OV - exec - Loop 59

	Loop Id: 59	Module: exec	Source: haloExchange.c:633-642	Coverage: 0.06%

0x40a830 MOV	(%RAX),%EBX    [6]

0x40a832 LEA	0x1(%R8),%R10

0x40a836 VMOVSD	0x8(%RAX),%XMM1    [6]

0x40a83b ADD	$0x70,%RAX

0x40a83f MOV	%EBX,(%R13,%R8,4)    [3]

0x40a844 MOV	-0x6c(%RAX),%ECX    [5]

0x40a847 MOV	%ECX,(%R14,%R8,4)    [4]

0x40a84b MOV	-0x38(%RAX),%R15D    [5]

0x40a84f VMOVSD	%XMM1,(%RDI,%RDX,1)    [1]

0x40a854 VMOVSD	-0x60(%RAX),%XMM14    [5]

0x40a859 VMOVSD	%XMM14,0x8(%RDI,%RDX,1)    [1]

0x40a85f VMOVSD	-0x58(%RAX),%XMM11    [5]

0x40a864 VMOVSD	%XMM11,0x10(%RDI,%RDX,1)    [1]

0x40a86a VMOVSD	-0x50(%RAX),%XMM0    [5]

0x40a86f VMOVSD	%XMM0,(%RSI,%RDX,1)    [2]

0x40a874 VMOVSD	-0x48(%RAX),%XMM10    [5]

0x40a879 VMOVSD	%XMM10,0x8(%RSI,%RDX,1)    [2]

0x40a87f VMOVSD	-0x40(%RAX),%XMM5    [5]

0x40a884 VMOVSD	%XMM5,0x10(%RSI,%RDX,1)    [2]

0x40a88a ADD	$0x2,%R8

0x40a88e VMOVSD	-0x30(%RAX),%XMM8    [5]

0x40a893 MOV	%R15D,(%R13,%R10,4)    [3]

0x40a898 MOV	-0x34(%RAX),%R9D    [5]

0x40a89c MOV	%R9D,(%R14,%R10,4)    [4]

0x40a8a0 VMOVSD	%XMM8,0x18(%RDI,%RDX,1)    [1]

0x40a8a6 VMOVSD	-0x28(%RAX),%XMM9    [5]

0x40a8ab VMOVSD	%XMM9,0x20(%RDI,%RDX,1)    [1]

0x40a8b1 VMOVSD	-0x20(%RAX),%XMM6    [5]

0x40a8b6 VMOVSD	%XMM6,0x28(%RDI,%RDX,1)    [1]

0x40a8bc VMOVSD	-0x18(%RAX),%XMM2    [5]

0x40a8c1 VMOVSD	%XMM2,0x18(%RSI,%RDX,1)    [2]

0x40a8c7 VMOVSD	-0x10(%RAX),%XMM4    [5]

0x40a8cc VMOVSD	%XMM4,0x20(%RSI,%RDX,1)    [2]

0x40a8d2 VMOVSD	-0x8(%RAX),%XMM7    [5]

0x40a8d7 VMOVSD	%XMM7,0x28(%RSI,%RDX,1)    [2]

0x40a8dd ADD	$0x30,%RDX

0x40a8e1 CMP	%R8D,%R12D

0x40a8e4 JG	40a830

/home/eoseret/qaas_runs_ZEN5/174-043-3878/intel/CoMD/build/CoMD/CoMD/src-openmp/haloExchange.c: 633 - 642

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

633:    for (int ii=begin, iTmp=0; ii<end; ++ii, ++iTmp)

634:    {

635:       atoms->gid[ii]   = tmp[iTmp].gid;

636:       atoms->iSpecies[ii] = tmp[iTmp].type;

637:       atoms->r[ii][0]  = tmp[iTmp].rx;

638:       atoms->r[ii][1]  = tmp[iTmp].ry;

639:       atoms->r[ii][2]  = tmp[iTmp].rz;

640:       atoms->p[ii][0]  = tmp[iTmp].px;

641:       atoms->p[ii][1]  = tmp[iTmp].py;

642:       atoms->p[ii][2]  = tmp[iTmp].pz;

Coverage (%)	Name	Source Location	Module
►13.66+	redistributeAtoms._omp_fn.0	timestep.c:154	exec
○	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►13.66+	redistributeAtoms._omp_fn.0	timestep.c:154	exec
○	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►13.52+	redistributeAtoms._omp_fn.0	timestep.c:154	exec
○	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►12.25+	redistributeAtoms._omp_fn.0	timestep.c:154	exec
○	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►12.11+	redistributeAtoms._omp_fn.0	timestep.c:154	exec
○	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►11.69+	redistributeAtoms._omp_fn.0	timestep.c:154	exec
○	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►10.70+	redistributeAtoms._omp_fn.0	timestep.c:154	exec
○	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►10.28+	redistributeAtoms._omp_fn.0	timestep.c:154	exec
○	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6

