OV - exec - Function loadAtomsBuffer

	Function: loadAtomsBuffer	Module: exec	Source: haloExchange.c:362-393 [...]	Coverage (incl. loops): 0.04% \| (excl. loops): 0.00%

/home/eoseret/qaas_runs_ZEN5/174-043-3878/intel/CoMD/build/CoMD/CoMD/src-openmp/haloExchange.c: 362 - 393

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

362: {

363:    AtomExchangeParms* parms = (AtomExchangeParms*) vparms;

364:    SimFlat* s = (SimFlat*) data;

365:    AtomMsg* buf = (AtomMsg*) charBuf;

366:

367:    real_t* pbcFactor = parms->pbcFactor[face];

[...]

373:    int nCells = parms->nCells[face];

374:    int* cellList = parms->cellList[face];

375:    int nBuf = 0;

376:    for (int iCell=0; iCell<nCells; ++iCell)

377:    {

378:       int iBox = cellList[iCell];

379:       int iOff = iBox*MAXATOMS;

380:       for (int ii=iOff; ii<iOff+s->boxes->nAtoms[iBox]; ++ii)

381:       {

382:          buf[nBuf].gid  = s->atoms->gid[ii];

383:          buf[nBuf].type = s->atoms->iSpecies[ii];

384:          buf[nBuf].rx = s->atoms->r[ii][0] + shift[0];

385:          buf[nBuf].ry = s->atoms->r[ii][1] + shift[1];

386:          buf[nBuf].rz = s->atoms->r[ii][2] + shift[2];

387:          buf[nBuf].px = s->atoms->p[ii][0];

388:          buf[nBuf].py = s->atoms->p[ii][1];

389:          buf[nBuf].pz = s->atoms->p[ii][2];

390:          ++nBuf;

391:       }

392:    }

393:    return nBuf*sizeof(AtomMsg);

0x20ab30 PUSH	%RBP

0x20ab31 MOV	%RSP,%RBP

0x20ab34 PUSH	%R15

0x20ab36 PUSH	%R14

0x20ab38 PUSH	%R13

0x20ab3a PUSH	%R12

0x20ab3c PUSH	%RBX

0x20ab3d MOV	%RCX,-0x38(%RBP)

0x20ab41 MOVSXD	%EDX,%RCX

0x20ab44 MOV	(%RDI,%RCX,4),%EDX

0x20ab47 TEST	%EDX,%EDX

0x20ab49 JLE	20ac7a

0x20ab4f MOV	0x10(%RSI),%R8

0x20ab53 MOV	0x18(%RDI,%RCX,8),%R10

0x20ab58 MOV	0x48(%RDI,%RCX,8),%RCX

0x20ab5d MOV	%RDX,%RAX

0x20ab60 MOV	%RSI,-0x30(%RBP)

0x20ab64 XOR	%R9D,%R9D

0x20ab67 XOR	%EDI,%EDI

0x20ab69 VMOVSD	0x48(%R8),%XMM0

0x20ab6f VMOVSD	0x50(%R8),%XMM1

0x20ab75 VMOVSD	0x58(%R8),%XMM2

0x20ab7b MOV	%R10,-0x40(%RBP)

0x20ab7f VMULSD	(%RCX),%XMM0,%XMM0

0x20ab83 VMULSD	0x8(%RCX),%XMM1,%XMM1

0x20ab88 VMULSD	0x10(%RCX),%XMM2,%XMM2

0x20ab8d MOV	0x18(%RSI),%RCX

0x20ab91 MOV	%RAX,%RSI

0x20ab94 MOV	0x78(%RCX),%R8

0x20ab98 JMP	20abac

0x20ab9a NOPW	(%RAX,%RAX,1)

(44) 0x20aba0 INC	%R9

(44) 0x20aba3 CMP	%RSI,%R9

(44) 0x20aba6 JE	20ac7c

(44) 0x20abac MOV	-0x40(%RBP),%RAX

(44) 0x20abb0 MOVSXD	(%RAX,%R9,4),%R10

(44) 0x20abb4 CMPL	$0,(%R8,%R10,4)

