OV - exec - Loop 98

	Loop Id: 98	Module: exec	Source: timestep.c:88-94	Coverage: 1.20%

0x40e903 MOVSXD	(%RCX),%RSI    [2]

0x40e906 MOVSXD	0x4(%RCX),%R12    [2]

0x40e90a VMULSD	(%RDX),%XMM0,%XMM15    [6]

0x40e90e ADD	$0x10,%RCX

0x40e912 ADD	$0x60,%RAX

0x40e916 ADD	$0x60,%RDX

0x40e91a SAL	$0x4,%RSI

0x40e91e VDIVSD	0x8(%RBX,%RSI,1),%XMM2,%XMM14    [4]

0x40e924 MOVSXD	-0x8(%RCX),%RSI    [1]

0x40e928 VFMADD213SD	-0x60(%RAX),%XMM14,%XMM15    [3]

0x40e92e VMOVSD	%XMM15,-0x60(%RAX)    [3]

0x40e933 VMULSD	-0x58(%RDX),%XMM0,%XMM1    [5]

0x40e938 VFMADD213SD	-0x58(%RAX),%XMM14,%XMM1    [3]

0x40e93e VMOVSD	%XMM1,-0x58(%RAX)    [3]

0x40e943 VMULSD	-0x50(%RDX),%XMM0,%XMM3    [5]

0x40e948 SAL	$0x4,%R12

0x40e94c VFMADD213SD	-0x50(%RAX),%XMM3,%XMM14    [3]

0x40e952 VMOVSD	%XMM14,-0x50(%RAX)    [3]

0x40e957 VMULSD	-0x48(%RDX),%XMM0,%XMM5    [5]

0x40e95c VDIVSD	0x8(%RBX,%R12,1),%XMM2,%XMM4    [4]

0x40e963 VFMADD213SD	-0x48(%RAX),%XMM4,%XMM5    [3]

0x40e969 VMOVSD	%XMM5,-0x48(%RAX)    [3]

0x40e96e VMULSD	-0x40(%RDX),%XMM0,%XMM6    [5]

0x40e973 SAL	$0x4,%RSI

0x40e977 VFMADD213SD	-0x40(%RAX),%XMM4,%XMM6    [3]

0x40e97d VMOVSD	%XMM6,-0x40(%RAX)    [3]

0x40e982 VMULSD	-0x38(%RDX),%XMM0,%XMM7    [5]

0x40e987 VFMADD213SD	-0x38(%RAX),%XMM7,%XMM4    [3]

0x40e98d VMOVSD	%XMM4,-0x38(%RAX)    [3]

0x40e992 VMULSD	-0x30(%RDX),%XMM0,%XMM9    [5]

0x40e997 VDIVSD	0x8(%RBX,%RSI,1),%XMM2,%XMM8    [4]

0x40e99d VFMADD213SD	-0x30(%RAX),%XMM8,%XMM9    [3]

0x40e9a3 VMOVSD	%XMM9,-0x30(%RAX)    [3]

0x40e9a8 VMULSD	-0x28(%RDX),%XMM0,%XMM10    [5]

0x40e9ad VFMADD213SD	-0x28(%RAX),%XMM8,%XMM10    [3]

0x40e9b3 VMOVSD	%XMM10,-0x28(%RAX)    [3]

0x40e9b8 VMULSD	-0x20(%RDX),%XMM0,%XMM11    [5]

0x40e9bd VFMADD213SD	-0x20(%RAX),%XMM11,%XMM8    [3]

0x40e9c3 VMOVSD	%XMM8,-0x20(%RAX)    [3]

0x40e9c8 MOVSXD	-0x4(%RCX),%R12    [1]

0x40e9cc VMULSD	-0x18(%RDX),%XMM0,%XMM13    [5]

0x40e9d1 SAL	$0x4,%R12

0x40e9d5 VDIVSD	0x8(%RBX,%R12,1),%XMM2,%XMM12    [4]