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	1.33
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	9.14
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.33
Bottlenecks	P6, P7, P8, P9,
Function	sortAtomsInCell
Source	haloExchange.c:633-642
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	8.00
CQA cycles if no scalar integer	6.00
CQA cycles if FP arith vectorized	8.00
CQA cycles if fully vectorized	0.88
Front-end cycles	4.63
P0 cycles	0.83
P1 cycles	0.83
P2 cycles	0.83
P3 cycles	0.83
P4 cycles	0.83
P5 cycles	0.83
P6 cycles	8.00
P7 cycles	8.00
P8 cycles	8.00
P9 cycles	8.00
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	6.00
P15 cycles	6.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	38.00
Nb uops	37.00
Nb loads	16.00
Nb stores	16.00
Nb stack references	0.00
FLOP/cycle	0.00
Nb FLOP add-sub	0.00
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	28.00
Bytes prefetched	0.00
Bytes loaded	112.00
Bytes stored	112.00
Stride 0	0.00
Stride 1	2.00
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	NA
Vectorization ratio add_sub	NA
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	NA
Vector-efficiency ratio all	10.94
Vector-efficiency ratio load	10.94
Vector-efficiency ratio store	10.94
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	NA

Metric	Value
CQA speedup if no scalar integer	1.33
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	9.14
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.33
Bottlenecks	P6, P7, P8, P9,
Function	sortAtomsInCell
Source	haloExchange.c:633-642
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	8.00
CQA cycles if no scalar integer	6.00
CQA cycles if FP arith vectorized	8.00
CQA cycles if fully vectorized	0.88
Front-end cycles	4.63
P0 cycles	0.83
P1 cycles	0.83
P2 cycles	0.83
P3 cycles	0.83
P4 cycles	0.83
P5 cycles	0.83
P6 cycles	8.00
P7 cycles	8.00
P8 cycles	8.00
P9 cycles	8.00
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	6.00
P15 cycles	6.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	38.00
Nb uops	37.00
Nb loads	16.00
Nb stores	16.00
Nb stack references	0.00
FLOP/cycle	0.00
Nb FLOP add-sub	0.00
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	28.00
Bytes prefetched	0.00
Bytes loaded	112.00
Bytes stored	112.00
Stride 0	0.00
Stride 1	2.00
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	NA
Vectorization ratio add_sub	NA
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	NA
Vector-efficiency ratio all	10.94
Vector-efficiency ratio load	10.94
Vector-efficiency ratio store	10.94
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	NA

Path /

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

Function	sortAtomsInCell
Source file and lines	haloExchange.c:633-642
Module	exec

The loop is defined in /home/eoseret/qaas_runs_ZEN5/174-043-3878/intel/CoMD/build/CoMD/CoMD/src-openmp/haloExchange.c:633-642.

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

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 8.00 to 6.00 cycles (1.33x speedup).

Workaround

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

Vectorization

Your loop is not vectorized. Only 10% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your loop, you can lower the cost of an iteration from 8.00 to 0.87 cycles (9.14x 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 fassociative-math (included in Ofast or ffast-math) 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:

reading data from caches/RAM (load units are a bottleneck)
writing data to caches/RAM (the store unit is a bottleneck)

By removing all these bottlenecks, you can lower the cost of an iteration from 8.00 to 6.00 cycles (1.33x speedup).

Workaround

Read less array elements
Write less array elements
Provide more information to your compiler:
- hardcode the bounds of the corresponding 'for' loop
- use the 'restrict' C99 keyword

No data for this section

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

No instructions are processing arithmetic or math operations on FP elements. This loop is probably writing/copying data or processing integer elements.

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

The binary loop does not contain any FP arithmetical operations. The binary loop is loading 112 bytes. The binary loop is storing 112 bytes.

Unroll opportunity

Loop is 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	38
nb uops	37
loop length	186
used x86 registers	13
used mmx registers	0
used xmm registers	12
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	4.63 cycles
front end	4.63 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	0.83	0.83	0.83	0.83	0.83	0.83	8.00	8.00	8.00	8.00	0.00	0.00	0.00	0.00	6.00	6.00
cycles	0.83	0.83	0.83	0.83	0.83	0.83	8.00	8.00	8.00	8.00	0.00	0.00	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	NA
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	4.63
Dispatch	8.00
Data deps.	1.00
Overall L1	8.00

Vectorization ratios

INT

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

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

INT+FP

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

Vector efficiency ratios

INT

all	6%
load	6%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

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

INT+FP

all	10%
load	10%
store	10%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
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 8.00 cycles. At this rate:

5% of peak load performance is reached (14.00 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
10% of peak store performance is reached (14.00 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: 40a830