(44) 0x20abb9 JLE	20aba0

(44) 0x20abbb MOV	-0x30(%RBP),%RAX

(44) 0x20abbf MOV	%R10D,%ECX

(44) 0x20abc2 SAL	$0x6,%ECX

(44) 0x20abc5 MOVSXD	%ECX,%RCX

(44) 0x20abc8 MOV	%RCX,%R12

(44) 0x20abcb MOV	0x20(%RAX),%RDX

(44) 0x20abcf MOVSXD	%EDI,%RAX

(44) 0x20abd2 IMUL	$0x38,%RAX,%R13

(44) 0x20abd6 ADD	-0x38(%RBP),%R13

(44) 0x20abda MOV	0x8(%RDX),%R11

(44) 0x20abde MOV	0x10(%RDX),%RBX

(44) 0x20abe2 MOV	0x18(%RDX),%R14

(44) 0x20abe6 MOV	0x20(%RDX),%R15

(44) 0x20abea LEA	(%RCX,%RCX,2),%RDX

(44) 0x20abee LEA	0x4(%RDX,%RDX,1),%RDX

(44) 0x20abf3 NOPW	%CS:(%RAX,%RAX,1)

  (45) 0x20ac00 MOV	(%R11,%R12,4),%EAX

  (45) 0x20ac04 INC	%EDI

  (45) 0x20ac06 MOV	%EAX,(%R13)

  (45) 0x20ac0a MOV	(%RBX,%R12,4),%EAX

  (45) 0x20ac0e INC	%R12

  (45) 0x20ac11 MOV	%EAX,0x4(%R13)

  (45) 0x20ac15 VADDSD	-0x10(%R14,%RDX,4),%XMM0,%XMM3

  (45) 0x20ac1c VMOVSD	%XMM3,0x8(%R13)

  (45) 0x20ac22 VADDSD	-0x8(%R14,%RDX,4),%XMM1,%XMM3

  (45) 0x20ac29 VMOVSD	%XMM3,0x10(%R13)

  (45) 0x20ac2f VADDSD	(%R14,%RDX,4),%XMM2,%XMM3

  (45) 0x20ac35 VMOVSD	%XMM3,0x18(%R13)

  (45) 0x20ac3b VMOVSD	-0x10(%R15,%RDX,4),%XMM3

  (45) 0x20ac42 VMOVSD	%XMM3,0x20(%R13)

  (45) 0x20ac48 VMOVSD	-0x8(%R15,%RDX,4),%XMM3

  (45) 0x20ac4f VMOVSD	%XMM3,0x28(%R13)

  (45) 0x20ac55 VMOVSD	(%R15,%RDX,4),%XMM3

  (45) 0x20ac5b ADD	$0x6,%RDX

  (45) 0x20ac5f VMOVSD	%XMM3,0x30(%R13)

  (45) 0x20ac65 ADD	$0x38,%R13

  (45) 0x20ac69 MOVSXD	(%R8,%R10,4),%RAX

  (45) 0x20ac6d ADD	%RCX,%RAX

  (45) 0x20ac70 CMP	%RAX,%R12

  (45) 0x20ac73 JL	20ac00

(44) 0x20ac75 JMP	20aba0

0x20ac7a XOR	%EDI,%EDI

0x20ac7c IMUL	$0x38,%EDI,%EAX

0x20ac7f POP	%RBX

0x20ac80 POP	%R12

0x20ac82 POP	%R13

0x20ac84 POP	%R14

0x20ac86 POP	%R15

0x20ac88 POP	%RBP

0x20ac89 RET

0x20ac8a NOPW	(%RAX,%RAX,1)

Coverage (%)	Name	Source Location	Module

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 /

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

The code analyzed by CQA in that panel excludes loops and represents 0.00% of application time for run aocc_1

Source file and lines	haloExchange.c:362-393
Module	exec

The function is defined in /home/eoseret/qaas_runs_ZEN5/174-043-3878/intel/CoMD/build/CoMD/CoMD/src-openmp/haloExchange.c:362-367,373-393.

