXATOMS*iBox,ii=0; ii<s->boxes->nAtoms[iBox]; ii++,iOff++)

111:       {

112:          int iSpecies = s->atoms->iSpecies[iOff];

113:          real_t invMass = 0.5/s->species[iSpecies].mass;

114:          kenergy += ( s->atoms->p[iOff][0] * s->atoms->p[iOff][0] +

115:          s->atoms->p[iOff][1] * s->atoms->p[iOff][1] +

116:          s->atoms->p[iOff][2] * s->atoms->p[iOff][2] )*invMass;

Path /

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	2.86
CQA speedup if fully vectorized	4.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	2.00
Bottlenecks	P8, P9,
Function	.omp_outlined..4
Source	timestep.c:110-116
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	5.00
CQA cycles if no scalar integer	5.00
CQA cycles if FP arith vectorized	1.75
CQA cycles if fully vectorized	1.25
Front-end cycles	2.33
DIV/SQRT cycles	1.25
P0 cycles	1.25
P1 cycles	1.00
P2 cycles	1.00
P3 cycles	0.50
P4 cycles	2.00
P5 cycles	2.00
P6 cycles	2.00
P7 cycles	2.50
P8 cycles	2.50
P9 cycles	0.00
P10 cycles	0.00
P11 cycles	0.50
P12 cycles	0.50
P13 cycles	5.00
Inter-iter dependencies cycles	4
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	15.00
Nb uops	14.00
Nb loads	5.00
Nb stores	1.00
Nb stack references	1.00
FLOP/cycle	1.60
Nb FLOP add-sub	0.00
Nb FLOP mul	1.00
Nb FLOP fma	3.00
Nb FLOP div	1.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	8.80
Bytes prefetched	0.00
Bytes loaded	36.00
Bytes stored	8.00
Stride 0	1.00
Stride 1	2.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	1.00
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	0.00
Vectorization ratio add_sub	NA
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	NA
Vector-efficiency ratio all	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	12.50
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	12.50
Vector-efficiency ratio div_sqrt	12.50
Vector-efficiency ratio other	NA

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	2.86
CQA speedup if fully vectorized	4.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	2.00
Bottlenecks	P8, P9,
Function	.omp_outlined..4
Source	timestep.c:110-116
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	5.00
CQA cycles if no scalar integer	5.00
CQA cycles if FP arith vectorized	1.75
CQA cycles if fully vectorized	1.25
Front-end cycles	2.33
DIV/SQRT cycles	1.25
P0 cycles	1.25
P1 cycles	1.00
P2 cycles	1.00
P3 cycles	0.50
P4 cycles	2.00
P5 cycles	2.00
P6 cycles	2.00
P7 cycles	2.50
P8 cycles	2.50
P9 cycles	0.00
P10 cycles	0.00
P11 cycles	0.50
P12 cycles	0.50
P13 cycles	5.00
Inter-iter dependencies cycles	4
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	15.00
Nb uops	14.00
Nb loads	5.00
Nb stores	1.00
Nb stack references	1.00
FLOP/cycle	1.60
Nb FLOP add-sub	0.00
Nb FLOP mul	1.00
Nb FLOP fma	3.00
Nb FLOP div	1.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	8.80
Bytes prefetched	0.00
Bytes loaded	36.00
Bytes stored	8.00
Stride 0	1.00
Stride 1	2.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	1.00
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	0.00
Vectorization ratio add_sub	NA
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	NA
Vector-efficiency ratio all	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	12.50
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	12.50
Vector-efficiency ratio div_sqrt	12.50
Vector-efficiency ratio other	NA

Path /

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

Function	.omp_outlined..4
Source file and lines	timestep.c:110-116
Module	exec

The loop is defined in /beegfs/hackathon/users/eoseret/qaas_runs/170-850-7424/intel/CoMD/build/CoMD/CoMD/src-openmp/timestep.c:110-116.

The related source loop is not unrolled or unrolled with no peel/tail loop.

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. 8 data elements could be processed at once in vector registers.

By vectorizing your loop, you can lower the cost of an iteration from 5.00 to 1.25 cycles (4.00x speedup).

Details

All SSE/AVX 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:
- 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 5.00 to 2.50 cycles (2.00x 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 5.00 to 4.00 cycles (1.25x speedup).

FMA

Detected 3 FMA (fused multiply-add) operations.

Slow data structures access

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

Details

Irregular (variable stride) or indirect: 1 occurrence(s)

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

Workaround

Try to remove indirect accesses. If applicable, precompute elements out of the innermost loop.

Type of elements and instruction set

5 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in scalar mode (one at a time).

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

The binary loop is composed of 8 FP arithmetical operations:

3: addition or subtraction (all inside FMA instructions)
4: multiply (3 inside FMA instructions)
1: divide

The binary loop is loading 36 bytes (4 double precision FP elements). The binary loop is storing 8 bytes (1 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.18 FP operations per loaded or stored byte.