0x40e9dc VFMADD213SD	-0x18(%RAX),%XMM12,%XMM13    [3]

0x40e9e2 VMOVSD	%XMM13,-0x18(%RAX)    [3]

0x40e9e7 VMULSD	-0x10(%RDX),%XMM0,%XMM14    [5]

0x40e9ec VFMADD213SD	-0x10(%RAX),%XMM12,%XMM14    [3]

0x40e9f2 VMOVSD	%XMM14,-0x10(%RAX)    [3]

0x40e9f7 VMULSD	-0x8(%RDX),%XMM0,%XMM15    [5]

0x40e9fc VFMADD213SD	-0x8(%RAX),%XMM15,%XMM12    [3]

0x40ea02 VMOVSD	%XMM12,-0x8(%RAX)    [3]

0x40ea07 CMP	%RCX,%R8

0x40ea0a JNE	40e903

/beegfs/hackathon/users/eoseret/qaas_runs_test/gmz12.benchmarkcenter.megware.com/177-374-1600/CoMD/build/CoMD/CoMD/src-openmp/timestep.c: 88 - 94

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

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

89:       {

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

91:          real_t invMass = 1.0/s->species[iSpecies].mass;

92:          s->atoms->r[iOff][0] += dt*s->atoms->p[iOff][0]*invMass;

93:          s->atoms->r[iOff][1] += dt*s->atoms->p[iOff][1]*invMass;

94:          s->atoms->r[iOff][2] += dt*s->atoms->p[iOff][2]*invMass;

Coverage (%)	Name	Source Location	Module
○99.60	gomp_thread_start	team.c:130	libgomp.so.1.0.0

min	med	avg	max

Percentile Index	10	20	30	40	50	60	70	80	90	100
Value

min	med	avg	max

Percentile Index	10	20	30	40	50	60	70	80	90	100
Value

Path /

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.58
CQA speedup if fully vectorized	8.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.14
Bottlenecks	P10, P11,
Function	advancePosition._omp_fn.0
Source	timestep.c:88-94
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	16.00
CQA cycles if no scalar integer	16.00
CQA cycles if FP arith vectorized	10.13
CQA cycles if fully vectorized	2.00
Front-end cycles	6.50
P0 cycles	2.00
P1 cycles	2.00
P2 cycles	2.00
P3 cycles	2.00
P4 cycles	2.00
P5 cycles	2.00
P6 cycles	11.00
P7 cycles	11.00
P8 cycles	11.00
P9 cycles	11.00
P10 cycles	14.00
P11 cycles	14.00
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	6.00
P15 cycles	6.00
DIV/SQRT cycles	16.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	53.00
Nb uops	52.00
Nb loads	32.00
Nb stores	12.00
Nb stack references	0.00
FLOP/cycle	2.50
Nb FLOP add-sub	0.00
Nb FLOP mul	12.00
Nb FLOP fma	12.00
Nb FLOP div	4.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	21.00
Bytes prefetched	0.00
Bytes loaded	240.00
Bytes stored	96.00
Stride 0	1.00
Stride 1	1.00
Stride n	4.00
Stride unknown	0.00
Stride indirect	0.00
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	1.58
CQA speedup if fully vectorized	8.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.14
Bottlenecks	P10, P11,
Function	advancePosition._omp_fn.0
Source	timestep.c:88-94
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	16.00
CQA cycles if no scalar integer	16.00
CQA cycles if FP arith vectorized	10.13
CQA cycles if fully vectorized	2.00
Front-end cycles	6.50
P0 cycles	2.00
P1 cycles	2.00
P2 cycles	2.00
P3 cycles	2.00
P4 cycles	2.00
P5 cycles	2.00
P6 cycles	11.00
P7 cycles	11.00
P8 cycles	11.00
P9 cycles	11.00
P10 cycles	14.00
P11 cycles	14.00
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	6.00
P15 cycles	6.00
DIV/SQRT cycles	16.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	53.00
Nb uops	52.00
Nb loads	32.00
Nb stores	12.00
Nb stack references	0.00
FLOP/cycle	2.50
Nb FLOP add-sub	0.00
Nb FLOP mul	12.00
Nb FLOP fma	12.00
Nb FLOP div	4.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	21.00
Bytes prefetched	0.00
Bytes loaded	240.00
Bytes stored	96.00
Stride 0	1.00
Stride 1	1.00
Stride n	4.00
Stride unknown	0.00
Stride indirect	0.00
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	advancePosition._omp_fn.0
Source file and lines	timestep.c:88-94
Module	exec