Warnings:
Ignoring paths for analysis

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 4.88 to 1.75 cycles (2.79x speedup).

Workaround

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

Vectorization

Your function is not vectorized. Only 11% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your function, you can lower the cost of an iteration from 4.88 to 0.39 cycles (12.48x 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 function can use their full power.

Workaround

Try another compiler or update/tune your current one
Make array accesses unit-stride:
- If your function 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

Type of elements and instruction set

3 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 function (in the source code) and the binary function

The binary function is composed of 3 FP arithmetical operations:

3: multiply

The binary function is loading 92 bytes (11 double precision FP elements). The binary function is storing 24 bytes (3 double precision FP elements).

Arithmetic intensity

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

General properties

nb instructions	41
nb uops	39
loop length	134
used x86 registers	15
used mmx registers	0
used xmm registers	3
used ymm registers	0
used zmm registers	0
nb stack references	3

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	4.88 cycles
front end	4.88 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	1.00	1.00	1.00	1.00	1.00	1.00	3.75	3.75	3.75	3.75	1.50	1.50	0.00	0.00	0.00	0.00
cycles	1.00	1.00	1.00	1.00	1.00	1.00	3.75	3.75	3.75	3.75	1.50	1.50	0.00	0.00	0.00	0.00

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

Cycles summary

Front-end	4.88
Dispatch	3.75
Overall L1	4.88

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	0%

all	0%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	0%
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	0%
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	0%

Vector efficiency ratios

INT

all	11%
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)
other	6%

all	12%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	12%
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	11%
load	12%
store	12%
mul	12%
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	6%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 4.88 cycles. At this rate:

7% of peak load performance is reached (18.87 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
3% of peak store performance is reached (4.92 out of 128.00 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). By removing all these bottlenecks, you can lower the cost of an iteration from 4.88 to 3.75 cycles (1.30x speedup).

ASM code

In the binary file, the address of the function is: 20ab30

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput	Vectorization
PUSH %RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
MOV %RSP,%RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
PUSH %R15	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %R14	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %R13	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %R12	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %RBX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
MOV %RCX,-0x38(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (12.5%)
MOVSXD %EDX,%RCX	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 (%RDI,%RCX,4),%EDX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
TEST %EDX,%EDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
JLE 20ac7a <loadAtomsBuffer+0x14a>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
MOV 0x10(%RSI),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV 0x18(%RDI,%RCX,8),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (12.5%)
MOV 0x48(%RDI,%RCX,8),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV %RDX,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV %RSI,-0x30(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (12.5%)
XOR %R9D,%R9D	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.25	scal (6.3%)
XOR %EDI,%EDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.25	N/A
VMOVSD 0x48(%R8),%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 0x50(%R8),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD 0x58(%R8),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
MOV %R10,-0x40(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (12.5%)
VMULSD (%RCX),%XMM0,%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	3	0.50	scal (12.5%)
VMULSD 0x8(%RCX),%XMM1,%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	3	0.50	scal (12.5%)
VMULSD 0x10(%RCX),%XMM2,%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	3	0.50	scal (12.5%)
MOV 0x18(%RSI),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV %RAX,%RSI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV 0x78(%RCX),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
JMP 20abac <loadAtomsBuffer+0x7c>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	1	N/A
NOPW (%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
XOR %EDI,%EDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.25	N/A
IMUL $0x38,%EDI,%EAX	1	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	3	0.33	N/A
POP %RBX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R12	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R13	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R14	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R15	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
RET	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50	N/A
NOPW (%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A

The code analyzed by CQA in that panel excludes loops and represents 0.00% of application time for run aocc_1

Source file and lines	haloExchange.c:362-393
Module	exec

The function is defined in /home/eoseret/qaas_runs_ZEN5/174-043-3878/intel/CoMD/build/CoMD/CoMD/src-openmp/haloExchange.c:362-367,373-393.