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P14	P15	Latency	Recip. throughput	Vectorization
MOV (%RAX),%EBX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (6.3%)
LEA 0x1(%R8),%R10	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.25	N/A
VMOVSD 0x8(%RAX),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
ADD $0x70,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
MOV %EBX,(%R13,%R8,4)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (6.3%)
MOV -0x6c(%RAX),%ECX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (6.3%)
MOV %ECX,(%R14,%R8,4)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (6.3%)
MOV -0x38(%RAX),%R15D	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (6.3%)
VMOVSD %XMM1,(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x60(%RAX),%XMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM14,0x8(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x58(%RAX),%XMM11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM11,0x10(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x50(%RAX),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM0,(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x48(%RAX),%XMM10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM10,0x8(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x40(%RAX),%XMM5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM5,0x10(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
ADD $0x2,%R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
VMOVSD -0x30(%RAX),%XMM8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
MOV %R15D,(%R13,%R10,4)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (6.3%)
MOV -0x34(%RAX),%R9D	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (6.3%)
MOV %R9D,(%R14,%R10,4)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (6.3%)
VMOVSD %XMM8,0x18(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x28(%RAX),%XMM9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM9,0x20(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x20(%RAX),%XMM6	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM6,0x28(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x18(%RAX),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM2,0x18(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x10(%RAX),%XMM4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM4,0x20(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x8(%RAX),%XMM7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM7,0x28(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
ADD $0x30,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
CMP %R8D,%R12D	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
JG 40a830 <sortAtomsInCell+0x1080>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A

Function	sortAtomsInCell
Source file and lines	haloExchange.c:633-642
Module	exec

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 8.00 to 6.00 cycles (1.33x speedup).

Workaround

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

Vectorization

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 fassociative-math (included in Ofast or ffast-math) 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:

reading data from caches/RAM (load units are a bottleneck)
writing data to caches/RAM (the store unit is a bottleneck)

By removing all these bottlenecks, you can lower the cost of an iteration from 8.00 to 6.00 cycles (1.33x speedup).

Workaround

Read less array elements
Write less array elements
Provide more information to your compiler:
- hardcode the bounds of the corresponding 'for' loop
- use the 'restrict' C99 keyword

No data for this section

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

No instructions are processing arithmetic or math operations on FP elements. This loop is probably writing/copying data or processing integer elements.

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

The binary loop does not contain any FP arithmetical operations. The binary loop is loading 112 bytes. The binary loop is storing 112 bytes.

Unroll opportunity

Loop is data access bound.

Workaround

General properties

nb instructions	38
nb uops	37
loop length	186
used x86 registers	13
used mmx registers	0
used xmm registers	12
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	4.63 cycles
front end	4.63 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	0.83	0.83	0.83	0.83	0.83	0.83	8.00	8.00	8.00	8.00	0.00	0.00	0.00	0.00	6.00	6.00
cycles	0.83	0.83	0.83	0.83	0.83	0.83	8.00	8.00	8.00	8.00	0.00	0.00	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	NA
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	4.63
Dispatch	8.00
Data deps.	1.00
Overall L1	8.00

Vectorization ratios

INT

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

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

INT+FP

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

Vector efficiency ratios

INT

all	6%
load	6%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

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

INT+FP

all	10%
load	10%
store	10%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
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 8.00 cycles. At this rate:

5% of peak load performance is reached (14.00 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
10% of peak store performance is reached (14.00 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: 40a830

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P14	P15	Latency	Recip. throughput	Vectorization
MOV (%RAX),%EBX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (6.3%)
LEA 0x1(%R8),%R10	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.25	N/A
VMOVSD 0x8(%RAX),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
ADD $0x70,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
MOV %EBX,(%R13,%R8,4)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (6.3%)
MOV -0x6c(%RAX),%ECX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (6.3%)
MOV %ECX,(%R14,%R8,4)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (6.3%)
MOV -0x38(%RAX),%R15D	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (6.3%)
VMOVSD %XMM1,(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x60(%RAX),%XMM14	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM14,0x8(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x58(%RAX),%XMM11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM11,0x10(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x50(%RAX),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM0,(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x48(%RAX),%XMM10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM10,0x8(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x40(%RAX),%XMM5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM5,0x10(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
ADD $0x2,%R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
VMOVSD -0x30(%RAX),%XMM8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
MOV %R15D,(%R13,%R10,4)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (6.3%)
MOV -0x34(%RAX),%R9D	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (6.3%)
MOV %R9D,(%R14,%R10,4)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (6.3%)
VMOVSD %XMM8,0x18(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x28(%RAX),%XMM9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM9,0x20(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x20(%RAX),%XMM6	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM6,0x28(%RDI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x18(%RAX),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM2,0x18(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x10(%RAX),%XMM4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM4,0x20(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD -0x8(%RAX),%XMM7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD %XMM7,0x28(%RSI,%RDX,1)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
ADD $0x30,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
CMP %R8D,%R12D	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
JG 40a830 <sortAtomsInCell+0x1080>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

Workaround

Slow data structures access

Details

Workaround

Type of elements and instruction set

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

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

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

Workaround

Slow data structures access

Details

Workaround

Type of elements and instruction set

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

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