The loop is defined in /beegfs/hackathon/users/eoseret/qaas_runs_test/gmz12.benchmarkcenter.megware.com/177-374-1600/CoMD/build/CoMD/CoMD/src-openmp/timestep.c:88-94.

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 16.00 to 2.00 cycles (8.00x 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 ftree-vectorize (included in O3) to enable loop vectorization (including 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 16.00 to 14.00 cycles (1.14x 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 16.00 to 12.00 cycles (1.33x speedup).

FMA

Detected 12 FMA (fused multiply-add) operations.

Slow data structures access

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

Details

Constant non-unit stride: 4 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)

Type of elements and instruction set

28 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 40 FP arithmetical operations:

12: addition or subtraction (all inside FMA instructions)
24: multiply (12 inside FMA instructions)
4: divide

The binary loop is loading 240 bytes (30 double precision FP elements). The binary loop is storing 96 bytes (12 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.12 FP operations per loaded or stored byte.

Unroll opportunity

Loop is potentially data access bound.

Workaround

Unroll your loop if trip count is significantly higher than target unroll factor and if some data references are common to consecutive iterations. This can be done manually. Or by recompiling with -funroll-loops and/or -floop-unroll-and-jam. Or with the unroll (resp. unroll_and_jam) directive on top of the inner (resp. surrounding) loop. You can enforce an unroll factor: #pragma GCC unroll N

General properties

nb instructions	53
nb uops	52
loop length	269
used x86 registers	7
used mmx registers	0
used xmm registers	16
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	6.50 cycles
front end	6.50 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	2.00	2.00	2.00	2.00	2.00	2.00	11.00	11.00	11.00	11.00	14.00	14.00	0.00	0.00	6.00	6.00
cycles	2.00	2.00	2.00	2.00	2.00	2.00	11.00	11.00	11.00	11.00	14.00	14.00	0.00	0.00	6.00	6.00

Execution ports to units layout:

P0: ALU
P1: ALU
P2: ALU
P3: ALU, BRU
P4: ALU, BRU
P5: ALU, BRU
P6 (512 bits): store address, load
P7 (512 bits): store address, load
P8 (512 bits): store address, load
P9 (512 bits): store address, load
P10 (512 bits): VPU, DIV/SQRT
P11 (512 bits): VPU, DIV/SQRT
P12 (512 bits): VPU
P13 (512 bits): VPU
P14 (512 bits): FP store data
P15 (512 bits): FP store data

Cycles executing div or sqrt instructions	16.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	6.50
Dispatch	14.00
DIV/SQRT	16.00
Data deps.	1.00
Overall L1	16.00

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 16.00 cycles. At this rate:

5% of peak load performance is reached (15.00 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
4% of peak store performance is reached (6.00 out of 128.00 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: 40e903