Warnings:
Ignoring paths for analysis

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 4.88 to 1.75 cycles (2.79x speedup).

Workaround

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

Vectorization

Details

Workaround

Try another compiler or update/tune your current one
Make array accesses unit-stride:
- If your function 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

Type of elements and instruction set

3 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 function (in the source code) and the binary function

The binary function is composed of 3 FP arithmetical operations:

3: multiply

The binary function is loading 92 bytes (11 double precision FP elements). The binary function is storing 24 bytes (3 double precision FP elements).

Arithmetic intensity

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

General properties

nb instructions	41
nb uops	39
loop length	134
used x86 registers	15
used mmx registers	0
used xmm registers	3
used ymm registers	0
used zmm registers	0
nb stack references	3

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	4.88 cycles
front end	4.88 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	1.00	1.00	1.00	1.00	1.00	1.00	3.75	3.75	3.75	3.75	1.50	1.50	0.00	0.00	0.00	0.00
cycles	1.00	1.00	1.00	1.00	1.00	1.00	3.75	3.75	3.75	3.75	1.50	1.50	0.00	0.00	0.00	0.00

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

Cycles summary

Front-end	4.88
Dispatch	3.75
Overall L1	4.88

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	0%

all	0%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	0%
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	0%
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	0%

Vector efficiency ratios

INT

all	11%
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)
other	6%

all	12%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	12%
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	11%
load	12%
store	12%
mul	12%
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	6%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 4.88 cycles. At this rate:

7% of peak load performance is reached (18.87 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
3% of peak store performance is reached (4.92 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

ASM code

In the binary file, the address of the function is: 20ab30

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput	Vectorization
PUSH %RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
MOV %RSP,%RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
PUSH %R15	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %R14	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %R13	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %R12	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %RBX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
MOV %RCX,-0x38(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (12.5%)
MOVSXD %EDX,%RCX	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 (%RDI,%RCX,4),%EDX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
TEST %EDX,%EDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
JLE 20ac7a <loadAtomsBuffer+0x14a>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
MOV 0x10(%RSI),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV 0x18(%RDI,%RCX,8),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (12.5%)
MOV 0x48(%RDI,%RCX,8),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV %RDX,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV %RSI,-0x30(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (12.5%)
XOR %R9D,%R9D	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.25	scal (6.3%)
XOR %EDI,%EDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.25	N/A
VMOVSD 0x48(%R8),%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 0x50(%R8),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
VMOVSD 0x58(%R8),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
MOV %R10,-0x40(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	4	0.50	scal (12.5%)
VMULSD (%RCX),%XMM0,%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	3	0.50	scal (12.5%)
VMULSD 0x8(%RCX),%XMM1,%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	3	0.50	scal (12.5%)
VMULSD 0x10(%RCX),%XMM2,%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	3	0.50	scal (12.5%)
MOV 0x18(%RSI),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV %RAX,%RSI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV 0x78(%RCX),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
JMP 20abac <loadAtomsBuffer+0x7c>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	1	N/A
NOPW (%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
XOR %EDI,%EDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.25	N/A
IMUL $0x38,%EDI,%EAX	1	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	3	0.33	N/A
POP %RBX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R12	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R13	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R14	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R15	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
RET	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50	N/A
NOPW (%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A

Name	Coverage (%)	Time (s)
▼loadAtomsBuffer–	0.04	0.01
▼Loop 44 - haloExchange.c:376-389 - exec–	0.01	0.10
○Loop 45 - haloExchange.c:380-389 - exec	0.03	0.28

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

Type of elements and instruction set

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

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

Type of elements and instruction set

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

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