General properties

nb instructions	15
nb uops	14
loop length	68
used x86 registers	7
used mmx registers	0
used xmm registers	5
used ymm registers	0
used zmm registers	0
nb stack references	1

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	2.33 cycles
front end	2.33 cycles

Back-end

	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5
uops	1.25	1.25	1.00	1.00	0.50	2.00	2.00	2.00	2.50	2.50	0.00	0.00	0.50	0.50
cycles	1.25	1.25	1.00	1.00	0.50	2.00	2.00	2.00	2.50	2.50	0.00	0.00	0.50	0.50

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	5.00
Longest recurrence chain latency (RecMII)	4.00

Cycles summary

Front-end	2.33
Dispatch	2.50
DIV/SQRT	5.00
Data deps.	4.00
Overall L1	5.00

Vectorization ratios

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

Vector efficiency ratios

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 5.00 cycles. At this rate:

7% of peak load performance is reached (7.20 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (1.60 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: 2121b0

Instruction	Nb FU	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP4	FP5	Latency	Recip. throughput
MOVSXD (%R9,%R15,4),%R12	1	0.25	0.25	0.25	0.25	0	0.33	0.33	0.33	0	0	0	0	1	0.33
VMOVSD -0x8(%R11),%XMM4	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50
VMOVSD -0x10(%R11),%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50
INC %R15	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	1	0.25
SAL $0x4,%R12	1	0	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50
VDIVSD 0x8(%R10,%R12,1),%XMM1,%XMM2	1	0	0	0	0	0	0.33	0.33	0.33	0.50	0.50	0	0	13	5
VMULSD %XMM4,%XMM4,%XMM4	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	3	0.50
VFMADD231SD %XMM3,%XMM3,%XMM4	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	4	0.50
VMOVSD (%R11),%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50
ADD $0x18,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	1	0.25
VFMADD213SD %XMM4,%XMM3,%XMM3	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	4	0.50
VFMADD231SD %XMM3,%XMM2,%XMM0	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	4	0.50
VMOVSD %XMM0,-0x30(%RBP)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0.50	0.50	1	1
CMP %R15D,%R8D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	1	0.25
JNE 2121b0 <.omp_outlined..4+0x1a0>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	1	0.50-1

Function	.omp_outlined..4
Source file and lines	timestep.c:110-116
Module	exec

The loop is defined in /beegfs/hackathon/users/eoseret/qaas_runs/170-850-7424/intel/CoMD/build/CoMD/CoMD/src-openmp/timestep.c:110-116.

The related source loop is not unrolled or unrolled with no peel/tail loop.

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. 8 data elements could be processed at once in vector registers.

By vectorizing your loop, you can lower the cost of an iteration from 5.00 to 1.25 cycles (4.00x speedup).

Details

All SSE/AVX 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:
- 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 5.00 to 2.50 cycles (2.00x 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 5.00 to 4.00 cycles (1.25x speedup).

FMA

Detected 3 FMA (fused multiply-add) operations.

Slow data structures access

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

Details

Irregular (variable stride) or indirect: 1 occurrence(s)

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

Workaround

Try to remove indirect accesses. If applicable, precompute elements out of the innermost loop.

Type of elements and instruction set

5 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in scalar mode (one at a time).

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

The binary loop is composed of 8 FP arithmetical operations:

3: addition or subtraction (all inside FMA instructions)
4: multiply (3 inside FMA instructions)
1: divide

The binary loop is loading 36 bytes (4 double precision FP elements). The binary loop is storing 8 bytes (1 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.18 FP operations per loaded or stored byte.

General properties

nb instructions	15
nb uops	14
loop length	68
used x86 registers	7
used mmx registers	0
used xmm registers	5
used ymm registers	0
used zmm registers	0
nb stack references	1

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	2.33 cycles
front end	2.33 cycles

Back-end

	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5
uops	1.25	1.25	1.00	1.00	0.50	2.00	2.00	2.00	2.50	2.50	0.00	0.00	0.50	0.50
cycles	1.25	1.25	1.00	1.00	0.50	2.00	2.00	2.00	2.50	2.50	0.00	0.00	0.50	0.50

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	5.00
Longest recurrence chain latency (RecMII)	4.00

Cycles summary

Front-end	2.33
Dispatch	2.50
DIV/SQRT	5.00
Data deps.	4.00
Overall L1	5.00

Vectorization ratios

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

Vector efficiency ratios

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 5.00 cycles. At this rate:

7% of peak load performance is reached (7.20 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (1.60 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: 2121b0

Instruction	Nb FU	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP4	FP5	Latency	Recip. throughput
MOVSXD (%R9,%R15,4),%R12	1	0.25	0.25	0.25	0.25	0	0.33	0.33	0.33	0	0	0	0	1	0.33
VMOVSD -0x8(%R11),%XMM4	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50
VMOVSD -0x10(%R11),%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50
INC %R15	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	1	0.25
SAL $0x4,%R12	1	0	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50
VDIVSD 0x8(%R10,%R12,1),%XMM1,%XMM2	1	0	0	0	0	0	0.33	0.33	0.33	0.50	0.50	0	0	13	5
VMULSD %XMM4,%XMM4,%XMM4	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	3	0.50
VFMADD231SD %XMM3,%XMM3,%XMM4	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	4	0.50
VMOVSD (%R11),%XMM3	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50
ADD $0x18,%R11	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	1	0.25
VFMADD213SD %XMM4,%XMM3,%XMM3	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	4	0.50
VFMADD231SD %XMM3,%XMM2,%XMM0	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	4	0.50
VMOVSD %XMM0,-0x30(%RBP)	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0.50	0.50	1	1
CMP %R15D,%R8D	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	1	0.25
JNE 2121b0 <.omp_outlined..4+0x1a0>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	1	0.50-1

Report Configuration

Vectorization

Details

Workaround

Execution units bottlenecks

Workaround

Expensive FP math instructions/calls

FMA

Slow data structures access

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

Vectorization

Details

Workaround

Execution units bottlenecks

Workaround

Expensive FP math instructions/calls

FMA

Slow data structures access

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