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P14	P15	Latency	Recip. throughput	Vectorization
MOVSXD (%RCX),%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	1	0.25	N/A
MOVSXD 0x4(%RCX),%R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	1	0.25	N/A
VMULSD (%RDX),%XMM0,%XMM15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADD $0x10,%RCX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD $0x60,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD $0x60,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
SAL $0x4,%RSI	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.33	N/A
VDIVSD 0x8(%RBX,%RSI,1),%XMM2,%XMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	13	4	scal (12.5%)
MOVSXD -0x8(%RCX),%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	1	0.25	N/A
VFMADD213SD -0x60(%RAX),%XMM14,%XMM15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM15,-0x60(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x58(%RDX),%XMM0,%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x58(%RAX),%XMM14,%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM1,-0x58(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x50(%RDX),%XMM0,%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
SAL $0x4,%R12	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.33	N/A
VFMADD213SD -0x50(%RAX),%XMM3,%XMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM14,-0x50(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x48(%RDX),%XMM0,%XMM5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VDIVSD 0x8(%RBX,%R12,1),%XMM2,%XMM4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	13	4	scal (12.5%)
VFMADD213SD -0x48(%RAX),%XMM4,%XMM5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM5,-0x48(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x40(%RDX),%XMM0,%XMM6	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
SAL $0x4,%RSI	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.33	N/A
VFMADD213SD -0x40(%RAX),%XMM4,%XMM6	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM6,-0x40(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x38(%RDX),%XMM0,%XMM7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x38(%RAX),%XMM7,%XMM4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM4,-0x38(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x30(%RDX),%XMM0,%XMM9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VDIVSD 0x8(%RBX,%RSI,1),%XMM2,%XMM8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	13	4	scal (12.5%)
VFMADD213SD -0x30(%RAX),%XMM8,%XMM9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM9,-0x30(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x28(%RDX),%XMM0,%XMM10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x28(%RAX),%XMM8,%XMM10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM10,-0x28(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x20(%RDX),%XMM0,%XMM11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x20(%RAX),%XMM11,%XMM8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM8,-0x20(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
MOVSXD -0x4(%RCX),%R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	1	0.25	N/A
VMULSD -0x18(%RDX),%XMM0,%XMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
SAL $0x4,%R12	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.33	N/A
VDIVSD 0x8(%RBX,%R12,1),%XMM2,%XMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	13	4	scal (12.5%)
VFMADD213SD -0x18(%RAX),%XMM12,%XMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM13,-0x18(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x10(%RDX),%XMM0,%XMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x10(%RAX),%XMM12,%XMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM14,-0x10(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x8(%RDX),%XMM0,%XMM15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x8(%RAX),%XMM15,%XMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM12,-0x8(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
CMP %RCX,%R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
JNE 40e903 <advancePosition._omp_fn.0+0x1c3>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	0.33-1	N/A

Function	advancePosition._omp_fn.0
Source file and lines	timestep.c:88-94
Module	exec

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 16.00 to 2.00 cycles (8.00x 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 ftree-vectorize (included in O3) to enable loop vectorization (including 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 16.00 to 12.00 cycles (1.33x speedup).

FMA

Detected 12 FMA (fused multiply-add) operations.

Slow data structures access

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

Details

Constant non-unit stride: 4 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)

Type of elements and instruction set

28 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 40 FP arithmetical operations:

12: addition or subtraction (all inside FMA instructions)
24: multiply (12 inside FMA instructions)
4: divide

The binary loop is loading 240 bytes (30 double precision FP elements). The binary loop is storing 96 bytes (12 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.12 FP operations per loaded or stored byte.

Unroll opportunity

Loop is potentially data access bound.

Workaround

General properties

nb instructions	53
nb uops	52
loop length	269
used x86 registers	7
used mmx registers	0
used xmm registers	16
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	6.50 cycles
front end	6.50 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	2.00	2.00	2.00	2.00	2.00	2.00	11.00	11.00	11.00	11.00	14.00	14.00	0.00	0.00	6.00	6.00
cycles	2.00	2.00	2.00	2.00	2.00	2.00	11.00	11.00	11.00	11.00	14.00	14.00	0.00	0.00	6.00	6.00

Execution ports to units layout:

P0: ALU
P1: ALU
P2: ALU
P3: ALU, BRU
P4: ALU, BRU
P5: ALU, BRU
P6 (512 bits): store address, load
P7 (512 bits): store address, load
P8 (512 bits): store address, load
P9 (512 bits): store address, load
P10 (512 bits): VPU, DIV/SQRT
P11 (512 bits): VPU, DIV/SQRT
P12 (512 bits): VPU
P13 (512 bits): VPU
P14 (512 bits): FP store data
P15 (512 bits): FP store data

Cycles executing div or sqrt instructions	16.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	6.50
Dispatch	14.00
DIV/SQRT	16.00
Data deps.	1.00
Overall L1	16.00

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 16.00 cycles. At this rate:

5% of peak load performance is reached (15.00 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
4% of peak store performance is reached (6.00 out of 128.00 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: 40e903

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P14	P15	Latency	Recip. throughput	Vectorization
MOVSXD (%RCX),%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	1	0.25	N/A
MOVSXD 0x4(%RCX),%R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	1	0.25	N/A
VMULSD (%RDX),%XMM0,%XMM15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADD $0x10,%RCX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD $0x60,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD $0x60,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
SAL $0x4,%RSI	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.33	N/A
VDIVSD 0x8(%RBX,%RSI,1),%XMM2,%XMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	13	4	scal (12.5%)
MOVSXD -0x8(%RCX),%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	1	0.25	N/A
VFMADD213SD -0x60(%RAX),%XMM14,%XMM15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM15,-0x60(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x58(%RDX),%XMM0,%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x58(%RAX),%XMM14,%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM1,-0x58(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x50(%RDX),%XMM0,%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
SAL $0x4,%R12	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.33	N/A
VFMADD213SD -0x50(%RAX),%XMM3,%XMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM14,-0x50(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x48(%RDX),%XMM0,%XMM5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VDIVSD 0x8(%RBX,%R12,1),%XMM2,%XMM4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	13	4	scal (12.5%)
VFMADD213SD -0x48(%RAX),%XMM4,%XMM5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM5,-0x48(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x40(%RDX),%XMM0,%XMM6	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
SAL $0x4,%RSI	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.33	N/A
VFMADD213SD -0x40(%RAX),%XMM4,%XMM6	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM6,-0x40(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x38(%RDX),%XMM0,%XMM7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x38(%RAX),%XMM7,%XMM4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM4,-0x38(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x30(%RDX),%XMM0,%XMM9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VDIVSD 0x8(%RBX,%RSI,1),%XMM2,%XMM8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	13	4	scal (12.5%)
VFMADD213SD -0x30(%RAX),%XMM8,%XMM9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM9,-0x30(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x28(%RDX),%XMM0,%XMM10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x28(%RAX),%XMM8,%XMM10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM10,-0x28(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x20(%RDX),%XMM0,%XMM11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x20(%RAX),%XMM11,%XMM8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM8,-0x20(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
MOVSXD -0x4(%RCX),%R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	1	0.25	N/A
VMULSD -0x18(%RDX),%XMM0,%XMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
SAL $0x4,%R12	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.33	N/A
VDIVSD 0x8(%RBX,%R12,1),%XMM2,%XMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	13	4	scal (12.5%)
VFMADD213SD -0x18(%RAX),%XMM12,%XMM13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM13,-0x18(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x10(%RDX),%XMM0,%XMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x10(%RAX),%XMM12,%XMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM14,-0x10(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMULSD -0x8(%RDX),%XMM0,%XMM15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
VFMADD213SD -0x8(%RAX),%XMM15,%XMM12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	4	0.50	scal (12.5%)
VMOVSD %XMM12,-0x8(%RAX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	1	0.50	scal (12.5%)
CMP %RCX,%R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
JNE 40e903 <advancePosition._omp_fn.0+0x1c3>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	0.33-1	N/A

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

Unroll opportunity

Workaround

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

Unroll opportunity

Workaround

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code